Microwave radar testing of tree trunks is one of the ways for the trunk interior evaluation. The interpretation of the radar images can be a very complex task -among others due to the roughness of the tree bark. This paper studies the influence of a surface roughness on radar data of observed cylindrical objects, trees in particular. During our study, we did numerical simulations and laboratory measurements to compare radar data obtained by testing a cylinder with a smooth and an irregular surface. Then, several real trees with different surfaces and internal structures were tested to validate our findings. Those experiments indicate that the presence of a rough and irregular bark can significantly inhibit our ability to study the internal structure of the tree with the radar. On the other hand, if the bark is smooth, it is possible to infer the internal composition of the tree even for highly heterogeneous specimens.