Data Analytics for a Changing Climate: Feature Engineering for the forecast of hydrometeorological events

Floods are becoming increasingly frequent, and consequently, the number of people and infrastructure affected by these events has increased. It is essential to have accurate models for the prediction of such hydrometeorological events, improving preparedness and decision making for damage reduction. The goal of this work was to determine the variables (features) that contribute the most to predicting hydrometeorological events. Feature engineering techniques were used to understand which factors are most helpful in predicting floods. The features were composed based on data from rain gauges, altimetry, and location of rivers and lakes. It was observed that the variables that had the greatest impact on improving the model were the data from rain gauges and altitude data. The predictive model proposed is part of a larger system being developed in the context of Smart Cities called ICARUS. The system is aimed at improving response time and up-time of critical infrastructure during extreme events.