Validation of the GOES-R ABI flood and standing water algorithm using gauging station measurements and interpretation maps

Validation is an important task in the development of satellite remote sensing products. Strategies for validation vary depending on the nature of the products. The validation process of the flood and standing water product (FSW) for the Geostationary Operational Environmental Satellite - R series (GOES-R) is presented in this paper. A major challenge in the validation of the FSW product is the lack of ground truth flood maps and similar reference products from other satellite systems and other sources. To overcome this limitation, a two-level validation scheme for the FSW product is developed using the Moderate-resolution Imaging Spectroradiometer (MODIS) data as a proxy. In the first level, gauging station data collected by the US. Geological Survey (USGS) are employed as ground truth flood point information on local scales to verify the effectiveness of the proposed algorithm for flood detection. Gauging station data collected during 34 flood cases that occurred in 2010 and 2011 in the continental US were validated and assessed according to the rate of correct detection. Results showed that 79.71% of flooding stations were accurately detected from the MODIS 1 km images by the proposed FSW algorithm. In the second level of the validation. FSW detection results using the proposed algorithm were compared to the reference flood maps, which were generated by a supervised support vector machine (SVM) classification followed by human interpretation and editing. Flood detection accuracies for three major flood events occurred in Asia and Australia in 2010 were evaluated. Confusion matrices were employed as the accuracy measurement for the second level of the validation. Commission errors for the three flood cases were 6.75%, 13.45% and 21.45%, respectively. Omission errors of flood pixels varied between 9.58% and 19.61%. The validation results suggest that the employed FSW algorithm is capable of producing flood and standing water maps in an operational environment, and it meets the required accuracy and execution time of the product. (C) 2012 Elsevier Inc. All rights reserved.