重庆市南川区土壤锰元素遥感反演

Remote Sensing Inversion of Soil Manganese in Nanchuan District,Chongqing

土壤中的锰元素对植物生长起着重要作用,土壤锰含量过高或者缺少都将对植物产生不良影响,因此快速监测土壤中的锰含量尤为重要。目前利用遥感技术监测土壤锰含量的相关研究主要集中在利用土壤光谱估算土壤锰含量,而对于植被常年覆盖的南方地区,难以从卫星影像中获取土壤光谱。因此,引入植被光谱,探索植被覆盖区域土壤锰元素的快速监测方法。首先从Landsat 8影像中提取11种植被光谱指标,并运用皮尔逊相关系数(Pearson correlation coefficient)结合方差膨胀因子(VIF)筛选出最佳植被光谱指标;在此基础上,利用偏最小二乘回归(PLSR)、多元逐步回归(MSR)和BP神经网络(BPNN)算法构建最佳植被光谱指标与土壤锰元素之间的光谱响应模型,分析比较三个模型的估算效果从而确定最佳反演模型;最后,基于最佳反演模型进行土壤锰含量空间制图。以重庆市南川区为例,研究结果表明:3个植被光谱指标(比值植被指数,归一化植被指数和可见光大气阻抗植被指数)被确定为土壤锰元素最佳的光谱响应指标;BPNN光谱响应模型(R^(2)为0.78,RMSE为334.24,RPD为2.13)为土壤锰含量最佳反演模型,其土壤锰含量的制图精度(R^(2)为0.69,RMSE为567.64,RPD为1.30)。表明通过植被光谱指标反演土壤锰含量可行,该研究为区域尺度的土壤锰含量监测开拓了新思路。

Manganese in soil plays an important role in plant growth,and high or low levels of soil manganese will have adverse effects on plants,so it is especially important to monitor soil manganese content quickly.At present,the studies related to the monitoring of soil Mn content by remote sensing technology mainly focus on the estimation of soil Mn content using soil spectra.At the same time,it is difficult to obtain soil spectra from satellite images in the southern region where vegetation covers all year round.Therefore,this paper introduces vegetation spectra to explore the rapid monitoring method of soil Mn elements in vegetation-covered areas.Firstly,ll vegetation spectral indicators were extracted from Landsat 8 images,and the best vegetation spectral indicators were selected by Pearson correlation coefficient combined with Variance Inflation Factor(VIF);based on this,Partial Least Squares(PLS)regression was used.(PLSR),Multiple Stepwise Regression(MSR)and BP Neural Network(BPNN)algorithms were used to construct the best vegetation spectral indicators.Finally,spatial mapping of soil Mn content was carried out based on the best inversion model.Taking the Nanchuan District of Chongqing City as an example,the results showed that three vegetation spectral indicators(specific vegetation index,normalized vegetation index and visible atmospheric impedance vegetation index)were identified as the best spectral response indicators of soil Mn.The mapping accuracy of soil Mn content(R^(2) was 0.69,RMSE was 567.64,and RPD was 1.30).The results showed that the inversion of soil Mn content by vegetation spectral indicators is feasible,and this study opens up new ideas for monitoring soil Mn content at the regional scale.