Using high-frequency surface data assimilation to improve short-term ensemble forecasts of convection initiation and evolution on 29 May 2012

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 5 February 2014: 4:15 PM
Room C202 (The Georgia World Congress Center )
Ryan A. Sobash, University of Oklahoma, Norman, OK; and D. J. Stensrud

Surface data assimilation (DA) has the potential to improve forecasts of convection initiation and subsequent convective evolution. Since the processes driving convection initiation occur on scales inadequately observed by traditional observations (e.g. METAR), mesoscale surface networks (i.e. mesonets) could be especially beneficial given their higher temporal and spatial resolution. This work aims to assess the impact of assimilating both METAR and mesonet surface observations on ensemble Kalman Filter (EnKF) generated analyses and forecasts of the 29 May 2012 convective event over the southern Great Plains. This event produced total losses estimated at 500 million U.S. dollars, primarily due to very large hail within the Oklahoma City metropolitan area.

An initial 50-member ensemble (dx = 3 km) was constructed at 18 UTC on 29 May 2012 by downscaling a mesoscale ensemble (dx = 15 km) produced in a cycled EnKF analysis system run at NCAR during Spring 2012. Then, METAR and MADIS mesonet observations (including both the West Texas and Oklahoma mesonets) were assimilated every 5-minutes into the 3-km ensemble using the EnKF. Following 3 hours of DA, 50-member, 6-hour ensemble forecasts were initialized using the 21 UTC ensemble analyses, approximately 30 minutes prior to convection initiation. For comparison, a control 50-member ensemble forecast (without DA) was initialized using the 18 UTC initial conditions.

Forecasts of convection initiation within the 21 UTC 23 UTC time period were improved by assimilating METAR and mesonet data, both in predicted timing and location. This primarily occurs due to better estimates of mass convergence along the simulated surface boundaries driving CI. Surface DA also decreases biases in the boundary-layer moisture field leading to reduced convective coverage in the forecasts and differences in predicted convective evolution. The control ensemble predicts an MCS that moves eastward across northern Oklahoma and southern Kansas, while the assimilation experiment predicts a convective system that moves southeasterly, more accurately reflecting the observed convective evolution. Analyses and forecasts from experiments where mesonet data were withheld possess similar errors to the control ensemble. Further, hourly (rather than 5-minute) assimilation of METAR and mesonet observations was also unable to produce analyses and forecasts that matched the observed convective evolution. These results illustrate the benefit of high spatial and temporal resolution surface data on analyses and forecasts of this particular convective event.