地物高光谱技术已被用于土壤有机质（SOM)等理化参数速测, 但由于含水量、 粗糙度等因素的影响, 基于遥感影像的SOM空间反演精度较低。 为此引入时相信息, 将时像信息与光谱信息结合对研究区SOM进行预测, 使预测模型精度显著提高。 以黑龙江典型黑土区（北安市南部、 海伦市中部、 绥化市东部、 绥棱县西南部、 望奎县中部)为例, 获取多期MODIS影像, 利用MODIS数据高时间分辨率的优势, 研究含水量对土壤反射光谱曲线的影响; 基于SOM与含水量对反射率的综合作用分析, 建立SOM遥感预测模型。 结果表明: (1)利用单期影像建立的SOM光谱预测模型, 未加入含水量变化对土壤反射光谱曲线的影响信息, 基于年积日（DOY)117, 119, 130, 140, 143单期影像建立的SOM预测模型, RMSE分别为0591, 0522, 0545和0553, R2分别为0505, 0614, 0562, 0568和0645, 模型精度及稳定性较低; (2)利用年积日119和143多时相影像建立的SOM预测模型, 考虑了含水量与SOM的综合作用, RMSE为0442, R2为0723, 模型精度、 稳定性得到显著提高。 研究成果对于区域土壤肥力评价、 土壤碳库储量估测、 精准农业发展有重要意义。

Soil organic matter (SOM) is one of the most important measuring indexes of soil fertility. How to predict SOM spatial distribution precisely has great significance to soil carbon storage estimation and precision agriculture development. Traditional measurement of SOM, although with higher accuracy, consumes a lot of labor resources and costs long-term monitoring period, therefore, it is hard to achieve dynamic monitor of SOM. Spectroscopy technique has been used in SOM and other soil physicochemical parameters quick measurement. However spatial inversion model accuracy of SOM based on remote sensing images is relatively lower than laboratory model accuracy due to the influence of soil moisture, roughness and so on. In recent years, most studies have not eliminated the effect of moisture. Since moisture has great influence on SOM spectra reflectance, this study introduced the temporal information combined with the spectral information in order to solve this problem. Soil moisture has differences in multi period remote sensing images, and the spectra reflectance is also different. Based on the combination of reflectance from of two periods remote sensing images, the spectral index was constructed to predict SOM in this study. MODIS images of study area acquired in this study area (Blacksoil zone) because of the advantage of high temporal resolution. Spectra reflectance of MODIS images were used to analyze the effect of moisture on soil spectral reflectance, and then the spectral prediction models of SOM were built based on the comprehensive impacts of SOM and soil moisture. The results shows that: (1) the accuracy of SOM prediction model based on single image was lower without consideration of moisture effect, The Root mean square error (RMSE) of SOM prediction model were 0591, 0522, 0545, 0553, and the determination coefficient (R2) were 0505, 0614, 0562, 0568, 0645 respectively based on the day of year (DOY) 117, 119, 130, 140, 143 single image. (2) Model with multi temporal images (DOY119 and 143) which considered the effect of moisture and SOM showed better predictive ability. RMSE was 0442 while R2 was 0723. Therefore the accuracy and stability of the model were significantly improved, and it can be used to predict the spatial distribution of SOM in regional scale. This study provides important information for regional soil fertility evaluation, soil carbon storage estimation, and precision agriculture development.