Improving Hydrologic Modeling Using Cloud-Free MODIS Flood Maps
Abstract
Flood mapping from satellites provides large-scale observations of flood events, but cloud obstruction in satellite optical sensors limits its practical usability. In this study, we implemented the Variational Interpolation (VI) algorithm to remove clouds from NASA’s Moderate Resolution Imaging Spectroradiometer (MODIS) Snow-Covered Area (SCA) products. The VI algorithm estimated states of cloud-hindered pixels by constructing three-dimensional space–time surfaces based on assumptions of snow persistence. The resulting cloud-free flood maps, while maintaining the temporal resolution of the original MODIS product, showed an improvement of nearly 70% in average probability of detection (POD) (from 0.29 to 0.49) when validated with flood maps derived from Landsat-8 imagery. The second part of this study utilized the cloud-free flood maps for calibration of a hydrologic model to improve simulation of flood inundation maps. The results demonstrated the utility of the cloud-free maps, as simulated inundation maps had average POD, false alarm ratio (FAR), and Hanssen–Kuipers (HK) skill score of 0.87, 0.49, and 0.84, respectively, compared to POD, FAR, and HK of 0.70, 0.61, and 0.67 when original maps were used for calibration.
Document Details
- Document Type
- Pub Defense Publication
- Publication Date
- Nov 01, 2019
- Source ID
- 10.1175/jhm-d-19-0021.1
Entities
People
- Hoang Tran
- Konstantinos M Andreadis
- Kuolin Hsu
- Mohammed Ombadi
- Phu Nguyen
- Soroosh Sorooshian
Organizations
- Army Research Office
- California Energy Commission
- National Oceanic and Atmospheric Administration
- National Science Foundation
- United States Department of Energy
- University of California
- University of California, Irvine
- University of Massachusetts Amherst