Abstract:
|
While large climate model ensembles are invaluable tools for physically consistent climate prediction, they also present a large burden in terms of computational resources and storage requirements. A complementary approach to large initial condition ensembles is to train statistical models on fewer runs. While simulations from these models cannot capture the complexity of climate model runs, they can address scientific questions of interest such as sampling the full variability of regional trends. We demonstrate this potential by comparing runs from a large ensemble and a statistical model trained with only four runs, and show that the variability of regional temperature trends is indistinguishable between the original simulations and the ones from the statistical model. Training statistical models on fewer runs might prove especially useful in the context of large climate model inter-comparison projects where creating large ensembles for each model is not possible.
|