ABSTRACT
Water resources management aims to solve problems arising from intensive use of water. The proper management of this resource is based on understanding water availability, often using information from hydrometric stations; flow data is the most important information. The availability of information on river flows is often insufficient for all regions of interest. A technique called hydrological regionalization can be an alternative for obtaining information on streamflow. The objective of this study was to review the main regionalization techniques used, their advantages and limitations, as well as perspectives for the future. Traditional and widely used methods for forecasting hydrological variable, such as spatial proximity and multiple linear regression, were addressed, as well as new technologies, such as the geostatistical approach, techniques using volume balance in watersheds based on remote sensing products, and machine learning techniques. These techniques allow working with several physical characteristics of basins, generally ensuring better performances than the multiple linear regression. Further advancements in this area of knowledge are expected shortly, as the great potential of machine learning has been explored only to a small extent for hydrological regionalization purposes.
Key words:
reference flows; watersheds; hydrological data
HIGHLIGHTS:
Nowadays, no one can think of regionalization without GIS tools.
Traditional methods remain the most widely used in practice.
Machine learning methods have great potential in this area.
Thumbnail
