Potential of remote sensing and open street data for floodmapping in poorly gauged areas: a case study in Gonaives, Haiti

摘要

The Hispaniola Island, in the Caribbean tropical zone, is prone to extreme flood events. Floods are caused by tropical springs and hurricanes and may lead to human losses, economical damages, and spreading of waterborne diseases. Flood studies based upon hydrological and hydraulic modelling are hampered by almost complete lack of hydrometeorological data. Thenceforth, and given the cost and complexity in the organization of field measurement campaigns, the need for exploitation of remote sensing data, and open source data bases. We present here a feasibility study to explore the potential of (i) high-resolution of digital elevation models (DEMs) from remote imagery and (ii) remotely sensed precipitation data, to feed hydrological flow routing and hydraulic flood modelling, applied to the case study of river La Quinte closed to Gonaives (585 km~2), Haiti. We studied one recent flood episode, namely hurricane Ike in 2008, when flood maps from remote sensing were available for validation. The atmospheric input given by hourly rainfall was taken from downscaled Tropical Rainfall Measuring Mission (TRMM) daily estimates, and subsequently fed to a semi-distributed DEM-based hydrological model, providing an hourly flood hydrograph. Then, flood modelling using Hydrologic Engineering Center River Analysis System (HEC-RAS 1D, one-dimensional model for unsteady open channel flow) was carried out under different scenarios of available digital elevation models. The DEMs were generated using optical remote sensing satellite WorldView-1 and Shuttle Radar Topography Mission (SRTM), combined with information from an open source database (OpenStreetMap). Observed flood extent and land use have been extracted using Système Pour l'Observation de la Terre-4 (SPOT-4) imagery. The hydraulic model was tuned for floodplain friction against the observed flooded area. We compared different scenarios of flood simulation and the predictive power given by model tuning. Our study provides acceptable results in depicting flooded areas, especially considering the tremendous lack of ground data, and shows the potential of hydrological modelling approach fed by remote sensing information in Haiti, and in similarly data-scarce areas. Our approach may be useful to provide depiction of flooded areas for the purpose of (i) flood design for urban planning under a frequency-driven approach and (ii) forecasting of flooded areas for warning procedures, pending availability of weather forecast with proper lead time.