Remote sensing prediction and characteristic analysis of cultivated land salinization in different seasons and multiple soil layers in the coastal area

Abstract

Soil salinization is the main threat to agricultural development. The ecological environment in the coastal plain is fragile and complex, and it is urgent to realize remote sensing monitoring of cultivated land salinization. This paper took the Kenli District in the Yellow River Delta as the study area, and measured the soil salinity content in four seasons: spring, summer, autumn, and winter, and four layers: surface layer (0–15 cm), superficial layer (15–30 cm), middle layer (30–45 cm), and deep layer (45–60 cm). The vegetation index and salinity index were extracted from Landsat images, and surface salinity inversion models in different seasons were constructed by a random forest algorithm. Then, the conversion model was constructed to achieve salinity prediction in multiple soil layers by simulating the salinity variation relationship, and the variation characteristics of soil salinization were analyzed. The results showed that: (1) Both vegetation and salinity indices are significantly correlated with surface salinity in different seasons, with R2 > 0.694 and RMSE < 3.358. (2) Salinity conversion models are quadratic or cubic, with R2 > 0.621 and RMSE < 5.281, showing strong universality. (3) The underground salinity prediction method based on the conversion model is superior to direct remote sensing modeling, R2 and RMSE can be improved by 0.110 and 0.411, respectively. The prediction effect in autumn and deep layer is relatively poor. But the prediction and interpolation results of salinity grades have similar spatial distribution, and the area error is generally < 6.84%. (4) The cultivated land salinization is the weakest in summer. From surface layer to deep layer, soil salinity increases continuously. The soil salinization of low-grade land is serious, restricting the development of cultivated land quality. This study proposes a remote sensing prediction method for cultivated land salinization in different seasons and multiple soil layers in the coastal plain, which can provide decision support for dynamic monitoring of land salinization and sustainable utilization of land resources.