My recommendation is to just perform Rowbotham's original experiments. They are very simple. Much simpler than the Wallace experiment where everything needs to be exactly level and aligned, and which just causes endless questions on whether the positioning and surveying methods were accurate.
The methods Rowbotham uses are based on a basic concept such as whether an object is visible in the distance or not due to the earth's curvature. Simple experiment.
But the problem with Rowbotham's experiment, as I'm sure everyody knows, is that he was too close to the water to escape the effects of refraction.
I think a slightly modified version of the Wallace method would work very well. His problem was that he didn't use enough targets. But if he'd had at least three, the results would have been much more clear.
I'm thinking with up to 10 markers it'll be even better.
I just did a quick scale model version and it seems pretty straightforward:
Statement #1: If the river is flat, the middle target (of three) can never appear higher than both the other two (at the same time).
Statement #2: If the river is curved, the three targets can never appear to be on the same plane (or level).
Tilting the camera doesn't make any difference to the result, or to the relation between the targets, it only changes where the targets appear (as a whole group) in the frame.
Raising or lowering the camera above or below the level of the targets does change the relationship between the targets, but still matches what we would expect, and is in accordance with the two statements above.
If the words aren't clear, apologies for that - probably the video of the (rough) scale model will be better; I'll let you know when it's finished uploading.
Thanks for the speedy reply. :-)