光谱分辨率对黑土有机质预测模型的影响

摘要

Laboratory reflectance of Black soil samples was re-sampled with different spectral resolution, and the correlation between soil organic matter (OM) and reflectance, spectral variables was analyzed to study the effect of spectral resolution on black soil OM predicting model. The results are as follows: the spectral response range of black soil OM is between 445 and 1 380 nm, high OM content shades the spectral effect of other soil properties. The precision of black soil OM predicting models increases and decreases with spectral resolution, and the maximum accuracy is at 50 nm, which is wider than hyperspectral resolution, and narrower than the bandwidth of multispectral sensors; with the derivative of logarithmic reflectance reciprocal as input variables, the optimal black soil organic matter predicting model shows high accuracy, with R2 =0. 799 and RMSE=0. 439; the results can provide the academic and technical support for soil organic matter remote sensing reversing and quick instrument developing.%高光谱遥感以其高光谱分辨率适于反射光谱特征复杂的地物识别与参数反演,但对于反射光谱特征平滑的地物,高光谱数据可能存在数据冗余问题.本研究对实验室测定的黑土高光谱反射率进行重采样,基于统计分析方法研究了光谱分辨率对黑土有机质预测模型精度的影响,结果表明:黑土有机质含量高,土壤有机质的光谱作用范围宽(445～1380 nm)；黑土有机质光谱预测模型精度随光谱分辨率降低,呈现先增后减的趋势,最优模型的光谱分辨率为50 nm,低于高光谱遥感波段设置,略高于多光谱传感器波段设置；黑土有机质光谱预测最优模型以倒数对数微分为自变量,模型决定系数R2=0.799,RMSE=0.439,研究成果为土壤有机质遥感反演、光谱速测仪器的研制,以及传感器波段设置提供理论基础与技术支持.