You are on page 1of 13

Forestry An International Journal of Forest Research

Forestry 2019; 00, 1–13, doi:10.1093/forestry/cpz057

A comparison of UAV laser scanning, photogrammetry and airborne


laser scanning for precision inventory of small-forest properties

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


Stefano Puliti1, *, Jonathan P. Dash2 , Michael S. Watt3 , Johannes Breidenbach1 and Grant D. Pearse2

1
Norwegian Institute for Bioeconomy Research (NIBIO), Division of Forest and Forest Resources. Høgskoleveien 8, 1433 Ås, Norway
2
Scion, Department of Forest Informatics. Sala Street, Private Bag 3020, Rotorua 3046, New Zealand
3
Scion, Department of Forest Informatics, 10 Kyle Street, Riccarton, Christchurch 8011, New Zealand

*Corresponding author. Tel: +47 936 78 891; E-mail: stefano.puliti@nibio.no

Received 0 Month 2019

This study addresses the use of multiple sources of auxiliary data from unmanned aerial vehicles (UAVs) and air-
borne laser scanning (ALS) data for inference on key biophysical parameters in small forest properties (5–300 ha).
We compared the precision of the estimates using plot data alone under a design-based inference with model-
based estimates that include plot data and the following four types of auxiliary data: (1) terrain-independent
variables from UAV photogrammetric data (UAV-SfM); (2) variables obtained from UAV photogrammetric data
normalized using external terrain data (UAV-SfMDTM ); (3) UAV-LS and (4) ALS data. The inclusion of remotely
sensed data increased the precision of DB estimates by factors of 1.5–2.2. The optimal data sources for top
height, stem density, basal area and total stem volume were: UAV-LS, UAV-SfM, UAV-SfMDTM and UAV-SfMDTM .
We conclude that the use of UAV data can increase the precision of stand-level estimates even under intensive
field sampling conditions.

Introduction order of a few centimetres and LS data can be collected at point


densities of between 60-1500 points per m2 (Puliti et al., 2015).
The use of unmanned aerial vehicles (UAV) or systems (UAS) However, due to range limitations, civilian UAVs are currently
within the civilian market has increased dramatically over the constrained for use over areas of up to 10 km2 when complete
last five years. Within the realm of environmental monitoring, coverage is required (Dandois and Ellis, 2013; Whitehead et al.,
UAVs have been used to collect data from laser scanning (LS) sen- 2014) using light-weight sensors and considerably smaller areas
sors, RGB imagery or even multispectral and hyperspectral sen- when deploying heavier and more energy consumptive LS sys-
sors Salamí et al. (2014). Although their use within commercial tems. Above this threshold, alternative platforms (e.g. manned
plantation forestry remains limited so far, there is considerable aircraft) become more cost-effective at obtaining complete
potential for use of UAVs within this sector. A number of potential coverage (Heaphy et al., 2017). This restriction may limit the use
applications have already emerged that include forest inventory of UAVs for major data acquisitions across large forests. However,
(Puliti et al., 2017; Puliti et al., 2015; Wallace et al., 2016), post- UAVs could play an important role in collection of data to support
harvest site assessment (Goodbody et al., 2017; Pierzchała et al., precision forestry practices in small forest properties, or within
2014; Puliti et al., 2018a; Talbot et al., 2017), characterization plantation stands of a given age for a particular purpose (e.g.
of weed competition (Watt et al., 2017) and detection of tree inventory, road planning) where an absence of high-quality data
disease (Dash et al., 2017). can be economically and environmentally costly. Recent research
Rapid growth in the use of UAVs is predicted to continue also shows that using partial UAV coverage in larger plantations
(Hugenholtz et al., 2012; Nex and Remondino, 2014; Watts could be a cost-effective option for forest inventory (Puliti et al.,
et al., 2012) as they offer greater operational flexibility compared 2018b). The use of UAVs for forest inventory is one of the most
to other platforms (e.g. manned airborne) and can be used for promising applications of UAV technology.
data collection under a wider variety of atmospheric conditions Within forest inventory applications three-dimensional point
(Whitehead and Hugenholtz, 2014). Among available aerial clouds are among the most useful form of remotely sensed data
platforms, UAVs can provide very fine resolution and data density. as these data have been reliably shown to accurately predict
Imagery collected from a UAV typically has a resolution in the key forest biophysical variables such as height and volume in

© Institute of Chartered Foresters, 2019. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.

1
Forestry

commercial forest settings (Dash et al., 2015; Næsset, 2002; sample is available to fit a model, as in our case study, some of
Watt and Watt, 2013). Point clouds can be generated either the considered MB variance estimators can also be interpreted as
through LS (UAV-LS) (Jaakkola et al., 2010; Wallace et al., 2012) being DB model-assisted (Mandallaz, 2013).
or through the use of imagery captured from digital still cameras The comparison of different auxiliary data types from UAV
(Dandois and Ellis, 2013; Lisein et al., 2013). LS systems emit is clearly important for determining the contribution towards

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


laser pulses and record the time-of-flight for reflected energy improvement of the precision of the inventory estimates and
to be returned to an onboard sensor. Combined with positional how this information can be most economically efficiently inte-
information on the sensor’s location, a three-dimensional point grated into forest inventory. However, in regard to UAVs very
cloud representation of the scanned area can be produced. Point little research has compared the precision of DB estimates, using
clouds can also be obtained using a combination of computer only plot data, to that of MB estimates. Within a 7330 ha for-
vision (Structure from Motion; SfM) and photogrammetry tech- est area Puliti et al. (2017) compared the precision of hybrid
niques. Unlike LS, SfM methods rely on passive optical sensors estimation using partial-coverage UAV data and field plots to
that do not penetrate the canopy (Lowe, 2004). Several authors two other methods. These two methods included DB inference
have reported the predictive accuracy of models linking either using only field plots and a MB approach that utilized wall-
UAV-SfM or UAV-LS data with ground reference forest biophysical to-wall airborne LS data collected from manned aircraft (air-
variables (Dandois and Ellis, 2013; Dandois et al., 2015; Lisein et borne laser scanning [ALS]). Compared to the use of only field
al., 2013; Puliti et al., 2015; Wallace et al., 2016). Recent research plots, the hybrid estimation using UAV data was 4.4 times more
also demonstrates the utility of explanatory variables extracted efficient. The largest precision was obtained through complete
from photogrammetry that do not require a digital terrain model coverage of ALS data that had a relative efficiency 1.6 times
(DTM), i.e. DTM-independent variables, for prediction of forest greater than that obtained through use of the partial-coverage
volume (Giannetti et al., 2018), which is an approach that could UAV data. However, we are unaware of any research that has
be useful in densely forested areas or other situations where it is undertaken a similar comparison within small-moderate sized
difficult to obtain a DTM. Despite the diverse range of UAV-based forests, with a large degree of stand uniformity, where com-
approaches that are available for forest inventory, there is still a plete coverage is available from a UAV. Furthermore, to date, no
lack of understanding of how the uncertainty of these estimates attempt has been made to compare the precision obtainable
compares among these methods and estimates obtained from either from UAV-SfM or UAV-LS data under rigorous estimation
only plot data. frameworks.
The increasing availability of remotely sensed data has been The research was undertaken in a Pinus radiata D. Don
matched by the development of inferential frameworks that plantation located in the central North Island, New Zealand. In
enable a rigorous estimation of forest biophysical parameters New Zealand, ∼30 per cent of the national plantation estate
(e.g. growing stock volume or above ground biomass). Tradi- is contained within fragmented small or medium sized forests
tional plot-based inventories adopt design-based (DB) inference, (5–300 ha). These forests represent a challenging sampling
whereby a probability sample is selected from the population problem as they can be quite variable, but growers are often
according to an appropriate sampling design. For DB estimators reluctant to invest in an adequate field plot sample size to
the uncertainty is usually expressed through the variance, or the properly characterize them due to their relatively small value.
standard error, which is a fixed quantity given the population, In this study, we sought to identify and compare suitable
sampling design and the estimator. The integration of point cloud datasets and inferential frameworks for incorporating UAV data
data collected from UAVs with other sources of inventory infor- for inventory of small forests. Remotely sensed data used in
mation raises challenges for appropriate statistical inference. The the analyses were extracted from the following four sources:
inferential framework adopted is governed, among others, by the (1) DTM-independent variables extracted from photogrammetry
kind of sample (probability vs non-probability) and the sample (UAV-SfM); (2) variables obtained from the UAV photogrammetric
size. For small forests, where complete UAV data coverage is pos- point cloud normalized using an accurate DTM from an ALS
sible, model-assisted and model-based (MB) variance estimators survey (UAV-SfMDTM ); (3) UAV point cloud data extracted from
can be applied. While the first requires a probability sample within LS (UAV-LS) and (iv) ALS point cloud data (see Table 1 for
the area of interest (i.e. internal data), the second allows for the summary).
use of external field plot data. Thanks to the possibility of using The objective of this study was to assess (1) the contribution
external data, MB inference is a more appealing inferential frame- of UAV and ALS data to the precision of the inventory estimates
work for UAV-based forest inventories since UAV acquisitions are and (2) if the precision increases, which type of UAV auxiliary
generally performed on small areas and are associated with the data leads to the most precise estimates. Such questions were
measurement of only few plots, if any. Thus, to provide practical addressed by comparing the estimated population parameters
insights in the use of UAVs in forest inventory, in this study, we (means) and their uncertainty for four key forest biophysical
assessed only the MB framework as it does not require an internal parameters (top height, stand density, basal area and total stem
probability sample of field plots. The uncertainty of the estimates volume) using the following two inventory approaches (1) DB
obtained from this form of inference is based on the uncertainty inference using plot data and (2) MB inference using plot data
of the model parameter estimates and the residual variance and wall-to-wall coverage of UAV-SfM, UAV-SfMDTM , UAV-LS or
(Gregoire, 1998; McRoberts, 2006; McRoberts et al., 2013). It may, ALS data. Furthermore, analytical and bootstrap estimators were
however, be interesting to note that, if an internal probability used for variance estimation.

2
Precision inventory of small-forest properties

Table 1 Description of datasets used to provide explanatory variables to model forest inventory attributes

Dataset Dataset description

UAV-SfM DTM-independent variables extracted from UAV photogrammetry

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


UAV-SfMDTM Variables obtained from UAV imagery and the height-normalized SfM point cloud using a DTM acquired from a
pre-existing conventional ALS survey
UAV-LS Variables describing point cloud data extracted from UAV LS
ALS Variables describing point cloud data extracted from conventional airborne LS collected from a manned aircraft

Materials Remotely sensed data


UAV LS data
Study area
LS data were acquired using a Velodyne HDL-32E (Velodyne, San
The study area, which was located close to Rotorua, New Zealand
Jose, USA) scanner embedded in a LidarUSA ‘Snoopy A-Series’
(38.235 S, 176.156 E, 380 m a.s.l.) was a 40 ha Pinus radiata
LS system (LidarUSA, Alabama, USA). The laser scanner was
plantation established in 1993. The stand was pruned to 6 m
mounted on an Altus ORC4 remotely piloted helicopter platform
above ground and thinned from an initial density of ∼600 stems
(Altus Ltd., Hamilton, New Zealand). A flying altitude of around
per hectare (sph) to around 350 sph in a single thin-to-waste
80 m above the local terrain was used and the flight plan ensured
silvicultural intervention at age 9.
that there was considerable side overlap to remove the possibility
of data voids. All flight manoeuvres and altitude adjustments
were made outside of the area of interest to avoid the possibility
of flight artefacts in the dataset. The laser scanner is only capable
Field data
of recording a single return but produced a pulse density of 712
A total of 30 bounded circular 0.06 ha slope-adjusted field plots pulses/m2 and a pulse spacing of 0.04 m. The LS of the area
were measured to provide data for this study resulting in a required a total of 78 min, which included three flights lasting
sampling intensity of 4.5 per cent. The locations of these plots 35, 20 and 23 min.
were selected according to a systematic grid with a spacing of The Velodyne HDL-32E laser scanner has a rotating array of
100 m × 100 m (Figure 1). 32 lasers, providing a maximum potential ‘scan angle’ of 180
Plot measurements, which were taken during June 2017, degrees from the Snoopy system. Only the inner laser beams
included diameter at breast height (DBH) measured at 1.4 m for are equiangular and returns acquired with a large off-nadir angle
all trees within the plot. Height was measured on a sub-sample of induce substantial artefacts and noise in the point cloud data
eight trees, covering the DBH range within the plot and being free when acquired from a UAV. To minimize these effects, only data
from excessive lean or malformation. These measurements were from the inner-most lasers (±8 degrees off nadir) were retained
used to fit a Petterson regression (Petterson, 1955) using ordinary and a sector reduction was applied to limit the effective scan
least square (OLS) between DBH and tree height for each plot that angle to +/−25 degrees across track.
was subsequently used to estimate tree height for all plot trees Flight line matching, ground classification, noise removal and
where tree height was not measured. Top height (H) was then the identification of overlap points were carried out in the Ter-
estimated as the average height of the six largest trees per plot, rasolid software (Terrasolid Oy, Espoo, Finland). Subsequently,
corresponding to the 100 largest trees per hectare, where largest points classified as overlap were removed to ensure a more even
was defined in terms of DBH. Plot basal area (G) and stand density density over the entire study area. A transect of the UAV-LS and
(N) were estimated from the plot measurements. A compatible UAV-SfMDTM point cloud is shown in Figure 2.
volume and taper model (Goulding and Murray, 1976) with A total of 28 ALS explanatory variables were calculated from
coefficients estimated from a nationwide dataset, was used to the UAV-LS point cloud data coincident with the field plots. The
predict stem volume and upper stem diameters based on the tree variables included height percentile (p10 , p20 , . . ., p100 ), density
measurements recorded. These data were processed using an variables (d10 , d20 , . . ., d100 ) and the mean of eight textural
optimal log-bucking algorithm embedded in a commercial forest variables (mean, variance, homogeneity, contrast, dissimilarity,
yield prediction software package (YTGEN, Silmetra, Tokoroa, New entropy, second moment and correlation) calculated based on
Zealand). A set of log-grade specifications and prices provided
the canopy height model using the R package ‘glcm’ (Haralick
by the forest manager were used to segregate trees into log-
et al., 1973; Zvoleff, 2016). The study area was then tessellated
products using the log-bucking algorithm designed to maximize using a grid of 600 m2 cells, which matched the plot area,
economic value. The total stem volume (TSV) was predicted
for which the same variables as used for the plots were then
for each plot by summing the total volume assigned to both
computed from the UAV-LS data.
merchantable and non-merchantable log-products.
Summary statistics for all of the plots included in the study
(Table 2) show relatively small variation in top height with a mean ALS data
of 36.3 m and range of 33.5– 38.6 m. Mean values for N, G and The ALS data for this study were collected in December 2017
TSV ranged almost twofold across the plots (Table 2). using a Trimble AX60i scanner with a Trimble AP50 inertial motion

3
Forestry

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


Figure 1 Overview of the sampling design for the acquisition of field data.

Table 2 Summary statistics for the ground reference data for top height (H), number of trees per hectare (N), basal area (G) and total stem volume
(TSV)

Variable Mean Standard deviation Range

Top height (m) 36.3 1.4 33.5– 38.6


Stand density (stems ha−1 ) 366 54.4 233– 450
Basal area (m2 ha−1 ) 65.7 7.4 45.9– 85.0
TSV (m3 ha−1 ) 781 92.08 519– 1007

system and GNSS receiver. Data capture was completed using UAV photogrammetric data
a manned aircraft system. Campaign settings included a flying Full coverage imagery was acquired using a DJI ZenMuse X3 12
height of 700 m above ground level, pulse rate of 360 kHz and a MP camera (DJI, Shenzhen, China) mounted on a DJI Matrice 600
swath side overlap of 60 per cent. The scanning system operated
UAV. Dense overlapping images were acquired for the generation
in multiple return mode and had an effective laser footprint
of point cloud data using Agisoft PhotoScan (Agisoft, 2017). Flight
size of 0.35 m. The resulting dataset had a pulse density of
altitude was set at 300 m above ground level and image overlap
10.2 pulses/m2 and an on-ground point spacing of 0.31 m. Initial
to 95–90 per cent front and side overlap. Imagery was collected
processing was carried out by the supplier using Terrasolid and
under clear sky conditions within a single flight lasting 20 min.
included ground and noise classification. Subsequent process-
Five ground control points were established on the ground in open
ing was carried out in the LAStools software package (Rapid-
areas and/or canopy gaps and located using a Trimble Geo 7X
Lasso GMBH, Gilching, Germany) and included triangulation of
a terrain model using the ground classified points, further noise differential GNSS system.
removal and normalization of elevations of non-ground returns Photogrammetric processing of the UAV imagery resulted in
to heights above the local ground level. The same variables a dense point cloud (approximately 60 points/m2 ) and an ortho-
that were extracted for the UAV-LS data were extracted for the mosaic with a ground sampling distance of 10 cm. A total of 31
ALS data, including textural variables from the canopy height point cloud explanatory variables were extracted including height
model. percentiles.

4
Precision inventory of small-forest properties

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


Figure 2 Detail of an area including a field plot to highlight the differences between the horizontal (a–c) and vertical (d) representations of the canopy
using: a) the UAV orthomosaic (shown for visual reference only); (b) the UAV-SfM canopy height model; (c) the UAV-LS canopy height model; and (d)
the height normalized UAV-SfMDTM and UAV-LS data.

Two sets of variables were calculated from the UAV pho- data and wall-to-wall coverage of UAV-SfM, UAV-SfMDTM , UAV-LS
togrammetric data, namely the UAV-SfM (i.e. DTM-independent) or ALS data. The DB estimate (i.e. using only plot data) was then
and the UAV-SfMDTM variables (from the normalized point used to assess the efficiency of using additional remotely sensed
cloud). The former included a total of 163 explanatory variables data. Within this study, we considered an efficient estimator
extracted from the raw point cloud without any operations to one that had a smaller variance estimate compared to that of
alter the height values (i.e. normalization). The use of these purely field-based estimates, while not showing obvious signs of
variables mainly aims at describing the variations in height and bias. Precision and accuracy for the four biophysical parameters
spectral response of the top of the canopy. Among the most were compared among the possible combinations of data source
relevant DTM-independent variables indicated by Giannetti et al. (UAV-SfM, UAV-SfMDTM , UAV-LS and ALS). The bias of the estima-
(2018) were textural variables calculated based on the digital tors of the mean was assessed by: (1) scrutinizing the model fit
surface model (i.e. top of the canopy), local maxima counts and graphically and by the mean of differences between observed
spectral variables. The UAV-SfMDTM variables included a total of 31 and predicted values; (2) comparing the first two moments of the
variables including height percentiles (p10 , p20 , . . ., p100 ), density distributions (for the sample and for the whole population) of the
variables (d10 , d20 , . . ., d100 ), the mean and standard deviation selected explanatory variables and (3) checking whether the MB
of textural variables calculated on the canopy height model and estimates were within DB confidence intervals. Furthermore, the
the mean of the red, green and blue digital number values (R, G uncertainty of the variance was assessed by comparing analyti-
and B). cal and bootstrap variance estimators.

Methods
Inference
Overview This section describes the methodology used to estimate the
In this study we estimated four forest biophysical parameters  μ̂)) for each biophysical variable
mean (μ̂) and variance (Var(
(i.e. H, N, G and TSV) according to MB inference using field plot using both DB and MB inference. In the latter case the variance

5
Forestry

was expressed as the mean squared error (MSE) for a population element is equal to 1/N and X U is a N × (r + 1) matrix of
U composed of j = {1, . . . , N} sampling units. explanatory variables from the remotely sensed auxiliary data.
For the DB estimate we assumed the field data (S) to represent The mean square error (MSE) estimator is:
a simple random sample of size n and a direct estimator of the
mean is:  
 p μ̂MB = ιU T X U Cov(
 β̂ S )X U T ιU

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


MSE ((5))

1 
n
μ̂DB = yi ((1)) where the subscript p indicates that this estimator incorporates
n
i=1
only the uncertainty of the estimates of the regression parame-
ters. In our case study, where the models are fitted on internal
where yi is equal to the ground reference value of the biophysical data (field plots from within the area of interest) based on a
variable of interest for the ith field plot (i = 1, . . ., n). The DB probability sample, this estimator can also be interpreted as
 (μ̂DB )) was estimated assuming simple
variance of the mean (Var being model-assisted (design-based) and suitable for internal
random sampling by the variance estimator: models (Mandallaz,
 2013). The covariance matrix of the model
parameters Cov β̂ S is estimated according to
  2
 μ̂DB = sY
Var ((2))
n
  ε̂TS ε̂S  T 
 β̂ S =
Cov XS XS +  ((6))
s2Y
where is the sample variance. n−r−1
For the MB inference, we fitted separate multiple linear
regression models for the different combinations of ground where ε̂S is a n-length vector of estimated residuals over the
reference biophysical parameters of interest and explanatory sample S and  is a matrix that describes the variance and
variables extracted from the remotely sensed data (UAV- correlation of the residuals (Pinheiro and Bates, 2006, p. 203). The
SfM, UAV-SfMDTM , UAV-LS, or ALS data). Prior to using these gls function of the R-package (R Core Team, 2017) nlme (Pinheiro
models for the estimation of population parameters, their et al., 2019) was used to estimate the model parameters. An esti-
validity was evaluated according to a visual assessment of the mator of the mean squared error (MSE  (μ̂MB )) that incorporates
scatterplots of observed vs predicted values. This assessment the uncertainty due to presence of heteroscedasticity and auto-
aimed at identifying potential signs of heteroscedasticity and correlation has been previously adopted by Breidenbach et al.
non-linearity. Further evaluation of the model included the (2016), McRoberts (2006) and McRoberts et al. (2018). Based on
assessment of the model fit (R2 Adj ), root mean square error the formulation by Breidenbach et al. (2016) such an estimator
(RMSE), mean difference (MD) and RMSE and MD values as can be expressed as:
percentage of the observed mean (RMSE% and MD% ). The RMSE
and MD were determined by leave-one-out cross validation as      
the square root of the mean of the squared residuals and the  prh μ̂MB
MSE =  p μ̂MB + ιU T Cov
MSE  σ̂ 2 ιU
ε
mean of the residuals, respectively.
  
N 
N
The linear models had the form: =  p μ̂MB + 1
MSE σ̂j σ̂k ˆ jk ((7))
N2
  j=1 k=1
yS = X S β S + εS , εS ∼ N 0, σε2 W ((3))
 
 σ̂ 2 is the N × N covariance matrix of estimated grid
where Cov
where X S is a n × (r + 1) matrix, r is the number of selected ε

explanatory variables, β S is a column vector of model parameters cell residual variances with the elements σ̂j σ̂k ˆ jk and the sub-
of length r + 1, and ε S is a column vector of random errors with script prhs indicates that the estimator accounts for parameter
zero expectation, of length n. The residual variance is expressed uncertainty, residual heteroscedasticity and spatial correlation.
as the product of the mean squared residual (σε2 ) and a n × In equation (7), the first component is estimated according to
n weight matrix W (Breidenbach et al., 2016). W is an identity equation (5) and incorporates the uncertainty due to model
matrix under homoscedasticity, a diagonal matrix with weights parameter estimation while the second component incorporates
based on a variance model under heteroscedasticity, or a matrix the uncertainty due to residual error. Specifically, σ̂j and σ̂k repre-
where the off-diagonal elements contain weights that result sent the residual variation estimated for the grid cells j = k = {1,
from a model describing the correlation pattern in the residuals . . ., N} while ˆ jk is the estimated correlation between two grid cell
in the case of autocorrelation. predictions j and k due to autocorrelation. See Breidenbach et al.
The population mean of a small forest property is estimated (2016) for more information on the estimation of ˆ jk . For each of
as: the possible combinations of variables of interest and auxiliary
data we fitted generalized least squares (GLS) models with and
without the correlation structure and with and without weights to
μ̂MB = ιU T X U β̂ S ((4))
accommodate heteroscedasticity. An improvement in the model
over OLS regression was defined if the inclusion of autocorrelation
where β̂ S is a vector of estimated coefficients and, as defined and heteroscedasticity in the GLS models resulted in a decrease
by Saarela et al. (2016a) ιU is a vector of length N where each of two AIC points.

6
Precision inventory of small-forest properties

In those cases where homoscedasticity (σ̂j σ̂k = σ̂ε2 ) and null Results
autocorrelation (ˆ jk = 1) are assumed by the model, the estima-
tor in equation (7). reduces to: Models
The visual assessment of the observed against predicted values
using the fitted OLS models (Figure 3) revealed that there was no

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


    2
serious systematic lack of fit, which justified the adoption of linear
MSE  μ̂MB + σ̂ε
 pr μ̂MB = MSE ((8))
N models.
Top height (H) was accurately predicted from all three terrain-
normalized data sources (UAV-SfMDTM , UAV-LS and ALS) with
where the subscript pr indicates that the MSE estimator accounts adjusted-R2 ranging from 0.64 to 0.73 and RMSE ranging from
for parameter and residual uncertainty and σ̂ε is estimated from 0.76 to 0.91 m (Table 3). The upper percentiles (p90 , p95 and p100 )
the model residual variance. were the most important predictors within these models. The
UAV-SfM dataset that was not normalized using terrain infor-
mation had a smaller goodness-of-fit (R2 Adj = 0.17) and larger
Estimation of sampling statistics RMSE (1.33 m) compared to the other models using different
The standard error (SE) is given as the square root of the esti- source of auxiliary data. In respect to stand density (N), the UAV-
mated variance or MSE of the mean estimate and standard SfMDTM model was the one with largest R2 Adj (0.69), followed
error as a percentage of the mean (SE% ). The relative efficiency by UAV-LS (0.57) and UAV-SfM (0.55). Metrics utilizing imagery-
(RE) is given as the ratio between the estimated variance under related features, as well as some mid-canopy point cloud metrics,
the DB inference and the MSE of a MB estimator (i.e. RE = were important in these models. In contrast, the model based on
Var  (μ̂MB )). The RE is a factor that describes by how
 (μ̂DB )/MSE ALS data used a single height percentile and produced a much
much the number of field plots would need to be changed in weaker model (R2 Adj = 0.34). For basal area (G), the use of UAV
the DB inference to obtain the same level of precision as using photogrammetric data (UAV-SfM or UAV-SfMDTM ) yielded larger
another estimator. Because the RE is a ratio of variances, it is R2 Adj ranging from 0.49 to 0.63 compared to UAV-LS or ALS data
only meaningful for approximately unbiased estimators. Thus, (R2 Adj ranging from 0.21 to 0.33). For the three datasets acquired
for the specific case where a total of n = 30 plots were used from the UAV, density and textural variables were most useful for
in case A, a value of RE = 2 would correspond to the need for predicting G while the model fitted to ALS data used only d100 .
twice the number of plots (n = 60) to obtain the same level of As with basal area, the plot-level UAV-SfMDTM and UAV-SfM total
precision. stem volume (TSV) models had better model fit (R2 Adj = 0.65 and
In addition to the precision we assessed the potential pres- R2 Adj = 0.57) than UAV-LS (R2 Adj = 0.45) or ALS (R2 Adj = 0.37).
ence of bias in the estimators. An estimator was considered A mixture of density, percentile and textural variables were
unbiased if the following three different criteria were met: (1) selected as predictors of TSV for the three UAV-acquired datasets
MD close to zero as expected for linear regression models; (2) (Table 3). According to a two-sided t test, the MD was not
equality of the first two moments (i.e. mean and variance) from significantly different from zero (P-value always > 0.05) for any
the distribution of the explanatory variables in the sample and of the combinations of biophysical variables and auxiliary data
in the population and (3) the estimated mean needed to be type.
within the 95 per cent confidence interval of the DB estimate. The inclusion of information on heteroscedasticity and spa-
For assessing the uncertainty of estimated variances we also tial autocorrelation did not improve the models (based on AIC).
adopted a non-parametric bootstrap (Efron and Tibshirani, 1994; However, the quantification of spatial autocorrelation can often
equation (6.6), also referred to as ‘bootstrapping pairs’, McRoberts be difficult with forest inventory data. Therefore, even though
et al., 2018), with nboot = 1000 iterations as an alternative to the including spatial autocorrelation did not improve the models we
analytical estimators described above. According to this method still included the standard error estimates considering spatial
at each bth iteration we (1) selected a simple random sample autocorrelation using the estimated autocorrelation parameters.
with replacement of size equal to that of the original data; The comparison of the analytical estimates with those
(2) refitted the model to the resampled data; (3) applied the estimated using the bootstrapping procedure revealed ratios
SE
model to the grid cells covering the entire area and (4) esti- ( SE analytic ) > 1.1 or <0.9 for half of the tested combinations
mate the mean (μ̂b ) according to the MB estimator. Finally, bootstrap

we estimated the bootstrapping mean and variance according of biophysical parameters and auxiliary data (see Table 4). This
to: suggested a lack of bias in the estimator or the parameter
estimates for half of the studied combinations. For the remaining
half, the results indicated that for H and N the analytical
SE
1
nboot
 estimator over-estimated the MSE ( SE analytic ranging from 1.3 to
bootstrap
μ̂boot = μ̂b ((9))
nboot 1.6). On the contrary, for G and TSV there may have been some
b=1 SE
small underestimation of the MSE ( SE analytic = 0.8). This indicates
bootstrap
that the analytical MSE estimators are generally applicable
nboot
 with the utilized linear models. Interestingly, the indication of
1 2
 μ̂boot ) =
Var( μ̂b − μ̂boot ((10)) potential presence of bias in the analytical MSE estimators was
nboot − 1 always associated with UAV-SfMDTM data. Because of the small
b=1

7
Forestry

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


Figure 3 Scatterplots matrix of the observed vs predicted values using OLS models for all variables of interest (columns) and remotely sensed data
sources studied (rows). The line represents the 1:1 line.

differences between analytical and bootstrapping estimates, noticeable for the H-model using UAV-SfM data and theN-model
only the analytical estimates were used for further comparison. using UAV-LS data (see Figure 4).
The inclusion of RS data increased the precision of population For the same models as in Figure 4 (H-model using UAV-SfM
parameter estimates compared to the DB approach (i.e. field data andN-model using UAV-LS data), the mean estimate was
data alone) in 10 out of the 16 (62 per cent) combinations of either outside or close to the edge of the 95 per cent confidence
response variables and data sources (Table 4). Overall, for 14 intervals of the DB estimate (see Figure 5). Further potential bias
of the 16 combinations of biophysical parameters and auxiliary was noticeable for the MB estimators of G and TSV using UAV-
data, the estimated mean was within the 95 per cent confidence SfM data, for which the estimated means were smaller than
interval of the DB estimate. In contrast to the assessment of the lower bound of the 95 per cent confidence interval for the
the MD (Table 3), which suggested no presence of significant DB estimate (see Figure 5). No differences in the sample and
systematic errors in the model predictions, for some explanatory population distributions of explanatory variables were found for
variables the mean and variance varied substantially between these models.
the sample and the population distributions, thus suggesting a When averaged across the four RS data sources, the model
possible presence of bias of the estimator. This was particularly based estimates resulted in an average RE for H, N, G and

8
Precision inventory of small-forest properties

Table 3 Summary of the developed models. The models that resulted in the most precise model-based estimates for each of the biophysical
parameters are highlighted

Variable Auxiliary Explanatory variablesa R2Adj RMSE RMSE% MD MD%


of interest data

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


H UAV-SfM nintensity1Q 0.17 1.33 3.68 0.00 −0.01
UAV-SfMDTM p95 , crown 0.73 0.76 2.08 0.01 0.02
UAV-LS p95 , p100 0.72 0.76 2.11 0.001 0.006
ALS p90 , ntrees 0.64 0.91 2.50 −0.02 −0.05
N UAV-SfM SDdissimilarity , maxgreen 0.55 38.22 10.44 0.70 0.19
UAV-SfMDTM d90 , blue,SDdissimilarity 0.69 34.06 9.34 0.16 0.04
UAV-LS p30 ,d50 , homogeneity, 0.57 39.8 10.87 −0.11 −0.03
second moment
ALS p60 0.34 45.96 12.55 −0.36 −0.09
G UAV-SfM mean, homogeneity 0.49 5.54 8.42 −0.05 −0.07
UAV-SfMDTM d90 , d100 , dissimilarity, 0.63 4.97 7.55 0.15 0.23
SDmean , SDvariance
UAV-LS variance 0.21 7.01 10.66 −0.008 −0.008
ALS d100 0.33 6.36 9.67 −0.05 −0.08
TSV UAV-SfM mean, homogeneity 0.57 62.75 8.04 −0.54 −0.07
UAV-SfMDTM p95 , d100 , mean, 0.65 60.57 7.76 −0.55 −0.07
dissimilarity
UAV-LS d10 , d90 , mean 0.45 77.58 9.93 −2.02 −0.25
ALS p30 0.37 75.99 9.73 −0.55 −0.07

an
intensity1Q : number of points within the first quartile of intensity value (see Giannetti et al., 2018), ntrees: average number of local maxima,
crown: average crown size, maxgreen : maximum green value, blue: mean blue value, p10 -p100 : height percentiles, d10 -d100 : density variables, mean,
homogeneity, second moment, dissimilarity, variance: mean textural variables, SDdissimilarity : standard deviation textural variables.

TSV of 1.4, 1.2, 1.2 and 1.5, respectively (see Table 4). When plots (for a total of 47 plots). The largest REs in this study, which
excluding models leading to potentially biased estimators, the were achieved using of remotely sensed auxiliary data, ranged
largest RE for H, N, G and TSV were obtained using the UAV- from 2.1 for total stem volume using UAV-SfMDTM data to 2.2 for
LS (RE = 2.2), UAV-SfM (RE = 1.5), UAV-SfMDTM (RE = 1.6) top height using UAV-LS data. This indicates more than twice
and UAV-SfMDTM data (RE = 2.1), respectively. For N the largest as many field plots would be required to provide an equivalent
RE was actually found for UAV-LS data, however the estimate level of precision to that obtained using auxiliary data from the
was determined as potentially biased (see Figure 4) and thus UAV. In addition to the economic value of these data, UAV-LS
excluded from further comparison. Gains in RE for H were largest in regions with little to no terrain data would offer substantial
using either the UAV-LS (RE = 2.2) or ALS data (RE = 1.5) but additional benefits such as improved information for harvest
non-existent when relying on the UAV photogrammetric data planning, while UAV imagery would offer a high-resolution map
(RE = 1.0). Other than for TSV, UAV-LS resulted in consistently of the area that may further aid in harvest planning and iden-
larger RE than ALS. The use of predictor variables extracted tifying areas of damage or abnormal yield (e.g. wind damage)
from a normalized photogrammetric point cloud (i.e. UAV- that may be missed using field plots alone. Imprecise forest
SfMDTM ) resulted in consistently larger RE than the use of inventory estimates remain a significant problem in the study
DTM-independent variables (i.e. UAV-SfM) for almost all biophysi- context where measurement of a larger sample size can be
cal parameters. Using UAV-SfMDTM data provided marked RE gains deemed too expensive for smaller forests. This can lead to sub-
over UAV-SfM for G (1.6 vs 1.0) and TSV (2.1 vs 1.1) (Table 4). optimal transactions during forest sales or inefficient market-
ing and logistics following harvest. Integrating UAV data in the
manner we have investigated offers higher precision forest inven-
tory estimates alongside valuable novel information for forest
Discussion managers.
Our results highlight the advantages of using UAVs to support Despite these benefits, the gains reported here using MB infer-
inventory data acquisition for small forests, in which it may not ence were relatively small compared to previous research that
be economically feasible to undertake traditional manned aerial has shown an RE of 2.5– 8.0 (Ene et al., 2012; Næsset et al., 2013;
survey. Field plots and UAV data for this study were measured Puliti et al., 2017) for volume using ALS data as the auxiliary vari-
by using a well-established commercial contractor. Using these able. These modest gains in RE reflect the relatively homogenous
costs as a basis for comparison, the cost of acquiring the UAV nature of the sampled stand and the relatively larger sampling
data could have paid for the collection of an additional 17 field intensity of field data compared to previous studies. The smaller

9
Forestry

Table 4 Estimated biophysical parameters according to the studied inferential frameworks (DB and model-based) and data sources in terms of
 μ̂DB )) or mean squared error (MSE(
estimated mean (μ̂), variance (Var(  μ̂MB )) of the mean, standard error (SE), standard error as a percentage of
 μ̂DB ) and MSE(
the estimated mean for case A (SE% ), relative efficiency as a ratio between Var(  μ̂MB ). The models that resulted in the most precise MB
estimates for each of the biophysical parameters are highlighted

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


 μ̂)or MSE(
 μ̂) b SEanalytic
Variable of Data used in Inferential μ̂ Var( SE SE% RE nc SEbootstrap
interest the estimation frameworka

H Field plots DB 36.3 0.1 0.26 0.7 - 30 -


UAV-SfM MB 36.1 0.1 0.3 0.7 1.0 31 1.1
UAV-SfMDTM 36.9 0.1 0.3 0.7 1.0 29 1.6
UAV-LS 36.3 0.0 0.2 0.5 2.2 66 1.3
ALS 36.4 0.0 0.2 0.6 1.5 46 1.3
N Field plots DB 366.1 98.7 9.9 2.7 - 30 -
UAV-SfM MB 355.9 65.7 8.1 2.3 1.5 45 1.1
UAV-SfMDTM 362.0 84.6 9.2 2.5 1.2 35 1.2
UAV-LS 350.5 59.4 7.7 2.2 1.7 50 0.9
ALS 361.0 176.9 13.3 3.7 0.6 17 1.5
G Field plots DB 65.8 1.8 1.4 2.1 - 30 -
UAV-SfM MB 62.6 1.8 1.3 2.1 1.0 31 1.0
UAV-SfMDTM 64.4 1.1 1.1 1.6 1.6 49 0.8
UAV-LS 64.6 1.8 1.4 2.1 1.0 30 0.9
ALS 64.7 1.5 1.2 1.9 1.3 38 1.0
TSV Field plots DB 780.9 282.6 16.8 2.2 - 30 -
UAV-SfM MB 739.1 246.8 15.7 2.1 1.1 34 0.9
UAV-SfMDTM 769.0 131.6 11.5 1.5 2.1 64 0.8
UAV-LS 786.5 203.6 14.3 1.8 1.4 42 0.8
ALS 766.2 218.4 14.8 1.9 1.3 39 1.0

a The adopted inferential frameworks consisted in either DB or MB.


b Forthe DB estimator the uncertainty is expressed as variance, while for the MB inference as mean square error.
c n = number of plots required to obtain the same level of precision as using field data alone.

Figure 4 distribution of the explanatory variables for the sample and for the entire population models for two example models. The first row represents
the H model fitted using UAV-SfMDTM data and the second row the N model fitted using UAV-LS data.

variation in stand biophysical variables, associated with a rather With such limited number of plots the uncertainty of the DB mean
large sampling intensity, resulted in great precision of the DB estimate is likely to increase substantially compared to this study.
estimate and reduced model performance. In such conditions, the MB framework may be attractive as it
In contrast to this study where a large number of field plots enables the use of external data to augment small local field
were measured to obtain a relatively small uncertainty of the data acquisitions for modelling purposes. Within the UAV realm,
field-based estimates, in operational conditions only a portion Puliti et al. (2017) showed how the use of external data in UAV MB
of these would be measured (i.e. one day of work corresponding and hybrid inference can substantially increase the RE. Thus, fur-
to approximately eight field plots in the study forest conditions). ther research should address possibilities to develop large scale

10
Precision inventory of small-forest properties

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


Figure 5 Point estimates and 95 per cent confidence intervals for the studied forest biophysical parameters, estimators and remotely sensed data
sources. The dashed horizontal lines represent the 95 per cent confidence interval for the DB estimate.

models using field and UAV data acquired over different areas predictive accuracy, particularly for H, but also to a lesser extent
and apply them to newly acquired UAV data and a small sample for all other variables including TSV. Given that UAV-SfM inven-
of field data. Such a possibility could be particularly interesting tories are not limited to areas where ALS DTMs are available,
for single tree species plantations as it could enable substantial further research should explore the use of this approach for the
reductions in the costs for field data acquisition. estimation of biophysical parameters across a range of forest
The use of UAV photogrammetric data (i.e. UAV-SfM and UAV- types and stand conditions.
SfMDTM ) provided more accurate model predictions than LS data The use of UAV-LS was notable for producing the largest
(i.e. UAV-LS and ALS) for all biophysical parameters and the RE for H and models that were nearly identical to the most
largest RE for all variables butH. This highlights the advantages accurate model obtained from UAV-SfMDTM . This broadly agrees
of fine resolution UAV imagery that can capture both structural with findings by Wallace et al. (2016) who showed UAV-LS to
information through SfM as well as the spectral and textural be more accurate (RMSE = 0.92 m) than UAV-SfMDTM (RMSE =
properties of the forest canopy. The most precise estimates were 1.30 m) for prediction of height. Results from this study extended
found when the SfM data were height normalized to a DTM the comparison between these two UAV data sources to N, G
obtained from ALS data even though for H, we found indica- and TSV which, in contrast to H, showed larger RE for models
tions of bias in the estimators. Giannetti et al. (2018) demon- developed from UAV-SfMDTM than those developed from UAV-
strated how UAV-SfM DTM-independent variables could be used LS data. The UAV-LS sensor used in this study was developed
to model growing stock volume with similar accuracy to ALS data. primarily for automotive applications. Consequently, the laser
Our study confirmed the results by Giannetti et al. (2018) and operates at less power and with larger beam divergence than
extended this approach to a broader range of inventory variables. many survey-grade ALS scanners, which strongly restricts the
Interestingly, for estimating N and discarding the estimator using ability of the laser beam to penetrate through small canopy gaps.
UAV-LS because potentially biased, UAV-SfM data led to the most Furthermore, the UAV-LS scanner used in this study only provided
precise estimates among the remaining alternatives. However, a single return; a multi-return sensor may better characterize
when using UAV-SfM to estimate G and TSV for the entire forest the lower canopy improving the depiction of some biophysical
area, our results revealed that not only the RE was close to one parameters. In terms of our results, the large precision found
but that the estimators also indicated potential negative bias. for H estimates is, therefore, likely to be due to the increased
Thus, it appears that there is need to better understand how probability of resolving actual tree top heights because of the
well DTM-independent variables (i.e. UAV-SfM) perform beyond very large point density. Further research should examine how
assessing their model predictive accuracy. When compared to estimates made using UAV-SfMDTM compare to those from more
UAV-SfMDTM , we found models using UAV-SfM to have reduced advanced survey-grade UAV-LS (e.g. Riegl VUX-1UAV), which offer

11
Forestry

both improved penetration, larger point densities and multiple


returns per pulse. References
A unique characteristic of this study in regard to the UAV Agisoft 2017 Agisoft PhotoScan user manual: professional edition. Version
literature is that we not only assessed the accuracy of the models 1, 3.
fitted using different combinations of biophysical variables and Breidenbach, J., McRoberts, R.E. and Astrup, R. 2016 Empirical coverage of

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


remotely sensed data sources but also reported the uncertainty model-based variance estimators for remote sensing assisted estimation
and potential bias of the MB estimators. MB estimation has the of stand-level timber volume. Remote Sens. Environ. 173, 274–281.
advantage that it does not require a probability sample from Dandois, J.P. and Ellis, E.C. 2013 High spatial resolution three-dimensional
the target area (Ståhl et al., 2016) and that MSE estimators are mapping of vegetation spectral dynamics using computer vision. Remote
not overly sensitive to geolocation mismatches between field Sens. Environ. 136, 259–276.
plots and the auxiliary data (Saarela et al., 2016b). It is therefore Dandois, J.P., Olano, M. and Ellis, E.C. 2015 Optimal altitude, overlap, and
important for further research to address the use of field and weather conditions for computer vision UAV estimates of forest structure.
UAV data external to the area of interest to capture the potential Remote Sens. (Basel), 7 (10), 13895–13920.
benefits of MB inference. Dash, J.P., Marshall, H.M. and Rawley, B. 2015 Methods for estimating
In this study, tree height was estimated for unmeasured trees multivariate stand yields and errors using k-NN and aerial laser scanning.
based on the relationship between tree height and DBH in a sub- Forestry 88, 237–247.
sample of representative trees within each plot. This is standard Dash, J.P., Watt, M.S., Pearse, G.D., Heaphy, M. and Dungey, H.S. 2017
practice in many forest inventory systems but may introduce Assessing very high resolution UAV imagery for monitoring forest health
some uncertainty into estimates of forest biophysical parameters during a simulated disease outbreak. ISPRS J. Photogram. Remote Sensing
such as the total stem volume. The uncertainty introduced by 131, 1–14.
allometric model prediction in this manner is increasingly rec- Efron, B. and Tibshirani, R.J. 1994 An Introduction to the Bootstrap. CRC
ognized and may have adverse effects on estimate precision. press.
A rigorous analysis of this additional uncertainty component Ene, L.T., Næsset, E., Gobakken, T., Gregoire, T.G., Ståhl, G. and Nel-
was beyond the scope of this paper, but as the same data and son, R. 2012 Assessing the accuracy of regional LiDAR-based biomass
predictive allometry was used for all estimators this should not estimation using a simulation approach. Remote Sens. Environ. 123,
579–592.
adversely affect the comparison presented.
Giannetti, F., Chirici, G., Gobakken, T., Næsset, E., Travaglini, D. and Puliti, S.
2018 A new approach with DTM-independent metrics for forest growing
Conclusion stock prediction using UAV photogrammetric data. Remote Sens. Environ.
213, 195–205.
This study demonstrated how UAV auxiliary data can effectively
Goodbody, T.R.H., Coops, N.C., Tompalski, P., Crawford, P. and Day, K.J.K.
support precision forestry practices for small-forest owners by 2017 Updating residual stem volume estimates using ALS- and UAV-
providing them with advanced data analytics otherwise available acquired stereo-photogrammetric point clouds. Int. J. Remote Sensing, 38
only for large forest enterprises. Our results support the use of (8-10), 2938–2953.
MB estimation in small forest properties that are homogenous in Goulding, C. and Murray, J. 1976 Polynomial taper equations that are
structure. UAV-SfMDTM data yielded some of the most precise esti- compatible with tree volume equations. NZJ For. Sci., 5 (3), 313–322.
mates out of the studied data sources and variables of interest, Gregoire, T.G. 1998 Design-based and model-based inference in sur-
highlighting the relevance of using UAV photogrammetric data vey sampling: appreciating the difference. Can. J. For. Res., 28 (10),
for estimation of key variables such as N, G and TSV. 1429–1447.
Haralick, R.M., Shanmugam, K. and Dinstein, I.H. 1973 Textural
features for image classification. IEEE Trans. Sys. Man Cybern. 6,
Acknowledgements 610–621.
Heaphy, M., Watt, M.S., Dash, J.P. and Pearse, G.D. 2017 UAVs for data
The authors would like to acknowledge the forest manager Jeff Tomble-
collection - plugging the gap. NZ J. Forestry Sci., 62 (1), 23–30.
son for providing access to the study forest and for supporting our
research. Ben Morrow and Robin Hartley of Scion are acknowledged for Hugenholtz, C.H., Moorman, B.J., Riddell, K. and Whitehead, K. 2012
their contribution to UAV data collection. Small unmanned aircraft systems for remote sensing and earth science
research. Eos, Trans Amn Geophysical Union, 93 (25), 236–236.
Jaakkola, A., Hyyppä, J., Kukko, A., Yu, X., Kaartinen, H., Lehtomäki, M.
et al. 2010 A low-cost multi-sensoral mobile mapping system and its
Conflict of interest statement feasibility for tree measurements. ISPRS J. Photogram. Remote Sensing,
None declared. 65 (6), 514–522.
Lisein, J., Pierrot-Deseilligny, M., Bonnet, S. and Lejeune, P. 2013 A pho-
togrammetric workflow for the creation of a Forest canopy height model
Funding from small unmanned aerial system imagery. Forests, 4 (4), 922.
Lowe, D.G. 2004 Method and apparatus for identifying scale invariant
Forest Growers Levy Trust, AGMARDT; The New Zealand Farm features in an image and use of same for locating an object in an image.
Forestry Association; the Neil Barr Foundation; Scion’s Strategic Google Patents.
Science Investment Funding within the ‘Improving Mandallaz, D. 2013 Design-based properties of some small-area estima-
Small-Plantation and Woodlot Inventory’ project; Norwegian tors in forest inventory with two-phase sampling. Can. J. For. Res., 43 (5),
Institute for Bioeconomy Research (NIBIO). 441–449.

12
Precision inventory of small-forest properties

McRoberts, R.E. 2006 A model-based approach to estimating forest area. Saarela, S., Holm, S., Grafström, A., Schnell, S., Næsset, E., Gregoire, T.G. et
Remote Sens. Environ., 103 (1), 56–66. al. 2016a Hierarchical model-based inference for forest inventory utilizing
McRoberts, R.E., Næsset, E. and Gobakken, T. 2013 Inference for lidar- three sources of information. Ann. Forest Sci., 73 (4), 895-910.
assisted estimation of forest growing stock volume. Remote Sens. Environ. Saarela, S., Schnell, S., Tuominen, S., Balázs, A., Hyyppä, J., Grafström, A.
128, 268–275. et al. 2016b Effects of positional errors in model-assisted and model-

Downloaded from https://academic.oup.com/forestry/advance-article-abstract/doi/10.1093/forestry/cpz057/5648988 by guest on 02 December 2019


McRoberts, R.E., Næsset, E., Gobakken, T., Chirici, G., Condés, S., Hou, Z. based estimation of growing stock volume. Remote Sens. Environ. 172,
et al. 2018 Assessing components of the model-based mean square error 101–108.
estimator for remote sensing assisted forest applications. Can. J. For. Res., Salamí, E., Barrado, C. and Pastor, E. 2014 UAV flight experiments applied
48 (6), 642–649. to the remote sensing of vegetated areas. Remote Sens. (Basel), 6 (11),
Næsset, E. 2002 Predicting forest stand characteristics with airborne 11051-11081.
scanning laser using a practical two-stage procedure and field data. Ståhl, G., Saarela, S., Schnell, S., Holm, S., Breidenbach, J., Healey, S.P.
Remote Sens. Environ., 80 (1), 88–99. et al. 2016 Use of models in large-area forest surveys: comparing model-
Næsset, E., Gobakken, T., Bollandsås, O.M., Gregoire, T.G., Nelson, R. and assisted, model-based and hybrid estimation. Forest Ecosys., 3 (1), 5.
Ståhl, G. 2013 Comparison of precision of biomass estimates in regional Talbot, B., Rahlf, J. and Astrup, R. 2017 An operational UAV-based
field sample surveys and airborne LiDAR-assisted surveys in Hedmark approach for stand-level assessment of soil disturbance after forest
County, Norway. Remote Sens. Environ. 130, 108–120. harvesting. Scandinavian J. Forest Res. 1–10.
Nex, F. and Remondino, F. 2014 UAV for 3D mapping applications: a Wallace, L., Lucieer, A., Malenovský, Z., Turner, D. and Vopěnka, P. 2016
review. Appl. Geomatics, 6 (1), 1–15. Assessment of Forest structure using two UAV techniques: A comparison
Petterson, H. 1955 Yield of coniferous forests. Medd. Stat. Skogsforsok- of airborne laser scanning and structure from motion (SfM) point clouds.
sanct Sweden, 45 IB. Forests, 7 (3), 62.
Pierzchała, M., Talbot, B. and Astrup, R. 2014 Estimating soil displacement Wallace, L., Lucieer, A., Watson, C. and Turner, D. 2012 Development of
from timber extraction trails in steep terrain: Application of an unmanned a UAV-LiDAR system with application to forest inventory. Remote Sens.
aircraft for 3D modelling. Forests, 5 (6), 1212. (Basel), 4 (6), 1519.
Pinheiro, J. and Bates, D. 2006 Mixed-Effects Models in S and S-PLUS. Watt, M.S., Heaphy, M., Dunningham, A. and Rolando, C. 2017 Use of
Springer Science & Business Media. remotely sensed data to characterise weed competition in forest plan-
Pinheiro, J., Bates, D., DebRoy, S. and Sarkar, D. 2019 nlme: linear and tations. Int. J. Remote Sensing 38, 2448–2463.
nonlinear mixed effects models. R Package 3, 1–140 Ed., CRAN. Watt, P.J. and Watt, M.S. 2013 Development of a national model of tree
Puliti, S., Ene, L.T., Gobakken, T. and Næsset, E. 2017 Use of partial- volume from LiDAR metrics for New Zealand. Remote Sens. Environ. 34,
coverage UAV data in sampling for large scale forest inventories. Remote 5892–5904.
Sens. Environ. 194, 115–126. Watts, A.C., Ambrosia, V.G. and Hinkley, E.A. 2012 Unmanned aircraft
Puliti, S., Ørka, H.O., Gobakken, T. and Næsset, E. 2015 Inventory of small systems in remote sensing and scientific research: classification and
forest areas using an unmanned aerial system. Remote Sens. (Basel), 7 considerations of use. Remote Sens. (Basel), 4 (6), 1671–1692.
(8), 9632–9654. Whitehead, K. and Hugenholtz, C.H. 2014 Remote sensing of the environ-
Puliti, S., Saarela, S., Gobakken, T., Ståhl, G. and Næsset, E. 2018a Com- ment with small unmanned aircraft systems (UASs), part 1: a review of
bining UAV and Sentinel-2 auxiliary data for forest growing stock volume progress and challenges 1. J. Unmanned Vehicle Sys., 2 (3), 69–85.
estimation through hierarchical model-based inference. Remote Sens. Whitehead, K., Hugenholtz, C.H., Myshak, S., Brown, O., LeClair, A., Tam-
Environ., 204 (Supplement C), 485–497. minga, A. et al. 2014 Remote sensing of the environment with small
Puliti, S., Talbot, B. and Astrup, R. 2018b Tree-stump detection, segmen- unmanned aircraft systems (UASs), part 2: scientific and commercial
tation, classification, and measurement using unmanned aerial vehicle applications 1. J. Unmanned Vehicle Sys., 2 (3), 86–102.
(UAV) imagery. Forests, 9 (3), 102. Zvoleff, A. 2016 glcm: Calculate Textures from Grey-Level Co-Occurrence
R Core Team 2017 R: A Language and Environment for Statistical Comput- Matrices (GLCMs). R package version 1.6.1. https://CRAN.R-project.org/pa
ing. R Foundation for Statistical Computing, Vienna, Austria. ckage=glcm

13

You might also like