Abstract
Abstract. In this paper we discuss climate model tuning and present an iterative automatic tuning method from the statistical science literature. The method, which we refer to here as iterative refocussing (though also known as history matching), avoids many of the common pitfalls of automatic tuning procedures that are based on optimisation of a cost function, principally the over-tuning of a climate model due to using only partial observations. This avoidance comes by seeking to rule out parameter choices that we are confident could not reproduce the observations, rather than seeking the model that is closest to them (a procedure that risks over-tuning). We comment on the state of climate model tuning and illustrate our approach through three waves of iterative refocussing of the NEMO (Nucleus for European Modelling of the Ocean) ORCA2 global ocean model run at 2° resolution. We show how at certain depths the anomalies of global mean temperature and salinity in a standard configuration of the model exceeds 10 standard deviations away from observations and show the extent to which this can be alleviated by iterative refocussing without compromising model performance spatially. We show how model improvements can be achieved by simultaneously perturbing multiple parameters, and illustrate the potential of using low-resolution ensembles to tune NEMO ORCA configurations at higher resolutions.
Highlights
The development of ocean, atmosphere, and coupled climate models represents a huge scientific undertaking that is happening simultaneously and relatively separately throughout the world’s modelling centres and within the many universities that collaborate with them
We have described and illustrated iterative refocussing for the ocean model NEMO run at 2◦ resolution and argued for the method to be used for tuning complex numerical models of the ocean, atmosphere, and climate
−1.0 −0.5 0.0 0.5 1.0 (a) rn_lc (b) climate models to data to define a region of model parameter space, which is consistent with the data we are using and allows us to focus the search for good models only in those subspaces
Summary
The development of ocean, atmosphere, and coupled climate models represents a huge scientific undertaking that is happening simultaneously and relatively separately throughout the world’s modelling centres and within the many universities that collaborate with them. This parametric uncertainty can be pertinent for studies of complex problems such as the stability of the Atlantic meridional overturning circulation (MOC) (Williamson et al, 2013) and, it should be quantified and, at least representative models, reported as the final step in a tuning exercise Even tuning methods such as Bayesian calibration (Kennedy and O’Hagan, 2001; Rougier, 2007; Sexton et al, 2011), which explicitly quantify parametric uncertainty in the form of a probability distribution for model parameters, can be described as forms of optimisation (they assume the existence of a single optimum or “best input” and perform inference for it) and suffer from some of the drawbacks stated above. We present a comparison of a model representative of the “tuned” parameter space with both observations and the numerical model’s standard configuration (Sect. 5), and conclude with a comment on and an example of the application of iterative refocussing to high-resolution models (Sect. 6) and a discussion (Sect. 7)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
