2011 Storm-scale Radar Data Assimilation Workshop
Monday, 17 October 2011
120 David L. Boren Blvd., University of Oklahoma, Norman, OK 73072
2011 Stormscale Radar DA Workshop Attendees.
The Storm-scale Radar Data Assimilation Workshop was held at the National Weather Center in Norman, Oklahoma, on 17-18 October 2011. The workshop was jointly sponsored by the NOAA National Severe Storms Laboratory, the Center for Analysis and Prediction of Storms at the University of Oklahoma, and the Global Systems Division of the NOAA Earth System Research Laboratory. Over 50 participants from across the United States and the United Kingdom attended to overview radar data assimilation successes and needed improvements, radar data and quality control, and the influence of model error on storm-scale forecasts. All of the speakers were invited and asked to discuss their experiences with radar data assimilation, low-hanging fruit that could lead to rapid progress, challenges faced, and obstacles to successful prediction. Numerous discussion sessions provided a helpful forum for surveying the field and the forthright exchange of information and opinions on how to make progress.
Results shown at the workshop make it clear that storm-scale radar data assimilation is becoming a mature discipline within meteorology. Currently, a two-step process is often used in which the mesoscale environment is analyzed first. The assimilation of radar observations then allows for deep convection to be initialized with a close fit of the model state to the observations within storm-scale models and provide very good short-range forecasts of convection for some events. However, the forecasts for a number of events lose skill quickly as the convection either decays or diverges from the observed evolution and/or spurious convection develops elsewhere. It also appears that radar data assimilation has a longer positive effect in the late afternoon hours for reasons we are just beginning to understand It is much easier to obtain a high-quality analysis of convection than an accurate forecast of convection, suggesting that numerous challenges remain. Model error certainly plays a role in the loss of forecast skill, especially with respect to microphysics and planetary boundary layer schemes, but the data assimilation methodology may also play a role. It is unclear how well the initial conditions are balanced after radar data assimilation is applied, and any imbalances within the initial state likely contribute to the loss of forecast skill as the model quickly acts to develop a balanced state that may deviate from reality. Error in the storm environment can also play a significant role, and this points to the need to develop integrated multi-scale data assimilation systems that analyze convective storms and storm environment in an optimal and consistent manner.
Other challenges revolve around radar data quality control. Radar observations are essential to successful storm-scale prediction, as they provide the only detailed in-storm observations currently available, yet these observations are difficult to quality control. One must quality control the data from each radar site separately, and then combine them into a coherent depiction of ongoing conditions. Numerous methods have been developed for quality control, with varying degrees of success. Many of the quality control algorithms used in real-time testing are very aggressive in removing questionable observations and only retain data within the core regions of the thunderstorms. The dual-polarization upgrade to the WSR-88D network should provide for improved quality control and lead to the improved use of radar observations in the assimilation process. Dual-polarization observations also provide new opportunities to assess storm-scale analyses and forecasts of hydrometeor characteristics.
The workshop discussion sessions led to a number of action items for the community to consider. Since radar data quality control is a community challenge, a centralized radar data processing code that could be used as a community resource was suggested. A community supplied forward operator for reflectivity, Doppler velocity, precipitation fall speed, and polarimetric variables could also be very beneficial to the data-assimilation community and to developers of microphysics schemes. An inter-comparison project in which different radar data assimilation approaches would be applied to identical cases, using identical data sets, models, and verification methods, was suggested and received strong support. Examining similar projects from the climate modeling community that were both successful and unsuccessful would be helpful to determine how to craft such a project. Limiting the first comparison to a single case might be wise to work out any problems before a large time commitment was made by participants. Funding for such a project was mentioned as a concern. The need for broader organizational support, such as could be provided by a Joint Center for Radar Data Assimilation, was emphasized as a way to help strengthen collaborations across the community and lead to stronger and more effective transitions from research-to-operations. Finally, improvements in the observational systems would be beneficial, such as providing more routine boundary layer and mesoscale information, and more direct involvement with radar engineers could lead to long-term gains for the community.
All of the participants are thanked for contributing to a successful workshop, with added thanks to the invited speakers. As there was strong interest in having another storm-scale radar data assimilation workshop in the next 2-3 years, we will look to organize another one in 2013 or 2014.
David Stensrud, National Severe Storms Laboratory
Ming Xue, Center for Analysis and Prediction of Storms
David Dowell, Global Systems Division