Integration of max-stable processes and Bayesian model averaging to predict extreme climatic events in multi-model ensembles
- 65 Downloads
Projections of changes in extreme climate are sometimes predicted by using multi-model ensemble methods such as Bayesian model averaging (BMA) embedded with the generalized extreme value (GEV) distribution. BMA is a popular method for combining the forecasts of individual simulation models by weighted averaging and characterizing the uncertainty induced by simulating the model structure. This method is referred to as the GEV–embedded BMA. It is, however, based on a point-wise analysis of extreme events, which means it overlooks the spatial dependency between nearby grid cells. Instead of a point-wise model, a spatial extreme model such as the max-stable process (MSP) is often employed to improve precision by considering spatial dependency. We propose an approach that integrates the MSP into BMA, which is referred to as the MSP–BMA herein. The superiority of the proposed method over the GEV–embedded BMA is demonstrated by using extreme rainfall intensity data on the Korean peninsula from Coupled Model Intercomparison Project Phase 5 (CMIP5) multi-models. The reanalysis data called Asian precipitation highly-resolved observational data integration towards evaluation, v1101 and 17 CMIP5 models are examined for 10 grid boxes in Korea. In this example, the MSP–BMA achieves a variance reduction over the GEV–embedded BMA. The bias inflation by MSP–BMA over the GEV–embedded BMA is also discussed. A by-product technical advantage of the MSP–BMA is that tedious ‘regridding’ is not required before and after the analysis while it should be done for the GEV–embedded BMA.
KeywordsBias-variance trade-off Bootstrap Composite likelihood L-moments Spatial extremes Variance estimation
The authors would like to thank two reviewers and Associate Editor for their valuable comments and constructive suggestions. This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (No. 2016R1A2B4014518), and also funded by the Korea Meteorological Administration Research and Development Program under Grant KMI2018-03414. Lee’s work was supported by the Basic Science Research Program through the NRF Funded by the Ministry of Education (2017R1A6A3A11032852).
- Casey T (1995) Optimal linear combination of seasonal forecasts. Aust Meteor Mag 44:219–224Google Scholar
- Cooley D, Cisewski J, Erhardt RJ, Jeon S et al (2012) A survey of spatial extremes: measuring spatial dependence and modeling spatial effects. Revstat 10(1):135–165Google Scholar
- Cressie N, Wikle C (2011) Statistics for spatio-temporal data. Wiley, HobokenGoogle Scholar
- Diggle PJ, Ribeiro PJ Jr (2007) Model-based geostatistics. Springer, New YorkGoogle Scholar
- Hosking JRM (1990) L-moments: analysis and estimation of distributions using linear combinations of order statistics. J R Stat Soc Ser B 52:105–124Google Scholar
- Nelsen RB (2006) An introduction to copulas, 2nd edn. Springer, New YorkGoogle Scholar
- Ribatet M (2013) Spatial extremes: max-stable processes at work. J Fr Stat Soc 154(2):156–177Google Scholar
- Ribatet M, Singleton R, Team RC (2013) SpatialExtremes: Modelling Spatial Extremes. R package version 2.0-1. http://spatialextremes.r-forge.r-project.org/docs/SpatialExtremesGuide.pdf. Accessed 10 Jan 2017
- Smith RL (1990) Max-stable processes and spatial extremes. unpublished manuscript. http://www.stat.unc.edu/postscript/rs/spatex.pdf. Accessed 5 Oct 2010