Search Results

You are looking at 1 - 5 of 5 items for :

  • Author or Editor: David D. Turner x
  • Weather and Forecasting x
  • All content x
Clear All Modify Search
Jeffrey D. Duda and David D. Turner

Abstract

The Method of Object-based Diagnostic Evaluation (MODE) is used to perform an object-based verification of approximately 1400 forecasts of composite reflectivity from the operational HRRR from April – September 2019. In this study, MODE is configured to prioritize deep, moist convective storm cells typical of those that produce severe weather across the central and eastern US during the warm season. In particular, attributes related to distance and size are given the greatest attribute weights for computing interest in MODE.

HRRR tends to over-forecast all objects, but substantially over-forecasts both small objects at low reflectivity thresholds and large objects at high reflectivity thresholds. HRRR tends to either under-forecast objects in the southern and central Plains or has a correct frequency bias there, whereas it over-forecasts objects across the southern and eastern US. Attribute comparisons reveal the inability of the HRRR to fully resolve convective scale features and the impact of data assimilation and loss of skill during the initial hours of the forecasts.

Scalar metrics are defined and computed based on MODE output, chiefly relying on the interest value. The object-based threat score (OTS), in particular, reveals similar performance of HRRR forecasts as does the Heidke Skill Score, but with differing magnitudes, suggesting value in adopting an object-based approach to forecast verification. The typical distance between centroids of objects is also analyzed and shows gradual degradation with increasing forecast length.

Restricted access
Matthew T. Bray, David D. Turner, and Gijs de Boer

Abstract

Despite a need for accurate weather forecasts for societal and economic interests in the U.S. Arctic, thorough evaluations of operational numerical weather prediction in the region have been limited. In particular, the Rapid Refresh Model (RAP), which plays a key role in short-term forecasting and decision making, has seen very limited assessment in northern Alaska, with most evaluation efforts focused on lower latitudes. In the present study, we verify forecasts from version 4 of the RAP against radiosonde, surface meteorological, and radiative flux observations from two Arctic sites on the northern Alaskan coastline, with a focus on boundary-layer thermodynamic and dynamic biases, model representation of surface inversions, and cloud characteristics. We find persistent seasonal thermodynamic biases near the surface that vary with wind direction, and may be related to the RAP’s handling of sea ice and ocean interactions. These biases seem to have diminished in the latest version of the RAP (version 5), which includes refined handling of sea ice, among other improvements. In addition, we find that despite capturing boundary-layer temperature profiles well overall, the RAP struggles to consistently represent strong, shallow surface inversions. Further, while the RAP seems to forecast the presence of clouds accurately in most cases, there are errors in the simulated characteristics of these clouds, which we hypothesize may be related to the RAP’s treatment of mixed-phase clouds.

Restricted access
Temple R. Lee, Michael Buban, David D. Turner, Tilden P. Meyers, and C. Bruce Baker

Abstract

The High-Resolution Rapid Refresh (HRRR) model became operational at the National Centers for Environmental Prediction (NCEP) in 2014 but the HRRR’s performance over certain regions of the coterminous United States has not been well studied. In the present study, we evaluated how well version 2 of the HRRR, which became operational at NCEP in August 2016, simulates the near-surface meteorological fields and the surface energy balance at two locations in northern Alabama. We evaluated the 1-, 3-, 6-, 12-, and 18-h HRRR forecasts, as well as the HRRR’s initial conditions (i.e., the 0-h initial fields) using meteorological and flux observations obtained from two 10-m micrometeorological towers installed near Belle Mina and Cullman, Alabama. During the 8-month model evaluation period, from 1 September 2016 to 30 April 2017, we found that the HRRR accurately simulated the observations of near-surface air and dewpoint temperature (R 2 > 0.95). When comparing the HRRR output with the observed sensible, latent, and ground heat flux at both sites, we found that the agreement was weaker (R 2 ≈ 0.7), and the root-mean-square errors were much larger than those found for the near-surface meteorological variables. These findings help motivate the need for additional work to improve the representation of surface fluxes and their coupling to the atmosphere in future versions of the HRRR to be more physically realistic.

Full access
Junjun Hu, Nusrat Yussouf, David D. Turner, Thomas A. Jones, and Xuguang Wang

Abstract

Due to lack of high spatial and temporal resolution boundary layer (BL) observations, the rapid changes in the near-storm environment are not well represented in current convective-scale numerical models. Better representation of the near-storm environment in model initial conditions will likely further improve the forecasts of severe convective weather. This study investigates the impact of assimilating high temporal resolution BL retrievals from two ground-based remote sensing instruments for short-term forecasts of a tornadic supercell event on 13 July 2015 during the Plains Elevated Convection At Night field campaign. The instruments are the Atmospheric Emitted Radiance Interferometer (AERI) that retrieves thermodynamic profiles and the Doppler lidar (DL) that measures horizontal wind profiles. Six sets of convective-scale ensemble data assimilation (DA) experiments are performed: two control experiments that assimilate conventional and WSR-88D radar observations using either relaxation-to-prior-spread (RTPS) or the adaptive inflation (AI) technique and four experiments similar to the control but that assimilate either DL or AERI or both observations in addition to all other observations that are in the control experiments. Results indicate a positive impact of AERI and DL observations in forecasting convective initiation (CI) and early evolution of the supercell storm. The experiment that employs the AI technique to assimilate BL observations in DA enhances the humidity in the near-storm environment and low-level convergence, which in turn helps forecasting CI. The forecast improvement is most pronounced during the first ~3 h. Results also indicate that the AERI observations have a larger impact compared to DL in predicting CI.

Full access
Irina V. Djalalova, Laura Bianco, Elena Akish, James M. Wilczak, Joseph B. Olson, Jaymes S. Kenyon, Larry K. Berg, Aditya Choukulkar, Richard Coulter, Harinda J. S. Fernando, Eric Grimit, Raghavendra Krishnamurthy, Julie K. Lundquist, Paytsar Muradyan, David D. Turner, and Sonia Wharton

Abstract

The second Wind Forecast Improvement Project (WFIP2) is a multiagency field campaign held in the Columbia Gorge area (October 2015–March 2017). The main goal of the project is to understand and improve the forecast skill of numerical weather prediction (NWP) models in complex terrain, particularly beneficial for the wind energy industry. This region is well known for its excellent wind resource. One of the biggest challenges for wind power production is the accurate forecasting of wind ramp events (large changes of generated power over short periods of time). Poor forecasting of the ramps requires large and sudden adjustments in conventional power generation, ultimately increasing the costs of power. A Ramp Tool and Metric (RT&M) was developed during the first WFIP experiment, held in the U.S. Great Plains (September 2011–August 2012). The RT&M was designed to explicitly measure the skill of NWP models at forecasting wind ramp events. Here we apply the RT&M to 80-m (turbine hub-height) wind speeds measured by 19 sodars and three lidars, and to forecasts from the High-Resolution Rapid Refresh (HRRR), 3-km, and from the High-Resolution Rapid Refresh Nest (HRRRNEST), 750-m horizontal grid spacing, models. The diurnal and seasonal distribution of ramp events are analyzed, finding a noticeable diurnal variability for spring and summer but less for fall and especially winter. Also, winter has fewer ramps compared to the other seasons. The model skill at forecasting ramp events, including the impact of the modification to the model physical parameterizations, was finally investigated.

Restricted access