土壤有机质与水分反射光谱响应特征综合作用模拟

土壤有机质与水分均对土壤反射光谱特征有显著的影响,但其作用机理难以定量描述。通过对黑龙江省典型黑土区土壤野外高光谱反射率的测定,研究了该区土壤的光谱反射特征;利用BP神经网络方法,以土壤有机质与水分数据作为输入层,以土壤光谱反射率一阶微分作为输出层,建立黑土有机质与水分的高光谱预测模型,并对模型的稳定性和预测能力进行检验。结果表明:土壤有机质与土壤表层0~20 cm含水量之间具有显著的相关性,相关系数为0.59;1570 nm波段处的一阶微分为输出层的模型精度最高,RMES达到0.017,平均绝对误差为0.014,平均相对误差为0.110;模型检验结果表明建立的BP神经网络模型具有良好的稳定性;土壤有机质、含水量对土壤光谱反射率的综合作用得到较准确的描述,可以用于野外土壤有机质与水分的速测。

Synthesis Simulation of Correlation Characteristics between Organic Matter and Moisture Reflectance Spectra

The soil spectral reflectance is comprehensively determined by soil parent material,texture and moisture,organic matter,iron oxide,etc.Therefore,it's hard to identify the single effect of each soil property.Previous researches analyzed single factor of soil reflectance spectrum,however,ignored the impact of the interactions of the above mentioned factors.In this paper,two major factors influencing black soil spectral characteristics,such as soil organic matter(SOM) and soil moisture,were considered to build the hyperspectral reflectance model.The research was carried out in black soil area in Heilongjiang,China.The result showed that the correlation coefficient between SOM and soil moisture in 0-20 cm was up to 0.59;the soil reflectance model with soil organic matter and soil moisture as input variable was established using Back Propagation(BP) artificial neural network.This model was proved to be of great stability,and its RMSE reached 0.0017.The comprehensive effects of organic matter and moisture to soil reflectance spectra were accurately described.The BP neural network model provides a theoretical support to predict soil physical and chemical parameters precisely and quickly.