Statistically approximating or "emulating" time series model output in parameter space is a common problem in climate science and other fields. There are many packages for spatio-temporal modeling. However, they often lack focus on time series, and exhibit statistical complexity. Here, we present the R package stilt designed for simplified AR(1) time series Gaussian process emulation, and provide examples relevant to climate modelling. Notably absent is Markov chain Monte Carlo estimation - a challenging concept to many scientists. We keep the number of user choices to a minimum. Hence, the package can be useful pedagogically, while still applicable to real life emulation problems. We provide functions for emulator cross-validation, empirical coverage, prediction, as well as response surface plotting. While the examples focus on climate model emulation, the emulator is general and can be also used for kriging spatio-temporal data.
Bibliographical noteFunding Information:
For their roles in producing, coordinating, and making available the CMIP5 model output, we acknowledge the climate modeling groups, the World Climate Research Programme's (WCRP) Working Group on Coupled modeling (WGCM), and the Global Organization for Earth System Science Portals (GO-ESSP). We thank Jong-Soo Shin for help with extracting Korean temperature output, and Patrick Applegate for sharing the SICOPOLIS ice sheet model output. We acknowledge financial support from National Research Foundation of Korea (NRF-2009-0093069, NRF-2018R1A5A1024958), and from the Institute for Basic Science (project code IBS-R028-D1). This work was also co-supported by the National Science Foundation through the Network for Sustainable Climate Risk Management (SCRiM) under NSF cooperative agreement GEO-1240507 and the Penn State Center for Climate Risk Management. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation, or any other foundation or entity
© 2018 The R Journal.
All Science Journal Classification (ASJC) codes
- Statistics and Probability
- Numerical Analysis
- Statistics, Probability and Uncertainty