titlegical Classification Framework
Characterizing Forest Succession in Central Ontario using LiDAR derived Indices
Karin van Ewijk, Paul Treitz and Neal Scott Department of Geography, Queen’s University, Kingston
Motivation
Motivation Oliver and Larson’s framework (1990) •
Two perspectives on stand development: 1. 2.
•
Structure Process
General patterns of forest development following both major and minor disturbances Stand initiation
Adapted from Oliver and Larson, 1990
Stem exclusion
Understory re‐initiation
Old growth
Research Goal To develop LiDAR derived structural indices that are suitable predictors of the four seral stages described by Oliver and Larson. LiDAR derived indices: Height DBH Percentiles Deciles Vertical structure Canopy density
Study Site Great Lakes – St. Lawrence Forest Region (GLSLFR), Ontario, Canada: Transitional zone between southern deciduous forest of eastern North America and predominantly coniferous boreal forest in the north.
Source: OMNR, 2004
Petawawa Research Forest: Species composition: eastern white pine, red pine, eastern hemlock, white cedar, white and black spruce, jack pine; yellow and white birch, poplars, sugar and red maple, basswood, red oak.
Data LiDAR data
Field data
ALTM 3100 sensor (Optimal Geomatics)
Forest variables
Altitude: 1000 m System PRF: 100 kHz Scan Freq: 54 Hz Scan Half Angle: 13° Cross Track Resolution: 0.499 m Down Track Resolution: 0.572 m
(plot size: 400 m2)
Stand types
Conifer dominated Intolerant Hardwoods Mixed Woods Tolerant Hardwoods
Seral stage Stand initiation (1) Stem exclusion (2) Understory re‐ initiation (3) Old growth (4)
N (total=66) 3 13
Six flight lines flown in August 2007 = 3 hits/m2
Nominal density (a subset of the flight lines had a density of 10 hits/m2)
DBH, Tree height, Species composition, Seral stage, Location: diff. corrected GPS
19 31
METHODOLOGY
Forest Variables LiDAR derived Variables HGT AVE
Average Height
HGT SD and AD
Standard deviation and Absolute deviation of Heights
HGT L where h is tree height and g is it’s basal area QMDBH where dbh is diameter at breast height, and n is stems per plot SKEW and KURTOSIS
Measures of asymmetry and “peakedness” of the probability distribution of height
P90
90th Height Percentile
D2 and D6
2nd and 6th Height Decile
DB
Canopy Density Index = Proportion 1st Veg. Returns / Proportion All Returns
VCI where HB is the total number of height bins, and pi is the proportional abundance of LiDAR returns in height bin i.
VCI Description of the four forest stand development stages
Stand initiation
Stem exclusion
Understory re‐initiation
LiDAR point clouds at 3 returns per m2
Frequency of LiDAR returns per height bin (2 m)
Old growth
Analyses A. Density and Height Bin effects on VCI B. Binomial Logit Model Assumptions: 1. 2. 3.
No multicollinearity (VIF
Karin van Ewijk, Paul Treitz and Neal Scott Department of Geography, Queen’s University, Kingston
Motivation
Motivation Oliver and Larson’s framework (1990) •
Two perspectives on stand development: 1. 2.
•
Structure Process
General patterns of forest development following both major and minor disturbances Stand initiation
Adapted from Oliver and Larson, 1990
Stem exclusion
Understory re‐initiation
Old growth
Research Goal To develop LiDAR derived structural indices that are suitable predictors of the four seral stages described by Oliver and Larson. LiDAR derived indices: Height DBH Percentiles Deciles Vertical structure Canopy density
Study Site Great Lakes – St. Lawrence Forest Region (GLSLFR), Ontario, Canada: Transitional zone between southern deciduous forest of eastern North America and predominantly coniferous boreal forest in the north.
Source: OMNR, 2004
Petawawa Research Forest: Species composition: eastern white pine, red pine, eastern hemlock, white cedar, white and black spruce, jack pine; yellow and white birch, poplars, sugar and red maple, basswood, red oak.
Data LiDAR data
Field data
ALTM 3100 sensor (Optimal Geomatics)
Forest variables
Altitude: 1000 m System PRF: 100 kHz Scan Freq: 54 Hz Scan Half Angle: 13° Cross Track Resolution: 0.499 m Down Track Resolution: 0.572 m
(plot size: 400 m2)
Stand types
Conifer dominated Intolerant Hardwoods Mixed Woods Tolerant Hardwoods
Seral stage Stand initiation (1) Stem exclusion (2) Understory re‐ initiation (3) Old growth (4)
N (total=66) 3 13
Six flight lines flown in August 2007 = 3 hits/m2
Nominal density (a subset of the flight lines had a density of 10 hits/m2)
DBH, Tree height, Species composition, Seral stage, Location: diff. corrected GPS
19 31
METHODOLOGY
Forest Variables LiDAR derived Variables HGT AVE
Average Height
HGT SD and AD
Standard deviation and Absolute deviation of Heights
HGT L where h is tree height and g is it’s basal area QMDBH where dbh is diameter at breast height, and n is stems per plot SKEW and KURTOSIS
Measures of asymmetry and “peakedness” of the probability distribution of height
P90
90th Height Percentile
D2 and D6
2nd and 6th Height Decile
DB
Canopy Density Index = Proportion 1st Veg. Returns / Proportion All Returns
VCI where HB is the total number of height bins, and pi is the proportional abundance of LiDAR returns in height bin i.
VCI Description of the four forest stand development stages
Stand initiation
Stem exclusion
Understory re‐initiation
LiDAR point clouds at 3 returns per m2
Frequency of LiDAR returns per height bin (2 m)
Old growth
Analyses A. Density and Height Bin effects on VCI B. Binomial Logit Model Assumptions: 1. 2. 3.
No multicollinearity (VIF
