ClimateBench: A benchmark for data-driven climate projections

Abstract

<p>Exploration of future emissions scenarios mostly relies on one-dimensional impulse response models, or simple pattern scaling approaches to approximate the physical climate response to a given scenario. Such approaches are unable to reliably predict climate variables which respond non-linearly to emissions or forcing (such as precipitation) and must rely on heavily simplified representations of e.g., aerosol, neglecting important spatial dependencies.</p><p>Here we present ClimateBench - a benchmark dataset based on a suite of CMIP, AerChemMIP and DAMIP simulations performed by NorESM2, and a set of baseline machine learning models that emulate its response to a variety of forcers. These surrogate models can skilfully predict annual mean global distributions of temperature, diurnal temperature range and precipitation (including extreme precipitation) given a wide range of emissions and concentrations of carbon dioxide, methane and spatially resolved aerosol. We discuss the accuracy and interpretability of these emulators and consider their robustness to physical constraints such as total energy conservation. Future opportunities incorporating such physical constraints directly in the machine learning models and using the emulators for detection and attribution studies are also discussed. This opens a wide range of opportunities to improve prediction, consistency and mathematical tractability.</p><p>We hope that by defining a clear baseline with appropriate metrics and providing a variety of baseline models we can bring the power of modern machine learning techniques to bear on the important problem of efficiently and robustly sampling future climates.</p>

FaIRGP: A Bayesian Energy Balance Model for Surface Temperatures Emulation

“…Conducting detection and attribution studies with emulators is not exclusive to FaIRGP, and could in principle be conducted with any emulator as discussed in Watson‐Parris et al. (2022). However, the strength of FaIRGP lies in its ability to input temperature ranges directly into a known probability density function, providing a precise probability between 0 and 1 of such temperatures to occur under a given emission scenario.…”

“…The data is obtained from the ClimateBench v1.0 (Watson‐Parris et al., 2022) climate emulation benchmark data set. ClimateBench v1.0 proposes a curated data set of local annual mean surface temperature (∼2° horizontal resolution), paired with annual emissions for four of the main anthropogenic forcing agents: carbon dioxide (CO 2 ), methane (CH 4 ), sulfur dioxide (SO 2 ) and black carbon (BC).…”

“…These results suggest that FaIRGP provides a better emulation of the global temperature response to anthropogenic forcing compared to both FaIR and the plain GP model. We include in Appendix C2 a comparison with the GP emulator from ClimateBench (Watson‐Parris et al., 2022).…”

“…Statistical modeling enjoys greater flexibility and has produced powerful emulators, capable of successfully approximating ESMs' spatially resolved climate projections with visually indistinguishable outputs. More recently, statistically driven emulators drawing from advances in statistical machine learning have demonstrated a remarkable capacity at regressing global emission profiles onto spatially resolved temperature and precipitation maps (Watson‐Parris et al., 2022).…”

AtmoDist: Self-supervised representation learning for atmospheric dynamics

“…The tasks evaluated in our work, that is, downscaling and missing data interpolation, are only two possible applications that we selected to demonstrate the usefulness of the learned representations of AtmoDist. Forecasting (Rasp et al, 2020;Bi et al, 2022;Lam et al, 2022;Pathak et al, 2022), climate response modeling (Watson-Parris et al, 2022), or the detection and modeling of extreme weather events (Racah et al, 2017;Blanchard et al, 2022), are other potential tasks that could benefit from improved representation learning.…”