Incorporating long-term climate information into stochastic models of annual hydrological data: A Bayesian hierarchical approach

Abstract

The impact of climate variability on water supply drought security is currently the subject of considerable uncertainty. Water resource planners feel ill-equipped to provide reliable estimates of future drought security due to the shortcomings of the current suite of stochastic models. For example, there is considerable evidence that hydrological data are dynamically modulated by long-term climatic processes such as the Inter-decadal Pacific Oscillation (IPO) and El Niño Southern Oscillation (ENSO). The traditional stochastic models used to simulate annual hydrological data do not have a mechanism to emulate this long-term climate variability. A Bayesian hierarchical framework that can utilise long-term climate indices, such as the IPO, to inform a stochastic model of annual hydrological data is presented. The top level of the hierarchy uses the climatic indices to inform a stochastic model that represents the long-term variability. This is combined with the hydrological data at the lower level in the hierarchy to ensure the short-term variability is captured. The jointly calibrated hierarchical model displayed some evidence of multi-decadal persistence in the upper level. However, tests of the simulation of rainfall data revealed some components of the model require refinement. Future research endeavours aim to extend this simple model and include more data sources such as paleo data to further improve simulations.