Forest growth monitoring of tropical acacia plantation by multi-year ALOS/PALSAR polarimetric data

摘要

This study assessed the possibility to monitor the growth of industrial plantation forests in Sumatra, Indonesia using L-band SAR data. Our previous analysis on a single year data of ALOS/PALSAR [1] showed linear correlations between decomposition powers from the four-component decomposition method [2] and field-measured forest parameters. However, increase of the canopy/volume scattering with the growth of forests showed inconsitency with the opitical ALOS/AVNIR data analysis results [3], that showed decreasing trend with tree age in the normalized difference vegetation index (NDVI) values for trees older than two years. However, the issues of inconsistency between microwave and optical data analysis results and backscatter mechanisms related to vegetation layer and are yet to be discussed. In this study, a general four-component scattering power decomposition method [4] was applied to multi-year ALOS/PALSAR data from 2007 to 2010 to provide a definite relationship between polarimetric SAR data and forest inventory data of, and a suggestive comprehension of backscattering mechanisms from, the targets consisting of fast-growing planted acacia trees. Statistical comparisons (linear correlation analysis) were performed between base 10 logarithms of in-situ measurement data (trunk diameter, tree height, standing tree volume) and the decomposition powers (surface, volume, double-bounce, helix scatteirngs) normalized by the total power, in order to evaluate the linear dependence of the polarimetric information onto the forest biometric parameters. The correlations indicated the same tendency both among the forest biometric parameters and in each decomposition power. The multi-year PALSAR data analysis revealed the following rational correlations. Significant negative correlations were observed in the surface scattering power (R = −0.65 ∼ −0.79), whereas positive correlations were found in canopy (R = 0.47 ∼ 0.71), double-b- unce (R = 0.42 ∼ 0.61) and helix (R = 0.59 ∼ 0.73) scattering powers. Furthermore, we attempted to investigate the decomposition power's transition which took place in the forest compartment (minimum management unit) area through the three years. The analysis was conducted over those compartments with a consistent permanent sample plots (PSPs) for the three years between 2007 and 2010: 9 of 32 plots in 2007 retained until 2010. Our tracking analysis results suggested that (1) the effect of harvesting was clearly shown on the double-bounce scattering, but not on the other scattering powers (2) According to increase of understory vegetation after the growing phase of the trees, the canopy scattering started to much include both forest canopy and understory effects and consequently the double-bounce scattering, which theoretically denotes the stem volume, decreaseed, (3) the helix scattering represented the density of branches to some extent and (4) the polarimetric decomposition powers have great potential to be utlized for the detection of forest degradation. Especially the detailed tracking analysis of the polarimetric decomposition powers with comparison to the forest inventory data has never been studied so far. Our findings indicated that the effect of understory environment has to be taken into account for the estimates of standing tree volume in this plantation or over sparsely vegetated areas. This study also shows the possibility of monitoring industry forests using microwave L-band SAR data.