Spatiotemporal Evolution Characteristics and Ecological Restoration Effectiveness Assessment of Mangroves Based on Multi-Source Remote Sensing Data: A Case Study of Mangroves in Guangdong Province from 2000 to 2024

Mangrove communities possess exceptional capacities for soil and water conservation and ecological protection, and their successional dynamics have long commanded global attention. Guangdong Province, located along the southern coast of China, hosts some of the country’s most extensive mangrove stands. Leveraging the Google Earth Engine (GEE) platform, this study integrates Landsat-5/7/8/9 and Sentinel-2 multi-source imagery spanning 2000–2024 and incorporates multiple ecological indicators to construct a Stacking ensemble machine-learning model that blends Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Multi-Layer Perceptron (MLP). The model was applied to 32 mangrove nature reserves and six core distribution areas in Guangdong. Among the base learners, XGBoost exhibited the highest accuracy, achieving 96.98% on the validation set. The optimized Stacking model subsequently revealed that the total mangrove area in the study region expanded continuously from 3,024.72 ha in 2000 to 7,708.83 ha in 2024—a net increase of 4,684.11 ha and an average annual growth rate of 5.8%. Spatially, the mangrove landscape shifted from “scattered fragmentation” to “contiguous aggregation,” with mean patch size rising from 1.64 ha to 6.04 ha. Across the six representative core zones – Futian (Shenzhen), Qi’ao–Dangan Islands (Zhuhai), Zhanjiang, Huidong (Huizhou), Zhenhai Bay, and Cheng Village (Yangxi) – significant area increases (4.4–9.7% yr−1) and progressive patch enlargement were observed. The seaward margins of these mangroves advanced at mean rates of 7.6–15.2 m yr−1. These findings demonstrate the outstanding effectiveness of ecological restoration efforts in Guangdong’s mangroves and provide a quantifiable “Guangdong Model” for nationwide mangrove conservation and rehabilitation.

Перевод пока недоступен