Air Traffic Analysis, Inc Using WITI for Airport Arrival Performance Analysis A report on work-in-progress December 2010
2 WITI and WITI-FA (Forecast Accuracy) WITI = Weather weighted by traffic En-route weather E-WITI uses actual convective Wx data, e.g. NCWD E-WITI-FA uses convective forecast data, e.g. CCFP, LAMP, … Both use the same scheduled traffic on major flows Convective forecast data is converted to quasi-NCWD format Terminal weather T-WITI uses actual surface Wx data (METARs) T-WITI-FA uses surface Wx forecast data (TAFs) Both use the same scheduled traffic at major airports TAF converted to quasi-METAR form, rolling look-ahead stream
3 VFR IFR VFR Weather Event Arrival Rates Time Airport Capacity Rate Actual Arrivals region of possible avoidable costs Framework for Quantifying Avoidable and Unavoidable Weather Impact
4 Arrival Rate Deficit: Illustration Example: LGA, 04/03/2009 Forecast called for rain, low ceilings and strong winds from the southwest which would have forced LGA into a single-runway operation with low arrival rate. Actual winds were much weaker. Ceilings lifted earlier than forecast. Deficit between scheduled and actually-achieved arrival rates needs to be measured Portion attributable to inaccurate weather forecast needs to be quantified This portion then needs to be split into two pieces: see next slide
5 Accidents, outages, VIP flights, security, etc Avoidable Unavoidable Reasonable risk mgmt Impact caused by other airports Inaccurate forecast Overly conservative TMI Actual AAR < TMI (in excess of risk mgmt) Deficit Bad Weather Avoidable and Unavoidable Portions of Arrival Rate Deficit: Breakdown Unavoidable portion of arrival rate deficit (gap between scheduled and maximum- achievable-given-actual-Wx arrival rates) needs to be subtracted from overall deficit. A reasonable risk mitigation factor (% arrival rate?) should also be subtracted. Impact caused by other airports, unrelated to this one, should be subtracted as well. Whats left is the avoidable portion. It can be subdivided into 4 categories: (a) Deficit caused by inaccurate forecast; (b) Overly conservative TMI – related to (a) and possibly partly caused by it; (c) Actual AAR below TMIs (even after discounting for risk factor), and (d) Transition. En-route to Terminal to Final Transition
6 Adding GDP Information GDP and non-GDP Periods Non-GDP GDP Non-GDP GDP
7 Non-GDP Arrival Rate Deficit Dissection Not counted (scheduled arrival rate is too low) Avoidable delay = 0 (actual arr rate >= scheduled) Over-forecast and actual arr rate < scheduled: all the deficit goes toward avoidable delay (Non-GDP Inefficiencies); a portion of the deficit is attributed to Wx forecast inaccuracy
8 GDP Arrival Rate Dissection No impact (GDP > scheduled rate) GDP Setup Inefficiency (but avoidable delay due to GDP Wx forecast inaccuracy = 0) GDP Execution Inefficiencies
9 Avoidable Delay/Cost analysis work is in progress and will continue in FY11 Another WITI application: airport delay prediction (Both programs are funded by the FAA ATO-P Aviation Weather Group and led by AvMet Applications, Inc)
10 Training WITI Model on Airport Delay Example: ATL, 2007 (WITI based on actual Wx) Once trained on historical data, the Airport WITI model can be used for delay prediction using forecast Wx and scheduled traffic for the day
11 Hourly Delay Forecast, 06/08/2008 Forecast: LAMP (convective); TAF (surface)
12 Hourly Delay Forecast, 01/28/2009 Forecast: TAF (surface)
13 Back-up Slides
14 WITI: Measuring Weather / Traffic Impact The Hand the NAS Is Dealt Every Day Traffic Demand National Airspace System (NAS) ATM, Airline Response Strategies Operational Outcomes Local Airport Weather En-route Convective Weather The Weather Index (WITI) expresses severity of weather impact on the NAS, weighted by air transportation service demands Capacity, Safety constraints
15 WITI is a weighted sum of three components: WITI Composition Weather Weighted by Traffic, Quantified – En-route Component: hourly frequency on major flows X amount of convective Wx that these flows cross – Terminal Component 15 Used by the FAA and NWS on a regular basis: Measure system performance in an objective manner – weekly reports Compare different seasons Wx/traffic impact with outcomes (e.g. delays) – Linear part: capacity degradation due to terminal weather impact, proportional to number of ops – Non-linear (Queuing Delay) part reflecting excess traffic demand vs. capacity
16 Method: Use Airport Arrival Rates Compute Arrival Rate Deficit We compare: Scheduled arrival rates from ASPM database Actual arrival rates, also from ASPM Model-generated arrival capacity based on METARs (i.e., actual weather data) Model-generated rates capacity based on TAFs (i.e., forecast weather data) Computed using a parametric model of airport capacity under different Wx conditions Use FAAs airport capacity benchmarks and historical data on actual airport throughput Any arrival rate deficit (possible minus actual) may be an indication of avoidable delays / cancellations
17 Re-Tooling WITI as an Airport Model Standard WITI is a NAS Wx Impact assessment tool A weighted sum of 3 components – Weights computed to provide best correlation between WITI and Delay for OEP34 airports combined WITI can be re-tooled as an airport model / delay predictor – Use an airport specific, much more detailed WITI metric and train it on that airports delay-vs-Wx-and-traffic-demand data – 12 components instead of 3 (ATL Wind WITI, EWR Snow WITI, ORD Convective WITI, etc) Calibrate WITI straight to minutes-of-delay for direct comparison with actual ASPM delays