AI-driven spatiotemporal quantification and prediction of soil salinity at European scale using the LUCAS database

Soil salinization, referring to the excessive accumulation of soluble salt in soils, adversely influences nutrient cycling, microbial activity, biodiversity, plant growth and crop production thus affecting soil health and ecosystem functioning. Soil salinity quantification is a major step toward mitigation of its effects. Therefore, developing quantitative tools to predict soil salinity at regional and continental levels under different boundary conditions and scenarios is crucial for sustainable soil management and security of natural resources (1-3). This study proposes an AI-driven soil salinity quantification and projection approach focused on EU soils using a set of environmental covariates which consist of soil properties, terrain attributes, climate, and remotely sensed variables. The soil salinity point data which was used for model training and validation, expressed as electrical conductivity, was obtained from the LUCAS survey for the years 2015 and 2018. The novelty of this work lies in the careful integration of LUCAS data point with AI-driven models aiming to produce soil salinity maps for EU soils. Different AI algorithms including Random Forest, LightGBM, and XGBoost were used in this study enabling us to evaluate the performance of each algorithm in predicting soil salinity across EU with the XGBoost algorithm producing the most accurate results. Feature engineering technique was applied to reduce the models’ collinearity; thus 17 covariates were selected as the most important model variables influencing soil salinity from the initial 34 covariates investigated in our analysis. The output of the predictive model will be a gridded dataset illustrating the spatial and temporal (yearly) distribution of soil salinity across the EU, accompanied by the corresponding uncertainty maps with the spatial resolution of 1-km. This information is crucial for identifying regions with elevated salinity levels and formulating necessary action plans to mitigate the situation.

References

• Hassani, A., Azapagic, A., Shokri, N. (2020). Predicting Long-term Dynamics of Soil Salinity and Sodicity on a Global Scale, Proc. Nat. Acad. Sci., 117(52), 33017-33027, https://doi.org/10.1073/pnas.2013771117
• Hassani, A., Azapagic, A., Shokri, N. (2021). Global Predictions of Primary Soil Salinization Under Changing Climate in the 21st Century, Nat. Commun., 12, 6663. https://doi.org/10.1038/s41467-021-26907-3
• Shokri-Kuehni, S.M.S., Raaijmakers, B., Kurz, T., Or, D., Helmig, R., Shokri, N. (2020). Water Table Depth and Soil Salinization: From Pore-Scale Processes to Field-Scale Responses. Water Resour. Res., 56, e2019WR026707, https://doi.org/10.1029/2019WR026707