services Anton Van Rompaey 1 , Veerle Vanacker 2 , Vincent Balthazar - - PowerPoint PPT Presentation
Forest transition in mountain areas: topographic correction, large scale mapping and modeling of ecosystem services Anton Van Rompaey 1 , Veerle Vanacker 2 , Vincent Balthazar 2 , Eric Lambin 2 , Jaclyn Hall 2 , Patrick Hostert 3 , Patrick
Belgian Earth Observation Day 2012 September 5 – 2012 - Bruges Anton Van Rompaey1, Veerle Vanacker2, Vincent Balthazar2, Eric Lambin2, Jaclyn Hall2, Patrick Hostert3, Patrick Griffiths3, Steven Vanonckelen1, Armando Molina1, Derek Bruggeman2, Marie Guns2
1: Geography Research Group, Dep. Earth and Environmental Sciences, KULeuven 2: Department of Geography, Earth and Life Institute , UCLouvain 3: Geographisches Institut, Humbolt-Universität zu Berlin
Forest transition Mountain areas Topographic correction Large scale mapping Ecosystem service assessment
Transition from a phase of net deforestation to a phase of
For a region: pathway through time driven by Economic development -> off-farm jobs -> land abandonment
Land degradation lowering of yields -> land abandonment
Mather, 1995: FT in France Regional patterns of FT
Forest transitions typically occur in mountain areas
Drome (SE-France):
1985
Remote sensing seems privileged
Transition are of subtle: gradual
Typically patchy landscape with
Understanding of mechanisms
Mountain environment:
Impact on ecosystem services is
To what extent is it possible to detect
Atmospheric and topographic correction
Most complex procedure not always
Development of automated
Monitoring ecosystem services in FT-
3 study sites Romanian Carpathians Buthan Himalayas Ecuadorian Andes
No Corr Band ratioing Cosine corr C-corr Minnaert corr No Corr Dark object subtraction Physical transfer function
Ls,λ is the uncorrected radiance value of the image and Lp represents the minimum radiance value of the image, calculated as the 1th percentile. Tr,λ is the Rayleigh scattering transmittance function, including the sea-level atmospheric pressure (P0; in mbar), the ambient atmospheric pressure (P; in mbar) and the band wavelength (λ). M is the relative air mass and θs is the solar zenith angle (in degrees). Tw,λ is the water-vapor transmittance function, calculated with the precipitable water vapor (W; in cm), relative air mass (M) and water-vapor absorption coefficients (aw).
θs is the solar zenith angle and β is the incident solar angle, cos β = cos θs cos θn + sin θs sin θn cos (ϕt – ϕa), θn is the slope angle of the terrain and k,λ is the slope of the regression between x = log(cos θn cos β) and y = log(ρT,λ cos θn). Parameter Cλ is the quotient of intercept (bλ) and slope (mλ) of the regression line between x and y, the h-factor represents a topographic parameter derived from the SRTM (h = 1-θn/π) and the h0-factor an empirical parameter derived from the regression line between reflectance and cos β (h0 = (π +2θs)/2π).
Efficient methods reduce the reflectance values between
(a) no AC or TC; (b) DOS without TC; (c) DOS with band ratio; (d) TF with cosine; (e) TF with PBM; (f) TF with PBC.
Reflectance (%)
Band 4
(c) cos β (a) (e) (f) (b) (d) cos β Reflectance (%) no AC or TC DOSwithout TC DOS with band ratio TF with cosine TF with PBM TF with PBC Reflectance (%) Reflectance (%) Reflectance (%)
y = -8.7x + 40.7 R² = 0.0282 5 10 15 20 25 30 35 40 45 50 0.0 0.2 0.4 0.6 0.8 1.0 P = 2.67E-11 y = 1.3x + 37.4 R² = 0.0003 5 10 15 20 25 30 35 40 45 50 0.0 0.2 0.4 0.6 0.8 1.0 P = 0.465 y = 14.639x + 16.12 R² = 0.0813 5 10 15 20 25 30 35 40 45 50 0.0 0.2 0.4 0.6 0.8 1.0 P = 1.78E-37 y = 12.6x + 19.4 R² = 0.0938 5 10 15 20 25 30 35 40 45 50 0.0 0.2 0.4 0.6 0.8 1.0 P = 5.08E-42 y = -2.7x + 19.6 R² = 0.0055 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.0 0.2 0.4 0.6 0.8 1.0 P = 0.001 y = 0.7x + 17.7 R² = 0.0003 5 10 15 20 25 30 35 40 45 50 0.0 0.2 0.4 0.6 0.8 1.0 P = 0.483 y = -2.7x + 19.6 R² = 0.0055 5 10 15 20 25 30 35 40 45 0.0 0.2 0.4 0.6 0.8 1.0 P = 0.001
Reflectance (%)
Efficient preprocessing results in better land cover
(a) (b) (c)
(a) no AC or TC; (b) TF with cosine; (c) TF with PBC
Efficient preprocessing results in better land cover
40 45 50 55 60 65 70 75 80 85 90 No TC Band ratio Cosine PBC PBM No TC Band ratio Cosine PBC PBM No TC Band ratio Cosine PBC PBM No AC DOS TF
0.05 0.1 0.15 0.2 0.25 0.3 0.35 BS BL CF MX GRASS WT NoTC_DOS NoTC_TF Bandratio_NoAC Bandratio_DOS Bandratio_TF Cosine_NoAC Cosine_DOS Cosine_TF PBC_NoAC PBC_NoAC PBC_TF
1.
16
Use:
Decision rule set: select best observation per pixel
composite image
17 17
18 18
Composite image (RGB=4/5/3) Score image (low=blue, high = red) Flag image (RGB = PathRow/Year/DOY) Mean image (RGB=4/5/3) Clear observation count (low=blue, high=red) STDV image (RGB=7/5/4)
19
20
Natural hazard regulation (mass movements) Water and erosion regulation (agricultural production,
Carbon storage and biodiversity (REDD/REDD+)
LINK WITH CHANGES IN LANDSLIDE
LINK WITH CHANGES IN DISCHARGE REGIMES LINK WITH CHANGES IN CARBON
Objective of creating landslide inventories for
Selection and validation of the method for a
Application of shape and spectral
Combination with morphometric
Statistical relation between landslide
LANDSAT ETM+: RGB 453 (30 m) LANDSAT ETM+: RGB 453 (15 m PAN-fusion) ASTER VNIR + SWIR: RGB (PCA1,PCA2, PCA3) LANDSAT ETM+ (P-S): RGB (PCA1, PCA2, PCA3)
Monthly rainfall (mm)
Monthly rainfall data Monthly streamflow data
Long-term trends corrected for intra and interannual varation Land Use
a. c. b. d.
a. Northern Costa Rica
Tempisque watershed
5414 km2
sea level - 1964 m
Sub-landscape 25 km2
b. Northern Vietnam
91,600 km2
~100 - 3100 m
Sub-landscape 400 km2
c. Valdivia Chile
11,775 km2
Sea level - 1222 m
Sub-landscape 25 km2
d. Highland Ecuador
Pangor watershed
282 km2
ranging from 1431 to 4333 m
Sub-landscape 2.25 km2
– Natural forest regeneration
– Natural forest regeneration
– Natural forest to plantation
–
– Páramo to plantation
Integration and automation of total process chain Preprocessing large scale mapping of LU and LUC
Validation at different levels Evaluation of added value of procedures and