A general regional frequency analysis framework for quantifying local-scale climate effects: A case study of ENSO effects on Southeast Queensland rainfall

- Development of a rigorous spatio-temporal regional frequency analysis framework.
- Enables verification of asymmetric ENSO impact on Queensland precipitation using multi-site local rainfall data.
- Enables predictions of multi-site local rainfall extremes conditional on ENSO, with uncertainties.
- During a strong La Niña, the most likely 1 in 100 year rainfall can be 20-50% higher than normal.
- Regionalizing the effect of ENSO strongly reduces uncertainty.

SummaryThere is increasing evidence that the distribution of hydrometeorological variables such as average or extreme rainfall/runoff is modulated by modes of climate variability in many regions of the world. This paper presents a general spatio-temporal regional frequency analysis framework that allows quantifying the effect of climate variability on the distribution of at-site hydrometeorological variables. Climate effects are described through the parameters of a pre-specified distribution, by using regression models linking parameter values with time-varying covariates, such as climate indices. For the regional model copulas are used to incorporate spatial dependency. A Bayesian framework is used for inference and prediction, which enables quantification of parameter and predictive uncertainties. A regional approach enables better identification of climate effects which can be subject to high uncertainty using only at-site (local) analysis. Lastly, model comparison tools enable considering competing statistical hypotheses on the nature of climate effects and selecting the most relevant one.This modelling framework is applied to two case studies assessing the effect of El Niño Southern Oscillation (ENSO) on summer rainfall in Southeast Queensland. The first case study focuses on summer rainfall totals while the second analysis focuses on extremes using summer daily rainfall maxima. The Southern Oscillation Index (SOI), a measure of ENSO, is considered as a time-varying covariate. In order to account for different effects during La Niña and El Niño episodes, an asymmetric piecewise-linear regression is used to analyse the rainfall data using both local and regional models. During La Niña episodes, SOI has a significant effect on both summer rainfall totals and maxima. Conversely, during El Niño episodes, the SOI has little effect on rainfall. It is found that, during a strong La Niña, the most likely 1 in 100 year summer maximum daily rainfall for different sites estimated with the local asymmetric model can be 5-33% higher than the estimates from a local symmetric linear model and 20-50% higher than the estimates from a stationary model, albeit with significant uncertainty. Results from regional and local models are also compared: the former shows a great advantage in terms of uncertainty reduction and allows a better quantification of the ENSO effect on summer rainfall totals and maxima.