An overview of streamflow prediction using random forest algorithm

Since the first application of Artificial Intelligence in the field of hydrology, there has been a great deal of interest in exploring aspects of future enhancements to hydrology. This is evidenced by the increasing number of relevant publications published. Random forests (RF) are supervised machine learning algorithms that have lately gained popularity in water resource applications. It has been used in a variety of water resource research domains, including discharge simulation. Random forest could be an alternate approach to physical and conceptual hydrological models for large-scale hazard assessment in various catchments due to its inexpensive setup and operation costs. Existing applications, however, are usually limited to the implementation of Breiman's original algorithm for extrapolation and categorization issues, even though several developments could be useful in handling a variety of practical challenges in the water sector. In this section, we introduce RF and its variants for working water scientists, as well as examine related concepts and techniques that have earned less attention from the water science and hydrologic communities. In doing so, we examine RF applications in water resources, including streamflow prediction, emphasize the capability of the original algorithm and its extensions, and identify the level of RF exploitation in a variety of applications.