Background
Over half the aircraft delays in the national airspace in the summer months are due to thunderstorms. Unfortunately, reliable detection and prediction of thunderstorms-- while essential to safe and efficient use of the nation's airspace--remains a difficult scientific challenge. To address this challenge the RAL Convective Weather group partners with MIT Lincoln Laboratory and NOAA/GSD on the FAA Aviation Weather Research Program (AWRP) Convective Weather Product Development Team. C. Mueller leads the RAL effort. The team is responsible for developing end-to-end convective weather detection and forecast systems used by airline dispatch offices, FAA traffic flow management, and pilots to make both tactical and strategic decisions. High- resolution, very-short-period (0-2 h) nowcasts are used for tactical rerouting needed to take advantage of jet routes that remain usable near and through areas of convection. Lower-resolution, longer-period (0-6 h) forecasts are used for strategic flight planning.
Current Activities
RAL has taken the lead in the development and deployment of the National Convective Weather Forecast (NCWF) system. NCWF provides probabilistic nowcasts of convective storm location and intensity and combines meteorological observations, feature detection algorithms, and numerical weather prediction model output to provide automated nowcasts of thunderstorm position. The current operational version of NCWF shows the convective hazard detection field and a binary forecast of storm location with a 1-h lead time (NCWF-1). This product was made available to users in 1998 and became operational in 2000. The NCWF-2 forecast is available on experimental ADDS and provides motion vectors and probabilistic forecasts. The NCWF-2 is a preliminary step towards a continuous 0-6 h, probabilistic forecast (NCWF-6).
Under the sponsorship of the National Weather Service RAL is collaborating with the Ft. Worth Weather Forecast Office (WFO) to demonstrate the value a human forecaster adds to automated graphical and gridded 1-h forecasts of storm initiation. The objective of the demonstration is to show that the forecaster can enhance the product without interfering with routine product generation. Forecasters with their knowledge of conceptual models and understanding of the limitations inherent in observations and numerical weather prediction models play a crucial role in integrating a variety of data. The NCAR Auto-Nowcast (ANC) system is unique in allowing WFO forecasters to enter lines of boundary-layer convergence and adjust storm initiation likelihood fields. The forecaster input is used in upstream processes that do not interfere with the timely output of the forecast product necessary for aviation services.
RAL also participates in the Naval Research Laboratory's NOWCAST program, providing nowcasts of convection at 1-2 hours for the region surrounding the Naval Air Station near Fallon, Nevada. This effort is aimed at understanding and simulating the behavior of the atmosphere on local, regional, and global scales and applying this knowledge to the development and implementation of objective analysis and prediction systems for the Navy and other DoD users.
Results
NCWF-2: Probabilistic Forecasts (click on image to enlarge).
J. Pinto, D. Megenhardt and N. Rehak transferred a new version of NCWF-2 to the NWS Aviation Weather Center (AWC). This product became available on the experimental Aviation Digital Data Service (ADDS) in Spring 2005. The installation provided an excellent opportunity to meet with forecasters and discuss how the product could be used to aid in the production of forecasts. Skill scores from the summers of 2004 and 2005 obtained using NOAA’s Real-Time Verification System (RTVS) demonstrated improvement in the forecast skill resulting from system upgrades in 2005. Dissemination was greatly enhanced by A. Braeckel and P. McCarthy’s work to incorporate NCWF-2 output (see Figure) into the ADDS Flight Path Tool, which allows overlaying multiple aviation-relevant fields, radar/forecast looping, and enhanced zoom capability.
NCWF-6: Blended Forecasts (click on image to enlarge).
Illustration of (a) Java realtime side-by-side display tool (b) 6 hour change in convective occurrence from WSR-88D climatology and (c) skill scores for the NCWF, RCPF and Merged systems as a function of lead time based on validation between 1-14 August 2005.
J. Pinto, C. Mueller and D. Megenhardt developed and tested an experimental system for blending extrapolation techniques and RUC model data to produce a 1-6 h probabilistic forecast of convection. The new system produced probabilistic forecasts with more skill than either technique alone (see Figure). Further statistical analyses revealed that the relative weight given to the two systems used to produce the blended forecasts should be a function of time of day, lead time and synoptic regime. For example, NWP forecasts should have greater weight at times (locations) when (where) convection typically initiates, because initiation forecasts are not currently possible with extrapolation. These findings support the development of an operational forecast system that will produce 1-6 h probabilistic forecasts of convection every 15 min. The system will be designed to take into account variability in relative forecast skill of extrapolation and NWP to optimize the way they are blended. G.Cunning and J.Pinto developed a web-based viewer and user manual that were used to facilitate comparison of the systems being blended and will serve as templates for product dissemination and training next year.
R. Roberts, D.Megenhardt, S. Dettling, N. Oien and D. Albo installed, tested, supported and maintained the NCAR Auto-Nowcaster (ANC) at the Dallas-Ft. Worth WFO. This was the first time that the ANC system was used operationally by NWS forecasters. R. Roberts, E. Nelson, D. Albo and D. Megenhardt worked directly with the NWS forecasters to ensure that the user interface to the ANC system was streamlined to enable forecasters to interact with the system and improve forecast quality. The team also visited the NWS and CWSU offices to provide training and support to the ANC system. Based on forecaster input, T. Saxen, H. Cai and D. Megenhardt made changes to the ANC fuzzy logic system by adding new data and modifying weights of existing fields. Initial results of the forecaster-computer mix are very encouraging; modifications based on these results are being made to the ANC and will be operational in May 2006. Finally, R. Roberts and D. Albo are working with the NWS-MDL to allow forecasters to enter boundaries and manipulate the interest fields from the NWS-AWIPS workstation. This transition will facilitate the forecaster’s interaction with the ANC system.
ANC at Dallas-Ft. Worth WFO
In FY2005, radar data from the Fallon SWR radar and nearby NEXRADs have been mosaiced and a storm extrapolation program run on the reflectivity data in a real-time system. This system also ingests the NRL COAMPS numerical model winds for steering level flow and lightning strike data. Results to date are very promising, and real-time validation is underway. Storm extrapolation polygons are imported into the NOWCAST system for display. This effort is led by C. Kessinger with major contributionsfrom D. Megenhardt and B. Hendrickson; J. Pinto, a recent addition to the team, will work to optimize the storm tracking system.