Probabilistic Prediction of Precipitation Deficits using Large-scale Circulation Indices as Precursors - The Role of The Pacific Decadal Oscillation

Climate variability patterns are the most common cause for atmospheric droughts. Precipitation in the southern US is influenced by two large scale circulation patterns: the El Niño Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO). Positive (negative) precipitation anomalies are generally associated with El Niños (La Niñas). The recent record drought in 2011 for example was strongly influenced by La Niña conditions. Moreover, PDO, when in phase with ENSO, tend to enhance the effect of El Niños or La Niñas. We hypothesize that the state of ENSO and PDO can be used to conditionally predict precipitation anomalies. We use copulas to (1) investigate the relationship between precipitation anomalies and ENSO and PDO and (2) use the information to predict precipitation anomalies based on the projected state of these climate patterns. Copulas offer a viable alternative for this study as marginals from different families can be combined into joint distributions. We first use bivariate copula to model the relationship between ENSO and precipitation anomalies and then extend to include PDO in a trivariate model. We choose from ten different copulas, belonging to the elliptical and Archimedian families, and found that the copula modeling the dependence structure changes across climate regions. The prediction capability of the bivariate model is compared to that of the trivariate model to assess if PDO can hone prediction capabilities. The proposed methodology is applied to the state of Texas, which has ten climate divisions exhibiting varying climatic conditions. The outcome of the study can provide advanced warning on the expected state of precipitation, based on projected ENSO and PDO conditions. Such warning may help trigger drought management plans in the state. The methodology developed in this research can be easily extended and applied to different regions.