Year: 2025
Type: article
Abstract: High-intensity human activity zones disproportionate pressure on ecologically sensitive ecosystems, accelerating landscape fragmentation and the degradation of ecosystem functions. Although existing land cover maps capture broad ecosystem patterns, they frequently lack the thematic granularity and spatial fidelity needed for fine-scale ecological research and management. This study proposes a hierarchical-modular ecosystem mapping framework. Its core first combines a hierarchical classification strategy (HCS) with probabilistic ensemble methods (PEM) to maximize class separability and reduce model uncertainty, increasing overall accuracy (OA) by 2.50-5.41 percentage points. A subsequent refinement step integrates a pixel- and object-based with knowledge (POK) algorithm, drawing on phenological, morphological, and locational rules, to enhance ecosystems delineation – focusing on forestland and wetland subtypes. Applying this framework, we generated the 2020 Yangtze River Delta ecosystem map (YRD_EM). Results show YRD_EM achieves 90.29% OA for 10 major ecosystem categories (e.g. forestland, irrigated cropland, marshland) and 88.50% OA for 15 fine categories (e.g. evergreen/deciduous forest, river/lake, inland/coastal marsh). Compared with existing products, YRD_EM significantly improves wetland connectivity representation and shrubland detection, and delineates ecologically complex transition zones more accurately. This framework shows potential for ecosystem mapping and monitoring, providing better data for ecosystem conservation planning.
Authors: Jingda Xin, Danyang Wang, Yunqi Wang, Yayi Tan, Jiaying Zheng, Xibao Xu, Yihan Lu, Weijie Huang, Zeqi Wei, Huan Zhang, Changbo Zhong, Zhaofu Li
Language: English
Subfield: Media Technology
Field: Engineering
Domain: Physical Sciences
SDG: Life on land
Open Access status: gold
Estimated APC paid: $2,390
Awards: G2443011419, G7431566414
OpenAlex ID: W4414451276