SECTION 1: PROJECT OBJECTIVES AND ACCOMPLISHMENTS
The project objective is to evaluate the operational utility of very high resolution (6 km) NWP model data for a spectrum of forecaster identified phenomena presenting significant QPF and aviation forecast challenges. Project outcomes were the following: (i) routine forecaster access to very high resolution model forecast data and concomitant improvements in forecast skill; (ii) development of methodologies that will ease the transition of NWS forecasters into the environment of ever-increasing data availability; (iii) further insight into the development of optimal national modeling strategies. To accomplish the project objective and secure these outcomes, the following overarching tasks were conducted: (a) evaluation of very high resolution NWP model data; (b) logistics for providing routine forecaster access to the high resolution model forecast data in realtime. Below, we summarize responsibilities, activities and accomplishments related to each of these tasks.
A. Evaluation of very high resolution NWP model data
Six major phenomenological types were identified to be of interest to the study: (1) warm season convection; (2) development, movement and intensity of lake effect snowbands (LES); (3) winter cyclones; (4) timing and inland penetration of lake/bay breezes; (5) local winds over the bay of Green Bay during the summertime for nearshore lake forecasts; (6) stratus and fog. The focus of the evaluation effort was intended to be on cases that occur under realtime constraints.
Both the UWM and NWS PIs identified a set of cases of convection (#1), LES (#2) and winter cyclones (#3) that occurred during the period in which the UWM realtime system has been operational and archived (Summer 1998 - present). A paper detailing the evaluation of realtime lake breeze forecasts (#4) has already been published in the AMS journal Wea. Forecasting (Roebber and Gehring 2000). Results of this study show the utility of the model in predicting the onset and relative westward extent of the lake breeze. Further improvement in this guidance will require (at the minimum) improved initialization of lake temperatures and better boundary layer physics. The last two categories, local winds over Green Bay (#5) and stratus/fog (#6) are being evaluated primarily through NWS forecaster feedback (see section 1.1b below).
To facilitate the verification effort for categories 1-4, an online database of the realtime forecasts was generated by transferring the archived data from DAT tape to hard disk.
Evaluation of the forecasts represented in this database was conducted with the help of two UWM masters degree students (James Sieveking and Michael Fowle) and an LTE employee who recently completed his masters degree under PI Roebber (Darrel Johnson).
Verification of convective forecasts (#1) suggests that the numerical guidance is useful for determining the existence, timing, location and mode of these events, particularly where there is larger-scale organization (e.g. squall lines or multicellular convective systems). As anticipated, individual storms cannot be predicted but their aggregate effects can, in many cases. The possibility of using FDDA techniques to overcome timeliness limitations was evaluated; however, this option was discarded as too costly relative to forecast data accuracy. A thesis on this topic was produced:
Fowle, M., 2001: Study of the convective forecast skill of very high resolution (6-km) NWP model data, M.S. Thesis, University of Wisconsin at Milwaukee, 93 pp. [Available from Department of Mathematical Sciences, University of Wisconsin at Milwaukee, WI 53201- 0413].
A draft version of this work for publication in the AMS journal Weather and Forecasting has been created. Submission is expected sometime in the next 6 months.
Results of the verification of LES (#2) suggest that the model can provide exceptional guidance for the development of long-axis wind parallel LES, including those involving embedded vortex structures (e.g. 19 Feb 1999). Results were also encouraging for the one identified case of short-axis wind parallel LES. These analyses will be continued by NWS-GRB forecasters under the guidance of the NWS PI.
The winter cyclone verification (#3) was comprised by a detailed case study of a specific cyclone event (the 2-3 January 1999 midwest blizzard; maximum snowfall rates of 1 to 1.5 inches per hour in embedded snowbands, 10-20 inch totals, winds to 60 mph and 8 foot drifts). A thesis from this study was produced:
Sieveking, J., 2001: 24-48h predictability of lake-enhanced snowbands in the 2-3 January 1999 midwest blizzard, M.S. Thesis, University of Wisconsin at Milwaukee, 66 pp. [Available from Department of Mathematical Sciences, University of Wisconsin at Milwaukee, WI 53201-0413].
A draft of this work is underway, with the goal of publication in the AMS journal Weather and Forecasting.
An additional, subjective verification of each of the six phenomenological types was conducted in realtime by NWS forecasters (see section 1.1b).
B. Logistics for providing realtime forecaster access to high resolution model forecast data
The UWM realtime modeling system was developed initially through the sponsorship of the National Science Foundation (NSF). As part of that system, basic forecast data were developed in two formats: web-based graphics and Vis5d. In the proposal stage, the plan was to develop the forecast data in multiple formats: modified web-based graphics, Vis5d (primarily for the SOO) and GRIB (the standard format for model data produced by NCEP). However, the NWS PI recognized that web-based graphics and Vis5d would not allow these data to be smoothly blended into the operational routine and would therefore restrict its adoption by forecasters, whereas the GRIB format insured that data organization would conform to operational custom with respect to both tools and aesthetics.
In the proposal, only the KGRB NWS office was to be engaged in the operational evaluation of the model forecasts. However, the NWS office at KMKX joined the study in an unfunded capacity, making it possible to conduct subjective evaluations by forecasters from two offices at the close of each day shift for which one of the above-noted phenomenological types occurs, via a one-page questionnaire. The questionnaire, jointly designed by the NWS and University PIs and vetted by forecasters, was easily filled out, yet bears directly on the objectives of this study. These subjective forecaster evaluations were begun on 2 October 2000 at KGRB and 1 November 2000 at KMKX. Evaluation of the collected evaluations is ongoing.
On-site training occurred at both KGRB and KMKX to establish forecaster facility in the use of very high resolution model data. This process was initiated with several on-site visits by the academic PI. The first took place at KGRB on 18 July 2000 (observation of operations), the second at KMKX on 10 August 2000 (observation of operations), the third at KGRB on 7 September 2000 (forecast seminar), the fourth at KMKX on 21 September 2000 (forecast seminar) and the fifth at KMKX on 16 November 2000 (forecast seminar).
Oral dissemination of study results was conducted in the form of two poster presentations at the AMS Mesoscale conference meeting in Fort Lauderdale, FL July 2001.
SECTION 2: SUMMARY OF UNIVERSITY/ NWS EXCHANGES
In November 2000, a comprehensive training seminar was developed and presented for scientists and emergency response teams for two local lakeshore nuclear power plants (Kewaunee and Point Beach) located within the NWS Green Bay office county warning area (CWA). The purpose of the seminar was to provide advanced training on: (1) identification and detection of lake breeze boundaries affecting eastern Wisconsin using satellite, radar and surface observations (ASOS and power plant wind towers); (2) the structure and physical characteristics of lake/land breezes and their influence on plume transport and (3) utilization of the Lake Breeze Index (LBI) and the realtime 6km MM5 model output to help forecast lake breeze onset and inland penetration. As a result of the training, the nuclear power plant scientists and emergency response teams not only have gained a better understanding of lake/land breeze influences on plume transport, but also plan to incorporate the realtime MM5 model forecasts currently available on the Web, into their emergency response planning in the event of an accidental release.
The original project PIs have established cooperation in this project with the NWS-KMKX office, allowing an extension of the evaluation beyond the original KGRB CWA. A verification study of realtime lake breeze forecasts was published in the AMS journal Wea. Forecasting (Roebber and Gehring 2000). This work was part of the general evaluation of the model, although not contained within the COMET proposal. The results of this evaluation will be useful to the current study and also helped to test evaluation methodologies. Additional seminars were conducted at KGRB and KMKX, to discuss the findings of the academic PI's work with respect to convective forecasting and the development of the lake breeze.
A paper investigating the potential utility of the MM5 in an extreme flooding event (21 June 1997) was published in Wea. Forecasting (Roebber and Eise 2001). The University PI presented several seminars at KGRB and KMKX to introduce forecasters to the model. Numerous other presentations have been made (see section 3.1) relating to this work.
SECTION 3: PRESENTATIONS AND PUBLICATIONS
Fowle, M.A. 2001. Study of the convective forecast skill of very high resolution (6-km) NWP model data. MS-Thesis, Department of Mathematical Sciences, University of Wisconsin-Milwaukee. 93 pp.
Fowle, M.A., 2001: Predictability of convection at 24-48h forecast range using a very-high resolution (6 km) NWP model. Ninth Conference on Mesoscale Processes, 30 July - 2 August 2001, Fort Lauderdale, FL.
Gehring, M.G, 1999: Realtime prediction of the lake breeze on the western shore of Lake Michigan. MS- Thesis, Department of Geosciences, University of Wisconsin-Milwaukee.
Johnson, D.A., 1999: A diagnostic case study of bent-back frontogenesis and associated severe weather over the U.S. Great Plains. MS-Thesis, Department of Mathematical Sciences, University of Wisconsin-Milwaukee, 95 pp.
Roebber, P.J., 1999: Storm-scale NWP in real-time for the upper midwest. Invited talk at the Department of Geography, Northern Illinois University, 19 March 1999.
Roebber, P.J., 1999: The role of Lake Michigan in modulating upstream weather conditions: experience with realtime forecasts using MM5 in warm and cold seasons and priorities for future model development. The Ninth Penn State/NCAR MM5 Users' Workshop, 23-25 June 1999, Boulder, CO.
Roebber, P.J., 1999: Synoptic input, sub-synoptic output: Model noise or reality? Invited talk at the Illinois State Water Survey, Champaign-Urbana, Illinois, 27 September 1999.
Roebber, P.J., 1999: Synoptic input, sub-synoptic output: Model noise or reality? Invited talk at the NWS Winter Weather Workshop, Sullivan, Wisconsin, 3 November 1999.
Roebber, P.J., 2000: The use of storm-scale resolution numerical weather prediction for operational forecasting of atmospheric convection. International Association for Mathematics and Computers in Simulation (IMACS) 2000 Conference, Milwaukee, Wisconsin, 25-27 May 2000.
Roebber, P.J., 2000: Synoptic input, sub-synoptic output: Model noise or reality? Invited talk at the National Severe Storms Laboratory, Norman, Oklahoma, 15 December 2000.
Roebber, P.J. and M.G. Gehring, 2000: Realtime prediction of the lake breeze on the western shore of Lake Michigan. Wea. Forecasting, 15, 298-312.
Roebber, P.J. and. D.A. Johnson, 2000: Analysis and model simulations of a case of bent-back frontogenesis and associated severe weather over the U.S. Great Plains. The Eleventh Extratropical Cyclone Workshop, 27-31 August, 2000, Pacific Grove, California.
Sieveking Jr., J.E. 2001. 24-48 hour predictability of lake-enhanced snowbands in the 2-3 January 1999 Midwest blizzard. MS-Thesis, Department of Mathematical Sciences, University of Wisconsin- Milwaukee. 65 pp.
Sieveking Jr., J.E. 2001. 24-48 hour predictability of lake-enhanced snowbands in the 2-3 January 1999 Midwest blizzard. Ninth Conference on Mesoscale Processes, 30 July - 2 August 2001, Fort Lauderdale, FL.
SECTION 4: SUMMARY OF BENEFITS AND PROBLEMS ENCOUNTERED
4.1 Benefits to the University
The single greatest benefit to the University partner is a formal, working (research) relationship with the NWS offices at KGRB and KMKX. This allows the study of more applied problems and exposure of graduate and undergraduate students to the problems and requirements of operations. Additionally, the University gains access to technologies being applied in the NWS offices (e.g. AWIPS) and obtains a greater appreciation and understanding of current forecast problems. The NWS office at KGRB has proven to be an effective and enthusiastic partner, while KMKX has provided unfunded support to this project through their voluntary participation.
In the past several years, the exposure of students at UWM to the operational environment has increased, through formal venues such as COMET and the informal opportunities that arise in conjunction with such efforts. A number of our recent graduates have gained employment in NWS offices in the Western and Central Region, which may in part be a reflection of this added exposure to operations.
This work has led to several masters theses and associated publications in the refereed literature. The benefits of this work have been sufficient to encourage the academic and NWS PIs to plan further collaborations on other outstanding forecast problems in the region.
A generic problem is associated with financing of COMET Cooperative projects. Because of the funding limits of approximately $30K/year, it is not possible to fully fund graduate students from these projects. They are partially funded during the academic year and/or summer months, with the remainder of their funding coming from other grants or teaching duties. This means that they cannot be as efficient in completing project tasks as they would otherwise be, since their time is divided.
Related to this is the issue of
computer equipment. Because the current equipment at UWM is a 4 CPU SGI O200,
we are providing 48-h 6-km forecasts once daily, with the first (second) 24-h
completing at UWM at approximately 2:30 am (7:00 am). There was some consideration
given to using FDDA from the 1800 UTC forecast cycle to "buy" some
time and make these data more available to the midnight shift, but this option
was not pursued in the interest of retaining the highest possible model skill.
The acquisition of computer equipment remains a substantial obstacle. Some means
for obtaining processing capability through the COMET program (perhaps with
some form of cost matching) should be entertained.
4.2 Benefits to the NWS office
As a result of the training provided to the local nuclear power plants mentioned in section 2.1 above, NWS Green Bay has gained access to wind data from a tower located at the Kewaunee Nuclear Power Plant located near the western Lake Michigan Shoreline. Access to wind data from a tower located at the Point Beach Nuclear Power Plant is also currently being pursued. This data, in combination with the realtime 6 km MM5 model forecasts, has helped to improve the NWS nearshore forecasts and warnings. The 00Z 6 km MM5 model forecasts have increased forecaster confidence in forecasting lake breeze onset and to a lesser extent, it's inland penetration. The NWS PI has provided additional training to the NWS Green Bay forecast staff on use of the Lake Breeze Index (LBI) in concert with the real-time MM5 model forecasts.
The MM5 model forecasts have provided useful information with respect to forecasting mid-lake and shore-parallel snowbands affecting eastern Wisconsin. The 6 km MM5 forecasts has proven particularly useful in forecasting the timing and inland penetration of weaker, more subtle LES events (e.g., snowfall amounts not necessarily reaching advisory criteria) that typically are not forecast well or missed entirely by the coarser resolution NCEP models. The 6 km MM5 forecasts have also provided insight into forecasting the potential for lake-enhanced snowfall along the eastern Wisconsin shoreline associated with synoptic-scale cyclones.
As mentioned in section 4.1 above, the primary limiting factor to maximum operational utilization of the MM5 model forecasts has been its timeliness. Although the first 24 hours of the model forecast are typically available by around 2:45 am, it is still a challenge for the forecaster to assess the entire suite of NCEP models and adequately assess the MM5 model output in time to meet the 4:00 am morning forecast package issuance deadline. This has somewhat discouraged full assessment and utilization of the MM5 model output operationally. However, even despite these limitations, even a "quick look" at the 6 km forecasts has proven to provide additional forecaster confidence, especially for LES events affecting eastern Wisconsin.
Also due primarily to limitations in computer power, forecasters have also been unable to utilize the 6 km MM5 forecasts to assess LES bands affecting the northern portion of the CWA off Lake Superior.