Estimating flood forecast performance using inundation data in Soroti, Uganda

Boelee, L. and Lumbroso, D. and Samuels, P.G. and Cloke, H. (2019) Estimating flood forecast performance using inundation data in Soroti, Uganda. In: AGU Fall Meeting 2019, 9-13 December 2019, San Francisco, USA.

PDF TP-064_AGU-2019-LGB-R1.pdf Download (972kB)

Abstract

This work explores the question “How can data on flood extents derived from Earth Observations (EO) be used to assess the performance of a global flood forecasting model in the ungauged catchment of the Okere and Okok Rivers in Uganda?”. The Global Flood Awareness System (GloFAS), jointly developed by the European Commission and the European Centre for Medium-Range Weather Forecasts (ECMWF), is a global hydrological forecast and monitoring system. In many parts of sub-Saharan Africa the performance of GloFAS has not been assessed. GloFAS is being used in some parts of Uganda to forecast floods. Recently Africa Risk Capacity has been developing a pan-African flood model for use in underpinning parametric flood insurance. The African Flood Extent Depiction Model (AFED) is a daily depiction of temporarily flooded areas everywhere in Africa over the past 20 years. The AFED uses satellite remote sensing from microwave sensors to map floods. The AFED data set was used to assess the performance of GloFAS for two rivers in Uganda. The AFED flood data consists of a flooded fraction per pixel which ranges from 0 to 1. This is not directly comparable to the river discharges produced by the GloFAS flood forecasting model. In order to compare both datasets and assess GloFAS’s performance, the following steps were taken: Extracting the flooded fraction of the Okok and Okere Rivers. Five methods were explored: Flooded fraction of the most downstream pixel; Catchment average flooded fraction for all non-zero pixels; Maximum flooded fraction in catchment; Number of pixels that are non-zero in the catchment; Sum of flooded fraction of all the pixels in the catchment. Comparing the recorded floods derived from newspaper articles with the EO data to establish if the AFED captures the flooding of the Okok and Okere Rivers. Establishing the range of the flood fraction that signifies flooding in recorded events. Extracting flood events using the peaks of the AFED data and the range of flooding from step 3. Assessing the performance of GloFAS and calculating its skill scores using this extended flood events. Results show that AFED data successfully identifies flooding for the two rivers and can be used to assess GloFAS’s performance.

Actions (for site administrators only - login required)

View Item