Submitted to: Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/18/2019
Publication Date: 1/27/2019
Citation: Belabid, N., Zhao, F., Brocca, L., Huang, Y., Tan, Y. 2019. Near-real-time flood forecasting based on satellite precipitation products. Remote Sensing. 11(252):1-18. https://doi.org/10.3390/rs11030252.
DOI: https://doi.org/10.3390/rs11030252

Interpretive Summary: Floods, storms and hurricanes are devastating natural disasters for human life and agricultural land. Accurate, timely flood forecasting is crucial to reduce damages from the disaster. In flood forecasting estimation of precipitation is a key component. Scientists of Beihang University, National Research Council of Italy, and USDA ARS Crop Production Systems Research Unit at Stoneville, Mississippi have collaboratively further developed a new approach to precipitation forecasting when the traditional method using the ground precipitation measuring stations is not efficient. This approach uses satellite precipitation products to cover large areas with high frequency of precipitation recordings. This study evaluated the approach with the data for the Ottawa watershed in Canada in the period from April to May 2017. The results show that with the satellite precipitation products the new approach produced the predicted values that matched the observed values overall and the developed model reproduced the shape of the observed data graphs. This study provides useful information for near real time discharge estimation to forecast flood disaster.

Technical Abstract: Floods, storms and hurricanes are devastating for human life and agricultural cropland. On flood disaster, near real time discharge estimation is crucial to avoid the damages. The key input data for discharge estimation is precipitation. Using the traditional ground stations to measure precipitation is not efficient especially during a huge rainstorm because precipitation varies even in the same region. This uncertainty might affect flood discharge estimation and forecasting models. For this reason, the use of satellite precipitation products (SPPs) provides a large area coverage of rainstorms, and higher frequency of precipitation data. In flood disaster reduction management, space technologies intervene mostly on emergency response, by providing earth observation satellites images acquired during the flood event. Analysts use visual interpretation or change detection methods for damage assessments by processing images acquired before and after flood disaster. In this paper a Near-Real Time (NRT) flood forecasting approach is proposed, based on SPPs permitting the use of space technologies more efficiently, by using satellite precipitation measurement from space. SPPs can cover large regions providing more frequent and accurate products around the world by using instruments designed to observe water in the atmosphere. The use of NRT remotely sensed precipitation products decreases the time of emergency response to flood disasters, saving human life and limiting damage caused by floods. In this study, we propose the use of new approach based on SPPs that gives the possibility to forecast flood using discharge hydrograph, then uses the results on flood extent mapping, by introducing SPPs into Rainfall-Runoff model named “Modello Idrologico Lumped in continuo” (MILc). In this research, firstly we evaluate the capacity of SPPs on flood discharge estimation, then their accuracy on flood extent mapping. Two high temporal resolution NASA’s SPPs were compared: 1. Integrated Multi-satellitE satellite Retrievals for Global precipitation measurement (IMERG), the real time version 3IMRGHH.05 product; 2. Tropical rainfall measurement mission Multisatellite Precipitation Analysis (TMPA), the real time version (3B42-RT). The two products are evaluated in the same study area, Ottawa watershed in Canada, and for the same period from 10 April to 10 May 2017. For TMPA, the results show that the comparison between observed and modeled discharges is not close to nominal value, i.e 0.5, with Nash-Sutcliffe efficiency (NSE) equal to -0.9241 and Adapted NSE (ANSE) during high flow conditions equal to -1.0048. The model does not reproduce the shape of the observed hydrographs. However, in case of IMERG, the difference between observed and modeled discharges is smaller, with NSE equal to 0.80387 and ANSE 0.82874. The model can reproduce the shape of the observed hydrographs, mainly during high flow conditions, which are interesting for SPPs being used as input data. After that, the modeled discharge hydrograph based on IMERG products is used as input on flood extent mapping. The results of the modeled flood extent map show that the error is mostly under one pixel compared with the observed flood benchmark data of Ottawa river acquired by RADARSAT-2 during the flood event. The developed flood forecasting approach based on SPPs offers a solution on flood disaster management for poorly or totally ungauged watershed regarding precipitation measurement. These findings could be referred to by others for NRT flood forecasting research and applications.