Rock riprapping revetment is a form of riverbank stabilization that is commonly used to protect road and bridge infrastructures from fluvial erosion. However, little is known about how these may affect habitat selection by riverine fish in streams impacted by multiple stressors stemming from agricultural activities. Nine streams in Southern Québec (Canada) served as study sites, where fish surveys and habitat attributes, both man-made (i.e. rock riprapping revetment) and natural (i.e. water depth, temperature etc.) were measured. Nested linear mixed-models using wetted stream width, water depth, and velocities as fixed effects, and spatio-temporal variables as random effects, were conducted to assess the ecological effects of rock riprapping revetment across agricultural streams. Results indicated that overall fish density was significantly explained by fixed effects (R2 = 35.7%) and by random effects (R2 = 24.2) which lead to a model explaining 59.9%. Nested linear mixed-models was assessed by cross-validation (jackknife leave-one-out procedure) and showed an R2CV of 57%. An interaction between rock riprapping revetment and wetted stream width on fish densities was also observed (p-value = 0.0195). These findings suggest that the rock riprapping revetment zones had greater fish densities if water depths were maintained, as these zones contributed to stream habitat complexities, providing suitable fish habitats that were not available elsewhere in these streams. The knowledge gained here may further improve on existing riprap zone designs implemented by the Ministry of Transportation of Quebec.