School of Meteorology (Defense)

Ensemble Prediction of Splitting Supercells and Hail on 10 May 2010

Jonathan Labriola

01 December 2015, 9:00 AM

National Weather Center, Room 4140
120 David L. Boren Blvd.
University of Oklahoma
Norman, OK

Severe hail is a major weather hazard, causing injuries and billions of dollars in damage each year. Though modern convective-scale NWP forecast ensembles have shown skill in predicting supercell thunderstorms, relatively little study has been performed on explicit short-term prediction of hail. On 10 May 2010, a left-splitting supercell occurred over southwestern Oklahoma, producing a swath of severe hail, as well as two anticyclonic tornadoes. Long-lived left-split supercells are relatively uncommon; accurate numerical prediction of such storms requires capturing the sheared environment in which they develop, making prediction difficult. Furthermore, left-moving supercells have been found to cause a disproportionate number of hail reports—predicting hail produced by splitting storms is thus an important area of study in improving hail forecasts. To predict the splitting storms of 10 May 2010 and their environment, multiple data sources, including NEXRAD and CASA radar data, Mesonet surface observations, as well as radiosonde and wind profiler data, were assimilated into a set of 40-member ensemble forecast experiments with 500 m horizontal grid spacing run using the ARPS EnKF system.

Practical predictability of the left-splitting storm is investigated, along with the forecast sensitivity to the choice of a single moment (LIN) or double moment (MY2) microphysical scheme. Overall, the MY2 ensemble skillfully predicts the left-splitting storm for lead times of up to 30 minutes; the LIN ensemble is skillful only at shorter lead times and produces a less favorable wind shear environment, as well as an overly intense cold pool. Hodographs and storm relative helicity (SRH) in the near storm environment are analyzed; SRH is highly sensitive to the storm motion vector, and varies substantially due to the assumptions used to obtain it. The Lakshmanan and Smith method is found to produce a more representative SRH field than the Bunkers method, helping to distinguish environments that produce strong, weak, and no splitting storms.
The model’s ability to forecast hail is also evaluated. Surface based hail reports are limited, so two radar products, maximum estimated size of hail (MESH) and a hydrometeor classification algorithm (HCA), are used as a proxy for hail observations. Both the LIN and MY2 ensembles overestimate predicted MESH in magnitude and spatial extent. Both ensembles exhibit greater skill when the model hail forecasts are verified against HCA-indicated hail swaths. The LIN scheme over-predicts the mass of hail in storm updrafts (a high bias in hail mass), while the MY2 scheme shows little bias in the hail forecast. Despite the biases of the LIN scheme, both ensembles produce skillful 0-60 minute hail forecasts.

School of Meteorology (Defense) Seminar Series website