Global Surface Water Delineation under Heterogeneous Backgrounds Based on A Fuzzy Clustering Approach

Abstract:

An accurate and robust method to extract terrestrial water bodies is critical to effectively manage the fundamental resources for terrestrial life. Conventional methods are frequently limited in terms of the uncertainty related to the coarse resolution of remotely sensed images and insufficient robustness caused by regional reflectance heterogeneity. Fuzzy clustering method (FCM) with local spatial information has proven to be capable of accounting for these limitations. This technique is however susceptible to immense false noise in original optical bands. A systematic surface water extraction method by synthesizing the water index (WI) method and a modified FCM (WIMFCM) was therefore proposed to improve the water extraction accuracy, the rationale of which is a background reflectance bias correction. Applications and validations were performed to sixteen background-heterogeneous sites in different parts of the world using the Landsat-8 OLI (Operational Land Imager) images, which differed from previous research focusing on water bodies with specific morphological features (e.g., linear rivers or regular lakes). Results obtained demonstrated that the use of the proposed WIMFCM could improve the accuracy of surface water body extraction without regional or seasonal limitations in comparison to alternative methods at both pixel and sub-pixel level. Additional tests carried out on MODIS (Moderate-resolution Imaging Spectroradiometer) data proved that the WIMFCM could be extended to global scale, as well as be used in near-real time water monitoring. The findings of this study are able to improve land cover mapping accuracy when using the optical satellite images under heterogeneous environments.