基于高光谱遥感土壤有机质含量预测研究

Study on the Prediction of Soil Organic Matter Content Based on Hyperspectral Remote Sensing

高光谱遥感技术是土壤养分含量监测的一种先进手段,能及时、准确地监测滨海滩涂土壤养分的动态变化,对滨海湿地生态环境保护具有重要意义。以辽河三角洲滨海湿地滩涂为研究对象,在对地面高光谱数据与实验室样本理化性质分析的基础上,经过微分以及连续统去除法相关分析,分别建立了多元线性回归模型和偏最小二乘法(PLSR)模型,并对模型结果进行分析、比较。结果显示,连续统去除法能有效的提高有机质含量与光谱反射率之间相关性,基于此变换所建立的多元线性回归模型R2达到了0.7545,RMSE为1.2216;偏最小二乘模型R2达到了0.8792,RMSE为1.2299。所建模型具有预测土壤有机质含量的潜力,其中偏最小二乘法优于多元线性回归法。

Hyperspectral remote sensing technology is an advanced technique in monitoring soil nutrient content. It is important to protect the coastal wetland environment by monitoring the dynamic changes of soil nutrients. Taking Liaohe River Delta beach as researching zone, based on hyperspectral ground data and laboratory analysis data of the samples, multiple linear regression model and partial least square(PLSR) model were established after correlation analysis between differentiation and continuum removal method. The examined models results conveyed that the continuum removal method could improve the correlation between organic matter content and spectral reflectance effectively. Based on this transformation, the R2 and RMSE of multiple linear regression models were 0.7545 and1.2216. The R2 and RMSE of partial least square regression model reached 0.8792 and 1.2299, respectively. It showed that the multiple linear regression model and the PLSR model had the potential to predict the soil organic matter content, and the partial least square was superior than linear regression method.