基于实测光谱和SRF的稀疏植被区MODIS积雪信息提取

Extraction of snow cover information in sparse vegetation area based on spectral measurement and SRF by using MODIS data

以新疆准噶尔盆地古尔班通古特沙漠为研究区,以中等分辨率成像光谱仪(MODIS 1B)数据为例,辅以MODIS光谱响应函数(SRF)和全波段光谱仪(ASD)准同步采集的雪面反射光谱,运用线性光谱混合模型(LSMM)实现了稀疏植被区积雪遥感信息提取。结果表明:①利用SRF对雪面反射光谱进行端元光谱到像元光谱的转换,生成对应于MODIS1—7波段的离散光谱,将其与用最小噪声分离(MNF)变换和像元纯度指数(PPI)法获得的MODIS影像端元光谱进行对比,发现MODIS1波段光谱值远大于转换光谱值,MODIS2—7波段光谱值与转换光谱值接近;②MODIS2—7波段影像端元光谱值适用于LSMM估算稀疏植被区积雪分量,积雪分量估算值与归一化差分积雪指数(NDSI)拟合结果显示,剔除MODIS1波段后估算的积雪分量与NDSI的相关性显著提高,表明所提取的积雪分量可以作为估算积雪的典型指数。

In this paper, the linear spectral mixture model（LSMM） was used for the spectral unmixing analysis of the Moderate Resolution Imaging Spectrometer（MODIS） data of the study area in Gurbantunggut desert. Using the spectral response function（SRF） of MODIS1 -7 bands,the authors transformed the end -member spectrum quasi - synchronously collected by the full - band spectrometer （ASD） to the pixel spectra, thus generating the discrete spectrum of MODIS1 -7 bands. Compared with the MODIS end -member spectra obtained by minimum noise fraction（MNF） transform and pixel purity index（PPI） ,the end -member spectral values of the first band of MODIS were much larger than the transformed spectrum values, but the spectral values of the MODIS2 - 7 bands were close to the transformed values. Therefore, selecting the image end - member spectral values of the MODIS2 - 7 bands, the authors used LSMM to estimate the abundance of snow in the sparse vegetation area appropriately. Fitting the estimated snow component value to the normalized difference snow index （NDSI）, the authors found that a significant correlation exists between them after excluding MODIS1 band. The correlation coefficients show that the snow component can be a typical index of the snow cover.