利用反射光谱及模拟多光谱数据定量反演北方潮土有机质含量

摘要

The present study aims to assess the feasibility of multi-spectral data in monitoring soil organic matter content .The data source comes from hyperspectral measured under laboratory condition ,and simulated multi-spectral data from the hyper-spectral .According to the reflectance response functions of Landsat TM and HJ-CCD (the Environment and Disaster Reduction Small Satellites ,HJ) ,the hyperspectra were resampled for the corresponding bands of multi-spectral sensors .The correlation between hyperspectral ,simulated reflectance spectra and organic matter content was calculated ,and used to extract the sensitive bands of the organic matter in the north fluvo-aquic soil .The partial least square regression (PLSR) method was used to estab-lish experiential models to estimate soil organic matter content .Both root mean squared error (RMSE) and coefficient of the de-termination (R2 ) were introduced to test the precision and stability of the modes .Results demonstrate that compared with the hyperspectral data ,the best model established by simulated multi-spectral data gives a good result for organic matter content , with R2 =0.586 ,and RMSE=0.280 .Therefore ,using multi-spectral data to predict tide soil organic matter content is feasible .%基于北京市52个潮土样本的高光谱数据和Landsat TM、环境减灾卫星（HJ）影像的波段响应函数，生成宽波段多光谱模拟数据，对比分析了室内实测光谱数据、宽波段模拟数据与土壤有机质含量的相关性，筛选敏感波段，利用偏最小二乘法构建北方潮土有机质含量预测模型。研究表明：在宽波段模拟数据建立的模型中，由Landsat TM 模拟数据的差值土壤指数（DSI）、比值土壤指数（RSI）、归一化土壤指数（NDSI）及其第3波段共同构建的模型最优，其决定系数与均方根误差分别为0.586和0.280；与实测光谱数据相比，模拟数据的最佳预测模型，均优于除一阶微分、弓曲差以外的其他10种高光谱模型。因此，利用多光谱数据预测潮土有机质含量是可行的。