Surface Water Extraction From Landsat 8 OLI Imagery Using the LBV Transformation

Surface water extraction from remote sensing images is crucial for monitoring water resources and studying global environmental change. This study introduced amethod for surface water extraction based on the LBV transformation (WE-LBV) from Landsat 8 Operational Land Imager (OLI) imagery. To avoid inadequate or redundant utilization of remote sensing data, input band combinations were selected using the band index method, and the LBV transformation equations for OLI images were subsequently derived. Characteristics of water and nonwater pixels were investigated after the LBV transformation. The established extraction rule stated that pixels should be classified as water if the B value is larger than the V value. For validation, the WE-LBV was used to extract water pixels from OLI images covering different water types and outside conditions and compared with other water indices on the global scale. Results showed that the producer accuracy (95.75%) and user accuracy (99.15%) of the WE-LBV were more stable and its overall accuracy (98.02%) and Kappa coefficient (0.9582) were higher than those of other methods. The WE-LBV classified shallow water and shadow pixels accurately, whereas other methods had problems. Water pixels with different colors and mixed pixels with water percentage higher than 80% could be extracted by the WE-LBV effectively. The WE-LBV did not require manually selected threshold, thus avoiding the effect of subjective human factors. Therefore, the WE-LBV is proposed as an accurate, simple, and robust method for surface water extraction, especially in situations when surface water needs to be identified in bulk automatically and accurately.