基于宽波段数据定量反演土壤有机质含量的研究

摘要

定量分析了北京顺义、通州区土壤高光谱反射特征,利用资源三号、高分一号、高分二号传感器的光谱响应函数,结合高光谱数据生成相应宽波段模拟数据;将土壤光谱数据、拟合宽波段数据分别与实测土壤有机质含量开展相关性分析,提取并筛选敏感波段,利用偏最小二乘法建立基于高光谱数据的土壤有机质含量预测模型;依据宽波段模拟数据和实测土壤有机质含量的相关性,提取并筛选敏感波段,建立土壤有机质含量预测模型.结果表明,在基于土壤高光谱数据建立的土壤有机质含量预测模型中,以对数的一阶微分为最优,其R和RMSE分别为0.697和0.195,偏最小二乘法得到的反演土壤有机质含量的模型是可靠的;在基于模拟宽波段构建的土壤有机质含量估测模型中,以高分一号的拟合精度最高,R和RMSE分别为0.334和0.240;受室外不可控因素的影响,模拟宽波段数据在估测北方地区土壤有机质含量方面仍需进一步研究.%In this study, the hyperspectral reflectance characteristics of soils in Shunyi and Tongzhou districts of Beijing city were quantitatively analyzed, and the corresponding spectral data were generated by using the spectral response function of resource No.3,high score one and high score two sensors,and the corresponding spectral data were generated by hyperspectral data. The method was used to analyze the soil organic matter content based on hyperspectral data. Based on the broad-band simulation data and the measured soil organic matter content, the soil organic matter content was calculated by using the par-tial least squares method. And the sensitive bands were extracted and screened to establish the soil organic matter content prediction model. The results show that the soil organic matter content is the best in the soil organic matter content prediction model based on soil hyperspectral data. The R and RMSE are 0.697 and 0.195, respectively. The soil organic matter content The model is reliable. In the soil organic matter content estimation model based on the simulation of wide-band construction, the fitting accuracy of high score number is the highest, R and RMSE are 0.334 and 0.240 respectively. By outdoor uncon-trollable factors, the data need to be further studied in estimating the soil organic matter content in the north.