The Conversation has posted an article discussing the challenges and benefits of developing a machine learning-based method to create street-by-street flood hazard models. This model would be extremely beneficial to local emergency managers and civilians operating during a flood by informing them of which roads are too dangerous to use. However, rainwater prediction programs require high levels of processing power and accuracy in precipitation forecasts is difficult. The writer describes how their group combined machine learning with a sophisticated flood model to train simpler models that capture a single variable. These models would be prepared ahead of any event, yet would be easy for the average computer to run on demand once flooding occurred. While the author has previously demonstrated the usefulness of this method in real-world conditions, there is still work to be done to deploy this method widely. Read more at The Conversation.