Estimating Soil Organic Matter Content Using Sentinel-2 Imagery by Machine Learning in Shanghai

Soil organic matter (SOM) plays an important role in the field of climate change and terrestrial ecosystems. SOM large areas, especially urban is difficult to monitor estimate by traditional methods. Urban land structure complex, soil a mixture inorganic constituents with different physical chemical properties. Previous studies showed that remote sensing techniques provide diverse data visible-near-infrared (VNIR)-shortwave infrared (SWIR) spectral range, are promising prediction content on scale. Sentinel-2 covers bands (VNIR-SWIR) for short revisit time. Thus, this article aimed evaluate capacity area (i.e., Shanghai). 103 bare samples filtrated from 398 at depth 20 cm were selected. Three methods, partial least square regression (PLSR), artificial neural network (ANN), support vector machine (SVM), applied. The root mean error (RMSE) modelling (mRMSE) coefficient determination ( R 2 ) (mR used reflect accuracy model. results show PLSR has poorest performance. ANN highest (mRMSE = 7.387 g kg $^{-1}$ , mR =0.446). RMSE (pRMSE) 4.713 (pR 0.723. For SVR, pRMSE 4.638 pR 0.732. SVR slightly higher than ANN. spatial distribution demonstrates value obtained closest range samples, performs better vegetation-covered areas. Therefore, can be method estimation.