How we evaluate models matters. Benchmarking 13 models showed that inferred model skill is more sensitive to how many observations are used and over what time period than to which observational dataset is used. agupubs.onlinelibrary.wiley.com/doi/10.1029/...
