1997-1998 FIELD TEST

Steven M. Zubrick

Steven J. Weiss

1. INTRODUCTION

The System for Convection Analysis and Nowcasting (SCAN) described by Smith et al. (1998) is an integrated suite of multi-sensor applications which detects, analyzes, and monitors convection and generates short-term probabilistic forecast and warning guidance for severe weather and flash floods automatically within the National Weather Service's (NWS's) Advanced Weather Interactive Processing System (AWIPS). The goals of SCAN are:

- to provide forecasters with accurate, timely, and consistent severe weather and flash flood guidance,

Corresponding author address: S. Smith, TDL/NWS, SSMC2, 1325 East-West Hwy., Silver Spring, MD, 20910 (Stephan.Smith@noaa.gov)

- to develop "smart" computer displays, menus, and graphical user interfaces that optimize the utility of the AWIPS Display 2 Dimensions (D2D; Biere 1998) and are compatible with the warning decision process,

- to develop multi-sensor databases to support the verification of thunderstorm, severe weather, and flash flood forecasts/warnings,

- to supplement forecaster event monitoring with multi-sensor, automated event monitoring,

- and, to accelerate the rate of technology transfer from research to operations.

Operational implementation of SCAN will result in longer lead times on warned events, fewer missed events, increased forecaster situational awareness, reduced forecaster fatigue during warning situations, rapid improvement in implemented techniques, and a well-defined focus for applied research.

During the 1997 and 1998 convective seasons, a SCAN prototype consisting of the National Severe Storms Laboratory's (NSSL's) Warning Decision Support System (WDSS), the National Center for Atmospheric Research's (NCAR's) Thunderstorm Autonowcaster and the AWIPS Thunderstorm Product/1-hr Quantitative Precipitation Forecast (QPF) was tested at the NWS forecast office in Sterling, Virginia. The purpose of the field test was to integrate the above software applications and to examine the usefulness of their combined products as tools for forecasters responsible for short-term forecasts and warnings. The Storm Prediction Center (SPC) provided probabilistic Day-3 convective outlooks (Hovis et al. 1998) for the SCAN test area primarily to assist in travel planning and staffing coordination for the larger convective events.

This paper describes the functionality of the SCAN prototype and the preliminary findings of the field tests as they pertain to severe weather and flash flood warnings. Additional information on the SCAN field tests, the SCAN AWIPS development schedule, and other SCAN-related items is available from the SCAN homepage at www.nws.noaa.gov/tdl/scan/scan2.html.

2. SCAN PROTOTYPE

The Research and Applications Program (RAP) of NCAR has over the past several years developed and refined an integrated system designed to produce automated guidance on the detection, tracking, and short-period (0-30 min) forecasting of thunderstorms. The system is conceptually based on a large compendium of operational and modeling research acquired over the years on the nature of convective storms.

The complete Thunderstorm Auto-Nowcast (TAN) system (Mueller et al. 1997) is comprised of about half a dozen major sub-systems and is designed to integrate and process data from various data platforms (radar, satellite, upper-air soundings, surface mesonet, etc.) using sophisticated algorithms for feature detection and tracking. These processed data are combined with boundary layer model output and heuristics of thunderstorm extrapolation and trending to produce area-specific, short-period (0-30 min) forecasts of convection (defined to be where radar echoes of 35 dBz will be located).

Similarly, NSSL has maintained a long-term effort to develop algorithms that utilize integrated meteorological data to identify severe weather phenomena. In support of these applications, NSSL has also designed and implemented new concepts for displaying critical information used in the warning decision process. The combined (algorithms and display) system, called the Warning Decision Support System (Eilts 1997), has been tested operationally for several years at selected NWS offices across the United States.

Over the past several years, the NWS's Techniques Development Laboratory (TDL) has engineered both algorithms and software applications focused on the diagnosis and prediction of thunderstorms, severe weather, and heavy precipitation. Two of the more recent of these, the AWIPS Thunderstorm Product (Churma and Smith 1998) and 1-hr Quantitative Precipitation Forecast (QPF), were chosen for testing within the SCAN prototype. The former provides automated thunderstorm detection at fixed geographic locations, while the later generates probabilistic 60-min forecasts of precipitation based on an extrapolative-statistical approach (Kitzmiller 1996).

Since WDSS was designed with warning decisions in mind, and we are principally exploring issues related to the implementation of new guidance products to assist forecasters in making warning decisions, WDSS was chosen as the vehicle on which the integration of all applications would take place. In the case of the TAN, all the various sub-systems ran on independent hardware and only a selected subset of the output was displayable on WDSS (the complete TAN output was available to the forecasters on the independent TAN workstation). For the Thunderstorm Product and the 1-hr QPF, however, the software was completely integrated into WDSS.

From a conceptual standpoint, WDSS and Thunderstorm Product components of the system provide detection and monitoring of severe and non-severe thunderstorms. The TAN and 1-hr QPF provide a limited predictive capability (as a necessary step in generating its forecasts, however, the TAN also detects boundaries and growing cumulus clouds).

Fundamentally, we hope to explore the utility of combining the basic information provided by "automated severe thunderstorm detection and monitoring" with that of "automated thunderstorm nowcasting" in order to improve the warning decision process. The linkage of these two key functionalities within AWIPS could provide the basis for an automated, "suggested" severe thunderstorm warning area.

Significant enhancements to WDSS and TAN occurred in 1998. Most notably, an automated anvil tracker based on GOES infrared satellite data (Zaras and Rabin 1998) and the Area Mean Basin Estimated Rainfall (AMBER; Jendrowski and Davis 1998) algorithm were added to WDSS. The full suite of TAN applications including a satellite-based cumulus detection algorithm, a correlation radar echo area tracker, an ingest of surface mesonet data, and an adjoint numerical model were implemented in 1998. As of this writing, the 1998 field test was not complete and therefore the findings reported here are based only on our limited experience from late summer of 1997. Initial results from the 1998 field test will be presented at the conference.

3. SUMMARY OF 1997 OPERATIONS

The WDSS component of the SCAN prototype was installed in May, with all staff forecasters trained on its use by the end of June. The Thunderstorm Product, 1-hr QPF ,and limited version Thunderstorm Auto-nowcaster components were installed in July, with all forecasters receiving additional training by mid-August. The integrated SCAN evaluation lasted from mid-August through the end of September, which unfortunately, for the purpose of the test, coincided with an extremely quiet period of convective weather across the Mid-Atlantic Region.

4. PRELIMINARY FINDINGS

A post-'97 field test survey of the Sterling forecasters indicated that the most-used products in the SCAN prototype were (grouped by application):

WDSS products

- Cell table

- Mesocyclone table

- Reflectivity cross sections

- Time history trends (especially cell and mesocyclone trends)

- Animation of various 88D graphics

TAN products

- Tracking Radar Echoes by Correlation (TREC) 2-dimensional wind fields

- Boundaries (detection and extrapolation)

- 30-min forecasts of thunderstorms

TDL products

- Probabilistic 1-hr QPF of > 1.00 in.

Viewed as an integrated system, the SCAN prototype was favorably received by the forecasters. Most of the initial feedback focused on ways that the utility of the TAN could be enhanced, both in the SCAN prototype itself and eventually in AWIPS. All forecasters liked having access to boundary detection/extrapolation information. The ability to overlay these products on standard radar products was deemed to be highly desirable. The Sterling staff felt the 30-min forecasts of thunderstorms would be useful in the warning decision process, but also expressed a desire for 60- or 120-min forecasts as a tool for the short-term or public forecaster. No other product generated as much excitement as the TREC 2-dimensional wind field from the TAN. The forecasters really felt this product helped them locate areas of convergence (divergence) and wanted it added immediately to the SCAN prototype display (it was not included originally). Although the lack of convective activity precluded the evaluation of the SCAN prototype as a tool for flash flood warnings, Sterling forecasters did find the probabilistic 1-hr QPF of > 1.00 in to be useful when heavy precipitation cells were detected by the radar.

SCAN participants (both research and operational meteorologists) found the probabilistic Day-3 convective outlooks provided by SPC to be useful in alerting for the potential of a major convective event. Users found that the most useful information was obtained from comparing the latest Day-3 with the previous Day-3's and with the climatological probabilities. In other words, lacking an innate feel for the absolute probabilities, the meteorologists quickly applied a personal calibration to them and, in this manner, derived value from the forecasts.

In general, it is clear that the SCAN 1997-1998 Field Test has been an extremely useful exercise for all parties involved. The NWS operational forecasters at Sterling were exposed to a number of state-of-the-art forecasting/warning tools that are targeted for implementation on AWIPS. The meteorologists and developers at NSSL, NCAR, TDL, and SPC were given the opportunity to test experimental applications and forecast products in a real-time operational environment. In so doing, they will be able to modify and enhance these applications in a manner that best lends itself to how the operational forecasters perform their jobs in the modernized NWS.

5. ACKNOWLEDGMENTS

The authors would like to thank the programmers and meteorologists at TDL, NSSL, and NCAR for their skilled work in programming and developing the SCAN prototype. We would also like to thank the forecasters and operational staff at Sterling for their cooperation and assistance in organizing and carrying out the 1997-98 field test. The SPC outlook forecasters are gratefully acknowledged for the their professional expertise in preparation of the SCAN Day-3 convective outlooks. SCAN is funded by the NWS, the National Science Foundation, and the U.S. Weather Research Program.

6. REFERENCES

Biere, M., 1998: The WFO-Advanced Two-Dimensional display software design. Preprints, 14^th International Conf. on Interactive Information and Processing Systems. Phoenix, Amer. Meteor. Soc., 376-379.

Churma, M.E., and S. B. Smith, 1998: Evaluation of the AWIPS Thunderstorm product. Preprints, 16^th Conf. on Weather Analysis and Forecasting, Phoenix, Amer. Meteor. Soc., 472-474.

Eilts, M. D., 1997: Overview of the Warning Decision Support System. Preprints, 28^th Conf. on Radar Meteorology. Austin, Amer. Meteor. Soc., 402-403.

Hovis, J. S., S. J. Weiss, H. E. Brooks, and J. S. Evans, 1998: Probabilistic Day 3 convective outlooks in support of the System for Convection Analysis and Nowcasting (SCAN) Project. Preprints, 19^th Conf. on Severe Local Storms., Minneapolis, Amer. Meteor. Soc., (this volume).

Jendrowski, P., and R. S. Davis, 1998: Use of geographic information systems with the Areal Mean Basin Estimated Rainfall algorithm. Preprints, Special Symposium on Hydrology. Phoenix, Amer. Meteor. Soc., 129-133.

Kitzmiller, D. H., 1996: One-hour forecasts of radar-estimated rainfall by an extrapolative-statistical method. TDL Office Note 96-1. 26 pp. (available from author).

Mueller, C., R. D. Roberts, and S. G. Henry, 1997: Thunderstorm automated nowcast system - real-time demonstrations. Preprints, 28^th Conf. on Radar Meteorology. Austin, Amer. Meteor. Soc., 406-407.

Smith, S.B., T. M. Graziano, R. A. Lane, W. O. Alexander. M. D. Eilts, J. T. Johnson, J. W. Wilson, R. D. Roberts, D. W. Burgess, D. H. Kitzmiller, R. E. Saffle, R. C. Elvander, S. M. Zubrick, J. T. Schaefer, S. J. Weiss, and D. A. Imy, 1998: The System for Convection Analysis and Nowcasting (SCAN). Preprints, 16^th Conf. on Weather Analysis and Forecasting, 14^th International Conf. on Interactive Information and Processing Systems. Phoenix, Amer. Meteor. Soc., J22-J24.

Zaras, D. S., and R. A. Rabin, 1998: Identification and tracking of cold cloud features in satellite imagery. Preprints, 19^th Conf. on Severe Local Storms., Minneapolis, Amer. Meteor. Soc., (this volume).