土壤有机质含量高光谱定量反演研究

Study on Quantitative Inversion of Soil Organic Matter Content by Hyperspectral

以太行山区北流河流域为研究对象,采用多元线性回归和随机森林回归构建土壤有机质反演模型,对表层土壤有机质含量进行反演估测,以期得出太行山区表层土壤有机质遥感定量反演的建模方式与相应技术参数。结果表明:1)Savitzky-Golay(SG)平滑和小波包分析等方法可以提高土壤光谱与土壤有机质含量的相关性。2)随机森林模型在反演土壤有机质含量研究中,模型精度与效果均优于多元线性回归模型,利用小波包分解得到的低频分量构建的随机森林模型在SOM含量反演中效果最好,R~2为0.812,实现了对土壤有机质的有效估算。3)影响土壤有机质含量反演模型精度的因素较多,在未来研究中需要从光谱数据源质量、光谱数据处理方式及建模方式等方面进行深入研究,进一步完善土壤有机质含量反演技术体系,提高土壤有机质含量反演精度。

Taking Beiliu River Basin in the Taihang Mountains as the research object,multiple linear regression and random forest regression were used to construct a soil organic matter inversion model to invert and estimate the surface soil organic matter content,in order to obtain the quantitative inversion and corresponding technical parameters of the surface soil organic matter in the Taihang Mountains.The results showed that,1)methods such as Savitzky-Golay(SG)smoothing and wavelet packet analysis could improve the correlation between soil spectra and soil organic matter content.2)In the research on inversion of soil organic matter content,the random forest model had better model accuracy and effect than the multiple linear regression model.The random forest model constructed by using the low-frequency components obtained by the wavelet packet decomposition had the best effect in the inversion of SOM content,and R~2 was 0.812,achieving an effective estimation of soil organic matter.3)There were many factors that affect the accuracy of the soil organic matter content inversion model.In the future research,in-depth research on the quality of spectral data sources,spectral data processing methods and modeling methods was needed to further complete the soil organic matter content inversion technology system and improve the inversion accuracy of soil organic matter content.