| A multivariate drought index (MDI) has been developed, with formulation and evaluation occurring within three diverse climate divisions in California. The MDI comprehensively considers all physical forms of drought—meteorologic, hydrologic, and agricultural—through selection of variables that are hydrologically-related and important to each drought type. Hydroclimatic monthly data for each climate division underwent principal components analysis (PCA), and retained components were eigenvalue-weighted to arrive at a single MDI value for each month. The MDI formulation considered numerous possibilities in its approach, including the type of PCA (covariance- or correlation-based), the number of components to retain (Kaiser or Jolliffe truncation rules), the sort of rotation to employ (varimax or no rotation), the type of data transformation to perform (Box-Cox power transformation or no transformation), and whether sparsely available observational data for pan evaporation were useful. Evaluation among these choices was accomplished through comparisons with the Palmer Drought Severity Index (PDSI) and the Vegetation Condition Index (VCI), and assessing the performance of the MDI formulations in resolving the 1977 and 1987–92 droughts. The superior MDI formulation was determined to utilize correlation-based PCA, with the Jolliffe retention rule; this formulation benefited from no rotation of components, no distributional transformations, and the inclusion of pan evaporation data. |
