In the context of water resources planning andmanagement, the prediction of fish distribution related tohabitat characteristics is fundamental for the definition ofenvironmental flows and habitat restoration measures. Inparticular, threatened and endemic fish species should bethe targets of biodiversity safeguard and wildlife conservationactions. The recently developed meso-scale habitatmodel (MesoHABSIM) can provide solutions in this senseby using multivariate statistical techniques to predict fishspecies distribution and to define habitat suitability criteria.In this research, Random Forests (RF) and LogisticRegressions (LR) models were used to predict the distributionof bullhead (Cottus gobio) as a function of habitatconditions. In ten reference streams of the Alps (NW Italy),95 mesohabitats were sampled for hydro-morphologic andbiological parameters, and RF and LR were used todistinguish between absence/presence and presence/abundanceof fish. The obtained models were compared on thebasis of their performances (model accuracy, sensitivity,specificity, Cohen’s kappa and area under ROC curve).Results indicate that RF outperformed LR, for bothabsence/presence (RF: 84 % accuracy, k = 0.58 andAUC = 0.88; LR: 78 % accuracy, k = 0.54 and AUC =0.85) and presence/abundance models (RF: 79 % accuracy,k = 0.57 and AUC = 0.87; LR: 69 % accuracy, k = 0.43and AUC = 0.81). The most important variables, selectedin each model, are discussed and compared to the availableliterature. Lastly, results from models’ application in regulatedsites are presented to show the possible use of RF inpredicting habitat availability for fish in Alpine streams.