Evaluation of compositing algorithms over the Brazilian Amazon using SPOT-4 VEGETATION data

摘要

The main objective of this study is to evaluate several algorithms to produce monthly composite images of the VEGETATION sensor onboard SPOT-4 over the Brazilian Amazon. The ability of the commonly used Normalized Difference Vegetation Index maximum value composite (MNDVI) and other compositing algorithms (i.e. one- and/or two-step algorithms), in terms of reducing the presence of clouds and cloud shadows and assessing spatial coherence of the composite images was tested. Among the one-step algorithms, the Soil Adjusted Vegetation Index maximum value composite (MSAVI), the minimum value composite of the red band (mRed) and the minimum value composite of the red band with an additional temporal persistence condition (mRedtp) were also analysed. The two-step algorithms included MNDVI or MSAVI followed by the minimum value of the viewing zenith angle, and mRed followed by MNDVI or by MSAVI. These eight compositing algorithms were used to produce monthly (August 2000) composite images of four regions (200 km x 200 km) representative of different landscape structures in the Brazilian Amazon. The results show that none of the compositing algorithms consistently performs best over all the regions. However, at this stage of the assessment, mRedMNDVI and mRedMSAVI are good candidates for SPOT-4 VGT monthly compositing over the Brazilian Amazon.