Forecasting Floods: A Global Challenge

Floods, the most prevalent natural disaster, wreak havoc worldwide, causing approximately $50 billion in annual damages. This figure, sadly, is not static; with climate change as a significant contributing factor, flood-related disasters have more than doubled since 2000. A staggering 1.5 billion people – nearly a fifth of the world’s population – face substantial risk from severe flooding. Improving early warning systems is not just a technological challenge; it’s a humanitarian imperative. Providing accurate and timely information can be the difference between life and death for thousands.

At ExecuteAI, we believe in the power of AI not to replace human ingenuity, but to amplify it. In 2017, driven by this belief, we began exploring how AI could make flood forecasting more reliable and accessible. This has been a multifaceted journey, involving continuous research and development, culminating in a real-time operational system providing alerts through Google Search, Maps, Android notifications, and our dedicated Flood Hub.

Scaling Globally with Machine Learning

Scaling flood forecasting globally, especially in regions where reliable local data is scarce, requires significant innovation. Our recent research, published in Nature, details how machine learning (ML) is revolutionizing flood prediction. We’ve significantly improved global-scale forecasting, particularly in data-sparse regions across Africa and Asia. Think of it like this: we’ve effectively extended the reliability of global nowcasts by an average of five days in these areas, bringing them closer to the level of forecasting currently enjoyed in Europe.

These advancements, validated in collaboration with the European Centre for Medium-Range Weather Forecasts (ECMWF), empower our Flood Hub to offer real-time river forecasts up to seven days in advance, spanning over 80 countries. This information is invaluable for individuals, communities, governments, and international organizations in taking proactive measures to protect vulnerable populations.

From Pilot Project to Global Impact

Our journey began with a pilot early warning system in India’s Ganges-Brahmaputra river basin. We hypothesized that ML could tackle the complexities of reliable flood forecasting at scale, and thankfully, the results spoke volumes. This pilot expanded, integrating real-time water level measurements, an elevation map, and hydrologic modeling. Collaboration with academic institutions, particularly the JKU Institute for Machine Learning, led to the development of sophisticated LSTM-based models, exceeding the accuracy of traditional hydrological models. This progress has extended our forecasting coverage to encompass all of India and Bangladesh.

The Power of Data and Open Science

A significant hurdle in flood forecasting is the lack of reliable data in many regions. Ironically, areas with lower GDPs are often most vulnerable to flood risks, yet also have the least amount of publicly available data. It’s a bit like trying to bake a cake without having all the ingredients – challenging, to say the least. This is where ML truly shines. By training a single model on all available river data, we can apply it to ungauged basins, essentially creating a recipe that can be adapted even with missing ingredients.

In the spirit of open science, we released a community-driven dataset for large-sample hydrology in Nature Scientific Data in 2023. This kind of data sharing is crucial for collective progress and building a more resilient future for all. It also ensures we aren’t all reinventing the wheel, which is a relief, especially as round ones work best, wouldn’t you agree?

Looking Ahead: A Collaborative Future

Our work demonstrates the transformative potential of AI in flood forecasting. However, the journey doesn’t stop here. We envision a future where even more precise forecasts are available across the globe, covering various flood-related events, from flash floods to urban inundations. We believe this future can only be achieved through continued collaborations with academic and expert communities, local governments, and industry partners. After all, as they say, a rising tide floats all boats (though hopefully not too high!).