@article{PAR00025420,
  title = {{A} posteriori learning for quasi-geostrophic turbulence parametrization},
  author = {{F}rezat, {H}. and {L}e {S}ommer, {J}. and {F}ablet, {R}. and {B}alarac, {G}. and {L}guensat, {R}edouane},
  editor = {},
  language = {{ENG}},
  abstract = {{T}he use of machine learning to build subgrid parametrizations for climate models is receiving growing attention. {S}tate-of-the-art strategies address the problem as a supervised learning task and optimize algorithms that predict subgrid fluxes based on information from coarse resolution models. {I}n practice, training data are generated from higher resolution numerical simulations transformed in order to mimic coarse resolution simulations. {B}y essence, these strategies optimize subgrid parametrizations to meet so-called a priori criteria. {B}ut the actual purpose of a subgrid parametrization is to obtain good performance in terms of a posteriori metrics which imply computing entire model trajectories. {I}n this paper, we focus on the representation of energy backscatter in two-dimensional quasi-geostrophic turbulence and compare parametrizations obtained with different learning strategies at fixed computational complexity. {W}e show that strategies based on a priori criteria yield parametrizations that tend to be unstable in direct simulations and describe how subgrid parametrizations can alternatively be trained end-to-end in order to meet a posteriori criteria. {W}e illustrate that end-to-end learning strategies yield parametrizations that outperform known empirical and data-driven schemes in terms of performance, stability, and ability to apply to different flow configurations. {T}hese results support the relevance of differentiable programming paradigms for climate models in the future.},
  keywords = {parametrization ; machine learning ; turbulence ; quasi-geostrophic},
  booktitle = {},
  journal = {{J}ournal of {A}dvances in {M}odeling {E}arth {S}ystems},
  volume = {14},
  numero = {11},
  pages = {e2022{MS}003124 [23 p.]},
  year = {2022},
  DOI = {10.1029/2022ms003124},
  URL = {https://www.documentation.ird.fr/hor/{PAR}00025420},
}