14. But WHY?
Source: Cape Bounty, 2009 Source: CAVM Team, 2003 Circu
0°
45
Cryptoga
W
°
°
E
45
Cryptoga
Noncarbo
Carbonat
Prostrate
Prostrate
Rush/gra
Graminoi
Nontusso
Tussock s
Erect dwa
Low-shru
90° E 180°
Sedge/gr
Source: NASA, 2011 Sedge, m
Sedge, m
Nunatak c
Glaciers
N Water
80 °
Lagoon
Ar
ct
Non-Arcti
ic
Cir
cle
13
5°
E
W
5°
13
La
Longi
Derived from: C
Vegetation Map
Flora and Faun
180° Service, Ancho
16. Research Objectives
1. Model spatial variability in plant
composition and percent cover for the
Sabine Peninsula, Melville Island, Nunavut
using multispectral satellite imagery
17. Research Objectives
1. Model spatial variability in plant
composition and percent cover for the
Sabine Peninsula, Melville Island, Nunavut
using multispectral satellite imagery
2. Develop and test new broad-band
spectral vegetation indices well-suited
for characterizing biophysical properties
of Arctic tundra vegetation
50. 8 • G.J. LAIDLER et al.
120
100
80
Percent Cover
60
% Cover La NDVI
40 R2 = 0.78
% Cover Ik NDVI
R2 = 0.72
20
% Cover Su NDVI
R2 = 0.74
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70
NDVI
FIG. 6. Linear relations between percent cover and NDVI values for surface
(Su), IKONOS (Ik), and Landsat (La) sensors.
DISCUSSION
Moisture and Percent Cover Field Estimates
In the majority of Arctic locations, the environmental
factor most closely correlated with vegetation type is soil
moisture (Oberbauer and Dawson, 1992). In areas of high
elevation, water is a limiting factor and an important determi-
nant of vegetation structure, productivity, and composition;
in lower areas, these aspects may not be controlled directly by
soil moisture, but rather by factors correlated with or affected FIG. 7. Landsat (top) and IKONOS (bottom) NDVI images for a sub-area
(approximately 6.8 km2) within the study area (around study plots P1–P3). Dark
by soil moisture, such as nutrient availability, thaw depth, soil areas represent regions of low NDVI (-1), while bright areas indicate high
aeration, redox potential, and pH (Oberbauer and Dawson, NDVI (+1), typically lush green vegetation.
1992). Micro-scale moisture gradients (across a few metres),
such as those in periglacial features (troughs to high-centre al., 1994; Walker et al., 1994; Murray, 1997; Henry, 1998;
polygons, frost boils, or stone stripes) or from wet meadows Young et al., 1999). The spatial heterogeneity of the study
to beach ridges, have great influence on the pattern and area, regarded as the irregularity of the physical environ-
51. 8 • G.J. LAIDLER et al.
120
100
80
Percent Cover
60
% Cover La NDVI
40 R2 = 0.78
% Cover Ik NDVI
R2 = 0.72
20
% Cover Su NDVI
R2 = 0.74
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70
NDVI
FIG. 6. Linear relations between percent cover and NDVI values for surface
(Su), IKONOS (Ik), and Landsat (La) sensors.
DISCUSSION
Moisture and Percent Cover Field Estimates
In the majority of Arctic locations, the environmental
factor most closely correlated with vegetation type is soil
moisture (Oberbauer and Dawson, 1992). In areas of high
elevation, water is a limiting factor and an important determi-
nant of vegetation structure, productivity, and composition;
in lower areas, these aspects may not be controlled directly by
soil moisture, but rather by factors correlated with or affected FIG. 7. Landsat (top) and IKONOS (bottom) NDVI images for a sub-area
(approximately 6.8 km2) within the study area (around study plots P1–P3). Dark
by soil moisture, such as nutrient availability, thaw depth, soil areas represent regions of low NDVI (-1), while bright areas indicate high
aeration, redox potential, and pH (Oberbauer and Dawson, NDVI (+1), typically lush green vegetation.
1992). Micro-scale moisture gradients (across a few metres),
such as those in periglacial features (troughs to high-centre al., 1994; Walker et al., 1994; Murray, 1997; Henry, 1998;
polygons, frost boils, or stone stripes) or from wet meadows Young et al., 1999). The spatial heterogeneity of the study
to beach ridges, have great influence on the pattern and area, regarded as the irregularity of the physical environ-
52. 10 • G.J. LAIDLER et al.
8 • G.J. LAIDLER et al.
Source: Laidler, Treitz, and Atkinson,
120
100
80
Percent Cover
60
% Cover La NDVI
40 R2 = 0.78
% Cover Ik NDVI
R2 = 0.72
20
% Cover Su NDVI
R2 = 0.74
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70
NDVI
FIG. 6. Linear relations between percent cover and NDVI values for surface
(Su), IKONOS (Ik), and Landsat (La) sensors.
DISCUSSION FIG. 8. Image of percent cover from IKONOS NDVI, calculated from the following regression equation: Y = 275.5 (IK_NDVI) + 16.07 (R2 = 0.716, p < 0.01).
linear, significant, and consistent across scales (i.e., R2 = the initial conditions for patch-scale models by inventory-
Moisture and Percent Cover Field Estimates 0.72 – 0.78; p < 0.01) (Fig. 6), corroborating trends re- ing landscape conditions and their relative proportions; ii)
ported in other Arctic environments. stratify landscapes into relatively homogeneous response
An image of percent cover derived from the IKONOS units for spatially distributed modeling of material and
In the majority of Arctic locations, the environmental NDVI data is presented in Figure 8, which portrays the energy transport; iii) extrapolate model simulations by
factor most closely correlated with vegetation type is soil relationship between percent cover variations (Fig. 3), mapping areas that are potentially sensitive to particular
topographic trends (Table 2), and associated moisture disturbances; and iv) assess landscape- and regional-scale
moisture (Oberbauer and Dawson, 1992). In areas of high regimes (Fig. 2). Percent cover increases along declining model simulations by comparative spatial pattern analy-
elevation, water is a limiting factor and an important determi- elevations and slopes as a reflection of increased vegeta- ses. Here, coefficients of determination for NDVI and
tion canopy density in areas of high moisture (i.e., water- percent cover were very similar for the IKONOS and
nant of vegetation structure, productivity, and composition; tracks, drainage channels, and areas with moderate to Landsat data. This similarity is a function of averaging the
in lower areas, these aspects may not be controlled directly by minimal exposure). Modeling percent cover over the en- IKONOS NDVI data to the plot level for correlation
soil moisture, but rather by factors correlated with or affected FIG. 7. Landsat (top) and IKONOS (bottom)tirevegetationprovidesaan interesting perspective on over-
study site
NDVI images for sub-area
all distribution and cover characteristics that
analysis. However, applying the model to the high-resolu-
tion data provides a more precise definition of the variabil-
(approximately 6.8 km2) within the study area (around study plots P1–P3). Dark
by soil moisture, such as nutrient availability, thaw depth, soil would otherwise be difficult to visualize. Although these
areas represent regions of low NDVI (-1), while bright areas indicate with caution (i.e., IKONOS
ity in vegetation percent cover across the landscape.
values must be interpreted high IKONOS data therefore show tremendous potential for
aeration, redox potential, and pH (Oberbauer and Dawson, NDVI (+1), typically lush green vegetation. NDVI explains 72% of the vegetation cover variance for tundra vegetation mapping at local scales: they are able to
1992). Micro-scale moisture gradients (across a few metres), the study plots), they provide important preliminary re- delineate percent cover trends and microsite variability
sults. Stow et al. (1993) suggest that data with high spatial throughout the study area (Fig. 8). At the same time,
such as those in periglacial features (troughs to high-centre al., 1994; Walker et al., 1994; Murray, 1997; Henry,would strengthen NDVI correla-
resolution (i.e., < 10 m) 1998; similarly accurate estimates of percent cover can be de-
polygons, frost boils, or stone stripes) or from wet meadows Young et al., 1999). The spatial heterogeneity ofvariables, making it easier to i) identify
tions to biophysical the study rived from Landsat data at intermediate or regional scales.
to beach ridges, have great influence on the pattern and area, regarded as the irregularity of the physical environ-
53. 10 • G.J. LAIDLER et al.
8 • G.J. LAIDLER et al.
Source: Laidler, Treitz, and Atkinson,
120
100
80
Percent Cover
60
% Cover La NDVI
40 R2 = 0.78
% Cover Ik NDVI
R2 = 0.72
20
% Cover Su NDVI
R2 = 0.74
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70
NDVI
FIG. 6. Linear relations between percent cover and NDVI values for surface
(Su), IKONOS (Ik), and Landsat (La) sensors.
DISCUSSION FIG. 8. Image of percent cover from IKONOS NDVI, calculated from the following regression equation: Y = 275.5 (IK_NDVI) + 16.07 (R2 = 0.716, p < 0.01).
linear, significant, and consistent across scales (i.e., R2 = the initial conditions for patch-scale models by inventory-
Moisture and Percent Cover Field Estimates 0.72 – 0.78; p < 0.01) (Fig. 6), corroborating trends re- ing landscape conditions and their relative proportions; ii)
ported in other Arctic environments. stratify landscapes into relatively homogeneous response
Many possible transfer
An image of percent cover derived from the IKONOS units for spatially distributed modeling of material and
In the majority of Arctic locations, the environmental NDVI data is presented in Figure 8, which portrays the energy transport; iii) extrapolate model simulations by
factor most closely correlated with vegetation type is soil relationship between percent cover variations (Fig. 3), mapping areas that are potentially sensitive to particular
topographic trends (Table 2), and associated moisture disturbances; and iv) assess landscape- and regional-scale
moisture (Oberbauer and Dawson, 1992). In areas of high regimes (Fig. 2). Percent cover increases along declining model simulations by comparative spatial pattern analy-
elevation, water is a limiting factor and an important determi-
functions
elevations and slopes as a reflection of increased vegeta- ses. Here, coefficients of determination for NDVI and
tion canopy density in areas of high moisture (i.e., water- percent cover were very similar for the IKONOS and
nant of vegetation structure, productivity, and composition; tracks, drainage channels, and areas with moderate to Landsat data. This similarity is a function of averaging the
in lower areas, these aspects may not be controlled directly by minimal exposure). Modeling percent cover over the en- IKONOS NDVI data to the plot level for correlation
soil moisture, but rather by factors correlated with or affected FIG. 7. Landsat (top) and IKONOS (bottom)tirevegetationprovidesaan interesting perspective on over-
study site
NDVI images for sub-area
all distribution and cover characteristics that
analysis. However, applying the model to the high-resolu-
tion data provides a more precise definition of the variabil-
(approximately 6.8 km2) within the study area (around study plots P1–P3). Dark
by soil moisture, such as nutrient availability, thaw depth, soil would otherwise be difficult to visualize. Although these
areas represent regions of low NDVI (-1), while bright areas indicate with caution (i.e., IKONOS
ity in vegetation percent cover across the landscape.
values must be interpreted high IKONOS data therefore show tremendous potential for
aeration, redox potential, and pH (Oberbauer and Dawson, NDVI (+1), typically lush green vegetation. NDVI explains 72% of the vegetation cover variance for tundra vegetation mapping at local scales: they are able to
1992). Micro-scale moisture gradients (across a few metres), the study plots), they provide important preliminary re- delineate percent cover trends and microsite variability
sults. Stow et al. (1993) suggest that data with high spatial throughout the study area (Fig. 8). At the same time,
such as those in periglacial features (troughs to high-centre al., 1994; Walker et al., 1994; Murray, 1997; Henry,would strengthen NDVI correla-
resolution (i.e., < 10 m) 1998; similarly accurate estimates of percent cover can be de-
polygons, frost boils, or stone stripes) or from wet meadows Young et al., 1999). The spatial heterogeneity ofvariables, making it easier to i) identify
tions to biophysical the study rived from Landsat data at intermediate or regional scales.
to beach ridges, have great influence on the pattern and area, regarded as the irregularity of the physical environ-
57. ‣ Final products: maps
of veg. % cover and
fxn group
composition
‣ Indices with yellow
and red edge bands
most effective at
identifying moss-
dominated veg. types
58. 1524 M . W I L L I A M S et al.
‣ Final products: maps
of veg. % cover and
fxn group
composition
‣ Indices with yellow
and red edge bands
most effective at
identifying moss-
dominated veg. types
‣ Non-linearity may
pose difficulties in
scaling
Source: Williams et al.,
Fig. 3 Relationships between estimated LAI (using both Skye NDVI and LI-COR LAI-2000 observations at 0.2 m resolution,
averaged for upscaling) vs. Skye NDVI at different spatial scales. Exponential model equations, R2, and root-mean-square error are
Satellite data. The comparison of ground-based NDVI different to Skye NDVI measurements (Fig. 6
with Landsat NDVI revealed a highly significant linear satellite data showed a peak in frequency to
relationship (Fig. 8, r2 5 0.20, Po0.0001). The intercept the low end of the measurement range, whi
59. 1524 M . W I L L I A M S et al.
‣ Final products: maps
of veg. % cover and
fxn group
composition
‣ Indices with yellow
and red edge bands
most effective at
identifying moss-
dominated veg. types
‣ Non-linearity may
pose difficulties in
scaling
Source: Williams et al.,
Fig. 3 Relationships between estimated LAI (using both Skye NDVI and LI-COR LAI-2000 observations at 0.2 m resolution,
averaged for upscaling) vs. Skye NDVI at different spatial scales. Exponential model equations, R2, and root-mean-square error are
Satellite data. The comparison of ground-based NDVI different to Skye NDVI measurements (Fig. 6
with Landsat NDVI revealed a highly significant linear satellite data showed a peak in frequency to
relationship (Fig. 8, r2 5 0.20, Po0.0001). The intercept the low end of the measurement range, whi