Integration of land use/land cover factors with machine learning in groundwater vulnerability assessment models for semi-arid regions Algeria
2025
Mohamed Azlaoui | Salah Karef | Atif Foufou | Nadjib Haied | Aziez Zeddouri | Djamal Bengusmia
This study presents a multi-methodological assessment of groundwater vulnerability in the Ain Oussera Plain, Algeria, utilizing three DRASTIC model approaches integrated with machine learning techniques. The study analyzed an area of approximately 3795 km² using standard DRASTIC, modified DRASTIC and DRASTIC-LULC models, validated through field measurements. The standard DRASTIC model revealed vulnerability indices ranging from 88 to 137, with 88.50 % of the area showing low vulnerability. The modified DRASTIC approach (90−141) demonstrated a shift toward higher vulnerability classifications, with moderate vulnerability zones increasing to 12.35 %. The DRASTIC-LULC model, incorporating land use factors, generated indices from 95 to 177 and uniquely identified high vulnerability zones (3.65 %) while showing a significant redistribution of vulnerability classes. Sensitivity analysis identified the impact of the vadose zone as the most important parameter (34.2 % effective weight). Model validation through nitrate (0–78.5 mg/L) and TDS (619.98–2833.62 mg/L) measurements demonstrated the superior performance of the DRASTIC-LULC model, with correlation coefficients of 0.56 and 0.75. The integration of Random Forest classification for land use mapping achieved 98.89 % accuracy, significantly outperforming traditional methods. The results provide crucial insights into groundwater protection strategies in semi-arid regions, emphasizing the importance of incorporating land-use factors in vulnerability assessments and establishing comprehensive monitoring programs.
Show more [+] Less [-]AGROVOC Keywords
Bibliographic information
This bibliographic record has been provided by Directory of Open Access Journals