Hydro-meteorological drought risk assessment using linear and nonlinear multivariate methods

Abstract

Drought is multivariate, yet several univariate and bivariate indices have been proposed for the evaluation of drought characteristics. However, limited attention has been paid to the mechanisms and performance of these drought indices. This study considered canonical correlation analysis (CCA), principal component analysis (PCA), and copula-based method to construct three composite hydro-meteorological indices, namely, JDHMI-CCA, JDHMI-PCA, and JDHMI-Copula. The main elements of these indices are rainfall and runoff for which historical data for the period of 1986–2016 at a 12-month time scale were collected for Kol-Mehran and Bandar-Sedij basins, Iran. Results demonstrated that the pattern of composite indices reflected the comprehensive moisture status well and was not affected by a single element. Analysis indicated that although the composite, univariate, and multivariate indices followed similar patterns in various parts but the behavior of runoff was the main source of inconsistency in the study region. While PCA and CCA ignored or assigned smaller weights to runoff, the copula method used the information of runoff in constructing the copula-based index. This study investigated the mechanisms of linear and non-linear methods in assessing the drought condition. The methods had a significant effect on the construction of composite indices which can provide useful information for drought monitoring.