The Longley-Rice model is a fairly common technique. However it's not a simple formula, and so requires some kind of software to implement. Unfortunately my favorite such tool, CRC CovWeb, has gone defunct.
For a more simplistic prediction, you can try making a simple link budget based on the Friis transmission equation, or if you want to get a little more sophisticated, the two-ray model.
You should additionally consider the radio horizon, and you may perform some simple math to determine how far your antenna can "see" based on an idealized spherical Earth. Or you might use some source of topographic information like heywhatsthat.
Your estimated coverage area is then all the areas that are close enough that free space path loss hasn't attenuated the signal beyond usability, and where terrain isn't in the way.
You will typically find even for modest powers around 5W, terrain and antenna height are most often the limiting factor. Higher powers and antenna gains are typically only useful if the station can be well placed, such as on a large tower, building, or mountaintop.