You are on page 1of 24

GIScience & Remote Sensing

ISSN: 1548-1603 (Print) 1943-7226 (Online) Journal homepage: https://www.tandfonline.com/loi/tgrs20

An integrated GIS tool for automatic forest


inventory estimates of Pinus radiata from LiDAR
data

Yang Chen & Xuan Zhu

To cite this article: Yang Chen & Xuan Zhu (2013) An integrated GIS tool for automatic forest
inventory estimates of Pinus�radiata from LiDAR data, GIScience & Remote Sensing, 50:6,
667-689, DOI: 10.1080/15481603.2013.866783

To link to this article: https://doi.org/10.1080/15481603.2013.866783

Published online: 05 Dec 2013.

Submit your article to this journal

Article views: 421

View related articles

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tgrs20
GIScience & Remote Sensing, 2013
Vol. 50, No. 6, 667–689, http://dx.doi.org/10.1080/15481603.2013.866783

An integrated GIS tool for automatic forest inventory estimates of


Pinus radiata from LiDAR data
Yang Chen* and Xuan Zhu

School of Geography and Environmental Science, Monash University, Clayton Campus, VIC 3800,
Australia
(Received 15 August 2013; accepted 11 November 2013)

This article presents an integrated GIS tool for automatic forest inventory of Pinus
radiata plantations from light detection and ranging (LiDAR) data. Built as a set of
tools running in the desktop GIS software package ArcGIS, it integrates spatial
analysis, LiDAR data analysis and image segmentation techniques as well as empirical
tree models to support forest inventories of Pinus radiata on an individual-tree basis.
The integrated GIS tool allows users to define or select plots to extract LiDAR data for
forest inventory, build and process canopy height models (CHMs) from the extracted
LiDAR data through surface modelling, delineate individual trees on the CHMs by
applying the marker-controlled watershed segmentation technique, and to derive forest
inventory estimates based on the CHMs and identified individual trees through spatial
analysis and tree modelling using the empirical models. It takes advantage of combin-
ing GIS and LiDAR to automatically conduct a forest inventory, build and manage a
forest inventory database and to spatially and statistically summarise and visualise the
inventory data. Although developed for forest inventories of Pinus radiata plantations
in Victoria, Australia, the integrated tool can be customised for forest inventories of
Pinus radiata plantations in other regions and for other types of plantations by
incorporating empirical tree models built and calibrated for those regions and those
species of trees.
Keywords: forest inventory; airborne LiDAR; GIS; Pinus radiata

1. Introduction
Forest inventories are the systematic collections of comprehensive information on the
quality and quantity of forest resources and are the fundamental tools for forest manage-
ment and planning (Köhl, Magnussen, and Marchetti 2006). Early concepts of sustainable
forest management and forest inventory focused on timber productivity (Cotta 1804;
Hartig 1975; Leblanc 2008). The recent inventory requirements have been gradually
expanded to multiple forest functions, forest biomass and biodiversity (Corona, Köhl,
and Marchetti 2004; Leblanc 2008; Lund and Smith 1997; Von Gadow, Nagel, and
Saborowski 2002). Forest inventories may have different purposes and various data
categories, and are often carried out at several points of time for the analysis of temporal
changes (Köhl, Magnussen, and Marchetti 2006). They may cover different spatial scales
from the global scale to national, regional, management-plot or compartment level and
finally stand level. Management-plot level inventories are mainly carried out in order to
facilitate harvesting and silvicultural decisions. The basic information in a management-

*Corresponding author. Email: yche158@student.monash.edu

© 2013 Taylor & Francis


668 Y. Chen and X. Zhu

plot level inventory often includes age, predominant stand height (PDH), diameter at
breast height (DBH), basal area (BA), stocking, site quality and timber volume.
A forest inventory could in principle be established based on a complete census of a
given area by measuring every tree. However, to manually measure every individual tree
is usually infeasible when large areas are involved. Traditional methods used in forest
inventories are based on sampling. They typically involve first grouping similar forest
stands according to stand age, structure, species and soil properties, and then selecting
representative sample plots (often circular plots), measuring inventory parameters or
attributes for all trees located in the selected sample plots, deriving means and totals for
measures of forest characteristics over different forest strata based on sample measure-
ments, and finally building a database for maintaining and managing the forest inventory
(Leblanc 2008). However, management-plot level forest inventories based on sampling
methods are labour-intensive and expensive, and do not provide information on how the
stratum information is distributed among the management plots (Holopainen and
Kalliovirta 2006; Leblanc 2008). Remote sensing has been applied to solve this problem
as it provides spatially explicit data from forest stand level to individual tree level for large
areas in a timely and economic fashion (Boyd and Danson 2005; Kätsch and Kunneke
2006; Lunetta 1999; Roller 2000). Remote sensing can be used to derive forest inventory
estimates based on either an image segment level or individual tree level analysis. Image
segmentation allows the identification and delineation of forest stands over large areas
according to tone, texture and other image characteristics on remote-sensing images,
based on which stand-wise average tree parameters can be determined (Pekkarinen
2002; Tuominen and Pekkarinen 2005). The accuracy of forest inventory estimates
based on image segmentation depends on the quality of the images, which also requires
successful correction of radiometric distortions in the images. Tree level analysis involves
identifying and measuring individual tree parameters using either digital photogrammetry
or laser scanning (Korpela 2004; Persson, Holmgren, and Soderman 2002).
Photogrammetric methods require stereo pairs of large-scale aerial photos or high-resolu-
tion satellite images as well as photogrammetric plotters. However, accurate photogram-
metry is impossible when the ground is obscured. In recent years, the laser scanning
technology light detection and ranging (LiDAR) was emerged as a rapid and an efficient
tool for forest inventories as it offers the advantage of efficient large-area coverage and
ability to retrieve three-dimensional information about individual trees, such as tree height
and crown closure (van Leeuwen and Nieuwenhuis 2010).
LiDAR is an active remote-sensing technique, and involves transmitting pulses of
laser light towards the ground and measuring the time of pulse return. Depending on the
altitude of the LiDAR instrument and the angle at which the pulse is sent, each pulse
illuminates a near circular area on the ground, which can generate one return or multiple
returns. These multiple returns can be analysed and classified to produce information
about objects above the ground as well as the bare ground surface. First returns can be
used to create digital surface models (DSMs) that include features above the ground
surface such as buildings and tree canopy. Intermediate returns are helpful in separating
vegetation from solid objects among the above ground features. Final returns are the first
approximation of the bare ground surface and can be used to build digital elevation
models (DEM). The use of LiDAR for forest inventory purposes began in the early 1980s
(Lim et al. 2003). LiDAR has been successfully used for characterising forest structures
(both horizontal and vertical) and estimating forest inventory parameters, including
canopy height (Andersen, McGaughey, and Reutebuch 2005; Brandtberg 2007; Harding
et al. 2001; Holmgren 2004; Hudak et al. 2002; Næsset and Bjerknes 2001), timber
GIScience & Remote Sensing 669

volume (Holmgren 2004; Nilsson 1996; Popescu 2007; Rowell et al. 2006; Solberg,
Naesset, and Bollandsas 2006), forest biomass (Lefsky et al. 1999; Nelson, Short, and
Valenti 2004; Popescu 2007; Suárez et al. 2005), vertical structure of forest stands
(Brandtberg 2007; Means et al. 1999), individual tree heights (Bienert et al. 2006;
Brandtberg 2007; García, Suárez, and Patenaude 2007; Leckie et al. 2003; Maas et al.
2008; Popescu, Wynne, and Nelson 2003; Suárez et al. 2005; Véga and Durrieu 2011),
DBH (Bienert et al. 2006; Maas et al. 2008) and site quality (Rombouts, Ferguson, and
Leech 2010). However, majority of the current studies on LiDAR for forest inventories
focus on methodological issues, i.e. trial and assessment of different statistical and image
processing methods for extraction of forest variables from LiDAR data. Some of these
methods are sophisticated and can only be implemented in specialised software packages.
As LiDAR goes far beyond the proof-of-concept phase and attracts much attention from
the forestry community, there is still a lack of a fully automated forest inventory system
built upon LiDAR data. This research aims to fill this gap.
In this article, we present an integrated GIS tool for automatic forest inventory of
Pinus radiata from LiDAR data, which was built by integrating a set of LiDAR data
processing and extraction tools and tree models (for DBH, BA and stem volume estima-
tion) within a GIS environment. It provides an integrated and easy-to-use GIS tool for
construction of a fully automated forest inventory system for Pinus radiata plantations. In
Sections 2 and 3, we provide a brief overview of the methods for forest inventory
estimates from LiDAR and the role of GIS in forest inventory. Section 4 describes the
functions of the integrated GIS tool. A forest inventory of Pinus radiata plantations in
Victoria, Australia, developed using the tool is introduced in Section 5. The article is
concluded in Section 6 with a discussion of the potential future development of the tool.

2. Forest inventory estimates from LiDAR


LiDAR allows the measurement of various structural forest characteristics from ground-
based, airborne or space-borne platforms. Airborne LiDAR is currently the most viable
option for forest inventories (van Leeuwen and Nieuwenhuis 2010), which is the focus of
this article. Two main groups of methods have been applied for measuring forest inven-
tory attributes using airborne LiDAR data: predictive modelling and image processing.
Predictive modelling involves the use of predictive models (such as regression models
and machine learning models) for predicting forest inventory parameters based on LiDAR
metrics (descriptive statistics of LiDAR point data). A forest inventory parameter, e.g.
mean tree height or site index, is a response or a dependent variable, while LiDAR
metrics, such as percentiles of the first return heights and proportion of ground returns, are
predictors or independent variables. For example, Andersen, McGaughey, and Reutebuch
(2005) applied linear regression analysis to derive average stand heights based on quantile
heights and the fraction of retrieved LiDAR returns from tree canopies in western
Washington State, USA. Rombouts, Ferguson, and Leech (2010) utilised LiDAR data
for site quality assessment of radiata pine plantations in South Australia. They first built
regression relationships between forest stand parameters (including PDH and stand
volume) and LiDAR metrics (including mean height, quadratic mean height, standard
deviation of heights, maximum height, average of the maximum height of first returns in
every plot, percentile heights of the ordered cumulative height distribution for first returns,
proportion of ground returns, proportion of returns in height–frequency distribution
classes, point density from first returns and scanning angle) at the plot level, and then
estimated the PDHs and stand volumes for each plot using the regression models, and
670 Y. Chen and X. Zhu

finally converted the estimated stand volumes to site quality classes or site indices using
age-dependent conversion tables and produced site quality maps. Zhao et al. (2011)
applied two machine learning techniques, support vector machines (SVM) and Gaussian
processes, to estimate canopy structural variables at the plot level over a temperate forest
in East Texas, USA, based on a set of LiDAR metrics, including mean heights, variances
and coefficients of variation for first/single returns, last returns, and all returns, 50
percentile heights of first/single and last returns and 80 canopy density metrics of various
types of returns. Predictive modelling is mainly used to derive stand averages of canopy
variables at the plot level.
Image processing techniques allow the delineation of individual tree crowns and the
estimation of forest inventory parameters from LiDAR data at the individual tree level.
Image processing on LiDAR data is often based on the canopy height model (CHM) (van
Leeuwen and Nieuwenhuis 2010; Zhao et al. 2011). A CHM is a gridded data layer
representing stand-level canopy heights above the around, which can be created by
subtracting the ground DEM from the DSM. Individual trees and tree crowns may be
identified and delineated by detecting local maxima (LM) and identifying the edges of
crowns around the LMs from a CHM (Maltamo et al. 2004; Persson, Holmgren, and
Soderman 2002; Popescu, Wynne, and Nelson 2002; Véga and Durrieu 2011). A number
of algorithms have been developed to optimise LM identification (i.e. to minimise errors
in locating individual trees), including conventional image filtering (Popescu, Wynne, and
Nelson 2002, 2003), template matching (Brandtberg et al. 2003) and height-based variable
window size filtering (Popescu and Wynne 2004). The major techniques for crown
segmentation on the CHM include morphological image analysis (Chen et al. 2006;
Kwak et al. 2007; Stereńczak, Będkowski, and Weinacker 2008), region growing
(Hyyppa et al. 2001; Persson, Holmgren, and Soderman 2002), spatial wavelet analysis
(Falkowski et al. 2006; Persson, Holmgren, and Soderman 2002) and fitting functions
(Popescu, Wynne, and Nelson 2002). The heights of the identified individual trees can be
simply extracted from the CHM according to their locations. Stocking, crown size and
canopy closure can be directly calculated based on the results of tree identification and
crown delineation from the CHM. DBH, site index, volume, biomass and other inventory
parameters cannot be measured directly from the CHM, but can be estimated based on tree
heights using empirical tree models.
Both predictive modelling and image processing have been widely used to provide
forest inventory estimates with varying degrees of success. In this research, we took the
image processing approach combined with empirical tree models.

3. GIS and forest inventory


While methods of data collection, contents and means of storage for forest inventories
vary widely, all comprise data that represent both the thematic and locational attributes of
forest features. Therefore, maps, imagery and related tabular data are the centrepiece of
forest inventories. GIS has been used to store, organise, document, retrieve, integrate,
summarise, update and map land and forest inventory data since the 1960s, when the first
GIS was developed for the Canada Land Inventory (Longley et al. 2005). Before the
introduction of GIS, forest inventories were stored in paper maps, orthophotographs and
tally sheets, and updated by aerial photography, field sampling methods and manual
drafting. With GIS, the forest inventory data are digitised and managed in the computer
system. A forestry inventory can be updated constantly and statistical summaries can be
GIScience & Remote Sensing 671

made instantly. GIS provides an efficient way for forestry agencies, organisations or
companies to manage and manipulate their forest inventory databases.
Currently, a forest inventory GIS is simply a geospatial database, consisting of a
collection of GIS data layers. The fundamental GIS data layer in a forest inventory contains
forest attribute data for each forest stand in a given area, including stand attributes of a
unique identifier, stand age, stand species, stand size, stocking, site index, the level of
reconnaissance (the method used to inventory the stand) and other forest attributes or
variables depending on the purpose(s) of the inventory. Other GIS data layers may represent
underlying topography, soils, logging roads and other relevant non-forest features. GIS
allows rapid updating of a forest inventory, recording and monitoring changes in forest
attributes, integrating inventory data with other spatial data and incorporating models to
support various forest management activities such as timber harvesting, silviculture and fire
management (Köhl, Magnussen, and Marchetti 2006). Today, GIS-based forest inventories
are the norm. However, a forest inventory GIS is still largely used for data management and
mapping, and for producing tabular inventory summaries. The forest inventory data col-
lected or derived from ground samples and remote-sensing imagery are still often required
to be entered into a GIS manually or through digitising. GIS itself does not provide tools to
automatically extract and estimate the forest features from samples and images. Since the
late 1990s, LiDAR has become an increasingly important data collection technique for
forest inventory (Næsset 1997; Næsset and Bjerknes 2001). The integration of GIS and
LiDAR provides an opportunity towards a fully automated forest inventory system, which
can automatically calculate forest inventory estimates from LiDAR using GIS spatial and
image analysis tools, and store them into a forest inventory GIS database. The integrated
GIS tool developed in this research aims to facilitate the development of such an automated
forest inventory system for Pinus radiata plantations in southern Australia.

4. An integrated GIS tool for forest inventory of Pinus radiata plantations


Our integrated GIS tool was developed for forest inventory of Pinus radiata plantations in
Victoria, Australia, owned and managed by Hancock Victorian Plantation Pty Limited
(HVP). It was designed and implemented with ArcGIS 10 (a GIS software package, http://
www.esri.com/software/arcgis/arcgis10) coupled with MCR (the MATLAB Compiler
Runtime for creating and running numerical and digital image processing applications,
http://www.mathworks.com.au/products/compiler/). The integrated GIS tool supports a
three-stage process for automatic forest inventory estimates from LiDAR data: data
extraction, data processing and inventory estimates (Figure 1). The first step, data extrac-
tion, is designed to extract LiDAR point data within the user-defined plots for forest
inventory; the second step, data processing, involves building and processing a CHM
from the extracted LiDAR data, and identifying individual trees and tree crowns on the
CHM; and the last stage, inventory estimates, involves estimating the forest inventory
parameters on an individual-tree basis and then summarising the estimates to the manage-
ment plots. Each step is implemented using a group of functions built as individual geo-
processing tools, and all the functions are organised as a toolset which are managed and
executed from ArcToolBox in ArcGIS (Figure 2).

4.1. Data extraction


LiDAR data sets are generally very large. It is impractical to process all LiDAR data
covering a large area. It is more efficient to capture forest inventory attributes from
672 Y. Chen and X. Zhu

LiDAR
Data

Select plot(s) Data Extraction

ArcGIS Extract LiDAR data

Build a CHM

Smooth a CHM Data Processing

MATLAB Compiler Identify individual


Runtime trees

Measure inventory
attributes
Inventory Estimates
Tree
Stocking
heights

ArcGIS PDH Height Site Dbhub


growth quality

Dbhob

Timber
BA
volume

Figure 1. Procedure of automatic forest inventory estimates from LiDAR.

LiDAR data according to management plots. The function of Select plot(s) of the tool
allows users to define or select single management plots or combinations of plots to
extract LiDAR data. After the plots are selected, the function Extract LiDAR data extracts
the LiDAR data covering the selected plots for data processing.

4.2. Data processing


After the LiDAR data are extracted for the selected plots, a DSM and a DEM can be
generated, which are then used to build a CHM. The function Build a CHM is designed
for this purpose. It first interpolates a DEM and a DSM from the first and ground returns,
respectively, using the Topo to Raster tool in ArcGIS based on the ANUDEM program
(Hutchinson 1989), and then creates a CHM by subtracting the DEM from the DSM.
A CHM might contain errors due to noise or local irregularities existing in the LiDAR
data, which may have resulted from errors in instrument measurement, atmospheric
radiation, ground surface undulation, ground surface curvature or misclassified low
vegetation as ground returns. These errors may create inaccurate segmentation or over-
segmentation when identifying individual trees using the image processing techniques
(Rafael and Richard 2008). Therefore, a CHM is often required to be filtered using image
smoothing techniques to reduce or eliminate the errors (García, Suárez, and Patenaude
GIScience & Remote Sensing 673

Arc Toolbox
3D Analyst Tools
Analysis Tools
Arc Hydro Tools
Cartography Tools
Conversion Tools
Data Interoperability Tools
Data Management Tools
Editing Tools
EdNVI Tools
Forest Inventroy Tools
Data Extraction
Extract LiDAR Data
Select Plot(s)
Data Processing
Build a CHM
Identify individual trees
SMooth a CHM
Inventory Estimates
BA
Dbhob
Dbhub
Height Growth
PDH
Site Quality
Stocking
Timber Volume
Tree Heights

Figure 2. Forest inventory tools in ArcGIS ArcToolBox.

2007). The function Smooth a CHM is used to smooth the generated CHM, using the low-
pass filter available in ArcGIS.
After a smoothed CHM is created, it can be used to identify individual tree tops and
delineate tree crowns by using the function Identify individual trees. This function
implements the marker-controlled watershed segmentation algorithm (Parvati, Rao, and
Das 2008; Serra 1982) to transform a smoothed CHM in order to separate individual trees.
Conceptually, the smoothed CHM is inverted as a grey scale (gradient) image, in which a
tree crown area is considered a watershed. Local minima in the image are identified as
individual trees, and watershed lines are delineated as tree crown edges. With marker-
controlled watershed segmentation, the local minima associated with tree tops are marked
as internal markers and the LM associated with the background or watershed lines are
represented as external markers (Rafael and Richard 2008; Soille 2003). The internal and
external markers are then used to segment the CHM. A procedure for the identification of
individual trees using marker-controlled watershed segmentation was proposed by Chen
674 Y. Chen and X. Zhu

(a) (b) (c)

Figure 3. (a) A smoothed CHM, (b) tree tops and (c) tree crowns.

and Zhu (2012). This procedure was coded in MATLAB language and implemented in
MCR. The function Identify individual trees produces two outputs from a smoothed
CHM: a tree top layer and a tree crown layer, as shown in Figure 3. The tree top layer
is the basis for estimating forest inventory attributes.

4.3. Inventory estimates


Forest inventory estimates are made at the individual tree level using a number of GIS
spatial analysis functions and application-specific empirical tree models with the inte-
grated GIS tool. The number and types of attributes included in a forest inventory depend
on the purpose of the forest inventory. For example, for harvesting and forest operations,
the volume and value of the growing stock and other stocking parameters are the
necessary inventory attributes. But for forest biodiversity assessment, the volume of
dead wood is also required. Currently, the integrated GIS tool contains models for
calculating eight forest inventory attributes: stocking, individual tree heights, height
growth, PDH, site quality, DBH, timber volume and BA. All these models or functions
were written in Python running in ArcGIS.

4.3.1. Stocking
Stocking is defined as the number of stems per hectare. Once individual trees are
identified, this function can be used to calculate stocking as the number of trees in each
plot divided by its area.

4.3.2. Tree heights


This function estimates heights of individual trees by overlaying individual tree tops
identified using the function Identify individual trees with the CHM and extracting the
height values from the CHM for each individual tree.

4.3.3. Predominant stand height (PDH)


PDH at a specific reference age, also known as site index (SI), is often used for site
quality assessment of Pinus radiata plantations in Australia, as it is closely correlated with
site productivity (Lewis and Ferguson 1993; Lewis, Keeves, and Leech 1976; Turner and
Holmes 1985). In southern Australia, PDH is usually defined as the mean height of the 75
tallest trees per hectare (Lewis, Keeves, and Leech 1976; Rombouts, Ferguson, and Leech
GIScience & Remote Sensing 675

2010). In typical forest inventory measurements, PDH was estimated by the mean of the
two tallest trees within 10 m of an assessor in the field (Lewis, Keeves, and Leech 1976).
In our integrated GIS tool, the PDH function overlays a grid of 20 m × 25 m over the user
selected plots, calculates PDH as the mean height of the four tallest trees in each grid cell
and outputs the grid with varying PDH values across the selected plots.

4.3.4. Height growth


Height growth models have been developed for estimating the height growth patterns for
Pinus radiata stands grown in Australia (Candy 1989; Carter and Martin 1992; Haywood
2009). The function of Height growth incorporates a suite of growth forecast models for
Pinus radiata in South Australia, based on locations and main soil types developed by
HVP (2004). Below is an example of a height growth model provided by the Height
growth function which predicts tree heights of Pinus radiata stands on the clay loam soil
(HVP 2004):
 
1  ea2 A1
H1 ¼ ðH2 þ a1 Þ  a1 (1)
1  ea2 A2

where H1 is the height at the measurement year A1, H2 represents the height at any update
age A2 and the parameters a1, and a2 cannot be provided due to the confidential
intellectual property of HVP Pty Ltd.

4.3.5. Site quality


The function is used to classify Pinus radiata stands into seven site quality classes
according to their PDH with reference age 11 years (HVP 2006b). The classification is
based on the yield curve of the total volume production to the age of Pinus radiata
established and revised by Jolly (1941, 1946), Lewis (1953, 1960, 1963c, 1968, 1970) and
Lewis, Keeves, and Leech (1972, 1973). Classes I to III have better soils and faster
growing rates; Class IV is less fertile and has shallower soils; the productivity further
reduces from Classes V to VI; Class VII represents the slowest-growing stands of
acceptable health (Lewis 1971).

4.3.6. Dbhub
This function calculates DBH under bark for Pinus radiata based on the following stem
taper equation developed by HVP (2006a):

Dub
¼ α 1  α 2 h  α 3 h2 (2)
Dbhub

where Dub is diameter under bark at height h and Dbhub is DBH under bark at 1.3 m. If
Dub is set to be Dbhub, the above equation will become

α1  1:3 α2  1:32 α3 ¼ 1 (3)

where assuming H is tree height in m,


676 Y. Chen and X. Zhu

α1 ¼ α2 þ α3

α2 ¼ ðST þ 2α3 Þ
" #
DJ þ ST ð1  HJ Þ
α3 ¼ 
ð1  HJ Þ2

1:3
BH ¼
H

In order to protect the proprietary right of HVP Pty Ltd, the parameters, HJ and ST cannot
be provided.
By solving the above equations, Dbhub can be derived.

4.3.7. Dbhob
The function Dbhob calculates DBH over bark for Pinus radiata based on the following
equation developed by HVP (2006a):

Dbhub ¼ Dbhob  exp½b1 þb2 lnðDbhobÞþb3 lnð AÞ (4)

where Dbhob is DBH over bark at 1.3 m, and A is the stand age, the parameters b1, b2 and
b3 cannot be provided due to confidentiality reasons, at the request of HVP Pty Ltd.
Therefore, this function uses the output from Dbhub as its input.

4.3.8. Basal area


Tree BA is the cross-sectional area of a tree trunk at breast height over bark, which is often
calculated based on Dbhob with the following equation (Hédl et al. 2009; Rowan and
Peter 2001):

BA ¼ 0:00007854Dbhob2 (5)

Stand BA is the sum of the BA of all live trees in a stand. The function BA allows the
calculation of either tree BA or stand BA according to plots defined by the user.

4.3.9. Timber volume


There are different ways to estimate timber volumes. For example, Rowan and Peter
(2001) used the following equation to estimate stand timber volumes in a typical forest
inventory:

SBAHT
SVOL ¼ (6)
3

where SVOL represents the stand timber volume, HT is the mean height of a forest stand
and SBA represents stand basal area. Næsset (1997) estimated SVOLs using LiDAR-
derived mean stand height and mean canopy cover density. Maltamo et al. (2004)
GIScience & Remote Sensing 677

conducted estimation of timber volume based on the weak relation between crown area
and individual tree height derived from LiDAR data. Our Timber volume function
implements the volume function developed by HVP (2006a), which is expressed as
 β 3
β1 β2 H2
TSVub ¼ exp Dbhob (7)
H  1:3

where TSVub represents stem volume under bark; β1 , β2 and β3 are dependent on soil
types; these variable are the confidential intellectual property of HVP Pty Ltd. Stand
volume can be computed by the summing up stem volumes of all live trees in a forest
stand. The Timber volume function allows the calculation of either stem volume under
bark or stand volume according to plots defined by the user.

5. A forest inventory of Pinus radiata plantations in Mt Worth, Victoria


5.1. Study area and data
The integrated GIS tool was applied to conduct a forest inventory of Pinus radiata
plantations in Mt Worth, Victoria, in south-eastern Australia. The study area is situated
on a private plantation managed by HVP, which covers approximately 300 ha (Figure 4).
The plantation consists mainly of Pinus radiata and eucalypts. The elevation ranges from
242 to 494 m a.s.l., the average annual rainfall is about 2000 mm and the main soil type is
clay loam. One hundred eighteen Pinus radiata management plots with six different ages

Planted Years
Roads
1970
1986
1987
1989
1997

0 200 400 m

Figure 4. The study area.


678 Y. Chen and X. Zhu

between 12 and 30 years were selected for the study. The purpose of the forest inventory
was to update the existing inventory for the assessment of current conditions of growing
stock and silvicultural decision making in the study area. It was also used to validate the
inventory estimates derived by the integrated GIS tool.
The LiDAR data used for the inventory were captured during an aerial campaign
conducted in October 2009, using the Airborne Optech Gemini LiDAR system, which has
a beam footprint of 0.3 m in diameter. The LiDAR system is a discrete return system,
recording up to four laser returns per laser pulse. The flying height was approximately
1100 m a.s.l., and the swath width was around 950 m. The horizontal and vertical
accuracies were 0.25 and 0.2 m, respectively. The LiDAR data conform to the LAS 1.2
specification (ASPRS 2008; Esri 2011), which already have ground and non-ground
returns separated. We assumed that the point classification was correct in the LiDAR
data. The LiDAR data covering the 118 management plots were extracted using the
Extract LiDAR data function, and a CHM was created with the resolution of
0.25 m × 0.25 m, using the Build a CHM function, which was then smoothed using the
Smooth a CHM function. Figure 5 shows the smoothed CHM.

5.2. Inventory results


The inventory attributes include stocking, tree height, PDH, site quality, DBH under bark,
DBH over bark, BA and timber volume. They were estimated by using the corresponding
functions in the integrated GIS tool. Figures 6–8 show the maps of stocking, site quality
and timber volume aggregated to management plots. Table 1 lists part of the inventory
attribute database generated using the integrated tool.

5.3. Validation
Individual tree heights, tree locations and DBH over bark were validated using the field
survey data.
Two field surveys were carried out to measure tree parameters for the accuracy
assessment of those calculated based on the LiDAR data using the integrated GIS tool.
The first one was conducted in September 2011, and the second was in December 2011.
In the two surveys, a total of 100 trees from six plots of Pinus radiata of different ages
were selected. Each plot covers an area of 20 m × 20 m (400 m2). The locations of six
reference points beside each plot were determined using a differential GPS. The locations
of the selected trees in a given plot were calculated from distance and angle measurements
taken from the corresponding reference point using a Vertex clinometer and a compass.
The heights and diameters over bark at breast height (1.3 m) of the surveyed trees were
measured with a Vertex clinometer and diameter tape, respectively. The results of valida-
tion are described below.

5.3.1. Tree height


As the LiDAR data were collected in 2009 and the field measurements were made in
2011, the tree heights estimated using the function of Tree heights based on the LiDAR
data were adjusted by using the growth forecast models for Pinus radiata developed by
HVP (2004). The adjusted tree heights were used for all other inventory estimates in this
case study. Since the major soil type in our study area is clay loam, the following equation
GIScience & Remote Sensing 679

0 200 400 m

Figure 5. The CHM of the study area.

was used to calculate the increment of tree height from 2009 to 2011, which was
transformed from Equation (1):

 
1  ea2 A2
I ¼ ðH1 þ a1 Þ  a1  H 1 (8)
1  ea2 A1

where I represents the increment of tree height from age A1 to age A2, H1 is the tree height
measured at age A1 and the variables a1 and a2 are confidential. Here, A1 and A2 are the
680 Y. Chen and X. Zhu

Stocking
(number of
tress/ha)
Roads
0–100
101–300
301–500
501–700
701–1000

0 200 400 m

Figure 6. Stocking of Pinus radiata.

ages for the years of 2009 and 2011, respectively. H1 is the tree height estimated from
LiDAR.
After the LiDAR-derived tree heights were adjusted using Equation (8), the root mean
square error (RMSE) was calculated using Equation (9):

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u N
uX ð xi  x0 Þ 2
RMSE ¼ t i
(9)
i¼1
N

where xi is the height of tree i measured in the field, xi′ represents the adjusted LiDAR-
derived height of tree i and N is the number of trees sampled in the field. The smaller the
RMSE is, the more accurate the LiDAR-derived tree heights are. An RMSE value of
0.42 m was obtained in our case, which indicates a high accuracy. Figure 9 shows the
scattergram with the fitted regression line for the adjusted LiDAR-derived tree heights and
field-measured tree heights. An R2 value of 0.98 and a slope value of 0.99 indicate that the
LiDAR-derived tree heights match the field measurements very well.
GIScience & Remote Sensing 681

Site Index
Roads
1
2
3
4
5
6
7

0 200 400 m

Figure 7. Site quality of Pinus radiata.

5.3.2. Tree locations


The accuracy of the tree locations identified from the LiDAR data was assessed using the
maximum horizontal distance between the 100 trees sampled in the field and their
corresponding LiDAR-derived treetops, which was 1.87 m. It should be noted that tree-
tops detected from the LiDAR data are not necessarily the exact locations of the trees.

5.3.3. DBH over bark


The accuracy of DBH over bark (Dbhob) estimated using the function Dbhob based on the
LiDAR data was also assessed using RMSE as expressed in Equation (9). Here, xi and xi′
represent Dbhob instead of tree height. The field-measured Dbhob values of the 100
sampled trees and their LiDAR-estimated Dbhob values were used with Equation (9),
which gave an RMSE value of 63.32 mm. They also showed a linear relationship with an
R2 value of 0.72, as illustrated in Figure 10. Therefore, the discrepancy between the field-
measured Dbhob values and the results derived from LiDAR data is negligible.
682 Y. Chen and X. Zhu

Timber Volume
(m3)
Roads
0–500
501–1000
1001–3000
3001–5000
5001–15000

0 200 400 m

Figure 8. Timber volume of Pinus radiata.

As we did not have field measurement data for other inventory attributes, including
stocking, PDH, site quality, DBH under bark, BA and timber volume, their LiDAR-
derived values were not directly validated. However, PDH and site quality were derived
based on tree heights; DBH under bark, BA and timber volume are the functions of tree
height and Dbhob. As tree heights and Dbhob values estimated from the LiDAR data
using the integrated GIS tool had a high accuracy as discussed above, the accuracies of the
estimates for these inventory attributes derived by the tool should be close to the
accuracies of the models used for their estimation. HVP (2006a) tested the stem taper
model expressed in Equation (2), using the 2005 independent measurement data in
Victoria and found that the mean error of estimates of DBH under bark by the model
was –0.39 cm on clay, –0.09 cm on sand and –0.07 cm on duplex sites. Their model for
timber volume as expressed in Equation (7) was tested using the same set of measurement
data and found to produce a mean error of –0.0002 m3 on clay, 0.0023 m3 on duplex and
0.0018 m3 on sand sites (HVP 2006a).
GIScience & Remote Sensing 683

Table 1. Part of the inventory attribute database.

Year Plots Area (ha) Stocking PDH (m) Site index Timber volume (m3) …

1979 1 1.71 47 16.31571 4 1516.71 …


2 0.9 430 15.40875 5 628.86 …
3 1.41 831 15.98326 4 2813.52 …
1986 4 0.96 457 12.97421 6 313.18 …
5 0.94 435 14.85731 5 360.21 …
6 0.44 532 17.5482 3 243.97 …
7 1.1 554 17.93241 3 637.57 …
8 0.12 750 14.02716 5 70.14 …
9 0.02 700 9.493073 7 3.05 …
10 0.74 464 13.46076 6 250.46 …
11 9.25 70 14.12255 5 476.68 …
12 1.97 626 14.17739 5 860.86 …
13 1.65 530 17.07773 3 831.12 …
14 0.27 522 17.52233 3 141.76 …
15 5.1 512 16.25533 4 2415.39 …
16 14.5 569 14.79701 5 6922.19 …
17 3.01 535 15.42441 5 1405.73 …
18 7.59 533 15.23338 5 3456 …
. . . . . . …
. . . . . . …
. . . . . . …

y = 0.44 + 0.99 * x
LiDAR-derived tree heights (m)

R-Square = 0.98
35.00

30.00

25.00
25.00 30.00 35.00
Field-measured tree heights (m)

Figure 9. Scattergram of tree heights.

6. Discussion and conclusions


This article has presented an integrated GIS tool for automatic forest inventory estimates
of Pinus radiata from LiDAR data. It provides a new automatic approach to forest
inventory through the integration of GIS and LiDAR. LiDAR has proven an effective
tool that can capture forest inventory measures accurately and consistently, and is chan-
ging the paradigm of forest inventory (Holopainen and Kalliovirta 2006; van Leeuwen
684 Y. Chen and X. Zhu

375 y = 165.86 + 0.54*x


R-Square = 0.72

LiDAR-derived Dbhob (mm)


350

325

315 350 385


Field-measured Dbhob (mm)

Figure 10. Scattergram of DBH over bark.

and Nieuwenhuis 2010; Zhao et al. 2011). While conventional forest inventories are based
on survey sampling and statistical inference techniques, LiDAR can enumerate the
population of trees within a large-area forest, directly measure forest variables and
automate forest inventory estimates in a timely and cost-effective manner. GIS provides
a platform for LiDAR data analysis, modelling, spatial database management and visua-
lisation. Therefore, our tool takes advantage of combining GIS and LiDAR to extract
forest information on an individual-tree basis, conduct a forest inventory by aggregating
individual tree parameters to user-defined management plots, automatically build and
manage a forest inventory database and to visualise the inventory data. Although it was
developed for forest inventories of Pinus radiata plantations in Victoria, Australia, the
integrated GIS tool can be customised for forest inventories of Pinus radiata plantations
in other regions and for other types of plantations by incorporating empirical tree models
built and calibrated for those regions and those species of trees.
Forest inventory, particularly operational forest inventory at the stand or plot level,
often requires high accuracy (Hyyppa and Hyyppa 2001). Hyyppa and Hyyppa (1999)
found that LiDAR-derived forest inventory attributes were more accurate than those
obtained with other optical remote-sensing inventory methods, and LiDAR was the only
remote-sensing method that met the requirements for accuracy in operative stand-based
forest inventories. Holmgren (2004) and Næsset (2004) reported that the accuracies of
LiDAR-derived estimates at the tree, plot and stand levels were very similar or even better
than those achieved in traditional field inventories. With the integrated GIS tool, the
accuracies of inventory estimates are highly dependent on the accuracies with individual
tree identification and tree height estimation. Several image processing methods have
been developed for single tree delineation based on LiDAR data as discussed in the
second section. Among them, the marker controlled watershed segmentation technique
has been shown to be a robust and flexible method for isolating individual trees (Ali,
Dare, and Jones 2008; Chen et al. 2006; Li et al. 2012; Liu and Shao 2012). It performs
watershed segmentation around user-specified or automatically generated markers rather
than the LM in the CHM built from LiDAR data, where each individual tree is segmented
with its crown boundary delineated as ridges. Marker-controlled watershed segmentation
GIScience & Remote Sensing 685

effectively avoids the over-segmentation problem with watershed segmentation (Soille


2003), and was found to be superior to conventional image filtering, height-based variable
window size filtering in delineating individual trees (Chen et al. 2006). The integrated GIS
tool applied this technique to identify individual trees and delineate tree crowns in a CHM
and achieved a minimum accuracy of 1.87 m for tree locations, an RMSE of 0.42 m for
tree heights and an RMSE of 63.32 mm for DBH over bark. However, it should be
acknowledged that our test area is covered by commercial Pinus radiata plantations,
which are dense and unthinned, but have a single species, simple structure and litter
understory. The tool may be well suited for such commercial forests, but it has yet been
tested in forest areas with mixed species and complex structure with intensive ground and
low vegetation.
In addition, the forest inventories of Pinus radiata plantations using the integrated GIS
tool currently focus on stocking parameters and timber volumes. As a continuation of the
development of the tool, new inventory attributes concerning forest biomass, ecological
biodiversity and forest health will be added. This will require the development or
assessment of new methods and new models for estimation of these attributes. It will
also involve the use of other remote-sensing data such as aerial photos and high-resolution
multispectral satellite images to provide more details of the spatial geometry and spectral
information for classifying tree species and assessing biodiversity and health. Further
studies will be conducted on methods for fusing LiDAR data with other types of remote-
sensing data and new tree models in the integrated GIS tool, as new inventory attributes
are introduced and the nature of forest inventory evolves. The integrated GIS tool can be
further exploited to incorporate new methods for integrating LiDAR data with other types
of remote-sensing data for BA, DBH and timber volume measurements as well as tree
growth modelling, thus removing the needs of empirical tree models for inventory
estimates that may be site-specific and contain uncertainties.

Acknowledgements
The authors would like to thank HVP Pty Limited for providing all LiDAR and related GIS data as
well as growth models, taper models and timber volume models. Special thanks go to Henry
Lieshout, Adam Newnham, Steve Eagle, Wendy Wright and Hong Xie, who provided valuable
support throughout the project.

References
Ali, S. S., P. Dare, and S. D. Jones. 2008. “Fusion of Remotely Sensed Multispectral Imagery and
LIDAR Data for Forest Structure Assessment at the Tree Level.” Paper presented at the Silk
Road for Information from Imagery, Beijing, July 3–11.
Andersen, H.-E., R. J. McGaughey, and S. E. Reutebuch. 2005. “Estimating Forest Canopy Fuel
Parameters Using LIDAR Data.” Remote Sensing of Environment 94 (4): 441–449. doi:10.1016/
j.rse.2004.10.013.
ASPRS. 2008. LAS Specification Version 1.2. Available at http://www.asprs.org/a/society/
committees/standards/lidar_exchange_format.html
Bienert, A., S. Scheller, E. Keane, G. Mullooly, and F. Mohan. 2006. “Application of Terrestrial
Laser Scanners for the Determination of Forest Inventory Parameters.” International Archives of
Photogrammetry, Remote Sensing and Spatial Information Sciences 36. http://www.isprs.org/
proceedings/XXXVI/part5/
Boyd, D. S., and F. M. Danson. 2005. “Satellite Remote Sensing of Forest Resources: Three
Decades of Research Development.” Progress in Physical Geography 29 (1): 1–26.
doi:10.1191/0309133305pp432ra.
686 Y. Chen and X. Zhu

Brandtberg, T. 2007. “Classifying Individual Tree Species Under Leaf-Off and Leaf-on Conditions
Using Airborne Lidar.” International Journal of Photogrammetry and Remote Sensing 61 (5):
325–340.
Brandtberg, T., T. A. Warner, R. E. Landenberger, and J. B. McGraw. 2003. “Detection and Analysis
of Individual Leaf-Off Tree Crowns in Small Footprint, High Sampling Density Lidar Data from
the Eastern Deciduous Forest in North America.” Remote Sensing of Environment 85 (3):
290–303.
Candy, S. G. 1989. “Growth and Yield Models for Pinus radiata in Tasmania.” New Zealand
Journal of Forestry Science 19 (1): 112–133.
Carter, P., and L. Martin. 1992. Models to Describe Height/Age Relationships in Conifer Species in
NSW. Internal Report. Sydney: Forestry Commission of New South Wales.
Chen, Q., D. Baldocchi, P. Gong, and M. Kelly. 2006. “Isolating Individual Trees in a Savanna
Woodland Using Small Footprint Lidar Data.” Photogrammetric Engineering Remote Sensing
72 (8): 923–932.
Chen, Y., and X. Zhu. 2012. “Site Quality Assessment of a Pinus radiata Plantation in Victoria,
Australia, Using LiDAR Technology.” Southern Forests: A Journal of Forest Science 74 (4):
217–227. doi:10.2989/20702620.2012.741767.
Corona, P., M. Köhl, and M. Marchetti. 2004. Advances in Forest Inventory for Sustainable Forest
Management and Biodiversity Monitoring. Dordrecht: Kluwer.
Cotta, H. 1804. Systematische Anleitung Zur Taxation Der Waldungen 1. Berlin: Sander.
Esri. 2011. Lidar Analysis in ArcGIS 10 for Forestry Applications. An Esri White Paper. Redlands,
CA: ESRI.
Falkowski, M. J., A. M. S. Smith, A. T. Hudak, P. E. Gessler, L. A. Vierling, and N. L. Crookston.
2006. “Automated Estimation of Individual Conifer Tree Height and Crown Diameter via Two-
Dimensional Spatial Wavelet Analysis of Lidar Data.” Canadian Journal of Remote Sensing
32 (2): 153–161. doi:10.5589/m06-005.
García, R., J. C. Suárez, and G. Patenaude. 2007. “Delineation of Individual Tree Crowns for
LiDAR Tree and Stand Parameter Estimation in Scottish Woodlands.” In The European
Information Society, edited by S. I. Fabrikant, and M. Wachowicz, 55–85. Berlin: Springer.
Harding, D. J., M. A. Lefsky, G. G. Parker, and J. B. Blair. 2001. “Laser Altimeter Canopy Height
Profiles: Methods and Validation for Closed-Canopy, Broadleaf Forests.” Remote Sensing of
Environment 76 (3): 283–297. doi:10.1016/s0034-4257(00)00210-8.
Hartig, G. L. 1975. Anweisung Zur Taxation Der Forste Oder Zur Bestimmung Des Holzertrags Der
Wälder. Gießen: Heyer.
Haywood, A. 2009. “Estimation of Height Growth Patterns and Site Index Curves for Pinus Radiata
Plantations in New South Wales, Australia.” Southern Forests 71 (1): 11–19. doi:10.2989/
Sf.2009.71.1.2.739.
Hédl, R., M. Svátek, M. Dančák, A. W. Rodzay, A. B. M. Salleh, and A. S. Kamariah. 2009. “A
New Technique for Inventory of Permanent Plots in Tropical Forests: A Case Study
from Lowland Dipterocarp Forest in Kuala Belalong, Brunei Darussalam.” Blumea –
Biodiversity, Evolution and Biogeography of Plants 54 (1–3): 124–130. doi:10.3767/
000651909x475482.
Holmgren, J. 2004. “Prediction of Tree Height, Basal Area and Stem Volume in Forest Stands Using
Airborne Laser Scanning.” Scandinavian Journal of Forest Research 19 (6): 543–553.
doi:10.1080/02827580410019472.
Holopainen, M., and J. Kalliovirta. 2006. “Modern Data Acquisition for Forest Inventories.” In
Managing Forest Ecosystems, edited by A. Kangas, and M. Maltamo, Vol. 10, 343–362.
Dordrecht: Springer.
Hudak, A. T., M. A. Lefsky, W. B. Cohen, and M. Berterretche. 2002. “Integration of Lidar and
Landsat ETM+ Data for Estimating and Mapping Forest Canopy Height.” Remote Sensing of
Environment 82 (2–3): 397–416. doi:10.1016/s0034-4257(02)00056-1.
Hutchinson, M. F. 1989. “A New Procedure for Gridding Elevation and Stream Line Data with
Automatic Removal of Spurious Pits.” Journal of Hydrology 106: 211–232.
HVP. 2004. Development of New Pinus radiata Growth Models for the GRP Estate (Confidential).
Victoria: HVP.
HVP. 2006a. Evaluation and Development of Volume and Taper Equations for the GRP Pine Estate
(Confidential). Resource information and business systems. Victoria: HVP.
GIScience & Remote Sensing 687

HVP. 2006b. Site Classification and Yield Forecasting of the HVP Pine Estate (Confidential).
Victoria: HVP.
Hyyppa, H., and J. Hyyppa. 1999. “Comparing the Accuracy of Laser Scanner with Other Optical
Remote Sensing Data Sources for Stand Attribute Retrieval.” The Photogrammetric Journal of
Finland 16: 5–15.
Hyyppa, H. J., and J. M. Hyyppa. 2001. “Effects of Stand Size on the Accuracy of Remote Sensing-
Based Forest Inventory.” Geoscience and Remote Sensing, IEEE Transactions on 39 (12):
2613–2621. doi:10.1109/36.974996.
Hyyppa, J., O. Kelle, M. Lehikoinen, and M. Inkinen. 2001. “A Segmentation-Based Method to
Retrieve Stem Volume Estimates from 3-D Tree Height Models Produced by Laser Scanners.”
IEEE Transactions on Geoscience and Remote Sensing 39 (5): 969–975.
Jolly, N. W. 1941. Yield Table for Unthinned Pinus radiata in the South-East, Ages 10–30 Years.
Adelaide: Woods and Forests Department.
Jolly, N. W. 1946. Yield Table for Unthinned Pinus radiata in the South-East. Adelaide: Woods and
Forests Department.
Kätsch, C., and A. Kunneke. 2006. “Forest Inventory in the Digital Remote Sensing Age.” The
Southern African Forestry Journal 206 (1): 43–49. doi:10.2989/10295920609505243.
Köhl, M., S. S. Magnussen, and M. Marchetti. 2006. Sampling Methods, Remote Sensing and GIS
Multiresource Forest Inventory. SpringerLink. Berlin: Springer.
Korpela, I. 2004. Individual Tree Measurements by Means of Digital Aerial Photogrammetry.
Helsinki: Finnish Forest Research Institute, Finnish Society of Forest Science.
Kwak, D. H., W. K. Lee, J. H. Lee, G. S. Biging, and P. Gong. 2007. “Detection of Individual Trees
and Estimation of Tree Height Using LiDAR Data.” Forest Research 12 (6): 425–434.
Leblanc, J. W. 2008. What Do We Own: Understanding Forest Inventory. Working in the Woods.
Hollister, CA: University of California Cooperative Extension.
Leckie, D., F. Gougeon, D. Hill, R. Quinn, L. Armstrong, and R. Shreenan. 2003. “Combined High-
Density Lidar and Multispectral Imagery for Individual Tree Crown Analysis.” Canadian
Journal of Remote Sensing 29 (5): 633–649.
Lefsky, M. A., D. Harding, W. B. Cohen, G. Parker, and H. H. Shugart. 1999. “Surface Lidar
Remote Sensing of Basal Area and Biomass in Deciduous Forests of Eastern Maryland, USA.”
Remote Sensing of Environment 67 (1): 83–98. doi:10.1016/s0034-4257(98)00071-6.
Lewis, N. B. 1953. Revised Yield Tables–Tentative Total Production Curves. Adelaide: Woods and
Forests Department.
Lewis, N. B. 1960. An Estimation of Log Yield Volumes by Royalty Class Assortments in Thinned
Stands on Rotations of 30, 40, 50, and 60 Years. Adelaide: Woods and Forests Department.
Lewis, N. B. 1963c, 1968, 1970. Provisional Yield Table for Pinus radiata – Thinned Stands in South
Australia (Total Production to Age). Adelaide: Woods and Forests Department.
Lewis, N. B. 1971. “Site assessment in South Australian pine plantations.” Paper presented at the
15th I.U.F.R.O. Congress, University of Florida, Gainesville, FL, March 15–20.
Lewis, N. B., and I. S. Ferguson. 1993. Management of Radiata Pine/N.B. Lewis and I.S. Ferguson
in Association with W.R.J. Sutton, D.G.M. Donald, H.B. Lisboa. Melbourne: Inkata Press.
Lewis, N. B., A. Keeves, and J. W. Leech. 1972. Provisional Yield Tales for Thinned Stands in South
Australia. Adelaide: Woods and Forests Department.
Lewis, N. B., A. Keeves, and J. W. Leech. 1973. Thinned Stand Yield Tables: Pinus Radiata: South-
East, South Australia. Adelaide: Woods and Forests Department.
Lewis, N. B., A. Keeves, and J. W. Leech. 1976. Yield Regulation in South Australian Pinus Radiata
Plantations. Adelaide: Woods and Forests Dept., South Australia.
Li, W., Q. Guo, M. K. Jakubowski, and M. Kelly. 2012. “A New Method for Segmenting Individual
Trees from the Lidar Point Cloud.” Photogrammetric Engineering & Remote Sensing 78 (1):
75–84.
Lim, K., P. Treitz, M. Wulder, B. Æ. St-Onge, and M. Flood. 2003. “LiDAR Remote Sensing of
Forest Structure.” Progress in Physical Geography 27 (1): 88–106. doi:10.1191/
0309133303pp360ra.
Liu, C., and Z. F. Shao. 2012. “Estimation of Forest Carbon Storage Based on Airborne LiDAR
Data.” Applied Mechanics and Materials 195–196: 1314–1320.
Longley, P. A., M. F. Goodchild, D. J. Maguire, and D. W. Rhind. 2005. Geographic Information
Systems and Science. 2nd ed. Chichester: John Wiley.
688 Y. Chen and X. Zhu

Lund, H. G., and W. B. Smith. 1997. “The United States Forest Inventory Program.” In Proceedings
of the FAO/ECE meeting of experts on Global Forest Resources Assessment, edited by A.
Nyssönen, and A. Ahti. Helsinki: Finnish Forest Research Institute.
Lunetta, R. S. 1999. “Remote Sensing Change Detection: Environmental Monitoring Methods and
Applications.” In Remote Sensing Change Detection: Environmental Monitoring Methods and
Applications, edited by R. S. Lunetta and C. D. Elvidge, 1st ed., 318. London: Taylor &
Francis.
Maas, H. G., A. Bienert, S. Scheller, and E. Keane. 2008. “Automatic Forest Inventory Parameter
Determination from Terrestrial Laser Scanner Data.” International Journal of Remote Sensing
29 (5): 1579–1593. doi:10.1080/01431160701736406.
Maltamo, M., K. Eerikainen, J. Pitkanen, J. Hyyppa, and M. Vehmas. 2004. “Estimation of Timber
Volume and Stem Density Based on Scanning Laser Altimetry and Expected Tree Size
Distribution Functions.” Remote Sensing of Environment 90 (3): 319–330. doi:10.1016/j.
rse.2004.01.006.
Means, J. E., S. A. Acker, D. J. Harding, J. B. Blair, M. A. Lefsky, W. B. Cohen, and W. A. McKee.
1999. “Use of Large-Footprint Scanning Airborne Lidar to Estimate Forest Stand Characteristics
in the Western Cascades of Oregon.” Remote Sensing of Environment 67 (3): 298–308.
doi:10.1016/s0034-4257(98)00091-1.
Næsset, E. 1997. “Estimating Timber Volume of Forest Stands Using Airborne Laser Scanner Data.”
Remote Sensing of Environment 61 (2): 246–253. doi:10.1016/s0034-4257(97)00041-2.
Næsset, E. 2004. “Practical Large-Scale Forest Stand Inventory Using a Small-Footprint Airborne
Scanning Laser.” Scandinavian Journal of Forest Research 19 (2): 164–179. doi:10.1080/
02827580310019257.
Næsset, E., and K. O. Bjerknes. 2001. “Estimating Tree Heights and Number of Stems in Young
Forest Stands Using Airborne Laser Scanner Data.” Remote Sensing of Environment 78 (3):
328–340. doi:10.1016/s0034-4257(01)00228-0.
Nelson, R., A. Short, and M. Valenti. 2004. “Measuring Biomass and Carbon in Delaware Using an
Airborne Profiling LIDAR.” Scandinavian Journal of Forest Research 19 (6): 500–511.
doi:10.1080/02827580410019508.
Nilsson, M. 1996. “Estimation of Tree Heights and Stand Volume Using an Airborne Lidar System.”
Remote Sensing of Environment 56 (1): 1–7. doi:10.1016/0034-4257(95)00224-3.
Parvati, K., B. S. P. Rao, and M. M. Das. 2008. “Image Segmentation Using Gray-Scale
Morphology and Marker-Controlled Watershed Transformation.” Discrete Dynamics in Nature
and Society 2008. doi:10.1155/2008/384346.
Pekkarinen, A. 2002. “A Method for the Segmentation of Very High Spatial Resolution Images of
Forested Landscapes.” International Journal of Remote Sensing 23 (14): 2817–2836.
doi:10.1080/01431160110076162.
Persson, Å., J. Holmgren, and U. Söderman. 2002. “Detecting and Measuring Individual Trees
Using an Airborne Laser Scanner.” Photogrammetric Engineering and Remote Sensing 68 (9):
925–932.
Popescu, S. C. 2007. “Estimating Biomass of Individual Pine Trees Using Airborne Lidar.” Biomass
and Bioenergy 31 (9): 646–655. doi:10.1016/j.biombioe.2007.06.022.
Popescu, S. C., and R. H. Wynne. 2004. “Seeing the Trees in the Forest: Using Lidar and
Multispectral Data Fusion with Local Filtering and Variable Window Size for Estimating Tree
Height.” Photogrammetric Engineering & Remote Sensing 70: 589–604.
Popescu, S. C., R. H. Wynne, and R. F. Nelson. 2002. “Estimating Plot-Level Tree Heights with
Lidar: Local Filtering with a Canopy-Height-Based Variable Window Size.” Computers and
Electronics in Agriculture 37 (1–3): 71–95.
Popescu, S. C., R. H. Wynne, and R. F. Nelson. 2003. “Measuring Individual Tree Crown Diameter
with Lidar and Assessing Its Influence on Estimating Forest Volume and Biomass.” Canadian
Journal of Remote Sensing 29: 564–577. Ottawa: Canadian Aeronautics and Space.
Rafael, C. G., and E. W. Richard. 2008. Digital Image Processing. 3rd ed. Harlow: Prentice Hall.
Roller, N. 2000. “Intermediate Multispectral Satellite Sensors.” Journal of Forestry 98 (6): 32–35.
Rombouts, J., I. S. Ferguson, and J. W. Leech. 2010. “Campaign and Site Effects in LiDAR
Prediction Models for Site-Quality Assessment of Radiata Pine Plantations in South
Australia.” International Journal of Remote Sensing 31 (5): 1155–1173. doi:10.1080/
01431160903380573.
GIScience & Remote Sensing 689

Rowan, R., and S. Peter. 2001. The Farmer’s Forest: Multipurpose Forestry for Australian Farmers.
Melbourne: Australian Master TreeGrower.
Rowell, E., C. Seielstad, L. Vierling, L. L. Queen, and W. Shepperd. 2006. “Using Laser Altimetry-
Based Segmentation to Refine Automated Tree Identification in Managed Forests of the Black
Hills, South Dakota.” Photogrammetric Engineering and Remote Sensing 72: 1379–1388.
Serra, J. 1982. Image Analysis and Mathematical Morphology. Vol. 1. London: Academic Press.
Soille, P. 2003. Morphological Image Analysis: Principles and Applications. 2nd ed. Berlin:
Springer.
Solberg, S., E. Naesset, and O. M. Bollandsas. 2006. “Single Tree Segmentation Using Airborne
Laser Scanner Data in a Structurally Heterogeneous Spruce Forest.” Photogrammetric
Engineering Remote Sensing 72 (12): 1369–1378.
Stereńczak, K., K. Będkowski, and H. Weinacker. 2008. “Accuracy of Crown Segmentation and
Estimation of Selected Trees and Forest Stand Parameters in Order to Resolution of Used DSM
and NDSM Models Generated from DENSE Small Footprint LiDAR Data.” The International
Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences Youth
Forum 38 (B6b): 27–33.
Suárez, J. C., C. Ontiveiros, S. Smith, and S. Snape. 2005. “Use of Airborne LiDAR and Aerial
Photography in the Estimation of Individual Tree Heights in Forestry.” Computers &
Geosciences 31 (2): 253–262. doi:10.1016/j.cageo.2004.09.015.
Tuominen, S., and A. Pekkarinen. 2005. “Performance of Different Spectral and Textural Aerial
Photograph Features in Multi-Source Forest Inventory.” Remote Sensing of Environment 94 (2):
256–268. doi:10.1016/j.rse.2004.10.001.
Turner, J., and G. I. Holmes. 1985. “Site Classification of Pinus Radiata Plantations in the Lithgow
District, New South Wales, Australia.” Forest Ecology and Management 12 (1): 53–63.
doi:10.1016/0378-1127(85)90136-7.
van Leeuwen, M., and M. Nieuwenhuis. 2010. “Retrieval of Forest Structural Parameters Using
LiDAR Remote Sensing.” European Journal of Forest Research 129 (4): 749–770. doi:10.1007/
s10342-010-0381-4.
Véga, C., and S. Durrieu. 2011. “Multi-Level Filtering Segmentation to Measure Individual Tree
Parameters Based on Lidar Data: Application to a Mountainous Forest with Heterogeneous
Stands.” International Journal of Applied Earth Observation and Geoinformation 13 (4): 646–
656. doi:10.1016/j.jag.2011.04.002.
Von Gadow, K., J. Nagel, and J. Saborowski. 2002. Continuous Cover Forestry: Assessment,
Analysis, Scenarios. Dordrecht: Kluwer Academic.
Zhao, K., S. Popescu, X. Meng, Y. Pang, and M. Agca. 2011. “Characterizing Forest Canopy
Structure with Lidar Composite Metrics and Machine Learning.” Remote Sensing of
Environment 115 (8): 1978–1996. doi:10.1016/j.rse.2011.04.001.

You might also like