A GSI-based data assimilation (DA) system, including three-dimensional variational assimilation (3DVar) and ensemble Kalman filter (EnKF), is extended to the multiscale assimilation of both meso- and synoptic-scale observation networks and convective-scale radar reflectivity and velocity observations. EnKF and 3DVar are systematically compared in this multiscale context to better understand the impacts of differences between the DA techniques on the analyses at multiple scales and the subsequent convective-scale precipitation forecasts. Averaged over 10 diverse cases, 8-h precipitation forecasts initialized using GSI-based EnKF are more skillful than those using GSI-based 3DVar, both with and without storm-scale radar DA. The advantage from radar DA persists for ~5 h using EnKF, but only ~1 h using 3DVar. A case study of an upscale growing MCS is also examined. The better EnKF-initialized forecast is attributed to more accurate analyses of both the mesoscale environment and the storm-scale features. The mesoscale location and structure of a warm front is more accurately analyzed using EnKF than 3DVar. Furthermore, storms in the EnKF multiscale analysis are maintained during the subsequent forecast period. However, storms in the 3DVar multiscale analysis are not maintained and generate excessive cold pools. Therefore, while the EnKF forecast with radar DA remains better than the forecast without radar DA throughout the forecast period, the 3DVar forecast quality is degraded by radar DA after the first hour. Diagnostics revealed that the inferior analysis at mesoscales and storm scales for the 3DVar is primarily attributed to the lack of flow dependence and cross-variable correlation, respectively, in the 3DVar static background error covariance.

Convection-allowing ensemble forecasts with perturbations to model physics, dynamics, and initial (IC) and lateral boundary conditions (LBC) generated by the Center for the Analysis and Prediction of Storms for the NOAA Hazardous Weather Testbed (HWT) Spring Experiments provide a unique opportunity to understand the relative impact of different sources of perturbation on convection-allowing ensemble diversity. Such impacts are explored in this two-part study through an object-oriented hierarchical cluster analysis (HCA) technique.In this paper, an object-oriented HCA algorithm, where the dissimilarity of precipitation forecasts is quantified with a nontraditional object-based threat score (OTS), is developed. The advantages of OTS-based HCA relative to HCA using traditional Euclidean distance and neighborhood probability-based Euclidean distance (NED) as dissimilarity measures are illustrated by hourly accumulated precipitation ensemble forecasts during a representative severe weather event.Clusters based on OTS and NED are more consistent with subjective evaluation than clusters based on traditional Euclidean distance because of the sensitivity of Euclidean distance to small spatial displacements. OTS improves the clustering further compared to NED. Only OTS accounts for important features of precipitation areas, such as shape, size, and orientation, and OTS is less sensitive than NED to precise spatial location and precipitation amount. OTS is further improved by using a fuzzy matching method. Application of OTS-based HCA for regional subdomains is also introduced. Part II uses the HCA method developed in this paper to explore systematic clustering of the convection-allowing ensemble during the full 2009 HWT Spring Experiment period.

Multiscale ensemble-based data assimilation and forecasts were performed in real time during the Plains Elevated Convection At Night (PECAN) field experiment. A 20-member ensemble of forecasts at 4-km grid spacing was initialized daily at both 1300 and 1900 UTC, together with a deterministic forecast at 1-km grid spacing initialized at 1300 UTC. The configuration of the GSI-based data assimilation and forecast system was guided by results presented in Part I of this two-part study. The present paper describes the implementation of the real-time system and the extensive forecast products that were generated to support the unique interests of PECAN researchers. Subjective and objective verification of the real-time forecasts from 1 June through 15 July 2015 is conducted, with an emphasis on nocturnal mesoscale convective systems (MCSs), nocturnal convective initiation (CI), nocturnal low-level jets (LLJs), and bores on the nocturnal stable layer. Verification of nocturnal precipitation during overnight hours, a proxy for MCSs, shows both greater skill and spread for the 1300 UTC forecasts than the 1900 UTC forecasts. Verification against observed soundings reveals that the forecast LLJs systematically peak, veer, and dissipate several hours before the observations. Comparisons with bores that passed over an Atmospheric Emitted Radiance Interferometer reveal an ability to predict borelike features that is greatly improved at 1-km, compared with 4-km, grid spacing. Objective verification of forecast CI timing reveals strong sensitivity to the PBL scheme but an overall unbiased ensemble.

Multiscale convection-allowing precipitation forecast perturbations are examined for two forecasts and systematically over 34 forecasts out to 30-h lead time using Haar Wavelet decomposition. Two small-scale initial condition (IC) perturbation methods are compared to the larger-scale IC and physics perturbations in an experimental convection-allowing ensemble. For a precipitation forecast driven primarily by a synoptic-scale baroclinic disturbance, small-scale IC perturbations resulted in little precipitation forecast perturbation energy on medium and large scales, compared to larger-scale IC and physics (LGPH) perturbations after the first few forecast hours. However, for a case where forecast convection at the initial time grew upscale into a mesoscale convective system (MCS), small-scale IC and LGPH perturbations resulted in similar forecast perturbation energy on all scales after about 12 h. Small-scale IC perturbations added to LGPH increased total forecast perturbation energy for this case. Averaged over 34 forecasts, the small-scale IC perturbations had little impact on large forecast scales while LGPH accounted for about half of the error energy on such scales. The impact of small-scale IC perturbations was also less than, but comparable to, the impact of LGPH perturbations on medium scales. On small scales, the impact of small-scale IC perturbations was at least as large as the LGPH perturbations. The spatial structure of small-scale IC perturbations affected the evolution of forecast perturbations, especially at medium scales. There was little systematic impact of the small-scale IC perturbations when added to LGPH. These results motivate further studies on properly sampling multiscale IC errors.

Twenty-member real-time convection-allowing storm-scale ensemble forecasts with perturbations to model physics, dynamics, initial conditions (IC), and lateral boundary conditions (LBC) during the NOAA Hazardous Weather Testbed Spring Experiment provide a unique opportunity to study the relative impact of different sources of perturbation on convection-allowing ensemble diversity. In Part II of this two-part study, systematic similarity/dissimilarity of hourly precipitation forecasts among ensemble members from the spring season of 2009 are identified using hierarchical cluster analysis (HCA) with a fuzzy object-based threat score (OTS), developed in Part I. In addition to precipitation, HCA is also performed on ensemble forecasts using the traditional Euclidean distance for wind speed at 10 m and 850 hPa, and temperature at 500 hPa.At early lead times (3 h, valid at 0300 UTC) precipitation forecasts cluster primarily by data assimilation and model dynamic core, indicating a dominating impact of models, with secondary clustering by microphysics. There is an increasing impact of the planetary boundary layer (PBL) scheme on clustering relative to the microphysics scheme at later lead times. Forecasts of 10-m wind speed cluster primarily by the PBL scheme at early lead times, with an increasing impact of LBC at later lead times. Forecasts of midtropospheric variables cluster primarily by IC at early lead times and LBC at later lead times. The radar and Mesonet data assimilation (DA) show its impact, with members without DA in a distinct cluster, through the 12-h lead time (valid at 1200 UTC) for both precipitation and nonprecipitation variables. The implication for optimal ensemble design for storm-scale forecasts is also discussed.

Forecasts generated by the Center for Analysis and Prediction of Storms with 1-and 4-km grid spacing using the Advanced Research Weather Research and Forecasting Model (ARW-WRF; ARW1 and ARW4, respectively) for the 2009-11 NOAA Hazardous Weather Testbed Spring Experiments are compared and verified. Object-based measures, including average values of object attributes, the object-based threat score (OTS), and the median of maximum interest (MMI) are used for the verification. Verification was first performed against observations at scales resolvable by each forecast model and then performed at scales resolvable by both models by remapping ARW1 to the ARW4 grid (ARW1to4). Thirty-hour forecasts of 1-h accumulated precipitation initialized at 0000 UTC on 22, 36, and 33 days during the spring of 2009, 2010, and 2011, respectively, are evaluated over a domain covering most of the central and eastern United States. ARW1, ARW1to4, and ARW4 all significantly overforecasted the number of objects during diurnal convection maxima. The overforecasts by ARW1 and ARW1to4 were more pronounced than ARW4 during the first convection maximum at 1-h lead time. The average object area and aspect ratio were closer to observations for ARW1 and ARW1to4 than for ARW4. None of the models showed a significant advantage over the others for average orientation angle and centroid location. Increased accuracy for ARW1, compared to ARW4, was statistically significant for the MMI but not the OTS. However, ARW1to4 had similar MMI and OTS as ARW4 at most lead times. These results are consistent with subjective evaluations that the greatest impact of grid spacing is on the smallest resolvable objects.

A prototype convection-allowing system using the Advanced Research version of the Weather Research and Forecasting (WRF-ARW) Model and employing an ensemble Kalman filter (EnKF) data assimilation technique has been developed and used during the spring 2016 and 2017 Hazardous Weather Testbeds. This system assimilates WSR-88D reflectivity and radial velocity, geostationary satellite cloud water path (CWP) retrievals, and available surface observations over a regional domain with a 3-km horizontal resolution at 15-min intervals, with 3-km initial conditions provided by an experimental High-Resolution Rapid Refresh ensemble (HRRR-e). However, no information on upper-level thermodynamic conditions in cloud-free regions is currently assimilated, as few timely observations exist. One potential solution is to also assimilate clear-sky satellite radiances, which provide information on mid- and upper-tropospheric temperature and moisture conditions. This research assimilates GOES-13 imager water vapor band (6.5 μm) radiances using the GSI-EnKF system to take advantage of the Community Radiative Transfer Model (CRTM) integration. Results using four cases from May 2016 showed that assimilating radiances generally had a neutral-to-positive impact on the model analysis, reducing humidity bias and/or errors at the appropriate model levels where verification observations were present. The effects on high-impact weather forecasts, as verified against forecast reflectivity and updraft helicity, were mixed. Three cases (9, 22, and 24 May) showed some improvement in skill, while the other (25 May) performed worse, despite the improved environment. This research represents the first step in designing a high-resolution ensemble data assimilation system to use GOES-16 Advanced Baseline Imager data, which provides additional water vapor bands and increased spatial and temporal resolution.

Neighborhood and object-based probabilistic precipitation forecasts from a convection-allowing ensemble are verified and calibrated. Calibration methods include logistic regression, one- and two-parameter reliability-based calibration, and cumulative distribution function (CDF)-based bias adjustment. Newly proposed object-based probabilistic forecasts for the occurrence of a forecast object are derived from the percentage of ensemble members with a matching object. Verification and calibration of single- and multimodel subensembles are performed to explore the effect of using multiple models. The uncalibrated neighborhood-based probabilistic forecasts have skill minima during the afternoon convective maximum. Calibration generally improves the skill, especially during the skill minima, resulting in positive skill. In general all calibration methods perform similarly, with a slight advantage of logistic regression (one-parameter reliability based) calibration for 1-h (6 h) accumulations. The uncalibrated object-based probabilistic forecasts are, in general, less skillful than the uncalibrated neighborhood-based probabilistic forecasts. Object-based calibration also results in positive skill at all lead times. For object-based calibration the skill is significantly different among the calibration methods, with the logistic regression performing the best and CDF-based bias adjustment performing the worst. For both the neighborhood and object-based probabilistic forecasts, the impact of using 10 or 25 days of training data for calibration is generally small and is most significant for the two-parameter reliability-based method. An uncalibrated Advanced Research Weather Research and Forecasting Model (ARW-WRF) subensemble is significantly more skillful than an uncalibrated WRF Nonhydrostatic Mesoscale Model (NMM) subensemble. The difference is reduced by calibration. The multimodel subensemble only shows an advantage for the neighborhood-based forecasts beyond 1-day lead time and shows no advantage for the object-based forecasts.