Merging HRRR Output into a Real-Time Gauge-Based Ensemble CONUS-Wide Dataset of Gridded Meteorological Fields

Gridded precipitation and temperature data products are uncertain due to a myriad of factors, including interpolation from a sparse observation networks, measurement representativeness and measurement errors. A lack of quantitative uncertainty estimates for hydrometeorological forcing fields limits the application of advanced data assimilation systems and other methods in land surface and hydrologic modeling.

We present a high-quality gridded, multi-year, sub-daily ensemble meteorological dataset for the contiguous US (CONUS) domain that improve upon prior gage-based ensemble datasets by incorporating model output fields from the High Resolution Rapid Refresh (HRRR) weather forecast system, using a spatial regression framework. The prior dataset used only static terrain features (aspect, elevation, slope, location) to predict time-varying local temperature and precipitation. This dataset augments these predictors using concurrent time-varying surface and upper atmosphere fields from HRRR analyses and nowcasts, essentially fusing gage observations and model-based estimates. The capability, which is provided in the Gridded Ensemble Meteorology Tool (GMET), can more generally merge in any gridded and point datasets. We report on progress and show early analyses of this dataset, which is designed to support studies of hydroclimatic variability, climate downscaling, hydrologic modeling and prediction, and infrastructure design and risk assessment.