Mangrove restoration increasingly seeks to recover biodiversity, ecosystem functioning and coastal resilience within wider blue-carbon landscapes. Species selection and planting locations are often guided by elevation envelopes or observations from individual reference forests. These approaches assume that species–environment relationships observed at one site or spatial scale are transferable to others, but this assumption is rarely tested across multiple sites within the same landscape.

We examined the consistency and transferability of mangrove tree species–elevation associations across 13 historically disturbed, regenerating mangrove sites in Singapore. Surface elevation was measured at individual trees and at random habitat points, allowing observed tree distributions to be evaluated relative to the availability of different elevations. We quantified landscape-level elevation associations for 12 tree species, compared their relative ordering among sites, and assessed whether pooled species–elevation relationships and landscape-level species abundances could reproduce local community composition.

Species showed distinguishable landscape-level elevation associations, confirming that elevation is an important axis of mangrove species sorting. However, species distributions overlapped broadly, and their relative positions along the elevation axis varied substantially among sites. Elevation also varied heterogeneously over short spatial distances and did not consistently form a simple seaward-to-landward gradient. Consequently, a coherent ordering emerged when observations were pooled across the landscape, but this ordering transferred only partially to individual sites and did not consistently reproduce local relative abundances or community composition.

These findings identify a bidirectional transfer problem. Individual reference sites may not represent broader landscape patterns, while landscape-level associations may be insufficient for predicting the composition of particular restoration sites. The observed variation could reflect site-specific environmental context, spatially structured or stochastic propagule supply, mass effects, disturbance history or demographic contingency; the present study does not distinguish among these mechanisms. Elevation and reference ecosystems therefore remain useful for initial restoration planning, but should not be treated as universally transferable templates. Replicated sampling and locally calibrated restoration targets are needed when restoration aims to recover diverse and resilient mangrove communities within blue-carbon ecosystems.