- Trees in drier landscape positions with lower density, steeper slopes, and southerly aspects were found to be more susceptible to mountain pine beetle infestation at this study site in Montana. A mixing model using NDVI and RGI indices from remote sensing data accurately classified pixels into green, red, and grey stages of infestation. Red-stage pixels showing trees currently infected by beetles were disproportionately located in positions with lower moisture, higher insolation, and sparser vegetation compared to unaffected trees. This pattern may affect future watershed hydrology as the outbreak progresses.

1. Green- Stage
-0.2 0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
Frequency
NDVI
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
RGI
Red- Stage
Grey- Stage
Frequency
Quick Bird Range
0.4
Figure 4: Probability distributions of infection stage relationship to
vegetation indices used for the mixing model, end members shown as bars
Introduction
Watershed ecohydrology is a function of the intersection
of vegetation pattern and landscape structure. The hydrologic
implications of vegetation disturbance depend on the spatial extent
and pattern of change, and the location on this ecohydrologic template.
Here we investigate this intersection with a focus on a recent mountain
pine beetle epidemic that is increasingly affecting areas in the Rocky
Mountains (Figure 1).The mountain pine beetle (MPB) invades
Lodgepole pines by boring and tunneling through the bark into the
phloem where they feed and lay their eggs. In addition, a blue stain
fungus carried by the beetle then effectively girdles the tree. The visible
signs of mountain pine beetle attack occur the spring and summer
following the attack. In the first stage the tree is killed but remains
green, referred to as green-stage. It then fades to yellow then to red
in the red-stage, and eventually enters the grey-stage when all of the
needles have dropped.
Objectives
Determine which forest stands are most susceptible to beetle infestation and how these
infestation patterns are related to hydrologic, topographic, and forest ecosystem compositional
characteristics. Our efforts to monitor vegetation mortality across space and time provide a context
for assessing the drivers of mountain pine beetle infestation and how outbreak patterns may affect
watershed ecohydrology via energy, water, and biogeochemical cycles.
Preliminary Hypothesis
Trees in drier landscape positions will be more water stressed and therefore more
susceptible to beetle infestation.
Prediction
Red-stage pixels will be found disproportionately in drier landscape positions.
Figure 1: Map of MPB
infested areas in the Rocky
Mountain west
500 600 700 800 900
0
0.2
0.4
0.6
0.8
1
Reflectance
Red-stage Signature
Wavelength (nm)
500 600 700 800 900
0.5
0.6
0.7
0.8
0.9
1
Reflectance
Grey-stage Signature
Wavelength (nm)
Frequency
0 10 20 30
0
0.25
0.5
0.75
1
Slope
5 10 15 20
0.25
0.5
0.75
1
Vegetation Height (m)
Frequency
0
0 0.25 0.5 0.75 1
0.25
0.5
0.75
1
Forest Density
Frequency
Red-StageGrey-Stage
Green-Stage
N
500 600 700 800 900
0
0.2
0.4
0.6
0.8
1
Reflectance
Green-stage Signature
Wavelength (nm)
SE N
N SE
RGI
NDVI
-0.1
-0.2 1
0.6
Green-­‐Red
Grey-­‐Red
Green-­‐Grey
Forest
Density
0.24
0.28
0.07
Aspect
0.19
0.27
0.08
Insolation
0.17
0.24
0.07
Slope
0.16
0.18
0.04
Elevation
0.14
0.12
0.09
TWI
0.13
0.10
0.04
Vegetation
Height
0.07
0.09
0.03
UAA
0.05
0.03
0.03
Insolation (kwh/year)
Red-stage
Infestation patterns develop across scales
Hillslope scale Plot scale
N
0.70 fR
0
2220 m
2140
0.70
0
0.75
0
0.54
0
0.70 fR
0
Frequency
6 8 10 12
0
0.25
0.5
0.75
1
TWI
Frequency
N E S W N
0
0.25
0.5
0.75
1
Aspect
Frequency
3000 3200 3400 3600
0
0.25
0.5
0.75
1
Insolation (kwh/year)
Ecohydrology of an outbreak: Impacts of vegetation pattern and landscape
structure on Mountain Pine Beetle disturbance
Kendra E. Kaiser1
, Brian L. McGlynn1
, Ryan E. Emanuel2
, Fabian Nippgen1
, John Mallard1
1
Montana State University,Watershed Hydrology Lab, Bozeman, Montana, 2
North Carolina State University
Acknowledgements
Special thanks to Galen Laird, Rebecca Kurnick, and Anna Bergstrom for
help with field work, and everyone in the watershed hydrology lab group for help
throughout the project. Undergraduate Scholars Program, Montana EPSCoR,
the College of Agriculture and the Land Resources Department. NSF award DEB-
1110742 for funding the hand-held spectrometer and EAR-0838193 for the Quick
Bird, NSF grant EAR 0837937, NSF grant EAR 0404130 NSF grant EAR 0943640.
Landscape Analysis andVegetation Stage Assessment
• Green and grey stage pixels showed similar cumulative distribution functions ( CDFs) indicative of the
spatial heterogeneity expected in an intact forest
• Red-stage pixels were concentrated in landscape positions distinct from the distributions of green-
stage and grey-stage pixels.
•These CDFs indicated that red-stage trees were preferentially located in less dense and drier landscape
positions (lower UAA, lowerTWI, steeper slopes, southerly aspects, higher insolation).
Mixing Model
NDVI and RGI were used as the two indices for a three-component mixing model.The three end members were
determined by the probability distribution functions of the stages of infestation and their correlation with the two
vegetation indices. This model was then applied to the Quick Bird Image returning the green-stage fraction fG
, red-
stage fraction (fR
), and grey-stage fraction (fD
) for each 1m2
pixel (Figure 5).
fG
+ fR
+fD
= 1
NDVIQB
= fG
(NDVIG
) + fR
(NDVIR
) + fD
(NDVID
)
RGIQB
= fG
(RGIG
) + fR
(RGIR
) + fD
(RGID
)
Discussion and Summary
Preliminary Hypothesis: Trees in drier landscape positions will be more
water stressed and therefore more susceptible to beetle infestation.
• Red-stage pixels had distinct landscape positions inTCEF
• Infestation is likely at an endemic (low) population level where beetles
mainly attack weakened trees
• AtTCEF trees in locations with lower vegetation density, steeper slopes,
southerly aspects, lower UAA, lowerTWI, lower elevation, and higher
insolation were more frequently impacted by the recent MPB infestation
than trees in other locations.
• However, infestation is a reflection of beetle outbreak dynamics which
could mask specific landscape metrics that describe tree susceptibility.
Future work:
• Monitor infestation atTCEF over time
• As the infestation progresses we expect evolving patterns of infestation
based on beetle population size:
• Steady population size: infestation frequency will increase in locations
with stressed trees
• Incipient Epidemic population: Beetles have the ability to mass attack
and overcome tree defences, thereby infesting healthier, taller trees
• Epidemic population: red-stage trees could be found more uniformly
across the landscape
Field Methods
Over the course of the fall of 2010, 700
samples of vegetation were recorded with a handheld
spectrometer (Apogee Instruments Model PS100, Logan, UT).
The average spectral signals for green and red Lodgepole pine
needles and defoliated branches were determined and used as
a proxy for MPB infestation stage (Figure 3). Green-stage (G)
indicates a healthy tree, Red-stage (R) an infected tree, and
Grey-stage (D) being a dead tree.
Figure 3: Average spectral signals for infestation stages shown with 25th and 75th percentiles
Distribution of Red-stage and Insolation
Site Description
TCEF was established in 1961 and is
located in the Lewis and Clark National Forest
in the Little Belt Mountains. It is in Meagher
County, Montana, about 64 km north of
White Sulfur Springs.TCEF is a 3693 ha set
of watersheds with elevation ranging from
2,179 to 2,286m. Lodgepole pines are the
dominant tree species in the watershed with
some subalpine fir and whitebark pine in the
higher elevations and Engelmann spruce in
the wetlands. Prior to 2009 the Lewis and
Clark National Forest had remained relatively
un-infested.
Figure 2: Quick Bird Image ofTCEF with sample collection
sites, map area ~22 km2
Figure 7: a) Output of mixing model showing individual fractions of red, green and grey respectively b) False-color
composite map of green-stage, red-stage, and grey-stage fractions showing spatial results of the mixing model
Figure 5: Mixing space
Figure 8: The following cumulative distribution functions show the landscape variables with the strongest
differentiation between red-stage and green or grey-stage.
6a)
• Insolation is an estimation of incoming solar radiation
• Higher insolation indicates higher potential evaporation which
can lead to lower soil moisture
•This pattern of disturbance can affect future water quantity and
quantity and biogeochemical cycling
Vegetation Indices
Various vegetation indices were calculated for the QuickBird image ofTCEF. The average
spectral signals of the three stages of infestation were used to determine which vegetation indices
would best separate the infection statuses for the experimental forest (Figure 4). The Normalized
DifferenceVegetation Index (NDVI) and the Red-Green Index (RGI) were found to best characterize
infection status.
6c)6b)
Statistical Analysis
Two-sample Kolmogorov-Smirnov (KS) tests were used to determine
how different the distributions of the three stages of infestation were.
Each distribution was significantly different than the others, bolded values
indicate metrics with the largest differences between infestation stages.
Map area ~22 km2
Higher frequency of infestation in locations with lower
topographic wetness index (TWI: ln [(A)/ tan(θ)])
Higher frequency of infestation in locations with
higher insolation
Higher infestation in Low density landscape positions
Higher frequency of infestation on steeper slopes
Higher frequency of infestation on slopes with SE aspect
fR
fG
fD
3576
2192
Figure 6: Inset from 7b: a) QuickBird classified fraction red-stage map b) LiDAR digital elevation model with vegetation c) Fraction red-stage overlaid on
DEM d) zoomed in example showing high resolution classification
7a)
Higher frequency of infestation in smaller trees
6d)
Landscape scale
7b)