ABSTRACT
This chapter is a review of some of the statistical and machine learning methodology, principles and techniques that are used for evaluating the quality of climate models. In the course of this review, we discuss interesting technical and methodological details, and important caveats noted by different teams of authors, and also include our own remarks. The current generation of probabilistic scoring and ranking of climate models is discussed in some detail. A new approach, based on empirical likelihood, and combining some of the strengths of the probabilistic scoring as well as the ensemble-based approach, is very briefly discussed. Limitations of current approaches and some future directions of this line of work are pointed out.
