So is Air Traffic decreasing in this COVID-19 world?
Background
As I write this, on March 19th 2020, we're about 3 full days into "properly" dealing with this damn COVID-19 business. By which I mean putting in place some self-isolation standards, not travelling unless we need to etc etc. And the news is full of stories of the impact of this, particularly on the travel industry and especially as it impacts Aviation.
It's also (coincidentally?) about 2 weeks since our major domestic independent airline -
flybe - went out of business (incidentally stranding me in Edinburgh but, hey, if you've got to be stranded somewhere, you can do a
lot worse!).
I wanted to test the hypothesis that if all of the above is real and active, we ought to be seeing a drop-off in air traffic.
Technology
I have a ADS-B receiver setup in my house. The summary of this is it's a radio that listens out for messages being broadcast from aircraft giving information on who they are and where they are - it's one of the more recent innovations in aviation safety and it's great. All commercial aircraft have it, and a lot of military and privately-owned aircraft do too.
My rig (a Raspberry Pi Zero and an off-the-shelf SDR-capable USB radio stick) is part of the
FlightAware global flight tracking network; all this data from feeds like mine is aggregated together and gives a pretty good global picture of what's in the sky at any given time. Since it's feeding that network, it's clearly capable of feeding other networks. So I've knocked up a bit of python to gather some summary stats on flight rates, shovel it all into an Influx time-series database, and then banged together some ugly but functional
jqPlot code to spew out graphs showing me how many aircraft I've detected, hour by hour. There's a bit of cleverness to ensure that I'm getting a count of distinct "flights" rather than "sightings" but it's not rocket science. And most importantly it's nothing that I can't already get out of FlightAware's own dashboards already but, hey, why use something that's already there if you can build it yourself eh?
Location, Location, Location
All of the above needs a little summary: We live about 20 miles from Southampton Airport which is was a major hub for Flybe (approx 80% of flights into/out of the airport). Thanks to things like curvature of the earth, I can't directly "see" aircraft at the airport, my setup picks them up as they climb above about 1,500ft altitude heading North, and a bit more heading South (because my antenna is blocked by the house to the South). We're also about 20 miles south of Farnborough which is one of the major executive jet airports for the region (and, since that region includes London, the country) so the same conditions apply.
We're also within easy detection range of the major airways and routes going into and out of London (Gatwick and Heathrow) so, again, I can't see them on the ground but I've got decent sight of most traffic going there out to about 5-10 miles from the airport. And there's a couple of decent East-West airways, and at least one North-South, from the Americas to the Continent that normally picks up a decent amount of traffic heading to and from the European Mainland.
I'm trying to say that I'm well placed to get numbers that show indicative rates of movements into these locations, but I'm not confident that I'm absolutely picking up 100% of all traffic here and there. Basically: good enough to show changes over time, is what I'm hoping for.
The Data
Here's a chart showing the hourly sum of detected flights as three overlaid lines: This week (more accurately: the last 7 days), Last week, and the Week Before Last. I'd like to go back further in time - same week last year would be great, a month ago would be better - but I only got the data gathering sorted, er, 3 weeks ago so that's all we've got. It's just a massive coincidence I started gathering data the week before all of the action started happening, but one uses what one can get. Anyway:
 |
| Aircraft Detection Rates to 19/03/2020, by week. |
Some features of interest:
- Ignore that drop-to-zero at the far right; it's an artefact of how I report on the data I haven't bothered to fix (I should "ignore last hour" as the data's incomplete or unrecorded)
- This Week (blue line) is significantly below previous weeks data, although not before about Tuesday (effects are less pronounced for Monday and back before - furthest right peak is today. Thursday, furthest left peak the tail end of last Thursday)
- It's possible that there's a drop-off in traffic associated with FlyBE going bust around the time of the arrow. Tuesday's peaks 2 weeks ago are higher than last week; Wednesday of both weeks are about the same (FlyBE ceased operations effectively overnight between Tuesday and Wednesday at that time). If that's the case, about 10%ish of flights I was detecting were FlyBE related.
Interpretation
Honestly I don't know whether I'm over-interpreting insufficient data.
I'm more confident in saying that, in the last 48-72 hours there's been a significant (up to 30%) drop off in air traffic within our area compared to the same time last week.
I've got some evidence that I was able to detect our regional airline going out of business by a 10% overnight drop-off of traffic that never recovered, but really I need more information than that.
Inspection of FlightAware's consoles - which plots "live" data like an air traffic control radar - shows that there's significantly fewer "small" aircraft (business jets, helicopters, private aircraft) than I'm used to seeing but it must be stressed that this is anecdotal observation, I don't have hard data to back that assertion up.
Further Work
What I haven't done is use the additional data available from look-up sources to categorise the data I've got. It might be useful to break down the data by aircraft type - commercial jet, business jet, helicopter, military etc - to work out which sectors (if any) are genuinely changing over time. Ah well, perhaps that's a project for another day.....
