Explore a pragmatically ambitious approach to multiscale global temperature reconstruction in this 41-minute conference talk by Finn Lindgren from Edinburgh. Delve into the complexities of uncertainty quantification and Gaussian Process emulation as applied to climate modeling. Learn about various data sources, challenges with data scarcity, and the intricacies of heat equation models and satellite observations. Examine model variability across different scales, non-Gaussian modeling techniques, and bias correction methods. Discover how daily temperature ranges are incorporated and various approximation techniques, including linear approximations and local approximate solves. Understand the application of Takahashi recursion, multiscale complements, and block update methods in improving model accuracy. Analyze error sources and the use of ensemble outputs to enhance the robustness of global temperature reconstructions.