Mapping the Levels of Soil Salination and Alkalization by Integrating Machining Learning Methods and Soil-Forming Factors

Accurate updating of soil salination and alkalization maps based on remote sensing images machining learning methods plays an essential role in food security, biodiversity, desertification. However, there is still a lack research using machine learning, especially one-dimensional convolutional neural networks (CNN)s, soil-forming factors to classify the salinization degree. As case study, study estimated by Random forests (RF) CNN 88 observations 16 environmental covariates Da’an city, China. The results show that: RF model (accuracy = 0.67, precision 0.67 for salination) with synthetic minority oversampling technique performed better than CNN. Salinity vegetation spectral indexes played most crucial roles alkalinization estimation Songnen Plain. spatial distribution derived from shows that 1980s 2021, areas increased at annual rate 1.40% 0.86%, respectively, size very high was expanding. degree change under various land-use types followed mash > salinate grassland dry land forest. This provides reference rapid mapping, evaluating, managing arid areas.