Affinity Propagation Approach for Catchment Classification Applied to
Arid Catchments
Abstract
One of the major issues in the arid region is the availability of
hydrological data for hydrological studies of the basins for water
resources projects. Since the Kingdom of Saudi Arabia (KSA) is a huge
country and contains many arid basins it is awfully expensive and
time-consuming to make hydrological networks for studying all these
basins. Therefore, the Affinity Propagation (AP) clustering technique is
proposed to cluster basins into groups that are similar in
morphological, hydrological, and landcover characteristics and defining
an exemplar (a representative basin) to each group. This basin is
utilized for the installation of a detailed hydrological network. The
hydrological response of that basin can be transferred and scaled
appropriately to other basins in the cluster since they are
hydrologically and morphologically similar. A pilot study is performed
on 18 sub-basins in the southwestern part of KSA. GIS software is used
to extract basin attributes and the clustering process is performed
using the AP cluster packages in R software. The results show that four
clusters are obtained based on the morphological attributes
(twenty-eight attributes), five clusters based on hydrological
attributes (twelve attributes), and three clusters based on land cover
and CN (three kinds of landcover as attributes). The AP clustering
technique was evaluated by the construction of a correlation matrix that
shows a high correlation of 0.817 to 0.999. This study provides a robust
technique that is effective and efficient to identify the similarity of
catchments and can help hydrologists to develop a catchment management
application in arid regions.