Multivariate delineation of rainfall homogeneous regions for estimating quantiles of maximum daily rainfall: A case study of northwestern Mexico

Lack of data in maximum daily rainfall frequency analysis can generate ineffcient estimates for design purposes. An approach to diminish these errors is to apply regional estimation techniques, which require that all stations be located at the same homogeneous region. In this paper, a delineation of homogeneous precipitation regions was made based on the multivariate methods of principal component analysis and hierarchical ascending clustering. A region in northwestern Mexico was selected to apply this methodology. It was concluded that only the coeffcients of variation of the L-moments, along with latitude, longitude and altitude at each climatological station are sufficient to define the homogeneous rainfall regions, and that either the inclusion or exclusion of information in the regional techniques has a direct impact on the estimation of events associated to different return periods.