The HydroFrame is a community platform that facilitates integrated hydrologic modeling across the United States. We design innovative workflow solutions that create pathways to enable hydrologic analysis for three target uses: modeling, analysis, and domain science. As part of our contribution to HydroFrame, we run HydroFrame workflows in the Kepler system, utilizing its automated workflow capabilities to perform end-to-end hydrology simulations involving data ingestion, preprocessing, analysis, modeling, and visualization.

Our focus helps ensure complete reproducibility of the workflows through the collection of provenance data, including input data and workflow provenance, using the built-in provenance framework of the Kepler system. We leverage machine learning techniques to predict workflow performance and forecast failures using an automatic performance collection component of the pipeline. This design optimizes the utilization of large-scale computational resources to adjust to the needs of all three user groups.