The accuracy of something like flighradar24 at low altitudes (where the difference between a landing and a crash is a matter of feet) is not high enough, and the data reporting interval is not small enough, to be used to do more than guess what happened. The article gives 3-4 data points per minute. You can't look at that data and draw any conclusions about the motion of an aircraft.
It's fun speculating, but please remember that the data from ADS-B is not designed to reflect the aircraft's motion over the time periods that matter in a crash. We have no idea what delays are present in the system, what delays flightradar's system adds in, and what the accuracy of the system is. It's very important to understand the limitations of any sensor data you analyse. It would be good to mention that in the article.
Edit: I notice that there's a little disclaimer at the bottom of the article saying that Flightradar24/flightaware data is "not 100% accurate", which is sort of covering the issue. However, when dealing with data like this it's important to remember that there's all sorts of ways data can be "accurate" or not. The uncertainty of the data itself, the frequency of data captured, the accuracy of the timestamps attached to the data and all sorts of other variables need to be considered when you're trying to understand the picture that sensor data is painting.
I would hope the difference between a landing and a crash is a little more than 'feet' otherwise we'd have a lot more crashes. Maybe the difference between a BAD landing and a crash is only feet :)
A typical airliner approach is a 3 degree angle touching down at 1000ft past the end of the runway. That means that they pass over the threshold at about 150ft. Let's say there's another 1000ft past that before things get problematic (at SFO it's considerably less, I believe), that means that the airliner passes over the dangerous bit at about 300ft when operating normally, and 300ft lower would cause a crash.
This isn't normally a problem, because detecting and ensuring that 300ft difference is not really hard.
It's fun speculating, but please remember that the data from ADS-B is not designed to reflect the aircraft's motion over the time periods that matter in a crash. We have no idea what delays are present in the system, what delays flightradar's system adds in, and what the accuracy of the system is. It's very important to understand the limitations of any sensor data you analyse. It would be good to mention that in the article.
Edit: I notice that there's a little disclaimer at the bottom of the article saying that Flightradar24/flightaware data is "not 100% accurate", which is sort of covering the issue. However, when dealing with data like this it's important to remember that there's all sorts of ways data can be "accurate" or not. The uncertainty of the data itself, the frequency of data captured, the accuracy of the timestamps attached to the data and all sorts of other variables need to be considered when you're trying to understand the picture that sensor data is painting.