Forecasting the colonization process is important for wildlife managers who supervise the reintroduction of endangered species or control the spread of invasive species. Patch connectivity is thus critical to predicting the fates of expanding populations. Connectivity in river networks results from river dendritic structure and dispersal modality of organisms. Both factors may strongly affect the colonization process and the efficiency of conservation action plans. Based on empirical data, we simulated, using a simplified model with limited number of parameters, the colonization of a large river network, the Loire River, by the native Eurasian beaver and the invasive African clawed frog. For each species, we inferred model parameters (dispersal behavior and distances) by comparing the simulated and the observed distributions. Using network theory, we evaluated the efficiency of alternative conservation strategies to prevent or promote colonization of the river network. Network robustness to fragmentation and disturbance was also assessed. The model acdurately predicted > 70% of the observed species ranges. Conservation strategies that selectively protect habitat patches with the highest connectivity values provide a weak advantage at preventing connectivity loss compared to random protection strategies. In contrast, the targeted destruction of highly connected patches had a much stronger effect on the fragmentation of the network than the random removal of habitat patches. Spatial network topology strongly contributes to determining colonization patterns of large river watersheds. Network theory allows tests for robustness of rivers to fragmentation and disturbance, and identification of strategies for conservation planning.