Abstract:
Abstract: Plant litter plays a critical role in controlling and protecting soil against water erosion and increasing soil organic carbon. The presence of plant litter efficiently reduces erosion and surface runoff, and influences the cycle of nutrients, carbon, and energy in ecosystem. Remote sensing can provide a new way to differentiate litter from soil, and spectral difference of plant litter and soil is the primary basis for the remotely sensed estimation of plant litter coverage. By using spectral measurement of the soil and litter samples of typical vegetation communities in the Yanhe River basin of Northern Shaanxi, the difference of spectral characteristics between soil and litter in the VIS-NIR (400-1 100 nm) and SWIR (1 100-2 500 μm) wavelengths and main impact factors were analyzed; the effectiveness of NDVI (normalized difference vegetation index) and typical senescent vegetation indexes such as NDSVI (normalized difference senescent vegetation index), NDTI (normalized difference tillage index) and CAI (cellulose absorption index) was evaluated to distinguish litter from soil. The results showed that the spectral behaviors of soil and litter were similar in the VIS-NIR wavebands, and the main difference between soil and litter was that the slopes of spectra of the litter samples were slightly greater than that of the soil samples. The two water absorption bands, centered at 1 400 and 1 900 nm, had the common spectral features in soil and litter within the SWIR waveband, while diagnostic features could be observed at 1 700 and 2 100 nm in the reflectance spectra of the dried litter samples, which were associated with the cellulose-lignin absorptions. Water content influenced the reflectance spectra of soil and litter samples obviously, and the reflectance of wet soil and litter was reduced by half compared to dry soil and litter. The cellulose-lignin absorption at 2100 nm obscured and disappeared in the reflectance spectra of wet litter samples, the spectra shape of the wet litter appeared very similar to that of wet soil, and hence it was indistinguishable between soil and litter. The existence of residue line which presents the linear regression relationship between any couple of TM bands was first verified with the litter samples. According to the relationship of soil line and residue line in feature space of two TM bands, residue line existed (R2=0.86) and was closed to soil line (R2=0.97) between TM3-TM4 wavebands, the NDVI values of soil samples were similar to the litter samples, and spectral differences of soil and litter can't be characterized by NDVI. The R2 of 0.81 illustrated the existence of soil line, but residue line (R2=0.10) can't be observed between TM3-TM5 wavebands, and the NDSVI values of soil and litter samples were mixed and featureless. Both soil line (R2=0.95) and residue line (R2=0.65) existed between TM5-TM7 wavebands, and NDTI values of soil and litter samples were still in proximity to each other. Due to low spectral resolution, there were limitations for multispectral indexes (like NDSVI and NDTI) to extract the information of litter. The spectral separability between soil and litter can be represented by the hyperspectral index CAI, which takes the advantage of obvious difference between soil and litter at 2 100 nm and thereby presents a good result for distinguishing litter from soil.