LEAD Poster

Slide1: 

Detection of Hazardous Weather Phenomena Using Data Assimilation Techniques Robert Fritchie, Kelvin Droegemeier, Mingjing Tong School of Meteorology and Center for Analysis and Prediction of Storms This work is supported primarily by the National Science Foundation under the following cooperative agreements: ATM03-31574, 31578, 31579, 31480, 31586, 31587, 31591, and 31594. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect those of the National Science Foundation. Overview The automated detection of tornadoes and other hazardous weather events involves using algorithms to identify patterns in “raw” Doppler radar reflectivity and velocity data. One such algorithm-based system is the NSSL Warning Decision Support System – Integrated Information (WDSS-II) One major limitation is that new detection algorithms must be created, or existing ones adapted, each time a new observation system is deployed. A tornado/parent mesocyclone will look very different when viewed by a WSR-88D with a gate size of 1km, as supposed to a Mobile doppler radar with a gate size of 50m. Another major limitation is that such algorithms operate principally on data directly measured by the radar (radial velocity and reflectivity) and thus do not make use of other important fields that are potentially available to them (e.g., pressure and temperature). An alternative approach involves using advanced data assimilation and retrieval techniques, applied to all available observations – especially those collected at fine scales by Doppler radar – to generate dynamically consistent, 3D gridded analyses of all key observed and unobserved meteorological quantities to which data mining tools can be applied. The potential advantages include the ability to interrogate quantities not available from raw data and the use of geometrically simple 3D grids. The most important advantage, however, is that the mining algorithms do not depend upon the data sources and do not have to be changed when new data sources are added (e.g., new types of radars). Research Objectives Examine tradeoffs of hazardous weather detection between conventional sensor-based algorithms and gridded data sets. Prove the ease of adding new instrumentation to detection algorithms that are based on the assimilated data only. Explore the computational requirements of data assimilation on a variety of scales and grid spacing combinations. Examine the physical signatures of various hazardous weather phenomena, particularly tornadoes, at various scales and grid spacings and compare to the detections with WDSS-II as well as ground truth. Investigate the value added to data sets through use of mobile or dynamic sensing platforms such as mobile radars or CASA radars. P12 Tools and Methodology: Compare detections produced by automated algorithms to features in assimilated analyses that are generated using ensemble Kalman filtering for an observed tornadic storm that occurred on 29 May 2004 and that was observed at reasonably close range by NEXRAD radar. Examine sensitivities to a variety of variable factors including, in WDSS-II, adaptable parameters and in ensemble Kalman filtering, grid spacing, data frequency, ensemble size, and quantities assimilated. Comparison between detected features, both through WDSS-II and Data Assimilation, and surveyed ground truth data will allow for a good assessment of relative skill of hazardous weather phenomena detection. Compute analyses with various resolutions and domains to compare their relative detections versus their computational requirements

Slide2: 

Detection of Hazardous Weather Phenomena Using Data Assimilation Techniques Robert Fritchie, Kelvin Droegemeier, Mingjing Tong School of Meteorology and Center for Analysis and Prediction of Storms This work is supported primarily by the National Science Foundation under the following cooperative agreements: ATM03-31574, 31578, 31579, 31480, 31586, 31587, 31591, and 31594. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect those of the National Science Foundation. P12 Early Results: 40-member ensemble of a square-root Kalman filter The filter uses observations from the KTLX radar of the May 29th case Assimilation was performed every 5 minutes (each volume scan) for a 1 hour period Built the storm to a physically consistent gridded set that closely resembles the actual storm itself The analysis grid, depicted in figure X, is a 180km x 120km x 16km block, with horizontal grid spacing of 1 kilometer and stretched vertical grid spacing with a minimum of 100 meters. The analysis ends at 0100UTC, or 8pm CDT. Shortly after that time, an anticyclonic tornado formed north of Calumet, OK. On the afternoon of May 29th, 2004, a long-lived supercell formed in Western Oklahoma and tracked through Central Oklahoma, north of Oklahoma City, through Northeast Oklahoma, including Tulsa, and finally dissipated west of the Arkansas state line. Throughout its approximately 9 hour lifetime, the storm produced at least 16 confirmed tornadoes. well organized cyclic supercell long-lived contained a variety of weather hazards produced a wide range of tornado intensities was very well-observed by mobile radars passed within close range of the WSR-88D located at Twin Lakes, Oklahoma. For these reasons and more, it is apparent that this would be a great case to utilize in testing dynamic data assimilation concepts. Figures 4 through 7 illustrate just a few fields that are available from having a dynamically consistent gridded data set produced by a Kalman filter. A Kalman filter provides the same fields as a numerical model, but are based on combining current observations from the radar with model physics in an optimal fashion. Highlighted results: Reflectivity derived from the assimilation closely resembles that displayed in WDSS-II. Several areas of intense vorticity maxima and minima exist, but only one is associated with a strong updraft. A relative minimum in pressure also indicates a rotating updraft. Baroclinic zones, associated with mini-fronts generated by the storm, also help to identify the greatest risk of high winds and tornadoes at their intersection. Future Work: Additional real-life case studies Perform assimilations at different resolutions Change domain coverages More quantitative comparisons between WDSS-II detections and phenomena indicated in assimilated data sets. Figure 5: Low-level cross-section of vertical vorticity with overlaid contours of strong positive vertical velocities (updrafts). Note the strong updraft is associated with negative vorticity, indicating a anti-cyclone. Several minutes later a anticyclonic tornado touched down. Time is 0100 UTC. Figure 6: Low-level cross-section of pressure with. Placement of the local minimum in pressure is in agreement with placement of the main anticyclonic updraft. Time is 0100 UTC. Figure 7: Cross-section of surface temperature with overlaid contours of vertical vorticity. Note that the intersection of the strong temperature gradients (mini-fronts) is associated with the strong negative vorticity area. Time is 0100 UTC. Figure 4: Cross-section of radar reflectivity derived from the assimilated data. Note the structure as compared to that displayed in figure 3. Time is 0100 UTC.