This is lifted from another thread with minimal editing, so it'll look familiar to some of you!
For this you need an accurate map. Now, right away, I know you'll be thinking "but the maps will have been doctored to give you the results you find". Well, if they were, this method would actually show it up with inconsistent results, but it doesn't, so you can put that thought out of your mind.
Equipment: 1 spirit level, 1 set square, and 1 ruler. You might also want something to set the spirit level up on, and some packing to level it.
Method: pick a mountain that is easily identifiable from a distance (I have a few to pick from where I am). Find a point on the map 30km away, 40km away, and 50km away (or any other largish distances, as long as you can measure them accurately. This is part of what negates any doctoring of the maps: picking arbitrary points to measure from). Set up your spirit level so that it points toward the mountain you picked, ensuring it is level. Set your ruler at a specific distance along the top of the level, standing up at right angles to it (that's what the set square is for). Sight from the edge of the level furthest from the mountain to the mountain top, and note where your line of sight crosses the ruler. The ratio of the measurement on the ruler to it's distance along the top of the level would be equal to the ratio of the mountain's height above your observation point to your distance to it (simple triangular geometry).
The height of the location the measurement is taken from is important, as you will get different results from different elevations at a given distance. Another reason why this cannot be faked or doctored.
In the below diagram, the ratio of h/d should be equal to r/s if the earth is flat, and should always be that way, no matter how far away from the mountain you are.
When I did this experiment though, there was a difference between h/d and r/s, and it got bigger and bigger the further away I got from the mountain. I did some extra maths to work out where the mountain top would be for a round earth:
apparent height on a round earth = square root of ((earth's radius + h)
2 - d
2) - (earth's radius + altitude of your position)
When I tested the results against those calculations, they matched perfectly. Remember: maths and geometry don't care what shape the earth is; they will always tell you the truth.
Results:
Mountain (Castle Top, in the Nandewar Ranges): 1075m AMSL
Point 1 (my back yard): 30.44km from peak, 215m AMSL
Point 2 (Kamilaroi Hwy): 40.46km from peak, 205m ASMSL
Point 3 (Wee Waa levee): 55.57km from peak, 192m AMSL
Using a 1100mm baseline on the spirit level, the following apparent heights were recorded:
Point 1: 28mm
Point 2: 18mm
Point 3: 11mm
Let's be pessimistic and say they're only within 1mm either way.
Here's what they should have been for a flat earth:
Point 1: 31mm
Point 2: 21mm
Point 3: 16mm
Note the increasing difference. This is due to the mountain 'going over the edge'.
Conclusion: The data collected disagrees with FE predictions*, and almost exactly matches RE predictions. The experiment is easy to conduct, and not prone to significant error, so the results can be considered reliable.
* Rowbotham's 'perspective' has been cited as a possible cause of the results. However, a passing familiarity with geometry should be sufficient to show that this 'effect' is not responsible for the rate at which the actual measurements diverge from what they should be for a flat earth.
Interesting Note: if you use the following equation, you can work out the difference between the apparent height for a round earth and a flat earth (with respect to the diagram):
rf - rr = s * tan((d / ce) * 360 / 2)
where rf is the apparent height on a flat earth, rr is the apparent height on a round earth, and ce is the circumference of the round earth. d and ce need to be in the same units as each other, and rf - rr will be in the same units as s.
Sorry for the long post! I thought it would be best to give plenty of detail, so that others can reproduce this experiment and obtain accurate results.