A new satellite-based technology called the Transient Artifact and Continuous Learning System (TACLS) is being developed to provide earlier warning of flash floods in the United States. The system uses machine learning to detect areas at risk of flooding sooner, potentially allowing the National Weather Service to issue alerts before communities are already inundated—a critical improvement, as floods are the second-deadliest weather event in America and increasingly common due to climate change. The motivation comes from cases like Laura Lin's experience in Indiana, where her town received over 20 centimetres of rain in hours, but warnings only arrived after residents were already flooded.

The current flood warning system relies on rain and stream gauges, satellite imagery, weather radar, and the expertise of staff across 122 forecast offices throughout the US and its territories. Forecasters manually assess rainfall amounts against pre-calculated thresholds based on local soil type and slope steepness to decide whether to issue alerts. However, this process is often too slow; just 15 centimetres of fast-moving water can knock over an adult, whilst 30 centimetres can displace larger vehicles. TACLS aims to streamline this process by using artificial intelligence to identify vulnerable areas automatically, enabling quicker decision-making and potentially saving lives.