The Effect Of Soil Moisture Content On Hyperspectral Prediction Of Soil Organic Matter Content

Soil organic matter(SOM) content is an important index to evaluate soil fertility. It can not only support crop nutrients, improve soil physical characteristics, but also keep soil water and fertilizers. It has important roles for farmland quality evaluation and the grain yield increasing. Soil spectral characteristics is the reflect of soil contents like SOM, soil moisture content, iron oxide content, soil texture, et al. It can be used to predicte SOM content. But soil moisture content has great influence on using spectrum analysis method to predict SOM content. The purpose of this article is to study this influence. Air dried soil, dry soil and moistened in steps of 5% to 40% samples were scanned in the laboratory with ASD Fieldspec Pro spectroradiometer. Spectra data include reflectance(R), first derivatives of reflectance(R’), and the logarithm of the inverse of the reflectance(Log(1/R)). Spectrum data for all the 63 soil samples’ and every soil type’s spectrum data were concerned to predict SOM content. Partial least squares regression(PLSR), support vertor machine(SVM) and PLSR-SVM combination were used to predict SOM content. The results are as follows:1. Taking 63 samples as all, the best predition accuracy for SOM content was air dried soil spectrum data by using PLSR and PLSR-SVM. When using SVM to predicte the SOM, the reflect data of 15% and 20% soil moisture content and air dry soil samples have better results.2. When the soil moisture content was or more than 25%, it was not suitable to be used for hyperspectral estimating SOM contents,because the influence of soil moisture content to soil spectral reflectance was higher than SOM.3. Log(1/R) was better for SOM spectral prediction when using PLSR and PLSR-SVM, while R were Log(1/R) are better when using SVM.4. SOM content prediction models for black soil showed better accuracy than the other soil types.