Understanding the response of species to global change requires disentangling the drivers of their distributions across landscapes. Colonization and extinction processes, shaped by the interplay of landscape-level and local patch-level factors, are key determinants of these distributions. However, disentangling the influence of these factors, when larger-scale processes manifest at local scales, remains a challenge. We addressed this challenge by investigating the colonization and extinction dynamics of the aquatic plant, Hydrilla verticillata, in a complex riverine rock pool system. This system, with hundreds of rock pools experiencing varying flooding frequencies, provided a natural laboratory to examine how a single landscape-level disturbance can differentially impact colonization and extinction depending on local patch characteristics to shape species distributions. Using 5 years of data across over 500 sites and more than 5000 surveys, we employed dynamic occupancy models to model colonization, extinction, and changes in Hydrilla patch occupancy while accounting for imperfect detection. Our results revealed that larger, infrequently flooded pools closer to the river were more likely to be colonized. In contrast, local extinction of Hydrilla was more likely in smaller pools closer to the river that flooded frequently. These findings underscore the importance of considering context-dependence in species distribution models. The same landscape-level disturbance (flooding) had opposing effects on colonization and extinction, with the direction and magnitude of these effects varying with local patch characteristics. Our study highlights the need for integrating local and landscape-level factors, and considering how larger-scale processes play out at the patch level, to understand the complex dynamics that shape species distributions.