Unveiling seaward expansion pattern in mangrove forests using UAV remote sensing and deep learning
Zhi Zhang, Xiaoxue Shen, Chunhua Yan, Ruili Li, Bing Li
Abstract
• An integrated remote sensing and deep learning framework was developed for automated species identification and canopy parameter extraction in mangroves. • The competition and succession dynamics of mangrove expansion were quantified. • Sequential species replacement and competition-driven community succession were identified as key characteristics of mangrove seaward expansion. • Novel integrated framework for mangrove monitoring. Mangrove wetlands in China are naturally expanding seaward due to rapid sediment accretion, yet the dynamics of species competition and stand structure during this expansion remain unclear. Here, we developed an integrated remote sensing framework combining satellite imagery, unmanned aerial vehicle (UAV) data, and deep learning to investigate species composition and stand structure in China’s largest contiguous mangrove forest. The algorithm achieved high accuracy in identifying two dominant species ( Aegiceras corniculatum and Avicennia marina , overall accuracy: 87.7 %) and extracting individual crown parameter (precision > 70 %). Results show that A. corniculatum initially colonizes new mudflats but is later outcompeted by A. marina , forming monospecific stands over time. Interspecific competition intensity peaks during early succession (1–7 years) and declines with stand age, accompanied by a shift in spatial distribution from random to uniform. These findings reveal a distinct pattern of sequential species replacement and community succession during mangrove seaward expansion. The proposed framework and ecological insights provide valuable guidance for near-natural mangrove afforestation and the restoration of degraded coastal ecosystems, contributing to sustainable wetland conservation strategies.