I had a quick look at the article and I think it is not a typical patch antenna. The water "patch" serves as a reflecting layer (due to high dielectric contrast between air and water) for the waves emanating from the feeding probe. So the probe is not generating conductive currents on the patch (as for the case of a standard patch). Therefore, there is no reason to make physical contact between the probe and the water. It might have impact on antenna parameters, but not on operating principle.
Why don't you use the same feeding as described in the paper?