When synthetic aperture radar (SAR) is used to obtain remote sensing images of forests, the images can be degraded by double reflections of the radar beam, e.g., by reflecting off water surface and then tree stems or by reflecting off a fallen tree and then by other fallen stems, which together act as a corner reflector. Here, we developed a method for correcting for this effect, called the double bounce effect. To demonstrate the method, we monitored changes in a peat swamp forest around Palangka Raya in Central Kalimantan, Indonesia using ALOS/PALSAR data. Biomass surveys were executed in plots of grass, shrub and forest and above ground biomass (AGB) was calculated. Five Fine Beam Double Polarization PALSAR images, which were obtained between July 2007 and December 2010, were used. The backscattering coef cient (BSC) increased asymptotically to the saturation level with increasing AGB. BSC of HH changed linearly with BSC of HV in areas with different biomass and without double bounce. The linear relationship between the BSC of HH and BSC of HV showed plant succession, and the change ratio of HH BSC to HV BSC show a contribution of vertical stems on BSC, since stem biomass was the main part of AGB in L-band. However, in areas such as riverside forests and re scars where BSC suffered from double bounce (DB areas) no such relationship was observed. Because signal of HH returns much stronger than that of HV by double bounce, we assumed that the double bounce changes the ratio of HH BSC and HV BSC inversely. The double bounce effect on HV BSC was corrected using a linear combination of two vectors which showed the contribution of vertical stems and the contribution of double bounce. In contrast to AGB, the original BSC was higher in DB areas than in high biomass forests. However, when we applied the above correction in DB areas, the BSC anomaly in HV disappeared. Curvilinear regression models were developed to estimate AGB using the corrected HV BSC of an image acquired in July 2007. The estimated AGB was not accurate when it exceeded 100 Mg ha^<-1> due to saturation of BSC, although it was reasonable in DB areas except in the case of an image acquired in December 2010. The method for correcting AGB should be suitable for monitoring of deforestation, regeneration and regrowth of trees in areas with low biomass in peat swamp forest.
抄録全体を表示