Combination of GF-2 high spatial resolution imagery and land surface factors for predicting soil salinity of muddy coasts

The distribution of soil salinity in coastal wetlands determines gradient distribution and succession of vegetation communities. Due to the inaccessibility of mudflats in coastal areas, soil salinity prediction and mapping is necessary for biodiversity protection and ecological integrity. Yancheng National Nature Reserve (YNNR) is experiencing a salinity-induced invasion of Spartina alterniflora and degradation of native vegetation (Suaeda salsa) since the 1980s, with Suaeda salsa facing the dilemma of disappearing from YNNR. This study was aimed to establish comprehensive land surface factors, such as salinity indices, vegetation characteristics, location, terrain attributes and tidal creeks, for soil salinity mapping from 2015 to 2019. Multiple linear regression (MLR) and random forest (RF) linked land surface factors and soil salinity, and RMSE of the soil salinity predictions were 0.83 ds/m for MLR and 0.72 ds/m for RF. The RF model for soil salinity mapping was more suitable. The value of soil salinity varied from 0.9 ds/m to 5.2 ds/m and decreased year by year, with average values of 3.09 ds/m, 2.87 ds/m and 2.67 ds/m for RF in 2015, 2017 and 2019. The distance from sea and vegetation type were the two most important indicators to predict the distribution of soil salinity in these mudflats, thus better predictions on how to predict the response of vegetation communities to shifting soil salinity.

Automated summary

The prediction and mapping of soil salinity in coastal wetlands is crucial for biodiversity protection and ecological integrity. This study successfully established soil salinity maps from 2015 to 2019 using multiple linear regression and random forest models, with the random forest model being more suitable. The distribution of soil salinity is influenced by factors such as distance from sea and vegetation type.