> it's very tricky to build data visualizations that work well in both Light and Dark Mode.
I'm sure only a small percentage of people will appreciate the difference, but I am one of those people. Thank you for putting in the extra effort making it easy to digest.
Interestingly, I was struck by the colors you used and instantly checked your code to find the palette. Your use of scale_fill_hue is beyond anything I’ve seen before. Where can I learn more?
One contributing factor is that the jetstream is stronger in winter months than summer months. Flights from BOS to SFO are up to 1 hour longer (heading into the jetstream) and 1 hour shorter from SFO to BOS (strong tailwinds).
In winter months, some airline's direct flights have to occasionally land and refuel mid-country[1]. I've experienced this first hand a couple times.
> Unlike the SFO → SEA charts, both charts are relatively flat over the years. However, when looking at seasonality, SFO → JFK dips in the summer and spikes during winter, while JFK → SFO does the complete opposite: dips during the winter and spikes during the summer, which is similar to the SFO → SEA route. I don’t have any guesses what would cause that behavior.
This one seems easy:
SF “summer” is actually foggy winter, whereas the winter months are mostly pretty mild. (Admittedly less so at SFO than in the city but it’s often still foggy.)
Airports tend to experience delays in wintery (wind and low visibility) conditions.
Thus the seasonal delay is from arriving at the airport experiencing “winter conditions.”
Could be, though with improved short-range forecasting and more sophisticated planning, bad weather at the destination increasingly manifests itself as departure delays.
Interestingly, and a tie between the another comment and this. The jet streams, thus winds aloft, I believe are stronger in the winter and more southerly.
SFO is even worse during the fog -- if there is insufficient visibility, they have to close one of the two runways because they're too close to use both using instruments.
It’s actually a bit more wild than that! SFO is always too close to land both runways at once via IFR alone. The second tail plane is required to land while maintaining separation visually, which can’t be done in bad weather. Also, since wakes are what they are, the second plane has to be the heavier one. Which is how you get extremely large planes landing at a major international airport under requirements in part for VFR.
The difference between eastbound and westbound speeds, and their inverse correlation, is to be expected, given that winds aloft are predominantly from the west. Just from eyeballing the speed charts, there seems to be about a 70mph difference on average, corresponding to average westerly winds of half that.
The graphs talking about flight speed are pretty nonsensical. The distance is fixed, and they already graphed time, so the only difference here that wouldn't be measured on the previous graph is route differences (assuming they're actually using flown distance and not just straight line distance).
The comment: "he metrics from JFK indicate there’s about a 20% drop in flight speed potentially due to wind resistance, which makes sense. " Also shows a lot of ignorance. Air speed doesn't change based on direction, but flying opposite directions sitting in the jet stream definitely does.
I'd like to see a version of this with data going back to the 1990's.
My memory is that before 9/11 and the aviation industry meltdown that planes used to burn more fuel to get from coast to coast quicker than they do now. But I'd like to see some actual numbers to determine if I remember that correctly.
The JFK → SFO data viz seems to track close enough with the average in that second chart (it's lower, likely due to less outlier skew). The SEA → SFO metric is the reverse of the displayed data visualization so that's not applicable.
This is really interesting, I wonder what caused the spike in the summer of 2017? I don't remember it being especially hot, any instance a strike, or anything like that.
I had used the ColorBrewer diverging palettes for the rainbow box plot color schemes, but they're too light and have low contrast when inverted.