Regional frequency analysis for mapping drought events in north-central Chile

Abstract

Summary Droughts are among the most important natural disasters, particularly in the arid and semiarid regions of the world. Proper management of droughts requires knowledge of the expected frequency of specific low magnitude precipitation totals for a variety of durations. Probabilistic approaches have often been used to estimate the average recurrence period of a given drought event. However, probabilistic model fitting by conventional methods, such as product moment or maximum likelihood in areas with low availability of long records often produces highly unreliable estimates. Recognizing the need for adequate estimates of return periods of severe droughts in the arid and semiarid region of Chile, a regional frequency analysis method based on L-moments (RFA-LM) was used for estimating and mapping drought frequency. Some adaptations to the existing procedures for forming homogeneous regions were found necessary. In addition, a new 3-parameter distribution, the Gaucho, which is a special case of the 4-parameter Kappa distribution, was introduced, and the analysis procedure was improved by the developments of two new software tools named L-RAP, to perform the RFA-LM analysis, and L-MAP, to map the resulting drought maps. Eight homogeneous sub-regions were delineated using the Gaucho distribution and used to construct return period maps for drought events with 80% and 40% precipitation of the normal. The study confirms the importance of a sub-regional homogeneity test, and the usefulness of the Gaucho distribution. The RFA-LM showed that droughts with a 40% precipitation of the normal have return periods that range from 4 years at the northern arid boundary of the study area to 22 years at the southern sub-humid boundary. The results demonstrate the need for different thresholds for declaring a drought than those currently in use for drought characterization in north-central Chile.