Cash before the flood, how AI can help get aid to at-risk populations

New model uses AI forecasts to trigger cash transfers before overflow, helping families evacuate, protect assets

With this year's floods devastating rural regions and displacing countless families, it is more important than ever to have systems in place that can anticipate and pre-emptively provide aid to affected populations.

The non-profit GiveDirectly is testing a new approach to flood relief, and that is to send money to households before water levels rise. Anticipatory cash assistance uses AI to forecast where floods will hit, and releases payments when certain risk thresholds are met.

GiveDirectly said it transferred US$105 to thousands of households in Nigeria within 48–72 hours of their forecasted flood thresholds in June, allowing families to evacuate, fortify homes and protect livestock ahead of inundation. The group also reported that "incomes more than doubled, food insecurity dropped by 90%, and most recipients felt better prepared for future floods".

Read: Record floods displace millions in Punjab

The model depends on Google’s Flood Hub, a forecasting platform that now covers over 80 countries and offers up to seven days of flood predictions. By leveraging Google's AI systems, aid agencies and governments can use it to receive village-level alerts, and Pakistan is among the countries covered.

Humanitarian agencies argue that paying early is both more dignified and more cost-effective than traditional aid that arrives after losses mount, and similar ideas have been piloted in Bangladesh and other regions, with studies suggesting anticipatory action can shorten recovery time and reduce humanitarian costs.

How Pakistan can implement a similar system

Pakistan has real-time river data through the Pakistan Meteorological Department's Flood Forecasting Division (PMD/FFD), and can use the Google Flood Hub (and its AI systems) to monitor parameters like predicted river discharge levels, inundation probabilities, and so on.

Through identifying and registering vulnerable households, like through the Benazir Income Support Program (BISP), the government can set up digital wallets to get funds to people as quickly as possible. BISP has already partnered with JazzCash for fund disbursements, so this can be adapted for flood relief.

Additionally, by connecting the PMDFFD's flood hub forecasts to Pakistan's Space and Upper Atmosphere Research Commission's (SUPARCO) Space Application Centre for Response in Emergency and Disasters (SACRED) satellites, we can refine the data for high accuracy and ensure precise targeting of settlements that are at risk.

By integrating “anticipatory action” into the National Disaster Management Authority's Proactive Monsoon Guidelines, and using the National Disaster Risk Reduction Strategy for 2025–2030, we can both plan for the floods and for taking care of our people before the floods, clarifying how at-risk people get payouts and the conditions for getting these payouts. There can also be a verification system to control errors and fraud integrated into the policy framework.

However, special attention would have to be given to people in areas that do not receive consistent mobile phone signals or those who are not digitally connected.

Trialling a system in Sindh’s lower Indus and southern Punjab, where flood risks are high, with the help of an independent partner, would be a good first step to a system that can save and support many.