Land cover change in national parks of the western...
Transcript of Land cover change in national parks of the western...
Land cover change in national parks of
the western U.S.
Robert E. Kennedy1, Zhiqiang Yang1, Justin Braaten1,
Peder Nelson1, Warren B. Cohen2, Eric Pfaff1
Landsat Science Team Meeting, January 19-21 2010
1 Department of Forest Ecosystems and Society,
Oregon State University2 USDA Forest Service Pacific Northwest Research Station
Kennedy et al. LCLUC 2010 April2
Take home messages
• To detect potential impacts of climate change
on vegetation we need to:
– Detect landcover changes that are both abrupt (fire,
etc.) and slow (drought, insects, encroachment)
– Distinguish between anthropogenic and naturalprocesses
– Use natural weather experiments at large scales to
improve inference based on robust hypotheses
• The Landsat archive can help in all three areas
Kennedy et al. LCLUC 2010 April3
Themes
• Abrupt and slow processes
• Attribution
• Climate cycles
Kennedy et al. LCLUC 2010 April4
Themes
• Abrupt and slow processes
• Attribution
• Climate cycles
Kennedy et al. LCLUC 2010 April5
Detecting abrupt and slow processes
Condition change
Time
Urban
“succession”
State Change Cyclical Change
Kennedy et al. LCLUC 2010 April6
Arbupt process: Harvest, fire
Project: Region 6
Effectiveness
Monitoring Program
for the Northwest
Forest Plan (NWFP)
Data: > 500
individual Landsat
scenes
TimeSync Interpretation
ongoing
Kennedy et al. LCLUC 2010 April7
Disturbance mapping: Time and magnitude
Year of disturbance Disturbance MagnitudeProject: Quantify trends in
harvest within Coastal Coho
ecologically-significant unit
Year of disturbance Disturbance Magnitude
LandTrendr
15713Not
Disturbed
1589Disturbed
Not
DisturbedDisturbed
Tim
eS
yn
c
Kennedy et al. LCLUC 2010 April8
Abrupt disturbance: Subtle and Not So Subtle
Harvest High
Harvest Low
Prior Harvest
Kennedy et al. LCLUC 2010 April9
Zion National Park
Bryce Canyon
National ParkFires
Elevation
Landscape dynamics on the Colorado Plateau
Disturbance associated
with the 2002 drought
Ongoing chronic
mortalityGrowth/Encroachment
Landsat captures a wide range of
disturbance and growth phenomena --
a map of the landscape as the
dynamic system we know it to be
Kennedy et al. LCLUC 2010 April10
Themes
• Abrupt and slow processes
• Attribution
• Climate cycles
Kennedy et al. LCLUC 2010 April11
Insect
Clearing
Conversion
Fire
Drought
Wind/Water
Snow
Natural vs. anthropogenic
Classifier
Patch size, shape
Landscape location
Pre-, post- condition
Sequence, duration
of change
Archive allows better or new information
Kennedy et al. LCLUC 2010 April12
Sequence, duration of change
a)
b)
c)
d)
e)
Change Labels:
A map of the
dynamic
landscape
The sequences
of change may
provide clues
about agent
Kennedy et al. LCLUC 2010 April13
First stabs at attribution
0
5
10
15
20
25
Protected
lands
Unprotected
lands
North Cascades Province: Percent
Area Disturbed 1985-2008
"Abrupt"
Slow mortality
Fire
Abrupt, but presumably
natural processes
Separate by duration and
known fire occurrence
Kennedy et al. LCLUC 2010 April14
We can bring more to bear…
High
magnitude
High
magnitude
Abrupt Slow
Use patch metrics and
landscape position
Add change labels
Kennedy et al. LCLUC 2010 April15
Insect
Clearing
Conversion
Fire
Drought
Wind/Water
Snow
Natural vs. anthropogenic
Classifier
Patch size, shape
Landscape location
Pre-, post- condition
Sequence, duration
of change
Archive allows better or new information
Kennedy et al. LCLUC 2010 April1616
NB
RB
rig
htn
ess
Gre
enn
ess
Wetn
es
s
LandTrendr: Temporal fitting
• Use segmentation of
one band to identify
“vertices” in time series
• Smooth between
vertices in other bands
• Result: “Pseudo-
images” with year-to-
year noise removed,
but actual change
retained
Kennedy et al. LCLUC 2010 April17
QuickTime™ and aYUV420 codec decompressor
are needed to see this picture.
Temporally consistent “pseudoimages”
• Consistent
spectral space
minimizes effects
of noise,
phenology,
atmosphere, etc.
• Allows
development of
yearly landcover
maps
See “Features” block at landtrendr.forestry.oregonstate.edu
for more fun movies
Kennedy et al. LCLUC 2010 April18
Yearly classification: Fire effects
1999 2002 2006
Area of detail Loss of vegetative cover
PartialFull
Canopy removal
Track fire effects using class labels familiar to users
Kennedy et al. LCLUC 2010 April19
Yearly classification: Fire effects1994
1997
2007
a)
b)
c)
d)
e)
f)
Kennedy et al. LCLUC 2010 April20
Themes
• Abrupt and slow processes
• Attribution
• Climate cycles
Kennedy et al. LCLUC 2010 April21
Climate variability
• Observing potential effects relative to cyclic
weather years may improve inference
Maps show departure of PRISM data relative to 1978-2009 record
1995 2001
Warmer
Cooler
National parks and
wilderness areas
Kennedy et al. LCLUC 2010 April22
Eventual goal: Relate attribution to climate series
-250
-200
-150
-100
-50
0
50
100
150
200
250
300
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1
10
100
1000
10000
tmin_1q
tmin_4q
Total Slow Mortality
North Cascades Ecoregion
Examine in terms of both overall trends and
year over year anomalies, compare to
Sierras, desert Southwest, etc.
Kennedy et al. LCLUC 2010 April23
Take home messages
• To detect potential impacts of climate change
on vegetation we need to:
– Detect landcover changes that are both abrupt (fire,
etc.) and slow (drought, insects, encroachment)
– Distinguish between anthropogenic and naturalprocesses
– Use natural weather experiments at large scales to
improve inference based on robust hypotheses
• The Landsat archive can help in all three areas
24
Thank you.