I've seen the number of induction coils increase drastically over the years in The Netherlands. They are quite easily recognizable, as there's typically a solar panel on a pole, that appears to provide power to the system.
Considering there's a wide variety of vehicles on the road, surely each variant must have (slightly) different characteristics when passing over the coil; especially when in a specific place at a specific time.
And given that they are situated at highway exits (see picture) or after entrances, and road users unable to exit and enter elsewhere, it would be trivial to track the bulk of a vehicle's trip.
This in context of ALPRs (in different forms) being in place at strategical locations (large junctions or at bridges or tunnels, and parking), and the address of vehicle's owner; you'd be able to connect the dots, and end up with a pretty complete picture.
I imagine that this is for traffic statistics and modeling, and isn't able to distinguish between different models of car
It's undoubtedly used for that too, but similar to cameras supposedly limited to traffic analysis, this system could also double as surveillance infrastructure. The readings might vary per vehicle passing over the coils: different concentrations of metal at different positions, the duration of a vehicle within range (which could be an indicator for length, speed, or both), etc. Measure these characteristics at many points along the way, while reading vehicle license plates at strategic locations, and it seems you'd have a pretty robust surveillance concept.
not very well though, while the raw data can vary car to car, they only use a simple vehicle detection count, and they could technically distinguish between large trucks, trailers, busses and family cars, but not very well and it would be based on signal duration so the traffic flow speed would have to be known too. A small electric car might have a similar signature to a large diesel truck just based on the inductive mass, and a hatchback could have the same signal duration as a bus if it's in traffic.
There's no useful data there for deeper analysis or correlation other than "hmm, road use is up in this area, increase road maintenance there in the next budget" or, "hmm, that intersection upgrade downstream has caused more traffic than predicted upstream", or "hmm, the timing on these fixed intersection lights need to be re-assessed or upgraded to automatic lights"
If they wanted to track car movements, they'd just use ALPR like everyone else
ALPRs are controversial, and governments know it. This concept would however be abstract enough for most (even around here it seems) , while effectively accomplishing the same.