OBJECTIVE WEATHER PREDICTION PROJECT (P. Dallavalle)

Short-Range Weather Forecasting (P. Dallavalle): Through the Collaborative Science, Technology, and Applied Research Program (CSTAR), the Office of Systems Development is funding statistical forecasting studies to be done at the recently-created Cooperative Institute at Pennsylvania State University. In a project entitled "Advanced Statistical Post-Processing Studies," principal investigators J. M. Fritsch and R. Vislocky propose to develop and test an enhancement of the MOS approach in which MOS guidance can be generated at any desired forecast location regardless of the availability of a historical record of observations. This "second-order" MOS forecast would be obtained by the use of MOS forecasts at nearby sites, an actual or estimate climatology of the forecast event, terrain elevation and aspect, land use values, and, possibly, output from the National Centers for Environmental Prediction's (NCEP's) numerical weather prediction model. Successful completion of this approach would enable generation of a gridded, digital database with fine-scale resolution that could be used to initialize the Interactive Computer Worded Forecast (ICWF). During the first year of the study, the feasibility of predicting the maximum/minimum (max/min) temperature, dew point, and wind speed in this fashion will be tested. Currently, the monthly gridded climatology is being generated for these weather elements. The 30-year NCDC climatology for cooperative stations is being used for the max/min temperature; TDL archives of hourly observations are being used for the dew point and wind speed. Eventually, the climatology will be mapped to the 20-km grid used in the ICWF program. A simple experiment was also conducted in which the 24-h forecast max/min temperatures at 12 current MOS sites were estimated by ignoring the actual MOS guidance at those sites and estimating "second-order" MOS forecasts from nearby MOS sites. Two schemes were tried. In the first,, a simple correction for the difference in climatic normals was used. In the second, the standard deviation of the departure from normal of the nearby MOS sites was applied to determine the forecast temperature at the "non-MOS" sites. In both instances, the "second-order" MOS forecasts were about 0.3 degrees F less accurate than the traditional MOS approach. Further tests will use more sophisticated correction schemes.

Development of the MOS-2000 system continued. This new system is designed to provide tools to develop statistical guidance more efficiently in an era when changes in the numerical weather prediction models are made frequently. A new database of observed hourly data has been created on the CRAY. We are now checking software that will put the observational data into the MOS-2000 format. Software to generate predictor data for use in regression analyses was written, and new processes to archive the Eta model forecasts in the MOS-2000 format are nearing completion.

M.Antolik presented a talk entitled: "Synoptic-scale statistical quantitative precipitation forecasts: Present and future" at the 21st Annual Meeting of the National Weather Association held in Cocoa Beach, Florida, in December. This discussion summarized the current state of the NGM MOS quantitative precipitation guidance and highlighted some of our plans for future development. Meetings with staff members of the office of Hydrology (OH) and the Office of Meteorology (OM) were also held to discuss TDL's contributions to a risk reduction exercise that will be held in the Eastern Region beginning in 1997. This exercise is designed to demonstrate the end-to-end forecast process required to produce probabilistic estimates of quantitative precipitation and subsequent river stage information. We will shortly begin development of NGM-based products required for the risk reduction. These products will be generated by post-processing the current NGM MOS quantitative precipitation forecasts.

Medium-Range Weather Forecasting (M. Erickson): Efforts to produce statistically-generated products from the output of the Medium-Range Forecast (MRF) model ensembles are continuing. Combining the MOS forecasts from the 11 NCEP MRF ensembles runs with the operational MRF-based MOS forecasts, we have 12 estimates of the surface weather conditions for projections out to 8 days in advance. From these values, we compute the mean, range, and standard deviation of the ensemble guidance. Efforts are now underway to evaluate the guidance, to establish a new archive of the ensemble model data, and to determine the most effective ways of disseminating the statistical information to interested users. Coordination meetings were held with NCEP's Hydrometeorological Prediction Branch and the NCEP Advisory Committee on Ensemble Forecasting.

National Verification Processing (V. Dagostaro): In support of the AFOS-Era Verification (AEV) program, data for the 1996 warm season were quality-controlled, and preliminary Max/min temperature and PoP verification summaries were produced. The summaries have not yet been distributed to the regional Scientific Services Divisions (SSD's) due to the need to check the precipitation amount observations collected by ASOS. For max/min temperature and PoP, we compared the local and NGM-based MOS forecasts for approximately 95 stations in the conterminous U.S.; no data were available for 6 stations in Alaska.

Work to convert and enhance the central AEV data processing software to run on a UNIX-based platform is continuing. Software to collect and temporarily store data on the CRAY has been tested and submitted for implementation in NCEP's operational job stream. Software to process and permanently store the data on the CRAY is now being tested in parallel with the existing HDSbased data archiving system. Quality control and data inventory software have been converted and tested on the CRAY. A procedure to disseminate electronically the AEV quality control and verification summaries has been defined and tested. Output that was formerly produced on paper or floppy disks and dissemineated by mail to the SSD's will now be disseminated electron ically via an Office of Systems Operations (OSO) file server. For test purposes, data inventories for April - June 1996 and verification summaries for ceiling height, visibility, wind speed, and wind direction were placed on the OSO file server.

Severe Weather Prediction (R. Reap): The operational NGM-based three-dimensional trajectory model was implemented on the CRAY mainframe in December. Alphanumeric and graphics products from this version of the trajectory model, including the new aircraft icing products for NCEP's Aviation Weather Center (AWC), will be transmitted to the field starting as soon as possible in early 1997. In support of the graphics implementation, the computer software used to produce automated convective outlooks of thunderstorms and severe local storms was extensively revised to generate the products in GRIB format for display on the N-AWIPS system. The new N-AWIPS displays portray geographical areas of expected thunderstorm and severe local storm occurrence, similar to the AFOS displays that were generated on the HDS mainframe. The outlook areas are, however, denoted by plotted numbers rather than by contours. For severe local storms, the value of the plotted number denotes the risk category associated with the outlook area, that is, 1 = approaching, 2 = slight risk, 3 = moderate risk, and 4 = high risk. The automated outlooks will be transmitted to AWC and NCEP's Storm Prediction Center (SPC) for use as guidance for the day 1 and day 2 forecasts. We are currently working on the graphics programs needed to produce the parcel trajectories and the automated convective outlooks for display on AFOS.

LOCAL TECHNIQUES DEVELOPMENT PROJECT (R. Reap)

0-3 Hour OPF and Severe Weather (D. Kitzmiller): We have added a page to the TDL web site for real-time 0-1 hour rainfall forecasts for the Sterling, Virginia, radar umbrella. The page features a categorical rainfall forecast, the probability of 0.1 inch or more of rain, and the cell-based 1-inch rainfall probability. The forecasts are updated every 30 minutes at 0 and 30 minutes past the hour. New forecasts are issued when the radar is in precipitation mode and at least 5 volume scans of data are available in order to estimate echo velocity through pattern matching.

All 0-1 h rainfall forecast products are now available through the prototype AWIPS Thunderstorm Product interface. The rainfall products are generated externally to the Thunderstorm Product's own modules, but are retrieved in gridded form and displayed at the user's request.

We have developed experimental algorithms for forecasting 0-3 h rainfall amounts by an extrapolative-statistical approach. Probability equations for point rainfall amounts in excess of 0.1, 0.5, and 1 inch have been developed. Statistical predictors include extrapolation forecasts of reflectivity from the national 10-km Radar Coded Message (RCM) mosaic, and humidity and stability indices from the NGM. The rainfall predictand data were derived from Stage III radar-gage estimates over the Central Plains during the June-September 1996 period. Probabilities are defined for boxes within the 40-km Manually-Digitized Radar (MDR) grid. The most important predictors for the 0.1-inch threshold are maximum forecasted reflectivity level and mean relative humidity; for the higher thresholds reflectivity and stability (the K index) are the most significant predictors.

Experimentation with 3-h extrapolation forecasts of reflectivity from the national RCM mosaic continues. We found that the extrapolated echo motions are consistent with the synoptic pattern. Some testing of the probability equations mentioned above was begun in order to check for consistency among the probability, radar echo, and humidity fields.

The radar-based nowcast algorithms for probabilities of large hail or any severe weather, running within the AWIPS Thunderstorm Product, were modified to account for radar site location. For sites west of 850 W, equations derived from Plains radar and severe weather data were used; for other sites, equations derived from data in the Mid-Atlantic states were used.

Thunderstorm Identification and Forecasting (S. Smith): We continued our real-time evaluation of the AWIPS Thunderstorm Product on the AWIPS computers, specifically with regard to data flow and general product performance. We also initiated our post-convective season evaluation of the product with a detailed analysis of the 24 June 1996 severe squall event that affected the Washington, D.C., metropolitan area. For this event, the AWIPS Thunderstorm Product achieved a probability of detection of 87%. Other convective days during 1996 are also being analyzed in order to generate a sufficiently large sample that will provide more reliable verification statistics on the product's performance.

In coordination with OM, we developed an initial set of requirements for the System for Convection Analysis and Nowcasting (SCAN). This system will integrate existing software packages in order to provide automated, objective warning guidance for severe weather and flash floods. A joint development effort for SCAN was subsequently organized with participating members from NCAR, NSSL, OSD, OH, and OM. Input from these members, by means of conference calls in November and December, led to further revision and clarification of the new SCAN requirements.

Final revisions were made to a paper entitled "Comments on 'An Interesting Mesoscale Storm-Environment Interaction Observed Just Prior to Changes in Severe Storm Behaviorl". The paper was accepted for publication in the Journal of Weather and Forecastina.

Local AWIPS MOS Program (LAMP) (J. Ghirardelli): We continued our work on the new cloud layer algorithm for LAMP, which will provide cloud height and amount forecasts for up to three layers of clouds. The main effort during this quarter was to develop as many of the algorithm thresholds as possible on the HDS mainframe computer before its removal. The persistence thresholds were completed for all three cloud elements (ceiling and lowest and highest cloud layers), for all eight start times, and for both warm and cool seasons. These thresholds will be used to decide whether to accept the LAMP-forecasted cloud height or merely persist the observations.

Substantial progress was also made on developing the layer separation thresholds; these are used to determine how many layers of clouds to forecast. All of the forecast data needed to develop the thresholds were created, including forecasts for the three cloud elements and best category forecasts of sky cover. Concurrent with the forecast data creation, the computer software to create the layer separation thresholds was modified and is nearing completion.

A talk entitled "The Local AWIPS MOS Program Graphical Display: Guidance for Short Term and First Period Forecasts" was given by S. Kelly in December, 1996, at the National Weather Association's 21st Annual Meeting held in Cocoa Beach, Florida.

Last quarter, we started producing LAMP forecasts at 0400 UTC, one hour prior to the regular start time of 0500 UTC, for use in issuing the 0600 UTC TAF. This quarter we calculated the temperature, dewpoint, and wind speed mean absolute error and bias for a selected group of 17 eastern stations in the U.S. for forecasts made at 0400 UTC and at 0500 UTC. Based on this limited sample, the new method appeared to produce reasonable forecasts. The wind speed errors were almost identical from the two different runs. However, the temperature forecasts were consistently too high around the time of sunrise. Aside from this "sunrise effect," the temperature and dewpoint forecasts from the 0400 UTC run were found to have mean absolute errors similar to those from the 0500 UTC run.

Work was started with PRC to obtain observations, grid, and MOS datasets for use by LAMP in the AWIPS era. A new program was also written to decode observations for the LAMP test system.

Heavy Precipitation Forecasting (J. Charba): The LAMP QPF system has been running since October on TDL's HP workstation in a continuous real-time mode with a 3-h cycle time. A system has also been implemented to archive all forecast products for all cycles. In December, new software was successfully implemented to generate the QPF products in GRIB format and to process them for transmission to NWS field offices in graphical form through the N-AWIPS communications system. This allows field offices with the capability to receive N-AWIPS products to have access to the QPF graphical products in real time.

MARINE ENVIRONMENTAL PREDICTION PROJECT (W. Shaffer)

Hurricane Storm Surge Forecasting (W. Shaffer): We are beginning to develop a new SLOSH basin for the Ocean Springs, Mississippi, area. This basin will allow better resolution of the areas between Mobile Bay, Alabama, and Lake Pontchartrain, Louisiana. Much of the work for this basin will be done within MapInfo, a Geographic Information System. We have acquired digital elevation data on a 30 m. by 30 m. grid. NOS has digitally scanned U.S.G.S. quadrangle maps of this area for us. The digital elevation data will be blended with data from our existing basins to give us a "first guess" for the SLOSH model elevation field. These data will be visually checked and adjusted by using the raster scanned quadrangle maps.

Mr. Stephen Baig, who recently joined NHC's storm surge group, visited our group to learn some of our modeling techniques. While he was with us, Stephen investigated the MapInfo program and our data entry techniques using this program.

Extratropical Storm Surge Forecasting (W. Shaffer): The operational versions of the extratropical model were moved from NOAA's CRAY-3 to the CRAY-4 computers in late December. We have examined the extratropical surge model's performance for the past 2 months over the Chesapeake Bay, for a period when no significant norleaster storms occurred. Water levels on the upper bay show a pronounced rise with strong southerly winds. The model's surge forecasts for the tide gage at Baltimore correspond favorably with the observed water levels. Timing of the event was excellent, but forecast surge levels were too high. We expect that tuning the AVN forecast winds for the bay will improve the surge forecasts there.

Coastal Wave Forecasting (C.S. Wu): With the assistance of Dr. Robert Jensen at the Corps of Engineers' Coastal Engineering Research Center, we ran the spectral ocean wave model WAM (cycle 4) for hurricanes Felix and Luis (1995). Numerical comparisons of marine winds and waves agreed well with field measurements. we are working with Dr. Jensen to implement the SLOSH hurricane wind formulation as the forcing for the ocean wave model. We have recently acquired the manual for the Delft SWAN shallow water wave module, which is compatible with the WAM model.

We formulated and coded two parametric wave models with input from the SLOSH wind fields. Preliminary tests with these models were conducted for hurricane Felix (1995), with model computations of highest wave heights comparing favorably with observed heights. In addition, the observations of coastal waves and inland high water marks during hurricane Fran's landfall were acquired and analyzed for later model verification.

We participated in planning the SandyDuck field experiment to be held at the Corps of Engineers' facility in Duck, North Carolina, from June to November, 1997. Much of the work we proposed on measuring wave setup Will be conducted by a group from the University of Delaware. Storm wave and sea water level data collected during the hurricane season will be analyzed to allow us to better model wave breaking and runup.

LOCAL TECHNIQUES DEVELOPMENT PROJECT (D. Ruth)

WFO Application Development (D. Ruth): Design, Development, and Testing teams continued to respond to Deficiency igeports (DRS) against hydrometeorological (HM) applications. AWIPS Release 1.1 distributed in early November included improved functionality and data availability for the Grid Processing and NEXRAD Image and Graphic applications. AWIPS "Emergencyll Release 1.1.1 provided to the field in late November eliminated NEXRAD product mapping errors, and added the ingest and display of NEXRAD Build 9.0 products. AWIPS Release 1.OTEF scheduled for release in early January includes Contour setup data fixes, improved functionality of the NEXRAD Image and Graphic application, and improved performance of the Grid Processing applications. AWIPS Release l.GRIDs scheduled for release in late January includes the addition of boundary layers for some derived parameters and the modifications necessary to accommodate changes to the NCEP data stream. Final Test Reports were completed for the Grid Processing, NEXRAD Image and Graphic, and Contouring applications.

Work continued on integrating the Fog Product and ICWF into AWIPS Build 2. These applications are being prepared for field implementation this spring. A draft HM User's Guide for the ICWF was completed in late December.

Development stopped on several other HM applications originally proposed for release in AWIPS Build 2. Instead, our effort has focused on understanding the internals of WFO-Advanced. DDT team members participated in training sessions on WFO-Advanced, Object Oriented Analysis and Design, and C++. Materials are being prepared for an AWIPS Application Integration Framework Manual which will support applications development for Build 3 and beyond.

TDL, OH, OSD, OSO, and GSC personnel continued to work on implementing configuration management software (PCMS). OSD is preparing a Build 3 CM Implementation Plan and a PCMS Control Plan.

Product Generation (M. Peroutka): The ICWF supported significant changes to operations at forecast offices in Boise, Idaho, and Charleston, West Virginia. The Boise office reconfigured their forecast zones, and the Charleston office expanded several of their forecast products to 5 days. Formatters for the Coded Cities Forecast (CCF) and the Revised Digital Forecast (RDF) tabular product were enhanced to include extended forecast periods.

Build 10.0 of the ICWF was delivered to Charleston for beta testing. ICWF Watch/warning/Advisory programs now support multiple W/W/Als in one product, certain non-W/W/A statements, and the generation of "rip and read" scripts for the NOAA Weather Radio. ICWF developers attended several training sessions during this period, including a one-day seminar on WFO-Advanced and its internals, Object oriented Analysis and Design, and C++.

Interactive Techniques Development (D. Ruth): An operational version of the ICWF terminal forecast editor was provided to the forecast office in Charleston, West Virginia, in ICWF Build 10. Aviation objects are created and modified on a time/height plot via a set of pop-up dialogues. New features including LAMP initialization and a TAF text window are currently under development for Build 11.

Several enhancements were made to the ICWF slider bar interface. Most significantly, forecasters can now link threshold adjustments to site-defined regions by use of geographic weights. An interface was developed which enables the local office to define these weights. The grid editor was also enhanced to allow weight functions when incrementing and decrementing continuous fields. For example, forecasters can increase precipitation amounts just on eastern slopes, or decrease temperatures in mountain valleys.