Solar Forecasts from the 3-km High-Resolution Rapid Refresh

The High-Resolution Rapid Refresh (HRRR) model is now being run operationally at the National Centers for Environmental Prediction (NCEP) in Silver Springs, Maryland. A more advanced version of the model is run hourly in real time at the Global Systems Division (GSD) of the Earth System Research Laboratory (ESRL) in Boulder, Colorado. Since 2013, the GSD version of the model has played an important role in the Department of Energy (DOE) Solar Forecast Improvement Project (SFIP). In this paper, we summarize some improvements to the HRRR that have been achieved through SFIP.

Numerous changes have occurred to the data assimilation system associated with the HRRR. We have incorporated hybrid ensemble-3DVAR data assimilation within GSI, resulting in an improved initial model state. Also, we have introduced a requirement for the conservation of virtual potential temperature in the model column, in order to improve sounding (and therefore cloud) structures in the low levels. We have also conducted initial tests with assimilating two new satellite data products from NESDIS, with the goal of increasing our use of satellite cloud data for clouds with tops greater than 1200 m above ground level without increasing our relative humidity bias.

Within the model itself, much work over the last six months has dealt with a longstanding warm and dry bias in the HRRR, which was found to be partly due to excess incoming shortwave radiation. The RRTMG radiation scheme was introduced, allowing for more complex cloud-radiation interactions. The scheme allows a climatological aerosol loading which reduces the incoming shortwave radiation. The introduction of the Grell-Freitas convection scheme in the RAP has improved the distribution of low-level clouds (particularly shallow cumulus) in the forecasts, further reducing the shortwave radiation reaching the ground. Finally, we have also added a namelist option allowing a smoother diurnal transition in the solar zenith angle, resulting in a more realistic diurnal radiation curve.