National Aviation Day seemed like a good excuse to publish this.
I pulled six months of the US DOT's Reporting Carrier On-Time Performance database for routes with one end in Hawaii.
88,387 flights. 68 city pairs flown in both directions. One row for each flight that actually operated during the period. I'm using median air times rather than means.
The westbound flight being slower isn't exactly a discovery. The part I wasn't expecting was the relationship with distance.
Table:
SFO 2,398 mi 22 min 8.0% of the return
LAX 2,556 mi 30 min 10.3%
SEA 2,677 mi 27 min 8.7%
DEN 3,365 mi 44 min 12.1%
ORD 4,243 mi 66 min 14.8%
ATL 4,502 mi 81 min 16.8%
EWR 4,962 mi 93 min 17.5%
SFO loses 8% of the return time. EWR loses 17.5%. It's roughly twice the distance.
For the 66 westbound pairs the correlation with distance is r = 0.94, with a negative intercept. Mean extra time for all 68 pairs is 38 minutes. Median is 32.5.
I thought the obvious explanation would be the jet stream. Longer flights go farther north and spend more time in the stronger winds. Then I looked at Seattle.
It's the northernmost airport in that table. It's also 121 miles farther than LAX and the penalty is three minutes smaller. So I don't know.
The DOT files have the flight times and distances but not winds aloft. If anybody knows of historical winds data that could be matched to these flights, I'd like to play with it.
There was another thing I got wrong on the first pass through the data. Two routes looked like exceptions because the flight to Hawaii was faster.
They're Guam and Pago Pago. Both are west and south of Hawaii. The flight to Hawaii from either one is going east.
I calculated the initial great-circle bearing for each airport and the two "exceptions" stopped being exceptions: 66 of 66 westbound routes are slower and 2 of 2 eastbound routes are faster.
That's also what led to the way the map is drawn. Honolulu is in the middle and the airports are placed by distance and initial bearing. Guam and Pago Pago end up on the other side of the circle.
It's an azimuthal equidistant projection. No basemap and no map library. The map is inline SVG. Distance from the center is true distance and the angle is true bearing.
Airport coordinates are from OurAirports and are public domain. No external requests.
CSV is at /data. Each row has its source and check date. The dataset is CC0.
Disclosure: the site is a Hawaii travel site and makes money from affiliate links. There aren't affiliate links on this page or the data pages.
National Aviation Day seemed like a good excuse to publish this.
I pulled six months of the US DOT's Reporting Carrier On-Time Performance database for routes with one end in Hawaii.
88,387 flights. 68 city pairs flown in both directions. One row for each flight that actually operated during the period. I'm using median air times rather than means.
The westbound flight being slower isn't exactly a discovery. The part I wasn't expecting was the relationship with distance.
Table: SFO 2,398 mi 22 min 8.0% of the return LAX 2,556 mi 30 min 10.3% SEA 2,677 mi 27 min 8.7% DEN 3,365 mi 44 min 12.1% ORD 4,243 mi 66 min 14.8% ATL 4,502 mi 81 min 16.8% EWR 4,962 mi 93 min 17.5%
SFO loses 8% of the return time. EWR loses 17.5%. It's roughly twice the distance.
For the 66 westbound pairs the correlation with distance is r = 0.94, with a negative intercept. Mean extra time for all 68 pairs is 38 minutes. Median is 32.5.
I thought the obvious explanation would be the jet stream. Longer flights go farther north and spend more time in the stronger winds. Then I looked at Seattle.
It's the northernmost airport in that table. It's also 121 miles farther than LAX and the penalty is three minutes smaller. So I don't know.
The DOT files have the flight times and distances but not winds aloft. If anybody knows of historical winds data that could be matched to these flights, I'd like to play with it.
There was another thing I got wrong on the first pass through the data. Two routes looked like exceptions because the flight to Hawaii was faster.
They're Guam and Pago Pago. Both are west and south of Hawaii. The flight to Hawaii from either one is going east.
I calculated the initial great-circle bearing for each airport and the two "exceptions" stopped being exceptions: 66 of 66 westbound routes are slower and 2 of 2 eastbound routes are faster.
That's also what led to the way the map is drawn. Honolulu is in the middle and the airports are placed by distance and initial bearing. Guam and Pago Pago end up on the other side of the circle.
It's an azimuthal equidistant projection. No basemap and no map library. The map is inline SVG. Distance from the center is true distance and the angle is true bearing.
Airport coordinates are from OurAirports and are public domain. No external requests.
CSV is at /data. Each row has its source and check date. The dataset is CC0.
Disclosure: the site is a Hawaii travel site and makes money from affiliate links. There aren't affiliate links on this page or the data pages.