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

Abstract

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

Towards the Improvement of Soil Salinity Mapping in a Data-Scarce Context Using Sentinel-2 Images in Machine-Learning Models

“…Overall, the RF model performed the best [23]. Convolutional Neural Network (CNN) and RF models were compared for soil salinity estimates across China and the results showed that the RF model was the most reliable [24]. In another study comparing the SVM, ANN and RF models to retrieve soil salinity estimates from Sentinel-2 images across Chinese arid areas, the SVM model reached the highest accuracy [10].…”

“…As soil salinity field measurements are costly and time consuming, most of the studies investigating machine-learning potential for soil salinity mapping relied on training sets gathering less than 100 samples [21,[23][24][25]28,29]. In this context, it is crucial to increase the training set size in order to improve the reliability of machine-learning models to estimate soil salinity.…”

Challenges and Opportunities in Remote Sensing for Soil Salinization Mapping and Monitoring: A Review

“…One of the primary challenges is the limited spatial coverage. As most sensors have low spatial coverage, this has restricted soil salinization studies mainly to a local scale [151].…”

Soil Salinity Estimation by 3D Spectral Space Optimization and Deep Soil Investigation in the Songnen Plain, Northeast China

“…For this reason, the use of RS as a rapid, non-contact, and low-cost technique was shown to be valuable for detecting SSS either directly on bare soils with efflorescence and white-crust or indirectly on covered soils through the biophysical characteristics of vegetation or soil moisture [4,20,21]. Some indirect environmental factors, such as terrain, hydrology, climate, and the like, could also be incorporated to characterize soil salinity [22][23][24][25], as they are mutually associated with one another. The commonly used RS inversion method is to choose a significant band or combine some sensitive bands to form different kinds of salinity indices to invert soil salinity.…”

“…Multispectral satellite images (such as MODIS, Landsat, and Sentinel) are frequently used because the multi bands of visible spectra (0.38~0.76 µm) and near-infrared spectra (0.76~3 µm) can provide the highest separability between salinized and non-salinized soils for distinguishing saline-alkaline soil regions [11,[26][27][28][29][30][31]. The Sentinel-2 multispectral sensor has a higher spatial resolution (10 m) in the red, green, blue, and near-infrared spectrum than MODIS or Landsat, in which the bands can be combined in different ways to obtain various salinity indices to invert SSS [24]. Overall, previous studies using RS mainly focused on SSS mapping, while saline-alkaline problems have extended far beyond surface soils and also affect deep soils.…”