Modeling the dynamics of competition and coexistence between species is cru- cial to predict long-term impacts of invasive species on their native congeners. However, natu- ral environments are often fragmented and variable in time and space. In such contexts, regional coexistence depends on complex interactions between competition, niche differentia- tion and stochastic colonization–extinction dynamics. Quantifying all these processes at land- scape scale has always been a challenge for ecologists. We propose a new statistical framework to evaluate metapopulation parameters (colonization and extinction) in a two-species system and how they respond to environmental variables and interspecific competition. It requires spatial surveys repeated in time, but does not assume demographic equilibrium. We apply this model to a long-term survey of two snails inhabiting a network of freshwater habitats in the West Indies. We find evidence of reciprocal competition affecting colonization or extinction rates, modulated by species-specific sensitivity to environmental variables. Simulations using model estimates allow us to predict species dynamics and explore the role of various coexis- tence mechanisms described by metacommunity theory in our system. The two species are pre- dicted to stably coexist, because niche partitioning, source–sink dynamics and interspecific differences in extinction–colonization parameters all contribute to reduce the negative impacts of competition. However, none of these mechanisms is individually essential. Regional coexis- tence is primarily facilitated by transient co-occurrence of the two species within habitat patches, a possibility generally not considered in theoretical metacommunity models. Our framework is general and could be extended to guilds of several competing species.