Comprehensive forecasting of geospatial changes in soil fertility in Cherkasy region using CLUE-S and ANN models

Abstract

Introduction. This study analyzes soil fertility dynamics in Ukraine’s Cherkasy region, considering natural, socio-economic and anthropogenic factors. Soil fertility underpins agricultural productivity, ecological stability and food security, yet is highly sensitive to climate variability, land-use changes, and economic decisions. The research focuses on developing a spatially explicit predictive model to forecast fertility trends, capturing interactions between land use, agrochemical properties (e.g., humus content, pH) and economic drivers of land management. By linking environmental and socio-economic processes, the study provides a tool for informed, sustainable land resource management.

Methods. To achieve this objective, an integrated modeling framework was developed, combining the CLUE-S model, which is designed for spatially dynamic analysis of land-use change, with an artificial neural network (ANN) capable of predicting agrochemical indicators based on climatic, agronomic, and socio-economic data. The CLUE-S component enabled the simulation of land-use transitions and the assessment of spatial patterns of change, while the ANN provided forecasts of humus content and pH levels using historical datasets. This approach allowed for the synthesis of geospatial modeling and data-driven prediction in a unified system. Special consideration was given to socio-economic parameters, including state agricultural policies, the availability of financial resources for the agricultural sector, and prevailing market conditions, ensuring that the modeled scenarios would be both scientifically valid and reflective of real-world constraints.

Results. The modeling revealed that humus content and soil pH are the primary determinants of long-term soil productivity. Under the business-as-usual scenario, where current land-use practices remain unchanged, humus content is projected to decrease by 8–12 % by 2050, while pH levels could fall to 5,7, resulting in a 15–20 % decline in fertility. In contrast, the sustainable land-use scenario, which incorporates optimized crop rotations and practices aimed at maintaining soil organic matter, is expected to stabilize humus content at 4 % and maintain pH levels between 6,3 and 6,5, leading to improved nutrient availability and more efficient use of resources. These results clearly demonstrate the significant potential of targeted land management strategies to mitigate future losses in soil fertility.

Conclusions. The findings confirm the effectiveness of an integrated spatial-predictive modeling approach that combines the CLUE-S model with artificial neural networks for forecasting soil fertility under different environmental and socio-economic conditions. A distinctive feature of this study is its comprehensive inclusion of socio-economic drivers alongside biophysical variables, enabling the development of realistic and policy-relevant projections. The proposed framework has considerable practical value for agricultural landscape planning, the design of sustainable land-use strategies, and the creation of land resource management policies. By enabling the assessment of long-term impacts of land-use change, the optimization of soil resource utilization, and the prevention of degradation, the CLUE-S + ANN model provides a strategic tool for ensuring agricultural productivity and ecological stability in the face of climate change and socio-economic transformations.