16 th Weather Squadron Fly - Fight - Win Air Force Weather Ensembles Scott Rentschler Fine Scale and Ensemble Models 16WS/WXN
Background Air Force Weather Decision Support: Weather impacts on specific missions Bombs on target! Mission planners consult humans or human made products Mitigation from weather threats Protect life and property Humans issue advisories/watches/warnings for specific locations for pre-defined criteria (i.e. severe weather) Fly - Fight - Win 2
AFW Ensemble Prediction Suite (AFWEPS) Global Ensemble Prediction Suite (GEPS) Combination of GFS, GEM, and NOGAPS ensembles Post-processed at AFWA Mesoscale Ensemble Prediction Suite (MEPS) 10 members of WRF-ARW with unique physics configurations Initial conditions are deterministic UM, GFS, GEM, and NOGAPS 20 km northern hemisphere and tropical stripe domains to 144 hours run once per day (18Z) with online dust Five re-locatable 4 km domains run once per day to 54 hours Fly - Fight - Win 3
Configuration Weather Research and Forecasting (WRF) model v3.2 10 members using various combinations of physics/lower boundary schemes IB/LBC s: ½ GFS (3), 1 NOGAPS (2),.6 GEM (2), ~¼ UM (3) No data assimilation WRFPost processor Fly - Fight - Win 4
AFWEPS Webpage https://weather.afwa.af.mil/host_home/dnxm/jefs/jefs.html Available ensemble products: - Global (GEPS) - Mesoscale (MEPS) Precipitation Amount Precipitation Type Snow Amount Cloud Cover Lightning Dust Lofting Severe Weather Blizzard Surface Wind Gust Visibility Wind Chill Heat Index Smoke Trapping Potential Haboob Index Severe Weather Realtime verification also available on webpage Fly - Fight - Win 5
Product Goals Air Force Weather Product Goals: Provide probabilistic information as close to decision space as possible Match weather variables to pre-defined warning criteria Allow mission information to interact with weather information Ensure any output is useful, easy and quick to comprehend Fly - Fight - Win 6
4 km MEPS PEP Huntsville, AL multiple tornadoes from 20-23Z Assuming hourly probabilities are independent (no correlation), there was a 85% chance of a tornado within 20NM of Huntsville over the 15Z-03Z period. Fly - Fight - Win 7
JPADS and Ensembles limit risk exposure minimize cost Weather/mission interaction example Currently, must fly pre-mission over drop area and take a measurement of winds model too inaccurate on average exposure to risk With ensembles, decision makers can see if measurement needs to be taken based on uncertainty that day EX 85% of simulated drops for 31 May 2011 land in acceptable range acceptable risk to skip pre-mission cost savings and improved safety Fly - Fight - Win 8
Aerostats and Ensembles limit risk exposure minimize cost Weather/mission interaction example Tethered balloons (aerostats) in Afghanistan provide intel Very weather sensitive winds, lightning, turbulence, snow One joint probability product probability of hazardous conditions for aerostats User can then use risk management principles to decide when aerostats need to be reeled in (or plan for that possibility) Fly - Fight - Win 9
Tools for improvement Develop better, ensemble-derived, decision support products through dialog between developers and users Air Force Weather Tools for better products: Convection allowing ensembles (4 km resolution) Weather uncertainty due to convection is primary problem Algorithms to diagnose sub-grid scale probabilities High-impact phenomena are still sub-grid even at 4 km Inclusion of dust online inside model Dust from convection is #1 problem to solve Also working on dust source regions and uncertainties Fly - Fight - Win 10
4 km vs 12 km Proof of Concept (Deterministic) A B C 4 km proof of concept #1 12 km proof of concept #1 A--Convective activity mid-simulation (not in 12 km) B Mesoscale wind perturbation (in both 4 km and 12 km) C Sea breeze front from Persian Gulf (in both 4 km and 12 km) Fly - Fight - Win 11
26 Apr 01Z Little Rock AFB, AR Probability of a tornado within 20 nautical miles during the preceding hour. Black lines are updraft rotation (supercell) from the individual ensemble members. This can be used to see if supercells will be strong and/or long-track Fly - Fight - Win 12
Dialog with Users Develop better, ensemble-derived, decision support products through dialog between developers and users Air Force Weather Dialog with users: Prototype web page with products in real-time Rapid prototyping when users make requests Users prefer to judge utility for themselves objective stats or theoretical arguments not compelling Site visits for us to learn what the user needs are, and to explain what we have to offer Solicit feedback and continue the conversation Fly - Fight - Win 13
Dialog with users Feedback from users of test information: SW Asia forecaster: We found ensemble information to be far superior to currently operational products inability to move U2 s during dust storms was costing $500K per day ensembles correctly pointed to small window of opportunity to get U2 s out during multi-day storm 21 OWS (Europe): We have started integrating some of your products into our daily forecast process and they are kicking ass. I/we are most impressed with this Do we sound excited? We are! 51 OSS/OSW (South Korea): Looks like the ensemble nailed the rain/snow event we had today. The <1 mile Vis probability was instrumental in deciding if the first goes of the day should be scrubbed which included the 7th AF CC flying. We have partners working with us on test and evaluation at NWS WFO Omaha, Wichita, Minneapolis, Tallahassee, Des Moines, Paducah, Duluth, Davenport, St. Louis, in addition to the Aviation Weather Center. Fly - Fight - Win 14
Dialog with users Feedback from users of test information: 45 WS (Cape Canaveral): Our launch weather officers have looked at the products, particularly the probability plots (winds, precip, and lightning). These products are definitely value-added, as they provide information in the format we need.we would like to see JEFS products made permanent over our AOR. 28 OWS (SW Asia): However, I applaud the efforts of the Ensemble folks and their ability to do, whatever they are doing to the model (feeding it crackers?) to make it more accurate as to severe weather potentials their model physics are proving to be a more and more valuable component over the model suites we currently use. 25 th OWS (SW US): we had a remarkable performance improvement during the convective season for the flight most affected by convection. The capability that the 06Z model run presented for us during the monsoon allowed us to issue the necessary WWA s with more lead time Metrics: 16% increase in warning forecast skill, 24% decrease in false alarms with ensembles compared to without ensembles Fly - Fight - Win 15
Future Production Air Force Weather Web Services (AFW-WEBS) Probability of lightning overlaid with observed strikes Web based accessible anywhere Overlay models, observations, non-weather information relevant to mission Long term ensemble information databased, allowing for dynamic probability requests For instance, joint probability of 2 inches of snow, 35 knot winds, or visibility less than 3 miles in a 24 hour period. Fly - Fight - Win 16
Verification Fly - Fight - Win 17
Summary Air Force Weather is actively pursuing operational ensemble systems Focus on first 48 hours, convective scale, dust inclusion Development goal is to improve Air Force decisions Information presented near decision space Integrate with non-weather variables for optimization Continuous dialog with users Ensures buy-in, relevance to mission Improves final products Fly - Fight - Win 18