Imperial College London

Experimental Investigation of the Effects of

Soil and Environmental Conditions on
Smouldering Wildfires

A thesis submitted in partial fulfilment of the

requirements for the degree of
Doctor of Philosophy
in
Mechanical Engineering
By
Eirik Gaard Christensen
2020

Supervised by Prof. Guillermo Rein

 Copyright

Eirik Gaard Christensen, 2020

All rights reserved

 Declaration of Originality

I declare that this thesis and the work described within have been completed solely
by myself under the supervision of Prof. Guillermo Rein. Where others have
contributed or other sources are quoted, full references are given.

Eirik Gaard Christensen

2020

i
 Copyright Declaration

The copyright of this thesis rests with the author. Unless otherwise indicated, its
contents are licensed under a Creative Commons Attribution-Non Commercial-No
Derivatives 4.0 International Licence (CC BY-NC-ND). Under this licence, you
may copy and redistribute the material in any medium or format on the condition
that; you credit the author, do not use it for commercial purposes and do not
distribute modified versions of the work. When reusing or sharing this work,
ensure you make the licence terms clear to others by naming the licence and
linking to the licence text. Please seek permission from the copyright holder for
uses of this work that are not included in this licence or permitted under UK
Copyright Law.

ii
 ABSTRACT

Experimental Investigation of the Effects of Soil and

Environmental Conditions on Smouldering Wildfires
by Eirik Gaard Christensen
Doctor of Philosophy in Mechanical Engineering
Imperial College London, 2020
Supervised by Prof. Guillermo Rein

Smouldering peat-wildfires are the largest fires on earth and are responsible for vast
economic damage, negative health effects, and significant quantities of greenhouse gas
emissions. Despite their importance, limited research has focused on understanding their
dynamics. Here I systematically study the influence of the three most important soil
conditions: moisture content, inorganic content, and density, as well as three prevalent
natural environmental conditions: wind, slope and temperature. Two novel experimental
rigs were developed to study these conditions: the shallow reactor, which facilitating the
simultaneous measurement of both horizontal and in-depth spread, and the Experimental
Low-temperature Smouldering Apparatus (ELSA) which enables, for the first time, the
experimental study of arctic wildfires by studying smouldering in low temperature
conditions. The data generated is such that I put forward a new unifying theory of the
governing parameters of smouldering spread which explains the influence of all three
major soil properties. Horizontal spread was found to be controlled by heat sink density
(the energy required to heat the soil to burning temperatures), while in-depth spread was
governed by the organic density. The study of wind and slope revealed that forward and
uphill spread significantly influenced horizontal spread rates due to improved heat
transfer and oxygen supply, while downhill slopes and wind opposite to the spread
direction had no significant effect. Evidence was found to suggest that spread on a slope
can be explained as a function of the angle of spread direction relative to a horizontal
plane. For the first time, I revealed that decreased soil temperatures resulted in deeper
depth of burning, and heat losses reduced the critical moisture content from 160% (with
an insulated reactor base) to 120% (with a cold reactor base). Spread rate and peak
temperature where negligibly affected by soil temperature, even sustaining in frozen soil
conditions. This thesis provides a comprehensive study of factors influencing smouldering

iii
wildfires, providing insight and data which support a new theory of smouldering spread,
improving our understanding of smouldering dynamics.

iv
ACKNOWLEDGEMENTS
First I would like to thank my supervisor, Prof. Guillermo Rein for giving me the
opportunity to pursue research into fire science, encouraging collaboration on projects
outside of the topic of my thesis, and supporting me to travel to attend conferences and
conduct large scale experiments.

A big thank you to all the members of Hazelab for many engaging and stimulating
conversations on both fire and non-fire related topics. I have greatly enjoyed our many
trips together and studying beside you. In particular I want to thank Yuqi Hu, Agung
Santoso, Fahid Amin, Wuquan Cui, and Dwi Purnomo for the incredible adventure that
we had in Indonesia while conducting Gambut.

Thank you to my wonderful family for encouraging me all my life and offering invaluable
advice when I have had difficult decisions to make. I will always be grateful for you.

Additionally, I wish to thank Grace O’Regan, for her unrelenting support through what
has been an incredibly challenging endeavour.

Finally, a thank you to you the reader. Thank you for taking the time to read my thesis,
which I have put nearly 4 year into creating and thank you for taking an interest in
smouldering wildfires.

v
Table of Contents
NOMENCLATURE ........................................................................................................ ix

PREFACE ...................................................................................................................... xi

Other publications.................................................................................................................xii

Chapter 1: Introduction to Smouldering Wildfires......................................................... 14

1.1 Smouldering Wildfires ................................................................................................... 14

1.2 Fundamentals of Smouldering Combustion ................................................................... 17

1.3 Thesis Outline ................................................................................................................. 23

Chapter 2: Experimental Methods and Scales in Smouldering Wildfires ...................... 24

Summary............................................................................................................................... 24

2.1 Introduction .................................................................................................................... 24

2.2 Micro-scale ..................................................................................................................... 26

2.3 Meso-scale: Laboratory Experiments ............................................................................. 27

2.3.1 Experimental Smouldering Reactors ........................................................................ 28

2.3.2 Horizontal Reactor ................................................................................................... 29

2.3.3 Column Reactors ...................................................................................................... 32

2.3.4 Shallow Reactors ...................................................................................................... 33

2.3.5 Ignition Device and Protocol.................................................................................... 35

2.3.6 Sensitivity of Experiments ....................................................................................... 36

2.4 Field Scale ...................................................................................................................... 40

2.5 Conclusion ...................................................................................................................... 42

Chapter 3: Experimental Methodology .......................................................................... 43

Summary............................................................................................................................... 43

3.1 Sample Material and Preparation ................................................................................... 43

3.2 Shallow Reactor .............................................................................................................. 44

3.2.1 Reactor ..................................................................................................................... 44

vi
 3.2.2 Soil Conditions ......................................................................................................... 45

3.2.3 Wind and Slope ........................................................................................................ 47

3.2.4 IR Image Analysis .................................................................................................... 48

3.3 Experimental Low-temperature Smouldering Apparatus (ELSA) ................................. 53

3.3.1 Experimental Set-up ................................................................................................. 53

3.3.2 Soil Conditions ......................................................................................................... 55

3.3.3 Spread Rate Analysis ............................................................................................... 56

Chapter 4: Influence of Moisture and Inorganic Content on Smouldering Dynamics .... 60

Summary............................................................................................................................... 60

4.1 Introduction .................................................................................................................... 61

4.2 Method ............................................................................................................................ 62

4.3 Results and Discussion ................................................................................................... 63

4.3.1 Smouldering Width .................................................................................................. 63

4.3.2 Horizontal Spread Rate ............................................................................................ 64

4.3.3 In-depth Spread Rate ................................................................................................ 65

4.3.4 Global Spread Rate................................................................................................... 67

4.3.5 Bifurcation; the Partial Extinction of Smouldering.................................................. 69

4.4 Conclusions: ................................................................................................................... 70

Chapter 5: Influence of Density on Smouldering Dynamics and a Unifying Theory of

Spread ........................................................................................................................... 72

Summary............................................................................................................................... 72

5.1 Introduction .................................................................................................................... 73

5.2 Method ............................................................................................................................ 74

5.3 Results and Discussion ................................................................................................... 74

5.3.1 Influence of Density on Smouldering Dynamics ..................................................... 74

5.3.2 Governing Spread Dynamics ................................................................................... 78

5.4 Conclusion ...................................................................................................................... 84

vii
Chapter 6: Influence of Environmental Condition on Smouldering Dynamics: A Study of
Wind and Slope ............................................................................................................. 86

Summary............................................................................................................................... 86

6.1 Introduction .................................................................................................................... 87

6.2 Method ............................................................................................................................ 88

6.3 Results and Discussion ................................................................................................... 88

6.3.1 Effect of Wind .......................................................................................................... 88

6.3.2 Effect of Slope .......................................................................................................... 93

6.4 Conclusion ...................................................................................................................... 98

Chapter 7: Arctic Fires and Smouldering Combustion- Influence of Cold Temperatures on

Fire Spread .................................................................................................................... 99

Summary............................................................................................................................... 99

7.1 Introduction .................................................................................................................. 100

7.2 Method .......................................................................................................................... 101

7.3 Results and Discussion ................................................................................................. 101

7.3.1 Influence of Smoke Extraction Rate ........................................................................ 58

7.3.2 Effect of Cold Temperatures on Smouldering ....................................................... 101

7.4 Conclusion .................................................................................................................... 112

Chapter 8: Conclusion ................................................................................................. 114

References: .................................................................................................................. 120

LIST OF FIGURES ..................................................................................................... 132

LIST OF TABLES ....................................................................................................... 139

Appendix 1:.................................................................................................................. 140

Appendix 2:.................................................................................................................. 143

A2.1 The Effect of Wind on Smouldering Dynamics ........................................................ 143

A2.2 The Effect of Slope on Smouldering Dynamics ........................................................ 144

viii
NOMENCLATURE
Symbols
ci Specific heat capacity of inorganic content (j/kgK)
co Specific heat capacity of organic density (j/kgK)
cp Specific heat capacity (j/kgK)
cw Specific heat capcity of water (j/kgK)
d Diameter of ignition coil access points (mm)
H0 Depth of sample (cm)
IC0 Inherent inorganic content (%)
k Conductivity (w/mK)
Lw Latent heat of evaporation (J/kg)
𝑚"̇ Burning rate (kg/m2h)
m Mass of sample (kg)
mp Mass of dry peat (kg)
ms Mass of sand (kg)
mw Mass of water (kg)
Sd In-depth spread rate (cm/h)
SG Global spread rate (cm/h)
Sh Horizontal spread rate (cm/h)
tb Burning time (h)
Tc Air temperature at the chamber (°C)
Tcond Condition temperature (°C)
Ti Air temperature at inlet (°C)
V Reactor volume (m3)
W Mean width of smouldering (cm)
 Slope of experiment (°)
β Effective slope (°)
Δ Compression (%)
∆𝐻P Change in enthalpy of pyrolysis (j/kg)
∆𝑇d Temperature increase of drying (°C)
∆𝑇H Temperature increase of dry matter (°C)
θs Direction of spread relative to slope (°)
ix
θw Direction of spread relative to wind (°)
ρ Density (kg/m3)
ρb Bulk density (kg/m3)
ρi Inorganic density (kg/m3)
ρo Organic density (kg/m3)
ρw Water density (kg/ m3)
ф Direction of global spread direction (°)

Abbreviations
MC Moisture content (%)
MCx Moisture content series
IC Inorganic content (%)
ICx Inorganic content series
TGA Thermo-gravimetric analysis
MLR Mass loss rate (g/s)
IR Infrared
EF Emission factor (g/g)

x
PREFACE
Chapter 1 is an introductory chapter which explains the impact of smouldering wildfires and
why they are necessary to study. It also includes an introduction to the fundamental principles
in smouldering spread. In addition it present the aim of this thesis.

Chapter 2 discusses the scales across which smouldering occur and reviews the methods used
in literature to study smouldering at it scale. This chapter emphasises the importance, and
unique benefit studies at each scale provides to the holistic understanding of smouldering
wildfires. I also provide guidance on best practice for experimentally studying smouldering in
the meso-scale. This chapter is based on:

E. Christensen, Y. Hu, F. Restuccia, M. A. Santoso, X. Huang, and G. Rein Experimental

Methods and Scales in Smouldering Wildfires. Chapter 19 in: ‘Fire effects on soils: State of
the Art and Methods’, CSIRO, Australia. (2019) Doi: 10.1071/9781486308149

Chapter 3 explains the methodology of the experiments used throughout this thesis. Detailing
the set-up of analysis frame work of the two experimental rigs used here: the shallow reactor
and the Experimental Low-temperature Smouldering Apparatus (ELSA).

Chapter 4 investigates the influence of moisture and inorganic content on smouldering

dynamics in the shallow reactor. The chapter is based on:

E. G. Christensen, N. Fernandez-Anez, G. Rein, Influence of soil conditions on the

multidimensional spread of smouldering combustion in shallow layers, Combustion
and Flame, 214 (2020) 361-370. Doi: 10.1016/j.combustflame.2019.11.001

Chapter 5 completes the data set of the most important physical soil parameters by studying
the effect of density on smouldering dynamics using the shallow reactor. The results from
chapter 4 and 5 are combined to study the underlying fundamental parameters affecting spread,
and a general theory of smouldering spread is presented. This chapter is based on:

E. G. Christensen, G. Rein, Experimental Evidence of a Unifying Theory of Spread for

Smouldering Fires, Combustion and Flame (submitted)
xi
Chapter 6 investigates the influence of wind and slope direction on smouldering dynamics
using the shallow reactor. This chapter is based on:

E. G. Christensen, Y. Hu, D. Purnomo, G. Rein, Influence of wind and slope on

multidimensional smouldering peat fires, Proceedings of the Combustion Institute,
(2020). Doi: 10.1016/j.proci.2020.06.128

Chapter 7 investigates the influence of temperature on smouldering dynamics using ELSA.

This is the first experimental work studying smouldering arctic fires. This chapter is based on:

E. G. Christensen, Y. Hu, G. Rein, Arctic fires and smouldering combustion: influence

of soil and air temperature on fire spread, (to be submitted)

Chapter 8 Summarizes and provides conclusions on the work presented in this thesis

Other publications
Below is a list of papers and book chapters that were co-authored over the course of this PhD,
but does not form a part of this thesis.

Y. Hu, E. Christensen, F. Restuccia, and G. Rein, Transient gas and particle emissions from
smouldering combustion of peat. Proceedings of the Combustion Institute, 37 (2019) 4035-4042. Doi:
/10.1016/j.proci.2018.06.008

Y. Hu, E. Christensen, H.M.F. Amin, T. E. L.Smith, and G. Rein, Experimental study of moisture
content effects on the transient gas and particle emissions from peat fires, Combustion and Flame,
209 (2019) 408-417. Doi: 10.1016/j.combustflame.2019.07.046

M. A. Santoso, X. Huang, N. Prat-Guitart, E. Christensen, Y. Hu, G. Rein, Smouldering Fires and

Effects on the Soil: The State of the Art. Chapter 14 in: ‘Fire effects on soils: State of the Art and
Methods’, CSIRO, Australia. (2019) Doi: /10.1071/9781486308149

xii
M. A. Santoso, E. G. Christensen, J. Yang, G. Rein, Review of the transition from smouldering to
flaming combustion in wildfires. Frontiers in Mechanical Engineering, 5 (2019). doi:
10.3389/fmech.2019.00049

M. A. Santoso, W. Cui, H. M. F. Amin, E. G. Christensen, Y. S. Nugroho, G. Rein, Laboratory study

on the suppression of smouldering peat wildfires: effects of flow rate and wetting agent. International
Journal of Wildland Fire (submitted)

E. Rackauskaite, M. Bonner, F. Restuccia, N. Fernandez Anez, E. G. Christensen, N. Roenner, W.

Wegrzynski, P. Tofilo, M. Heidari, P. Kotsovinos, I. Vermesi, F. Richter, Y. Hu, C. Jeanneret, R.
Wadhwani, G. Rein, Fire Experiment inside a Very Large and Open-Plan Compartment: x-ONE. Fire
Technology (submitted)

xiii
Chapter 1: Introduction to Smouldering
Wildfires
1.1 Smouldering Wildfires
Wildfires have become an increasing concern over the past decades, increasing in frequency
and severity [1]. They pose a significant threat to society, causing large scale destruction,
changes to natural habitats, large economic loss, severe degradation of air quality, and
contribute significantly to global carbon emissions [2,3]. This increase in risk of fires stems
mainly from a combination of poor land management and a changing climate [4]. Research
into wildfires is essential to understand the current changes in fire regimes, predict future trends
and develop mitigating strategies.

Wildfires can broadly be split into two categories; flaming and smouldering, each of which
show markedly different behaviours, and have the ability to impact the environment in distinct
ways. Flaming is a high temperature (~1200°C) , homogeneous gas-phase combustion [5], and
can travel quickly across landscapes, consuming surface fuels and tree crown material,
changing the landscape, and in some cases facilitate a change in vegetation [6,7]. Additionally,
flaming wildfires often generate strong winds due to their high temperatures and the strong
buoyancy of their plumes. As a result, firebrands (small burning, typically smouldering,
fragments of fuel) are often carried significant distances downstream and are capable of
igniting houses at the wildland urban interface [8,9].

In contrast, smouldering is the slow, low temperature, heterogeneous form of combustion,

which propagates through porous fuels [10]. Smouldering fires are notorious for being both
easier to ignite and more difficult to supress than flaming fires, surviving heavy rain fall and
even fire-fighting efforts [10]. These fires contribute significantly to the total amount of
biomass consumption in wildfires [11,12], burning deep into organic soils for periods lasting
up to several weeks [10]. Smouldering wildfires are the largest fires on earth, with respect to
their total fuel consumption and are most common in peatlands. Peat is defined as a soil with
an organic content greater than 20% [13]. It is comprised of partially decomposed plant matter
that accumulates over centuries to millennia [14], developing in areas where the rate of plant

14
growth is greater than the rate of decomposition. These conditions occur where the ground is
too wet, or the environment is too cold for rapid decomposition [15]. These peatlands exist
across the globe from tropical regions such as Indonesia, to the arctic regions of Serbia and
Alaska. Historically, these systems have behaved as a carbon sink; although they only cover
3% of the global land area, peatlands contain 25% of the global carbon storage [4,16].

The smouldering of these ancient layers of organic-rich soils provides a rapid means of
releasing stored carbon into the atmosphere [2]. These fires have become a regular occurrence
in Indonesia, which has large expanses of peatlands. In many regions peatlands are drained to
expand human activity such as agriculture, decreasing the water table and drying the peat, thus
making the soil more prone to burning [17–19]. Such activities are exacerbated by the effect
of climate change which have resulted in more extreme weather events such as more severe
El Nino seasons [20] in south east Asia and hotter, dryer, and longer summer seasons in boreal
and arctic regions [4,21]. Temperature increases aid in drying peatlands, lower watering table
levels, thus making peatlands more flammable [2] and can result in mega-fires. An example of
one such mega-fire is the 1997 Indonesian wildfire where between 0.81 and 2.57 Gt of carbon
was released, which is said to be up to 40% of annual global anthropogenic emissions [2]. A
similarly large scale fire also occurred in 2015, following another extreme el Nino summer [3].
These fires often burn 50 cm into the soil [2,22] and can reach peat which dates back over a
millennia old [23]. Thus due to the long timescale required to re-sequester the lost carbon it is
considered a positive contribution to the carbon emissions [4,16]. This is different from flaming
fires, whereby the majority of the lost carbon may be rapidly captured within a decade by new
plant growth.

Smouldering is a characteristically incomplete form of combustion and forms high levels of

carbon monoxide (relative to flaming fires) and a broad range of organic compounds and
particulate matter. Due to the low temperature of the smouldering wildfires, the smoke plumes
are weakly buoyant, and so remain close to the ground. This smoke may enter populated
communities, significantly deteriorating the local air quality in events known as haze. Haze
events have frequently plagued south-east Asia as they are responsible for large economic
losses, and have severe negative health impacts [24–26]. For example Koplitz et al. [24],
estimated the smoke exposure from the 2015 megafire was responsible for over 100,000 excess
deaths in South East Asia.

15
Recent events have also highlighted a change in the fire regime of arctic regions, with fires
occurring in Alaska [27], Siberia [28], and even Greenland [29], where significant swaths of
land in regions with infrequent cases of fires have burned. These northern peatlands are closely
connected with permafrost, and are an important natural system. Permafrost is defined as any
ground that is frozen for a continuous period of at least 2 years [30]. These cold conditions
reduce the rate of plant matter decomposition, promoting the accumulation of plant debris,
resulting in ideal conditions for peat formation. The peat, in turn, acts as an effective insulator,
protecting the permafrost. Tarnocai et al. [31] reports that 1700Gt of carbon are stored in
permafrost, which is equivalent to 2 times the mass of carbon in the atmosphere. Permafrost is
overlaid by the active layer; a layer that thaws and freezes seasonally. The thickness of both
the active layer and the permafrost are a function of several variables including seasonal air
temperature, snow deposition, and soil thermal properties [32]. With climate change increasing
average seasonal temperatures and longer dry seasons, these peatlands are transitioning to
becoming a net emitter of greenhouse gases [33]. Drier and warmer weather increase the depth
of the active layer and the natural rate of plant matter decomposition, releasing carbon dioxide
and methane, the latter is a particularly potent greenhouse gas [34]. Besides providing a means
of rapidly releasing the stored carbon into the atmosphere, fire can also have a long lasting
damaging effects on the underlying permafrost [35,36], which may facilitate increased rates of
natural peat decomposition and further greenhouse gas emissions. These emissions in turn
contribute positively to climate change warming the region further [37]. These effects are
exacerbated by a trend that shows that temperatures in northern regions are increasing at twice
the rate of the global average [38]. As mentioned, warmer and dryer seasons also increase the
vulnerability of these soils to ignition [39]. Additionally, increased temperatures also increase
lightning activity which is the most common ignition source in northern fires [40]. With
continued climate changes, fires in these northern regions are predicted to continue to increase
in frequency and severity [41]. As such, the health of these peatlands are crucial, if the goal of
keeping global temperature rise to below 1.5°C is to be met [42].

Furthermore, there are reports of fires that have ignited during the summer, survived the winter
season by burning below ground, and returned to the surface once the snow has melted. This
is said to be correlated with summers with very high temperature and low precipitation
followed by mild winters [43]. These conditions enable the fire to penetrate deep into the soil,

16
which in turn insulates the fire from winter conditions. Such summers are increasing in
frequency with longer fires seasons, enabling deeper burn depths [33]. Fires that resurface after
winter have been referred to as zombie fires, as they are thought to die in the autumn, only to
come alive in the spring [44]. However, they are more commonly referred to as holdover-fires
or overwintering fires. Such fires, once they resurface, may transition to flaming under
favourable conditions and spread much more rapidly [45]. Overwintering fires may have
significant implications on total burn area and total emissions in artic regions. Fires in boreal
and arctic peatlands pose a particular challenge due to their remoteness, making a firefighting
response difficult. The fire dynamics of arctic fires are not often discussed due to their remote
occurrence, with much research occurring post extinction or via remote sensing. However, it
is highly likely the dominant mode of combustion in arctic regions is smouldering because of
the prevalence of peat as well as smouldering fires ability to surviving at low levels of oxygen
concentration [46,47], enabling sustained burning below the surface and likely even below a
layer of snow or ice. No experiments to date have studied the dynamics of smouldering arctic
wildfires.

Despite their global importance, the fire science community has focused on flaming wildfires,
leaving smouldering a still poorly understood phenomena. Some lab-scale experimental work
has been performed to gain a fundamental understanding of the smouldering dynamics of these
wildfires and will be discussed further in the next section. Such research is essential to inform
mitigation strategies.

1.2 Fundamentals of Smouldering Combustion

Smouldering generically propagates through fuels in 3 distinct fronts. These fronts are
identified as drying, pyrolysis and oxidation (Figure 1). The drying front is marked by a
temperature threshold of around 100°C which is associated with the phase change of water.
Locally, this temperature will not be exceeded until the moisture is evaporated. This process
presents a significant barrier to the propagation of the smouldering due to the large latent heat
of water. Although this is not easily visible, it occurs in the left most part of Figure 2. Once the
moisture has been removed the temperature of the peat steadily increases until a temperature
of approximately 250°C [48]. Above 250°C the fuel thermally decomposes in a process called
pyrolysis which does not require oxygen to occur. This is an endothermic process that produces
black carbonaceous char and pyrolysates [46]. Flaming combustion occurs when these
17
pyrolysates oxidize; as such, smouldering has been shown to be capable of transitioning to
flaming [45,49]. Notably, char can also be formed via an alternative pathway that is not
presented in the simplified schematic in Figure 1: peat oxidation. This reaction, which requires
the presence of oxygen, is thought to be exothermic and increases the rate of char formation
(relative to an inert environment), but the precise nature of this pathway is still a matter of
research [50,51]. This pathway, when included in computation models, has been shown to
improve simulation accuracy [52]. The pyrolysis front is identifiable in Figure 2 by the
boundary between the char and the fresh peat. The char continues to increase in temperature
until approximately 300°C, when it begins to oxidize and release energy [53]. This heat drives
the propagation forward and provides the necessary heat for the previous steps to occur. It is
also responsible for the majority of the mass lost during smouldering [54]. The residue from
the oxidation front is ash; the nonreactive elements in the original material and is clearly visible
as the white powdery substance in the rightmost part of Figure 2.

Figure 1 - Demonstrating a simplified schematic of the fundamental fronts in smouldering combustion in soils.

Figure 2 - Smouldering through commercial peat sample, illustrating characteristic smouldering fronts; fresh peat includes a
region of drying. The char and ash region can also clearly be seen by the colour and structure change of the fuel, resulting in
a cloud of visible emissions.

Examining the generalized reactions pathways (Figure 1) it is evident that heat losses and
oxygen supply are the two fundamental mechanisms that control smouldering propagation [48].

18
Oxygen supply is crucial for the production of energy and so the rate of oxygen supply governs
the heat release. Heat losses influence spread by removing the energy generated from oxidation,
delaying the advancement of the drying and pyrolysis fronts and limiting the spread rate. This
balance between heat generation and loss is the focus of much research in smouldering fires,
as a broad range of parameters influence these mechanisms. In particular, they guide the study
and understanding of smouldering phenomena in organic soils [55].

MC, IC and density are the three soil conditions which are considered to have the greatest
influence on smouldering dynamics [10]. Each influence the mechanisms of oxygen supply
and heat losses uniquely. MC is widely considered the single most important factor in
controlling the behaviour of smouldering fires, influencing the temperatures generated in the
soil and spread rates [39,56] due to water having a high specific heat capacity and latent heat
of vaporization, which behaves as a significant heat sink, consuming a large proportion of the
energy produced by the oxidation front [10]. Moisture has been found to decrease horizontal
spread rate significantly [53,56,57] but increase the in-depth spread rate. The increase in in-
depth spread rate was explained by the increase in porosity of peat due to its expansion when
absorbing moisture which was hypothesised to improve oxygen flow. However, in all cases,
spread rate decreases with depth due to the growing ash layer limiting oxygen supply [58–60].
Similar to moisture, IC does not participate in the combustion, but acts as a heat sink absorbing
energy as it increases in temperature alongside the peat. Additionally the IC, fills the pore space,
reducing oxygen supply as well as increasing the conductivity of the bulk material. Very few
studies have considered the influence of IC on smouldering dynamics. The few studies that
have been performed report that a 30 % increase in IC reduced the horizontal spread by around
35% [61] while in-depth spread has been shown to be independent to IC [62]. Finally, density
also directly influences the controlling mechanism of smouldering. Increasing density results
in decreased oxygen supply, and alters the internal and external heat transfer. Both horizontal
and in-depth spread rates have been shown to decrease with increasing density, as a greater
volume of material is require to heat up and be consumed [56,59]. The sensitivity of horizontal
spread rate to density has been found to increase with MC [56] attributed to the increasing
energy required to vaporise water, and the diminished oxygen supply.

Smouldering can only be self-sustaining if the energy evolved from this char oxidation is
sufficient to overcome the energy requirements of the first two fronts (i.e drying, and pyrolysis

19
fronts) and the heat losses to the environment [39] (Figure 1). In conditions where heat losses
are too great, such as high MC, the smouldering front will stop or sustained ignition will not
be possible [39,63–65]. Frandsen [39,63] pioneered this crucial concept by performing lab
experiments with commercial peat, adjusting both the MC, by adding water, and IC, by adding
sand, until the sample would no longer sustain smouldering combustion. This threshold to
ignition was found to have a negative linear relationship. These findings were later successfully
replicated by Huang et al. [65] using a computational model. Hungerford et al. [55] replicated
Frandsen’s study using natural samples of a range of plant species, and found a similar trend
but shifted to higher values. Hartford [66], found that the linear threshold to ignition between
IC and MC is decreased with increasing organic densities. Similarly, Hu et al. [61] found that
the critical moisture content of peat decreased from 160% to 50% by increasing density by
25%. While Huang and Rein [67] found that, at very low densities, an initial increase in critical
MC with increasing density occured before a turning point was achieved and increasing density
resulted in decreased critical MC. This was explained by a relative change in dominant heat
transfer modes. Garlough and Keyes [68] and Prat-Guitart [56], however, found that density
had no effect on ignition probability. This illustrates that the effect of density is complex and
still poorly understood. This may be due to confounding variables, such as porosity, material
density, and thermal properties, as density has a multifaceted interaction with the two key
driving factors of smouldering, namely, oxygen supply and heat transfer [69]. The
understanding of critical conditions are crucial for informing the risk of ignition of soils.

The balance between oxygen supply and heat losses to maximize energy generation is naturally
sought by a smouldering fire. This process can lead to a long misunderstood phenomena;
overhang. Smouldering peat in wet conditions has been experimentally and computationally
shown to migrate deeper into the soil as heat losses are reduced due to insulation provided by
the soil above [70]. Moving downwards however also results in diminishing oxygen supply. It
was found that with increasing MC the depth at which peak temperatures could be found would
increase. Figure 3 is an image showing the char layer existing well below the free surface. This
propagation results in periodic collapse of the above peat as the smouldering consumes the soil
mass below the surface and eliminating the support. This phenomenon illuminates earlier
findings that discovered critical thicknesses of samples. Palmer [71], an early researcher of
smouldering, studied the critical sample thickness of dust, by laying a train of dust on an
inclined surface, maintaining a level top surface such that the thickness of the sample decreased

20
across the its length. Smouldering would cease at the location where the depth resulting in heat
losses greater than the energy generated. Similarly, Miyanishi and Johnson [72] successfully
explained patterns in forest ground fires (duff) by experimentally studying the relationship
between soil moisture and critical depth, finding a greater critical depth for higher MCs.

Figure 3 - Image of a smouldering peat experiment with 120% moisture content. It can be seen that the smouldering front as
indicated by the char layer (the black material indicated with an arrow) exists far below the surface of the peat.

Further to this, environmental conditions have a direct influence on smouldering dynamics, the
most notable of which is wind. The strength and orientation of wind strongly interacts with the
combustion as it directly affects the oxygen supply. Two spread categories have broadly been
defined based on the direction of the air flow relative to the spread; forward and opposed spread
[48]. Forward spread is when the oxygen supply is in the same direction as the spread. This
enhances the spread mechanism by rapidly providing oxygen to the oxidation zone, which in
turn produces hot gaseous products which are blown in the direction of spread resulting in
enhanced convective heating, aiding in the preheating of the fuel. Opposed spread, on the other
hand, is when the oxygen supply flows against the direction of smouldering spread. In this
instance the preheated region is convectively cooled, and the char begins to oxidize as soon as
it is formed, creating a significant overlap between the pyrolysis and oxidation zone.
Comparing the two modes of wind controlled spread it has been found that forward spread
results in greater spread rates and broader smouldering waves, while opposed spread only
experiences a slight increase in spread rate relative to quiescent conditions, but is far less
sensitive to increases in wind speed [71]. Additionally, Valdivieso and Rivera [73] showed that
forward wind conditions increase the critical MC of pine needles. Slope can similarly create

21
forward and opposed wind conditions, as buoyant plumes encourage uphill airflow [71].
However, the influence of slope on smoulder spread in peat has never been studied.

Soil and environmental conditions interact dynamically in nature, with soil conditions changing
both in-depth and horizontally. MC, IC, and density all typically increase with depth to varying
degrees. This often results in limiting the downward propagation of smouldering by exceeding
critical smouldering conditions [53,64]. MC is known to vary from over 300% down to below
20% in drought scenarios [74]. Further non-uniformity such as stage of decomposition has also
been found to influence smouldering dynamics [66]. While environmental conditions have high
spatial and temporal variability. Human behaviour also furthers the variability of soil
conditions as the drainage of peatlands results in changes to the characteristic of the peat such
as MC, density and even stage of decomposition [75]. Such natural variability makes the study
of smouldering dynamics on the field scale a challenge as variables cannot be isolated. Further
discussion of variability and its importance to smouldering fires is presented in Chapter 2.

The research in literature into key physical and environmental parameters affecting
smouldering wildfires is fragmented and has been conducted across a variety of rigs and soil
sources with horizontal and in-depth spread studied individually. This has made creating a
unifying theory on smouldering wildfires, understanding the relative importance of each
parameter, and gaining an insight into the relationship between horizontal and in-depth spread
a challenge. In this thesis I aim to address this challenge by proposing two novel reactors
which can be adopted to study smouldering. The first of which, enables the simultaneous study
of both horizontal and in-depth spread. Using this reactor the influence of 5 conditions will be
explored: MC, IC, density, wind and slope, and the common controlling parameters of these
variables will be explored. These studies will not only to further our understanding of the
fundamentals of smouldering dynamics but also illustrate the value of adopting a common rig
design. The second experimental set up aims at providing an effective means of studying the
influence of temperature to address the rising concern of arctic fires. I will also present data
using this rig examining the interaction between condition temperatures and smouldering
dynamics. Thus the overarching mission of this thesis is to present a comprehensive
experimental study on the dominant natural conditions to gain a greater understanding of their
effects on smouldering dynamics, contributing to a deeper fundamental understanding of

22
smouldering wildfires and promote the adoption of these reactors to facilitate comparable
results.

1.3 Thesis Outline

The outline of this thesis is as follows:

Chapter 1 introduces the global importance of smouldering wildfires and the fundamental
dynamics of smouldering combustion.

Chapter 2 reviews the methods employed to study smouldering wildfires across scales and
discusses the importance of research at each scale.

Chapter 3 presents the experimental methodologies used in this thesis.

Chapter 4 presents an experimental study of the effects of moisture and inorganic content on
the multi-dimensional smouldering spread in peat.

Chapter 5 presents an experimental study of the effect of density on multi-dimensional

smouldering spread in peat. I then combine the results from chapter 3 and 4 to conduct a holistic
analysis of the controlling parameters of smouldering spread and present a unifying theory of
smouldering spread.

Chapter 6 presents an experimental study of the effect of both wind and slope on multi-
dimensional smouldering spread in peat.

Chapter 7 presents the first experimental study of smouldering spread in frozen soil, and studies
the effect of soil temperature on spread dynamics.

Chapter 8 summarises and gives conclusions on the work presented in this thesis.

23
Chapter 2: Experimental Methods and
Scales in Smouldering Wildfires
Summary1
This chapter discusses the experimental methods used in literature to understand smouldering
across a wide variety of scales. Smouldering is characterized by an oxidation reaction on the
micro-scale, its dynamics are affected by heat transfer which is subject to the natural variation
of soil and environmental conditions on a meso-scale. In wildfires this interplay occurs
simultaneously across large swaths of land on the field scale. Thus smouldering is studied at
each scale to gain a holistic understanding of the dynamics. The micro-scale is often studied
via processes such as thermo-gravimetric analysis (TGA) to understand the kinetics of
pyrolysis and oxidation. This is essential for our understanding of the underlying chemistry as
well as to inform computer models which give insight not available through experimentation.
While the meso-scale, allows for rapid experimentation and the isolation of the influence of
individual variables providing insights into the fundamental dynamics of these fires. Finally
the field-scale, presents challenges with its great spatial and temporal variability, but is
essential to study to understand the significance of smouldering fires and inform what
behaviours are realistic and spur further research questions.

2.1 Introduction
Smouldering occurs across a vast range of scales: from the heterogeneous chemistry at the
micro-scale, to the interaction between environmental conditions and fire dynamics. Much is
still to be understood about this type of wildfire. The study of smouldering wildfires is
complicated by the natural variability found in the ecosystem. Topography, micro-topography,
weather conditions, soil condition, and soil depth, all have an influence on smouldering
behaviour, and often change rapidly in space and in time [76]. These variables confound the
understanding of the wildfire behaviour, and makes isolating the influence of individual

1
Chapter is based on: E. Christensen, Y. Hu, F. Restuccia, M. A. Santoso, X. Huang, and G.
Rein Experimental Methods and Scales in Smouldering Wildfires. Chapter 19 in: ‘Fire effects
on soils: State of the Art and Methods’, CSIRO, Australia. (2019) Doi:
10.1071/9781486308149
24
variables important, but difficult, and hinder the development of a more fundamental under-
standing of fire. Experimental studies are at the centre of research activity for this topic, in
order to develop theories and models.

Experiments can be performed at a variety of scales, each capable of answering different

questions. The scale of the experiments can be divided into three categories: micro-scale, meso-
scale and field scale. Micro-scale studies are performed on milligram samples examining the
fundamentals of chemistry, independent of bulk behaviour and transport phenomena. The
meso-scale, which is typically studied in a laboratory, is performed on gram to kilogram
samples. It is used to learn about the fundamentals of fire dynamics, but greatly simplifies the
vast complexity of natural ecosystems making specific predictions at the real scale difficult.
However, this simplification allows all important conditions to be controlled so that the
relationship between variables can be discerned and measured, providing vital insight into the
general response of a system to a known change. Fundamental knowledge derived from meso-
scale studies, in turn, allows field data to be understood in context and for it to be better related
to measurements at other sites. Field-scale studies gather information from fires in situ, and
generally span a few metres to kilometres. Although satellite studies make little to no
distinction between flaming and smouldering fires, and field studies on smouldering are few
and far between, they provide invaluable data from realistic fires and conditions that could be
applied to other fires.

Studies at each of the three scales are essential for an integral and robust understanding of the
fire phenomena. The micro-scale provides detailed information about the chemistry that
informs computational models. Then, the model is coupled with heat and mass transfer
mechanisms and validated against meso-scale experiments, identify the controlling variables
of the results measured in the laboratory. The fundamental understanding at the smaller scale
can then be extrapolated and applied to the large scale to provide context and deeper meaning.
Figure 4 illustrates how the combination of studies across all scales is a powerful strategy
towards the understanding of smouldering wildfires, with each scale informing, explaining and
putting into context the knowledge gained from the others.

The experimental methods used at each scale and their limitations will be discussed in the
following sections. Particular attention will be given to the meso-scale. This is because the
micro-scale has more standardised procedures and equipment, and the field scale has very few

25
experiments focusing on smouldering dynamics. On the other hand, the meso-scale has been
studied using several experimental set ups [39,59,77,78], but little specific guidance can be
ascertained from the current literature on how the experimental set up is designed and why.
Central to their use is the key understanding of the assumptions made in the design of the
experimental set up, and the key mechanisms that control smouldering. It is therefore beneficial
to communicate experimental experiences and establish a framework for best practice as a
reference for future studies. Several design factors will be highlighted and their significance
will be explained, with justifications. The intention of this chapter is to provide a guidance of
the available methods for the study of smouldering wildfires, particularly the fire in organic
soils.

Figure 4 - Scale matters. Experiments are conducted on all scales from sub-millimetre scale up to metre and kilometre
scale. Each scale provides useful insights into understanding the large-scale wildfire phenomena. (Micro-scale image
by David Lazaro)

2.2 Micro-scale
The study of thermal decomposition and combustion of soil at the smallest scale (milligrams)
focuses on the chemistry of pyrolysis and oxidation and the development of reaction schemes
to best describe the chemical pathways taken in the burning of soils. One of the most common
techniques is thermo-gravimetric analysis (TGA). This method exposes milligram-scale
samples to selected temperature ramps, in the order of 5–30°C/min [79,80]. The mass of the
sample is recorded and plotted against the temperature. The relationship between mass loss and
temperature indicates the characteristic temperature of each chemical reaction. These
experiments are typically conducted in air or nitrogen (inert) environments to allow for the
distinct reactions dependency to temperature and oxygen to be determined. This analysis relies

26
on the fundamental assumption that the sample size is sufficiently small so that the temperature
is the same as the surrounding gas and that no temperature gradient exists within the sample.
Any large temperature gradient across the sample would result in the decoupling of the
temperature measurements and the chemistry response, misrepresenting the onsets of chemical
reactions. Additionally the analysis assumes that mass transfer is instantaneous, meaning that
any gaseous products are immediately released. Because these experiments are conducted on
such a small scale, obtaining samples that are representative in composition of the bulk material
is particularly challenging in non-homogenous soils.

TGA results have been crucial in understanding fundamental chemistry involved in the
combustion of peat and for constructing more accurate computational models [52]. Most
commonly, computational models have been applied to meso-scale samples and allow for the
individual examination of the influence of parameters, such as conductivity and specific heat
capacity, which cannot be independently varied in the field or in the laboratory [65].

2.3 Meso-scale: Laboratory Experiments

Meso-scale experiments provide an effective means of simplification to study fundamental
behaviour of smouldering through isolating and controlling variables. Experiments at this scale
can vary, allowing for creative approaches to studying a complex phenomenon. Although all
experiments are limited by the chosen set up; in the study of smouldering dynamics, open-
topped reactor have had great success [39,54,56,57]. A few of the experimental set ups in the
literature will be presented and discussed, outlining their strengths and limitations. It is hoped
that, through a detailed discussion of the experimental framework, widespread adoption might
follow, allowing for consistency of results and form a foundation for a relatable database,
discussion and comparison.

The current state of the art in the study of fundamental behaviour of smouldering has preferred
to use commercial peat due to its consistency and availability. Such peat is more uniform in
density and particle size, which aids in replicability of experiments. The most commonly used
type is Irish sphagnum peat (Shamrock Irish moss peat, Bord na Mona Horticulture Ltd)
[56,57,70,81,82]. A common peat sample allows for the direct comparison of results between
experiments. However, some experiments have taken pedons or other samples of natural soils
and smouldered them in the laboratory [55,64,77,83,84]. Such samples pose additional

27
challenges such as quantifying variability between and within samples. On the other hand,
natural samples provide a meaningful and necessary intermediary step between commercial
samples analysed at laboratory scale and field scale experiments.

Commercial peat is well suited to study the most influential parameter, moisture content (MC),
by controlling the addition of water [39,57,77]. In experiments commercial peat is typically
dried first, such that a controlled quantity of water can be added, targeting a desired MC. Drying
is performed at 80–90°C until no mass change can be seen. Water is then added and the sample
is mixed vigorously by hand before allowing the sample to sit in a sealed container for at least
24 h, allowing the moisture to homogenise and be fully absorbed into the peat. Before starting
an experiment, the sample is again vigorously mixed to remove any moisture gradients caused
by water movement due to gravity. It is recommended that at the time of the experiment, a
subsample of around 100 g should be collected and dried to accurately measure the moisture
content.

Meso-scale experiments allow for several diagnostics to be used simultaneously to study

smouldering. The most frequent of these are weighing scales, thermocouples, infrared (IR) and
visual imaging. Weighing scales allow for the rate of loss of mass of the sample to be calculated.
Thermocouples provide the ability to measure a temperature profile across a sample. The
specific location of thermocouples is left to the experimenter, but, if a generic profile
temperature profile is of interest, the tip of the thermocouples should be placed into the centre
plane of the reactor to minimize boundary effects. Multiple thermocouples can measure
smouldering spread rate by noting the time interval taken between successive readings at a
given temperature threshold [70] and dividing by the distance between the thermocouples.
Caution should be taken in adding too many thermocouples because this may conduct heat
away from the smouldering front, as well as influence the structure of the soil as it recedes.
Cameras provide a visual record of the shape of the front, as well as the rate of spread
[56,84,85]. This set of diagnostics provides a robust range of measurements for the analysis of
smouldering dynamics in meso-scale experiments.

2.3.1 Experimental Smouldering Reactors

Smouldering reactors are insulated open-top cuboids in which smouldering experiments are
performed. The open-top design allows for natural oxygen supply from the free surface, and is
constructed from insulating material, used for its low conductivity and non-flammability
28
[39,54]. Most commonly, the reactor is filled to the rim with soil, such that a repeatable quantity
is used and the density of the sample can be calculated. The specific dimensions and method
of ignition vary depending on the direction of spread being studied.

Smouldering spread can be divided into two dimensions: horizontal and vertical (in-depth), as
shown in Figure 5. Each dimension has been found to show different behaviour in response to
changes in soil conditions [57,59]. For this reasons, it is typical to study these two modes of
spread separately using a unique reactor for each. Three primary reactor types will be discussed:
(1) the horizontal reactor, used to primarily study horizontal spread; (2) the column reactor,
used for in-depth spread; and (3) the shallow reactor. Each provide unique abilities to study the
fundamentals of smouldering soil dynamics. The dimensions of reactors have an influence on
the behaviour of the smouldering soil. Smaller reactor will thermally and physically interact
more strongly with the smouldering fronts. Ideally an infinitely large reactor would be used,
but a balance must be struck between realism and consumption of time, space and material.
Additionally, it would be useful to have standardised reactor dimensions, because any change
to the reactor dimensions introduces uncertainties, and a consistent methodology would allow
results to be compared in the literature.

Figure 5 - Schematic of smouldering front in horizontal and in-depth spread in peat fires. Sh and Sd indicate the horizontal
spread and in-depth spread rate. Source: [52], CCBY

2.3.2 Horizontal Reactor

For the study of horizontal spread, the internal dimensions of the current version of the reactor
are 20 × 20 × 10 cm. The sample is ignited using a heated coil along one side of the reactor at
5 cm depth [70,86]. In order to install the coil igniter, small holes are punctured through the
front wall of the reactor at 5 cm below the top and 1 cm from each of the side walls. Thin

29
ceramic tubing is inserted snuggly into the holes to prevent wear and expansion by the coil.
The horizontal reactor is illustrated in Figure 6.

The reactor is designed such that it is inexpensive and easily constructed. It is built using
insulating walls (fibreboards). The side walls join together on top of the base board, with an
additional baseboard laid inside the reactor, fitting between the walls (Figure 6). The boards
are joined tightly using staples and reinforced using aluminium tape. Aluminium tape or foil is
attached to the outside of the reactor to reduce radiative heat losses and to seal the junctions
where the boards meet, limiting oxygen ingress through the gaps.

Given that the sample is ignited along the full length of one side (Figure 6) the smouldering
front tends to spread parallel to the heater and so propagating one dimensionally horizontally.
The dimension of reactor and the heat transfer through walls will influence shape of the spread
front. The thermal conductivity of the insulation walls should be much less than that of the soil,
such that it does not conduct heat forwards, heating the sample. The width of the reactor must
be large enough such that the centreline spread can behaves independent of the wall effects. A
20 cm width has been found to be suitable, as illustrated by linear front exhibited in the infrared
time-lapse photos of the surface of dry smouldering peat in Figure 7.

Figure 6 - (a) A reactor for testing horizontal spread; and (b) an exploded view for the purposes of construction.

30
Figure 7 - Infrared image seen from above the smouldering spread through dry peat in the horizontal reactor. The igniter is
buried below the surface at the boundary at the top of the images.

The interaction between the length and depth of the reactor and the smouldering dynamics is
also important. The front propagates across the length of the reactor, with heat transferring to
the various fronts of smouldering. As the smouldering approaches, the far wall begins to
influence the spread. During the early stages of the interaction, the boundary may cause lower
heat losses because the wall replaces wet soil, meaning no moisture is required to evaporate.
During the later stage, as the pyrolysis and oxidation front approaches, the heat losses through
the wall may be significant due to the high temperature gradient across the wall. In early studies,
smouldering experiments were performed in reactors of 10 × 10 cm with a broad variation in
depth [39,55,68,77,87]. This is not large enough to allow for the smouldering structure to form
in the absence of the boundary effect. To illustrate this, Figure 8 is a temperature profile taken
from a smouldering dry peat sample using thermocouples placed at mid depth, 3, 7, 11 and 15
cm away from the front wall. It was ignited using a heat coil for 30 min. It can be seen that, at
the end of the ignition period (after 30 min), the temperature rise only propagates between 7 to
11 cm. It is evident that in reactors smaller than 11 cm, the boundary will begin to influence
the dynamics before the ignition protocol has concluded. The temperatures begin to rise at 15
cm from the front wall after ~1.3 h, almost 1 h after the ignition protocol ended. From this it
can be deduced that a reactor of 20 cm appears appropriate to study the smouldering dynamics
with minimal influence from the walls. It should be noted that examining the behaviour in dry
peat is the most challenging scenario, because dry peat has the greatest spread rate and is thus
most rapidly influenced by the walls. However, this behaviour illustrates the broad validity of
the proposed dimension. Larger dimensions are not considered in the literature because the
duration of the experiments is excessive and the moisture of unburned peat may be dried
naturally. The depth of the reactor is also significant, because it is a possible heat sink. However,
for smouldering fires in scenarios where heat losses are high (high moisture content or
inorganic content), the smouldering front tends to recede deeper into the peat, increasing the
depth of the burning [70]. This is because below the surface because an optimal balance is

31
found between minimising heat losses and maximising oxygen supply [70]. Limiting the depth
of the reactor limits smouldering depth. As mentioned before, it has been shown that decreasing
sample depths also reduce the critical moisture content of horizontal spread [71,72]. As such,
using an insulated base minimises heat loss, providing conditions favourable to smouldering.
For the purpose of studying ignition and horizontal spread, an insulating boundary will behave
as an upper limit of heat retention in natural conditions for a given depth. However further
research is required to understand the influence of depth on smouldering dynamics.

Figure 8 - Temperature profile of smouldering dry peat in a horizontal reactor, measured at 5 cm depth. Temperatures 11 cm
from the front wall can be seen to rise only 15 min after the 30 min ignition period.

2.3.3 Column Reactors

Experiments that specifically study in-depth smouldering require a substantial increase in-
depth [58,59,62,64,88] to allow for prolonged study of propagation solely in the in-depth
direction. It has been shown that in-depth and horizontal spread are affected differently to
changes in soil properties. In-depth spread is much slower than horizontal spread and both
Huang and Rein [59] and Roulston et al [58] showed that the rate of in-depth spread progres-
sively decreases with depth as the thickness of the ash layer increases on the surface reducing
the oxygen supply to the oxidation front. However, it was shown that (unlike horizontal spread)
in-depth spread rate increased with increasing MC [59] and that the upward spread of peat fire
can be much faster than the in-depth spread [89]. These findings illustrate the values of studying
these horizontal and in-depth spread independently.

32
A reactor specifically designed to study in-depth spread should ideally be constructed such that
the front movement can be considered one-dimensional. Most commonly, column reactors
have an internal dimension measuring 10 × 10 × 26–30 cm (Figure 9)[59,88], although
experiments have also been performed using wider samples, as the 60 × 40 cm sample size
used by [64], or cylindrical in shape [62]. Unfortunately there is a limited number of
experiments focused on this spread component, despite its importance in controlling the
severity of burn and total emissions in wildfires.

Figure 9 - Illustration of a column reactor. Typical dimensions of which are 10 × 10 × 30 cm, with a coil igniter submerged
5 cm below the free surface.

2.3.4 Shallow Reactors

Another method of studying smouldering at the meso-scale is that of a shallow reactor. This
set up allows for the study of both horizontal and in-depth spread. The reactor measures 40 ×
40 × 1.6 cm, with the ignition coil being 1 cm long, with a coil diameter of 1 cm to promote a
radial propagation for 20 cm in all directions (Figure 10). The sample is made a uniform thick-
ness by scraping away excess soil above the height of the walls.

As discussed earlier, in thicker samples with a high MC, smouldering tends to recede deeper
into the soil making observation of the smouldering difficult. However using this shallow
reactor the smouldering is maintained near the surface and using an IR camera, a lead and

33
trailing edge can be clearly defined (Figure 11). Each of the edges can be uniquely associated
with the two directions of propagation. The leading edge propagated horizontally into virgin
soil and can be used to measure the horizontal spread. The trailing edge, however, is an
indicator of where smouldering has successfully consumed the full depth of the soil. Because
both the depth of the layer and burning time are known, the in-depth spread rate can be
calculated. Additionally, the value of this set-up is the reduced experimental time, requiring
only between 2-5hr to complete, compared to a minimum of 8 hrs in the horizontal reactor.
Additionally, the shallow layer allows for the isolation of the influence of the condition on the
smouldering spread, independent of depth effects such as changing burning depth and overhang.
Providing a powerful technique for examining the influence of soil properties and wind on
spread rates and the relationship between horizontal and in-depth spread. However, due to the
dimensions of the reactor, it comes with significant limitations. The substantial reduction of
depth introduces a strong boundary effect, e.g. the insulation wall, promotes fast forward and
in-depth propagation. As such, direct application of results to the field scale should be done
with caution. This method will be presented in greater detail in the following chapter (Chapter
3).

Figure 10 - A shallow reactor measuring 40 × 40 × 1.6 cm, with a short 1-cm long ignition coil placed at the centre of the
reactor wired through the base via small ceramic tubes.

34
Figure 11 - Infrared image taken from above the shallow reactor showing the spread of a leading and trailing edge in a layer
of dry peat.

2.3.5 Ignition Device and Protocol

For all reactors, it is recommended to ignite using a heated metal coil to accurately control the
amount of energy supplied to the sample. However, critical for ignition, enough energy must
be supplied through the coil such that all moisture is evaporated from the soil locally, so the
soil can be heated above the temperature necessary for pyrolysis (>250°C) and trigger oxidative
reactions (>300°C) [52,79,80]. Ideally, the ignition protocol is sufficiently strong that results
become independent of ignition. Ultimately, if an ignition protocol is too weak, the
smouldering may not successfully initiate despite sustained smouldering being feasible. The
minimum application of 100 W for a duration of 30 min has been found to successfully ignite
samples of <160% MC and is considered to be equivalent to a strong exposure to a flaming fire
[77]. A 0.05 mm thick wire is wound with an internal diameter of 1 cm, such that a wrap density
of 1 coil per centimetre is achieved. The coil should be buried to maximise the heat transfer to
the sample and to minimise energy lost through the free surface. Other methods have also been
used such as adding a section of dry soil, which the coil ignites and smouldering spreads into
the natural sample [55], applying a heated coil briefly to the surface of the sample [39,90], or
using a heat gun until the sample is flaming before smother the flames and allowing
smouldering to propagate [84] . Another alternative in the case of in-depth spread is to expose
the surface to radiant heat flux, as done by Benscoter et al. [64] who exposed natural organic
soil samples to a heat flux of ~6 kW m-2 for 5 min. Notably, this soil sample was relatively dry
so it is uncertain is this heat flux is reliable for other MC.

35
2.3.6 Sensitivity of Experiments
Because of large experimental uncertainty it is beneficial to communicate the sensitivity of
experiments, such that important parameters can be identified and controlled, attempting to
reduce variability and uncertainties between studies. Heat transfer has largely been taken into
account in the construction of the reactors and soil moisture content. For oxygen supply,
experiments have been performed to examine the sensitivity of the smouldering experiments
to changes in the oxygen arrival by changing the wind conditions on the free surface. A
horizontal reactor with 800 g of dry peat was used as a representative case to illustrate the
impact of these factors. The mass loss rate (MLR) was measured by a balance as a powerful
descriptor of the smouldering dynamics to quantify the changes in behaviour.

First, the extraction rate of the hood surrounding the reactor was adjusted, because as it has
been shown that wind has a significant effect on the behaviour on the smouldering dynamics
[69,70,73]. Figure 12 illustrates the difference between using a high versus low extraction rate.
The vertical air flow speed, measured 30 cm above the sample ,was found to be 0.7 and 0.3
m/s for the high and low extraction, respectively. The initial peak in Figure 12 at 0.5 h, is the
result of the coil heater. The mass loss dips briefly as the heat of the coil dissipates. The next
stage of linear increase is the result of the horizontal surface spread, increasing the extent of
the smouldering. The peak is reached once the smouldering front reaches the far wall and
smouldering begins to propagate downwards across the entire reactor. Clearly, the time for
smouldering to reach the far side of the reactor reduces, suggesting an increased horizontal
spread rate, and the peak mass consumption increases with ventilation. This emphasises that it
is important to keep the ventilation consistent throughout the experiment.

36
Figure 12 - Graph illustrating the influence of wind flow on mass loss rate of 800 g of dry peat in the horizontal reactor. At
30 cm above the reactor, the vertical airflow velocity was measured to be 0.7 and 0.3 m s–1 for high and low ventilation
settings, respectively. All lines are smoothed over 30 min and the clouds indicate the range of experimental results.

Small amounts of air could also enter via the coil access points and have a significant influence
on the smouldering dynamic. Three sets of two repeated experiments were conducted with
increasing size of the coil holes. The sizes tested were: 0 mm, where by the wire was simply
punctured through the insulating board, giving minimal space around the wire; 1.5 mm, which
is the proposed standard protocol using an alumina tube with an external diameter of 3 mm; a
10 mm hole, which was used to test the extreme behaviour. The MLR of each experiment is
presented in Figure 13. The size of the coil access points have a marked influence on the initial
peak of MLR by the end of ignition at 0.5 h. The difference in the initial peak is attributed to
the increased availability of oxygen through the coil access points. Beyond this stage, both the
experiments with the 1.5 and 10 mm hole behave almost identically, while the 0 mm reactor
was slower to reach the peak. The explanation for this becomes apparent when examining the
IR footage when the igniter is terminated, as shown in Figure 14. By the end of ignition, for
both the 1.5 and 10 mm coil access point samples the smouldering has breached the surface
across the full width of the sample, at which stage oxygen is predominantly recruited from the
free surface and becomes independent from the small amount of oxygen that arrives through
the coil access points. Smouldering in the 0 mm coil access points samples reach the surface
15 min later before following a similar trend to the other samples. However, for the experiments
with the 10 mm access point, the smouldering reaches the surface above the access points much
earlier. While the 1.5 mm appears to reach the surface simultaneously across the entire width
of the reactor. For this reason the 1.5mm access point is recommended to encourage a strong
ignition while allowing for a uniform smouldering front to develop. The small ceramic inlays
are inserted to limit the accidental growth of the holes while.

37
Figure 13 - Graph illustrating he influence of the coil access point size on the mass loss rate of 800 g of dry smouldering
peat in the horizontal reactor. Three sizes were experimented with; 0 mm, 1.5 mm, and 10 mm. The lines are fitted
against the smoothed average mass loss rates of 3 repeated experiments for every condition, and the clouds indicate
the range of the experimental results.

Figure 14 - Comparing infrared images of sample surface 30 min after ignition of smouldering dry peat in the horizontal
reactor. Both samples with coil access points greater than 0 mm show smouldering reaches the surface across the width
of the box by the end of the ignition protocol.

Sample density influences the oxygen flow through the medium, but also affects the heat losses.
It is known to have a large impact on the behaviour of the smouldering front [56,59,66,91].
Between experiments this variable can change due to how samples are handled. The influence
of this was explored as shown in Figure 15, which compares the average MLR curve of dry
samples that were shaken and unshaken, with an approximate density difference of 10%.
Notably, the horizontal spread rate of the unshaken sample was greater than that of the shaken
peat as evidenced by the earlier arrival of the second peak. This serves to emphasise that it is
valuable to be conscientious when handling samples to limit variability. The MLR of all
experiments examined in this chapter, representing a broad range of possible conditions, are
collated to form a band of possible smouldering dynamics. The upper and lower limit of this
band correspond to the maximum and minimum MLR curves of the experimental dataset. This

38
band serves to illustrate the range of experimental uncertainty (Figure 16). It is important to
emphasise that this is not an exhaustive range, and will be dramatically changed by MC, IC,
and possible different soil types and origins.

Figure 15 - Average MLR plots of dry peat, examining the influence of sample preparation and resulting density changes.
The peak values are shown to remain constant between the experiments, but the second peak of the unshaken sample is
40 min earlier than that of the shaken. The clouds indicate the range of the experimental results.

This exploration into the sensitivity of meso-scale experiments also acts as an illustration of
the ability to target the effect of a single parameter on smouldering dynamics, which can be
used to understand behaviour at the larger scale. On the other hand, although correlational
relationship can be established, it can be challenging to identify the dominant mechanism
controlling the behaviour accurately and confidently, because diagnostic capabilities are
limited. However, computational models can be used in combination with experiments to arrive
at such a conclusion. This fundamental understanding can be applied to the behaviour at the
large scale, and put findings made at the field scale into context and give greater meaning.

39
Figure 16 - ‘Cloud of variability’, encompassing all experimental data presented on dry peat in the horizontal reactor.

2.4 Field Scale

Unfortunately, very few field-scale experiments have been conducted to study smouldering
wildfires, typically studies have focused on gas emissions or plant and soil
consumption/recovery post-fire [12]. Often satellite data is used in the study of fires, although
limited or no information can be derived about the smouldering dynamics from these images
because the coarse resolution (size of one pixel is 500 m x 500 m or 12 km x 12 km) is
comparable to the size of a field experiment.

Emissions have been studied using portable gas analysers, using laser scanning techniques.
Little attempt is made to relate these results to smouldering dynamics in the field scale. The
most typical values calculated from these concentration measurements are emission factors (EF)
and modified combustion efficiencies [92]. EFs describe the quantity of a gas species emitted
per gram of dry soil. Despite that, the mass loss rate of soil cannot be directly measured in a
field setting. However, after measuring the carbon content of the soil, it is assumed that the
total carbon content is emitted and conserved. Thus, after measuring the carbon content of the
soil and the emissions, the emission factors of CO2, CO, and particulate matter can be
determined [92]. Despite providing limited information about the smouldering dynamics these
results are valuable in providing insights into the averaged chemistry as well as the health risks
posed by smouldering soils.

40
Figure 17 - A typical horizontal distribution of a burned surface peat layer with several projections of the fire front line
at Plot-2 on 10 August 2002. Source: [53].

Satellite images have been used to attempt to quantify the extent and depth of soil consumed
to understand the emissions released during a fire episode. Data examining the extent of fire
damage typically use vegetation indices to map a recent fire [93]. Informative studies have
examined the relationship between fire size and environmental conditions such as fuel moisture
or vapour pressure [94,95]. Unfortunately, satellite footage provides little or no means to
distinguish between smouldering and flaming fires, and the ability of satellite IR cameras to
detect smouldering is still questionable due to the low temperature and in-depth spread of
smouldering– therefore study of these fires are very limited. However efforts have also been
made to use drones to collect data using IR and Lidar measurements to measure the depth of
soil consumption [96] by finding the difference in ground level between the surrounding area
and the burned area.

To date, only a single planned smouldering field burn experiment can be found in the literature
[97]. This study attempted to ignite several small plots measuring 0.5 × 0.5 m by filling a
partially buried metal tray with charcoal and igniting the fuel. Unfortunately, the soil tested
was beyond the threshold of sustained ignition described by Frandsen [63] and so the research-
ers failed to sustain ignition. However, this does present a movement towards performing
planned experimental field experiments. Usup et al. [53] studied real accidental smouldering
fires, by investigating fire events in Kalimantan, Indonesia, and examining fuel consumption,
spread rates and soil temperatures. Several samples were taken from near the fire to represent
the fuel before burning (studying fuel properties), while thermocouples embedded into the peat
provided temperature measurements, and iron poles placed known distances apart provided a

41
scale for which spread rate could be estimated. Figure 17 shows the mapping of the horizontal
smouldering spread from this study. It can be seen that the smouldering front moves in
seemingly arbitrary directions, looking for the conditions most conducive to smouldering. Such
field experiments provide a valuable insight into smouldering behaviour in natural conditions,
and may provide a basis for informed land management strategies.

2.5 Conclusion
Ultimately, the field-scale smouldering fires are the most relevant but the most challenging to
study. The large variability both locally and globally makes garnering an understanding of
smouldering dynamics and making predictions on future behaviour a significant challenge. It
is clear that smouldering fires are, in physical and chemical terms, complex and occur at a
range of scales. Research is needed at all scales to advance the understanding of the
phenomenon. The behaviour of smouldering fires is dependent on the micro-scale chemistry,
influenced by the meso-scale dynamics and highly variable across the field scale.

Although limited in its application, the micro-scale provides necessary understanding about the
fuels kinetics, providing the needed parameters to model the behaviours. More studies at this
scale will help improve the accuracy of computer models, which, combined with meso-scale
experiments, will provide key insights into fundamental smouldering wildfires. Meso-scale
experiments, such as the ones presented in this chapter, allow for study of smouldering
dynamics and control of variables such that generalised behavioural correlations can be
established. Although meso-scale experiments do not accurately replicate the natural
environment, they clarify the interaction between variables. It is suggested that the
methodology presented here should be adopted to enable research in different locations to be
compared, and so synergistically progress this field of study. Additionally, field-scale data pro-
vides necessary benchmarking and information about realistic behaviour. A greater number of
studies at the large scale are necessary for geographical comparisons and to provide sorely
needed qualitative observations that can steer the research at smaller scales.

As illustrated, an integral understanding of smouldering fires is possible only by studying the

phenomena on each of the three scales. This understanding of the controlling mechanism gov-
erning smouldering wildfires may lead to improved land management, better use of fire-
fighting resources, and reduce the global burden of these wildfires.

42
Chapter 3: Experimental Methodology
Summary
This chapter describes the experimental methodology used in this thesis. It can broadly be split
between two different set-ups: the shallow reactor and the Experimental Low-temperature
Smouldering Apparatus (ELSA). Both rigs are novel and have been designed to accomplish
unique tasks. The shallow reactor, measuring only 1.6 cm deep, allows for rapid
experimentation and enables a large set of soil and environmental conditions to be explored.
As such, the influence of MC, IC, density, wind and slope, were studied using this set-up.
ELSA, on the other hand, was designed for the purpose of studying the smouldering dynamics
of Arctic wildfires by studying the influence of soil temperature. The following sections details
the sample preparation procedure used for all experiments as well as outlines the experimental
rigs used and the specific conditions studied.

3.1 Sample Material and Preparation

The sample preparation procedure used is the same as presented in Chapter 2. In all experiments
commercial milled Irish Sphagnum peat (Shamrock Irish Moss Peat, Bord na Mona
Horticulture) was used, due to its batch consistent properties in the long term, ease of
accessibility, and frequent use in scientific literature [56,57,70,82]. The peat contained
54.1/5.1/1.3/0.5% mass fraction of C/H/N/S respectively and had an inherent IC (IC0) of 2.5 ±
0.6 %, measured through the use of a muffle furnace by heating the sample to 1000 ºC and
recording the mass every hour until there was no mass lost between subsequent measurements.
Before conditioning, peat was dried in an oven at 80 ºC [82] until no mass loss was seen in
measurements 6 h apart. Where the IC was to be altered, sand was added to the dry peat and
vigorously mixed. The sand used was fine, washed, commercial sand (Hanson Play Pit Sand),
similar to that used in [82]. MC was then altered by adding the desired quantity of water to the
peat and vigorously mixed again. The sample was left in a sealed bucket for a period of at least
24 hours, to allow the peat to absorb the moisture and to homogenise. Immediately before use,
the conditioned peat would again be vigorously mixed and a 100g subsample was removed and
dried to measure the final MC of the sample. Both IC and MC are calculated on a dry basis
(Eqs 1 and 2).
43
 𝑚𝑤
𝑀𝐶 = 𝑚 % (1)
𝑝 +𝑚𝑠

𝑚𝑠 +𝐼𝐶0 𝑚𝑝
𝐼𝐶 = % (2)
𝑚𝑝 +𝑚𝑠

Where mw is the mass of the water (g), ms is the mass of the added sand (g), IC0 is the inherent
inorganic content of the unconditioned peat (%), and mp is the mass of the unconditioned peat
(g).

3.2 Shallow Reactor

3.2.1 Reactor
Chapter 4 to 6 uses a novel shallow reactor as briefly referenced in Chapter 2. The shallow
reactor is an open-top reactor with an internal dimension of 40 × 40 × 1.6 cm (Figure 18). The
reactor was built using insulating ceramic boards (k=0.7 w/mK, ρ=310 kg/m3, cp=1090 J/kgK).
Due to the shallow nature of the reactor the dried peat was sieved at 8.5 mm to remove
excessively large material, improving the homogeneity of the fuel and sample depth uniformity.
Samples were prepared by filling the reactor with conditioned soil and levelling the surface of
the sample to the rim of the reactor walls, removing any excess material without compacting
or tamping, thus creating a flat and repeatable sample thickness of 1.6 cm. The mass of the
sample was measured using a mass balance (±0.1 g) so that the bulk and organic density could
be calculated. The organic density (ρo) is the density of combustible material in a sample and
was calculated using Eq. 3; where ρb is the bulk density and is found by dividing the sample
mass by the volume of the reactor (2560 cm3).

44
Figure 18 - Illustration of the experimental setup, displaying a central ignition using a wound coil and key features observable
in IR imaging.

1−𝐼𝐶
𝜌𝑜 = 𝜌𝑏 𝑀𝐶+1 (3)
The samples were ignited at the centre of the reactor using a helical coil, 1 cm long, with a
diameter of 1 cm, and made of a 0.5 mm thick nichrome wire. 20 W were supplied to the sample
through the coil for 30 min. This is a greater flux than used in literature, which is typically 100
W for 30 min across an 18 cm helical coil [57,70]. For this reason, the ignition protocol can be
considered a reliable and strong ignition source, sufficient to ignite all samples studied. The
experiments, from ignition to burnout, lasted between 2 and 5 h. Smouldering spread was
monitored by taking images with an IR camera (FLIR A655sc) at a frequency of 1 frame per
minute, and was placed directly above the centre of the reactor. From these images a leading
and trailing edge could be clearly observed as shown in Figure 18.

3.2.2 Soil Conditions

Three soil conditions were explored using the shallow reactor: MC, IC and density. MC and
IC were explored together in three series: MCx, ICx and mixed. The MCx series consisted of
samples with no additional IC added, and target MC of 0 to 160 % in increments of 20 %, with
an addition set at 110%. The ICx series was modified by solely altering the IC, while
maintaining a minimum MC (< 5%). Seven IC values were explored between 2.5 (IC 0) and
90 %. Finally, the mixed series, had both IC and MC modified. The IC was fixed at 40 %, and
MC between 0 and 80% in increments of 20% as well as an additional set at 70 % was prepared.
Each soil condition was studied 3 times, resulting in a total of 63 experiments. The organic

45
densities measured here matched well with the samples presented by Huang and Rein [59]
(Figure 19a). The organic density decreased with MC as shown in Figure 19a, indicating that
the peat swells with the addition of water, while it is relatively insensitive to IC up to 75% after
which it sharply decreases. This suggests the sand is accommodated in the pore space between
the peat particles (Figure 19b) at low ICs. Notably, the organic density of the Mixed series (40%
IC) decreases more rapidly than the MCx series (2.5% IC). This is because MC is calculated
as a percentage relative to the dry mass (Eq. 1); which as the IC increases, the proportion of
organic matter decreases. Thus, at the same MC, each unit of organic matter has more moisture
available at higher IC, enabling greater swelling at lower MC.

Figure 19 - Measurements and the relationship between organic density and soil properties, a) showing the influence of
moisture content and compare to result presented by Huang et al. 2016 [70] b) the effect of inorganic content on the organic
density. Error bars are calculated from measurement uncertainties.

The influence of density on smouldering dynamics was studied in samples of 80% MC with no
alterations made to its IC. 80% MC was used as it represents a realistic MC, and provides a
sample that allows for a large range of achievable densities due to its swelling. Variations in
density was achieved by compressing the peat to varying degrees. Five levels of compression
were targeted, from 0 to 40% in increments of 10%. 40% compression was the maximum
attainable density increase. Compression was calculated relative to the density of an
uncompressed sample (157.5 kg/m3). A metal plate was used to uniformly compress the sample
varying distances to achieve the desired degree of compression. The soil was then levelled with
the height of the reactor. The mass of the sample was measured and the sample density was
calculated to measure the achieved compression. If density varied more than 5% from the
targeted value, the reactor was emptied, and the process was repeated. Three experiments were
conducted for each compression condition. The average MC, density and organic density of all
soil condition series are provided in Table A1.1 for posterity.

46
3.2.3 Wind and Slope
21 experiments were conducted with wind and slopes as the main variables and compared to
that from experiments with no wind and slope. For wind, MCs of 0, 40, 80, and 120% were
studied. This covers the span of MC that can self-sustain smouldering, as will be shown in
Chapter 4, and allows for comparison with results found from the MCx series with no wind or
slope. For experiments on slope, an 80% MC was targeted. All samples were within 5% of
their targeted MC. Each MC condition was repeated 3 times, to capture natural variability and
reduce uncertainty. The reactor was filled with the conditioned soil and the mass of each sample
was recorded. The bulk density (b) and organic density (o) were then calculated as before.

Wind was applied along the surface of the peat, using two, 30 cm wide, tangential fans placed
on top of each other. The fans were placed 20 cm from the edge of the reactor to allow the
boundary layer to develop. Characterization of the flow was performed by measuring wind
speeds between 10 and 60 cm from the fan, 0 to 15 cm from the centreline of the fan, and at
several heights using a hotwire anemometer (Figure A1.1). The measurement of smouldering
dynamics were limited to within 14 cm of the ignition where wind speed was found to be
consistent. The wind speed measured within the region considered for study (30-50 cm from
the fan; 0 and 7.5 cm from the centreline, 1 cm above the surface, illustrated in Figure A1.2)
averaged 0.86 ± 0.12 m/s. This is a similar order of magnitude to smouldering experiments in
literature [70,73]. Higher wind velocities resulted in smouldering particles of peat being carried
forward altering the spread mechanism. Only the maximum wind speed was selected as it
would have the greatest effect on the smouldering dynamics. The wind was applied after the
end of the ignition protocol.

For the study of slope, the reactor was inclined using a lab jack, and the slope was measured
using a protractor (Figure 20). Three slopes (), 10º, 20º, and 30º relative to horizontal were
studied and compared to spread in a horizontal configuration (=0º) from the MCx series. Each
inclination was performed 3 times, to capture natural variation of the peat bed and uncertainties.
The average MC, density and organic density of samples studied under these conditions are
also provided in Table A1.1 for posterity.

47
Figure 20 - Illustration of experimental set up for studying the influence of slope on smouldering dynamics. When when θ s is
equal to 0 and 180°, smouldering spreads uphill or downhill, respectively.

3.2.4 IR Image Analysis

The key features observed in the IR images were the leading and trailing edge. Measurements
of the leading edge provided the horizontal spread rate while the trailing edge was used to
calculate the in-depth spread rate by dividing the thickness of the soil bed (H) by the
smouldering burning time (tB). I define burning time as the time difference between the trailing
edge and leading edge reaching any given location. The assumption made in this calculation is
that all of the fuel is consumed. However, in all cases where the sample successfully ignited,
the residual mass at the end of the experiment was within ~5% of the mass due to the IC.
Horizontal spread rate in the slope experiments refers to spread in the same direction as the
surface of the peat, while in-depth spread is perpendicular to the surface (into the bulk of the
material).
𝐻
𝑆𝑑 = 𝑡 (4)
𝐵

It is also important to note that, the bottom boundary influences the in-depth spread, but the
trend in behaviour is valuable for insight and understanding the fundamentals of smouldering
dynamics. Additionally, the shallow depth of the reactor minimized the influence from the
effects of overhang as smouldering is maintained near the surface.

48
 Figure 21 –A cross section illustrating the key features of smouldering dynamics and framework of analysis.

For analysing the spread, the IR images were converted to gray scale and in the case of
experiments studying the effect of slope, the images were corrected for perspective by
identifying the corners of the reactor and remapping the image such that the corners make up
a perfect square as they do in reality. This allows horizontal distances to be measured accurately
without further correction. 60 equally spaced transects were drawn radially from the centre of
the ignition coil. Each transect returned a pixel intensity profile similar to the example profile
in Figure 22. Figure 22 also demonstrates how this pixel profile changes with time. These pixel
values were normalized to the range 0 to 0.15 W/cm2-sr, and an arbitrarily low threshold value
of 0.2 was selected and used to define the location of both the leading and trailing edge. Their
movement was tracked every 10 min. The results are insensitive to the specific threshold value
due to the steep gradients of the intensity profile. The value 0.2 was selected as it was suitably
close to the minimum heat flux measured by the IR camera to capture the smouldering fronts,
while also high enough to reduce the influence of noise. The scale of each pixel was defined
by dividing the real length of the reactor (40 cm) by the length of the reactor in pixels. This
gave a scale of approximately 0.8 mm/pixel.

49
Figure 22 - Illustrating the method of IR image analysis using example images from a sample with 40% MC and 2.5% IC. The
plot present the pixel intensity profile from the same transect at 40, 80 and 120 min after ignition (top images) and how using
a normalized pixel intensity value of 0.2 was used to track the leading edge (grey line).

Although correcting for perspective allows for horizontal spread rates to be measured without
further adjustment [85] some corrections were required to calculate the in-depth spread rates
on a slope. This is because in-depth spread is measured by measure the time difference between
the leading and trailing edge passing a given location. Because the leading edge is measured
on the surface of the peat, and the trailing edge is measured on the base of the reactor. These
points are offset from each other vertically introducing an error in recording burning time (tb)
error due to the perspective, as illustrated in Figure 23. This error is corrected by Eq 5, which
uses the measured smouldering width and  to quantify the error in the burning time and give
a corrected burning time (tb,corrected). Where β is calculated using Eq. 6, and is the angle between
any transect and the horizontal plane such that  =  when s = 0º (uphill, as illustrated in
Figure 20), and  = 0º when s = 90º (sidehill). DO is the observed distance spread, DT is the
true distance spread, and t is the time.
𝐻 𝐻
𝑆𝑑,𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑒𝑑 = 𝑡 = 𝐻 (5)
𝑏,𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑒𝑑 𝑡𝑏 ×( tan(𝛽))
𝑊

50
 𝛽 = sin−1(cos 𝜃𝑠 sin 𝛼) (6)

Figure 23 - Illustrating the source of error in calculating in-depth spread rate from spread on a slope. Top illustration is the
time at which the burning time is started, bottom is when burning time is terminated.

Although the method used here allows for the measurement of transient behaviour, it was found
that the self-sustaining spread rates was steady in all cases, as shown in Figure 24, indicating
that average values are a good representation of the results and will therefore be used in the
discussion. This is true for wind and slope conditions as well as shown in Figure 25. Small
fluctuations of spread rate during the steady state spread were observed and is likely due to
natural variation in soil properties or airflow. The results of the horizontal spread rate were
averaged between 35 min (5 min after the end of ignition) and the time at which the
smouldering front reached 17 cm from the coil (3 cm from reactor walls) to avoid any influence
of the boundary. For experiments with wind, spread rate analysis was limited to within 14 cm
of the center, as this is where the wind speed was found to be consistent, as described earlier.
In the case of in-depth spread, which considers both the leading and trailing edge, values are
averaged from when the trailing edge passed 2 cm from the coil to avoid any sensitivity to the
ignition or IR signal from the hot coil.

51
Figure 24 - Illustrating the independence of both leading and trailing spread rates with time. Figure a) shows the distance
from the centre of the leading and trailing edge with time and figure b) exemplifies the consistent spread rate of the leading
and trailing edge, allowing for averaged values to be used for discussion. Plots are generated from an experiment with 40 %
MC and 40% IC. The clouds represent the standard deviation across all 60 transect

Figure 25 – Presenting examples of the consistent spread rate in all directions relative to a) wind and b) slope. Notably,
wind is applied after the end of the ignition protocol, as evidenced by the increase after 0.5h.

52
3.3 Experimental Low-temperature Smouldering Apparatus
(ELSA)
3.3.1 Experimental Set-up
The effect of cold soil temperatures on smouldering dynamics was explored by conducting 2
series of small-scale experiments using a novel rig. The rig was designed to control 3
parameters relevant to the soil temperature in arctic regions: the ambient air temperature, the
temperature of the bottom boundary of the reactor in which the peat was burned to represent
the temperature of deep soil layers, and the soil temperature.

To control the air temperature a large chamber with an internal dimensions of 120 × 80 × 80
cm was constructed using vermiculite and lined with a fibreboard insulating material (Figure
26). A 40 × 40 cm extraction hood was mounted through the top of the chamber with a variable
speed, mechanical fan installed to control the extraction rate of the smoke produced from the
smouldering samples. A fine metal mesh (0.5 mm) skirt extends down from the hood with an
access door, as illustrated in Figure 26. The skirt reduced any significant air currents from
developing which would affect the smouldering dynamics. A platform was constructed to
support the reactor such that experiments could be performed from within the mesh skirt. A
compressed airline was passed through a vortex tube, which is a device that allows for the
manual adjustment of air temperature. The vortex tube directed air through an inlet at bottom
of the chamber and allowed for some adjustment of the inlet air temperature. This, in turn,
offered some control of the chamber temperature, which was measured by a thermocouple, TC,
placed within the mesh cage at the same height as the top edge of the reactor. As a result, the
chamber temperature could be varied from 20°C down to 10°C (Figure 27) by adjusting the air
inlet temperature down -7 ± 2°C. The air inlet temperature was measured by the thermocouple
(Ti) at the inlet as illustrated in Figure 26.

The reactor in which peat was burned was a modified version of the horizontal reactor
presented in Chapter 2 and is named the cold reactor. It is an open top reactor with internal
dimensions of 20 × 20 × 10 cm. The walls were constructed from an insulating fibreboard
(k=0.7 w/m·K, ρ=310 kg/m3, cp=1090 J/kg·K). The bottom of the rector is a thin aluminium
plate so to allow for the effective conduction of heat. To control the temperature of the bottom
boundary of the reactor it was placed onto a large aluminium base (22 × 22 × 2.5cm) with an

53
inset tubing through which a temperature-controlled water – glycol mixture was pumped. The
temperature of the glycol could be varied and maintained by a benchtop recirculating chiller
(Polyscience LS-series compact chiller). A water-glycol mixture was used to allow of below
zero liquid temperatures could be used. A thermo-couple was mounted to the base to measure
the temperature (Tb) as indicated in Figure 26 to ensure that the base was maintaining the
desired temperature.

Where the initial temperature of sample was required to be altered, the reactor and sample was
placed in a fridge or freezer for at least 12 h before starting the experiment. The samples were
ignited using an 18cm, helically wound nichrome coil, 1 cm in diameter through which 100 W
of power is supplied, which has been shown to be a strong ignition source as described in
Chapter 2 [77]. A scale is used to measure the mass of the sample every minute. 16
thermocouples were inserted along the centreline of the reactor at 3, 7, 11 and 15 cm from the
ignition side and at 2, 4, 6 and 8 cm depths, as illustrated in Figure 26, to capture the
temperature development through the sample. This is a similar thermocouple array to that used
in [70]. Both IR (FLIR Duo R) and visual cameras (GoPro) were mounted above the reactor to
capture the behaviour and spread of the smouldering. A smaller IR camera than used for the
shallow reactor was required such that it could be mounted within the hood.

Figure 26 - Illustration of the Experimental low-temperature smouldering apparatus (ELSA). On the left shows key features
of the ractor which is filled soil. The internal dimension of the reactor measures 20 × 20 × 10 cm. On the right, key features
of the rig is highlighted. The dimensions of the chamber is 120 × 80 × 80 cm.

54
3.3.2 Soil Conditions
To explore the effect of soil temperature on smouldering dynamics, 3 temperature conditions
were targeted: below zero, near zero, and room temperature. This covered a range of
temperatures from frozen conditions to summer conditions in Arctic regions, where the active
layer is warm [98]. Experiments with MC of 50% and 100% were conducted at all three
condition temperatures, with one to three experiments conducted per soil condition. The base
temperature (Tb), chamber temperature (Tc) and soil temperature (T0) were recorded for all
three condition temperatures (Tcond) and are presented in Figure 27. The initial soil temperatures
were set by placing the samples in a freezer at -15°C, a fridge at 2°C or in a sealed box and left
at room temperature (20°C), for 12 h before ignition as mentioned previously, to obtain
homogenous soil temperatures in the three target temperature regions described above. The
initial soil temperature of the samples were measured by averaging the temperature of all 16
sample thermocouples until t =1 min of the experiment (before any thermal effects of ignition
are felt). These values are shown in Figure 27 as solid markers at t=0. The base temperature
and chamber temperature were averaged for each condition and summarized together with the
initial soil temperature in Table 1. The average base temperature will be used as the reference
temperature for describing each condition. The condition temperatures will be referred to from
here on as -7, 2, and 21°C. It is evident that the deviation between base temperature, chamber
temperature and initial soil temperature increases with decreasing condition temperature.
However, each condition temperature achieves a distinctly different temperature in alignment
with the desired target temperature range (below zero, near zero, and room temperature).
Additionally, the difference in air temperature and conditions temperature is not expected to
have a significant impact on the results of the experiment as the specific heat capacity of air is
small, and during the experiment the incoming air is mixed and heated by the rising emissions
from smouldering before reaching the fuel.

Table 1 - Summary of average temperatures for the three key temperature conditions: -7°C, 2°C, and 21°C. The uncertainties
reported are the standard deviation of measured temperature across all experiments at the given condition temperature.

Condition temperature (Tcond) -7°C 2°C 21°C

Base temperature (Tb) -6.8 ± 0.6°C 2.1 ± 1.3°C 20.7 ± 0.3°C
Chamber temperature (Tc) 10.1 ± 1.4°C 9.5 ± 2.2°C 21.8 ± 1.5°C
Initial soil temperature (T0) -11.3 ± 1.5°C 3.8 ± 0.4°C 18.0 ± 0.1 °C

55
Figure 27 - Average temperatures of the base temperature, chamber temperature, and initial soil temperature for all 3
experimental condition temperatures (Tcond). The error bars show the range of measured values across all experiments.

3.3.3 Spread Rate Analysis

Spread rates were measured in two ways. The first using thermocouples to monitor the
movement of the drying front by using the 100°C threshold as the marker and dividing the
distance between subsequent thermocouples by the time difference between the arrivals of the
threshold temperature. This method has been used in literature and described as an effective
way of monitoring spread rate [57,70]. This method has the benefit of resolving spread rates
with depth. The second is using the IR images, which capture the radiate heat emissions from
the burning soil to measure the surface spread in a very similar manner to that described for
the shallow layer.

The IR images are first gray scaled and adjusted for perspective, before 7 evenly spaced
transects are made in the direction of spread, an example of this is presented in Figure 28. Each
transects return pixel intensity along the transect as demonstrated in Figure 29a. Using a
threshold value of 200 to mark the leading edge the position and spread rate can be calculated.
Gray-scale images have pixel values between 0 and 255 and so are non-dimensional. Notably,

56
the camera used is designed for measuring radiant emission from objects up to 250°C
(assuming black body emissions) and so become quickly saturated (as is evident by the plateau
around a pixel value of 240), however as only the leading edge is of interest, this is sufficient.
The threshold was selected such that it was suitably above noise. The results are insensitive to
the specific value used, due to the steep increase in pixel intensity coinciding with smouldering
front. Figure 29b is an example of the measured position of the leading edge along all 7
transects. A clear linear slope is evident, implying a steady spread of the smouldering front.
The vertical dotted lines indicate the time between which spread was averaged. As the trend is
suitably linear, the difference in position between the dotted lines is divided by the
corresponding difference in time to yield an average spread rate. The specific times over which
spread was calculated was manually selected for each experiment. As the IR measures the
position of the smouldering front by observing the surface, the spread rates account for the
changes in burning depth, as opposed to thermocouples which are fixed at certain depths.

Figure 28 - Demonstrating the method of capturing the spread rates using IR imagery. Left image shows an example grayscale
image. Middle image is after the image orientation has been corrected, and the transect lines have been drawn. The width and
length of the reactor are used to calibrate the scale of the pixels. The image on the right demonstrates the pixel saturation
associated with the high radiative heat emissions from burning of peat, and the clear definition of the leading edge (spreading
from top to bottom).

57
Figure 29 - Demonstrating the analysis method in measuring spread rate from IR imagery. a) depicting the grayscale pixel
intensity along a transect at 110 and 350 min after ignition b) shows the measured position of the smouldering front across 7
transects with time, and illustrates the steady horizontal spread rate of the smouldering front.

3.3.4 Characterization: Influence of Smoke Extraction Rate

A set of preliminary experiments were conducted first to characterize the effect of the chamber
and the extraction rate. Experiments were conducted with sample of 50% MC at room
temperature outside the chamber, as well as inside the chamber with two different extraction
rates: 1 and 2 m/s. These extraction rates are similar to that used in [57,86]. In all cases the
base temperature and the initial soil temperature was conditioned to room temperature. Two
experiments were conducted in each condition.

The air flow within the chamber was observed to be more quiescent than outside, regardless
of the extraction rate. This was observed by the lack of perturbation to the emissions.
Particularly in the early stage where the smoke remains attached to the surface of the peat.

Figure 30 shows the averaged mass loss rate of each characterization condition. It can be seen
that all conditions behave reasonably similarly. Examining the peak temperatures it was found
that they were, on average, 7% and 13% higher for extraction rate of 1 and 2 m/s compared to
that of open combustion, respectively. Using the temperature data, spread rates were calculated
as done in [57,70]; by dividing the horizontal distance between thermocouples (4 ± 0.5 cm) by
the difference in time of between thermocouples reaching 100°C as described in the previous
section. In both cases the spread rate within the chamber was lower than that in open air (Table
2); with 1 m/s extraction rates resulting in an average spread rate 16% less than in open
combustion compared to 23% less at 2 m/s. However, the spread rates are reasonably similar
between the two extraction rates, with a slightly improved performance in the 1 m/s extraction

58
rate. Considering the above it appears that the results are largely insensitive to the extraction
rate with an extraction speed of 1 m/s seeming marginally preferable. For this reason it was
used in all following experiments.

Figure 30 - Mass loss rate influenced by varying extraction conditions. The cloud demonstrate the range of repeat results.

Table 2 – Summarizing average spread rates at varying extraction rates and comparing to that in open combustion

Horizontal Spread rate (cm/h)

Depth
Open 1 m/s 2 m/s
(cm)
2 4.37 2.99 3.13
4 3.72 3.32 3.14
6 3.61 2.82 2.88
8 3.44 3.64 2.47
Average 3.79 3.19 2.90

59
Chapter 4: Influence of Moisture and
Inorganic Content on Smouldering
Dynamics
Summary2
In this chapter, I systematically conducted 63 experiments on multidimensional smouldering
spread, studying the individual and combined influence of two key factors affecting
smouldering in organic soils: moisture content (MC) and inorganic content (IC). Both
horizontal and in-depth smouldering spread were investigated using a novel shallow reactor.
This shallow depth allows a greater number of experiments to be performed in a short period
of time compared to deeper samples. Horizontal spread was found to decrease linearly with
moisture content (R2 > 90%); while in-depth spread rate increased linearly up to 300% in
samples with a moisture content of 140%. Increased inorganic content linearly decreased the
horizontal spread rate but had little influence on in-depth spread in drier samples. Interestingly,
in wetter samples, in-depth spread was in fact sensitive to inorganic content. A novel approach
combining horizontal and in-depth spread rates as vector components, revealed that the global
spread is independent of moisture content, with an average spread rate of 8.7 and 8.4 cm/h for
2.5 and 40% IC, with changes in direction according to moisture content; going in-depth for
wet soils, and horizontally for dry soils. Similarly, increasing the IC encouraged downward
spread for wet samples. I also report observations of a bifurcation of horizontal spread, where
spread would locally extinguish where the in-depth spread was greater than the horizontal
spread. These findings provide previously unknown insight into the relationship between
horizontal and in-depth spread in smouldering fires, ultimately improving the fundamental
understanding of such fires.

2
This chapter is based on: E. G. Christensen, N. Fernandez-Anez, G. Rein, Influence
of soil conditions on the multidimensional spread of smouldering combustion in
shallow layers, Combustion and Flame, 214 (2020) 361-370. Doi:
10.1016/j.combustflame.2019.11.001
60
4.1 Introduction
Research to date has found that smouldering spread is controlled by two key mechanisms: heat
transfer and oxygen supply [99]. Moisture content (MC), and inorganic content (IC), are
considered two of the most influential factors in peat fires, affecting the susceptibility of soil
to ignition [39,63,65] as well as smouldering dynamics [10] by influencing these two
mechanisms. MC is the measure of the amount of water in the soil, while IC is the quantity of
non-combustible material in the soil (both expressed as a percentage relative to the dry soil
mass). IC varies between soil types and location. Smouldering wildfires most commonly occur
in peat, which is defined as any soil with an organic content greater than 20%, thus the
inorganic content of such soils can range between 0 to 80% [22]. Moisture content is a highly
variable soil condition both spatially and temporally. MC is known to vary from over 300%
down to below 20% in drought scenarios [74]. Prat-Guitart [76] demonstrated the significant
spatial variation in soil MC by studying 8 plots of 1.5 x 1.5m and found MC could change at a
rate of 10% every cm. The presence of tree roots which absorb water and tree canopies limiting
rainfall onto the soil floor has been shown to cause smouldering to preferentially burn round
the base of trees [12,72,100]. Such observations demonstrate the significant impact MC has on
smouldering severity and spatial distribution of fuel consumption, emphasizes the importance
of understanding its influence on smouldering dynamics. MC influence on smouldering spread
is explained by its effect on heat transfer and heat losses. Smouldering spread propagates both
horizontally and in-depth [10] (Figure 1), where horizontal spread (Sh) is along the soil layer
and in-depth spread (Sd) is into the bulk of the material.

Figure 31 - Illustration of the stages and key features of smouldering. Horizontal spread is along the peat layer, and in-depth
spread is downwards. The vector combination of the two spread components give the global spread rate (S G) with a direction
(ф). The leading edge is the first location of char oxidation and the trailing edge is the last location of char oxidation.

61
Horizontal spread is known to decrease with moisture [22,56,57,101], owing to the latent heat
of water evaporation requiring energy and time to dry the soil. Additionally, in wet peat,
horizontal spread rates are faster below the surface than at the top. This results in a phenomena
known as overhang, where deeper layers of peat are consumed more rapidly than those closer
to the surface, creating a section of peat that overhangs a void. Eventually, the overhang loses
its structural integrity and due to its weight, it collapses and accelerates the burning.
Smouldering propagation is then characterized by periodic cycles of overhang formation and
collapse [70]. The depth of this overhang increases with MC [70] until a critical MC is reached,
and smouldering can no longer be sustained because the energy required to dry the soil is
greater than that provided from the char oxidation [39]. The influence of depth is complex and
still poorly understood. In contrast to horizontal spread, in-depth spread increases with MC as
observed experimentally in [59]. Huang and Rein [59] explained that this is because of the
improved oxygen supply due to swelling. IC, on the other hand, has no influence on in-depth
spread, according to both experiments and modelling [62,102] while horizontal spread is
decreased [61]. Notably, as the fire moves downward, in-depth spread rate decreases with depth
due to the accumulation of ash above the oxidation zone reducing the oxygen supply [59,62,89].

As illustrated above, all research to date has focused on horizontal and in-depth spread
individually. The influence of both MC and IC, the two most significant soil condition in the
context of smouldering spread, are studied in combination and individually in a comprehensive
study. In this paper, for the first time, these spread components are combined to understand
the dynamics of a single global spread rate. A novel experimental set-up is used, making it
feasible to conduct a much larger number of experiments, due to the reduced experimental time
required, and gain insights that are not possible with deeper reactors.

4.2 Method
Experiments were conducted using the shallow reactor, as described in Chapter 3. To explain
briefly, commercial peat was dried before modifying it through the addition of water and sand
to achieve a desired the MC and IC respectively. A reactor with internal dimensions of 40 × 40
× 1.6 cm and was constructed from an insulating material and filled with conditioned soil. The
surface of the sample was levelled with the edge of the reactor by scraping away excess material
to create a consistent sample depth. The samples were ignited at the center, and spread was

62
measured by analysing IR images from an IR camera mounted directly above the center of the
reactor.

Experiments were conducted in 3 series: MCx, ICx, and mixed. The MCx series was for the
study of the influence of MC on smouldering spread. It consisted of samples with MC between
0 and 160% in increments of 20% with an additional set at 110% and an IC of 2.5% (inherent
IC). The ICx series was for the study of the effect of IC on smouldering dynamics. The IC was
varies from 2.5%, 20%, 40%, 60%, 75%, 80% and 90% in dry samples (MC<5%). Finally, the
samples in the mixed series, were altered to have a fixed IC of 40% and MC between 0 and
80% in increments of 20% as well as an additional set at 70 %. This allowed for a comparison
between the MCx and mixed series: two series with a fixed IC and varying MC. Each soil
condition was studied 3 times, resulting in a total of 63 experiments.

4.3 Results and Discussion

4.3.1 Smouldering Width
The distance between the leading and trailing edges, called the smouldering width (Figure 31),
was found to vary with soil conditions. Figure 32a, reveals that with increasing moisture the
smouldering width shortens. Increasing the IC of the soil from 2.5 to 40% results in a systematic
decrease of the smouldering width, on average, by 1.7 cm. At low MCs (Figure 32b), the
increase of IC bi-linearly decreases the width of the smouldering front; with the first linear
trend between zero and 60% IC, followed by a steep decrease in slope from 60 to 90% IC. The
widths drop to zero once spread could no longer be sustained as the sample contained too much
MC, IC, or a combination of both, crossing a critical threshold [39,63,65].

63
Figure 32 - Shows the change in smouldering width as affected by a) moisture content for two inorganic contents and b)
inorganic content for two moisture contents.

4.3.2 Horizontal Spread Rate

Horizontal spread measurements are presented in Figure 33a. Apart from the behaviour of peat
with both low MC and IC, a linear downward relationship was found between horizontal spread
rate and MC (red line; R2=0.94). Increasing the IC of the soil to 40% maintains the linear
relationship with moisture (blue line; R2=0.90) (Figure 33a), but resulted in lower spread rates,
because IC acts as a sink of sensible heat while not contributing to combustion.

Similar results were noted for the ICx series (Figure 33b) where a clear linear relationship
between horizontal spread rate and IC was observed (black line; R2=0.99) [61]. This linear
decrease in spread rate is attributed to the heat capacity of sand. Moisture can be seen to shift
this trend downwards. Figure 33c, provides a visual representation of the spread rates
throughout the variable space, illustrating the trade-off between IC and MC. The soil conditions
at which spread rate drops to zero define a threshold of ignition, beyond which smouldering
cannot be sustained. The values found here are higher than that found by Frandsen [39,63], and
on the upper end of what was computationally predicted by Huang et al. [65].

This decrease in horizontal spread rate has been noted in literature [22]. Good agreement exists
between current results and horizontal spread rates reported for dry peat (MC <10%) [70],
despite the deeper samples (10cm). This is likely because at low MCs, smouldering spreads
faster along the surface, while at high MCs the overhang phenomena occurs and influences the
spread rate [101]. Hu et al. [57] recently reported a linear decrease in horizontal spread rate
with MC using the same peat as in this study. The slope is almost identical to the one reported
64
here. However, the spread rates reported here are 2 to 3 times greater than Hu et al. [57],
possibly because of their deeper samples. Santoso et al. [22] also collated horizontal spread
rates from the literature and found that for MC > 10%, spread rate averaged ~3 cm/h with a
negative relationship between spread rate and MC. Notably, this comes from a range of
different soils, techniques and measurements conducted at a variety of depths. The significant
difference between spread rates reported here and those found in the literature may be due to
the shallow design of the reactor, which minimizes the heat losses to deeper layers of soil and
limits the formation of overhang. This emphasizes the importance of experiments in deep layers,
and although valuable fundamental understanding can be found in the study of shallow layers,
it is important to complement findings with corresponding research in deep layer setups.

Figure 33 - The influence of soil properties on horizontal spread rate: a) The effect of MC on horizontal spread rate for soils
with two IC values, b) the effect of IC on the horizontal spread rate for two different MC values c) plot of the influence of both
MC and IC on horizontal spread rate. (Slopes: 2.5% IC, -0.026cm/h·%MC, omitting the first data point; 40% IC, -0.027
cm/h·%MC; 0% MC, -0.106 cm/h·%IC)

4.3.3 In-depth Spread Rate

In-depth spread rates were measured and are reported here for the first time alongside
horizontal spread rate values for the same experiments. This is a major advantage of the shallow
65
reactor design; providing further insight into the dynamics of smouldering. Converse to
horizontal spread, in-depth spread rate increased with MC, up to tripling the spread rate from
0 - 140% MC (Figure 34a). This agrees with results previously reported in the literature by
Huang and Rein [59]. However, in-depth spread rates measured here are 2 to 3 times larger
than Huang and Rein [59] at a depth of 5 cm. This is again likely to be due to the shallow depth
of the reactor.

Here I also show the influence of IC. The increase of IC from 2.5% to 40% resulted in higher
spread rates across all MCs apart from MC <10%. Figure 34b shows that at low MC, in-depth
spread rate is insensitive to the changes in IC, except for IC > 75% where it steeply increases.
This independent relationship is in agreement with [62]. Yang and Chen [62] found no change
to the in-depth spread rate of the oxidation front with changing IC, but noted that the spread
rate of the pyrolysis front decreased linearly. Here, at a MC of 40% (Figure 34b), a difference
in spread rate does appear with increasing IC. This suggests that in-depth spread is insensitive
to IC at low MCs, but increases in sensitivity with MC.

Figure 34 - Influence of soil properties on in-depth spread rate: a) shows increasing relationship between MC and in-depth
spread for two different IC values, b) shows independence of in-depth spread on IC at low MCs, but changes at high MCs.

The positive effect of MC on in-depth spread rate can be explained by a decrease in organic
density with MC (Figure 19). The decrease in organic density means that there is less fuel per
unit depth and so less time is required to consume the full sample depth. This also explains the
variation in response between samples of 2.5% IC and 40% IC. As discussed in section 3.2.2
Soil Conditions, because MC is calculated as a percentage relative to the dry mass, as IC
increases the ratio between organic matter and moisture decreases. Thus, at the same MC, there
is more water available to absorb per unit of peat at increasing IC.

66
However, the decreased organic density alone does not explain the difference. Examining the
burning rate per unit area 𝑚"̇, which is a measure of how quickly mass is consumed per area
and is calculated (Eq 7) by multiplying the in-depth spread rate, Sd (cm/h), by the organic
density, ρo (kg/m3), it can be seen that a increasing linear relationship exists (Figure 35). This
shows that the rate of fuel consumption increases with MC, a finding that is counter intuitive
but aligns well with the conclusion drawn from Huang and Rein [59]. It suggests that the
decreased organic density improves the oxygen supply, which results in an increase in char
oxidation. This consistency in trend also acts as a qualitative validation of the experimental set
up.
𝑚"̇ = 𝜌𝑜 𝑆𝑑 (7)

Figure 35 - Linear relationship between burning rate and MC for two values of IC. Burning rate is a measure of the rate of
mass consumed per unit horizontal area, and the critical MC is the threshold beyond which smouldering cannot be sustained.

Burning time, which is defined as the duration that smouldering was present in any horizontal
location, is also a metric of interest, and is used in calculating the in-depth spread rate (Eq 4).
As the burning time varies inversely to the in-depth spread rate and will exhibit an inverse trend
to Figure 34, it can be seen that, counter intuitively, burning time decreases with increasing
MC. In other words, dryer peat will heat a region for longer than moist peat. This may have
significant implication on long term soil damage [22,77].

4.3.4 Global Spread Rate

Each of the spread rates are vector components of a single global spread with an associated
direction, ф (Figure 31). I show, for the first time, that this global spread rate of peat fires is, at
least superficially, independent of MC (Figure 36a), but changes in direction. The average
global spread rate was 8.7 (omitting the behaviour of dry peat) and 8.4 cm/h for 2.5 and 40%
67
IC, respectively, which also indicates a mild influence of IC. Figure 36b shows a positive linear
relationship between ф and MC (R2 > 0.99), showing that for wetter soils spread tends to
become in-depth, while spreading more horizontally in drier soils. Additionally, as with in-
depth spread, the change in ф with MC is greater at a higher IC (40%). This may have
significant implication for deep soils with varying in-depth moisture profiles. Examing the
influence of IC in dry samples (ICx, Figure 37), it can be seen that global spread linearly
decreases, with the direction of spread remaining constant until higher IC (75 and 80%). This
trend may be limited to dry soils, because,as I have shown, IC has a minimal effect on global
spread at MC greater than 20% and has an increasing effect on the direction of spread with
increasing MC (Figure 36).

Figure 36 - Plotting a) global spread rate, which is the combination of horizontal and in-depth spread, against MC, indicating
its independence; and b) the linear change of direction of spread with MC. Error bars show the range in experimental data.
Black line are results calculated from literature values of horizontal [101] and in-depth [59] spread rates.

To further explore the global spread, it was calculated from reported values of horizontal [101]
and in-depth [59] spread rates reported in literature. Prat et al. [101] studied horizontal spread
in a horizontal reactor, with a 5 cm thick peat bed, while Huang and Rein [59] studied in-depth
spread independently in a 30 cm deep column reactor. Notably, both used the same peat as that
used in this experiment. The two experiments did not study the same MC values, so to make
them comparable the spread rates were linearly interpolated between the two nearest MC values.
The dashed black line in Figure 36a reveals a trend consistent with the experimental data
presented here; a global spread that is independent of MC. The average global spread rate of
the literature values was 4.0 cm/h, which is noticably lower than that which has been calculated
for the 2.5 % IC samples using the shallow reactor. The significance of this behaviour is not
yet understood, and so further study is required to understand this trend, particularly in deeper
soils and in the context of overhang.

68
Figure 37 - Global spread and direction of spread affected by changing inorganic content in dry samples. Error bars show
the range in experimental data.

4.3.5 Bifurcation; the Partial Extinction of Smouldering

As MC was increased, a bifurcation of behaviour was exhibited (Figure 38 a and b) where
smouldering became locally extinct (Figure 38c), resulting in an unstable and meandering
smouldering front. The y-axis, on the right of Figure 38 a and b, provides the average
proportion of transects where smouldering stopped. This was calculated by dividing the number
of transects that did not exhibit pixel values above the threshold (0.2) by the total number of
transects used, which was increased to 600 so to improve the resolution of spread detection.
This proportion of the smouldering perimeter that had become extinct increased with MC above
the MC threshold values of 110 and 60 %, for ICs of 2.5 and 40% respectively, until
smouldering could no longer be sustained and became completely extinct. This bifurcation has
never been reported before, and occurs when MC is close to critical MC [39]. As can be seen
in Figure 38 a and b, bifurcation first occurs when horizontal spread becomes less than or equal
to in-depth spread. This corresponds with a ф angel of 45°. Thus the intersection of the
decreasing trend of horizontal spread rate and increasing trend of in-depth spread rate, mark
the critical threshold for bifurcation. It is notable that increasing the IC, which decreases the
horizontal spread rate and increase the in-depth spread rate, causes criticality to occur at a lower
MC. This is consistent with findings in literature, which state that the critical MC is decreased
through the additon of inorganic material [39]. Further research is needed to understand the
importance of the interplay between in-depth and horizontal spread rate, and how it can be

69
applied to deeper samples of peat, where complex dynamics such as overhang can occur more
prominentely.

Figure 38 – a) and b) Illustrate how once in-depth spread exceeds the horizontal spread rate smouldering begins to extinguish
locally and exhibit a bifurcation behaviour. c) An IR image illustrating bifurcation, with regions of spread and extinction as
well as illustrating the use of transects in detecting extinction and measurement of the extinction perimeter.

4.4 Conclusions:
A new experimental rig was developed and used to simultaneously study, for the first time,
horizontal and in-depth spread rates across a range of moisture contents and inorganic contents.
The shallow reactor allowed for decreased experimental time and allowed for a greater number
of conditions to be examined. Some of the trends found are consistent with previous
experiments, which act as qualitative validation of the set-up. Both moisture and inorganic
content linearly decreased horizontal spread due to the additional energy required to evaporate
the moisture and heat the inorganic content. Conversely, moisture content increased the in-
depth spread rate by up to 300%, due to the decrease of organic density as a consequence of
the swelling of peat, which decreased the total organic mass in depth as well as increasing the

70
burning rate. Inorganic content showed little influence on horizontal spread at low moisture
content but was more influential at higher moisture content, which has never been studied
before. The intersection of the increasing trend of in-depth spread and the decreasing trend of
horizontal spread, reveal a critical point at which bifurcation can be observed, and extinction
becomes possible, highlighting the value of studying the fundamental dynamics of smouldering.
Furthermore, for the first time it has been revealed that the global spread vector, a combination
of both horizontal and in-depth spread rate, is insensitive to moisture content. The direction of
this spread is dependent on moisture content, spreading more downwards in wet soils. From
this study, it is evident that horizontal spread and in-depth spread, although distinctly different,
are closely related and require further study to understand the implications and influence in
deeper soils. These findings provide valuable new insights into smouldering dynamics and a
better understanding of the fundamental interaction between soil conditions and burning
dynamics in shallow layers, which can help inform policy makers and fire-fighting practices.

71
Chapter 5: Influence of Density on
Smouldering Dynamics and a Unifying
Theory of Spread
Summary3
Soil conditions such as moisture content, inorganic content and density, are known to change
the dynamics of smouldering. However a generalised theory that describes the influence of all
of these soil conditions is lacking. Here I experimentally study the influence of density on
smouldering dynamics and bring together the results presented in Chapter 4 so to, for the first
time, create a complete data set of experiments on the effect of the 3 most significant soil
properties on smouldering spread. This enabled an analysis of the effect soil properties on
smouldering dynamics and the discovery of a unifying theory on smouldering spread. Density
was studied by conducting 15 experiments using commercial peat samples in a shallow open
top reactor with varying densities from 157 to 223 kg/m3 (0 – 40% compression). Density was
found to decrease both horizontal and in-depth spread rate by 1.8 cm/h with a 40 % increase
in density, representing a decrease of 24 and 33% in spread rate respectively. However, the
burning rate of the peat is independent of density. On the field scale, this means that the carbon
emissions of ongoing wildfires would be unaffected by the density or compaction of the soil,
but may be more difficult to ignite and spread slower. By examining this completed data set of
experiments, it was found that across all soil conditions, in-depth spread is controlled by the
inverse of the organic density of the soil, while horizontal spread is controlled by the inverse
of the heat sink density, which is the energy required to heat the soil to peak smouldering
temperature. These findings have significant implications on our understanding of smouldering
dynamics as they provide a theory for interpreting the influence of soil condition on
smouldering spread.

3
This chapter is based on journal manuscript: E. G. Christensen, G. Rein, Experimental
Evidence of a Unifying Theory of Spread for Smouldering Fires, (to be submitted)
72
5.1 Introduction
Soil properties are known to significantly influence the smouldering fronts and spread rates
[22]. The most influential of these parameters are moisture content (MC), which is the amount
of water in the soil; inorganic content (IC), which is the amount of non-combustible material
in the soil (both represented as a percentage of the dry mass); and bulk density [10]. The
influence of MC and IC on smouldering spread were presented in Chapter 4. Density, however
is a highly important variable, with little research dedicated to understanding its effect on
wildfires. Density naturally varies in nature, typically increasing with depth [49,64]. Density
can also be varied inadvertently by human actions such as drainage, leading to subsistence [75]
or intentionally through compaction which has been proposed as a means of improving the
sustainability of plantation, although this has been met by scepticism by the scientific
community [19]. Thus understanding the impact of variation in density has on smouldering
dynamics is essential to predicting and understanding the smouldering behaviour on the field
scale.

In peat MC was shown to decrease horizontal spread rate [56,57] and increase in-depth spread
rate [59], while IC decreased the horizontal spread rate [61] but had no influence on the in-
depth spread rate[62,102]. The influence of density on smouldering spread is not clear in peat.
Experimental studies show density decreases horizontal spread [56] and reduced in-depth
spread rate [59]. However only two experiments were conducted in studying the effects of
density on in-depth spread. Huang and Rein [59] computationally predicted a negative inverse
relationship between density and in-depth spread, which also experimentally found by He et al.
[60] in samples of biomass. Both of which concluded that oxygen supply and fuel density
where controlling variables. Smucker et al. [103] also found a decrease in both horizontal and
in-depth spread with density in samples of synthetic biomass. Frandsen [39]revealed a negative
linear relationship exists between MC and IC defining a threshold of ignition whereby and
increase in MC or IC would result in an inability to sustain smouldering. However, Hartford
[66] found that increasing the organic density (density of the dry fuel) of these soils results in
a decrease in the critical MC and IC, while other have found an increased critical MC at greater
densities [87] or no change on probability of ignition [68]. Kohlenberg et al. [83] reported that
in experiments studying in-depth burning of natural samples greater organic densities resulted
in greater burn depths. This illustrates the fields immature understanding of the complex effect
density has on smouldering spread, and the necessity of further research into this key parameter.
73
Notably, research has commonly focused on studying soil conditions individually, as a result
there is a lack of a generalised theory that describes the influence of all of these soil conditions.

In this work, the influence of density on smouldering dynamics was experimentally

investigated using a shallow reactor. In combination with the results presented in Chapter 4,
this completes the data set of the three most significant soil parameters and allowed for a
holistic analysis these parameters to determine a unifying theory enabling an explanation of
the influence of soil properties on smouldering dynamics.

5.2 Method
Experiments were conducted using the shallow reactor, as described in Chapter 3. To explain
briefly, the internal dimensions of the reactor measured 40 × 40 × 1.6 cm and was constructed
from an insulating material. Samples were prepared by first drying the peat in an oven at 80°C
before adding water, sufficient to achieve a MC of 80% (on dry basis). The IC of the samples
were left unchanged from the inherent content (2.5%). Densities were altered by compressing
the soil. Five levels of compression were targeted, from 0 to 40% in increments of 10%.
Compression was calculated relative to the density of a fresh sample (157.5 kg/m3). If density
varied more than 5% from the targeted value, the reactor was emptied, and the process was
repeated. Three experiments were conducted for each compression condition. The samples
were ignited at the center, and spread was measured by analysing IR images from an IR camera
mounted directly above the center of the reactor.

5.3 Results and Discussion

5.3.1 Influence of Density on Smouldering Dynamics
Figure 39 is an IR image series illustrating the influence of density on smouldering dynamics.
Each image is from samples of different densities, all of which are taken 140 min after ignition.
Notably, each image shows an even and circular smouldering front, suggesting the peat
properties were spatially uniform. It can be seen that the smouldering slows down significantly
with density, and the smouldering width is weakly affected by it. Smouldering width is defined
as the distance between the leading and trailing edge. Figure 39 shows a slight linear increase
in width with density. Fundamentally this width is dictated by the relative spread rates of
horizontal and in-depth spread (Figure 21) as they control the movement of the leading and

74
trailing edge respectively. A relative decrease in in-depth spread compared to horizontal spread
will lead to a wider smouldering front. The spread rate of these components are therefore
examined to understand this increase in smouldering width.

Figure 39 -IR image series illustrating the influence of density on smouldering spread in peat. Each image is taken 140 min
after ignition. The crosses indicate location of ignition.

Figure 40 - Influence of density on the smouldering width. A slight linear increase is noticeable with a slope of 0.0046
cm·m3/kg1 (R2 = 0.92).

A 40% increase in density had a significant influence on both horizontal and in-depth spread,
decreasing both by 1.8 cm/h as shown in Figure 41. The decrease of both horizontal and in-
depth spread were found to be parallel, with slopes of -0.028 and -0.029 cm·m3/kg·h,
respectively. The ratio of in-depth to horizontal spread rate decreases from 0.77 to 0.68 between
158 and 222 kg/m3 (0 and 40% compression). This change in relative spread rates is the reason
for the increase in smouldering width with density.

75
Figure 41 – Influence of density on both horizontal and in-depth spread rates. Both spread rates decrease with density.

These findings, are in agreement with the previous literature. For example, Prat-Guitart [56]
found an inverse relationship between horizontal spread rate and organic density, which is the
density of combustible material in the peat. In-depth spread was shown to be slower in samples
with greater densities by Huang and Rein [59], who measured a 40% decrease between samples
with a 40% difference in density. This is similar to the 33% decrease in spread rate found here
across the same change in density. Smucker et al. [103], in studies of cellulose and hemicellulos,
also found that both in-depth spread and horizontal spread decrease with density.

Combining the horizontal and in-depth spread as the two vector components of a global spread
with an associated direction , as first presented in Chapter 4, shows a linearly decreasing
global spread rate with density. The global spread rate decreased a total of 27% over a 40%
increase in density. This is in contrast to the effect of moisture found in Chapter 4, which found
global spread to be independent of MC. The direction of the global spread, however, is weakly
influenced by density, becoming slightly more horizontal with . The decreasing trend in  was
found to be statistically significant with the p-value of 0.005 (<0.05). In the previous chapter,
a decreasing  has been associated with more robust smouldering, as extinction, in this reactor,
occurs once  >45°. Smucker et al. [103] found that increasing density resulted in higher peak
temperatures. Here, however, I measured a 5% decrease in average radiation from the
smouldering; which is an insignificant change across a wide increase in density.

76
Figure 42 – Change of global spread rate and associated direction () with density. Global spread decreases linearly with
density with a slope of -0.039 cm·m3/ kg·h (R2= 0.94). The direction of spread is decreased by 9% for 0 to 40% increase in
density..

Figure 43 examines the burning rate (𝑚̇ ") which is the rate of fuel consumption per unit area,
calculated by multiplying the measurements of in-depth spread rate by the organic density (Eq
7). The relationship between burning rate and density was found to be statistically insignificant
with a p-value of 0.1 (>0.05). A relatively constant burning rate of 4.66 ± 0.16 kg/m2·h was
measured. Similarly, Frandsen [104] who calculated burning rates by dividing the organic mass
in the reactor by the duration of the experiment, reported that density did not significantly
influence the burning rate in samples of peat.

Figure 43 – Calculated changes in burning rate as a function density are independent, with a p-value of 0.1. Burning rate is
calculated by Eq.7 . Average burning rate is 4.66 ± 0.16 kg/m2·h.

The findings on the effect of density on smouldering dynamics are significant in the context of
mitigating smouldering wildfires. Compaction has been promoted as a mean of protecting

77
peatlands, although limited scientific support exists to support this notion [19]. The decreasing
horizontal spread rate suggests that ignition may be more challenging at higher bulk densities
due to the greater energy required to dry and heat the soil as such requiring a great ignition
source [66]. However, once ignited, this study suggests that the burning rate is largely
unaffected by compaction, having little to no sensitivity to changes in density, meaning the
quantity of emissions will be unaffected by compression. The horizontal extent of smouldering
may be reduced due the slower fire spread. Thus, less peat may be affected before the fire is
extinguished. Ultimately, this may support the notion that compressing degraded peatlands
may increase the difficulty of ignition, and the slower rate of spread results in less total mass
consumed in the event of a peat wildfire.

5.3.2 Governing Spread Dynamics

To further study the governing dynamics of in-depth spread, in-depth spread is plotted against
organic density (Figure 44). For comparison, data was also gathered Huang and Rein [59], who
studied the effect of MC and a 40 % increase in density on in-depth spread in deep samples (30
cm) (blue diamonds in Figure 44). In addition, I compare the three experimental series that I
presented in Chapter 4; varying MC in samples with 2.5% IC (MCx; right triangle), varying
MC in samples with an IC 40% (mixed; up triangle), and varying IC in dry samples (ICx; left
triangle). Figure 44 is the first time data from samples across a range of soil conditions are
brought together. Figure 44 shows a strong trend between in-depth spread rate and organic
density, despite the experiments varying in MC, IC, and density. The results found using the
shallow reactor are 2 to 3 times faster than spread rates taken from deeper reactors. However,
the trend is consistent across varying reactors. This common non-linear trend supports the
computational findings by Huang and Rein [59], which predicted the in-depth spread rate of
soils with different MC are controlled by the organic density. High linearity is found between
in-depth spread and the inverse of organic density (Figure 44b), with R2 values of 0.97 and
0.89 for the shallow reactor and Huang and Rein [59] respectively. Due to these findings I
propose that in-depth spread is depends on the inverse of organic density. This is a highly
significant finding as it explains the in-depth spread dynamics across the three most significant
soil conditions, MC, IC, and density, through a single common parameter.

78
Figure 44 – a) Showing the common relationship between in-depth spread and organic density in samples with varying soil
conditions. b) Demonstrates the linearity of the relation between in-depth spread and 1/organic density. The data is compared
together with 3 other experimental series using the shallow reactor: right triangle (MCx) – varying MC in samples with
IC=2.5%; up triangle (mixed) – varying MC in samples with IC=40%; and left triangle (ICx) – varying IC with samples of
low MC (<5%), as well as an experimental series conducted by Huang and Rein,2017 [59] varying MC and density in deep
672.4
peat layers. The equation that describes the shallow reactor data is 𝑆𝑑 = − 1.892.
𝜌𝑜

Organic density, however, does not capture variations in horizontal spread rate as shown in
Figure 45. Instead horizontal spread rate appears to correlate with heat sink density (ΔHs). In
this thesis the heat sink density is the enthalpy change, per unit volume, to a typical peak
smouldering temperature for peat (550 °C) [70]. The organic density (𝜌𝑜 ), density of moisture
(𝜌𝑤 ), and inorganic density (𝜌𝑖 ) are calculated using Equation 3, 8 and 9 for each soil condition,
respectively, and substituted into Equation 10. Additionally, heat of pyrolysis and latent heat
of vaporisation were also taken into account. The values for these and the thermal properties
are given in Table 3, and were selected due to their successful use in literature in modelling
smouldering spread in peat [59,65]. A similar method was applied by Prat-Guitart et al. [56],
Lukenbach et al., [105] and Benscoter et al.[64]. Prat-Guitart et al., calculated the energy
required to heat peat to a pyrolysis temperature (150°C) and noted there was an increase in
energy requirement with increasing MC and density. While both Lukenbach et al., and
Benscoter et al. used it to explain and predict extinction by comparing it to the energy released
during smouldering. Here I more closely explore the connection between heat sink density and
horizontal spread rate.

79
Figure 45 - Illustrating the poor correlation between organic density and horizontal spread rate. Comparing samples with
varying experimental set up and soil conditions from multiple sources such as Hu et al.,2019[57] , Prat et al.,2016[56] and
Cowan et al.,2020[84]. Error bars represent the range of experimental data.

𝑀𝐶
𝜌𝑤 = 𝜌𝑏 𝑀𝐶+1 (8)
𝐼𝐶
𝜌𝑖 = 𝜌𝑏 𝑀𝐶+1 (9)

∆𝐻𝑠 (𝑢𝑛𝑖𝑡𝑠 𝑜𝑓 𝐽/𝑚3 ) = 𝜌𝑤 (𝑐𝑤 ∆𝑇𝑑 + 𝐿𝑤 ) + 𝜌𝑜 (𝑐𝑝 ∆𝑇𝐻 + ∆𝐻𝑝 ) + 𝜌𝑖 𝑐𝑖 ∆𝑇𝐻 (10)

Table 3 - List of values used for calculating heat sink density. Thermal properties used are found in [65]

Variable Value Unit Description

cw 4186 J/kg·K Specific heat of water
co 1840 J/kg·K Specific heat of organic matter
ci 880 J/kg·K Specific heat of inorganic matter
∆𝑇d 80 °C Temperature increase to 100°C (water drying)
∆𝑇H 530 °C Temperature increase to peak value in smouldering
Lw 2256 kJ/kg Latent heat of water vaporisation
∆𝐻P 500 kJ/kg Heat of pyrolysis

Horizontal spread rats from 3 other experiments from literature, the results from Chapter 4 and
the spread rates presented in this chapter are plotted against each soil condition’s heat sink
density in Figure 46 [56,57,84]. Chapter 4 studied both horizontal and in-depth spread across
a broad range of soil conditions, as explained earlier. Hu et al. [57] conducted horizontal spread
experiments varying MC in a reactor with the dimensions of 20 × 20 × 10 cm using commercial
peat. While Prat-Guitart et al. [56] also used the same commercial peat, and varied both MC
80
and density in a horizontal reactor measuring 22 × 18 × 5 cm. More recently, Cowan et al. [84]
gathered natural samples of duff from below ponderosa pine trees. They defined an upper and
lower duff layer by the extent of decomposition. With the upper layer being less decomposed
than the bottom layer. For all samples they measured the MC, IC, and density of both layers
before igniting the surface of the samples and studying their smouldering spread rates. They
noted that their MC measurements poorly captured the average sample MC in samples with
upper duffer layers greater than 9 cm. For this reason I limited the data to only samples with
upper duff layers less than or equal to 9 cm deep. It can be seen that there is a clear trend for
each of the experimental data sets, with the results obtained using the shallow reactor collapsing
onto a single trend. This convergence of experiments conducted in the shallow reactor to a
single trend across a range soil conditions suggests that horizontal spread is predominately
controlled by changes in heat sink density. Although the results gathered from literature each
form unique trends, they all demonstrate an inverse relationship between horizontal spread rate
and heat sink density. This is further illustrated in Figure 46b, which reveals the strong linear
dependence of horizontal spread rate on the inverse of the soils heat sink density with an R2
value of 0.87 for all experiments conducted in the shallow reactor. A similar analysis on in-
depth spread showed no common trend (Figure 47). Notably, some outlier exist for the data set
presented by Cowan et al. [84], the source of this variability is unknown, but may be due to the
difficulty of capturing the average soil properties of natural samples. Despite this, a clear trend
is still observable. These outliers were omitted from the linear fit.

This dependency of horizontal spread rate on a soils heat sink, although intuitively alluded to
in many papers [39,56,84,106,107], has never been presented. It provides a common factor
upon which all soil properties can be generalized to explain spread rates. Further research is
required to understand what influences the constant of proportionality, as each experimental

81
appears to result in a unique relationship. It may be related to such things as thermal properties,
boundary condition, or heat of combustion.

Figure 46 – a) presents the relationship between horizontal spread and heat sink density of the soil. b) demonstrating the
linearity between horizontal spread rate and the inverse of heat capacity. Results are compared with results from Chapter 4,
Hu et al.,2019 [57], Prat et al.,2016 [56], and Cowan et al.,2020 [84]. Varying across a broad range of MC, inorganic content
3.276×109
and density. The equation of the of the linear fit for the shallow layer data is 𝑆ℎ = − 2.825.
∆𝐻𝑠

Figure 47 - Illustrating the poor correlation between in-depth spread rate and heat sink density. Results are compared across
a range of soil conditions and experimental set up(Huang & Rein, 2017 [59]) . Error bars represent the range of experimental
data.

These findings show that in-depth spread and horizontal spread are controlled by two separate
variables; organic density and heat sink density, respectively. This is the first time the
differences in governing dynamics of horizontal and in-depth spread have been clearly shown.
The difference in the controlling spread mechanism may be explained by each spread directions
access to oxygen. Considering in-depth spread to be forward spread (as oxygen travels in the
same direction as spread) Ohlemiller [108] showed that in such conditions the spread rate of
82
the oxidation front was limited by the rate of char consumption. This is because the penetration
of oxygen into the fuel is limited as it is rapidly consumed in-depth through reactions with the
char [67]. Oxygen can therefore only penetrate deeper into the soil through the consumption of
a layer of char. Increasing the organic density will thereby increase the quantity of fuel that
must be consumed before the front may advance, and may also influence the oxygen
penetration depth. However, the endothermic processes (drying and pyrolysis) can occur
without oxygen and can propagate independently. For this reason the thickness of the char layer
may be far greater than the oxidation zone [67]. This was also observed experimentally by He
et al. [60] in an in-depth spread experiment with biomass, where the drying sub front was found
to spread in-depth an order of magnitude faster than the oxidative front. The presence of MC
reduced the spread of the drying front while having no effect on the oxidative spread rate.
Similar results were found in the smouldering of organic waste [109]. Palamba et al. [110],
using thermocouples to measure smouldering in-depth spread rates in peat, reported the spread
rate of the 250°C threshold (the pyrolysis front) spread more rapidly than the 350°C threshold
(oxidation front). The greater rate of spread of the endothermic front compared to the
exothermic fronts is a logical prerequisite for sustained smouldering combustion in a
homogenous fuel, because if the rate at which char is formed (dried and pyrolysed) is slower
than the rate at which char is oxidized and consumed, then spread cannot be self-sustained. So
it follows that the spread rate of the endothermic processes must be equal to, or greater than,
the spread rate of the oxidation front. It may be possible to describe the threshold for self-
sustained smouldering as a function of these spread rates. This was analogously shown in
Chapter 4 where once the in-depth spread was greater than the horizontal spread, which are
controlled by oxidation and heat transfer respectively, the smouldering failed to sustain.

𝑚"̇
Furthermore, examining equation 7, whereby in-depth spread can be calculated by 𝑆𝑑 = 𝜌𝑜
,

the primary factors influencing the in-depth spread rate are evident: organic density and the
burning rate. The effect of organic density has been explained by increasing the fuel required
to consume before the smouldering front may advance. The second parameter, burning rate,
deserves further study. However, factors such as surface area to volume ratio [99], air flow
[108], and a number of other factors may, depending on the rate limiting step, all increase the
burning rate. Smouldering has been primarily identified as an oxygen supply limited process.
As such oxygen has a particularly strong influencing factor, as found in literature [108]. For
instance, as spread continues to propagate downward, it has been experimentally been found
83
to decrease with depth, in part, due to the formation of an ash layer restricting oxygen supply
to the oxidation zone [58,59]. Similarly increasing the oxidizer flow increases the reaction rate
resulting in a greater quantity of char consumed per unit time and a greater rate of spread
[46,47,111].

Horizontal spread, on the other hand, travels perpendicular to the oxygen supply which means
that, broadly, oxygen can react with the char upon creation. The spread might therefore be
governed by how quickly energy can be transferred and the heat sink density of the soil can be
overcome. With this in mind, the differences in trend of each results from literature in Figure
46 may be due to different heat loss rates. Equally, changes in oxygen supply, such as wind,
will similarly change the heat release rate of the smouldering as well as the heat transfer
mechanism, thus affecting the spread rate [71]. Horizontal spread, for this reason, appears to
be affected by more complex interaction of factors than in-depth spread as it is affected by the
same parameters as in-depth spread, but also sensitive to heat transfer effects. This proposed
theory for understanding the governing dynamics of smouldering gives a new insight into
wildfires, and may allow for more focused research.

5.4 Conclusion
For the first time, the effect of density on horizontal and in-depth spread in peat is
simultaneously studied. It was found that a 40% increase in density decreased horizontal spread
rate by up to 24% and in-depth spread rate by up to 33%. The experiments also showed that
the rate at which fuel is consumed in-depth is independent of density, suggesting that the
decrease in in-depth spread rate is due to the increased mass that is required to burn. This
indicates that the method of compaction, as proposed as a means of protecting degraded
peatlands, once ignited, does not reduce peat consumption. However, the decreased horizontal
spread rate means that the horizontal growth of the fire may be limited before firefighting action
or heavy rains occur. Additionally, it implies that initial ignition may be more difficult as more
energy is required to locally heat the fuel to ignition. The global spread of peat was also shown
to decrease and trending to spread more horizontally with density. These findings provide a
greater understanding of the impact of artificial compaction of peatlands and natural variation
in soil density on smouldering dynamics.

84
By comparing results found here to those from several studies varying moisture content,
inorganic content and density, I show that in-depth spread can be understood by a soil’s organic
density, regardless of its soil condition. As char oxidation is restricted to the surface due to
limited diffusion, the rate of oxidation and the quantity of fuel controls how quickly the front
progresses. I propose a functional inverse relationship between in-depth spread and organic
density. I also found supporting evidence in literature to suggest that this is may be a universal
behaviour of smouldering fuel layers. In contrast, horizontal spread was dependent on the heat
sink density of the soil, which is the amount of energy required to heat the soil to smouldering
temperatures. Changes in soil properties are reflected in the heat sink density of the soil and
the spread rate was found to be inversely correlated with this value. This suggests that while
in-depth spread is governed by the rate at which the organic matter can be consumed, horizontal
spread is controlled by heat transfer. These are significant findings in the field of smouldering
wildfires and provides a powerful new fundamental theory of the governing mechanisms.
Furthermore, this holistic analysis speaks to the benefit of conducting experiments in a
consistent rig, and should encourage future researchers to adopt the reactor used here such that
experimental results can be directly compared.

85
Chapter 6: Influence of Environmental
Condition on Smouldering Dynamics: A
Study of Wind and Slope
Summary4
This chapter studies the influence of wind and slope on multidimensional smouldering spread,
using a shallow open reactor. I studied the influence of wind direction (forward, perpendicular,
opposed) and slope (uphill, side hill, downhill) on five smouldering dynamics: horizontal
spread rate, in-depth spread rate, burning rate, global spread, and direction of spread. These
conditions are known to be controlling variables in spread dynamics of flaming wildfires;
however, wind and slope are rarely studied for smouldering wildfires. I conducted 21
experiments and compare the data to an additional 15 experiments from Chapter 4. Where
airflow was concurrent with spread (forward wind and uphill), both wind and slope increased
the horizontal spread rates by to 101% and 28%, respectively, from quiescent and flat
conditions. When wind was perpendicular to spread, horizontal spread rate was increased by
wind (up to 21%). While horizontal spread in the opposite direction to the wind, side hill, and
downhill was negligibly influenced. In-depth spread on the other hand was seen to increase in
positions with forward and perpendicular wind, but decrease when spreading downhill. In all
other conditions there was no statistically significant influence on spread. Broadly, it was found
that slope was a less influential factor than wind. It was also found that spread in any direction
on a slope can be evaluated as a function of the angle of the spread relative to the horizontal
plane, regardless of the slope of the terrain. These findings provide new insight into important
field conditions affecting smouldering wildfires and provide a better understanding of their
spread.

4
This Chapter is based on: E. G. Christensen, Y. Hu, D. Purnomo, G. Rein, Influence of
wind and slope on multidimensional smouldering peat fires, Proceedings of the
Combustion Institute, (2020). Doi: 10.1016/j.proci.2020.06.128
86
6.1 Introduction
Little is known about the influence of environmental conditions such as wind and slope of the
ground on the smouldering of peat. This is despite these factors being prevalent in nature,
present to varying degrees in all wildfires, and are heavily studied parameters in flaming
wildfires [112,113]. In particular, no studies have examined the effect of wind or slope on in-
depth spread; the parameter which governs the emissions per unit area for smouldering
wildfires [59]. The little research to date focused on smouldering spread exposed to wind or
slopes has found that horizontal spread is increased by wind when in the same direction as
spread, as found in studies on pine needles [73] and dust layers [71], because the unburned
material is heated convectively, assisting the drying and preheating process. This mechanism
has been shown to increase the maximum fuel MC through which smouldering can be sustained
in samples of pine needles [73]. This concurrent orientation of airflow and spread direction is
known as forward spread [108]. Opposed spread is when airflow and spread are in opposite
directions [114]. Palmer [71] found that opposed flow, over piles of dust, increased horizontal
spread to a small degree but was insensitive to wind speeds. Wind has also been demonstrated
to enable the transition from smouldering to flaming [115], which can frustrate fire-fighting
efforts and presents an increase in spread rate and hazard [45,49].

For comparison, flame spread rates increase non-linearly with increasing uphill gradients or
concurrent winds, and decrease slightly downhill or in opposing wind [112,113]. This
behaviour is explained by the heat transfer from the flame to the fuel. Both slope and wind
cause the flame to tilt towards the fuel, increasing radiative and convective heat transfer and
dramatically increasing spread rates [112,116]. Thus the dynamics found for flaming fires
cannot be applied to smouldering fires.

In this chapter I, for the first time, study the effect of slope and wind direction on both
horizontal and in-depth spread rate of smouldering peat by systematically performing
experiments using the shallow reactor, and applying wind to samples with varying MC and
altering the slope of the peat bed. Thus I aim to improve our fundamental understanding of
these two prevalent factors in smouldering wildfires that have received little attention to date
by addressing this gap in knowledge.

87
6.2 Method
To study the influence of environmental conditions experiments were conducted using the
shallow reactor with some modification to the standard set-up, as described in Chapter 3. To
explain briefly, the internal dimensions of the reactor measured 40 × 40 × 1.6 cm and was
constructed from an insulating material. Samples were prepared by first drying the peat in an
oven at 80°C before adding water to achieve a desired MC, however in all cases the IC of the
samples were left unchanged from the inherent content (2.5%). The samples were ignited at
the center, and spread was measured by analysing IR images from an IR camera mounted
directly above the center of the reactor.

For experiments on wind, samples of 0%, 40%, 80%, and 120% MC were prepared as this
spanned the range within which smouldering was should to sustain as found in Chapter 4. A
wind speed of 0.86 ± 0.12 m/s was applied over the surface of the peat at the end of the ignition
protocol. For experiments on slope, the samples were condition to 80% MC and a lab jack was
used to pitch the reactor at 3 different angles (α); 10°, 20°, 30°. Three experiments were
conducted for each condition to capture the variability in the experiments. These experiments
were compared against the results from the MCx series with corresponding MC presented in
Chapter 4 as they were conducted with no slope or wind. The central ignition enabled the
simultaneous study of spread in all directions relative to both wind and slope.

6.3 Results and Discussion

6.3.1 Effect of Wind
Figure 48 illustrates the influence of wind direction relative to smouldering spread (θw), and
MC on 4 major metrics of smouldering dynamics, horizontal spread rate (Sh), in-depth spread
(Sd), burning rate (𝑚̇"), which is the measures of how quickly fuel is consumed (calculated
using Eq. 7), and global spread (SG), which combines the horizontal and in-depth spread rate
as spread components and was first discussed in Chapter 4. To focus the discussion, I examine
three directions: forward (θw=0°), perpendicular (θw=90°, 270°), and opposed (θw=180°). These
direction are presented in Figure 49 and compared to results with no wind from Chapter 4. I
consider the significance of the influence of spread direction relative to no wind conditions by
conducting an ANCOVA, using MC as a co-variate and report trend when a significance level
is <0.05.

88
Figure 48 –Polar plot of the influence of wind direction and MC on a) horizontal spread rate, b) in-depth spread, c) burning
rate, and d) global spread rate. Radial axis is spread rate (cm/h) while the angular axis is the direction θ w. Where θw= 0, 90,
and 180° is forward perpendicular and opposed respectively.

As reported in Chapter 4, MC decreases horizontal spread and increases in-depth spread.

However, both horizontal and in-depth spread rates in forward spread mode were greatly
increased by wind, relative to no-wind conditions (Figure 49 a and b) (p<0.05). At 120% MC,
horizontal and in-depth spread rate was 100% and 37% faster than in quiescent conditions
respectively, while in dry peat the increase was only 44% and 22% respectively. This shows
that as a proportion both horizontal and in-depth spread rate are increasingly affected by wind
with increasing MC. However, the slope of the linear relationship between spread rates and
MC (as shown in chapter 4) are insignificantly affected by wind (p>0.05), thus the influence
of wind is insensitive to changing MC. This is in contrast to findings presented by Huang et al.
[70] and Valdivieso and Rivera [73]. Huang et al. [70] found a ~140% increase in horizontal
spread between wind speeds of 0 and 1.2 m/s in 5% MC peat, but only a ~30% increase in
89
samples of 100% MC, suggesting that horizontal spread rate was less sensitive to wind with
greater MC. The difference in results may be due to a greater sample depth (10 cm) and an
increase in burning depth with MC which protects the smouldering from the wind.

Figure 49 - Influence of MC and wind direction on a) horizontal spread rate, b) in-depth spread rate, c) burning rate, and d)
global spread rate. The results are offset from MC values and each moisture condition is separated by dotted lines. The error
bars show the range of experimental results from three repetitions. The results for no wind conditions were taken from Chapter
4.

Where the wind was perpendicular, horizontal and in-depth spread rates increased slightly
(p<0.05), while no change in spread was observed where wind was in the opposite direction
(p>0.05) (Figure 49 a and b). This is in agreement with [71], which observed a minor increase
in horizontal spread rate with opposed wind and insensitivity to wind speed in studies on dust
(grass, cork, and beech sawdust). Ohlemiller and Lucca [108], however, found a small linear
increase with opposed wind speed in samples of cellulosic insulation. For comparison flaming
fires, also see a decrease in opposed spread rates due to convective cooling and reduced
radiative heating as the flame is tilted away from the fuel and see a strong positive influence in
forward wind conditions due to increased convective and radiative heating [116].

90
For horizontal spread rate, the difference between forward and opposed wind was 9% at 0%
MC and increases to 98% at 120% MC. While in-depth spread rates changed from a 2%
difference at 0% MC to 30% at 120% MC. The difference between forward and opposed
horizontal spread can be attributed to preheating by smoke flowing over the peat ahead of the
pyrolysis front, as observed by Valdivieso and Rivera [73] with smouldering pine needles. In
contrast, for opposed flow, airflow convection cools the smouldering front.

The burning rate was found to increase with MC (Figure 49 c). This is in consistent with [59]
who conducted in-depth smouldering spread experiments in a 30cm column of peat, and
Chapter 4 which is used as the comparison for no-wind conditions. The burning rate was
observed to be greater in the forward condition compared to opposed. Both forward and
perpendicular wind directions increased the burning rate compared to no-wind conditions
(p<0.05), while the opposed flow did not result in any increase in burning rate (p>0.05). The
increase in burning rate in the forward spread condition suggests an improved oxygen supply
leading to a greater rate of char oxidation and thus a greater rate of energy release. For the
condition with higher burning rates, I also observed an increased radiative heat flux measured
by the IR camera. It was found that forward spread radiated more heat than opposed spread by
up to 53%. The heat flux from the forward spread was greater than in quiescent conditions but
insensitive to MC, while heat flux decreased with MC in opposed spread and was similar to
that than in quiescent conditions (Figure 50). The difference in heat flux could also be visually
observed as forward spread often glowed, suggesting a higher local temperature. This may
suggest that the increase in horizontal spread rate is not only increased by the improved heat
transferred offered by the forward spread condition but also a greater energy release rate due
to greater oxygen supply increasing the burning rate.

91
Figure 50 – a) Showing the variation of normalized heat flux with wind direction. Demonstrating the independence of heat
flux from forward spread and decreasing heat flux from opposed spread with increasing MC. The results are average over 3
repeats for each MC conditio. b) demonstrates the influence of MC on peak radiant flux in quiescent conditions under no-wind
conditions. The data for the right figure is of the MCx series presented in Chapter 4. All clouds show the range of results
measured.

Combining the horizontal and in-depth spread rates as vector components gives the global
spread rate (SG) with a direction . Figure 49 d shows that the global spread rate was influenced
by wind direction with the forward and perpendicular spread increasing compared to no wind
conditions (p<0.05), while opposed spread was negligibly affected by wind (p>0.05). The
global spread appears to be independent of MC (p>0.05), apart from the unusual behaviour of
dry peat. This independence of global spread rate with MC was also reported in Chapter 4, but
is shown here to also be true for all wind conditions. The direction of spread was also found to
be more horizontal in the forward (a smaller angle of ) than in quiescent conditions (p<0.05)
and went more in-depth with increasing MC (Figure A2.1). An overview of the smouldering
dynamics as a function of wind direction is provided in Table 4.

Table 4 – Summarizing the effect of wind direction on smouldering dynamics. Percentage differences are noted relative to the
no-wind condition for 0, 40, 80 and 120% MC respectively. Arrows upwards means an increase, arrow downward indicates
a decrease, while ~ means a negligible change. Trends are based on the results from an ANCOVA with MC as the covariant
and a difference in population marginal mean with a p value <0.05.

Horizontal spread, Sh In-depth spread, Sd Burning rate, ṁ" Global spread, SG Direction of spread,Φ

Forward ↑ (44, 55, 47, 101%) ↑ (22, 33, 19, 37%) ↑ (19, 25, 21, 24%) ↑ (43, 52, 38, 65%) ↓ (-15, -13,-15, -20%)

Perpendicular ↑ (18, 14, 3, 21%) ↑ (25, 25, 8, 32%) ↑ (22, 17, 10, 20%) ↑ (19, 15, 5, 27%) ~ (5, 9, 4, 6%)

Opposed ~ (32, 4, -6, 2%) ~ (20, 18, 3, 5%) ~ (16, 11, 4, -4%) ~ (31, 6, -2, 3%) ~ (112, 12, 6, 4%)

92
6.3.2 Effect of Slope
Polar plots showing the influence of slope relative to spread direction (θs) of four major
smouldering dynamics are presented in Figure 51 a-d: horizontal spread, in-depth spread,
burning rate, and global spread rate respectively. Notably, on slopes of 30°, char uphill was
observed to slip, resulting in unreliable measurements and so spread dynamics affected by this
behaviour are removed from the results. The slipping material however did not affect the spread
perpendicular or downhill.

Overall, slope is found to be a weaker influence on spread dynamic compared with wind, but
share some similar responses due to the similarities in their fundamental mechanisms. For
instance, compared to spread on a flat fuel bed (α=0°), horizontal spread increased uphill (θs=0°)
by 20% for α=20°, but a linear regression revealed that it was statistically insignificantly
affected downhill (θs=180°) (p>0.05) (Figure 51a and Figure 52a). The buoyancy of the
emissions causes hot air to move along the fuel surface uphill creating forward spread
conditions (as with concurrent wind), increasing the rate of heat transfer [71]. It can also be
noted that uphill spread rates increases with increasing slope (p<0.05). These findings for
horizontal spread are similar to results presented by Palmer [71] with insulation fibre boards.
Palmer [71] found an average increase of 26% in horizontal spread between 0° and 30°
inclination, but and an average increase of 8% for spread down a 30° slope. By comparison the
effect of slope is more significant in flaming fires where heat transfer includes radiation,
significantly increases spread uphill as the uphill fuel is nearer the flame. However, flame
spread uphill is typically unsteady and has an increasing spread rate (>10°) [113].

93
Figure 51 - Polar plot of the influence of slope on a) horizontal spread rate, b) in-depth spread rate, c) burning rate, and d)
global spread rate. Radial axis is spread rate (cm/h) and the angular axis is the direction θ s. Where 0, 90, and 180° is uphill
(forward), sidehill (perpendicular) and downhill (opposed), respectively. Spread rates affected by the slippage of peat were
removed. Each slope is averaged between 3 repeats, and all samples here are condition to 80% MC.

94
Figure 52 – The influence of slope on smouldering dynamics on a) horizontal spread, b) in-depth spread, c) burning rate, d)
global spread rate, due to spread relative to the slope. Cases affected by the slippage of peat were removed. Values are
averaged across 3 repeats while error bars provide the maximum and minimum spread of the repetitions.

A linear regression revealed that uphill slope appeared to have no significant influence on in-
depth spread (Figure 52b) or burning rate (Figure 52c) (p>0.05). This implies that the airflow
does not significantly contribute to a greater rate of oxidation and therefore the increase in
horizontal spread with slope may be solely due to improved heat transfer to the peat rather than
an increase in heat release rate. This is in contrast with wind which also showed an increase in
burning rate (by proxy heat release rate) which may also have contributed to increasing the
forward spread rate. Because the horizontal spread rate increased with increasing slope, while
in-depth spread was unaffected, the global spread rate increased (Figure 52d) and the direction
of the global spread () decreased, with a greater uphill slope (Figure A2.2).

95
Horizontal spread on downhill slopes showed a very different behaviour, with no statistically
significant change in spread rate with slope (p>0.05) and both in-depth spread and burning
rate decreasing with slope (p<0.05). It is difficult to accurately attribute what causes this
decrease. One hypothesis is that this decrease in in-depth spread may suggest a decreased
availability of oxygen supply and may be the explanation as to the difference in behaviour
opposed spread and downhill spread. Global spread rate downhill statistically significantly
decrease downhill by up to 14% (Figure 52b, c, and d) and the direction of the global spread
rate tended more in-horizontal (decreasing ; Figure A2.2) with increasing downhill slopes
(p<0.05). Downhill spread experiences buoyancy induced air flow that transports smoke away
from the front, causing cold air to travel over the surface of the peat, increasing heat losses.
However, as horizontal spread rate was not significantly reduced, it suggests that this cooling
effect does not have a strong influence. Spread sideways to the hill also no considerable effect
on the smouldering dynamics, behaving similarly to that of no slope (Figure 52). An overview
of these findings are provided in Table 5. Further experiments may be required to test the
validity of applying a linear regression to the data generated. However assuming linearity was
thought reasonable based on the limited results and the research performed in other materials
[71]. Additionally, a greater number of experiments would be required to justify a more
complex model and due to the limited range of slopes on which the peat in this reactor is stable
this was not necessary. Regardless, what is clear is that slope, compared to wind has a much
weaker effect, likely due to the low velocity of the air current produced by buoyant smoke
compared to driven winds.

Table 5 - Summarizing the effect of slope relative to spread on smouldering dynamics. Arrows indicates the relative change in
behaviour to no-slope conditions as confirmed by the p-value < 0.05 of a linear regression. Percentages provide the average
change from 0° to 10, 20, and 30° respectively. Where x is given, no value was found due to slippage of peat.

Horizontal spread, Sh In-depth spread,Sd Burning rate, ṁ" Global spread, SG Direction of spread, Φ

Uphill ↑ (8, 20, x%) ~ (-4,-4 , x%) ~ (-1, 7, x%) ↑ (4, 12, x%) ↓ (-9,-16, x%)

Sidehill ~ (3, 2, 9%) ~ (0, -13, 3%) ~ (3, -3, 2%) ~ (2, -3, 5%) ↓ (-2, -11, -9%)

Downhill ~ (-9, -9, -6 %) ↓ (-11,-24, -15 %) ↓ (-8, -16, -10%) ↓ (-10,-14 ,-9 %) ↓ (-2, -13, -8%)

It was found that, regardless of the slope of the fuel bed, smouldering dynamics may be
evaluated as a function of the spread angle  from the horizontal plane. Calculated using Eq. 6
presented earlier, and is such that  =  when s = 0º (uphill, as illustrated in Figure 20), and

96
 = 0º when s = 90º (sidehill). Figure 53a and A2.3 shows how the spread rate of smouldering
behaves similarly regardless of the overall slope (). This helps our understanding of how
smouldering spreads on slopes. Plots of in-depth spread rate, burning rate and global spread
rate as a function of the spread angle , are given in Figure 53b, c and d respectively. Flaming
fires, which accelerate uphill, will not have a constant behaviour with  making this analysis
unfeasible [116]. However, where spread travels downhill or on gentle gradients the spread
angle  may be useful and so deserve further research.

Figure 53 - Plotting averaged a) horizontal spread rate, b) in-depth spread rate, c) burning rate, and d) global spread rate as
a function of β, which is the angle at which smouldering spreads relative to the horizontal plane. Spread affected by the
slippage of peat were removed. The cloud error bars show the range in spread from the repetitions.

97
6.4 Conclusion
In this chapter I present a systematic experimental study of both the influence of wind and
slope on horizontal spread rates, in-depth spread rates, burning rates, and global spread rates.
Both horizontal and in-depth spread rates increased with forward wind up to 101 and 37%
respectively between 0.84 m/s wind speed and no-wind conditions due to improved heat
transfer and oxygen supply respectively. Opposed wind did not influence the spread dynamics,
causing behaviour similar to that of no-wind conditions. Notably, varying moisture content was
found to have no significant effect on the influence of wind. Slope, compared to wind, was
less influential on smouldering dynamics. A 20° increase in slope increased the horizontal
spread rates uphill by 20%, a significant increase relative to flat conditions. Horizontal spread
downhill and side hill spread no significant difference when compared to no-slope. In-depth
spread on the, on the other hand, decreased when spreading downhill, and was negligibly
affected in all other directions. It was also found that the effect of slope on smouldering
dynamics could be analysed as a function of the spread angle, , which helps of smouldering
spread on slopes. Applying these findings to the field scale; forward wind and uphill slope,
compared to in quiescent conditions and flat topographies, increase spread, and are expected
to result in greater fuel consumption and emissions. These finding gives an insight into the
fundamental interaction between smouldering fires and two important environmental
conditions.

98
Chapter 7: Arctic Fires and Smouldering
Combustion- Influence of Cold
Temperatures on Fire Spread
Summary5
Recent wildfires in arctic regions have burned unprecedented swaths of land, demonstrating a
worrisome change in the arctic fire regime and release large volumes of greenhouse gasses into
the atmosphere. These events highlighting the vulnerability of these biomes to climate change
and the importance of protecting them. Arctic fires are poorly understood due to their remote
location and the difficulty in detecting them. Here I investigate the smouldering dynamics
involved in these fires by presenting a novel experimental rig, which allows for controlled cold
conditions to study the influence of low temperatures. The rig consists of a chamber with low
air temperatures, a reactor with a temperature controlled base to imitate the influence of
permafrost, and samples preconditioned to a desired temperatures. I study the influence of 3
condition temperatures covering a range of realistic Arctic soil temperatures; -7°C, 2°C, and
21°C. For the first time I experimentally show that smouldering can be sustained in soil
temperatures below freezing. Spread rates were found to increase with temperature, increasing
14% in samples of 50% moisture content and 18% in samples of 100% moisture content,
between -7°C and 21°C. The depth of burning was also found to increase, migrating deeper
into the soil, with decreasing temperatures (from 21°C and -7°C). Furthermore, the critical
moisture content was between 110% and 120% for all condition temperatures. This is lower
than results found in literature which used a reactor with an insulated base (160% MC) that
minimized heat losses compared to the chilled base used here. This suggests that as the soil
warms with climate change, the maximum moisture content in which smouldering can sustain
increases. In turn, this may lead to more frequent occurrence of fires in arctic regions. These
finding are the first experimental work on smouldering Arctic wildfires and improve our

5
This chapter is based on journal manuscript: E. G. Christensen, Y. Hu, G. Rein, Arctic
fires and smouldering combustion: influence of soil and air temperature on fire spread,
(to be submitted)
99
understanding of the effect of cold temperatures and present a novel method of studying these
fires at laboratory scale.

7.1 Introduction
The fire dynamics of Arctic fires are poorly discussed in literature due to their remote
occurrence, with most research occurring post extinction or via remote sensing. However, it is
highly likely the important mode of combustion is smouldering as it is more persistent and
difficult to extinguish than flaming [99]. These characteristics enable the sustained burning
below the surface and even below a layer of snow or ice [44]. Fires have been reported to burn
through the winter deep below ground and under snow cover, before emerge in in the spring
[44]. Such fires have been called holdover fires. Smouldering wildfires in Arctic region are
increasing in frequency and severity, and are transitioning what was once a historic carbon sink
to a net emitter [33], releasing significant quantities of greenhouses gases into the atmosphere
[117]. This increase in fire activity has been attributed to a rise in temperatures, resulting in
warmer and drier fuels. The arctic is particularly susceptible to climate change as it experiences
temperature increases at a much greater rate than the rest of the globe [38]. It is for this reason
that there is interest in understanding the relationship between smouldering dynamics and
temperature.

Once ignited, smouldering spreads both horizontally and in-depth with the spread rates affected
by both soil and environmental conditions [10]. The most influential parameter affecting
smouldering dynamics is moisture content (MC) [39], the measure of the mass of water in soil,
which decreases the horizontal spread rate due to the additional energy required to dry the soil
[56], and increases the in-depth spread rate, which is governed by the rate of char oxidation
[59]. A continued increase in MC beyond a critical point results in smouldering failing to self-
sustain, as the energy required to dry the peat becomes greater than the energy produced from
char oxidation [39,55]. Moisture has also been shown to increase the depth of burning whereby
smouldering occurs below the surface leaving the surface negligibly affected. This increased
burning depth also creates formations known as overhang, whereby a section of peat is
suspended over a void created by the rapid consumption of fuel. This formation eventually
collapses and begins to burn [70]. Thus smouldering spreads with cyclical pattern of overhang
formation and collapse. This phenomena is believed to be because an optimal balance between
heat losses and oxygen supply exists below the surface. The temperature of the soil and heat
100
losses to cold deep layer peat, such as permafrost may have an impact on this balance. However,
smouldering dynamics have, primarily, been studied to understand wildfires in tropical and
temperate peatlands. No laboratory experiments have yet explored the dynamics of fires
occurring in arctic or permafrost regions. Here I introduce a novel experimental set-up suitable
to address this gap in knowledge and demonstrate its usefulness in studying smouldering arctic
fires to gain a greater understanding of these fires and promote the rigs adoption by researchers.
To this end I, for the first time, conduct experiments varying both temperature and MC to
understand their effect of smouldering wildfires.

7.2 Method
Here I studied the influence of different condition temperatures on the smouldering dynamics.
The experiments conducted in this chapter used the cold reactor, which is a modified version
of the horizontal reactor presented in Chapter 2, and is described in full in Chapter 3. To
summarize I built a chamber in which the air temperature can be reduced to around 10°C, and
use an open top reactor with internal dimensions of 20 × 20 × 10 cm. The bottom boundary of
the cold reactor is an aluminium plate that is liquid cooled and temperature controlled such that
it imitates cold soil and permafrost layers. The samples are conditioned to desired temperatures
by placing them in fridges or freezers for a minimum of 12 h before use. The condition
temperatures examined was -7°C, 2°C, and 21°C. Capturing a range that is realistic for soil
condition in the arctic [98]. The samples were conditioned to MC of 50 and 100% and the
dynamics were studied using, a scale to measure mass loss, K-type thermocouples to measure
the temperature development within the sample, and IR imagery to observe the smouldering
spread.

7.3 Results and Discussion

7.3.2 Effect of Cold Temperatures on Smouldering
Mass Loss Rate
The mass loss rates from 50% MC samples at the 3 temperature conditions are presented in
Figure 54. Most notable is that smouldering was found to successfully sustain in condition
temperatures below zero, showing that the energy from smouldering is sufficient to overcome
both the additional sensible heat required as well as the latent heat of melting. This is the first
time smouldering has been experimentally shown to propagate through frozen soil which

101
illustrates the how wildfires can sustain in permafrost regions and winter seasons [43,44]. 4
stages can be identified as also outlined by Hu et al. [86]. Stage I is the ignition stage, and lasts
until the power to the ignition coil is turned off (0.5 h). Stage II is the growth stage, where the
smouldering front develops and grows in size. The spread stage (III) is characterized by a more
steady mass loss rate and steady spread. The final stage (IV) is the decay phase, where the
smouldering has reached the boundary of the reactor and its spread is limited by the fuel
availability. During stage I and II, large amount of variability appears to exist, but no consistent
trend is visible with condition temperature. However, in stage III the trends converge with no
significant differences in the mass loss rate between the various condition temperatures and all
reach a very similar peak.

Figure 54 - Showing the influence of condition temperature on mass loss rate and 4 stages of spread. I is the ignition stage, II
is the growth stage, III is the spread stage, while IV is the decay phase.

Temperature Evolution
Figure 55 shows the temperature development measured by the thermocouples in samples of
50% MC at -7 °C and 21 °C. Examining the bottom 4 plots (6 and 8 cm depth), a clear spike
occurs at the end of stage I (ignition period, 0.5 h), 3 cm from the ignitor. The temperature
begins, initially, to decrease as heat is diffused through the peat as external energy from the
igniter is no longer provided. Following this, the temperature beings to increase again once the
smouldering front approaches. The thermocouples 6 cm and below reach a consistent peak
temperature between 500-600°C before decreasing as the smouldering front moves away from
the thermocouple location. Notably, clear periods of stagnation are apparent in all plots of
Figure 55 around 100°C as the smouldering front overcomes the latent heat of evaporating the
moisture [53]. A characteristic difference can be observed between the temperature profiles
102
taken above 4 cm depth and below 6 cm depth; the deeper temperature profiles present a cleaner
trace, with fewer jumps, and the temperatures are markedly higher. This is because, as observed
in previous studies [70], smouldering occurs below the surface of the peat, resulting in the top
layer of peat often being minimally effected by smouldering until overhang becomes too great
and the column collapses. However, this means that the peat temperatures at 2 and 4 cm below
the initial surface often do not reach significant temperatures. Comparing the effect of the
condition temperature, the shallow layers (≤ 4cm) reach higher temperatures at higher
condition temperatures, while deeper layers are seemingly unaffected.

Figure 55 - Comparison between temperature developments of thermocouples at various horizontal and in-depth positions in
samples of 50% MC. Dotted lines represent the approximate transition between 4 stages: I is the ignition stage, II is growth
stage, III is stead stage, IV is the decay stage.

Of particular interest is the temperature during the ignition stage (I) as the power input from
the igniter is constant. Figure 56, demonstrates the temperature residence time of the two
thermocouples nearest the ignition coil (4 and 6 cm depth) during the ignition period (0.5 h).
Residence time is a measure of how long the thermocouple remained above a given temperature.
The trace of each condition temperature start at the initial soil temperature at the top left corner
of the plot. It is evident that the time taken for the thermal front to propagate to the
thermocouples increases with decreasing condition temperature, as the residence time descends
vertically for a longer period. Once the thermal front does reach the soil, it begins to heat
rapidly as indicated by a more horizontal line, until around 80°C, where the temperature
stagnates. This is associated with the process of evaporating the moisture in the soil as

103
mentioned earlier. Interestingly, in the -7°C condition, no plateau appears at 0°C, which would
be expected due to the latent heat of melting ice. This can be explained by the difference in
melting and vaporizing energy requirements. Vaporization is around 4.5 times more energy
intensive than melting, and as such features more prominently and so present more of a barrier
to overcome.

Figure 56 - Temperature residence vs time in samples of 50% MC at the two nearest thermocouple to the ignition coil during
stage I. a) 4 cm below the source, and b) 6 cm below the surface, both thermocouples are approximately 2 cm horizontally
from the ignition

Notably, the peat at 4 cm below the free surface does not fully dry, as the peak temperature at
the end of the ignition remained below 100°C. At 6 cm below the surface, on the other hand,
the fuel is dried and temperature continue to steadily increase until the end of the ignition. As
discussed this is likely due to the reduced heat losses with depth [70]. This characteristic
difference between these two depths exists through-out the smouldering spread as evidenced
by the temperature profiles in Figure 55, which shows a robust smouldering front at 6cm and
lower.

Comparing the residence time profiles of each condition temperature it appears that the there
is a simple shift due to their difference in initial temperature; whereby an increased amount of
time and energy is required to reach the drying phase. The 21°C condition temperature samples
reach the drying temperature more quickly than the 2°C condition which in turn is quicker than
the -7°C condition. The peak temperature during the ignition period is noted by the residence
time profile intersection with the x-axis. The peak ignition temperatures are plotted in Figure
104
57. Here it can again be noted that shallow layers (2 and 4cm) do not exceed the drying phase
(100°C). For this reason the peak temperature reached during phase I varies by only 7°C
between 21 and -7°C condition temperatures. The deeper layers experienced peak stage I
temperatures of 250°C and above with a slightly positive influence of condition temperature
(although a linear regression revealed no significant influence; p>0.05). This is also evident in
Figure 55b. At 6 cm depth the peak stage I temperature was 22 °C higher in the 21°C condition
temperature samples compared to the -7°C samples. This is very similar to the temperature
difference in the condition temperature, and so illustrates how the soil temperature delays the
temperature increase and shifts the temperature residence profile.

Figure 57 - Showing the effect of condition temperature on peak temperatures measured during the stage I (ignition period)
as measured by thermocouples nearest the ignition coil in samples with 50 % moisture content. Uncertainties are taken from
range of repeated experiments.

Similar behaviour was observed in peak temperature during the spread stage (III). Figure 58a
shows the effect of condition temperature on peak temperature. Here I also compared the effect
of MC due its significant effect on smouldering dynamics and importance in nature. Notably,
in deep layers the difference in peak temperature with MC is negligible, while more shallow
layers show a greater deviation at higher condition temperatures. For both MC, in the deep
layers, the peak temperature decreased slightly with increasing condition temperature, with
samples of 50% MC decreasing by 6.1% and 100% MC samples decreasing by 2.5%. While in
shallow layers the peak temperature increased by 90.5% and 22.5% for 50 and 100% MC,
respectively between -7°C and 21°C. It should be noted that a linear regression showed that
condition temperature had no significant effect on the peak temperature at any depth, with all
p-values being greater than 0.05. This may, in part, be due to the data. A greater number of

105
experiments should be conducted in the future to reveal the dependence of peak temperature
on condition temperature.

Figure 58b gives greater insight to the maximum temperature profile with depth for each MC
and condition temperature. It can be seen that the change in MC noticeably changes the profile
of the peak temperatures. In samples of 100% MC, soil at 4 cm depth or above remain around
100°C, this implies they do not fully dry before the peat column collapses. The global
maximum temperature in samples of 100% MC occurs at 8cm or below, but for 50 % MC it
occurs at around 6 cm depth. This difference in the depth of the peak temperature indicates that
burning is occurring at deeper depths with greater MC. This echoes the findings presented in
literature [53,70] which found that burning depth increases with MC. An estimate of the
burning depth was made by using a 200°C threshold as an indicator of pyrolysis occurring, as
done by Huang et al. [70]. The depth at which the peak temperature is 200°C therefore marks
the top of the charred region. This depth is found by linearly interpolating between the peak
temperatures of adjacent thermocouples and finding the location of the 200°C threshold. Where
the 200°C occurred above 2 cm, the profile water interpolated assuming that the surface of the
peat was equal to the condition temperature. This method is represented in Figure 58b), where
depth of burning is taken as where the peak temperature profiles cross the vertical dotted line
which corresponds to the 200°C threshold. The smouldering depths are presented in Table 6
along with the burning depth found by Huang et al. [70].

Figure 58 - a) The effect of ambient temperature on peak temperature, comparing samples of 50% and 100% MC and at
different depths (shallow, 2 cm; deep, 8cm.) b) Presenting the profile of peak temperature by depth and comparing both the
effect of MC and ambient temperature. Dotted line represents threshold of pyrolysis and used as an indicator of burning depth.

106
Table 6 - Estimated burning depth by using 200 C as an indicator of char formation and linearly interpolating between peak
temperatures measured by thermocouples. Depth of burning measured by Huang et al.,2016 [70], is also presented for
comparison.

Depth of burning (cm)

-7 °C 2°C 21°C Huang et al.,2016
50% MC 2.5 ± 0.2 1.8 ± 0.3 1.5 ± 0.3 3±1
100% MC 5.4 ± 0.6 4.6 ± 0.02 4.8 ± 0.5 9±1

Two effects are noticeable in the depth of burning; both condition temperatures and MC,
change the depth of smouldering. Firstly, decreasing condition temperatures increases the
burning depth, with a linear model confirming the significance of this trend (p<0.05). This is
consistent with the patterns in Figure 58a which showed decreasing condition temperatures
resulted in increasing deep layer temperatures and decreasing shallow layers temperatures. At
50% MC the smouldering depth increased by 66% between 21°C and -7°C, while in samples
of 100% MC the increase was 13%. This change in depth of burning is likely due to the
additional energy required to melt and heat the soil causing the optimal balance between
oxygen supply and heat losses to exist deeper within the soil. Secondly, comparing the depth
of burning estimated here to that presented by Huang et al. [70] who conducted experiments at
room temperature, using the same source of peat and a similar reactor but with an insulated
bottom face, it is evident that the effect of the energy loss through the base results in shallower
burning depths. This difference is enhanced by increasing MC. In literature the burning depth
at 100% MC was reported to be at 9 cm below the free surface (1 cm above the base of the
reactor) while here I present a smouldering depth of nearly half of this: 4.8 cm. Smouldering
within a cm of the base in this experiment would likely result in substantial heat losses. It was
observed that at the end of the experiments a thin layer at the bottom of the reactor, about 0.5
cm thick, would remain frozen and uncharred. This excess energy loss results in moving the
optimal conditions of oxygen supply and heat loss closer to the surface. Samples of 50% MC,
on the other hand, with a burning depth of 3 cm below the surface, experienced little change in
burning depth due to the change in boundary condition because of the larger distance from the
base. As such, there are two opposite trends. The cooled base moves the smouldering front
upwards, while decreasing soil and air temperatures promotes a deeper burning depth. This
finding suggests that the presence of permafrost may significantly influence the dynamics of
smouldering wildfires.

107
Figure 59 demonstrates the temperature residence time across the full duration of the
experiment in samples of 50 % MC, which is different to Figure 56, which considered the
ignition phase. Results from literature [57] are also presented here for comparison. Hu et al.
[57] conducted experiments with identical soil conditions to those presented here and used the
standard horizontal reactor, whereby the base is insulated. The thermal residence time is
measured using the second column of thermocouples (located 7 cm from the ignition side) as
they are minimally affected by the ignition and all thermocouples measured the peak
smouldering temperature, with many of them returning to ambient temperature. However, for
many experiments, including those in [57], the local temperatures in the deep layer (8cm) were
observed to still be high (>200°C) when the experiment was terminated. To correct for this, the
cooling period was modelled by linearly extrapolating the measured cooling rate down to the
condition temperature. A linear model is a suitable approximation for the cooling as evidence
by the near linear trends in the deep layers in Figure 55. In Figure 59a, there is a clear
distinction between temperature conditions in the shallow peat layer. Increasing condition
temperatures is shown to shift the residence time profiles, such that the soils experience higher
temperature for greater durations in soils with higher condition temperature. The experiments
by Hu et al. [57], on the other hand, has the lowest residence time at high temperatures, but
greatest residence times at low temperature when compared to the results measured in here.
These trends might be best explained by the burning depth. As shown in Table 2, burning depth
decreases with increasing condition temperature, as such the smouldering occurs closer to the
surface which increases the heating to the shallow layer peat, while the standard horizontal
reactor, as used by Hu et al. [57] and Huang et al., [70], observed the deepest burning depth,
and so experienced the lowest temperatures in shallow layers. Examining the temperature
residence time in deep layers (8 cm) (Figure 59b) a similar trend can be seen, with increasing
condition temperature increasing the residence times of nearly all temperatures. However,
residence profile converge towards similar peak temperatures (where the residence profile
intersects the x axis), resulting in minor differences in peak temperature as discussed before.
In deep layers, because the low temperatures are resident for less time at low condition
temperature, and the peak temperature are quite similar, it implies that the width of the
smouldering becomes narrower with decreasing condition temperatures. This may be because
the colder soil temperatures requires more energy and time to heat and melt the ice, resulting
in a compression of the smouldering width. The deep layer residence time calculated from [57]
show a higher peak temperature as well as a higher residence time across all temperature, this

108
can be attributed to the insulated base reducing heat losses. It can also be noted that all profiles
show a kink at around 80°C due to the drying step.

Figure 59 –thermal residence times showing at which thermocouples remained above a given temperature and the influence
of ambient temperature at a) shallow layers (2 cm) and b) deep layers (8 cm). The thermocouples used are located 7 cm from
the ignition side. The thermal residence time from Hu et al., 2019 [57] was provided for comparison which was conducted in
a reactor of the same dimension but with an insulated base.

Critical Moisture Content:

Additional experiments were performed to explore how the condition temperature influenced
the critical MC of samples. Critical MC is the greatest MC in which smouldering fire can be
sustained. It was found that smouldering could be successfully sustained in MC of 110%
however would fail at 120%. This was true across all condition temperatures. By comparison,
this is lower than values found by Hu et al. [57], who used the horizontal reactor which has an
insulated base and found the critical MC to be 160%. As such, the condition temperature does
not appear to have any effect on the critical MC, although further refinement of the MC
between 110% and 120% may reveal a difference. On the other hand, the difference between
an insulated base and the cooled base is significant. The heat losses through the base, have been
shown to reduce the depth of burning, which in turn likely increases the heat losses of the
smouldering to the environment ultimately reducing the energy available for advancing the

109
smouldering front. This suggests that the critical MC of peatlands underlain by permafrost or
otherwise cold soil, may have a significantly lower critical MC than when the below ground
soil is warmer. This may have significant implications on the frequency of Arctic fires and
related emissions, as warming will not only result in dryer soils, it also leads to increased soil
temperatures and deeper active layers which may increase the critical MC. These fires are
expected to increase during the 21st century [21,28], and may be further increased by an
increased critical MC.

Horizontal Spread Rate

Spread rates were measured in two ways. The first was using thermocouples to track the arrival
of a 100°C [57,70] as described earlier. This has the benefit of resolving spread rates with depth.
The second is using the IR images to measure the surface spread as described in Chapter 3 and
is similar the method used in Chapter 4-6.

Figure 60a presents the spread rates measured using the IR images and shows a slightly
decreasing relationship with decreasing condition temperature. This is likely due to the
increased sensible energy required to heat the fuel from lower temperatures and to melt the ice.
MC has a more substantial effect on spread rate, with horizontal spread rates decreasing with
increasing MC. For samples with 50 % MC the spread rate decreased 14% between 21°C and
-7°C conditions. This difference increased to 18% in samples of 100% MC. As such, condition
temperature was found to have a minor but insignificant influence on horizontal spread rates
as confirmed by a nonsignificant linear regression (p>0.05), but was significantly influenced
by MC as confirmed by a significant 2-sample t-test (p<0.05). This trend was similarly found
by Krause and Schmidt [118], who studies smouldering spread rates in samples of wood and
cork dust with starting temperature ranging from 20 to 140°C. They found that spread rates
would approximately double between temperature differences of 120°C. This temperature
difference is vast and incomparable with natural temperature ranges. Additionally, the
experiments conducted by Krause and Schmidt were conducted by igniting a bulk sample at its
centre and measuring the time required to reach the surface. This is a distinctly different
condition than what is studied in this chapter, as upward spread has both a different oxygen
supply path (in the opposite to the direction of spread) and heat loss conditions compared to
horizontal spread. For this reason it is difficult to make a direct comparison between these two
experiments. However, considering the significant temperature range required to obtain a

110
doubling in spread, it supports the notion that the condition temperature has a small, but
present, effect on smouldering spread rates. This is also evident in the spread rates measured
using thermocouple (Figure 60b), which show a minimal difference in spread profile between
condition temperatures, and a larger difference between MC conditions. The spread profile
with depth in each case was largely uniform. No spread rate could be measured at 2 cm depth
in samples of 100% MC because the thermocouples never reached the threshold value of
100°C.

Figure 60 - a) illustrating the effect of condition temperature on spread rate, measured using infrared imaging, in sample of
50% and 100% moisture content. Results are also compared against IR spread measurements of a similar reactor with an
insulated base at room temperature presented by Amin et al.,2020 [85]. IR data was not available for samples with 100%
moisture at 2C. b) showing the effect of depth on spread rate on both samples with 50% and 100% moisture content at various
ambient temperature and compared to spread rates presented by Huang et al.,2016 [70]. Error bars are based on uncertainties
in measurement. Temperature at 2 cm depth in samples of 100% moisture never exceeded 100C and so could not be measured.

For both IR and thermocouple measurements of spread the results are compared against values
found in literature that used the horizontal reactor and the same source of peat as used here
[70,85]. The key difference is that the reactor used in the literature had an insulated base and
was conducted at room temperature, and so the results are most analogous to the 21°C
condition. Results for both samples of 50% and 100% MC were compared. Both the IR and
thermocouples measurements show that horizontal spread rates found here are marginally
smaller than those in literature, which may be due to the quiescent conditions inside the
chamber or the reduced heat losses because of the insulated base used in the standard horizontal
reactor. The horizontal spread rates found by Huang et al. [70] were roughly constant with
depth, in agreement with the results observed here. These similarities between the results found
here and those in reported in literature are despite the additional heat losses and changes to the
depth of burning. This may suggest that the changes burning depth compensate for the changes
in condition temperature. Additionally, the cooling effect of the base may not be significantly

111
felt away from the cold plate. Peat is often referenced as a highly insulating material, having a
profound effect on reducing the active layer thickness, bringing the permafrost closer to the
surface by limiting the heat transfer of the summer heat into the ground [30,119]. Similarly,
peat may limit the heat losses from a smouldering fire into permafrost until they are close
proximity.

Smouldering is shown to be a persistent combustion process despite changes to conditions

temperatures, with minimal changes to spread rates. This provides context to a study performed
by Harden et al. [98] who conducted field studies of regions of permafrost and non-permafrost
sites, and burned and unburned sites. They found that regardless of the presence of permafrost
the burn severity, a measure of fuel consumption was similar. Suggesting an agreeing with our
statement that once smouldering has ignited the dynamics are marginally changed with
temperature. It is clear that further research is required into the relationship between soil
temperature and burning depth.

7.4 Conclusion
A novel new rig for studying smouldering Arctic fires under controlled laboratory conditions
was presented and the effect of condition temperature on smouldering dynamics was explored.
Here I show, for the first time, experimentally, that smouldering wildfires release sufficient
energy to propagate through frozen fuel beds. This illustrates the robust and persistent nature
of smouldering wildfires and explains reports of smouldering fires surviving through the winter
season and resurfacing once the snow has melted [44]. This highlights the importance of
preventing such fires, because once ignited they become difficult to stop. The depth of
smouldering was found to recede significantly deeper into the soil with decreasing condition
temperatures. In samples of 50% MC the depth of burning increased 66% from 21 and -7°C.
This is likely because heat losses decrease with depth, compensating for the increased heat
losses due to colder condition temperatures. Despite this change in dynamics the condition
temperature was shown to insignificantly change the smouldering spread rates. In samples of
50% moisture content horizontal spread rates decreased by 14% and in samples of 100% MC
the spread rate decreased by 18%. This suggests that burning depth is an effective mean of
compensating for changes in conditions such as moisture or temperature.

112
Comparisons were made to the critical moisture content found in experiments at higher
temperatures. The critical moisture content found here was 120% and is lower than the 160%
reported in literature. This decrease is because of the additional heat losses to the cold soil and
cooled bottom boundary. Smouldering fires in arctic region may therefore similarly have lower
critical moisture contents and be more difficult to ignite than warmer regions. This underscores
the importance of protecting these systems, as an increase in soil temperature may increase the
susceptible of soils to burning by expand the moisture contents in which smouldering can
sustained. These findings provide the first experimental insight into the possible dynamics of
arctic fires and demonstrate smouldering fires ability to propagate through frozen soil, thus
improving our understanding of Arctic fires. Additionally, this study demonstrates the value
of this new rig which can be used to conduct further research into the fundamentals of
smouldering arctic fires. Further experiments should be conducted to increase the robustness
of the statistical test to determine the dependence of various smouldering dynamics on
condition temperature.

113
Chapter 8: Conclusion
Smouldering wildfires are a rising global challenge due to their significant carbon emission,
severe health impact, damaging effect on soils, and negative economic impact. However,
research into the smouldering dynamics of peat fires is limited. Smouldering fires is a
phenomena that stretches across scales, from the microscale to the field scale which spans
kilometres. The importance of research at each of these scales has been discussed, as each scale
informs and provides context to studies at the other scales. The focus of this thesis has been on
the mesoscale, as it allows for isolation of variables and identification of trends. In literature
research at this scale is fragmented. With experiments being exploring limited number of
parameters using unique experimental methods. This has made comparing results and gaining
a holistic understanding of smouldering wildfires difficult. To remedy this, in this thesis, I
perform a comprehensive study of the influence of the most significant variables that influences
smouldering spread. I systematically study the three most influential physical soil properties
and 3 prevalent environmental conditions on smouldering dynamics. A crucial element of my
work was the development and presentation of two novel rigs that uniquely allow for the
fundamental study of these parameters: the shallow reactor and the Experimental Low-
temperature Smouldering Apparatus (ELSA).

The shallow reactor enabled decreased experimental times compared to conventional reactors
depths. Additionally, the shallow depth (1.6 cm) allowed the effects of soil and environmental
conditions on smouldering spread to be isolated from depth effects such as changing burning
depth and overhang collapse. Spread was measured using an infrared (IR) camera. I developed
a framework that made the calculation of both horizontal and average in-depth spread rate from
the IR images possible. As such this is the first rig through which both horizontal and in-depth
spread can be easily captured for the same experiment, giving a unique insight into their
relationship. This experimental method can be used by future researchers as an effective means
of isolating the influence of single variables as was done this thesis. The consistent use of this
rig would enable cross experiment comparison providing a large database of results which can
be analysed to further develop fundamental theories on spread.

Experiments with ELSA, on the other hand, used a deeper reactor (10 cm), a liquid cooled
based (between -7 and 21°C), and a chamber with reduced air temperature (~10°C) to

114
experimentally study arctic fires for the first time. This is the first experimental rig designed to
study smouldering arctic fires, a sorely understudied topic. This experiment used
thermocouples, mass balance, and an IR camera to study the spread dynamics. The depth of
the reactor was increased for these experiment so that dynamics such as changing burning depth
could be included and the results would more closely approximate Arctic conditions. This
thesis demonstrates the rigs effective use and could be adopted by future researchers to
investigate the dynamics of smouldering in the Arctic, and to reduce further fragmentation of
experimental work.

The three key physical soil conditions studied were moisture content (MC), inorganic content
(IC), and density. The study of these conditions in literature is incomplete and fragmented
across several different reactors, soil origin, as well as each spread component is studied
separately. Here, these conditions were studied in the same reactor with horizontal and in-depth
spread rate measured simultaneously. Horizontal spread rate decreased linearly with MC, IC,
and density. In-depth spread, on the other hand, increased with MC, decreased with and density,
and was independent to IC (at low MC). The opposite responses of horizontal and in-depth
spread to increasing MC resulted in a unique finding; the global spread rate, a spread rate which
combines both horizontal and in-depth spread was independent to MC, but spread increasingly
downwards. This is first time a global spread rate has been presented. This independent trend
was also confirmed by combining result from literature which studied horizontal and in-depth
spread separately. A continued increase in MC created a behaviour, coined bifurcation,
whereby spread would meander across the reactor, moving in a non-uniform fashion,
extinguishing locally, while spreading at a steady rate elsewhere. This behaviour occurred once
the direction of the global spread rate equalled 45° which corresponds to when the in-depth
spread rate is equal to the horizontal spread rate. This result may be limited the shallow reactor
and deserves further study in deeper reactors.

The use of a single rig to explore all three soil parameters allowed for a holistic analysis of the
effects of soil condition, enabling the development of a novel unifying theory of smoulder
spread. Analysis of the data showed that that horizontal spread and in-depth spread were
controlled by two different parameters. Horizontal spread was governed by the heat sink
density, and were inversely related (Figure 61 a). Heat sink density is the energy required to
heat the soil to the peak smouldering temperature. In-depth spread, on the other hand, was

115
controlled by organic density, the amount of combustible material in the soil, and were also
inversely related (Figure 61b). I theorize that the fundamental difference between these two
spread directions is related to the rate limiting process. Horizontal spread has oxygen readily
available as spread is parallel to the surface of the fuel. As such, I hypothesise that spread is
limited by the time it takes for the peat to be dried, pyrolysed and heated. In-depth spread, I
hypothesise, is limited by how quickly char can be consumed. Oxygen diffusion occurs in the
same direction as the smouldering spread and as such oxygen must penetrate hot reactive char
which limits its penetration depth and so propagation can only advance through the
consumption of the char. This improves our understanding of smouldering spread and the effect
of soil conditions as there is no previous theory of smouldering spread. This new theory allows
the influence of each of these soil condition to be understood through common parameters,
making soils of different compositions comparable.

116
Figure 61 – Combination of all experimental series presented in this thesis. Comparing a) horizontal spread rates as a function
of the heat sink density and b) in-depth spread as a function of organic density. The grey shadow give the range of results of
the soil condition series (MCx, mixed, ICx, and density)

The shallow reactor was also used to study the effect of wind and slope, two prevalent factors
in nature but have received little study in literature. These two factors influence smouldering
dynamics in a very similar manner. Wind concurrent with the spread direction increased both
horizontal and in-depth spread rates. This can ben see in Figure 61 a and b. Both horizontal and
in-depth spread, for all MC, is shifted vertically, above the common trend of the soil condition.
Horizontal spread was increased due to hot gasses convectively heating the fuel thus decreasing
the time required to overcome the heat sink density. In-depth spread on the other hand increased
due to the enhanced oxygen supply provided by the wind increasing the rate of oxidation, as
was evidenced by the increase in burning rate. In opposed conditions, where air flow was in
the opposite direction to spread, wind was shown to have negligible effect on either horizontal
117
or in-depth spread and so falls within the cloud associated with the experiments on soil
condition (Figure 61 a and b). In contrast the influence of slope was found to be weak. Spread
uphill was found to increase horizontal spread, likely as the buoyancy of the emission create
forward spread conditions. However in-depth spread was not significantly influenced.
Spreading downhill, on the other hand resulted in no significant change in horizontal spread
rate and a decrease in in-depth spread rate. This implied a decrease in oxygen supply to the
char. I also, for the first time, show that spread on a slope can be considered as a function of its
angle of spread relative to the horizontal plane. These finding contribute to our understanding
of smouldering wildfires, as they often occur on gradients and are subject to wind.

The final environmental condition, temperature, was conducted in a separate rig; ELSA.
Chapter 7 presented the first experimental evidence that smouldering can produce sufficient
energy to propagating through frozen soils. I showed that soil temperature had only a minor
influence on the peak temperature and smouldering spread rate of the peat fires. However, it
greatly changes the depth of burning. Decreasing soil temperatures resulted in deeper depths
of burning. When comparing between experiments with the cooled based (ELSA) to that of an
insulated base (horizontal reactor), the depth of burning was dramatically more shallow.
Decreasing from 9 cm depth of burning with the insulated base to 4.5 cm. Additionally, it
resulted in a decrease in critical MC from 160% to 120%. This speaks to the importance of
considering the heat losses into deeper soil layers when studying smouldering. The spread rates
measured using IR imagery of these experiments are also plotted in Figure 61a. For the frozen
conditions the heat sink density calculation were modified to include the latent heat of melting
ice (334000 J/kg) and made accommodation for the specific heat capacity of ice (2108 J/kgK),
which is less than water. It is notable that the trend exhibited by these data points is similar to
that in the shallow reactor, only shifted to lower spread rate values. This is likely due to the
deeper nature of this reactor. Further research is required to consider the influence of sample
depth, as this parameter seems important. This is the first experiment which studies the
smouldering dynamics of arctic wildfires.

This thesis has presented two experimental rigs, and a comprehensive study of 6 key conditions
and their effect on smouldering spread. It also presents novel new parameters and a theory on
the governing parameters controlling horizontal and in-depth spread which greatly improves
our fundamental understanding of smouldering wildfires. Some caution is suggested to

118
applying these finding directly to field scale, as the influence of depth is not yet fully
understood, and how this theory should be applied to section of deep peat layer requires further
research. Additionally, the relationship between in-depth spread and horizontal spread deserves
more research as this has been found to be important for self-sustained spread. Furthermore it
provides the first experimental evidence that smouldering fires can propagate through frozen
soils.

119
References:
[1] E.S. Kasischke, M.R. Turetsky, Recent changes in the fire regime across the North
American boreal region - Spatial and temporal patterns of burning across Canada and
Alaska, Geophys. Res. Lett. 33 (2006).

[2] S.E. Page, F. Siegert, J.O. Rieley, H.-D. V. Boehm, A. Jaya, S. Limin, The amount of
carbon released from peat and forest fires in Indonesia during 1997, Nature. 420 (2002)
61–65.

[3] V. Huijnen, M.J. Wooster, J.W. Kaiser, D.L.A. Gaveau, J. Flemming, M. Parrington, A.
Inness, D. Murdiyarso, B. Main, M. Van Weele, Fire carbon emissions over maritime
southeast Asia in 2015 largest since 1997, Sci. Rep. 6 (2016) 1–8.

[4] M.R. Turetsky, B. Benscoter, S. Page, G. Rein, G.R. van der Werf, A. Watts, Global
vulnerability of peatlands to fire and carbon loss, Nat. Geosci. 8 (2015) 11–14.

[5] D. Drysdale, An Introduction to Fire Dynamics: Third Edition, 2011.

[6] S. Payette, L. Filion, White spruce expansion at the tree line and recent climatic change,
Can. J. For. Res. 15 (1985) 241–251.

[7] L. Viereck, L. Schandelmeier, Effects of fire in Alaska and adjacent Canada: a literature
review, 1980.

[8] S.L. Manzello, E.I.D. Foote, Characterizing Firebrand Exposure from Wildland-Urban
Interface (WUI) Fires: Results from the 2007 Angora Fire, Fire Technol. 50 (2014) 105–
124.

[9] S.E. Caton, R.S.P. Hakes, D.J. Gorham, A. Zhou, M.J. Gollner, Review of Pathways for
Building Fire Spread in the Wildland Urban Interface Part I: Exposure Conditions, Fire
Technol. 53 (2017) 429–473.

[10] G. Rein, Smouldering Fires and Natural Fuels, in: Fire Phenom. Earth Syst. An
Interdiscip. Guid. to Fire Sci., 2013: pp. 15–33.

[11] I. Bertschi, R.J. Yokelson, D.E. Ward, R.E. Babbitt, R.A. Susott, J.G. Goode, W.M. Hao,
Trace gas and particle emissions from fires in large diameter and belowground biomass
fuels, J. Geophys. Res. Atmos. 108 (2003) 8472.

120
[12] G.M. Davies, A. Gray, G. Rein, C.J. Legg, Peat consumption and carbon loss due to
smouldering wildfire in a temperate peatland, For. Ecol. Manage. 308 (2013) 169–177.

[13] P.-T. Huang, M. Patel, M.C. Santagata, A. Bobet, Classification of Organic Soils, 2009.

[14] Z. Shi, S.D. Allison, Y. He, P.A. Levine, A.M. Hoyt, J. Beem-Miller, Q. Zhu, W.R.
Wieder, S. Trumbore, J.T. Randerson, The age distribution of global soil carbon inferred
from radiocarbon measurements, Nat. Geosci. (2020).

[15] D.L. Mokma, ORGANIC SOILS, in: Encycl. Soils Environ., Elsevier, 2005: pp. 118–
129.

[16] X.J. Walker, J.L. Baltzer, S.G. Cumming, N.J. Day, C. Ebert, S. Goetz, J.F. Johnstone,
S. Potter, B.M. Rogers, E.A.G. Schuur, M.R. Turetsky, M.C. Mack, Increasing wildfires
threaten historic carbon sink of boreal forest soils, Nature. 572 (2019) 520–523.

[17] T. V. Glukhova, A.A. Sirin, Losses of Soil Carbon upon a Fire on a Drained Forested
Raised Bog, Eurasian Soil Sci. 51 (2018) 542–549.

[18] M. Warren, S. Frolking, Z. Dai, S. Kurnianto, Impacts of land use, restoration, and
climate change on tropical peat carbon stocks in the twenty-first century: implications
for climate mitigation, Mitig. Adapt. Strateg. Glob. Chang. 22 (2017) 1041–1061.

[19] L.S. Wijedasa, J. Jauhiainen, M. Könönen, M. Lampela, H. Vasander, M.C. Leblanc, S.

Evers, T.E.L. Smith, C.M. Yule, H. Varkkey, M. Lupascu, F. Parish, I. Singleton, G.R.
Clements, S.A. Aziz, M.E. Harrison, S. Cheyne, G.Z. Anshari, E. Meijaard, J.E.
Goldstein, S. Waldron, K. Hergoualc’h, R. Dommain, S. Frolking, C.D. Evans, M.R.C.
Posa, P.H. Glaser, N. Suryadiputra, R. Lubis, T. Santika, R. Padfield, S. Kurnianto, P.
Hadisiswoyo, T.W. Lim, S.E. Page, V. Gauci, P.J. Van Der Meer, H. Buckland, F.
Garnier, M.K. Samuel, L.N.L.K. Choo, P. O’Reilly, M. Warren, S. Suksuwan, E.
Sumarga, A. Jain, W.F. Laurance, J. Couwenberg, H. Joosten, R. Vernimmen, A.
Hooijer, C. Malins, M.A. Cochrane, B. Perumal, F. Siegert, K.S.H. Peh, L.P. Comeau,
L. Verchot, C.F. Harvey, A. Cobb, Z. Jaafar, H. Wösten, S. Manuri, M. Müller, W.
Giesen, J. Phelps, D.L. Yong, M. Silvius, B.M.M. Wedeux, A. Hoyt, M. Osaki, T.
Hirano, H. Takahashi, T.S. Kohyama, A. Haraguchi, N.P. Nugroho, D.A. Coomes, L.P.
Quoi, A. Dohong, H. Gunawan, D.L.A. Gaveau, A. Langner, F.K.S. Lim, D.P. Edwards,
X. Giam, G. Van Der Werf, R. Carmenta, C.C. Verwer, L. Gibson, L. Gandois, L.L.B.
Graham, J. Regalino, S.A. Wich, J. Rieley, N. Kettridge, C. Brown, R. Pirard, S. Moore,
121
 B.R. Capilla, U. Ballhorn, H.C. Ho, A. Hoscilo, S. Lohberger, T.A. Evans, N. Yulianti,
G. Blackham, Onrizal, S. Husson, D. Murdiyarso, S. Pangala, L.E.S. Cole, L. Tacconi,
H. Segah, P. Tonoto, J.S.H. Lee, G. Schmilewski, S. Wulffraat, E.I. Putra, M.E. Cattau,
R.S. Clymo, R. Morrison, A. Mujahid, J. Miettinen, S.C. Liew, S. Valpola, D. Wilson,
L. D’Arcy, M. Gerding, S. Sundari, S.A. Thornton, B. Kalisz, S.J. Chapman, A.S.M. Su,
I. Basuki, M. Itoh, C. Traeholt, S. Sloan, A.K. Sayok, R. Andersen, Denial of long-term
issues with agriculture on tropical peatlands will have devastating consequences, Glob.
Chang. Biol. 23 (2017) 977–982.

[20] B. Wang, X. Luo, Y.M. Yang, W. Sun, M.A. Cane, W. Cai, S.W. Yeh, J. Liu, Historical
change of El Niño properties sheds light on future changes of extreme El Niño, Proc.
Natl. Acad. Sci. U. S. A. 116 (2019) 22512–22517.

[21] M. Flannigan, A.S. Cantin, W.J. De Groot, M. Wotton, A. Newbery, L.M. Gowman,
Global wildland fire season severity in the 21st century, For. Ecol. Manage. 294 (2013)
54–61.

[22] M.A. Santoso, X. Huang, N. Prat-guitart, E. Christensen, Fire Effects on Soil Properties,
CSIRO Publishing, 2019.

[23] E.B. Wiggins, C.I. Czimczik, G.M. Santos, Y. Chen, X. Xu, S.R. Holden, J.T. Randerson,
C.F. Harvey, F.M. Kai, L.E. Yu, Smoke radiocarbon measurements from Indonesian
fires provide evidence for burning of millennia-aged peat, Proc. Natl. Acad. Sci. 115
(2018) 12419–12424.

[24] S.N. Koplitz, L.J. Mickley, M.E. Marlier, J.J. Buonocore, P.S. Kim, T. Liu, M.P.
Sulprizio, R.S. DeFries, D.J. Jacob, J. Schwartz, M. Pongsiri, S.S. Myers, Public health
impacts of the severe haze in Equatorial Asia in September–October 2015:
demonstration of a new framework for informing fire management strategies to reduce
downwind smoke exposure, Environ. Res. Lett. 11 (2016) 094023.

[25] A.L. Hinwood, C.M. Rodriguez, Potential health impacts associated with peat smoke:
A review, J. R. Soc. West. Aust. 88 (2005) 133–138.

[26] S.K. Uda, L. Hein, D. Atmoko, Assessing the health impacts of peatland fires: a case
study for Central Kalimantan, Indonesia, Environ. Sci. Pollut. Res. 26 (2019) 31315–
31327.

122
[27] M.C. Mack, M.S. Bret-Harte, T.N. Hollingsworth, R.R. Jandt, E.A.G. Schuur, G.R.
Shaver, D.L. Verbyla, Carbon loss from an unprecedented Arctic tundra wildfire, Nature.
475 (2011) 489–492.

[28] F.S. Hu, P.E. Higuera, P. Duffy, M.L. Chipman, A. V. Rocha, A.M. Young, R. Kelly,
M.C. Dietze, Arctic tundra fires: Natural variability and responses to climate change,
Front. Ecol. Environ. 13 (2015) 369–377.

[29] V. Myroniuk, R. Paugam, K. Stebel, N. Evangeliou, A. Kylling, A. Stohl, S. Zibtsev, S.

Eckhardt, Open fires in Greenland in summer 2017: transport, deposition and radiative
effects of BC, OC and BrC emissions, Atmos. Chem. Phys. 19 (2019) 1393–1411.

[30] H.M. French, Permafrost, in: Periglac. Environ., John Wiley & Sons Ltd,., West Sussex,
England, 2013: pp. 83–115.

[31] C. Tarnocai, J.G. Canadell, E.A.G. Schuur, P. Kuhry, G. Mazhitova, S. Zimov, Soil
organic carbon pools in the northern circumpolar permafrost region, Global Biogeochem.
Cycles. 23 (2009) 1–11.

[32] T.E. Osterkamp, C.R. Burn, Permafrost, Permafrost. (2003) 1717–1729.

[33] M.R. Turetsky, E.S. Kane, J.W. Harden, R.D. Ottmar, K.L. Manies, E. Hoy, E.S.
Kasischke, Recent acceleration of biomass burning and carbon losses in Alaskan forests
and peatlands, Nat. Geosci. 4 (2011) 27–31.

[34] E. a. G. Schuur, B. Abbott, Climate change: High risk of permafrost thaw, Nature. 480
(2011) 32–33.

[35] K. Yoshikawa, W.R. Bolton, V.E. Romanovsky, M. Fukuda, L.D. Hinzman, Impacts of
wildfire on the permafrost in the boreal forests of Interior Alaska, J. Geophys. Res. 108
(2002) 16–17.

[36] C.M. Gibson, L.E. Chasmer, D.K. Thompson, W.L. Quinton, M.D. Flannigan, D.
Olefeldt, Wildfire as a major driver of recent permafrost thaw in boreal peatlands, Nat.
Commun. 9 (2018).

[37] A.H. MacDougall, C. a. Avis, A.J. Weaver, Significant contribution to climate warming
from the permafrost carbon feedback, Nat. Geosci. 5 (2012) 719–721.

[38] M.C. Serreze, R.G. Barry, Processes and impacts of Arctic amplification: A research

123
 synthesis, Glob. Planet. Change. 77 (2011) 85–96.

[39] W.H. Frandsen, The influence of moisture and mineral soil on the combustion limits of
smoldering forest duff, Can. J. For. Res. 17 (1987) 1540–1544.

[40] S. Veraverbeke, B.M. Rogers, M.L. Goulden, R.R. Jandt, C.E. Miller, E.B. Wiggins, J.T.
Randerson, Lightning as a major driver of recent large fire years in North American
boreal forests, Nat. Clim. Chang. 7 (2017) 529–534.

[41] A.M. Young, P.E. Higuera, P.A. Duffy, F.S. Hu, Climatic thresholds shape northern
high-latitude fire regimes and imply vulnerability to future climate change, Ecography
(Cop.). (2016) 1–12.

[42] A. Goldstein, W.R. Turner, S.A. Spawn, K.J. Anderson-Teixeira, S. Cook-Patton, J.

Fargione, H.K. Gibbs, B. Griscom, J.H. Hewson, J.F. Howard, J.C. Ledezma, S. Page,
L.P. Koh, J. Rockström, J. Sanderman, D.G. Hole, Protecting irrecoverable carbon in
Earth’s ecosystems, Nat. Clim. Chang. 10 (2020) 287–295.

[43] R. Scholten, S. Veraverbeke, Fires can overwinter in boreal forests of North America,
in: 2020.

[44] B.A. Vaughan, ’ Zombie ’ fires are burning the Arctic after smouldering under snow,
New Sci. (2020) 1–5.

[45] M.A. Santoso, E.G. Christensen, J. Yang, G. Rein, Review of the Transition From
Smouldering to Flaming Combustion in Wildfires, Front. Mech. Eng. 5 (2019).

[46] R.M. Hadden, G. Rein, C.M. Belcher, Study of the competing chemical reactions in the
initiation and spread of smouldering combustion in peat, Proc. Combust. Inst. 34 (2013)
2547–2553.

[47] X. Huang, G. Rein, Interactions of Earth’s atmospheric oxygen and fuel moisture in
smouldering wildfires, Sci. Total Environ. 572 (2016) 1440–1446.

[48] G. Rein, Smoldering Combustion, in: SFPE Handb. Fire Prot. Eng. Fifth Ed., 2016: pp.
581–603.

[49] M.À. Xifré-Salvadó, N. Prat-Guitart, M. Francos, X. Úbeda, M. Castellnou,

Smouldering Combustion Dynamics of a Soil from a Pinus halepensis Mill. Forest. A
Case Study of the Rocallaura Fires in Northeastern Spain, Appl. Sci. 10 (2020) 3449.

124
[50] V. Mamleev, S. Bourbigot, J. Yvon, Kinetic analysis of the thermal decomposition of
cellulose: The main step of mass loss, J. Anal. Appl. Pyrolysis. 80 (2007) 151–165.

[51] F. Richter, G. Rein, Heterogeneous kinetics of timber charring at the microscale, J. Anal.
Appl. Pyrolysis. 138 (2019) 1–9.

[52] X. Huang, G. Rein, Smouldering combustion of peat in wildfires: Inverse modelling of

the drying and the thermal and oxidative decomposition kinetics, Combust. Flame. 161
(2014) 1633–1644.

[53] A. Usup, Y. Hashimoto, H. Takahashi, H. Hayasaka, Combustion and thermal

characteristics of peat fire in tropical peatland in Central Kalimantan, Indonesia, Tropics.
14 (2004) 1–19.

[54] G. Rein, S. Cohen, A. Simeoni, Carbon emissions from smouldering peat in shallow and
strong fronts, Proc. Combust. Inst. 32 (2009) 2489–2496.

[55] R. Hungerford, W.H. Frandsen, K.C. Ryan, Ignition and Burning Characteristics of
Organic Soils, in: S.I. Cerulean, R.T. Engstrom (Eds.), Proc. 19th Tall Timbers Fire Ecol.
Conf. Fire Wetl. a Manag. Perspect. Tall Timbers Res. Stn., Tallahassee, 1995: pp. 79–
91.

[56] N. Prat-Guitart, G. Rein, R.M. Hadden, C.M. Belcher, J.M. Yearsley, Propagation
probability and spread rates of self-sustained smouldering fires under controlled
moisture content and bulk density conditions, Int. J. Wildl. Fire. 25 (2016) 456–465.

[57] Y. Hu, E.G. Christensen, H.M.F. Amin, T.E.L. Smith, G. Rein, Experimental study of
moisture content effects on the transient gas and particle emissions from peat fires,
Combust. Flame. 209 (2019) 408–417.

[58] C. Roulston, C. Paton-Walsh, T.E.L. Smith, A. Guérette, S. Evers, C.M. Yule, G. Rein,
G.R. Van der Werf, Fine Particle Emissions From Tropical Peat Fires Decrease Rapidly
With Time Since Ignition, J. Geophys. Res. Atmos. 123 (2018) 5607–5617.

[59] X. Huang, G. Rein, Downward spread of smouldering peat fire: the role of moisture,
density and oxygen supply, Int. J. Wildl. Fire. 26 (2017) 907.

[60] F. He, W. Yi, Y. Li, J. Zha, B. Luo, Effects of fuel properties on the natural downward
smoldering of piled biomass powder: Experimental investigation, Biomass and

125
 Bioenergy. 67 (2014) 288–296.

[61] Y. Hu, W. Cui, G. Rein, Haze emissions from smouldering peat: The roles of inorganic
content and bulk density, Fire Saf. J. 113 (2020) 102940.

[62] J. Yang, H. Chen, Natural Downward Smouldering of Peat : Effects of Inorganic

Content and Piled Bed, Fire Technol. 54 (2018) 1219–1247.

[63] W.H. Frandsen, Ignition probability of organic soils, Can. J. For. Res. 27 (1997) 1471–
1477.

[64] B.W. Benscoter, D.K. Thompson, J.M. Waddington, M.D.D. Flannigan, B.M. Wotton,
W.J.J. de Groot, M.R. Turetsky, Interactive effects of vegetation, soil moisture and bulk
density on depth of burning of thick organic soils, Int. J. Wildl. Fire. 20 (2011) 418.

[65] X. Huang, G. Rein, H. Chen, Computational smoldering combustion: Predicting the

roles of moisture and inert contents in peat wildfires, Proc. Combust. Inst. 35 (2015)
2673–2681.

[66] R.A. Hartford, Smoldring combustion limits as influenced by moisture, mineral content,
and organic bulk density, University of Montana, 1993.

[67] X. Huang, G. Rein, Computational study of critical moisture and depth of burn in peat
fires, Int. J. Wildl. Fire. 24 (2015) 798–808.

[68] E.C. Garlough, C.R. Keyes, Influences of moisture content, mineral content and bulk
density on smouldering combustion of ponderosa pine duff mounds, Int. J. Wildl. Fire.
20 (2011) 589–596.

[69] T.J. Ohlemiller, Smoldering combustion propagation through a permeable horizontal

fuel layer, Combust. Flame. 81 (1990) 341–353.

[70] X. Huang, F. Restuccia, M. Gramola, G. Rein, Experimental study of the formation and
collapse of an overhang in the lateral spread of smouldering peat fires, Combust. Flame.
168 (2016) 393–402.

[71] K.N. Palmer, Smouldering combustion in dusts and fibrous materials, Combust. Flame.
1 (1957) 129–154.

[72] K. Miyanishi, E.A. Johnson, Process and patterns of duff consumption in the mixedwood
boreal forest, Can. J. For. Res. 32 (2002) 1285–1295.
126
[73] J.P. Valdivieso, J. de D. Rivera, Effect of Wind on Smoldering Combustion Limits of
Moist Pine Needle Beds, Fire Technol. 50 (2014) 1589–1605.

[74] R.W. Wein, Fire behaviour and ecological effects on organic terrain., Role Fire North.
Circumpolar Ecosyst. (1983) 81–93.

[75] A.L. Sinclair, L.L.B. Graham, E.I. Putra, B.H. Saharjo, G. Applegate, S.P. Grover, M.A.
Cochrane, Effects of distance from canal and degradation history on peat bulk density
in a degraded tropical peatland, Sci. Total Environ. 699 (2020) 134199.

[76] N. Prat-Guitart, C.M. Belcher, D.K. Thompson, P. Burns, J.M. Yearsley, Fine-scale
distribution of moisture in the surface of a degraded blanket bog and its effects on the
potential spread of smouldering fire, Ecohydrology. 10 (2017) 1–11.

[77] G. Rein, N. Cleaver, C. Ashton, P. Pironi, J.L. Torero, The severity of smouldering peat
fires and damage to the forest soil, CATENA. 74 (2008) 304–309.

[78] N. Prat-Guitart, G. Rein, R.M. Hadden, C.M. Belcher, J.M. Yearsley, Effects of spatial
heterogeneity in moisture content on the horizontal spread of peat fires, Sci. Total
Environ. 572 (2016) 1422–1430.

[79] D. Cancellieri, V. Leroy-Cancellieri, E. Leoni, A. Simeoni, A.Y. Kuzin, A.I. Filkov, G.

Rein, Kinetic investigation on the smouldering combustion of boreal peat, Fuel. 93
(2012) 479–485.

[80] J. Yang, H. Chen, W. Zhao, J. Zhou, Combustion kinetics and emission characteristics
of peat by using TG-FTIR technique, J. Therm. Anal. Calorim. 124 (2016) 519–528.

[81] R.M. Hadden, G. Rein, C.M. Belcher, Study of the competing chemical reactions in the
initiation and spread of smouldering combustion in peat, Proc. Combust. Inst. 34 (2013)
2547–2553.

[82] F. Restuccia, X. Huang, G. Rein, Self-ignition of natural fuels: Can wildfires of carbon-
rich soil start by self-heating?, Fire Saf. J. 91 (2017) 828–834.

[83] A.J. Kohlenberg, M.R. Turetsky, D.K. Thompson, B.A. Branfireun, C.P.J. Mitchell,
Controls on boreal peat combustion and resulting emissions of carbon and mercury,
Environ. Res. Lett. 13 (2018).

[84] D.A. Cowan, W.G. Page, B.W. Butler, D.L. Blunck, Effects of fuel characteristics on

127
 horizontal spread rate and ground surface temperatures of smouldering duff, Int. J. Wildl.
Fire. (2020) 1–15.

[85] H.M.F. Amin, Y. Hu, G. Rein, Spatially resolved horizontal spread in smouldering peat
combining infrared and visual diagnostics, Combust. Flame. 220 (2020) 328–336.

[86] Y. Hu, E. Christensen, F. Restuccia, G. Rein, Transient gas and particle emissions from
smouldering combustion of peat, Proc. Combust. Inst. 37 (2019) 4035–4042.

[87] J. Reardon, R. Hungerford, K. Ryan, Factors affecting sustained smouldering in organic

soils from pocosin and pond pine woodland wetlands, Int. J. Wildl. Fire. 16 (2007) 107–
118.

[88] C. Zaccone, G. Rein, V. D’Orazio, R.M. Hadden, C.M. Belcher, T.M. Miano,
Smouldering fire signatures in peat and their implications for palaeoenvironmental
reconstructions, Geochim. Cosmochim. Acta. 137 (2014) 134–146.

[89] X. Huang, G. Rein, Upward-and-downward spread of smoldering peat fire, Proc.

Combust. Inst. 37 (2019) 4025–4033.

[90] D. Wilson, S.D. Dixon, R.R.E. Artz, T.E.L. Smith, C.D. Evans, H.J.F. Owen, E. Archer,
F. Renou-Wilson, Derivation of greenhouse gas emission factors for peatlands managed
for extraction in the Republic of Ireland and the United Kingdom, Biogeosciences. 12
(2015) 5291–5308.

[91] C. Ryu, Y. Bin Yang, A. Khor, N.E. Yates, V.N. Sharifi, J. Swithenbank, Effect of fuel
properties on biomass combustion: Part I. Experiments - Fuel type, equivalence ratio
and particle size, Fuel. 85 (2006) 1039–1046.

[92] Y. Hu, N. Fernandez-Anez, T. E.L Smith, G. Rein, Review of emissions from

smouldering peat fires and their contribution to regional haze episodes, Int. J. Wildl.
Fire. 27 (2018) 293–312.

[93] B.M. Jones, C. a. Kolden, R. Jandt, J.T. Abatzoglou, F. Urban, C.D. Arp, Fire Behavior,
Weather, and Burn Severity of the 2007 Anaktuvuk River Tundra Fire, North Slope,
Alaska, Arctic, Antarct. Alp. Res. 41 (2009) 309–316.

[94] A. V Rocha, M.M. Loranty, P.E. Higuera, M.C. Mack, F.S. Hu, B.M. Jones, A.L. Breen,
E.B. Rastetter, S.J. Goetz, G.R. Shaver, The footprint of Alaskan tundra fires during the

128
 past half-century: implications for surface properties and radiative forcing, Environ. Res.
Lett. 7 (2012) 044039.

[95] F. Sedano, J.T. Randerson, Multi-scale influence of vapor pressure deficit on fire
ignition and spread in boreal forest ecosystems, Biogeosciences. 11 (2014) 3739–3755.

[96] U. Ballhorn, F. Siegert, M. Mason, S. Limin, Derivation of burn scar depths and
estimation of carbon emissions with LIDAR in Indonesian peatlands, Proc. Natl. Acad.
Sci. 106 (2009) 21213–21218.

[97] E. Pastor, I. Oliveras, E. Urquiaga-Flores, J.A. Quintano-Loayza, M.I. Manta, E. Planas,

A new method for performing smouldering combustion field experiments in peatlands
and rich-organic soils, Int. J. Wildl. Fire. 26 (2017) 1040.

[98] J.W. Harden, K.L. Manies, M.R. Turetsky, J.C. Neff, Effects of wildfire and permafrost
on soil organic matter and soil climate in interior Alaska, Glob. Chang. Biol. 12 (2006)
2391–2403.

[99] G. Rein, Smoldering Combustion, in: SFPE Handb. Fire Prot. Eng., Springer New York,
New York, NY, 2016: pp. 581–603.

[100] M.G. Hille, S.L. Stephens, Mixed Conifer Forest Duff Consumption during Prescribed
Fires : Tree Crown Impacts, For. Sci. 51 (2005) 417–424.

[101] N. Prat, C. Belcher, R. Hadden, G. Rein, J. Yearsley, A laboratory study of the effect of
moisture content on the spread of smouldering in peat fires, Flamma. 6 (2015) 35–38.

[102] J. Yang, H. Chen, N. Liu, Modeling of Two-Dimensional Natural Downward

Smoldering of Peat, Energy & Fuels. 30 (2016) 8765–8775.

[103] B.D. Smucker, T.C. Mulky, D.A. Cowan, K.E. Niemeyer, D.L. Blunck, Effects of fuel
content and density on the smoldering characteristics of cellulose and hemicellulose,
Proc. Combust. Inst. 37 (2019) 4107–4116.

[104] W.H. Frandsen, Burning Rate of Smouldering peat, Northwest Sci. 65 (1991) 166–172.

[105] M.C. Lukenbach, K.J. Hokanson, P.A. Moore, K.J. Devito, N. Kettridge, D.K.
Thompson, B.M. Wotton, R.M. Petrone, J.M. Waddington, Hydrological controls on
deep burning in a northern forested peatland, Hydrol. Process. 29 (2015) 4114–4124.

[106] Y. Hu, E.G. Christensen, H.M.F. Amin, T.E.L. Smith, G. Rein, Experimental study of
129
 moisture content effects on the transient gas and particle emissions from peat fires,
Combust. Flame. 209 (2019).

[107] T.J. Ohlemiller, Modeling of smoldering combustion propagation, Prog. Energy

Combust. Sci. 11 (1985) 277–310.

[108] T.J. Ohlemiller, D.A. Lucca, An experimental comparison of forward and reverse
smolder propagation in permeable fuel beds, Combust. Flame. 54 (1983) 131–147.

[109] L. Yermán, H. Wall, J. Torero, J.I. Gerhard, Y.-L. Cheng, Smoldering Combustion as a
Treatment Technology for Feces: Sensitivity to Key Parameters, Combust. Sci. Technol.
188 (2016) 968–981.

[110] P. Palamba, R. Allo, E. Kosasih, Y. Nugroho, Effect of surface dry layer thickness on
the smoldering combustion of a stratified moisture content peat layer, Int. J. Adv. Sci.
Technol. 29 (2020) 2117–2139.

[111] L. Yermán, H. Wall, J.L. Torero, Experimental investigation on the destruction rates of
organic waste with high moisture content by means of self-sustained smoldering
combustion, Proc. Combust. Inst. 36 (2017) 4419–4426.

[112] D.X. Viegas, Slope and wind effects on fire propagation, Int. J. Wildl. Fire. 13 (2004)
143.

[113] B.W. Butler, W.R. Anderson, E.A. Catchpole, Influence of Slope on Fire Spread Rate,
USDA For. Serv. Proc. (2007) 75–82.

[114] J.L. Torero, A.C. Fernandez-pello, M. Kitano, Opposed Forced Flow Smoldering of
Polyurethane Foam, Combust. Sci. Technol. 91 (1993) 95–117.

[115] S.L. Manzello, T.G. Cleary, J.R. Shields, J.C. Yang, On the ignition of fuel beds by
firebrands, Fire Mater. 30 (2006) 77–87.

[116] J.L. Dupuy, Slope and fuel load effects on fire behaviour :Laboratory experiments in
pine needles fuel beds, Int. J. Wildl. Fire. 5 (1995) 153–164.

[117] M.C. Mack, M.S. Bret-Harte, T.N. Hollingsworth, R.R. Jandt, E. a G. Schuur, G.R.
Shaver, D.L. Verbyla, Carbon loss from an unprecedented Arctic tundra wildfire.,
Nature. 475 (2011) 489–92.

[118] U. Krause, M. Schmidt, The influence of initial conditions on the propagation of

130
 smouldering fires in dust accumulations, J. Loss Prev. Process Ind. 14 (2001) 527–532.

[119] J.P. Fisher, C. Estop-Aragonés, A. Thierry, D.J. Charman, S.A. Wolfe, I.P. Hartley, J.B.
Murton, M. Williams, G.K. Phoenix, The influence of vegetation and soil characteristics
on active-layer thickness of permafrost soils in boreal forest, Glob. Chang. Biol. 22
(2016) 3127–3140.

131
LIST OF FIGURES

Figure 1 - Demonstrating a simplified schematic of the fundamental fronts in smouldering

combustion in soils. ................................................................................................................. 18
Figure 2 - Smouldering through commercial peat sample, illustrating characteristic
smouldering fronts; fresh peat includes a region of drying. The char and ash region can also
clearly be seen by the colour and structure change of the fuel, resulting in a cloud of visible
emissions. ................................................................................................................................. 18
Figure 3 - Image of a smouldering peat experiment with 120% moisture content. It can be seen
that the smouldering front as indicated by the char layer (the black material indicated with an
arrow) exists far below the surface of the peat. ....................................................................... 21
Figure 4 - Scale matters. Experiments are conducted on all scales from sub-millimetre scale up
to metre and kilometre scale. Each scale provides useful insights into understanding the large-
scale wildfire phenomena. (Micro-scale image by David Lazaro) .......................................... 26
Figure 5 - Schematic of smouldering front in horizontal and in-depth spread in peat fires. Sh
and Sd indicate the horizontal spread and in-depth spread rate. Source: [52], CCBY ............ 29
Figure 6 - (a) A reactor for testing horizontal spread; and (b) an exploded view for the purposes
of construction. ........................................................................................................................ 30
Figure 7 - Infrared image seen from above the smouldering spread through dry peat in the
horizontal reactor. The igniter is buried below the surface at the boundary at the top of the
images. ..................................................................................................................................... 31
Figure 8 - Temperature profile of smouldering dry peat in a horizontal reactor, measured at 5
cm depth. Temperatures 11 cm from the front wall can be seen to rise only 15 min after the 30
min ignition period. .................................................................................................................. 32
Figure 9 - Illustration of a column reactor. Typical dimensions of which are 10 × 10 × 30 cm,
with a coil igniter submerged 5 cm below the free surface. .................................................... 33
Figure 10 - A shallow reactor measuring 40 × 40 × 1.6 cm, with a short 1-cm long ignition coil
placed at the centre of the reactor wired through the base via small ceramic tubes. ............... 34
Figure 11 - Infrared image taken from above the shallow reactor showing the spread of a
leading and trailing edge in a layer of dry peat. ....................................................................... 35
Figure 12 - Graph illustrating the influence of wind flow on mass loss rate of 800 g of dry peat
in the horizontal reactor. At 30 cm above the reactor, the vertical airflow velocity was measured

132
to be 0.7 and 0.3 m s–1 for high and low ventilation settings, respectively. All lines are
smoothed over 30 min and the clouds indicate the range of experimental results. ................. 37
Figure 13 - Graph illustrating he influence of the coil access point size on the mass loss rate of
800 g of dry smouldering peat in the horizontal reactor. Three sizes were experimented with;
0 mm, 1.5 mm, and 10 mm. The lines are fitted against the smoothed average mass loss rates
of 3 repeated experiments for every condition, and the clouds indicate the range of the
experimental results. ................................................................................................................ 38
Figure 14 - Comparing infrared images of sample surface 30 min after ignition of smouldering
dry peat in the horizontal reactor. Both samples with coil access points greater than 0 mm show
smouldering reaches the surface across the width of the box by the end of the ignition protocol.
.................................................................................................................................................. 38
Figure 15 - Average MLR plots of dry peat, examining the influence of sample preparation and
resulting density changes. The peak values are shown to remain constant between the
experiments, but the second peak of the unshaken sample is 40 min earlier than that of the
shaken. The clouds indicate the range of the experimental results. ......................................... 39
Figure 16 - ‘Cloud of variability’, encompassing all experimental data presented on dry peat in
the horizontal reactor. .............................................................................................................. 40
Figure 17 - A typical horizontal distribution of a burned surface peat layer with several
projections of the fire front line at Plot-2 on 10 August 2002. Source: [53]. .......................... 41
Figure 18 - Illustration of the experimental setup, displaying a central ignition using a wound
coil and key features observable in IR imaging. ...................................................................... 45
Figure 19 - Measurements and the relationship between organic density and soil properties, a)
showing the influence of moisture content and compare to result presented by Huang et al. [70]
b) the effect of inorganic content on the organic density. Error bars are calculated from
measurement uncertainties. ...................................................................................................... 46
Figure 20 - Illustration of experimental set up for studying the influence of slope on
smouldering dynamics. When when θs is equal to 0 and 180°, smouldering spreads uphill or
downhill, respectively. ............................................................................................................. 48
Figure 21 –A cross section illustrating the key features of smouldering dynamics and
framework of analysis. ............................................................................................................. 49
Figure 22 - Illustrating the method of IR image analysis using example images from a sample
with 40% MC and 2.5% IC. The plot present the pixel intensity profile from the same transect

133
at 40, 80 and 120 min after ignition (top images) and how using a normalized pixel intensity
value of 0.2 was used to track the leading edge (grey line). .................................................... 50
Figure 23 - Illustrating the independence of both leading and trailing spread rates with time.
Figure a) shows the distance from the centre of the leading and trailing edge with time and
figure b) exemplifies the consistent spread rate of the leading and trailing edge, allowing for
averaged values to be used for discussion. Plots are generated from an experiment with 40 %
MC and 40% IC. The clouds represent the standard deviation across all 60 transect ............. 52
Figure 24 – Presenting examples of the consistent spread rate in all directions relative to a)
wind and b) slope. Notably, wind is applied after the end of the ignition protocol, as evidenced
by the increase after 0.5h. ........................................................................................................ 52
Figure 25 - Illustration of the Experimental low-temperature smouldering apparatus (ELSA).
On the left shows key features of the ractor which is filled soil. The internal dimension of the
reactor measures 20 × 20 × 10 cm. On the right, key features of the rig is highlighted. The
dimensions of the chamber is 120 × 80 × 80 cm. .................................................................... 54
Figure 26 - Average temperatures of the base temperature, chamber temperature, and initial
soil temperature for all 3 experimental condition temperatures (Tcond). The error bars show the
range of measured values across all experiments. ................................................................... 56
Figure 27 - Demonstrating the method of capturing the spread rates using IR imagery. Left
image shows an example grayscale image. Middle image is after the image orientation has been
corrected, and the transect lines have been drawn. The width and length of the reactor are used
to calibrate the scale of the pixels. The image on the right demonstrates the pixel saturation
associated with the high radiative heat emissions from burning of peat, and the clear definition
of the leading edge (spreading from top to bottom). ............................................................... 57
Figure 28 - Demonstrating the analysis method in measuring spread rate from IR imagery. a)
depicting the grayscale pixel intensity along a transect at 110 and 350 min after ignition b)
shows the measured position of the smouldering front across 7 transects with time, and
illustrates the steady horizontal spread rate of the smouldering front. .................................... 58
Figure 29 - Illustration of the stages and key features of smouldering. Horizontal spread is
along the peat layer, and in-depth spread is downwards. The vector combination of the two
spread components give the global spread rate (SG) with a direction (ф). The leading edge is
the first location of char oxidation and the trailing edge is the last location of char oxidation.
.................................................................................................................................................. 61

134
Figure 30 - Shows the change in smouldering width as affected by a) moisture content for two
inorganic contents and b) inorganic content for two moisture contents. ................................. 64
Figure 31 - The influence of soil properties on horizontal spread rate: a) The effect of MC on
horizontal spread rate for soils with two IC values, b) the effect of IC on the horizontal spread
rate for two different MC values c) plot of the influence of both MC and IC on horizontal
spread rate. (Slopes: 2.5% IC, -0.026cm/h·%MC, omitting the first data point; 40% IC, -0.027
cm/h·%MC; 0% MC, -0.106 cm/h·%IC) .................................................................................... 65
Figure 32 - Influence of soil properties on in-depth spread rate: a) shows increasing relationship
between MC and in-depth spread for two different IC values, b) shows independence of in-
depth spread on IC at low MCs, but changes at high MCs. ..................................................... 66
Figure 33 - Linear relationship between burning rate and MC for two values of IC. Burning
rate is a measure of the rate of mass consumed per unit horizontal area, and the critical MC is
the threshold beyond which smouldering cannot be sustained. ............................................... 67
Figure 34 - Plotting a) global spread rate, which is the combination of horizontal and in-depth
spread, against MC, indicating its independence; and b) the linear change of direction of spread
with MC. Error bars show the range in experimental data. Black line are results calculated from
literature values of horizontal [101] and in-depth [59] spread rates. ....................................... 68
Figure 35 - Global spread and direction of spread affected by changing inorganic content in
dry samples. Error bars show the range in experimental data. ................................................ 69
Figure 36 – a) and b) Illustrate how once in-depth spread exceeds the horizontal spread rate
smouldering begins to extinguish locally and exhibit a bifurcation behaviour. c) An IR image
illustrating bifurcation, with regions of spread and extinction as well as illustrating the use of
transects in detecting extinction and measurement of the extinction perimeter. ..................... 70
Figure 37 -IR image series illustrating the influence of density on smouldering spread in peat.
Each image is taken 140 min after ignition. The crosses indicate location of ignition. .......... 75
Figure 38 - Influence of density on the smouldering width. A slight linear increase is noticeable
with a slope of 0.0046 cm·m3/kg1 (R2 = 0.92). ........................................................................ 75
Figure 39 – Influence of density on both horizontal and in-depth spread rates. Both spread rates
decrease with density. .............................................................................................................. 76
Figure 40 – Change of global spread rate and associated direction () with density. Global
spread decreases linearly with density with a slope of -0.039 cm·m3/ kg·h (R2= 0.94). The
direction of spread is decreased by 9% for 0 to 40% increase in density.. .............................. 77

135
Figure 41 – Calculated changes in burning rate as a function density are independent, with a P
value of 0.1. Burning rate is calculated by Eq.5 . Average burning rate is 4.66 ± 0.16 kg/m2·h.
.................................................................................................................................................. 77
Figure 42 – a) Showing the common relationship between in-depth spread and organic density
in samples with varying soil conditions. b) Demonstrates the linearity of the relation between
in-depth spread and 1/organic density. The data is compared together with 3 other experimental
series using the shallow reactor: right triangle (MCx) – varying MC in samples with IC=2.5%;
up triangle (mixed) – varying MC in samples with IC=40%; and left triangle (ICx) – varying
IC with samples of low MC (<5%), as well as an experimental series conducted by Huang and
Rein,2017 [59] varying MC and density in deep peat layers. The equation that describes the
shallow reactor data is 𝑆𝑑 = 672.4𝜌𝑜 − 1.892. ..................................................................... 79
Figure 43 - Illustrating the poor correlation between organic density and horizontal spread rate.
Comparing samples with varying experimental set up and soil conditions from multiple sources
such as Hu et al.,2019[57] , Prat et al.,2016[56] and Cowan et al.,2020[84]. Error bars represent
the range of experimental data. ................................................................................................ 80
Figure 44 – a) presents the relationship between horizontal spread and heat sink density of the
soil. b) demonstrating the linearity between horizontal spread rate and the inverse of heat
capacity. Results are compared with results from Chapter 4, Hu et al.,2019 [57], Prat et al.,2016
[56], and Cowan et al.,2020 [84]. Varying across a broad range of MC, inorganic content and
density. The equation of the of the linear fit for the shallow layer data is 𝑆ℎ = 3.276 ×
109∆𝐻𝑠 − 2.825. .................................................................................................................... 82
Figure 45 - Illustrating the poor correlation between in-depth spread rate and heat sink density.
Results are compared across a range of soil conditions and experimental set up(Huang & Rein,
2017 [59]) . Error bars represent the range of experimental data. ........................................... 82
Figure 46 –Polar plot of the influence of wind direction and MC on a) horizontal spread rate,
b) in-depth spread, c) burning rate, and d) global spread rate. Radial axis is spread rate (cm/h)
while the angular axis is the direction θw. Where θw= 0, 90, and 180° is forward perpendicular
and opposed respectively. ........................................................................................................ 89
Figure 47 - Influence of MC and wind direction on a) horizontal spread rate, b) in-depth spread
rate, c) burning rate, and d) global spread rate. The results are offset from MC values and each
moisture condition is separated by dotted lines. The error bars show the range of experimental
results from three repetitions. The results for no wind conditions were taken from Chapter 4.
.................................................................................................................................................. 90

136
Figure 48 – a) Showing the variation of normalized heat flux with wind direction.
Demonstrating the independence of heat flux from forward spread and decreasing heat flux
from opposed spread with increasing MC. The results are average over 3 repeats for each MC
conditio. b) demonstrates the influence of MC on peak radiant flux in quiescent conditions
under no-wind conditions. The data for the right figure is of the MCx series presented in
Chapter 4. All clouds show the range of results measured. ..................................................... 92
Figure 49 - Polar plot of the influence of slope on a) horizontal spread rate, b) in-depth spread
rate, c) burning rate, and d) global spread rate. Radial axis is spread rate (cm/h) and the angular
axis is the direction θs. Where 0, 90, and 180° is uphill (forward), sidehill (perpendicular) and
downhill (opposed), respectively. Spread rates affected by the slippage of peat were removed.
Each slope is averaged between 3 repeats, and all samples here are condition to 80% MC. .. 94
Figure 50 – The influence of slope on smouldering dynamics on a) horizontal spread, b) in-
depth spread, c) burning rate, d) global spread rate, due to spread relative to the slope. Cases
affected by the slippage of peat were removed. Values are averaged across 3 repeats while error
bars provide the maximum and minimum spread of the repetitions. (Spread rate values are
offset from their respective slopes and each slope conditions is separated by dotted lines.) .. 95
Figure 51 - Plotting averaged a) horizontal spread rate, b) in-depth spread rate, c) burning rate,
and d) global spread rate as a function of β, which is the angle at which smouldering spreads
relative to the horizontal plane. Spread affected by the slippage of peat were removed. The
cloud error bars show the range in spread from the repetitions. .............................................. 97
Figure 52 - Mass loss rate influenced by varying extraction conditions. The cloud demonstrate
the range of repeat results. ....................................................................................................... 59
Figure 53 - Showing the influence of condition temperature on mass loss rate and 4 stages of
spread. I is the ignition stage, II is the growth stage, III is the spread stage, while IV is the decay
phase. ..................................................................................................................................... 102
Figure 54 - Comparison between temperature developments of thermocouples at various
horizontal and in-depth positions in samples of 50% MC. Dotted lines represent the
approximate transition between 4 stages: I is the ignition stage, II is growth stage, III is stead
stage, IV is the decay stage. ................................................................................................... 103
Figure 55 - Temperature residence vs time in samples of 50% MC at the two nearest
thermocouple to the ignition coil during stage I. a) 4 cm below the source, and b) 6 cm below
the surface, both thermocouples are approximately 2 cm horizontally from the ignition ..... 104

137
Figure 56 - Showing the effect of condition temperature on peak temperatures measured during
the stage I (ignition period) as measured by thermocouples nearest the ignition coil in samples
with 50 % moisture content. Uncertainties are taken from range of repeated experiments. 105
Figure 57 - a) The effect of ambient temperature on peak temperature, comparing samples of
50% and 100% MC and at different depths (shallow, 2 cm; deep, 8cm.) b) Presenting the profile
of peak temperature by depth and comparing both the effect of MC and ambient temperature.
Dotted line represents threshold of pyrolysis and used as an indicator of burning depth. .... 106
Figure 58 –thermal residence times showing at which thermocouples remained above a given
temperature and the influence of ambient temperature at a) shallow layers (2 cm) and b) deep
layers (8 cm). The thermocouples used are located 7 cm from the ignition side. The thermal
residence time from Hu et al., 2019 [57] was provided for comparison which was conducted in
a reactor of the same dimension but with an insulated base. ................................................. 109
Figure 59 - a) illustrating the effect of condition temperature on spread rate, measured using
infrared imaging, in sample of 50% and 100% moisture content. Results are also compared
against IR spread measurements of a similar reactor with an insulated base at room temperature
presented by Amin et al.,2020 [85]. IR data was not available for samples with 100% moisture
at 2C. b) showing the effect of depth on spread rate on both samples with 50% and 100%
moisture content at various ambient temperature and compared to spread rates presented by
Huang et al.,2016 [70]. Error bars are based on uncertainties in measurement. Temperature at
2 cm depth in samples of 100% moisture never exceeded 100C and so could not be measured.
................................................................................................................................................ 111
Figure 60 – Combination of all experimental series presented in this thesis. Comparing a)
horizontal spread rates as a function of the heat sink density and b) in-depth spread as a function
of organic density. The grey shadow give the range of results of the soil condition series (MCx,
mixed, ICx, and density) ........................................................................................................ 117

138
LIST OF TABLES
Table 1 - Summary of average temperatures for the three key temperature conditions: -7°C,
2°C, and 21°C. The uncertainties reported are the standard deviation of measured temperature
across all experiments at the given condition temperature. ..................................................... 55
Table 2 - List of values used for calculating heat sink density. Thermal properties used are
found in [65] ............................................................................................................................ 80
Table 3 – Summarizing the effect of wind direction on smouldering dynamics. Percentage
differences are noted relative to the no-wind condition for 0, 40, 80 and 120% MC respectively.
Arrows upwards means an increase, arrow downward indicates a decrease, while ~ means a
negligible change. .................................................................................................................... 92
Table 4 - Summarizing the effect of slope relative to spread on smouldering dynamics. Arrows
indicates the relative change in behaviour to no-slope conditions. Percentages provide the
average change from 0° to 10, 20, and 30° respectively. Where x is given, no value was found
due to slippage of peat. ............................................................................................................ 96
Table 5 – Summarizing average spread rates at varying extraction rates and comparing to that
in open combustion .................................................................................................................. 59
Table 6 - Estimated burning depth by using 200 C as an indicator of char formation and linearly
interpolating between peak temperatures measured by thermocouples. Depth of burning
measured by Huang et al.,2016 [70], is also presented for comparison. ............................... 107

139
Appendix 1:
Appendix 1 includes additional information about Chapter 3: Experimental Methodology.
Detailing the sample conditions for all experiments performed in the shallow reactor.
Additionally, figures providing additional detail on the characterisation of the air flow applied
to the shallow reactor when studying the influence of wind.

140
Table A1.1 - summary of the average soil conditions for experiments performed using the shallow reactor. The condition code describes the target MC and IC, for example I2.5M80 indicates the
sample has 2.5 % IC and 80% MC. Each condition was performed 3 times.

Condition I2.5M0 I2.5M20 I2.5M40 I2.5M60 I2.5M80 I2.5M100 I2.5M110 I2.5M120 I2.5M140 I2.5M160
3
ρb (kg/m ) 162 186 170 165 158 168 182 196 181 191
MCx
3
ρo (kg/m ) 154 151 120 101 86 81 85 87 74 73
MC (%) 3 20 38 60 79 101 109 120 138 154
Condition I40M0 I40M20 I40M40 I40M60 I40M70 I40M80
ρb (kg/m3) 312 245 211 225 245 225
Mixed
3
ρo (kg/m ) 184 120 90 84 86 75
MC (%) 2 22 41 60 70 80
Condition I20M0 I40M0 I60M0 I75M0 I80M0 I90M0
ρb (kg/m3) 245 312 457 610 679 884
ICx
3
ρo (kg/m ) 192 184 180 151 135 88
MC (%) 2 0 1 1 1 0
Condition 10%Δ 20%Δ 30%Δ 40%Δ
Density
ρb (kg/m3) 176 191 203 222
(2.5% IC,
3
80% MC) ρo (kg/m ) 97 103 112 121
MC (%) 77 81 77 79
Condition I2.5M0 I2.5M40 I2.5M80 I2.5M120
3
Wind ρb (kg/m ) 156 162 160 177
(0.86 m/s) ρo (kg/m3) 149 112 87 78
MC (%) 2 40 79 121
Condition α=10° α=20° α=30°
Slope 3
ρb (kg/m ) 164 172 170
(2.5% IC,
80% MC) ρo (kg/m3) 89 96 92
MC (%) 80 76 82

141
Figure A1.1 – Characterization of airflow profile. a) wind speed with height as various distances from the fan, down the
centreline, b) wind speed with height as various distances from the fan, 7.5 cm from centreline.

Figure A1.2 – Characterizing the wind speed effecting the smouldering spread. The illustration on the left is a top view of the
reactor, where the intersection of the dotted lines indicate the location of wind speed measurements. The circled intersections
are the locations that are averaged to yield the average wind speed affecting the spread. The study of smouldering spread was
limited to within 14 cm of the ignition, indicated by the green circle. The figure on the right shows the wind speed 1 cm above
the ground at various distances from the fan and at distanced from the centreline (CL) of the fan.

142
Appendix 2:

Appendix 3 includes additional figures about Chapter 6: Influence of Environmental Condition

on Smouldering Dynamics: A Study of Wind and Slope

A2.1 The Effect of Wind on Smouldering Dynamics

Figure A2.1 – Showing the change in direction of global spread with wind direction and comparing to no wind conditions
taken from chapter 4. The results are offset from MC values and each moisture condition is separated by dotted lines. The
error bars show the range of experimental results from three repetitions.

143
A2.2 The Effect of Slope on Smouldering Dynamics

Figure A2.2 – Showing the influence of spread direction relative to the slope on the global spread direction φ. Values are
averaged across 3 repeats while error bars provide the maximum and minimum spread of the repetitions. (Spread rate values
are offset from their respective slopes and each slope condition is separated by dotted lines.) Black horizontal line is the
average direction of spread in samples with no slope.

Figure A2.3 – Shows how the smouldering dynamics; a) in-depth spread b) burning rate c) global spread rate and d) direction
of spread can be considered by the spread angle β.

144