1. Challenges in Urban Meteorology: A Forum for Users and Providers (September 21-23, 2004) Panel 4: Research and Development for Urban Weather and Climate Applications Lloyd Treinish IBM Thomas J. Watson Research Center Yorktown Heights, NY lloydt@us.ibm.com http://www.research.ibm.com/weather http://www.research.ibm.com/people/l/lloydt

2. Panel 4: Research and Development for Urban Weather and Climate Applications We are a small weather modelling group at IBM Research focusing on understanding of mesoscale weather as it relates to weather-sensitive operations and decision support, and developing solutions for such applications We have developed an operational, prototype NWP-based system, which has provided regular forecasts for the NYC-metropolitan area at 1 km resolution as a testbed for over 3 years (“Deep Thunder”) –To evaluate both meteorological and “business” value beyond physical realism –To evaluate the level of practicality and usability at reasonable cost –To develop an operational end-to-end infrastructure and automation with focus on high- performance computing, visualization and system integration –To prototype “business” applications with actual end users Additional testbeds at 2 km resolution for Chicago and Kansas City established earlier this year Although work is on-going and capabilities have limitations, it is sufficiently evolved for some practical urban applications and to assess additional needs

3. Current Focus and Issues  Several specific urban (and suburban) applications with short-term (three to 18 hours) weather sensitivity –Emergency management, homeland security, energy (distribution, operations and generation), road maintenance and operations, traffic management and airport terminal operations  Results to date are sufficiently compelling to enable discussions with local government agencies and private companies for each of these applications in several urban areas in the US  Unfortunately, these interactions are ad hoc and duplicative because –No formal mechanisms exist to engage such groups or to transfer technology or services –There is no central clearing house of information for urban decision makers to become aware of such potential capabilities  Methods are needed to properly validate the utility of such forecasting systems for these applications –Traditional meteorological (statistical) verification is not appropriate for many urban applications –True end-user metrics are often ill-defined or simply qualitative at present  What level of capability is “good enough” for an effective deployment ? –Need for a “heads-up” for severe weather event, even with opportunity for phase errors vs. standard zone forecasts, which may be too vague –More focused dissemination since standard meteorological products are a clear mismatch with end-user decision makers, whose expertise is in applications and understanding the impact of weather as opposed to meteorology

4. An Example (Among Many): NYC Metropolitan Area Heavy Rainfall Event -- 8 September 2004  Remnants of Hurricane Frances moved into the NYC metropolitan area early in the morning of September 8  The heaviest rainfall occurred in an area stretching from northeastern New Jersey through central Westchester County, NY with amounts in excess of 5" in some areas  There was widespread disruption of transportation systems (e.g., road closures, flooded subways, airport delays) and significant flooding in several regions  Evening NWS zone forecast (2130 EDT, 7 September) for the next day: "showers and a slight chance of thunderstorms, rain may be heavy at time in the morning"  Revised NWS zone forecast (0440 EDT, 8 September), adding: "locally heavy rain possible"  NWS issued a flash flood watch at 0748 EDT

5. Forecasted rainfall totals through 24 hours: 4 km nest (above) and 1 km nest (below)  Heavy rainfall predicted for the morning with similar distribution to reported rainfall, although some differences in totals  Forecast initiated with data from 2000 EDT (0Z) with results available about midnight  Significantly different forecast compared to NWS forecast (available about 8 hours before flash flood warning)  Despite some error, significant "heads-up" for event Forecast Results 07 September 2004 Late Evening

6. Measured Rainfall (Inches) JFK2.76 Mamaroneck 3.73 LGA3.83 Central Park3.75 Norwalk4.25 White Plains 5.85 Fair Lawn1.50 Bethpage5.20 Orange2.30 EWR 2.07 Hoboken3.87 A More Focused Presentation of the Forecast May Be Appropriate for Specific Applications (e.g., road maintenance and operations, traffic management) Road Forecast of Rainfall Totals (1 km Nest)

7.  Interest in surface and upper air winds dictates entirely different presentation  “Virtual wind profilers” at two locations within 1 km nest enhanced with trajectories to show forecasted propagation Forecast Results for Other Applications

8. Needs and Future Directions  Despite promising results to date, improvements are clearly needed in several key areas –Improved model representations of boundary layer physics and cloud microphysics –Better observing (sampling) strategies coupled with data assimilation to reduce errors in initial conditions –Continued advances in overall system cost-effectiveness (performance, throughput and usability)  Further prototyping and development of systems that can be used with confidence is required –End-to-end tailoring for specific application focus (throughput, physics and dissemination) – forecast products when they are needed in the way that they are needed –While R&D continues, deploy now even with limitations, to enable earlier understanding of operational constraints and issues as well as to develop user- oriented metrics –Establish additional testbeds for specific urban weather issues or application sensitivity –Emphasize further two-way education and collaboration with potential beneficiaries

9. Needs and Future Directions  Appropriate visualization is critical to enable meteorological data to be usable for urban applications – methods are available but not widely utilized –Understanding of how weather data need to be used and why (e.g., human factors concerning how users work and interact) –Understanding of how users perceive and interpret weather visualizations –Data must be made relevant for different classes of users using their terminology, and thus, expressible in terms that can be readily understood in real-time without expert interpretation  Effective coupling to derived modelling and analysis for proactive planning is necessary but much more work needs to be done –Meteorology needs to be reasonably correct (“good enough”) first, which has been the primary focus for R&D –Direct match to relevant physical problem (e.g., pavement, dispersion) –Direct match to relevant operational problem (e.g., crew and equipment optimization [scheduling and routing] that is impacted by weather)