We address the question of how to use a machine learned (ML) parameterization in a general circulation model (GCM), and assess its performance both computationally and physically. We take one particular ML parameterization (Guillaumin & Zanna, 2021,
Coupled climate simulations that span several hundred years cannot be run at a high‐enough spatial resolution to resolve mesoscale ocean dynamics. Recently, several studies have considered Deep Learning to parameterize subgrid forcing within macroscale ocean equations using data from ocean‐only simulations with idealized geometry. We present a stochastic Deep Learning parameterization that is trained on data generated by CM2.6, a high‐resolution state‐of‐the‐art coupled climate model. We train a Convolutional Neural Network for the subgrid momentum forcing using macroscale surface velocities from a few selected subdomains with different dynamical regimes. At each location of the coarse grid, rather than predicting a single number for the subgrid momentum forcing, we predict both the mean and standard deviation of a Gaussian probability distribution. This approach requires training our neural network to minimize a negative log‐likelihood loss function rather than the Mean Square Error, which has been the standard in applications of Deep Learning to the problem of parameterizations. Each estimate of the conditional mean subgrid forcing is thus associated with an uncertainty estimate–the standard deviation—which will form the basis for a stochastic subgrid parameterization. Offline tests show that our parameterization generalizes well to the global oceans and a climate with increased
- Award ID(s):
- 1912357
- NSF-PAR ID:
- 10449853
- Publisher / Repository:
- DOI PREFIX: 10.1029
- Date Published:
- Journal Name:
- Journal of Advances in Modeling Earth Systems
- Volume:
- 13
- Issue:
- 9
- ISSN:
- 1942-2466
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract https://doi.org/10.1002/essoar.10506419.1 ) and evaluate the online performance in a different model from which it was previously tested. This parameterization is a deep convolutional network that predicts parameters for a stochastic model of subgrid momentum forcing by mesoscale eddies. We treat the parameterization as we would a conventional parameterization once implemented in the numerical model. This includes trying the parameterization in a different flow regime from that in which it was trained, at different spatial resolutions, and with other differences, all to test generalization. We assess whether tuning is possible, which is a common practice in GCM development. We find the parameterization, without modification or special treatment, to be stable and that the action of the parameterization to be diminishing as spatial resolution is refined. We also find some limitations of the machine learning model in implementation: (a) tuning of the outputs from the parameterization at various depths is necessary; (b) the forcing near boundaries is not predicted as well as in the open ocean; (c) the cost of the parameterization is prohibitively high on central processing units. We discuss these limitations, present some solutions to problems, and conclude that this particular ML parameterization does inject energy, and improve backscatter, as intended but it might need further refinement before we can use it in production mode in contemporary climate models. -
more » « less
Abstract Attempts to use machine learning to develop atmospheric parameterizations have mainly focused on subgrid effects on temperature and moisture, but subgrid momentum transport is also important in simulations of the atmospheric circulation. Here, we use neural networks to develop a subgrid momentum transport parameterization that learns from coarse‐grained output of a high‐resolution atmospheric simulation in an idealized aquaplanet domain. We show that substantial subgrid momentum transport occurs due to convection. The neural‐network parameterization has skill in predicting momentum fluxes associated with convection, although its skill for subgrid momentum fluxes is lower compared to subgrid energy and moisture fluxes. The parameterization conserves momentum, and when implemented in the same atmospheric model at coarse resolution it leads to stable simulations and tends to reduce wind biases, although it over‐corrects for one configuration tested. Overall, our results show that it is challenging to predict subgrid momentum fluxes and that machine‐learning momentum parameterization gives promising results.
-
more » « less
Abstract A promising approach to improve climate‐model simulations is to replace traditional subgrid parameterizations based on simplified physical models by machine learning algorithms that are data‐driven. However, neural networks (NNs) often lead to instabilities and climate drift when coupled to an atmospheric model. Here, we learn an NN parameterization from a high‐resolution atmospheric simulation in an idealized domain by accurately calculating subgrid terms through coarse graining. The NN parameterization has a structure that ensures physical constraints are respected, such as by predicting subgrid fluxes instead of tendencies. The NN parameterization leads to stable simulations that replicate the climate of the high‐resolution simulation with similar accuracy to a successful random‐forest parameterization while needing far less memory. We find that the simulations are stable for different horizontal resolutions and a variety of NN architectures, and that an NN with substantially reduced numerical precision could decrease computational costs without affecting the quality of simulations.
-
more » « less
Abstract Many chemical processes depend non‐linearly on temperature. Gravity‐wave‐induced temperature perturbations have been shown to affect atmospheric chemistry, but accounting for this process in chemistry‐climate models has been a challenge because many gravity waves have scales smaller than the typical model resolution. Here, we present a method to account for subgrid‐scale orographic gravity‐wave‐induced temperature perturbations on the global scale for the Whole Atmosphere Community Climate Model. Temperature perturbation amplitudes
consistent with the model's subgrid‐scale gravity wave parameterization are derived and then used as a sinusoidal temperature perturbation in the model's chemistry solver. Because of limitations in the parameterization, we explore scaling of between 0.6 and 1 based on comparisons to altitude‐dependent distributions of satellite and reanalysis data, where we discuss uncertainties. We probe the impact on the chemistry from the grid‐point to global scales, and show that the parameterization is able to represent mountain wave events as reported by previous literature. The gravity waves for example, lead to increased surface area densities of stratospheric aerosols. This increases chlorine activation, with impacts on the associated chemical composition. We obtain large local changes in some chemical species (e.g., active chlorine, NOx, N2O5) which are likely to be important for comparisons to airborne or satellite observations, but the changes to ozone loss are more modest. This approach enables the chemistry‐climate modeling community to account for subgrid‐scale gravity wave temperature perturbations interactively, consistent with the internal parameterizations and are expected to yield more realistic interactions and better representation of the chemistry. -
The Summer Heat Balance of the Oregon Inner Shelf Over Two Decades: Mean and Interannual Variabilitymore » « less
Abstract Summer temperature and velocity measurements from 14 years in 15 m of water over the inner shelf off Oregon were used to investigate interannual temperature variability and the capacity of the across‐shelf heat flux to buffer net surface warming. There was no observable trend in summer mean temperatures, and the standard deviation of interannual variability (0.5°C) was less than the standard deviation in daily temperatures each summer (1.6°C, on average). Yet net surface heat flux provided a nearly constant source of heat each year, with a standard deviation less than 15
of the interannual mean. The summer mean across‐shelf upwelling circulation advected warmer water offshore near the surface, cooling the inner shelf and buffering the surface warming. In most years (11 out of 14), this two‐dimensional heat budget roughly closed with a residual less than 20 of the leading term. Even in years when the heat budget did not balance, the observed temperature change was negligible, indicating that an additional source of cooling was needed to close the budget. A comparison of the residual to the interannual variability in fields such as along‐shelf wind stress, stratification, and along‐shelf currents found no significant correlation, and further investigation into the intraseasonal dynamics is recommended to explain the results. An improved understanding of the processes that contribute to warming or cooling of the coastal ocean has the potential to improve predictions of the impact of year‐to‐year changes in local winds and circulation, such as from marine heat waves or climate change, on coastal temperatures.
