Intermittent Mediterranean rivers lose most of their dilution capacity in the dry season, yet their monitoring data are rarely analysed with tests suited to skewed, paired and small datasets. We analysed twelve physicochemical parameters (annual means) and three groups of faecal indica-tor bacteria (total coliforms, faecal coliforms and faecal streptococci) measured during four seasonal campaigns in 2024 at 26 stations of the Inaouène watershed (north-eastern Morocco), using non-parametric, multivariate and machine-learning methods. The waters are dominated by a Na–Cl facies (16 of 26 stations), and it is salinity, rather than sodicity, that limits irrigation suitability (17 stations in salinity classes C3–C4). Bacterial counts follow the order winter ≤ spring ≤ autumn ≤ summer at all 26 stations for total and faecal coliforms (Friedman test, p < 10⁻¹⁴; the only ties are zero counts), and the median summer/winter ratio of faecal coliforms is 11.7. Orthophosphate and ammonium track faecal contamination after controlling for conduc-tivity (partial ρ = 0.61 and 0.52). An arithmetic weighted water quality index (WQI) ranks the stations stably (ρ ≥ 0.94) but tracks faecal coliforms as well as the nutrients alone do (ρ = 0.75 versus 0.73 for phosphate). An exploratory random forest, evaluated on stations withheld from training, predicts faecal coliforms with a mean R² of 0.32 (range 0.30–0.34 over ten random seeds), which rises to 0.57 (0.56–0.59) when nitrate is added; no external validation set was available. No direct marker of faecal origin was measured; attribution to wastewater therefore remains inferential. Seasonal sampling is nonetheless indispensable, because annual means mask the summer peaks that matter for irrigation.