Skillful prediction of high-impact rainfall and streamflow events at lead times effective for proper hazard mitigation remains a significant challenge in nearly every region of the world. Additionally, rapid landscape change and shifting hydroclimate systems further complicate prediction problems as they reduce or even eliminate statistical stationarity assumptions upon which historical and current hydrologic prediction is based.
Over the last decade, sophisticated modeling systems within the disciplines of hydrology and meteorology have emerged, however, their use in integrated, cross-discipline prediction system development by research and operational agencies remains limited.
Challenges in current prediction systems include narrowly focused model development efforts, limited data discovery opportunities, labor intensive pre- and post-processing efforts, and sever limitations in community-wide access to sufficient computational capacity. To improve predictions, two major challenges must be addressed in order to bring state-of-the-art disciplinary science into inter-disciplinary predication practice:
- Scientific: Complexities of multiple physical processes occurring at a range of spatial and temporal scales require new approaches to predictive modeling of high impact hydrometeorological events
- Cyberinfrastructure: Heterogeneous data structures and semantics limit efficient discovery, processing and ingest of critical environmental data. There is significant unrealized potential in distributed (cloud) computing resources for predictive modeling