Storm Based Warning Polygons: Guidelines and Recommendations including issues related to Nonlinear Storm Motion Version 1.0 – 12/3/2009 Mr. Ashley H Kells OST32 – Development Branch, System Engineering Center (SEC), Office of Science and Technology Executive Summary This Whitepaper summarizes the recommendations of two independent teams commissioned by the Corporate Board in the area of Short Duration Warnings (such as Tornado Warnings, etc.) and provides a multi-phased approach (short, mid-term and long term actions) to improve WFO services and efficiency in Short Duration Warnings. The first team was a cross-regional team (Polygon Action Team), lead by Pete Browning (CR SSD Chief). Their task was to provide guidance on the area, size and length of warnings to ensure optimum balance between what science and technology allows versus what can efficiently be managed in the WFO and/or assimilated by media, etc., in disseminating warnings to public. The membership of the Polygon Action Team included William Bunting, SR; Claudia Cox, WR; Steve Naglic, ER; Bill Ward, PR; John Dragomir, AR; Kevin Scharfenberg, OCWWS; Russell Schneider, NCEP; and Peter Browning, CR. The team was assisted by John Ferree, OCWWS, Darone Jones, WR and Daryl Herzmann, Iowa State. The second team, lead by Ashley Kells of the Systems Engineering Center/OST, examined the issues related to Non-linear Storm Motion as described by Bill Bunting, MIC of WFO Forth Worth, in a briefing titled: “Performance Assessment: Some Trends, Challenges and Recommendations” dated May 12, 2009. Their mission was to coordinate with development organizations, OCWWS and the field a review of the warning generation process software’s adequacy in multiple storm environments including the assessment of non-linear movement situations, and provide recommendations to improve the forecasting tools. The team included Carl Bullock and Jim Ramer – OAR/GSD; John Ferree and Kevin Woodworth – OCWWS; Eric Howieson – SR; William Bunting – WFO FWD; and Ashley Kells – SEC. The recommendations are stratified into a multi-stage approach. Short term recommendations will be realized by adjusting the training on how the forecasters can include storm information in warnings and services. Mid term goals will be realized with minor format changes to NWS Policy Directives and warning generation software changes to the AWIPS II baseline to assist in standardizing services for the short duration warning environment. Long term goals will be realized with the deployment of the Next Generation Warning Tool including the use of Scientific Algorithms and Short Term Mesoscale Models to develop guidance for use by the warning forecaster and improve warning services. A complete list of recommendations is provided in Appendix A. Issues and Recommendations Issue 1: Training and Warning Proficiency Two training modules have been developed for Storm Based Warnings by the Warning Decision Training Branch (WDTB). Both modules provide excellent guidelines for the issuance of SBW polygons for a variety of warning decision circumstances. The second module (Advanced Storm-Based Warning Training) focuses on providing guidelines for issues identified during the first season of SBW polygon warnings. This module addresses many of the issues identified in this report, including the size of the polygon, length of warning text, CWA boundary issues and the movement/anticipation of the threat area. o Recommendation 1.1 (S): NWS Training Policy should be changed to ensure that each year all forecasters complete both WDTB training modules and Science Officers should include “polygonology” elements in seasonal training using the WES software capabilities. This action should be assigned to OCWWS and implemented by 2nd QTR FY10. o Recommendation 1.2 (S): Each WFO should institute an active post-event evaluation process for identification of operational deficiencies and best practices and then share polygon best practice and deficiency findings among Regions and WFOs. The http://nwschat.weather.gov/vtec web site allows offices to review each warning with storm reports and radar history. This tool can help illustrate how forecasters can improve by examining actual warning situations. This action should be assigned to each WFO/MIC and implemented by 2nd QTR FY10. Issue 2: CWA and Coastline Boundaries By far, the largest issue identified has to do with the issuance of SBW polygons near County Warning Area (CWA) and coastline boundaries. NWS Policy dictates that SBW polygons cannot cross over into another office’s CWA. Policy also dictates that SBW polygons are not allowed to cross over coastlines (extend from land to water to include water areas). Limiting polygons from crossing CWA boundaries is primarily due to the way we issue our warnings. NWS statements and warnings are tied to specific text products through the use of product identifiers and WMO headers. These are unique to each NWS office which enables emergency managers, the media and the public to focus on those specific products for their local area. Over water, the NWS issues a different suite of products to inform the public about severe weather threats. Thus a separate warning process is required to generate warnings when a severe weather threat area crosses either the CWA or a coastline boundary. To adhere to this policy, AWIPS software was designed to prevent polygons from being drawn over water or into the neighboring CWA. The forecaster can issue a reasonable polygon (using 4 to 6 corner points) to define the threat area, but when the warning is generated, the software will “snap” the warning polygon line back to the CWA boundary or the coastline. Since these boundaries may be rather complex, the polygon may now be defined by many more coordinate points as the software tries to match the complex boundary. However, the software has intentionally limited the maximum number of points to 20 in response to user feedback concerning the number of points that can reasonably be included in the text warning. The result is that the software is unable to properly define the complex boundary and produces unintended polygon shapes. Even without the reduction in polygon points requested by NWS customers, the warning shapes in these situations can be complex and potentially difficult to interpret. o Recommendation 2.1 (S): Each WFO should be required to collaborate with neighboring office(s) when severe weather threatens areas adjacent to the CWA boundary. This collaboration will help remove any gaps to the warning service and provide ample lead time to areas immediately adjacent to the CWA boundary. This action should be assigned to the WFO/MIC and implemented by 2nd QTR FY10. OCWWS should ensure that the collaboration is occurring. Recommendation 2.2 (L): During the 2nd QTR FY10, OCWWS will take the action to review the operational concept for polygons that cross CWA boundaries or land/marine boundaries for improvement. If any improvements are recommended NWS Policy and AWIPS software should be changed. An OSIP task should be initiated to define the requirements for an AWIPS release using the SREC process. The date implemented will depend on the SREC prioritization of that requirement. AWIPS II releases begin August 2011. Issue 3: Drawing Polygons o Issue 3.1: County Line influence on warning area County lines can also influence the shape and size of SBW polygons. The NWS warning text product includes the VTEC dissemination coding which identifies the warning using a numbering system and ties these numbers to the UGC county code numbers. VTEC coding in warnings provide a capability for NWS partners to better automate the dissemination of warnings by matching the UGC codes to those that reflect their area of responsibility. The NWS NOAA Weather Radio (NWR) utilizes the UGC codes to tone alert and disseminate warning information. Many communities can only issue tornado warning sirens on a county-wide basis, since these systems have not been designed to warn sectors of the county. As a result, forecasters may often utilize one or more county boundaries to shape their polygons as a way of mitigating dissemination issues that utilize the UGC codes. For example, a forecaster may choose to trim back an area of the polygon that includes a small part of a county – knowing that sirens will sound across the whole county if that small part remained in the warning. Another example is the drawing of the warning area larger to avoid the potential need for issuing another warning polygon within the same county. o Recommendation 3.1.1 (S): Forecasters should be trained to identify the threat area and include those portions of counties that are at risk. This action should be assigned to OCWWS and implemented by 2nd QTR FY10. o Issue 3.2: Proper layout of polygons We know there are limits to our ability to forecast where severe thunderstorms and tornadoes will occur. We issue warnings today based on radar observations and real-time storm reports which are usually validated by radar observations, but forecasters can use diagnostic and forecast data to help anticipate the evolution of a severe weather event. Application of best practices described in available NWS training can also help forecasters in anticipating how the severe threat area will evolve. These nonlinear storm behaviors are difficult to warn for. Even with NWS training and the use of near-storm environment tools, there can still be substantial uncertainty in how the severe weather threat will evolve. In these situations, forecasters may focus the SBW polygon on an existing severe storm based on radar signatures or observations from spotters, but can fail to adequately anticipate the severe weather hazard in adjacent areas. This may result in the need for an additional warning adjacent to or overlapping the first warning as the severe threat moves out of the first warning polygon. o Recommendation 3.2.1 (S): Forecaster should be trained to, despite the difficulty, focus not only on the extrapolation of the current severe weather threat, but also allow for new development or deviant storm motion so that the polygon properly matches the severe threat during the entire warning time frame. This action should be assigned to OCWWS and implemented by 2nd QTR FY10. o Recommendation 3.2.2 (S): Forecasters should be trained to include current warning polygon outlines for their own and neighboring CWAs on their warning generation display screen. This will help forecasters keep ahead of the severe weather events, issuing warnings with appropriate lead time and ensuring that no gaps occur between warning polygons. This action should be assigned to OCWWS and implemented by 2nd QTR FY10. o Recommendation 3.2.3 (S/M/L): As a short term goal, forecasters should be trained to add a fifth vertex (elongated pentagon) to the polygon if time and circumstances allow. This action should be assigned to OCWWS and implemented in conjunction with the software change. Forecaster should be trained to follow the following process when issuing a warning. Starting with an elongated pentagon, consider three different motion vectors of the severe threat. One vector is to the left of the storm motion, one along the current storm motion and one to the right of the storm motion. Then answer the question - what is the likelihood that the severe threat will move along these three vectors? This action should be assigned to OCWWS and implemented by 2nd QTR FY10. As a mid term goal, AWIPS software should be changed to use a new confidence or probability approach for forecasters to consider when drawing SBW polygons. The team proposes the use of a 5 point polygon (elongated pentagon) as a starting point for defining the warning area. The Elongated Pentagon provides 5 vertices which allow the warning area to be better manipulated to reflect the likelihood of the persistence and direction of the severe threat. An OSIP task should be initiated with a SON to define the requirements and assign to an AWIPS II release using the SREC process. The date implemented will depend on the SREC prioritization of the requirement. AWIPS II releases begin August 2011. As a longer term goal, AWIPS software should be changed to automatically use this elongated polygon. Also, AWIPS software should be changed to provide the forecaster with vectors that are derived from the local environment (use Bunkers Method or Corfidi Vectors as appropriate) and/or climatology of storm motions. The forecaster looks at these vectors and the recent behavior of the storms to quickly assess the likelihood values and adjust the pentagon accordingly. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. Issue 4: Inclusion or Omission of small portions of counties When forecasters draw polygons, the boundary of the polygon may clip a small portion of a county and be included in the warning text product. For dissemination systems that rely on the county UGC code, this will trigger alerts across that county. Often, the small portion is sufficiently small such that it can be omitted from the warning and not affect the warning service (when probability of severe weather occurring in that small area is unlikely). To account for this, WARNGEN has been designed to identify small portions of counties using 2 methods - a percentage of the county area or a specific minimum number of square miles of county area that is set in the warning template. WARNGEN will automatically exclude any area within a county that is less than the configured percentage. Offices typically choose 1, 2 or 3 percent of county area values for warnings issued by their office. For example, if an office has configured WARNGEN for 3 percent, then any county that has area within the polygon drawn by the forecaster that is less than 3 percent of the total size of that county will be automatically removed during warning generation time. Alternatively, the warning template can be configured to use a fixed minimum value instead of the percentage method. WARNGEN will then measure the size of the area in each county that the warning polygon includes and compare it to the minimum square mileage parameter. If the size is less than the value in the template, then the software will automatically exclude all area from that county. The value provided in the tornado warning template on the Central Region WARNGEN template best practice repository is 50 square miles. This value may be too large or too small depending on the typical size of counties in the office’s county warning area. Recommendation 4.1 (S): Each WFO should check WARNGEN configuration and ensure that the minimum warning area parameter is set sufficiently small to only remove counties when the warning inadvertently touches a county. The average size of a tornado warning polygon is less than 340 sq. miles. The average size of a county in the U.S. is 1126 sq. miles. If the percentage method is used and 3% is selected, then up to approximately 10% of the average tornado warning size is automatically excluded. In the Southeast U.S., many counties are much smaller than the average county size. This action should be assigned to the WFO/MIC and implemented by 2nd QTR FY10. Recommendation 4.2 (L): Warning software should be modified to provide notice to the forecaster when adjustments will be made to the polygon, and allowing the forecaster to confirm or refuse these changes. Today, forecasters do not know what the polygon will look like unless they remember to click the “Redraw box from warned/hatched area” button in WARNGEN. If the software makes automatic changes, then the forecaster may need to waste valuable time in redoing the warning area to override these changes. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. Issue 5: Warning Text Dissemination o Issue 5.1: Warning Size The size of a warning polygon can create problems with the dissemination of the text warning product. When WARNGEN creates the text product, it includes county names (including the direction portion of the county), state names (including the direction portion of the state), cities affected and cities in the path (if the pathcast is included) of the severe weather. For large polygons which typically include portions of many counties, the length of the warning text product becomes a dissemination problem as the delivery of the warning information is delayed, or in some cases not delivered at all due to a truncation of the message. o Recommendation 5.1 (S): Forecasters should be trained that large threat areas are best handled with multiple warnings to mitigate dissemination issues. WDTB training modules specify no more than 12 counties/parishes in a warning. Ideally it should be less than twelve to effectively communicate the location of the threat. The forecaster should focus on making the individual warnings be roughly the same scale as the phenomena being warned for. Tornadoes are small scale phenomena, so tornado warnings should be on the storm cell scale, tracking the expected path of the tornado and allowing for appropriate deviant motion of rotating thunderstorms and cyclical evolution. Severe thunderstorm events may also be on the storm cell scale for large hail or multiple times larger when considering high wind events like derechos. It is always important to review warning polygon size and configuration within the context of the event. This action should be assigned to OCWWS and implemented by 2nd QTR FY10. o Issue 5.2: Wording for Portions of Counties The wording and format of the counties included in the warning is very repetitious. The format lists each county on a separate line using the following format: Portion of county, county, in portion of state, state (e.g., “EXTREME SOUTHWEST JACKSON COUNTY IN NORTHWEST MISSOURI”). Given a polygon of a reasonable size that touches portions of 4 or 5 counties in the same part of the state the software created warning text will list numerous compass directions with excessive replication of the portion of the affected State. The use of compass directions to identify the portion of the county can be misleading to the public. In some cases, the portion of the county identified may cause people in the warning to believe the warning does not include them. When listing the counties affected, WARNGEN can also provide wording to list a prominent city that lies within the county (i.e., “including the city of XYZ” immediately following the county name where this city is located). This capability is configurable in the warning template and will add length to the warning text. Since cities affected are described in the pathcast and cities included sections of the warning, it may become redundant to include these cities here. o Recommendation 5.2.1 (S): AWIPS software should be changed to remove the use of the “including the city of …” from the WARNGEN template to better streamline the text product. The city can appear later in the warning in the cities affected and/or pathcast sections. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. o Recommendation 5.2.2 (L): WARNGEN templates can be configured to minimize the repetitious nature of listing multiple counties, however, when these changes are made, quality control errors occur when the warning is generated. The generation of text warning product needs a redesign to provide sectioned information in a more conversational style. The warning text product today IS the only official warning vehicle of the NWS and attempts to provide both a readable text and serve computer decoding software. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. o Issue 5.3: Too many Warnings Partners have complained that they are receiving too many warnings since we have moved to SBW polygons. This perception does not match reality, as the overall number of warnings has not increased dramatically. What may be feeding this perception is that we now can (and do) issue multiple warnings for the same county, which are in effect at the same time (for different parts of the county), based on different severe thunderstorms or storm clusters. If your dissemination system relies on the county coding, then a new warning for the same county which is already under a warning may seem redundant or excessive. However, if the warnings are graphically displayed, you can more easily see the threat areas as independent warnings. Overlapping warnings can cause similar confusion for the public. Policy dictates that an existing warning area can only be reduced in size as the threat diminishes. A warning area cannot be moved or expanded into a new area. The forecaster can issue a new warning to spread the severe threat downstream. This new warning may overlap the existing warning, resulting in having multiple warnings in effect for the same area at the same time. There is no means for replacing a warning with a new warning that redefines the area. The forecaster can issue a new warning and cancel or let the previous warning expire. However, cancelation or expiration of a warning can lead to confusion for customers who rely on county based warning dissemination. The public can be misled when they hear that the original warning has been cancelled or has expired when in fact, a warning remains in effect. o Recommendation 5.3.1 (S): Forecasters should be trained to minimize multiple overlapping polygons if possible, and stay ahead of the threat by issuing timely warnings downstream. They should not wait for the current polygon to nearly expire before issuing the next polygon downstream. They should use updated environmental parameters to assist in determining if the threat will continue downstream or weaken. This action should be assigned to OCWWS and implemented by 2nd QTR FY10. o Recommendation 5.3.2 (S): Forecasters should be trained to issue frequent SVS’s to refine polygon boundaries, update the pathcast, and include reports. When more than one warning is in effect for the same county, or a new warning is overlapping a previous warning, it is often better to let the warnings expire instead of canceling one warning while another is still in effect in the same county. WCMs can recommend to emergency managers that they monitor severe weather warnings graphically on radar displays to best understand the threat and avoid confusion when multiple warnings affect their county. This action should be assigned to OCWWS and implemented by 2nd QTR FY10. o Recommendation 5.3.3 (L): AWIPS software should change to use VTEC to handle instances when one warning expires but another warning remains in effect for the same general portion of a county or parish. This will allow better management of polygons without the confusion surrounding cancelling/expiration of one warning in a county while another is still in effect. A new VTEC code can be instituted to allow a forecaster to replace a warning with an overlapping polygon. This would allow the warning to progress seamlessly forward as the threat area evolves. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. o Recommendation 5.3.4 (L): Develop the capability to include latitude-longitude information in Weather Radio digital tone bursts to encourage the development of new technology to display polygon information with NWR and EAS. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. Issue 6: Pathcast o Issue 6.1: Identifying Cities/Towns in Path and Warning Area During the warning generation, forecasters can choose to include a pathcast which identifies approximate times of arrival of the severe weather threat at cities and towns based on the identified storm path. Several configurable parameters control how the software identifies and displays towns or cities in the path of the storm. The city or town needs to be within a configurable set distance from the identified storm path (i.e., 5 miles). Also, the maximum number of cities (15), time groups (8) and number in groups (5) in the pathcast can be configured. These parameters will affect warning length and/or the use of the “over mainly rural areas” wording if no towns or cities meet the thresholds. If the forecaster chooses to just list cities and towns in the warning area, a maximum number of these locations can also be configured (20). (Values in parentheses reflect those in the tornado warning template on the Central Region WARNGEN template best practice repository.) The cities used for the storm location and pathcast refer to cities or towns identified as having priority 1 or 2 in the wwa_warn_city file. Offices need to review this file and make appropriate adjustments to ensure proper city references are used. Recommendation 6.1.1 (S): Each WFO should review local AWIPS configuration files and adjust the city/town priorities based on geopolitical and demographic needs (the tier 1 and tier 2 priorities) for appropriate use by WARNGEN templates. They should update WARNGEN templates using the optimized versions from the WARNGEN template repository and test these configurations using the WES. This action should be assigned to the WFO/MIC and implemented by 2nd QTR FY10. o Issue 6.2: Storm Motion The way that WARNGEN identifies storm movement can also be problematic. The software uses a linear approach to determine the speed of movement (straight line between 2 points) and to then project into the future. This provides a first guess polygon for the forecaster based on this movement and warning length. There have been a number of cases where rotating supercells and/or new development on the storms flank have resulted in the severe threat deviating from the indicated pathcast and even moving outside the polygon. The resultant pathcast is not accurate and may not properly include arrival times for cities in the actual storm path. WARNGEN provides no alternative storm motion vector which can provide guidance to forecasters on expected deviant motion to account for right (left) movement or propagation of the severe threat. o Recommendation 6.2.1 (L): AWIPS software should be developed to better anticipate nonlinear storm processes on the warning time scale (~45 minutes), and communicate these findings to forecasters in a timely fashion. This research can lead to better storm motion tools to improve pathcasts and polygon placement. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. o Recommendation 6.2.2 (M): AWIPS software should be changed to allow forecasters to edit the storm motion vector to better identify cities in the path. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. Issue 7: Issues related to Line of Storms o Issue 7.1: Isolated storms developing very close to and/or merging with a severe line of thunderstorms. o Recommendation 7.1 (S/M/L): OCWWS will provide training in the short term to increase the awareness of the forecasters to add other storm information. OCWWS will also update their Directives in the mid-term to provide further guidance for inclusion of other severe storms in our products and services. In the longer term, requirements will be added to the OSIP Next Generation Warning Tool project to allow its Threat Definition Tool (TDT) the ability to allow for multiple storm/line tracks in a single threat area (polygon). This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. o Issue 7.2: Segments of a line that have different storm motion (i.e., direction and speed). o Recommendation 7.2 (L): AWIPS software should be developed to handle different segments of the storm line have different storm motion, using the TDT. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. Issue 8: Multiple severe storms or near severe storms developing close together. Recommendation 8: Forecasters should be trained in the short term to increase the awareness of the forecasters to add other storm information. This action should be assigned to OCWWS and implemented by 2nd QTR FY10. Policy Directives should be updated in the mid-term to provide further guidance for inclusion of other severe storms in our products and services. This action should be assigned to OCWWS and implemented by 4th QTR FY10. In the longer term, requirements will need to be added to the OSIP Next Generation Warning Tool project to allow its Threat Definition Tool (TDT) the ability to allow for multiple storm tracks in a single threat area (polygon). This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. Issue 9: Cells with Deviant Motion (Right or Left Turning Supercells) Recommendation 9 (M/L): AWIPS software should be developed to allow the warning forecaster to manipulate the storm threat area for storms with deviant motion, using the TDT. GSD has worked on proto-type in WarnGen. This should be assigned to the OSIP NGWT IWT team and included in the Next Generation Warning Tool deployment. Issue 10: Provide non-linear storm motion guidance Scientific Algorithms and Short Term Mesoscale Models could be developed to provide non-linear storm motion guidance to the warning forecaster in the Next Generation Warning Tool. Recommendation 10 (L): The longer term goal would be to provide non-linear storm motion guidance to the forecasters. An OSIP task should be initiated to investigate and incorporate non-linear storm motion guidance into the warning generation process and assign to an AWIPS II release using the SREC process. This task will need a Research and Development Project and into the “Warn on Forecast” Concept. The date implemented will depend on the SREC prioritization of the requirement. AWIPS II releases begin August 2011. This should be assigned to the OCWWS and implemented by 2nd QTR FY10. Issue 11: Additional Research and Development Required o Recommendation (L): The development of the Next Generation Warning Tool should utilize the NOAA Hazardous Weather Testbed to engage social scientists, human factors and ergonomics experts, and key national partners/stakeholders to identify ways to make long-term improvements to storm-based warnings. This should be assigned to the OSIP NGWT IWT team and implemented by 2nd QTR FY10. o Recommendation (L): Deploy AWIPS software to the NOAA Hazardous Weather Testbed to “game” polygon best practices during live events and provide a forum for sharing results to the operational and development communities. This action should be assigned to the OST/SEC and implemented by 2nd QTR FY10. o Recommendation (L): The NWS should prominently display graphical and KML versions of active warnings on NWS web pages. This can be done using a GIS-enabled version of the WWA map which provides the radar data as a layer with warning polygons and reports. This action should be assigned to the OCIO and implemented by 2nd OTR FY10.