Native habitat affinities predict fish invasions with post-invasion habitat shifts.

Identifying habitat affinities and demographic characteristics associated with high invasion potential is a key ecological challenge. However, testing invasion hypotheses exclusively in the invaded range is inherently flawed, as it conflates the intrinsic pre-adaptations that allowed the species to invade with the habitat shifts they adopted post-invasion. To elucidate how invasion potential is related to habitat affinities and post-invasion ecological dynamics, we investigated the unique paired-basin system of the Red Sea and Eastern Mediterranean. We captured in situ realized niche data (e.g. abundance, depth limits and habitat affinities) using standardized stereo underwater video surveys across 179 populations. We then applied a three-comparison framework to evaluate native-range invasion potential, post-invasion habitat shifts and recipient-community integration. Within the native Red Sea, future introduced species were distinct from non-introduced natives primarily due to their lack of reliance on coral habitats, shallower minimum depth limits, and broader habitat and depth ranges. Comparing the native and invaded ranges of introduced species revealed post-invasion niche shifts. Rather than shifting uniformly, introduced species exhibited high habitat lability, successfully abandoning affinities for tropical substrates to utilize available Mediterranean habitats. Within the Mediterranean, introduced species were assimilated into the recipient community, becoming indistinguishable from natives in habitat use rather than occupying distinct habitats. By synthesizing linear discriminant and mixed-modelling approaches into a cross-validated ensemble index, we prioritized specific Red Sea species as high-risk candidates for future Mediterranean invasion. Ultimately, this in situ framework highlights that while specific native-range habitat filters predict invasion potential, post-invasion habitat lability means that relying solely on native affinities could severely underestimate their spread and impact in novel environments.