Wireless underground communications is mainly characterized by the effects of soil moisture on antenna return loss and bandwidth as well as path loss. In this paper, the impacts of soil moisture, especially on underground channel capacity, are analyzed for underground wireless communications. It is shown that for a given antenna and soil moisture level, there exits an optimal operation frequency that maximizes channel capacity. While existing research on wireless underground communication is focused on fixed-frequency systems, this paper motivates the use of cognitive radio systems, which can adjust operation frequency in a wide range, for efficiency for wireless underground communication. Moreover, it is shown that soil type significantly affects the channel capacity and the capacity can be improved by using longer antennas that allow lower operation frequencies. However, the size of the antenna is also limited by other factors, such as device size and deployment challenges.