optimizing wfo aviation service & forecast performance dan shoemaker aviation curmudgeon wfo fwd
TRANSCRIPT
Optimizing WFO Aviation Service & Forecast Performance
Dan ShoemakerAviation CurmudgeonWFO FWD
Optimizing WFO Aviation Service & Forecast Performance
• Background—NWS performance.• Improving Service. – My / FWD’s “TAF Philosophy.”– False Alarm Rate.– Individual Feedback.–Aviation Statistics/Climatology.
NWS TAF Performance: Are we any good?
• Last year I put together a “report card” for a visit with Southwest Airlines.– Airlines have little/no access to TAF stats.
• 2000/3 % improvement over LAMP; 0-6 hours; for every SWA/AirTran city.– We have to be better than an automatic product.– Shows which offices are “paying attention”.
TUSSAN
ONTLA
XSFO SM
FRNO
SEABOI
DENM
SPM
DWCM
H CLE BUFBOS
BDLLG
ABW
IORF
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.021085.2 57.161.5 67.2
CSI % Impv over LAMP 0-6hr 2000/3 FY2010
TAF PerformanceWestern Region SWA Cities
Eastern Region
OMA MSP MKEMDW IND CMH SDF CLE PIT BUF MHT BOS PVD BDL ISP LGA PHL BWI IAD ORF0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
CSI % Impv over LAMP 0-6hr 2000/3 FY2010
Southern Region
-20
-15
-10
-5
0
5
10
15
20 21.9 21.0
-0.7
17.2
-0.8
-16.5
-38.5
CSI % Impv over LAMP 0-6hr 2000/3 FY2010
AirTran Cities
-20.0
-15.0
-10.0
-5.0
0.0
5.0
10.0
15.0
20.0
-0.3-1.4
21.4
-26.3
-1.9
15.2
CSI % Impv over LAMP 0-6hr 2000/3 FY2010
FWD 1000/3 Performance • From Stats on Demand:
“… A review of previous years forecast performance showed that the percent frequency of occurrence of IFR conditions is a reasonable first estimate predictor for overall forecast performance for a given time period, or region...”
% Occurrence of IFR & BLO National POD/FAR Results
10 63 / 39 8 58-62/42-45 6 57-60/42-45 4 56-59/44-46
FWD: 5 70 / 32
Optimizing WFO Aviation Service & Forecast Performance
The Problem?• Office culture can be “stuck” in bad habits.• Forecasters don’t see effects of their TAFs.– AFP must become “lobbyist” for Customers.– Long term advocate for improvement…there are
no quick fixes. • Forecasters must take ownership of their TAFs.– Individual feedback!
My Forecast Philosophy• Optimism beats pessimism; or “when in doubt,
leave it out.” (more later)• On “cautious” TAFs:– If TAFs could crash airplanes, the country would be
littered with aluminum (copper theft would decrease).
• On too much detail:– TAFs are not a portrait of the atmosphere —
they’re a stick figure. As long as they’re anatomically correct, they’re doing their job.– (this is PPTAF philosophy as well).
More Philosophy
• 2000/3 is the most important TAF category!!– TAF is mandated by FAA for airlines, not GA. – 1000/3 is a GA category, not an airline category.– Costs the airlines money.– Occurs more often than lower categories.
• This is where AFPs should place greatest emphasis.– 2000/3 focus will not degrade 1000/3 performance.
False Alarm Rate (FAR) (cig/vis)• 4 possible outcomes for a TAF (consider 2000/3):– Hit: fuel added, good forecast.– Correct Negative: no extra fuel, plane lands at
destination.– Missed event: no extra fuel, plane lands at
destination anyway (vast majority of cig/vis cases).– FA: fuel always added, not needed.
• 100% of FA’s penalize airlines – FA’s always add cost – much worse than missed event.
Optimism beats pessimism (FAR)• If you are “on the fence” about an occurrence, and you
bet on the lower category, you’re basically betting 50/50 on a low probability event. By the way, Las Vegas wants you to stop by on your way home.
• DFW/DAL annual % occurrence: (AVNFPS Climo tools) (SOD)
– VFR 82% Above 2000/3 90%– MVFR 12% Below 2000/3 10%– IFR 5%– LIFR & Below <1%
• Caution/pessimism affects FAR much more than POD.
FWD TAF Rules:
• Forecast probable weather, not possible weather (use AFD for possibilities).
• Don’t cross thresholds unless conditions will probably occur.– 2000/3 is most important category, then 600/2
(400/1), 1000/3 then others. • No TEMPOs in the Outlook Period. If you have
confidence, write a new line.
More FWD TAF Rules:
• No ¼ SM in the main body. Period.– Not our job to shut down operations. We forecast
weather, & let airlines decide whether to fly. • No PROB30s. (more later)• Forecast TSRA during probable times. Use VCTS &
CB for when less likely.• Use the AVNFPS cig/vis trend tool.
• The TAF police will always find you!
Individual Feedback
• Individual feedback is key to lowering FAR & improving performance. – Many think the problem lies elsewhere.
• You’re going to hear excuses…facts will trump them.– Shoe’s 4 stages of feedback — whining, crying,
grumbling, & finally acceptance.• Many forms of feedback — choose what
works best for you.
Individual Feedback
• Run individual TAF stats—at least a year’s worth to get a good sample size.– Annotate/circle/comment/point out good & bad.– The biases are a good place to find where individual
improvements can be made (pessimistic forecasters).
• Show collated statistics so each individual can see where they stand in the group.– Note: LAMP performance can vary considerably
person by person.
Individual Feedback
A
B
C
D
E
F
G
H
I
J
K
L
M FWD
-20 -10 0 10 20 30 40 50 60 70 80
4.0
-4.7
-1.3
-0.8
19.9
18.6
9.8
-3.0
4.2
0.8
2.2
7.9
4.0
22.1
-8.8
70.5
20.7
10.1
-5.3
19.0
-3.5
28.3
4.2
0.0
38.0
12.5
1000&32000&3
A
B
C
D
E
F
G
H
I
J
L
M
AVG
0 50 100 150 200 250 300 350 400 450
Hours of TEMPO Forecast
% Improvement vs LAMP 2009 Fcstr “D” had no clue they wrote that many TEMPOs
Individual Feedback—Forecaster “D”Did not beat MOS his/her first year at FWD
Issued a 24 hr TAF with 18 hours of TSRANote how low FAR improves scores & POD actually increased.
Forecaster “D” LAMP
Uh Oh!
TAF KDAL 082333Z 0900/0924 11014G24KT P6SM BKN050 FM090200 11014KT P6SM BKN050 FM090500 07015KT P6SM OVC025 FM090900 05012KT P6SM -RA OVC015 FM091100 36018G28KT 1/2SM SN OVC006 FM091800 34021KT 3SM -SN OVC010 FM092200 35015KT P6SM OVC025
PROB30 (SR)May – Apr
PROB30 (FWD) May – Apr
Warm Season Afternoon ThunderstormsAfternoon precip occurs @ KDFW 4% of time (SOD).
When precip occurs (AVNFPS climo distribution):
17Z 18Z 19Z 20Z 21Z 22Z 23Z 00Z 01Z 02Z 03Z0
5
10
15
20
25
30
35
40
VFRMVFRIFRLIFRVLIFR
Thunderstormstwo TAFs (warm season):
TAF KDFW 061800/072400 18012KT P6SM SCT060CB TEMPO 0620/0623 3SM TSRA BKN015CB FM070100….
TAF KDFW 061800/072400 18012KT P6SM SCT060CB TEMPO 0620/0623 6SM TSRA BKN050CB FM070100….
What’s the difference? #1 takes 3 hour hit on category.
Warm Season Afternoon Thunderstorms
We’re not ignoring the flight category, we’re forecasting the probable category.
Part of MFL problem?
17Z18Z
19Z20Z
21Z22Z
23Z00Z
01Z02Z
03Z05
10152025303540
VFRMVFRIFRLIFRVLIFR
Feedback Works
FY 2011MFL CSI
15.4 % Improvement
over LAMP
FY 2010MFL CSI
18.8% Worse
than LAMP
MFL LAMP
Miscellaneous
D2D Map backgroundAll ASOS/AWOS voice lines. Get current conditions (no
specials from AWOS). SVR WX wind gusts as well.
Summary• Stress optimistic forecasting (improve FAR) &
PPTAF principles. Focus on 2000/3, but don’t forget other important categories. Individual feedback works.
• Dig into aviation climatology to find areas ripe for improvement (AVNFPS climo tools & SOD).
• Every office has bad forecasts…learn from them. • Improvements will take sustained effort,
especially in the beginning. The office culture will change, & service/scores will improve.
Summary• Constant abrasion produces the pearl; it’s a disease of the oyster.
Lenny Bruce
• Questions/Comments?