Abstract

Globally, over 833 million hectares (9% of land) are salt-affected, especially in arid and semi-arid zones. In Nigeria’s semi-arid areas, irrigation enables dry season farming but contributes to salinity build-up, threatening soil health and yield. This study applied the SALTMED model, supported by R-based machine learning, to evaluate soil salinity and moisture dynamics in Kadawa, KRIP. The model was calibrated using local data: sandy loam texture, bulk density (1.748 - 1.925g/cm³), moisture (0.19 - 0.23m³/m³), field capacity (14.1 - 18.5%), wilting point (11.5 - 14.75%), and saturated hydraulic conductivity (165.51 - 516.33 mm/day). Between April and July 2024, salinity increased from 0.30 to 1.56 dS/m, while moisture dropped below 0.13m³/m³. SALTMED showed strong accuracy with R² values of 0.95 and 0.93 (salinity), and 0.86 and 0.95 (moisture). NRMSE, MAE, and CRM values confirmed good model performance. A strong inverse correlation (r as –0.96, R² as 0.93) indicated salinity increased by 1.96 dS/m for every unit drop in moisture. ARIMA forecasted further moisture decline (1.2 dS/m) by October 2024. The study recommends leaching, soil amendments, and precision irrigation to manage salinity sustainably.

Key words: soil salinization, salinity trend prediction, soil moisture dynamics, SALTMED model