Improving the interpretation of flood forecasts to support humanitarian operations

Researchers including Fathom’s Dr Jeff Neal, Dr James Savage and Dr Andrew Smith  recommend how best to improve the production and use of global flood forecasts during hazard events.

Forecasting plays a crucial role in coordinating humanitarian response to hazards such as tropical cyclones and floods – especially when disasters occur in regions lacking the local capacity or information for decision-making.

This study presents the results of cyclone forecasting provided by a multidisciplinary team at the UK Foreign and Commonwealth Development Office (FCDO), a consortium including Fathom, for two cyclones: Hurricane Iota in Central America (2020) and Cyclone Eloise in Mozambique (2021). 

It presents the successes and challenges experienced during both, and concludes with recommendations to improve future forecasts.

The disasters: Hurricane Iota and Cyclone Eloise

Hurricane Iota hit Nicaragua in November 2020. Humanitarian response for Iota was combined with aid for a hurricane that hit a nearby region just two weeks earlier; together, the two events impacted 5.2 million people, causing loss of life and damage to property and infrastructure.

Tropical Cyclone Eloise hit Mozambique in January 2021, impacting an area still recovering from a) a tropical storm that hit three weeks prior, and b) Cyclone Idai which hit in 2019. Eloise brought significant winds, rainfall, and river flooding, impacting nearly 500,000 people.

The results: Reflecting on wins and losses

The FCDO forecasters identified Iota as a potentially severe flood hazard five days before landfall, with the first bulletin produced four days later. Reflecting on Hurricane Iota, the researchers identify that:

• Despite the potential of flood models to produce earlier bulletins, these would have been unlikely to change the timing of actions taken by humanitarians; however, a short written forecast delivered in the lead-up to the event was valuable as an early indication of potential impact.
• Landslides and surface water and mountainous river flooding, as seen during Iota, are poorly represented by the chosen model (GloFAS) but of concern to responders; focusing on access roads offered a pragmatic solution.
• Real-time weather data, satellite imagery, and news reports were used as a ‘sense check’, enabling forecasters to include an assessment of forecast confidence alongside their bulletins.

For Cyclone Eloise, there was already an active humanitarian presence in the area. Eloise’s potential for ‘impactful flooding’ was identified nine days before landfall, and the first full bulletin warning of high winds, heavy rainfall, and coastal flooding issued the day before. Reflecting on Cyclone Eloise, the researchers identify that:

• The aid response benefited from existing relationships with organizations and humanitarian teams on the ground, which have strengthened since Hurricane Idai.
• Flexibly providing contextual information to users – for instance, how potential flood extent compared to that brought by Idai – enabled teams to better understand the level of response needed.
• Combining global and local forecasts for relevant rivers provided more reliable flood peak predictions.

The recommendations: More cooperation, more data, more user support

The researchers propose ways to improve the global forecasting systems involved in disaster prediction and response. These include:

• Increased cooperation and partnership, with improved synergy between global models and local contexts;
• Additional data and expertise to support how global models are interpreted;
• Clear documentation to support decision-makers juggling multiple sources of information; 
• Development of user-relevant metrics to assess the skill and trustworthiness of global models, so users can feel confident in their conclusions.