Monitoring the spatiotemporal dynamics of soil salinization in Qingtongxia Irrigation District using multi-temporal Landsat TM/OLI data with multi-parameter optimization

Qingtongxia Irrigation District in Ningxia is a typical Yellow River irrigation zone, where natural‑human induced soil salinization impacts 49.67% of cultivated land and threatens regional agricultural‑ecological sustainability. To unravel the spatiotemporal patterns of soil salinization, and underpin site-specific salinity mitigation strategies, this study proposes a multi-source data fusion framework for high-accuracy soil salinity retrieval. Using 35-year Landsat time-series and field surface-salinity measurements, five regression algorithms including Multiple Linear Regression (MLR), Stepwise Regression (SR), Partial Least Squares Regression (PLSR), Support Vector Regression (SVR), and Random Forest Regression (RFR) are constructed for surface soil salinity prediction. The preliminary comparisons show that unoptimized RFR achieves cross-validation R 2 of 0.132, whereas other baseline models yield negative R 2 values. Four smoothing filters were applied to denoise remote sensing spectra and improve inversion precision. Cross-comparison confirms that Gaussian Filter RFR (GF-RFR) achieves optimal predictive capacity with cross-validation R 2 of 0.717, superior to moving average-RFR (MA-RFR, R 2 =0.432), Savitzky-Golay-RFR (SG-RFR, R 2 =0.502), and median filter-RFR (MF-RFR, R 2 =0.541) for regional-scale surface soil salinity mapping. The optimized GF‑RFR model generates 1989–2023 soil‑salinity spatial maps, with 35‑year inversions revealing broadly declining salt levels across most of the irrigation district. Substantial salinity decreases occur along the central northeast‑southwest corridor and eastern‑western zones, while local salt accumulation dominates northern portions. This study provides theoretical and technical support for long‑term monitoring, mitigation and integrated management in arid Yellow River irrigation regions to promote sustainable soil utilization and food security.