five years of land surface phenology in an arctic landscape
TRANSCRIPT
Five Years of Land Cover Reflectance in a Large
Scale Hydrological Manipulation in
an Arctic Tundra Landscape
1Santonu Goswami2John A. Gamon
1Craig E. Tweedie
1Environmental Science and Engineering
University of Texas at El Paso2Dept. of Earth and Atmospheric Sciences
University of Alberta
-
.
Importance of Surface Hydrology (Soil
Moisture) in ecosystem structure and function
• Many observed and modeled change responses of
arctic terrestrial ecosystems are related to surface
hydrology:
– Surface energy budgets (Chapin et al. 2005, Euskircken et
al. 2007)
– Land-atmosphere carbon exchange (Merbold et. al 2009,
Wolf et al. 2008)
– Geomorphic processes (Lawrence and Slater 2005,
McNamara and Kane 2009)
– Provision of ecosystem goods and services (Milennium
Ecosystem Assessment)
– Plant phenology and response to warming (Arft et al. 1999,
Walker et al. 2006)
Plants as an indicator of change
• Detecting biotic responses to a changing environment is essential for understanding the consequences of global change.
• Plants can work as effective indicators of• changing conditions and, depending on the nature of the
change, change,
• respond by increasing or decreasing amounts of green-leaf biomass, chlorophyll, and water content.
• important effects on ecosystem processes such as NPP and nutrient cycling.
• Plant phenology is detectable using remote sensing technologies and therefore scaleable over space and time (e.g. NDVI).
Study Site - Biocomplexity Experiment,
Barrow, Alaska
Experimental Setup: Biocomplexity
Experiment
• Large-scale Hydrological Manipulation experiment to study the effect of varying Soil Moisture on ecosystem Carbon Balance in the Arctic Flooded
Drained
Controlled
Tramlines (300m long)
Barrow (71º19'N 156º37'W)
Atqasuk
100km
Brooks Range
Coastal Plain
Dikes
Eddy covariance towers
The Tram System
The Tram System
The Tram System
Near surface remote sensing technology
• Measurements are made at every meter along each of the
three 300 meter tramlines (precise and repeatable)
• Collects hyperspectral data in the vis-nir region and digital
images of vegetation
• More than 200,000 data files collected in 5 years
• Has the ability to carry various sensors packages
• Can be used to study change in phenology, sun angle
The Tram System
effect on vegetation etc
• Standardized sampling design is part of a larger effort
(SpecNet) to compare ecosystem responses using a
combination of spectral reflectance and flux
measurements
Tramline system in the biocomplexity experiment is the
largest ground based spectral data monitoring system
that we are aware of internationally
Measurement of Reflectance
• Reflectance
= ITarget / IDownwelling
= Radiance/Irradiance
• Reflectance correction
= (RTarget / RDownwelling) X (RDownwelling/ RPanel)= (RTarget / RDownwelling) X (RDownwelling/ RPanel)
• NDVI (Normalized Difference
Vegetation Index) calculation
= (R800-R680)/ (R800+R680)
• Also measures water table depth, thaw
depth.
Questions being asked in this study
• How does land-surface phenology (i.e. surface
reflectance properties) change inter-annually?
• Is there any detectable effect of experimental • Is there any detectable effect of experimental
treatment (flooding and draining) on surface
reflectance?
Transition Throughout the Seasonavg corrected reflectance south 070613
0
0.5
1
350 450 550 650 750 850 950 1050
wavelength
refl
ec
tan
ce
avg corrected reflectance south 070618
0.2
1st Week of June H2O Band
0
0.1
350 550 750 950
wavelength
refl
ec
tan
ce
0
0.1
0.2
0.3
350 550 750 950
Re
fle
cta
nce
Wavelength
Corrected reflected south 080806
3rd Week of June
1st Week of August
Chl Feature
Chl Feature
Inter-annual variability of NDVI in the treatment areas for
pre-treatment years (2005, 2006, 2007) and treatment years
(2008 and 2009)
0.50
0.60
0.70Peak growing time
0.50
0.60
0.70
Peak growing time
-0.20
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
150 170 190 210 230 250
ND
VI
Day of Year
NDVI North 2005 NDVI North 2006 NDVI North 2007
NDVI North 2008 NDVI North 2009
Snow-melt
Heavy flooding event
-0.20
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
150 170 190 210 230 250
ND
VI
Day of Year
NDVI Central 2005 NDVI Central 2006 NDVI Central 2007
NDVI Central 2008 NDVI Central 2009
Snow-melt
Snow fall
FLOODED DRAINED
Flooding treatment effect
Comparison of inter-
annual variability of
NDVI to water table
-20
-15
-10
-5
0
5
10
15
20
25
160 170 180 190 200 210 220 230
Wa
ter
Tab
le D
ep
th (
cm)
Day of Year
-30
-15
0
Th
aw
De
pth
(cm
)
Flooding treatment
Increased thaw in 2009
2009
20082007
Flooded
NDVI to water table
depth and thaw depth
in the flooded section
for 2007 (pre-
treatment) and 2008
and 2009 (treatment
years)
-45
-30
160 170 180 190 200 210 220 230
Th
aw
De
pth
(cm
)
Day of Year
-0.20
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
160 170 180 190 200 210 220 230
ND
VI
Day of Year
Flooding treatment in 2008
Snow-melt
Heavy flooding event
-20
-10
0
10
20
30
160 170 180 190 200 210 220 230
Wa
ter
Tab
le D
ep
th (
cm)
-30
-15
0
Th
aw
De
pth
(cm
)
Comparison of inter-
annual variability of
2009
20082007
Drained
-45
160 170 180 190 200 210 220 230
Th
aw
De
pth
(cm
)
-0.20
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
160 170 180 190 200 210 220 230
ND
VI
Day of Year
annual variability of
NDVI to water table
depth and thaw depth
in the drained section
for 2007 (pre-
treatment) and 2008
and 2009 (treatment
years)
Snow-melt
Snowfall
Comparison of inter-
annual variability of-30
-15
0
Th
aw
De
pth
(cm
)
-20
-10
0
10
20
30
160 170 180 190 200 210 220 230
Wa
ter
Tab
le D
ep
th (
cm)
Dry year 2007
2009
20082007
Control
annual variability of
NDVI to water table
depth and thaw depth
in the control section for
2007 (pre-treatment)
and 2008 and 2009
(treatment years)
-45
-30
160 170 180 190 200 210 220 230
Th
aw
De
pth
(cm
)
-0.20
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
160 170 180 190 200 210 220 230
ND
VI
Day of Year
Snowfall
Snow-melt
0.430.40
0.33
0.00
0.10
0.20
0.30
0.40
0.50
11% surface
water
28% surface
water
83% surface
water
ND
VI
Standing surface water
has a significant effect on
reflectance spectra.
This has profound effect
on the NDVI values.
The spectral range (400-1000nm) offers the ability to detect
several things, including water. Using a spectral index (NDSWI)
capable of capturing multi-dimensional surface hydrological
dynamics can improve the sampling capabilities of this system
NDSWI (Normalized
Difference Surface
Water Index) =Low
Quickbird 2002 Quickbird 2008 Goswami et. al – submitted, JGR
Biogeosciences
Water Index) =
(R490-R1000)
(R490+R1000)
High
Summary and Conclusion
• Changes in vegetation cover during the growing season was
clearly detectable using surface reflectance.
• Seasonal patterns of the NDVI values showed some
differences with flooding. The flooded section showed bigger
differences than the drained section.
• The seasonal patterns in NDVI was least different for the • The seasonal patterns in NDVI was least different for the
control section among the years.
• NDVI values were compromised by varying water cover
(among other things), so simultaneously tracking NDVI and
NDSWI might be a better approach in this landscape.
• Alternatively, other methods like Spectral Mixture Analysis
can probably help here.
Acknowledgements
• National Science Foundation (ASSP-0421588).
• Barrow Arctic Science Consortium.
• CH2M Hill Polar Field Services
• Ukpeagvik Inupiat Corporation – landholder
• Spectral Network. • Spectral Network.
• UTEP Cybershare Center of Excellence.
• Sergio Vargas, Perry Houser, Christian Andresen, Adrian
Aguirre, Sandra Villarreal, Ryan Cody, David Lin.
• Systems Ecology Laboratory, University of Texas at El Paso.