Minimum streamflow regionalization in a Brazilian watershed under different clustering approaches

Area of Science:

• Hydrology
• Water Resource Management
• Environmental Science

Background:

• Estimating minimum streamflows is crucial for water resource management.
• Gauged data is insufficient in Brazil, necessitating hydrological regionalization.
• The study focuses on the minimum streamflow equaled or exceeded 90% of the time (Q90).

Purpose of the Study:

• To compare hierarchical and non-hierarchical clustering for delimiting hydrologically homogeneous regions in Rio Grande do Sul, Brazil.
• To regionalize Q90 using multivariate statistics and regression analysis.
• To assess the effectiveness of morphoclimatic attributes in regionalization.

Main Methods:

• Applied multivariate statistics and regression analysis for Q90 regionalization.
• Utilized morphoclimatic attributes from 100 southern Brazilian watersheds as independent variables.
• Compared various hierarchical (e.g., Ward algorithm) and non-hierarchical clustering techniques.

Main Results:

• Clustering techniques successfully defined hydrologically homogeneous regions for Q90 estimation.
• The Ward algorithm with Manhattan distance showed strong potential.
• Drainage area, perimeter, centroids, and mean annual rainfall were key variables for accurate clustering.
• Developed mathematical models demonstrated excellent performance for estimating Q90.

Conclusions:

• Hydrological regionalization using clustering is a viable method for estimating Q90 in ungauged watersheds.
• The identified homogeneous regions and predictive models can support water resource planning in Rio Grande do Sul.
• Morphoclimatic data significantly enhances the accuracy of regionalization models.