The High-Resolution Rapid Refresh (HRRR) is a convection-allowing implementation of the Weather Research and Forecasting model (WRF-ARW) with hourly data assimilation that covers the conterminous United States and Alaska and runs in real time at the NOAA National Centers for Environmental Prediction. Implemented operationally at NOAA/NCEP in 2014, the HRRR features 3-km horizontal grid spacing and frequent forecasts (hourly for CONUS and 3-hourly for Alaska). HRRR initialization is designed for optimal short-range forecast skill with a particular focus on the evolution of precipitating systems. Key components of the initialization are radar-reflectivity data assimilation, hybrid ensemble-variational assimilation of conventional weather observations, and a cloud analysis to initialize stratiform cloud layers. From this initial state, HRRR forecasts are produced out to 18 h every hour, and out to 48 h every 6 h, with boundary conditions provided by the Rapid Refresh system. Between 2014 and 2020, HRRR development was focused on reducing model bias errors and improving forecast realism and accuracy. Improved representation of the planetary boundary layer, subgrid-scale clouds, and land surface contributed extensively to overall HRRR improvements. The final version of the HRRR (HRRRv4), implemented in late 2020, also features hybrid data assimilation using flow-dependent covariances from a 3-km, 36-member ensemble (“HRRRDAS”) with explicit convective storms. HRRRv4 also includes prediction of wildfire smoke plumes. The HRRR provides a baseline capability for evaluating NOAA’s next-generation Rapid Refresh Forecast System, now under development.

A set of mesoscale convective systems (MCSs) was simulated using the Weather Research and Forecasting model with 3-km grid spacing to investigate the skill at predicting convective initiation and upscale evolution into an MCS. Precipitation was verified using equitable threat scores (ETSs), the neighborhoodbased fractions skill score (FSS), and the Method of Object-Based Diagnostic Evaluation. An illustrative case study more closely examines the strong influence that smaller-scale forcing features had on convective initiation. Initiation errors for the 36 cases were in the south-southwest direction on average, with a mean absolute displacement error of 105 km. No systematic temporal error existed, as the errors were approximately normally distributed. Despite earlier findings that quantitative precipitation forecast skill in convectionparameterizing simulations is a function of the strength of large-scale forcing, this relationship was not present in the present study for convective initiation. However, upscale evolution was better predicted for more strongly forced events according to ETSs and FSSs. For the upscale evolution, the relationship between ETSs and object-based ratings was poor. There was also little correspondence between object-based ratings and the skill at convective initiation. The lack of a relationship between the strength of large-scale forcing andmodel skill at forecasting initiation is likely due to a combination of factors, including the strong role of small-scale features that exert an influence on initiation, and potential errors in the analyses used to represent observations. The limit of predictability of individual convective storms on a 3-km grid must also be considered. ABSTRACT A set of mesoscale convective systems (MCSs) was simulated using the Weather Research and Forecasting model with 3-km grid spacing to investigate the skill at predicting convective initiation and upscale evolution into an MCS. Precipitation was verified using equitable threat scores (ETSs), the neighborhoodbased fractions skill score (FSS), and the Method of Object-Based Diagnostic Evaluation. An illustrative case study more closely examines the strong influence that smaller-scale forcing features had on convective initiation.Initiation errors for the 36 cases were in the south-southwest direction on average, with a mean absolute displacement error of 105 km. No systematic temporal error existed, as the errors were approximately normally distributed. Despite earlier findings that quantitative precipitation forecast skill in convectionparameterizing simulations is a function of the strength of large-scale forcing, this relationship was not present in the present study for convective initiation. However, upscale evolution was better predicted for more strongly forced events according to ETSs and FSSs. For the upscale evolution, the relationship between ETSs and object-based ratings was poor. There was also little correspondence between object-based ratings and the skill at convective initiation. The la...

Two methods for assimilating radar reflectivity into deterministic convection-allowing forecasts were compared: an operationally used, computationally less expensive cloud analysis (CA) scheme and a relatively more expensive, but rigorous, ensemble Kalman filter–variational hybrid method (EnVar). These methods were implemented in the Nonhydrostatic Multiscale Model on the B-grid and were tested on 10 cases featuring high-impact deep convective storms and heavy precipitation. A variety of traditional, neighborhood-based, and features-based verification metrics support that the EnVar produced superior free forecasts compared to the CA procedure, with statistically significant differences extending up to 9 h into the forecast. Despite being inferior, the CA scheme was able to provide benefit compared to not assimilating radar reflectivity at all, but limited to the first few forecast hours. While the EnVar is able to partially suppress spurious convection by assimilating 0-dBZ reflectivity observations directly, the CA is not designed to reduce or remove hydrometeors. As a result, the CA struggles more with suppression of spurious convection in the first-guess field, which resulted in high-frequency biases and poor forecast evolution, as illustrated in a few case studies. Additionally, while the EnVar uses flow-dependent ensemble covariances to update hydrometers, thermodynamic, and dynamic variables simultaneously when the reflectivity is assimilated, the CA relies on a radar reflectivity-derived latent heating rate that is applied during a separate digital filter initialization (DFI) procedure to introduce deep convective storms into the model, and the results of CA are shown to be sensitive to the window length used in the DFI.