I think the way to think about an antenna is that it transforms electromagnetic wave power into electric (or EM) wave power on a conductor(s) which we want to flow between a port(s) connected to our circuits.
The EM wave in freespace can be described equivalently as a power density (dBm/m^2), electric field density or magnetic field density. They are all directly related through the freespace impedance, so any of these is valid (Thevenin and Norton work for freespace too).
The important thing to remember when modeling antennas is that there is an lossless impedance transformation from freespace to our electrical port that is related to how well the antenna is coupling to the EM wave (radiation resistance). The radiation resistance looks like an internal source impedance in series (Thevenin) with any resistive losses in the antenna, but the radiation resistance is not a thermal resistance (no noise contribution if you are worried about modeling antenna SNR). The antenna resistive losses do add kT thermal noise.
An ideal antenna would have 100% effeciency in converting the EM wave to the electrical wave measured at the port, so there would be just a radiation resistance in series with the source (or shunt for Norton). If there are losses, then those need to be added.
Here is a link to page that has a good description of the model for LF applications. It has a good SPICE model showing all the different parts of the model. I have tried it out and it seems to work well.
http://sidstation.lionelloudet.homedns.org/antenna-theory-en.xhtmlNotice that they have a conversion between electric and magnetic field strength (since they are considering the loop antenna to convert a B field into an EMF at the antenna port). But you could just as easily plug in a power density and calculate the E or B field.
I hope others have comments too, I'm trying to understand this better myself.