Hydro-meteorological data analysis for the characterisation of urban floods in Alexandria, Egypt

The city of Alexandria in Egypt and other Middle Eastern and North African (MENA) cities are at risk of flooding due to heavy rainfall, insufficient drainage capacity, and a lack of preparedness. Flood studies entailing detailed hazard modelling and risk analysis have agreeably become part of the blueprint for the selection and implementation of appropriate mitigation measures. Unfortunately many cities lack the data, resources and capacity to complete the required studies in a timely manner. Early warning systems using rainfall thresholds have been proposed as a cost-effective solution to increase preparedness. Several studies have successfully used rainfall thresholds in early warning system as a predictors to inundation and thus reducing immediate data requirements, complexity and computation time associated with flood modelling. When a probabilistic approach is considered, there is the advantage of considering the likelihood of occurrence of an event. This study aimed to use such an approach to characterise and predict urban floods by assessing rainfall threshold likely to cause flooding combined with ensemble precipitation forecast in Alexandria Egypt. Firstly the points over threshold method was used to evaluate the rainfall frequency of extreme events and found when compared to other studies this method yielded higher estimates for the 2yr return period which the current sewer system is designed. Rainfall thresholds were inferred by associating observed rainfall and historical flood information. These thresholds we then adjusted based on, extreme rainfall frequency and local drainage capacity. ECMWF, TIGGE ensembles were bias corrected and verified using various statistical measures for different lead times before being compared against rainfall thresholds. Results of this analysis and the probability of occurrence was then incorporated into a colour- coded hazard matrix to jointly evaluate likelihood and hazard severity. Events were evaluated for the period 2013-2015 and it was found that 3 out of 4 severe events analysed could have been predicted with a high likelihood of significant or severe events up to 48 hrs and 24hrs respectively before the event occurred. Meaning, had such a system been implemented, warnings and adequate measures could have confidently been implemented to prevent damage and loses associated with those events. However, the results of the ensemble verification and bias correction method indicated that while the bias was reduced there is the potential to improve results using an alternative bias correction method. The research concluded, while data was limited this practical approach was able to characterise and give warnings of possible floods. This inevitably contributes toward resilience building and increased preparedness which can be a model for upscaling in the region