One approach to improving the accuracy of a coarse-grid global climate model is to add machine-learned (ML) state-dependent corrections to the prognosed model tendencies, such that the climate model evolves more like a reference fine-grid global storm-resolving model (GSRM). Our past work demonstrating this approach was trained with short (40-day) simulations of GFDL's X-SHiELD GSRM with 3 km global horizontal grid spacing. Here, we extend this approach to span the full annual cycle by training and testing our ML using a new year-long GSRM simulation. Our corrective ML models are trained by learning the state-dependent tendencies of temperature and humidity and surface radiative fluxes needed to nudge a closely related 200 km grid coarse model, FV3GFS, to the GSRM evolution. Coarse-grid simulations adding these learned ML corrections run stably for multiple years. Compared to a no-ML baseline, the time-mean spatial pattern errors with respect to the fine-grid target are reduced by 6\%–26\% for land surface temperature and 9\%–25\% for land surface precipitation. The ML-corrected simulations develop other biases in climate and circulation that differ from, but have comparable amplitude to, the no-ML baseline simulation.
Machine-Learned Climate Model Corrections From a Global Storm-Resolving Model
language
en
file
Full Text PDF:/Users/pascal/Zotero/storage/P8KEGMMV/Kwa et al. - 2023 - Machine-Learned Climate Model Corrections From a G.pdf:application/pdf;Snapshot:/Users/pascal/Zotero/storage/B9NJ3SEK/2022MS003400.html:text/html
%0 Journal Article
%1 kwa_machine-learned_2023
%A Kwa, Anna
%A Clark, Spencer K.
%A Henn, Brian
%A Brenowitz, Noah D.
%A McGibbon, Jeremy
%A Watt-Meyer, Oliver
%A Perkins, W. Andre
%A Harris, Lucas
%A Bretherton, Christopher S.
%D 2023
%J Journal of Advances in Modeling Earth Systems
%K climate coarse-graining, ecomodelling global learning, machine model modeling, storm-resolving
%N 5
%P e2022MS003400
%R 10.1029/2022MS003400
%T Machine-Learned Climate Model Corrections From a Global Storm-Resolving Model: Performance Across the Annual Cycle
%U https://onlinelibrary.wiley.com/doi/abs/10.1029/2022MS003400
%V 15
%X One approach to improving the accuracy of a coarse-grid global climate model is to add machine-learned (ML) state-dependent corrections to the prognosed model tendencies, such that the climate model evolves more like a reference fine-grid global storm-resolving model (GSRM). Our past work demonstrating this approach was trained with short (40-day) simulations of GFDL's X-SHiELD GSRM with 3 km global horizontal grid spacing. Here, we extend this approach to span the full annual cycle by training and testing our ML using a new year-long GSRM simulation. Our corrective ML models are trained by learning the state-dependent tendencies of temperature and humidity and surface radiative fluxes needed to nudge a closely related 200 km grid coarse model, FV3GFS, to the GSRM evolution. Coarse-grid simulations adding these learned ML corrections run stably for multiple years. Compared to a no-ML baseline, the time-mean spatial pattern errors with respect to the fine-grid target are reduced by 6\%–26\% for land surface temperature and 9\%–25\% for land surface precipitation. The ML-corrected simulations develop other biases in climate and circulation that differ from, but have comparable amplitude to, the no-ML baseline simulation.
%Z e2022MS003400 2022MS003400