Remote Sensing of Savannah Vegetation Structure Using the Geoscience Laser Altimeter System and L-Band Synthetic Aperture Radar

Savannahs are globally important but not well understood systems. They consist of the coexistence of trees and grasses. A major challenge in savannah ecosystems studies is the need of maps of vegetation structure over large areas where the field data collection is impractical and time consuming. Active remote sensing such as spaceborne LiDAR and RADAR has experienced limited use in these complex heterogeneously vegetated systems. This thesis examines the ability of spaceborne ICESat-GLAS data to retrieve vegetation parameters from these more structurally and sparsely complex vegetated ecosystems. For this main purpose, two methods of retrieving savannah vegetation heights from GLAS data were explored based on Gaussian decomposition. Results showed that the direct method works well over flat areas (R=0.63 and RMSE=1.32m; R=0.68and RMSE=2.61m; n=12) for GLA01 and GLA14 respectively, while sloped areas need statistical methods to remove the effect of terrain slopes on the waveform extent for better estimation of maximum vegetation height (R²=0.78 and RMSE =14.5m; R²=0.67 and RMSE=17.5m; n=6) for using terrain index and waveform width as a terrain correction factor in regression models. This work has compared the estimation of vegetation parameters derived from airborne LiDAR data and field measured data with results from spaceborne GLAS LiDAR data. GLAS estimated stand density produced R=0.98 with those estimated from airborne LiDAR data. Moreover, GLAS ratio which related to the amount of woody cover in each footprint shows a strong relationship with those extracted from LiDAR airborne data (R=0.93) and also presents a good correlation (R=0.85) when compared with woody cover field observations. In addition, ALOS PALSAR RADAR data was used to evaluate the results of estimating woody cover from GLAS LiDAR waveform parameters in another site at the Kruger National Park. This comparison showed a significant correlation between GLAS estimated woody cover and ALOS L-band HH backscatter (R=0.59). This work provides an important source of knowledge for the South Africa National Parks (SANParks) Authority with a powerful tool for an adequate spatially-consistent monitoring and assessment of the woody vegetation structure of savannah landscapes.