Plant species mapping across the Miankaleh biosphere reserve and successional patterns in its drying wetland: A remote sensing-based assessment

Coastal habitats such as dune shrublands are critical for biodiversity, human livelihoods, and ecosystem services, yet they are increasingly threatened by climate change and human pressures. Effective monitoring is therefore essential for adaptive management and long-term resilience of these ecosystems. Miankaleh peninsula, a biosphere reserve and Ramsar site located on the southern coast of the Caspian Sea, northern Iran, is of great ecological significance but remains poorly studied despite recent climatic and anthropogenic pressures. This study used a multi-temporal Sentinel-2 dataset from 2023 (15 dates), including 10 spectral bands and 28 vegetation indices, along with extensive field observations, to train a Random Forest model to classify ten plant species and five land cover types. A Sentinel-2 and Landsat time series (2010−2023) was further used to map wetland and dry-wetland extent and to derive pixel-level drying age. The occurrence of different plant species in the dried wetland and their distribution within different drying ages were investigated. The classification achieved solid performance with an overall accuracy of 93.7%. Dominant plant species such as Punica granatum , Rubus sanctus , Juncus spp., Salicornia iranica , and Artemisia campestris revealed particularly high F1-scores (> 0.940). Annual wetland maps reached overall accuracies of 98.7–99.8% and indicated that about 20,952 ha (40.7%) transitioned to dry wetland. Plant community composition showed patterns consistent with ecological succession across wetland drying ages. Salicornia iranica dominated the early stages (years 1–4), but its proportion declined after year 6. Simultaneously, Juncus spp. increase notably, forming a significant share after year 10. The study provided a detailed plant species map and found patterns consistent with ecological succession in a drying wetland, using a reproducible framework that supports effective monitoring and management in similar ecosystems. • Species-level vegetation classification achieved an overall accuracy of 93.7%. • A reproducible framework to identify wetland drying age at 10 m spatial resolution was developed. • Clear vegetation succession patterns were observed across the wetland drying ages.