CMStatistics 2015: Start Registration
View Submission - CMStatistics
B0488
Title: Spatial estimating equations and application to changes in climate extremes Authors:  Jun Yan - University of Connecticut (United States) [presenting]
Abstract: To detect changes in climate extremes, no fully satisfactory analog of the widely used optimal fingerprinting method for mean climate states has been available. The state-of-the-art method incorporates the signals into the location parameters of generalized extreme value (GEV) distributions. The existing profile method is computing intensive combined with a bootstrap procedure for inferences, which is prohibitive for multiple signals. Further, it discards spatial dependence which may lead to low efficiency in estimation and, hence, low power in detection. We propose a combined score equation (CSE) method that combines the score equations of the GEV model at each grid box such that an approximate correlation function of the scores is used to improve the estimation efficiency of the signal effect. Unlike the pairwise likelihood (PL) method assuming max-stable processes the CSE method does not need full specification of spatial dependence. It provides a close analog to the optimal fingerprinting in detection and attribution of changes in climate extremes. The method is applied to extreme temperature in Australia under a perfect model setting and in Northern Europe with real data.