Discussion on "A combined estimate of global temperature"

While the real-world has warmed in one unique way, the available data, which is spatio-temporally incomplete and contains biases of unknown nature and timing, means this quantity can only ever be estimated. Craigmile and Guttorp (2022) propose an approach that optimally combines the range of existing products to gain a refined estimate. Implicit in this or, indeed, any similar approach, are assumptions about the available estimates representing both an unbiased and representative draw from the population of potential plausible estimates that could have been created. There are well-founded reasons to doubt that this is the case. An alternative approach, used in the IPCC report, is to employ much simpler approaches which, in particular, hedge against underestimation of the true uncertainty in the evolution of global surface temperatures, and thus serve to future-proof current estimates against subsequent dataset innovations arising from future improvements in our understanding.