Traffic Flow Impact (TFI) Michael P. Matthews 27 October 2015 Sponsor: Yong Li, FAA ATO AJV-73 Technical Analysis & Operational Requirements Distribution Statement A. Approved for public release; distribution is unlimited. This work is sponsored by the Federal Aviation Administration under Air Force Contract #FA8721-05-C-0002. Opinions, interpretations, recommendations and conclusions are those of the author and are not necessarily endorsed by the United States Government.
Example Weather Impact on Flows 11 September 2013 Light Moderate Heavy Intense Severe Extreme Traffic Flow Impact - 2
Example Weather Impact on Flows 11 September 2013 Light Moderate Heavy Intense Severe Extreme Traffic Flow Impact - 3
Example Weather Impact on Flows 11 September 2013 Major air traffic delays Severe impact in NY and significant delays across the whole travel system (69 diversions, 72 taxi-backs in NYC, Holding, Cancelled flights, etc.) Convective weather never shut down JFK, LGA, or EWR Impact was to the primary arrival/departure corridor to the west Event was well forecast Weather forecast models predicted storms Placement, Timing, Intensity were consistent from pre-dawn hours Forecast information was never used to mitigate delays Return to orderly operations took hours Heavy Demand Diverted Flows Limited Demand 17:50 UTC (Onset) 21:10 UTC 00:30 UTC Traffic Flow Impact - 4
Constraints and Traffic Management Initiatives (TMI) Severe constraints Significant constraints Moderate constraints Remove the demand (e.g. Playbook Reroutes) Reduce the demand via pre-departure delay (e.g. Airspace Flow Programs, Ground Delay Programs) Manage impacts locally at impact time (e.g. Miles-in-Trail, Deviations, Tactical Reroutes) Traffic Management Initiatives Manage Tactically Scale of impacts determines TMI scope (Moderate, Significant, Severe) Larger scope requires longer forecast / planning horizons, which increases uncertainty Traffic Flow Impact - 5
Decision-based Weather Impacts Airspace Flow Program Example Resources Decisions Flow Constrained Area (FCA) Airspace Flow Program (AFP) FCAOB1 Used to reduce demand for airspace in the presence of impacting weather Flow is throttled across FCA When to start the AFP? When to stop the AFP? What flow rate to set for the AFP? Strategic decisions are FLOW based Traffic Flow Impact - 6
Defining an Air Traffic Management Event 100 Tactical Management Onset Duration Permeability Limited Strategic Management Aggressive Strategic Management Maximum Impact 0 0 1 2 3 4 5 6 7 8 9 10 11 12 Forecast Period (hr) Strategic planners focus on ONSET, DURATION, and IMPACT Traffic Flow Impact - 7
4-D Weather Forecast Decision Support: Traffic Flow Management Translation of Weather into Air Traffic Management Decision Making With the current weather forecasts Decision Support Tools Storm Intensity Manage tactically? Storm Height Displays of weather forecasts how do you make these decisions? Flow-Based Capacity Impact Model Program Guidance Air Traffic Management Decision Making and Execution Develop and decide on air traffic management plan Execute plan Traffic Flow Impact - 8
Outline Background Weather translation into flow impact Computing permeability of airspace Validating impact with observed flow rates Forecasting permeability and uncertainty Traffic Flow Impact Tool Summary Traffic Flow Impact - 9
Convective Weather Avoidance Model (CWAM) WEATHER DATA Storm Intensity WEATHER AVOIDANCE FIELD Deviation Probability Lookup Table Storm Height Spatial Filters Flight Altitude Storm Height Area Coverage of Storm Weather Avoidance Field (WAF) DEVIATION DATABASE Non-Deviation Deviation Statistical Pattern Classifier Deviation Probability Flight Altitude Storm Height Area Coverage of Storm Traffic Flow Impact - 10
Weather Translation Process Translator Route & Flow Impact Model Encounter Length 1.0 0.8 0.5 0.0 Permeability Translated Metric September 11, 2013 ZOB/ZNY Transition Airspace Weather situation (WAF from VIL, EchoTops) Flow crossing Encounter Intensity Permeability Flow boundary Time of Day (UTC) Defined airspace resource Airspace traversing trajectories Permeability = Aggregation of all route impacts Traffic Flow Impact - 11
Validating Flow-Based Capacity Impact Model Validation requires a method of measuring an observed air traffic flow rate Flow Rate = (# AC / median transition time) * 15min Aircraft in Flow Aircraft NOT in Flow Non-weather event 06 August 2013 23:45 Light Moderate Heavy Intense Severe Extreme Traffic Flow Impact - 12
Weather Impacted Day 11 September 2013 18:45 UTC Light Moderate Heavy Intense Severe Extreme Without proper planning, tactical operations become inefficient - Over delivery of aircraft at onset Excessive tactical workload Favor arrivals over departure - Reactionary response Stops all future demand 01:00 UTC Light Moderate Heavy Intense Severe Extreme Traffic Flow Impact - 13
Weather Impacted Day 11 September 2013 20:00 UTC Light Moderate Heavy Intense Severe Extreme Rarely will the flow rate drop to zero - Long haul flights Departed many hours ago - Aircraft holding Anticipation of clearing Maintaining some demand 21:30 UTC Light Moderate Heavy Intense Severe Extreme Traffic Flow Impact - 14
Statistical Analysis of Flow-Based Capacity Impact Model Operational Decisions (Good and Bad) affect quality of analysis Assembled large database of weather impacted and non-impacted days Weather impacted days (70 2011 through 2015) Non-weather days (46 2013 through 2015) 116 Days of five minute flow rates Flow Rate (non-impact days) ZOB/ZNY Transition Airspace 15 Minute Flow Rate Historical time period of day with Highest Demand Time of Day Traffic Flow Impact - 15
Correlation of Airspace Permeability and Flow Rate ZNY/ZOB Transition - Airspace Permeability vs Flow Rate 36 32 Flow Rate (ac/15 min) 28 24 20 16 12 8 4 Maximum observed value Most likely range 10 th to 90 th Percentile Median value Modeled Permeability and Observed Flow Rates are correlated Translation of weather forecasts into meaningful flow impacts has not been possible until now! 0 0-19 20-39 40-59 60-79 80-99 100 Airspace Permeability Traffic Flow Impact - 16
Outline Background Weather translation into flow impact Computing permeability of airspace Validating impact with observed flow rates Forecasting permeability and uncertainty Traffic Flow Impact Tool Summary Traffic Flow Impact - 17
Weather Forecast Translation Process Extrapolation Forecast Translator Translated Metrics Feature Extraction 3 rd order translated metric (TMI onset time, duration, rate) Time lagged HRRR Forecast LAMP Forecast SREF-CTP Forecast Identification of similar scenarios Probability Distribution Function 2 nd order translated metric (e.g., flow rate) Permeability Forecast Uncertainty (Spread) Flow Rate 1 st order translated metric (e.g., permeability) Increasing specificity Tactical Management: No action Limited Management: Reduce Demand Permeability Aggressive Management: Major Flow Reduction Categorical Impact Traffic Flow Impact - 18
Machine Learning Approach to Forecast Uncertainty Estimation Uncertainty defined as bounds of true permeability given a forecast Model derived from statistical analysis of past forecast performance Remove forecast bias and estimate uncertainty bounds Permeability forecast model trained from Summer 2013/2014 cases Trained on Northeast FCAs Validation results obtained using ten-fold cross validation Neural network trained for quantile regression Uncertainty bounds: 20 th, 50 th, 80 th percentile TFI Permeability (%) 100 80 60 40 20 Traffic Flow Impact Forecast Permeability Forecast Traffic Flow Impact - 19 Training FCAs HRRR = High Resolution Rapid Refresh model FCA = Flow Constrained Area 0 Time (UTC)
Forecast Uncertainty Example Extrapolation Forecast 100 Time lagged HRRR Forecast TFI Permeability (%) 80 60 40 20 LAMP Forecast 0 1hr 2hr 3hr 4hr 5hr 6hr 7hr 8hr 9hr 10hr 11hr 12hr SREF-CTP Forecast TFI Permeability (%) 80 60 40 20 0 1hr 2hr 3hr 4hr 5hr 6hr 7hr 8hr 9hr 10hr 11hr 12hr Traffic Flow Impact - 20
Quantified Accuracy of Translated Forecast Components Forecast Performance Forecast Skill (R 2 between forecast and observed impact) Forecast Lead Time, hours Formerly notional curves now quantified Traffic Flow Impact - 21
Outline Background Weather translation into flow impact Computing permeability of airspace Validating impact with observed flow rates Forecasting permeability and uncertainty Traffic Flow Impact Tool Summary Traffic Flow Impact - 22
Decision Support for TMI Planning: Traffic Flow Impact (TFI) Forecasts FY14: Informal field evaluation at ATCSCC Very strong positive reaction by planners and ATCSCC operations manager ZNY001 FY15: Expanded evaluation Algorithm improvements funded by AJV-73 Field evaluations: AJM-33 ATCSCC 4 Enroute Centers 5 Airlines Report on expanded 2015 field evaluation pending Clicking displays confidence plot Predicted permeability category I think your data yesterday was spot on. I wrote to (FAA NE Tactical Operations) and I want to make certain we discuss this after the summer. It seems to be providing really good results. Thank you! - ATCSCC Planner Traffic Flow Impact - 23 TMI = Traffic Management Initiative ATCSCC = Air Traffic Control System Command Center
Summary Developed method to translate convective weather forecasts into flow-based capacity impacts for decision makers Impact category, permeability, flow rate Includes translated forecast confidence Validated model by measuring observed flow rates during clear air and convective storms Several promising operational/research initiatives Operational use and feedback by the command center Impact modeling in serious gaming: NASPlay Confidence modeling for decision making Traffic Flow Impact - 24