I have some papers that you can look at, and we have evaluated TCP performance and UDP) using empirical data for 802.11. However we did not look at SNR.
A problem that you will find is that the S/N can change very quickly, you can find that the value at the beginning of the connection is one and at the end of the connection is other, you need a more stable metric, for example the mean S/N in a period. In other hand, it is the routing protocol how usually handles this type of information with the objective of to find a stable route. This is important with TCP that the route is stable, if the route change many times, there is a high probability that the packets could arrive out of order and in this case the TPC protocol performance drop.
The transmission will broken, if the SNR decreased and sometimes goes below the threshold. When the SNR goes back manly above the threshold, the transmission will resume itself. The modified layer should be somewhere between network and transport layers. probably, it must have a control over the higher and the lower threshold.