Feasibility Study of C- and L-band SAR Time Series Data in Tracking Indonesian Plantation and Natural Forest Cover Changes

Tropical coverage by Envisat ASAR is sparse in space and time and has limited value for monitoring deforestation. The only available dual-polarized multitemporal dataset over Riau province, Indonesia (nine images in a single year), is used to distinguish and monitor tropical plantations and their dynamics and is compared with annual L-band PALSAR data and land cover maps derived from Landsat data. For the ASAR data, both VV and VH are important in discriminating different types of forest cover; whereas, at L-band, most of the relevant information is in the cross-polarized channel. The ASAR VV (but not the VH) backscatter from acacia plantations is strongly affected by whether the underlying soil is peat or nonpeat, which affects the separability of acacia from oil palm. Maximum likelihood classification of the C-band data gave overall accuracies of 86.2% and kappa coefficient of 0.78 by comparison with land cover maps derived from optical data. This was not improved by combining C- and L-band data. Classification of the C-band time series allows the rotation cycle of acacia plantations to be tracked. The available 4-year annual L-band time series shows potential for monitoring these dynamics, but the 1-year time spacing increases the risk of missing changes masked by the rapid growth of acacia.