A review is made of the main methods for observing root systems, with emphasis on the minirhizotron method used widely, and the relative merits are discussed. It is pointed out that from the needs of weather forecasting, root system observation can be used to correct the errors in land surface models resulted from the lack of the essential parameters of land processes, to improve the accuracy of regional microclimate prediction, study airground interaction, and grasp the regularity of weather variation. The research achievements at home and abroad are reviewed, and it is pointed that while selecting observation methods, the purposes and scope of a study should be considered fully, and any method has its own limitations. Before making a decision, it is necessary to understand the applicable conditions of a method, and made comparison carefully. Maybe the optimum solution is to use more than one methods.