Spatiotemporal variations of soil salinization in China's West Songnen Plain

Preventing soil salinization and improving soil productivity play vital roles in achieving the Sustainable Development Goals (SDGs), the premise of which is understanding the spatiotemporal patterns of soil salinity in various ecosystems. To improve regional ecosystem management decisions, we examined the spatial variations and temporal changes of soil salinity in China's West Songnen Plain by separately developing prediction models of soil salinity for vegetated and non-vegetated areas (VEA and NEVA) through the combination of field samples, LANDSAT series images, and the partial least square regression (PLSR) method. Twelve variables, encompassing spectral bands, vegetation, and soil salinity indices, were extracted from LANDSAT satellite images. Among the investigated variables, the enhanced vegetation index (EVI), red band, and digital elevation model (DEM) yielded the best results for assessing the soil salinity in VEA. In contrast, the green band and SI2 exhibited the most significant correlation with soil salinity for NEVA. Based on 10-fold cross-validation, models predicted soil salinity with R-2 values of 0.812 and 0.803. Results reveal that the salt-affected areas in the West Songnen Plain were estimated at 1.23 x 10(6) hm(2) in 2016 with evident heterogeneous distribution. Intensively salt-affected area (IA) and moderately salt-affected area (MA) are mainly distributed in the low-lying central and northern regions of the plain. From 1991 to 2016, the salt-affected area experienced a rapid increase followed by a slight decline with a peak in 2000. This suggests effective soil amelioration due to ecological projects and a wetter climate, however, the salt-affected area still had a net increase of 0.32 x 10(6) hm(2). The variation of different salt-affected levels was characterized by an increase in slightly salt-affected area (SA), MA, and IA. Cropland was the dominant land cover type affected by soil salinization, with the affected area expanding from 0.39 x 10(6) hm(2) in 1991 to 0.57 x 10(6) hm(2) in 2016. MA and SA were the dominant levels of soil salinization in grassland, with MA expanding by 1.3-times from 1991 to 2016. Although the area of soil salinization increased from 1991 to 2016, the salt-affected area was reduced, and the salt-affected level was downgraded in some parts of the study area after 2000. These results could provide support for monitoring soil salinization in other regions, with the findings expected to benefit ecosystem conservation and restoration for regional sustainable development and provide support for evaluating SDGs.