Categorizing aircraft more effectively
Methodology update
Hello readers! I promised to update here if we changed our methodology significantly, and I wanted to come here and talk about some changes we’re making in how we categorize aircraft.
Like our methodology page points out, today we pull data in from ADSB networks, from government aircraft registries, and other public sources. In a few examples recently, we’ve been tracking aircraft that we assumed were clean powered that were not.
Here’s what happened:
Some of our flight data providers share with us the ICAO type for the aircraft. In most cases, we ignore it, as we have the aircraft type from a registry, for instance the FAA or Transport Canada or France’s DGAC. In other cases, though, a flight data provider might be telling us about an aircraft we don’t have from an official registry.
When this happens, we take the flight data network at its word. In most cases, this is fine - there are plenty of flights globablly we would not be able to report on if we didn’t do this.
However, in some cases the flight data providers are wrong about a particular airplane.
I am also running down some issues where I suspect a government registry itself is wrong - so, flight data providers may just be encountering this same issue that we experience directly. At least one aircraft that the FAA lists as a Pipistrel Velis Electro, and lists the engine type as electric, we regularly observe to fly hundreds of kilometers.
Given that this is true, we might be over-focusing above on “the flight data provider had an incorrect ICAO type” - it might be that their own integrations with government registries led them astray (just like this happens with us). The aircraft we’ve spotted so far did come in from flight data providers, though, so we can’t confirm yet either way.
I’m not sure how this happens - my assumption is that the owner mis-identified the aircraft when they registered it, some kind of paperwork issue - but I’m still chasing this down.
The impact?
We were showing BRM Aero Bristell’s (gas powered) in Europe as though they were H55 powered Bristell Energic, in Spain and in Germany. Now, H55 is doing some amazing electrified work in the USA and elsewhere with their HB-SXD and HB-SXC, and we’re still showing those flights - but we’ve removed the gas burning flights that we miscategorized in Spain and in Germany.
A similar situation led to us listing a ZeroAvia owned Dornier DO-228 as a Hydrogen aircraft. I do believe this aircraft will be converted, but we' listed it as a Hydrogen powered aircraft before that conversion took place. We’ve removed these flights, which were between ZeroAvia’s home operating base in Cotswold Airport and La Roche sur-Yon, Lelystad, and Wolfsburg.
We’ve now removed the non-electrified flights above.
I want to point out that this will happen when depending on public data; however, I’m thinking hard about how to improve a) detection of such issues, and b) prevention of them in the first place. If you have thoughts on how we might do better, please share! kyle@skyzero.io
Update: we now maintain a list of aircraft we know are not electric.
We have confirmed that both flight data networks and registries sometimes feed us bad data and we’ve updated our logic accordingly. If we see registries or flight data providers telling us that an aircraft is electric when we know its not, we can now add it to this list, which will prevent us from showing it. If the registry changes this registration in future we will look at the aircraft again.