P1.11 HIGH IMPACT GRIDDED WEATHER FORECASTS Steve Amburn*, Steve Piltz, Brad McGavock, and J. M. Frederick

NOAA/NWS, Tulsa, OK

1. INTRODUCTION

Forecasting high impact weather presentssignificant issues for the National WeatherService (NWS). There is no doubt theseforecasts can be prepared and issued, but atwhat cost to NWS labor resources? In order toserve all customers, the NWS needs to becomeincreasingly efficient in its production of thevarious high impact weather forecasts requiredby its customers. For this reason, the WeatherForecast Office in Tulsa, OK, (WFO TSA) hasadopted a philosophy of forecasting the basicweather parameters (with a few additions) andallowing internal software to compute andproduce the high impact forecasts and forecastgraphics, as well as detailed observed fields. Forecasters at WFO TSA produce theirforecasts, as do all other NWS WFOs, using theGridded Forecast Editor or GFE (GlobalSystems Division, 2006). Within GFE, WFOTSA has added numerous components andtools to create additional gridded forecasts andobserved fields which can provide importantdecision-making information to the usercommunity. High impact elements such asprobability grids for severe thunderstorm andtornado threats have proven to be very valuableto users. Other hazard grids such as maximumwind gust, maximum heat index, flood and firespread potential, allow users to easily locateareas of concern in the seven-day forecast.

2. GRIDDED FORECAST PHILOSOPHY

With the advent of gridded forecastdatabases, the mission of the NWS has addedsignificant responsibilities at the WFO level.Inthe early 1990s, the public forecast function waseasily completed by grouping a handful ofcounties into a zone, then typing a three or fourperiod forecast for those four, five, or maybe asmany as eight zone groups. The area forecast,which replaced the “state forecast,” was also a --------------------------------------------------------------Corresponding author address: Steven A.Amburn, National Weather Service, 10159 E.11th Street, Suite 300, Tulsa, OK 74128; Email: Steve.Amburn@noaa.gov

relatively short product that took little time totype. However, the desire to provide more detailedforecasts for longer time frames has resulted insignificant changes at WFOs. To accomplishthese changes, the NWS employed griddedforecast fields and software to produce computerworded forecasts. This change in operationshas provided the NWS with new methods andopportunities to provide an unprecedented levelof service to our user community. This changein operations and the new technology hasallowed the local forecast office to address asmany customer needs and concerns aspossible. The only way to do this is to useautomation, where practical, in combination with“current” grid-based analyses and forecasts ofthe set of basic weather parameters. At WFO TSA, the primary use and purposeof the GFE is to provide this set of analyses andforecasts of meteorological parameters asaccurately as possible. From this set,parameters can be combined in numerous waysto produce a wide variety of high impact griddedanalyses and forecasts. These can then beprovided to the user community on a broadscale through the internet and through theHazardous Weather Outlook (HWO) textproduct.

3. ANALYSIS AND SHORT RANGEFORECAST GRIDS

Analysis grids play an important role inkeeping weather forecasts current and also inkeeping forecasters situationally aware. At WFOTSA, the basic set of weather parameters areupdated each hour, and include T, Td, Wind,Wind Gust, and Sky (cloud cover). From thosebasic parameters, other fields are updated,including Max T, Min T, RH, Heat Index, andWind Chill. PoP (probability of precipitation) andWeather grids are updated at the discretion ofthe forecaster as required. All these parametersare then blended into the existing forecastthrough an interpolation technique. Max T andMin T are either adjusted manually or areallowed to adjust automatically as necessary toensure the most accurate forecast possible.

This interactive process within GFE helpsforecasters identify significant changes in theatmosphere that may have a high or significantimpact on customers. The surface analysis grids are also combinedwith the latest model forecasts aloft to providethe latest and most accurate volumetric analysisof the atmosphere each hour (McGavock, 2004). A forecaster may choose to use data aloft fromeither the RUC or NAM models in combinationwith the latest hour’s surface fields. Theinternal parameters of choice for WFO TSAgenerally tend toward convection or thepreviously mentioned parameters, including thefollowing grid fields:

Storm Motion CAPE Cin (local cap index) Lifted Index Lifted Condensation Level Level of Free Convection 0-6 km shear 0-1 km shear 0-3 km helicity 0-1 km helicity Moisture Convergence Energy Helicity Index Max forecast hail size (local index) Max wind gust potential (local index) CAPE to 500mb CAPE to 700mb Windex WIIB (local summary index) Conditional probability of Severe Weather

(“Local” grids are primarily generated based onlocal studies and may not apply to other parts ofthe CONUS.)

In addition to basic analysis grids,forecasters can provide special analyses in amore interactive process within GFE. This willgenerally occur when some portion of the countywarning and forecast area (CWFA) hassignificantly different values than even astandard mesoscale analysis can provide. Forecasters can use GFE tools to makeassessments of how the environment will affectindividual storms. Storm scale analyses forhelicity are prime examples. A forecaster canadjust the storm motion for a cell with a deviantmotion (Figure 1) to see how that may affect thehelicity for that thunderstorm. After running the

GFE tool, a newly calculated helicity grid isavailable (Figure 2). In this particular case, theforecaster’s situational awareness is improved,allowing him to anticipate mesocyclone or eventornadic development from the thunderstorm.

Figure 3 shows a comparison of the GFEmesoscale analysis and the Storm PredictionCenter (SPC) RUC analysis for the same area. The two images on the left show zero to 1kmhelicity output from the WFO TSA GFE tool, onehour apart. The image on the right is the SPC

Figure 1. Manually updated storm motion field.

Figure 2. Updated helicity field from GFE tool.

analysis. The hourly updated surface grids andthe resolution of GFE paint a considerablydifferent picture than the more synoptic scaleanalysis to the right. Note the rather largepositive values in the SPC analysis while theGFE values showed a transition from negative topositive values from west to east. In fact, asupercell thunderstorm moving east across thearea exhibited an anticylonic circulation on itsnorth side, then transitioned to a cycloniccirculation on the south side as it moved to theeast side of the boxed area. The detail in GFEhelped warning forecasters understand thestorm evolution and helped forecasters positionspotters around the storm. Another example where local GFEmesoscale analyses are helpful are the Cin (Capindex, Figure 3), which is also a WFO TSAcreated tool for GFE. This tool is very helpful in identifying the potential for convective

Figure 3. Comparison of WFO GFE analyzed helicity and SPC mesoscale helicity.

Figure 4. WFO detailed Cap index.

development. These grids also seem to providebetter information than a generic RUC analysis. Strong capping is shown in red, while blue areas

indicate very little cap. Another helpful tool used in storm-scaleanalyses is the GFE hodograph tool (Figure 5),

Figure 5. Locally developed GFE point hodograph tool output, valid for near storm analysis.

developed at WFO TSA. This tool allowsforecasters to generate hodographs “on-the-fly”in proximity to specific thunderstorms. Withhourly updated surface grids and manually orautomated storm motions, this hodographoutput provides warning forecasters with theinformation necessary to anticipatethunderstorm evolution.

4. SHORT RANGE FORECAST GRIDS/WEBIMAGES

However, hourly updated parameters are alsoused to generate products, mostly graphical, formore specific weather related risks. In additionto convective threats, other risk analyses aredisplayed on the WFO TSA web page forhazardous weather. Specific analysis andforecasts on the hazardous weather page showrisk analysis information for tornadoes, severethunderstorms, flash flood, heavy rain, lightning,dense fog, strong winds, fire danger and heatindex or wind chill. These are all produced fromthe GFE basic set of forecast grids. However,elements such as the conditional probability ofsevere thunderstorms can be manuallyoverridden by the forecaster. The graphics areautomatically generated, allowing forecasters tomonitor and forecast the weather, and not beconsumed in graphics generation. The risk analysis graphics are generatedeither directly from the forecast grids, or bycombining grids and algorithms. Work is beingtested to use GFE grids to create warningprobabilities for the WFO TSA web page(McGavock, 2006). The current list of risk fieldsin the WFO TSA hazardous weather DecisionSupport Page are discussed below and providedecision makers with an easily viewable set ofsignificant weather parameters for a variety ofpurposes.

4.1 Tornado

The Tornado threat level (Figure 6) is basedon the probability of being in a Tornado Warningin an area. The probability of a TornadoWarning in an area is forecast based on:• The Probability of Thunderstorms (local

grid derived from the Weather and PoPgrid)

• The Probability of a Thunderstormbecoming Severe once it has developed(local grid of Conditional Probability)

• The Probability of a Thunderstormbecoming Tornadic once it has becomeSevere (local grid of ConditionalProbability)

A “local” grid is one which has been created atWFO TSA for the WFO TSA GFE and is notpart of the national GFE basic set of grids.

4.2 Severe Thunderstorm

The severe thunderstorm threat level (similarto Figure 6) is the probability of being in asevere thunderstorm warning sometime duringthe remaining portion of the hazard outlook. The probability of a Severe ThunderstormWarning in an area is forecast based on:• The Probability of Thunderstorms (local

grid)• The Probability of a Thunderstorm

becoming Severe once it has developed(local grid of Conditional Probability)

4.3 Flash Flood

The flash flood threat analysis is acombination of parameters consistent with thefollowing bulleted list. The final grid is based onthe probability that the flash flood guidance willbe exceeded in a three hour period. Theprobability of exceedance (POE) is computedfrom the QPF (the mean of the expecteddistribution of rainfall amounts) and theprobability density function of the exponential

Figure 6. The 24-hour probability of being in atornado warning.

equation. Details on the calculation of POE canbe found in Amburn (2006). • The Flash Flood threat level is based on

Flash Flood guidance from the RiverForecast Centers and the WFO QPF.

• The QPF is based on expectedcoverage, intensity and duration ofThunderstorms.

4.4 Heavy Rain

The threat for heavy rain is measured as theprobability to exceed 1 inch of rainfall in a six-hour period (Figure 9). Again, this is based onthe probability of exceedance computed fromthe QPF (the mean of the expected distributionof rainfall amounts) and the probability densityfunction of the exponential equation.

Figure 7. Heavy rain potential graphic, shownas the 24-hour QPF total.

4.5 Lightning

The lightning threat level is based strictlyon the probability of thunderstorms in an area. The probability of rain (GFE PoP grid) andprobability of thunderstorms (GFE Weather grid)are used to derive this high impact element inGFE. More detailed lightning information forongoing thunderstorms is also available at theWFO TSA web site through GFE analysis tools.

4.6 Dense Fog

The fog threat level is based on the forecastof fog (basic GFE Weather grid) and occurrenceas predicted by the forecaster. The graphicdepicts the weather type and time of beginningand ending of event. This graphic is “shipped”directly from GFE to the web page to providegraphical information to the users. Again, theforecasters only need to maintain accurate grids.

4.7 Strong Winds

This grid is the maximum wind speed anddirection obtained from the hourly forecast windgrids or wind gust grids. These are also basicGFE grid fields available from any WFO (Figure12). When these are highlighted in the DecisionSupport page, it provides a reminder to theforecasters to look for meeting possible watch orwarning criteria.

4.8 Fire Danger

Deteriorating fire weather conditions meanthat fires of any origin would have an increasingpotential to spread (Figure 13). A fire “spreadindex” is used to assess the threat of firesburning out of control. This spread index uses a

Figure 8. Max wind forecast for 24-hour period

from sustained winds and gusts.

combination of temperature, humidity, wind, andfine fuel state to calculate its value. All theseinputs are basic GFE grids except fine fuel statewhich has been added to the WFO TSA GFEdatabase for creation of this fire danger index. Information seen in this graphic is mostappropriately used for grass fire potential as theindex is based on fuels that can dry within onehour (grasses). The fuel state is a seasonalinput that is based on whether the grasses inthe area are green, dormant and dry, or somestate of transition in between. (Pharo, et. al.)

4.9 Heat Index / Wind Chill

The Heat Index or Wind Chill Index graphicsare also shown on the Decision Support page.These are the extreme values from each hourlygrid in the forecast area throughout the validtime of the forecast. These can be very useful toschools, coaches and other decision makerswho are responsible for the health and welfare ofpeople who may be working or playing outdoors.

4.10 Convective Watch Enhancement

The GFE convective analyses, tornado andsevere thunderstorm probabilities are also beingused to provide enhancements to SPC

convective watches. This is a step towardprobabilistic warnings. On a given day, the SPCoutlook may indicate an increased risk for severeconvection. When the watch is issued, that riskor probability has increased by some increment. At WFO TSA, mesoscale analyses through GFEare then used to indicate which areas within thewatch are more favorable for severe convectionthan others. Figure 16 shows the forecaster’s GFE gridfor tornado probability. Figure 17 shows atornado watch covering much of the CWFA. However, within the SPC watch, the WFO TSAforecaster was able to identify and effectivelydownscale the watch to show areas which wouldhave a higher probability of tornadoes. Redareas had the highest likelihood of tornadoesaccording to the GFE and forecaster analyses. This kind of refinement clearly provided an extralayer of information to the customers.

Figure 9. Wild fire spread index, available out 7days.

Figure 10. WFO TSA tornado warningprobabilities from GFE grids.

Figure 11. SPC tornado watch with WFO enhanced threat probabilities.

5. LONG RANGE FORECAST GRIDS

Longer range forecast grids also provide riskassessment graphics on the WFO TSAHazardous Weather Outlook (HWO) web pageand text product. Again, forecasters’ efforts areon providing the best possible set of basicforecast grids. From those, forecasts of risk forselect weather phenomena are produced withinGFE for use in the HWO and on the web. Thereis no duplication of effort to draw the hazards. Neither is there any risk of having forecasts andhazards that do not match. For severe weather risks, upper air featuresfrom the model of choice are combined with theforecasters’ set of basic surface grids togenerate forecast graphics of risk for thehazardous weather outlook for days two throughseven. A set of parameters are checkedautomatically in GFE which determines thethreat level and therefore identifies the possiblehigh impact weather for our customers. Similarly, QPFs are summed for each day ofthe hazardous weather outlook to derive thepotential of heavy rain. Total seven-day rainfalland POEs are generated from the basic gridsand sent to the web page for use in the HWOtext to notify customers of possible high impactweather situations (Figure 18). Regarding fire weather, the wild fire spreadindex (shown earlier) is computed out throughday seven to provide concerned persons withinformation about possible weather impacts. Again, within GFE a combination of basicforecast elements of wind, humidity, temperatureand vegetation state are used to generate theforecasts. Other long range high impact forecastsinclude lightning, severe thunderstorm, fog, firedanger, strong winds, heavy rain, and heatindex or wind chill. These are all derived fromthe basic set of forecast grids.

6. INTERNAL USES

High impact analyses and forecasts are usedinternally at WFO TSA to maintain forecastersituational awareness and also to assist in theforecast process. The tools available in GFEcan generate grid fields to show forecasterswhen combinations of forecast parameters arenearing or passing critical thresholds for suchrisks as heavy rain, severe thunderstorms, firedanger and heat index. These tools can quickly

and efficiently scan the entire grid database todetermine if watches, long-fuse warnings oradvisories should be issued. These tools alsoensure forecast integrity. Examples mentioned earlier include severethunderstorm and fire weather grid fields. Otherexamples include checks for high wind events,heat index and wind chill thresholds, and RedFlag criteria. Tools developed within GFE arerun and provide forecasters with information onwhen to issue or consider issuing watches forthese high impact events.

7. HIGH IMPACT EXAMPLES

WFO TSA has had numerous occasions toprovide high impact weather forecasts andanalyses to its customers by making efficientuse of the basic set of GFE grid elements. Emergency managers, fire weather respondentsand others can readily obtain weatherinformation that can be critical to their particularsituation. Also, by maintaining accurate analysisand forecast grids, specialty products can becreated “on-the-fly” to help provide briefings forweather-critical events. This has been done inthe past for large wildfire events, heavy rainpotential from the remnants of tropical systems,and numerous high probability severe weatherdays. In fact, a well maintained GFE griddatabase can provide an abundant set ofweather information to the customer. This isbeing tested at WFO TSA now in the form of theOutdoor Hazard Monitoring and ResponseSystem (OHMARS).

7.1. Special Forecast for Fire Weather

In November 2005, conditions were warm,dry and windy. This followed a period of lowrainfall and therefore resulted in extreme wildfiredanger across much of Oklahoma. The state ofOklahoma required on-site briefings at theirheadquarters offices. Although this took ourWarning Coordination Meteorologist out of theoffice, the high impact gridded analyses andforecast were automatically available on theWFO TSA web site and provided the neededinformation for briefing state officials. Inaddition, the fire spread index forecasts areavailable from our web site for the entire seven-day forecast period, allowing fire officials to starttaking actions several days in advance.

7.2. Special Forecast for Heavy Rain

In the late summer of 2005, remnants ofHurricane Rita were forecast into central andeastern Oklahoma. Although winds were notexpected to be a factor, 24-hour rainfallforecasts were expected to be over seven inchesin some areas. Emergency managers wereobviously concerned and needed frequentbriefings on the potential. GFE grids allowedthe continued production of rainfall exceedanceprobabilities (Amburn, 2006) for briefingpurposes, instead of just the average QPF. Standard exceedance probabilities(~probabilistic QPFs) are available for amountsof 0.10, 0.50, 1.00 and 2.00 inches. However,for very heavy rainfall forecasts typicallyassociated with land falling or decaying tropicalsystems, rainfall POEs can be computed simplyby changing a parameter in a GFE tool, thensending that image to the web to briefemergency managers and other officials. Figures 19 through 21 show an example of PoP,QPF and POEs for a decaying tropical event inthe forecast area.

Figure 12. Probability of rain for decayingtropical system.

Figure 13. QPF for decaying tropical system.

Figure 14. Probability to exceed (POE) 2inches of rainfall, corresponding to figures 19and 20.

7.3. OHMARS

The National Weather Service (NWS) inTulsa is working with local emergency managersto develop a system named OHMARS, theOutdoor Hazard Monitoring And ResponseSystem. This is an application available on theweb, generated fromGFE and other data atWFO TSA. The concept of OHMARS iscurrently under development and a workingprototype has been in existence only a shorttime. The goal of the project is to tieall available NWS information together into oneinterface and support emergency managementofficials in dealing with weather threats atemergency response locations, e.g. hazardousresponse sites, and large outdoor public venues.At such locations, the response and evacuationtimes are often much larger than the lead time ofan official NWS warning. Therefore decisionsmust come before the warning, and with lessdefinite information. By tying all availableinformation into a single interface, the decisionsupport to emergency mangers is more efficient,and a more complete analysis of a situation canbe accomplished. In OHMARS, the emergency official sets thethresholds and actual response information inthe system.  At the heart of the system is theuse of gridded forecasts from WFO Tulsa oftornadoes, severe thunderstorms, lightning, heat

index, wind gusts, rainfall, and sky cover.  NWSdata are fed into the interface, and alarms inOHMARS are activated when thresholds arecrossed.  This results in bulleted text from theemergency manager’s response plan to bedisplayed.  OHMARS monitors observed data,digital forecast data, and NWS text products totrigger alarms.  Using an example of a tornado threat, theemergency manager is alerted to tornadopotential for the public venue's location first thingin the morning by OHMARS after processinginformation from the nearest data point from thegridded tornado potential forecast issued byWFO Tulsa.  As confidence grows in the threat,an update to the tornado potential is made andOHMARS, accessing the gridded forecast,updates and alerts for the change in potential. A Tornado Watch issued by the NWS StormPrediction Center for the venue will alarm in thesystem.  As upstream thunderstormdevelopment is monitored and its evolutionforecast, the tornado potential for the venueincreases beyond the threshold, and OHMARSdisplays the emergency manager's own wordsthat call for the evacuation of the venue, before aTornado Warning is ever issued. Additionally, OHMARS displays relatedgraphics (some may be hand-drawn), along withcurrent conditions, links to text products, hourlyforecast information, and an audio briefing.  Itmay be possible to include text messaging tocell phones from OHMARS when alarmthresholds have been exceeded. WhileOHMARS is currently a demonstration project, itis expected that various entities having accessto NWS data sets will create applications withenhanced capabilities tailored for specific needs.

8. Summary

The National Weather Service WeatherForecast Office in Tulsa, Oklahoma, routinelyuses the Gridded Forecast Editor to computeand otherwise produce a variety of high impactgridded forecast products for distribution to stateand local decision makers. Emergencymanagers, city officials and the general publiccan obtain the information from the WFO TSAweb site hazardous weather page and apply thatinformation to their specific problem of the day. The simplicity of providing high impactweather analyses and forecasts through thegridded forecast database is that forecasters

Figure 15. Probability to exceed (POE) 5 inches

of rainfall, corresponding to figures 19 and 20.

need only maintain the “basic” GFE grids ofmeteorological parameters. Tools available inGFE help forecasters accomplish that, both forobserved and forecast grid fields. This makesforecast and graphics production very efficient.Algorithms access those grids throughout theseven-day forecast to produce specific highimpact weather forecasts and graphics withminimal additional input required of theforecaster. These high impact griddedobservational and forecast fields also serve toenhance forecasters’ situational awareness ofrapidly changing weather conditions, both on theshort range mesoscale and also in the longrange synoptic scale that may negatively impactour customers.

9. REFERENCES

Amburn, S. and J Frederick, 2006: AnExperiment in Probabilistic QuantitativePrecipitation Forecasting. Extended abstracts,18th Conference on Probability and Statistics inthe Atmospheric Sciences, Atlanta, GA, Amer.Meteor. Soc.

Global Systems Division, 2006: InteractiveForecast Preparation System (IFPS), GFE SuiteTechnical Review,http://www.fsl.noaa.gov/projects/development.html.

McGavock, J.B., G. Mathews, J. Frederick,2006: Utilizing experimental graphical severeweather warning probabilities to supplement theHazardous Weather Outlook. Extendedabstracts, Symposium on the Challenges ofSevere Convective Storms, Atlanta, GA, Amer.Meteor. Soc.

McGavock, J.B., S. Piltz, and J. Frederick.2004. Interactive mesoscale objective analysis intheNational Weather Service’s Graphical ForecastEditor, Preprints, 22nd Conference on SevereLocal Storms. Amer. Meteor. Soc.

Pharo, J.A., L.G. Lavdas, P.M. Bailey, dateunknown: Southern Forestry SmokeManagement Guidebook, Southern Forest FireLaboratory, Southeastern Forest ExperimentStation, USDA Forest Service, Macon, GA.