Flow signatures and catchment’s attributes for HCA clustering in a hydrologic similarity assessment (Tunisian case)

Abstract

Partitioning methods such as cluster analysis are advantageous in pooling catchments into hydrometric similar regions. They help overcome data shortage in ungauged catchments, which is a common problem in Sud Mediterranean zones. Without accurate forecasts, it is difficult to assess and manage water resources efficiently this situation won't be of any assistance to hydrology decision-makers. This paper illustrates a Tunisian application case, that aims to pool catchments with a hierarchical clustering algorithm (HCA) based on distances calculated in multidimensional physiographical and hydrometric space. The homogeneity of generated clusters is checked by the silhouette index. Then the distances efficiencies are compared. Nineteen semi-arid Tunisian catchments monitored since 1992 are studied. Twelve physiographical attributes, nine rainfall and streamflow signatures are considered in the HCA with two clusters. Correlation distance provides the most homogeneous clusters. Statistically the: percentage of area affected by anti-erosive practices, percentage of forest cover and catchment area are the most discriminating attributes. However, hydrometrical signatures appear to be irrelevant. These partitions highlight two different hydrological behaviors that must support forecasting. Results are promising in the Sud-Mediterranean case, where the shortage of hydrometrical data is an ongoing problem. They have the advantage of enabling hydrologic forecasting without requiring heavy information.