Search Results

You are looking at 11 - 16 of 16 items for :

  • Author or Editor: Craig S. Schwartz x
  • Monthly Weather Review x
  • Refine by Access: All Content x
Clear All Modify Search
Craig S. Schwartz
,
Glen S. Romine
,
Kathryn R. Fossell
,
Ryan A. Sobash
, and
Morris L. Weisman

Abstract

Precipitation forecasts from convection-allowing ensembles with 3- and 1-km horizontal grid spacing were evaluated between 15 May and 15 June 2013 over central and eastern portions of the United States. Probabilistic forecasts produced from 10- and 30-member, 3-km ensembles were consistently better than forecasts from individual 1-km ensemble members. However, 10-member, 1-km probabilistic forecasts usually were best, especially over the first 12 h and at rainfall rates ≥ 5.0 mm h−1 at later times. Further object-based investigation revealed that better 1-km forecasts at heavier rainfall rates were associated with more accurate placement of mesoscale convective systems compared to 3-km forecasts. The collective results indicate promise for 1-km ensembles once computational resources can support their operational implementation.

Full access
Craig S. Schwartz
,
May Wong
,
Glen S. Romine
,
Ryan A. Sobash
, and
Kathryn R. Fossell

Abstract

Five sets of 48-h, 10-member, convection-allowing ensemble (CAE) forecasts with 3-km horizontal grid spacing were systematically evaluated over the conterminous United States with a focus on precipitation across 31 cases. The various CAEs solely differed by their initial condition perturbations (ICPs) and central initial states. CAEs initially centered about deterministic Global Forecast System (GFS) analyses were unequivocally better than those initially centered about ensemble mean analyses produced by a limited-area single-physics, single-dynamics 15-km continuously cycling ensemble Kalman filter (EnKF), strongly suggesting relative superiority of the GFS analyses. Additionally, CAEs with flow-dependent ICPs derived from either the EnKF or multimodel 3-h forecasts from the Short-Range Ensemble Forecast (SREF) system had higher fractions skill scores than CAEs with randomly generated mesoscale ICPs. Conversely, due to insufficient spread, CAEs with EnKF ICPs had worse reliability, discrimination, and dispersion than those with random and SREF ICPs. However, members in the CAE with SREF ICPs undesirably clustered by dynamic core represented in the ICPs, and CAEs with random ICPs had poor spinup characteristics. Collectively, these results indicate that continuously cycled EnKF mean analyses were suboptimal for CAE initialization purposes and suggest that further work to improve limited-area continuously cycling EnKFs over large regional domains is warranted. Additionally, the deleterious aspects of using both multimodel and random ICPs suggest efforts toward improving spread in CAEs with single-physics, single-dynamics, flow-dependent ICPs should continue.

Free access
Craig S. Schwartz
,
Zhiquan Liu
,
Xiang-Yu Huang
,
Ying-Hwa Kuo
, and
Chin-Tzu Fong

Abstract

The Weather Research and Forecasting Model (WRF) “hybrid” variational-ensemble data assimilation (DA) algorithm was used to initialize WRF model forecasts of three tropical cyclones (TCs). The hybrid-initialized forecasts were compared to forecasts initialized by WRF's three-dimensional variational (3DVAR) DA system. An ensemble adjustment Kalman filter (EAKF) updated a 32-member WRF-based ensemble system that provided flow-dependent background error covariances for the hybrid. The 3DVAR, hybrid, and EAKF configurations cycled continuously for ~3.5 weeks and produced new analyses every 6 h that initialized 72-h WRF forecasts with 45-km horizontal grid spacing. Additionally, the impact of employing a TC relocation technique and using multiple outer loops (OLs) in the 3DVAR and hybrid minimizations were explored.

Model output was compared to conventional, dropwindsonde, and TC “best track” observations. On average, the hybrid produced superior forecasts compared to 3DVAR when only one OL was used during minimization. However, when three OLs were employed, 3DVAR forecasts were dramatically improved but the mean hybrid performance changed little. Additionally, incorporation of TC relocation within the cycling systems further improved the mean 3DVAR-initialized forecasts but the average hybrid-initialized forecasts were nearly unchanged.

Full access
Glen S. Romine
,
Craig S. Schwartz
,
Judith Berner
,
Kathryn R. Fossell
,
Chris Snyder
,
Jeff L. Anderson
, and
Morris L. Weisman

Abstract

Ensembles provide an opportunity to greatly improve short-term prediction of local weather hazards, yet generating reliable predictions remain a significant challenge. In particular, convection-permitting ensemble forecast systems (CPEFSs) have persistent problems with underdispersion. Representing initial and or lateral boundary condition uncertainty along with forecast model error provides a foundation for building a more dependable CPEFS, but the best practice for ensemble system design is not well established.

Several configurations of CPEFSs are examined where ensemble forecasts are nested within a larger domain, drawing initial conditions from a downscaled, continuously cycled, ensemble data assimilation system that provides state-dependent initial condition uncertainty. The control ensemble forecast, with initial condition uncertainty only, is skillful but underdispersive. To improve the reliability of the ensemble forecasts, the control ensemble is supplemented with 1) perturbed lateral boundary conditions; or, model error representation using either 2) stochastic kinetic energy backscatter or 3) stochastically perturbed parameterization tendencies. Forecasts are evaluated against stage IV accumulated precipitation analyses and radiosonde observations. Perturbed ensemble forecasts are also compared to the control forecast to assess the relative impact from adding forecast perturbations. For precipitation forecasts, all perturbation approaches improve ensemble reliability relative to the control CPEFS. Deterministic ensemble member forecast skill, verified against radiosonde observations, decreases when forecast perturbations are added, while ensemble mean forecasts remain similarly skillful to the control.

Full access
Ryan A. Sobash
,
David John Gagne II
,
Charlie L. Becker
,
David Ahijevych
,
Gabrielle N. Gantos
, and
Craig S. Schwartz

Abstract

While convective storm mode is explicitly depicted in convection-allowing model (CAM) output, subjectively diagnosing mode in large volumes of CAM forecasts can be burdensome. In this work, four machine learning (ML) models were trained to probabilistically classify CAM storms into one of three modes: supercells, quasi-linear convective systems, and disorganized convection. The four ML models included a dense neural network (DNN), logistic regression (LR), a convolutional neural network (CNN), and semisupervised CNN–Gaussian mixture model (GMM). The DNN, CNN, and LR were trained with a set of hand-labeled CAM storms, while the semisupervised GMM used updraft helicity and storm size to generate clusters, which were then hand labeled. When evaluated using storms withheld from training, the four classifiers had similar ability to discriminate between modes, but the GMM had worse calibration. The DNN and LR had similar objective performance to the CNN, suggesting that CNN-based methods may not be needed for mode classification tasks. The mode classifications from all four classifiers successfully approximated the known climatology of modes in the United States, including a maximum in supercell occurrence in the U.S. Central Plains. Further, the modes also occurred in environments recognized to support the three different storm morphologies. Finally, storm mode provided useful information about hazard type, e.g., storm reports were most likely with supercells, further supporting the efficacy of the classifiers. Future applications, including the use of objective CAM mode classifications as a novel predictor in ML systems, could potentially lead to improved forecasts of convective hazards.

Significance Statement

Whether a thunderstorm produces hazards such as tornadoes, hail, or intense wind gusts is in part determined by whether the storm takes the form of a single cell or a line. Numerical forecasting models can now provide forecasts that depict this structure. We tested several automated algorithms to extract this information from forecast output using machine learning. All of the automated methods were able to distinguish between a set of three convective types, with the simple techniques providing similarly skilled classifications compared to the complex approaches. The automated classifications also successfully discriminated between thunderstorm hazards, potentially leading to new forecast tools and better forecasts of high-impact convective hazards.

Restricted access
Craig S. Schwartz
,
John S. Kain
,
Steven J. Weiss
,
Ming Xue
,
David R. Bright
,
Fanyou Kong
,
Kevin W. Thomas
,
Jason J. Levit
, and
Michael C. Coniglio

Abstract

During the 2007 NOAA Hazardous Weather Testbed (HWT) Spring Experiment, the Center for Analysis and Prediction of Storms (CAPS) at the University of Oklahoma produced convection-allowing forecasts from a single deterministic 2-km model and a 10-member 4-km-resolution ensemble. In this study, the 2-km deterministic output was compared with forecasts from the 4-km ensemble control member. Other than the difference in horizontal resolution, the two sets of forecasts featured identical Advanced Research Weather Research and Forecasting model (ARW-WRF) configurations, including vertical resolution, forecast domain, initial and lateral boundary conditions, and physical parameterizations. Therefore, forecast disparities were attributed solely to differences in horizontal grid spacing. This study is a follow-up to similar work that was based on results from the 2005 Spring Experiment. Unlike the 2005 experiment, however, model configurations were more rigorously controlled in the present study, providing a more robust dataset and a cleaner isolation of the dependence on horizontal resolution. Additionally, in this study, the 2- and 4-km outputs were compared with 12-km forecasts from the North American Mesoscale (NAM) model. Model forecasts were analyzed using objective verification of mean hourly precipitation and visual comparison of individual events, primarily during the 21- to 33-h forecast period to examine the utility of the models as next-day guidance. On average, both the 2- and 4-km model forecasts showed substantial improvement over the 12-km NAM. However, although the 2-km forecasts produced more-detailed structures on the smallest resolvable scales, the patterns of convective initiation, evolution, and organization were remarkably similar to the 4-km output. Moreover, on average, metrics such as equitable threat score, frequency bias, and fractions skill score revealed no statistical improvement of the 2-km forecasts compared to the 4-km forecasts. These results, based on the 2007 dataset, corroborate previous findings, suggesting that decreasing horizontal grid spacing from 4 to 2 km provides little added value as next-day guidance for severe convective storm and heavy rain forecasters in the United States.

Full access