City of Pemba: development of an automatic prediction tools for pluvial hazard assessment

Mozambique is one of the countries in Africa most seriously affected by tropical cyclones, which bring heavy rains and flooding causing severe damage and often exacerbating human-driven emergencies. Coastal cities, like Pemba (Cabo Delgado), are exposed to cyclone-triggered urban flooding events. 

In the framework of the ECHO funded project “REDE-EDUCAMA Disaster Reduction and Education in Cabo Delgado and Manica)”, an innovative (open-source) hydraulic modelling tool was adapted to recreate flooding scenarios caused by heavy rainfall in the peninsula of Pemba (85 square kilometres) with the aim of identify the area most prone to pluvial flooding and implementing an operational tool to inform Disaster Risk Management Authorities (DRMA) with reliable forecasts to issue timely early warnings (EWs) in case of cyclones and heavy rain affecting this area. For improving the sustainability of the tool, the operational chain implemented in co-operation with local authorities, is based on the use of open-source free software and models. The hydrodynamic model of rainfall-runoff (Broich et al., 2019), available in Telemac-2D and adapted to deal with time-variant grid-based rainfall input, was used.  

A preliminary collection of available data was carried out for the definition of the inputs needed to feed the model: a topographical base-map and precipitation. The map was derived integrating the results of a high-resolution drone survey (performed together with local authorities on 14 km2) with the Copernicus DSM satellite product (30m), to ensure the hydrological continuity needed.  

Concerning the rainfall input, the historical precipitation data series from the Pemba weather station, provided by INAM Cabo Delgado was analysed to identify the maximum rainfall depth for certain hourly intervals (24, 48 and 72 hours). Following this analysis,  33 rainfall events (hyetographs), different in timing and intensity, were generated and used to feed the ponding model, to produce 33 urban flooding scenarios. For warning purposes, 2 representation modalities of the outputs were investigated: a 200-metre grid aggregation (selecting medium-high percentiles) and a neighborhood-scale aggregation (selecting high percentiles and using the neighborhood map provided by the Municipality of Pemba). 

Modelled inundation maps were shared and commented with the local community in Pemba, with the dual objective of receiving feedback and increasing flood risk awareness. 

The full pluvial flooding forecasting chain for the Pemba urban area was then operationally implemented by connecting the flooding scenarios with the operational weather forecasts, by means of FloodPROOFS open-source modelling system (https://github.com/c-hydro). Daily forecasts of rainfall over Pemba are extracted from freely available global models (GFS 0.25), considering a set of pixels surrounding Pemba to account for uncertainty. A tailored tool connects the forecast rainfall with the most similar rainfall scenario, activating the corresponding urban flooding scenario, was developed. Operational forecasts are made available to DRMA officers through the www.myDEWETRA.world EW platform. 

The application in Pemba demonstrated the goodness of the approach based on innovation and co-operation with local authorities, enabling the replication on other cities of the country.