NRWI: a novel spectral index optimized for waterbody extraction from high-resolution GF-2 satellite imagery

Remote sensing-based water indices are well-suited for rapid, large-scale waterbody extraction. However, most existing indices rely on short-wave infrared (SWIR) bands, making them incompatible with four-band high-resolution satellite imagery (visible and near-infrared only). Moreover, the few indic...

Full description

Saved in:
Bibliographic Details
Main Authors: Peng Zhang, Shanchuan Guo, Hong Fang, Pengfei Tang, Cong Lin, Hui Yang, Biao Wang, Yanlan Wu
Format: Article
Language:English
Published: Taylor & Francis Group 2025-08-01
Series:International Journal of Digital Earth
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/17538947.2025.2531846
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Remote sensing-based water indices are well-suited for rapid, large-scale waterbody extraction. However, most existing indices rely on short-wave infrared (SWIR) bands, making them incompatible with four-band high-resolution satellite imagery (visible and near-infrared only). Moreover, the few indices applicable to four-band data (e.g. the Normalized Difference Water Index) struggle to distinguish water from spectrally similar features such as buildings and shadows in complex environments. To address these limitations, we propose the Near-Infrared-RGB Water Index (NRWI), a novel spectral index optimized for four-band imagery. Systematic spectral analysis of 2,400 pure pixels across 12 land cover categories demonstrates NRWI’s enhanced ability to separate water from urban features (buildings/shadows). Validation using 1-m GF-2 imagery from 16 heterogeneous sites in Anhui Province, China, shows that NRWI achieves a mean Intersection over Union (IoU) of 0.788 (±0.096), outperforming NDWI by 0.211. Cross-sensor tests (GF-1, Sentinel-2) confirm NRWI’s adaptability, while comparisons with deep semantic segmentation models (UNet/DeepLab v3+) reveal a marginal accuracy gap (≤0.025) despite requiring zero training samples. NRWI provides a robust solution for rapid flood monitoring and precision water resource management.
ISSN:1753-8947
1753-8955