Characterization of links between hydro‐climate indices and long‐term precipitation in Brazil using correlation analysis

Abstract

Some teleconnection patterns characterized through various hydro‐climate indices (HCIs) have been shown to influence regional rainfall regimes over a range of time scales in Brazil. However, our current knowledge of how various HCI–rainfall relationships vary across regions in Brazil is still incomplete. In this study, simple correlation analysis is performed using sliding window sizes and lag times on the order of months to years to substantially reduce the effects of high‐frequency variability inherent in HCI and precipitation time series and reveal any lower‐frequency relationships that may exist. This analysis is applied to monthly rainfall and HCI data during the time‐span of 1961–2015. The strength and significance of each correlation were tested. HCIs considered in this analysis include those characterizing the El Niño–Southern Oscillation (ENSO), those characterizing the Madden–Julian Oscillation (MJO), as well as the Tropical Southern Atlantic index (TSA), the Caribbean index (CAR), and several others. A cluster analysis based on the k ‐medoids algorithm was also applied to explore the relationship between local factors and the distribution of HCIs exhibiting maximum correlation at each site. A 60‐month sliding window and a 12‐month lag time were found to optimize the results of the correlation. The MJO was found to have a strong and significant link to annual rainfall throughout much of eastern Brazil, while other HCIs showing strong influence in other regions include ENSO, the CAR, and the TSA. The cluster analysis revealed a more spatially homogeneous response over the Amazon River basin and the southern regions than other regions of Brazil with complex topography, where the orographic effect attenuates the influence of the main atmospheric mechanisms at a local scale. In general, the rainfall–HCI relationship could be exploited for some operational applications (e.g., a regional drought early warning system), where the strongest correlations were observed.

Bibliographic Information