Evaluation of Different Environmental Covariates Performance for Modeling Soil Salinity Using Digital Soil Mapping in a Susceptible Irrigated Rural Area

Soil is an indispensable resource for the development of the ecosystems, also working as a support for the human activities, being essential for the agricultural productivity. There are many soil degradation risks that cause a quality deterioration. One of the major risks is soil salinity, caused by the accumulation of salts both naturally and anthropically. For this reason, prevention measures are needed. To this end, soil properties inference and modelling result essential. Thus, the main objective of this research is to find the most useful environmental covariates for modeling soil salinity through the application of the Digital Soil Mapping (DSM) methodology in an irrigated rural area in Castile and León (Spain). For this purpose, 132 soil samples from two different laboratories were used, which contained electrical conductivity measured in saturated paste (EC_x). In addition, several environmental covariates related to soil salinity were employed to perform a statistical analysis through the combination of multiple linear regression (MLR) and generalized linear models (GLM). Afterwards, the best prediction model and its explanatory covariates were selected. The MLR showed R² values between 0.382 and 0.581 for the laboratories analyzed. In turn, all the models almost had the same main covariates, which were associated to remote sensing indices and topographic variables. Finally, it was concluded that the method is useful to determine the most important variables for modeling soil salinity, allowing more accurate predictions, identifying which susceptible areas need preventive measures and helping to achieve those SDGs targets that involve soil’s conservation.