LAUSR

Abstract Mapping soil salinity in coastal wetlands is important for revealing the spatiotemporal characteristics of soil salinization but has proved difficult due to complex influencing factors. An improved comprehensive land surface factor system (ICLSFS) was established based on remote sensing for mapping soil salinity in coastal wetlands. The ICLSFS comprised six classes of predictors, namely, spectral value, salinity index, vegetation index, heat and moisture index, spatial location, and land cover. The Yellow River Delta, a typical coastal wetland in North China, served as the study area to demonstrate the ICLSFS in different seasons. Results revealed the high accuracy and low uncertainty of the ICLSFS methods, and partial least squares regression achieved the highest accuracy. The root mean squared error was lowest in winter of all seasons when the soils were bare in North China. The easily-accessed data provided by remote sensing and the repeatability of simulation process render ICLSFS highly applicable in soil salinity mapping in different coastal wetlands. The mean values of soil salinity in the study area were 8.00, 7.53, 7.83, and 6.90 g/kg in spring, summer, autumn, and winter, respectively. Soil salinity exhibited distinct spatial heterogeneity and was high in shoreline areas, and low in the central part of the study area along the Yellow River. In different land cover types, salterns possessed the highest soil salinity, whereas wetland vegetation and farmland had low soil salinities. The ICLSFS was proven to possess high accuracy, low uncertainty, and good applicability, and the accuracy differed across seasons and was the highest in winter. The spatiotemporal characteristics of soil salinity were influenced by natural and anthropogenic factors. Human activities generally aggravated the soil salinization by saltern constructions, saline water irrigation, and groundwater use, while countermeasures to mitigate the soil salinization have achieved good effects in recent years.