A generalizable framework for urban wetland training samples generation and migration: A case study of global Ramsar Wetland Cities

A generalizable framework for urban wetland training samples generation and migration: A case study of global Ramsar Wetland Cities

Accurate urban wetland mapping requires reliable training samples, yet the cost reduction and efficiency enhancement of sample production in complex urban backgrounds remains challenging. This study introduces an automated framework Fusion Knowledge Rules and Spectral Matching (FKRSM) for training s...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhe Yang, Weiguo Jiang, Xiaogan Yin, Ziyan Ling, Xiaoya Wang, Miaolong Lin, Shuhui Lai, Xiao Li, Qiaozhen Guo, Zhijie Xiao, Ze Zhang, Qiuling Li, Peiyu Yang, Shihui Huang, Xiang Long, Keyi Yang, Kaifeng Peng, Yongbiao Yu, Xuan Liu, Yaheng Sheng, Xiaorui Ren, Xiangdong Yang, Haicheng Tian
Format: Article
Language:English
Published: Elsevier 2025-09-01
Series:Ecological Indicators
Subjects:
Urban wetland sample
Knowledge rules
Spectral matching
Google Earth Engine
Ramsar Wetland City
Online Access:http://www.sciencedirect.com/science/article/pii/S1470160X25010106
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurate urban wetland mapping requires reliable training samples, yet the cost reduction and efficiency enhancement of sample production in complex urban backgrounds remains challenging. This study introduces an automated framework Fusion Knowledge Rules and Spectral Matching (FKRSM) for training sample generation and migration. Implemented on Google Earth Engine (GEE), FKRSM applies dense Sentinel-2 time series and multi-source classification products to extract vegetation-hydrology dynamics, inundation frequency patterns, and geometric attributes of urban wetlands. A hybrid strategy combining index-threshold reclassification with morphological purification is used to delineate class-specific sample generation zone and to generate corresponding samples. A dual-constraint spectral matching method based on Spectral Angle Distance (SAD) and Euclidean Distance (ED) was developed to reduce the manual effort required for determining unchanged sample thresholds and to enable dynamic migration of urban wetland training samples. Tested across 43 Ramsar Wetland Cities (RWCs), FKRSM achieved automatic generation of 726,482 samples in the 2022 reference year with a sampling accuracy of 97.72 %. Migrated samples across 2016, 2018, 2020, and 2024 maintained an average accuracy of 92.53 %. Compared with recent research, FKRSM achieved 216.85 % of the performance (production time and classification accuracy) of prior methods in China’s first batch of 6 RWCs. FKRSM, with its demonstrated spatiotemporal generalizability, offers a scalable solution for ongoing fine-scale urban wetland mapping and further supports both six-year RWC re-accredited and new city accreditations.
ISSN:1470-160X

Similar Items