PNNL

Abstract

We propose a comprehensive solution to the problem of climate model output emulation. Our emulator, STITCHES, uses existing archives of Earth System Models’ (ESMs) scenario experiments to construct new scenarios, or enrich existing initial condition ensembles. STITCHES extends the idea of time-sampling - by which climate outcomes are stratified by the global warming level at which they occur, irrespective of the scenario and time associated to them - to the construction of a continuous Global Surface Air Temperature (GSAT) trajectory over the whole 21st century. By stitching together decade-long windows within a model simulation when GSAT has similar characteristics to the target GSAT trajectory, STITCHES creates a recipe, a series of pointers, to existing scenarios and times in the ESM archived output, and the emulator can thus recover any type of output, at any frequency and spatial scale available from the original ESM’s experiments.We show that for the variable that we expect to be smoother and less noisy than many variables commonly used for impact analysis, GSAT, the stitching, 10 in the great majority of cases, even under criteria for the creation of the recipe that are not tailored to the specific ESM emulated, does not introduce artifacts. We expect that most other surface atmospheric variables commonly used for impact analysis would be similarly not affected by the stitching procedure. Of course, a few exceptions surface when we explore the method’s performance over many CMIP6-ScenarioMIP-participating ESMs and experiments. These defective outcomes are mainly associated with a scarcity of the archived simulations from which to gather building blocks for the emulated trajectory. In the great majority of cases, however, STITCHES performance does not raise red flags according to our standards. The method therefore can be used to fill in the space between existing scenarios and to add to the size of existing initial condition ensembles. There are aspects of our emulator immediately recognizable as show-stoppers for specific applications, like when climate information is needed whose characteristics result from accumulated quantities over windows of times longer than those used as building blocks. But for many applications, we argue that a stitched product will satisfy the needs of impact researchers and thus open up the possibility of designing the next scenario experiments within CMIP7 according to new principles, relieved of the need to produce a number of similar trajectories that vary only in the strength of the radiative forcing produced.