Emerging Aviation Weather Research at MIT Lincoln
Laboratory*
Haig Iskenderian
19 November 2015
*This work was sponsored by the Federal Aviation Administration under Air Force Contract No. FA8721-05-C-0002. Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the United States Government.
2015 FPAW 2Iskenderian
• Offshore Precipitation Capability
• Convective Weather Avoidance Polygons
• Forecast Confidence
Outline
2015 FPAW 3Iskenderian
Aviation Weather Information Shortfall:Limited Offshore Observations and Forecasts
NEXRAD Radar Coverage
Good Coverage
Degraded Coverage
NoCoverage
Current Radar Analysis
No Weather Radar
Precipitation Intensity
Sample Flight Tracks Current Forecast Domain
CurrentForecast Domain
The Offshore Precipitation Capability (OPC) is being developed to provideoperational radar-like view of weather beyond radar coverage
2015 FPAW 4Iskenderian
OPC Input Data
Lightning Geostationary Satellite Numerical Weather Prediction Models
Visible Infrared
Update rate: ~10 – 30 minutes
Update rate: Forecasts issued hourly
4 Channels, 4 km res1 km resolution
Update rate: Every second
Lightning Flashes, Oct 14th, 2014Temperature, Pressure, & Winds
Earth Networks Total Lightning Network Sensor Locations
CG Flash+
GFS Forecast
RAP Model
2015 FPAW 5Iskenderian
OPC Processing
Inputs Feature Extraction Machine Learning(Random Forest)
.
.
.
.
)(1)( xtN
xFi
i∑=x
)(1 xt
)(2 xt
)(xtN
Output
Satellite
Lightning
Model
Multiple, heterogeneous inputs are flexibly accommodated and optimally combined
2015 FPAW 6Iskenderian
Merging OPC and RadarRadar
Machine Learning
Merged Radar + OPC
OPC is used to fill areas without weather radar coverage
2015 FPAW 7Iskenderian
OPC Mosaics and LightningRadar + Satellite + Lightning Radar + OPC + Satellite
2015 FPAW 8Iskenderian
Offshore Forecast
CoSPA Domain
OPC Analysis 17 UTC 3 October 2015
OPC is the starting point for the offshore forecast
HurricaneJoaquin
2015 FPAW 9Iskenderian
1 Hour Offshore Forecast
Valid 18 UTC 3 October 2015
Offshore Forecast blends OPC extrapolation with RAP numerical model forecast
2015 FPAW 10Iskenderian
8 Hour Offshore Forecast
Valid 01 UTC 4 October 2015
Offshore Forecast blends OPC extrapolation with RAP numerical model forecast
2015 FPAW 11Iskenderian
8 Hour Offshore Forecast Comparison
Valid 01 UTC 4 October 2015
Offshore Forecast extends the range of current forecast
Current 8 Hour Forecast Offshore 8 Hour Forecast
Outside currentforecast domain
2015 FPAW 12Iskenderian
• Offshore Precipitation Capability
• Convective Weather Avoidance Polygons
• Forecast Confidence
Outline
2015 FPAW 13Iskenderian
Weather Avoidance FieldIdentifying Storms that Pilots Avoid
Convective Weather Avoidance ModelWEATHER DATA
Spat
ial F
ilter
s
DEVIATION DATABASENon-Deviation Deviation
Area Coverage of Storm
Deviation Probability
StatisticalPattern
Classifier
Flig
ht A
ltitu
de –
Stor
m H
eigh
t
WEATHER AVOIDANCE FIELD
Weather Avoidance Field (WAF)
Deviation ProbabilityLookup Table
Storm Intensity
Storm Height
Area Coverage of Storm
Flig
ht A
ltitu
de –
Stor
m H
eigh
t
2015 FPAW 14Iskenderian
• In en route airspace, pilots avoid storms, not pixels
• Convective Weather Avoidance Polygon (CWAP) combines edges in the echo top field with WAF
• Identify the boundaries of storms that pilots tend to avoid
Convective Weather Avoidance PolygonsIdentifying Clusters of Storms that Pilots Avoid
Deviation probability
CWAP
08 August 2015
15 – 60 minute forecasts of CWAP can alert dispatchers, pilots, and ATC to growing storms that should be avoided
Precipitation intensity
01:40Z 02:10Z
2015 FPAW 15Iskenderian
CWAPs, Flights and Weather
2015 FPAW 16Iskenderian
Offshore CWAPs based on OPC
WAF Probability (%
)
2015 FPAW 17Iskenderian
• Offshore Precipitation Capability
• Convective Weather Avoidance Polygons
• Forecast Confidence
Outline
2015 FPAW 18Iskenderian
How Much Can I Trust the Forecast?
Forecasts issued 12 UTC 14 June 2015
6 hr HRRR ForecastValid @ 18Z
11 hr HRRR ForecastValid @ 23Z
LAMP ForecastValid @ 18Z
LAMP ForecastValid @ 23Z
Users often compare and contrast forecasts
2015 FPAW 19Iskenderian
Translate Weather to PermeabilityImportant for Enroute Decisions
CWAM
Medium Impact
Low Impact
High Impact
Observed Weather@ 23Z
Observed Weather@ 18Z
Perm
eabi
lity
(%)
Time (UTC)
100
60
40
20
0
80
2015 FPAW 20Iskenderian
• Developing machine learning methods to combine multiple forecasts of varying skill to provide the confidence in a permeability forecast
Forecast ConfidenceCombine Multiple Forecasts
Time (UTC)
Perm
eabi
lity
(%)
100
60
40
20
0
80
Traffic Flow Impact (TFI) Forecast
80th Percentile
50th Percentile
Black Line:Observed Permeability
80th Percentile
20th Percentile
Mean
Spread of 80th and 20th percentiles indicates forecast confidence
Extrap HRRRTime-lag
SREF LAMP
2015 FPAW 21Iskenderian
Perm
eabi
lity
(%)
100
60
40
20
0
80
Traffic Flow Impact (TFI) Forecast
Time (UTC)
Assess individual forecasts
Extrap HRRRTime-lag
SREF LAMP
• Developing machine learning methods to combine multiple forecasts of varying skill to provide the confidence in a permeability forecast
Forecast ConfidenceCombine Multiple Forecasts
2015 FPAW 22Iskenderian
Time (UTC)
Perm
eabi
lity
(%)
100
60
40
20
0
80
Traffic Flow Impact (TFI) Forecast
Observed Permeability
Observed permeability should tend to fall within shading
Extrap HRRRTime-lag
SREF LAMP
• Developing machine learning methods to combine multiple forecasts of varying skill to provide the confidence in a permeability forecast
Forecast ConfidenceCombine Multiple Forecasts
2015 FPAW 23Iskenderian
Quantified Accuracy of Translated Forecast Components
*Forecast Skill: coefficient of determination (𝑹𝑹𝟐𝟐) between forecasted and observed permeability
Combination of forecasts outperforms any single forecast
2015 FPAW 24Iskenderian
• Offshore Precipitation Capability (OPC)– Creating radar-like analyses for regions beyond radar– Forms the starting point for offshore forecast– Potential operational platforms for OPC include FAA NextGen Weather
Processor and NWS Multi-Radar/Multi-Sensor (MRMS) system
• Convective Weather Avoidance Polygons (CWAP)– Leverages Convective Weather Avoidance Model – CWAP defines a region of airspace that pilots tend to avoid – Included as part of NextGen Weather Processor technical transfer
• Forecast Confidence– Combines multiple forecast models to provide forecast confidence for
enroute planning– Included as part of NextGen Weather Processor technical transfer
Summary
Top Related