Table 1 Summary of recent works on water quality and irrigation suitability (methodology, findings, and limitations).

M’nassri et al., 2022³¹

IWQI, ANN, SVM

ML models improved IWQI prediction accuracy (R² > 0.9).

Focused on single index (IWQI) without multivariate integration.

Tunisia (central)

Jafar et al., 2023³²

MLR, ML (SVM, RF)

ML models outperformed regression in predicting drinking WQI.

Limited dataset; no irrigation-focused indices.

Iraq (Al-Seine Lake)

Gaagai et al., 2023³³

PCA, ANN, GIS

Combined hydrochemistry and ML for groundwater quality mapping.

Did not evaluate multiple IWQIs or model generalization.

Algeria (Doucen Plain)

Trabelsi & Bel Hadj Ali, 2022²⁹

ANN, IWQI

Demonstrated ANN superiority for irrigation water quality.

No integration with geostatistics or multi-index validation.

Tunisia (Medjerda Basin)

Palabıyık & Akkan, 2024²⁴

MLR, MLP, PCA

MLP achieved best WQI prediction (R² = 0.97).

Focused on single index (WQI); lacked spatial prediction.

Turkey (Aksu Creek)

Our Study

PCA, HAC, IWQIs, ANN, SVM, MLR, EBKRP

Integrated statistical-ML-geostatistical approaches; achieved R² = 0.97 for IWQI.

Introduces cross-model validation and high-resolution spatial mapping, which was the first application in the region.

Algeria (Ouled Djellal)