Observation Definitions and Their Implications in Machine Learning-Based Predictions of Excessive Rainfall

The implications of definitions fi nitions of excessive rainfall observations on machine learning model forecast skill are assessed using the Colorado State University Machine Learning Probabilities (CSU-MLP) forecast system. The CSU-MLP uses historical observations along with reforecasts from a global ensemble to train random forests to probabilis- tically predict excessive rainfall events. Here, random forest models are trained using two distinct rainfall datasets, one that is composed of fi xed-frequency (FF) average recurrence intervals exceedances , fl ash fl ood reports and the other a com- pilation of fl ooding and rainfall proxies [Unified fi ed Flood Verification fi cation System (UFVS)]. Both models generate 1-3-day fore- casts and are evaluated against a climatological baseline to characterize their overall skill as a function of lead time, season , region. Model comparisons suggest that regional frequencies in excessive rainfall observations contribute to when and where the ML models issue forecasts and subsequently their skill and reliability. Additionally, the spatiotemporal distribu- tion of observations has implications for ML model training requirements, notably, how long of an observational record is needed to obtain skillful forecasts. Experiments reveal that shorter-trained UFVS-based models can be as skillful as longer-trained FF-based models. In essence, the UFVS dataset exhibits a more robust characterization of excessive rainfall and impacts, and machine learning models trained on more representative datasets of meteorological hazards may not re- quire as extensive training to generate skillful forecasts.