Final Results: ALOS PALSAR Based Estimation of Above Ground Woody Biomass in African Savanna Woodlands
Paradzayi, Charles1; Annegarn, H. J.2; Schmullius, C.3
1Midlands State University, ZIMBABWE; 2UJ, SOUTH AFRICA; 3Friedrich Schiller University, GERMANY
Estimations of available fuelwood resources in communal savanna woodlands are widely based on conventional terrestrial and optical remote sensing approaches, which are constrained by limited geographic footprints and the use of leaf area indices and normalised difference vegetation indices, as surrogates for above ground biomass. As a result, reliable information about the location and estimated quantities of available woody biomass is scarce at local, national and global scales. Recent developments have shown that classification of backscatter information contained in full polarimetric radar retrievals from satellite borne sensors can discriminate between woody and non-woody vegetation. The intensity of the backscattered signal has been shown to be sensitive to above ground biomass density. However, no such studies have been reported across African savanna woodlands. This paper presents the final results of a study which used full polarimetric ALOS PALSAR retrievals to map and quantify fuelwood resources in communal savanna woodland in Welverdiend village, South Africa. Unsupervised entropy/alpha angle Wishart classification and maximum likelihood classification procedures are used to characterise the scattering classes from the ALOS PALSAR retrievals into eight major terrain scattering mechanisms. Five vegetation classes (random anisotropic, forest double bounce, vegetation, dihedral and dipole) are identified that are closely related to backscattering from woody vegetation components. Correlations between backscatter intensities acquired under dry and wet conditions with above-ground biomass densities estimated from field surveys are investigated to derive equations for predicting biomass densities. The regression analysis supports findings of similar studies where the HV backscatter intensity showed moderately strong relationship (R square >0.6) with above ground biomass densities. The inverted regression equations were used to estimate the biomass densities for areas covered with woody vegetation. Knowledge about the location and distribution of woody biomass has significant implications for fuelwood management and carbon sequestration initiatives. A combination of woodland management interventions, coupled with the transition to modern energy sources, has the potential of turning communal woodlands into carbon sinks.