Investigating relationships between Landsat TM/ETM+ surface reflectance data and stand-specific forest inventory parameters using empirical regression and forest radiative transfer models.
Khaldoun
Rishmawi, Department of Geography, University of Maryland, College Park, MD, rishmawi@umd.edu
(Presenting)
Nancy
Thomas, Department of Geography, University of Maryland, College Park, MD, nthomas1@umd.edu
Samuel
N
Goward, Department of Geography, University of Maryland, College Park, MD, sgoward@umd.edu
Accurate information on the rates of change of forest characteristics after disturbance is essential for understanding North American forest carbon dynamics. Landsat TM/ETM+ imagery accumulated since the year 1984 provide a unique data source for estimating forest characteristics. Through a NASA funded project – “North American Forest Disturbance and Regrowth since 1972 (NAFD)”, Landsat time series stacks (LTSS) have been assembled for 29 locations selected to represent United States forest types. Forest inventory parameters for these locations, like tree height (m), breast-height trunk diameter (DBH) (cm), stand density (trees m-2 ), and species composition were obtained from the USDA Forest Inventory and Analysis (FIA) database.
Surface reflectance values were derived for the 29 stacks by the Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS) which applies an atmospheric correction scheme based on the 6S radiative transfer code. The effects of variations in solar zenith angle (SZA) between Landsat acquisition dates on top of the canopy Landsat reflectance values were investigated and found to be significant. As the anisotropic properties of forests would introduce “noise” into empirical relationships between forest inventory parameters and top of canopy reflectance, an absolute-normalization approach was applied in an attempt to normalize for these effects. The absolute-normalization approach matches images in a time-series to an atmospherically corrected reference image using pseudo-invariant features and reduced major axis (RMA) regression.
This poster reports on the results of utilizing linear and surface fitting regression models between forest inventory parameters in early forest succession and normalized Landsat spectral values. Alternatively, top of canopy reflectance values will be modeled using a geometric–optical canopy reflectance model (GEOSAIL) and a three-dimensional forest light interaction model (FLITE). Model inputs include FIA measurements and other ancillary data estimated from published allometric models (e.g. foliage mass and crown radius regression models). Modeled values will be evaluated against Landsat surface reflectance value. Goodness of fit (r^2) and root mean squared error (RMSE) values calculated for the empirical regression method and for the forest radiative transfer models will be reported.
NASA Carbon Cycle & Ecosystems Active Awards Represented by this Poster: