Practical Applications of
Probability in Aviation
Decision Making
Haig Iskenderian
22 October 2014
Portfolio of TFM Decisions
Playbook Reroutes
Ground Stops
Ground Delay
Programs
Airspace Flow
Programs
National
9+
2014 FPAW 2
Iskenderian
Arrival & Departure
Fix Metering
Regional
6
3
Planning Horizon (Scale & Hours)
Local
0
Sample Strategic TFM Decisions
Sample Strategic Forecast Tools
Strategic Decisions
Manage tactically?
Throttle traffic flow?
LAMP
(Probabilistic)
CoSPA
(Deterministic)
Strategic reroutes?
TFM decisions require high-fidelity forecasts of storm
structure and characteristics
2014 FPAW 3
Iskenderian
Reduce
airport
arrivals?
Approaches to Probabilistic Forecasts
for Aviation
Model Ensemble
Thunderstorm Probability
?
Don’t
Know
50%
Historical Data
Translate to
probability
Translate to
decision
No
Yes
0%
Take
Action
100%
AFP!
Model Ensemble
Ensemble Airspace
Classification
Probability of Impact
0.8
0.6
Historical Data
Translate to
probability
Translate to
decision
Take
Action
0.4
0.2
0
Low Impact
Medium Impact
High Impact
2014 FPAW 4
Iskenderian
Low
Impact
Medium
Impact
High
Impact
Decision-based Uncertainty
Define Resources
Define Decisions
Provide Uncertainty
FCA
Airspace Availability
NY Metro FCA
1 hr
FCAOB1
2 hr
3 hr
4 hr
5 hr
6 hr
Most Likely Availability &
Range
7 hr
8 hr
FCAOB1
Blockage
FCAOB1
FCABW1
FCAA01
Low Impact
Medium Impact
High Impact
Low Impact
Medium Impact
0 hr
Time
8 hr
High Impact
Enroute
Terminal
Gate Availability
0.5 hr
N
E
S
1 hr
1.5 hr
2 hr
2.5 hr
3 hr
West Gates
3.5 hr
4 hr
Blockage
NY Metro Gates
Low Impact
Medium Impact
High Impact
W
0 hr
Terminal
2014 FPAW 5
Iskenderian
Time
4 hr
FCA Blockage and Uncertainty
Forecast Issued 12 UTC
FCA Blockage
0%
“Best Case”
High
Forecast
Scenario
Uncertainty
Low Impact
20%
Ensemble of CoSPA Forecasts
Valid 8 Hours in the Future (20Z)
40% Medium Impact
60%
TMI ?
80%
High Impact
“Worst Case”
Scenario
100%
Time (UTC)
Most likely blockage
Possible range of blockage
• Blockage forecast and uncertainty based on:
– Model to translate storm intensity and height to blockage
– Time-lagged ensemble of CoSPA forecasts
– Historical forecast performance given scale, orientation,
intensity of weather
– Location, issue time, valid time
2014 FPAW 6
Iskenderian
FCA Blockage and Uncertainty
Ensemble of CoSPA Forecasts
Valid 8 Hours in the Future (20Z)
Forecast Issued 12 UTC
FCA Blockage
0%
Low Impact
High Forecast
Uncertainty
20%
40% Medium Impact
60%
TMI ?
80%
High Impact
100%
Time (UTC)
Forecast Issued 15 UTC
FCA Blockage
0%
Ensemble of CoSPA Forecasts
Valid 5 Hours in the Future (20Z)
Low Impact
20%
TMI !
40%
Low Forecast
Uncertainty
Medium Impact
60%
80%
High Impact
100%
Observed blockage
2014 FPAW 7
Iskenderian
Time (UTC)
Regular forecast updates needed to revisit decisions
Forecast Event Start, Duration and
Recovery
Event Start
Forecast Issued 15 UTC
FCA Blockage
0%
Low Impact
Event start
20%
40%
Ongoing event
Medium Impact
60%
80%
High Impact
Ongoing Event
100%
Time (UTC)
Forecast Issued 22 UTC
FCA Blockage
0%
20%
40%
Ongoing
event
60%
Event
recovery
80%
100%
Time (UTC)
2014 FPAW 8
Iskenderian
Event Recovery
Summary
• Traffic Flow Management decisions encompass a portfolio of
time and space scales in the NAS
• Probabilistic forecasts cast in TFM decision space could
yield considerable operational benefit for strategic planning
– Convert high-fidelity weather forecasts to aviation impact,
then provide probability of impact
– Provide event start, severity, and duration
– Focus here on enroute airspace, but similar approach
could be used for other airspace and time horizons
• Probabilistic forecasts need to update regularly so that TFM
decisions can be revisited and revised
2014 FPAW 9
Iskenderian
Future Work
• Map FCA blockage to flow rate
• Consider uncertainty for broader portfolio of decisions
• Incorporate other models in decision-based uncertainty
(SREF, LAMP)
2014 FPAW 10
Iskenderian