When the Weather Is at the Other End

Summary

A flight that left a normal-weather airport for one in bad weather was cancelled 3.2% of the time and diverted 1.8%, against 1.0% and 0.1% in the same airport, calendar month and scheduled hour with clear weather. Measured on 14,324,742 flights between covered airports.

Every scheduled flight between two of the covered airports is joined to the weather at its origin in its scheduled departure hour and to the weather at its destination in its scheduled arrival hour. 14,324,742 flights have weather at both ends. An hour is bad weather when it is IFR or worse, or has a thunderstorm, snow or ice, fog, heavy rain or high wind, and normal when it is clear flying weather with none of those. 9,908,575 flights had normal weather at both ends and 875,396 left a normal airport for one in bad weather. 954,501 left bad weather for a normal airport, and 164,078 had bad weather at both ends.

Flights between two covered airports, by the weather at the origin in the scheduled departure hour and at the destination in the scheduled arrival hour. The lateness columns count flights that were neither cancelled nor diverted. 2,422,192 flights with marginal weather at either end are in neither group.

A flight that left a normal-weather airport for one in bad weather was cancelled 3.2% of the time, against 1.0% for flights in the same airport, calendar month and scheduled hour with clear weather at both ends: 2.2 points more, or 21.5 extra cancellations in every 1,000 flights. A flight that left bad weather for a clear airport was cancelled 3.4% of the time against 1.0%, 2.4 points more. In cancellations, weather at the destination costs about as much as weather at the origin. Of the extra cancellations in the three bad-weather groups, 36% were flights into bad weather from a clear airport, 43% flights out of bad weather to a clear one and 21% flights with bad weather at both ends. Counting the last group, bad weather at the destination was part of 57% of the extra cancellations and bad weather at the origin 64%.

Expected is the rate among flights in the same stratum (airport, calendar month and scheduled hour: the destination’s for the first and third rows, the origin’s for the second) whose two ends were both in normal weather. A stratum with fewer than 120 such flights gives no expectation and its bad-weather flights are left out.

A flight into bad weather from a clear airport left 15 or more minutes late 25.4% of the time, 5.6 points above its expectation, and arrived 15 or more minutes late 29.7% of the time, 10.8 points above. Most of what the destination’s weather added to the lateness appeared between pushback and arrival. It also diverted: 1.76% of these flights landed somewhere else, against 0.13% expected, 13 times as often. Flights out of bad weather diverted at 0.15%, close to the 0.12% expected. Flights with bad weather at both ends left late 30.2% of the time and arrived late 38.5% of the time.

Snow or ice at the destination cost the most cancellations of the conditions listed, 7.3 points above expectation among 100,009 flights from clear origins. In 4 of the 7 conditions with enough flights at both ends, weather at the destination cost at least 85% of what it cost at the origin; the largest shortfall was thunderstorm, 3.4 points into it and 5.3 out of it. The condition that diverted the most flights was heavy rain, 7.7 points above expectation. The conditions overlap, and each is compared with flights in the same airport, calendar month and scheduled hour in normal weather.

Flights into the condition left a normal-weather origin; flights out of it arrive at a normal-weather destination. The conditions overlap, so a flight in fog and low visibility is in both rows. A dash is a row under 120 flights.

Across the 62 covered airports with at least 500 flights from a clear origin into their bad weather, DFW was highest, with 58.2 extra cancellations per 1,000 flights (7.3% cancelled against 1.5% expected). LAX was lowest, with 2.1 (0.8% against 0.6%). Bad weather means different things at different airports, and the airlines and routes into them are not the same, so this describes the flights that arrived there and does not rate the airport.

Covered airports with at least 500 such flights and a cancellation rate at least 2 standard errors from its expectation. The first rows are the airports where these flights are cancelled most above expectation and the last rows the ones where they are cancelled least.

Only flights between two covered airports have weather at both ends, so airlines that mostly fly to smaller airports are under-represented. The weather is the worst any report gave in a clock hour at the airport, so a flight scheduled early in an hour is matched to reports made later in it. The scheduled arrival hour is the hour the flight was due and a delayed flight reaches the airport later, perhaps in better or worse weather. The data has no announcement time, so it cannot say whether a cancellation was decided before the weather arrived, and a flight cancelled because of weather somewhere else on its route is matched only to its two ends. A flight with marginal weather at either end is left out of the four groups.