Roberts, Gareth James; (2003) Monitoring land cover dynamics using linear kernel-driven BRDF model parameter temporal trajectories. Doctoral thesis (Ph.D), UCL (University College London).

Text Monitoring_land_cover_dynamics.pdf Download (81MB)

Abstract

Acquiring information on land cover dynamics using remotely sensed data has typically involved vegetation indices (Vis). Such methods, however, do not account for or utilise angular reflectance variations caused by surface 3-D geometric structure. Surface reflectance anisotropy is a function of viewing and illumination angles, and can be a source of error in WFOV or multi-temporal data. In addition to normalising reflectance, parametric BRDF models provide parameters describing the surface anisotropy which are used to derive estimates of albedo. Validation of the MOD43 product is carried out over sites in southern Africa using MISR and AirMISR data. Results indicate discrepancies exist between the MISR / AirMISR and MOD43 model retrievals on a per pixel basis. However, the MOD43 and MISR / AirMISR data are consistent over a wide spatial extent. A Kalman filter methodology is developed to derive multi-temporal model parameter estimates. Model parameter temporal trajectories are comparable to those of LS fitting although are 'noisier' in appearance. This results from the method applied to propagate the model parameters and associated uncertainty over time. Furthermore, LS inversion provides lower RMSE estimates. The temporal stability of the linear BRDF model retrievals is assessed. Analysis of the model parameter temporal stability is carried out in relation to SVIs using AVHRR data acquired over the UK and southern Africa. Results indicate that the model parameters are less stable than the SVIs and the observed reflectance. An investigation of the model parameter trajectories, in relation to the surface dynamics and their ability to derive biophysical parameter estimates is carried out over southern Africa. Results using AVHRR data are inconclusive due to uncertainties in the data. MODIS model parameter trajectories evolve in accordance with the surface phenology. Biophysical parameter estimates between methods are similar in trajectory although different in magnitude.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as XMLJSONCSV

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item