TiPES scientists Michael Ghil and Valerio Lucarini (together with Gabriele Vissio and Valerio Lembo) propose a metric to compare and evaluate climate simulations in Geophysical Research Letters.
No climate model simulates all climatic processes correctly. Each IPCC-class model has its strengths and weaknesses. This of course leads to uncertainty about future climate change. Which results should we trust the most, and in which respect?
In the successive IPCC reports, this problem is addressed by presenting the results of a range of climate simulations. But in theory, it should be possible to harvest the strengths of the many models, disregard their respectively weaker aspects, and thus be able to construct improved climate predictions with less uncertainty based on their many somewhat different results.
That is what G. Vissio et al. attempt in the article “Evaluating the performance of climate models based on Wasserstein distance” published recently in Geophysical Research Letters.
Based on novel and sophisticated ideas from the theory of dynamical systems, as reported in Ghil & Lucarini (Rev. Mod. Phys., 2020), the authors have developed an evaluation method of model performance that takes into account a model’s skill in simulating several aspects of climate, like temperatures, precipitation, and sea ice extent.
In their paper, a range of models is evaluated on how they estimate such aspects. This improved evaluation method can help to provide better composite climate projections.
- Article in Geophysical Research Letters