LCY8737

Having had a little too much time in recent days (weeks? months?) I played around with travel/aviation related APIs again and thought some of you may find this interesting/useful:

BA Flight Tracker
Allows you to analyse historic BA flight data (starts in Nov 2020 & up to 10 days ago) for delays, cancellations, operating aircraft & gate information. Can be useful if you want to predict what gate your flight is departing from or what the cancellation/delay chances are on a particular flight/route. Warning: There are a few smaller data consistency issues for code share flights.

BA Fleet Tracker
Checks which BA aircraft are in the air and what their position/direction is. Quite sobering to see that a large proportion of the fleet is grounded...

Let me know what you think...

notakeenflyer

Interesting stuff, thanks. I was looking to see how often A350 was scheduled for LHR-BOS.

Looks like something's telling you an incorrect aircraft type (there's a whole load more ZB** listed further down, too.


Seems you're getting the opposite problem sometimes too, when filtering on B789

But XWBA shows correctly in the list of A35K flights, too.

Happy debugging!

FamilyOf6

Great resource, thanks.

I've noticed this in FR24 too, but wondered why sometimes an aircraft is listed as "777" and sometimes a "B772" or "B77L" This can be for the same frame. Presumably you are just reporting back what is in the API, so not something you control.

LCY8737

Thank you notakeenflyer and FamilyOf6 , I did some digging in the database and it looks like you identified a pretty wide-spread issue. Some aircraft registrations are shown as up to 6 different aircraft types.

I have opened a ticket with FlightAware (they are the API provider) to see if they can do something about this. Worst case I will clean the data as it comes in (thankfully BA don’t have that many different aircraft types).

RB211

This is great - haven't had time to play with it at all but is very interesting - good job!

rb211.

No5

Probably different codings due to what sort of the aircraft itself it is, and also what sort of configuration the particular aircraft has.

PGberkshire

Awesome, great contribution

FamilyOf6

I think LCY8737 has noted this, but even the same actual aircraft reg is appearing with different codes. eg:
777 (G-VIIH) for one flight and B772 (G-VIIH) for another.
Then: 777 (G-VIIR) for one flight and B772 (G-VIIR) for another.

I have seen this in FR24 and OP is extracting from Flight Aware. What I have noticed is that sometimes, a flight might show the aircraft sub-type AND reg in advance of date of travel, then on other times just 777 with no reg. I can only assume the airline submits this info.

LCY8737

My favourite is G-YMMS: 90% of the time it showed as a B772, but the API also claimed it was an A388, B788, 777, B77L, & B789 during the last 4 months.

I have cleaned up the database now, so you shouldn't have further issues with wrong aircraft types. If I get an interesting response from FlightAware regarding the source of aircraft types, I'll update this thread.

fartoomanyusers

Some interesting infromation, thank you.

For the aircraft that are "likely on the ground" - have you considered taking the last position and saying which airport they are likely at ?

golfmad

Nice work. Have you considered showing the airborne fleet on a map?

LCY8737

Great idea, I'm now keeping the last known position. It'll take a few days until the data is updated for the whole fleet, but the first 10 registrations are already there.

The airport where the aircraft is likely on the ground is a little more tricky because technically I only know that there is no position data being received. That is mostly because the aircraft is on the ground, but it can also be that the aircraft is somewhere without receivers (high arctic or open ocean).

I'm now extracting the flight number from the call sign (where possible) and check the destination for that flight. For G-ZBKI that worked great, last seen over the Mediterranean as BAW165 operating LHR-TLV. No longer transmitting position data, so my code correctly guessed that she is on the ground in TLV.

Same logic does not work for G-ZBKP though (BAW49 LHR-SEA), last seen over Ireland & my code incorrectly predicts that she is at SEA when the real position is somewhere over the North Atlantic.

LCY8737

I'm a little hesitant doing that because that is a feature I have to pay for. I don't really mind the cost as such (a few hundred google maps api calls are not that expensive), it's more the potential for abuse that could land me with a hefty bill if the page would be loaded very frequently (e.g. by a bot or someone who thinks that is funny).

notakeenflyer

Assuming the position data has a timestamp, could you apply some logic relative to ETA of the flight? Simplifying it, in pseudo code
IF lastSeenTime > lastFlightETA
THEN position = arrivalAirport
ELSE position = "Unknown, out of coverage"

This probably opens a whole can of worms for early arriving flights. Do you extend the logic test above to say within 30 minutes of ETA, assume it's arrived. Might need to build in some geographic logic too, if it's after, or within 30 minutes of, ETA AND less than 10nm from destination, then assume it's arrived (would also fix your G-ZBKP issue above)

fartoomanyusers

It would be quite amusing if your code decides that half the fleet is on the ground in Shannon (because that was the last airport it overflew) !

Could your code take into account the last known speed to decide whether or not it was likely on the ground ?

In the "Last Known Position" section, the "Call Sign" isn't coming through as a link. Although maybe this is deliberate?

For the aircraft that are stored - maybe you could seed your data with this information ?
Covid-19 : Where will BA park their planes?