Deep learning to represent subgrid processes in climate models

Rasp, Stephan; Pritchard, Michael S.; Gentine, Pierre

doi:10.1073/pnas.1810286115

Cited by 649 publications

(742 citation statements)

References 36 publications

Supporting

Mentioning

676

Contrasting

Unclassified

Order By: Relevance

“…While the correlation and MSE summary statistics are very similar for both the DNN and Random Forest regression, the Random Forest prediction time is 27% slower than the DNN. Though the model prediction time is not perfectly optimized, these results are consistent with previous work demonstrating rapid computation from DNNs (Rasp et al, ), and slower implementations of Random Forests (Keller & Evans, ). Given the computational challenges many models already face, the lack of process‐based information currently available in models of V d , and the great potential for DNN model portability and retraining (Chollet & Allaire, ), we believe that the application of a DNN for this purpose is well justified.…”

Section: Resultssupporting

confidence: 87%

A Deep Learning Parameterization for Ozone Dry Deposition Velocities

Silva

Heald

Ravela

et al. 2019

Geophysical Research Letters

View full text Add to dashboard Cite

The loss of ozone to terrestrial and aquatic systems, known as dry deposition, is a highly uncertain process governed by turbulent transport, interfacial chemistry, and plant physiology. We demonstrate the value of using Deep Neural Networks (DNN) in predicting ozone dry deposition velocities. We find that a feedforward DNN trained on observations from a coniferous forest site (Hyytiälä, Finland) can predict hourly ozone dry deposition velocities at a mixed forest site (Harvard Forest, Massachusetts) more accurately than modern theoretical models, with a reduction in the normalized mean bias (0.05 versus ~0.1). The same DNN model, when driven by assimilated meteorology at 2° × 2.5° spatial resolution, outperforms the Wesely scheme as implemented in the GEOS‐Chem model. With more available training data from other climate and ecological zones, this methodology could yield a generalizable DNN suitable for global models.

show abstract

Section: Resultssupporting

confidence: 87%

A Deep Learning Parameterization for Ozone Dry Deposition Velocities

Silva

Heald

Ravela

et al. 2019

Geophysical Research Letters

View full text Add to dashboard Cite

show abstract

“…Several aspects of our coarse‐graining procedure could cause the mean state to drift, which Rasp et al () did not observe with their SP data. First, unlike in SP, NG‐Aqua contains many subgrid‐scale sources of momentum, which we neglected altogether.…”

Section: Resultsmentioning

confidence: 99%

Spatially Extended Tests of a Neural Network Parametrization Trained by Coarse‐Graining

Brenowitz

Bretherton

2019

J Adv Model Earth Syst

125

168

View full text Add to dashboard Cite

General circulation models (GCMs) typically have a grid size of 25–200 km. Parametrizations are used to represent diabatic processes such as radiative transfer and cloud microphysics and account for subgrid‐scale motions and variability. Unlike traditional approaches, neural networks (NNs) can readily exploit recent observational data sets and global cloud‐system resolving model (CRM) simulations to learn subgrid variability. This article describes an NN parametrization trained by coarse‐graining a near‐global CRM simulation with a 4‐km horizontal grid spacing. The NN predicts the residual heating and moistening averaged over (160 km)2 grid boxes as a function of the coarse‐resolution fields within the same atmospheric column. This NN is coupled to the dynamical core of a GCM with the same 160‐km resolution. A recent study described how to train such an NN to be stable when coupled to specified time‐evolving advective forcings in a single‐column model, but feedbacks between NN and GCM components cause spatially extended simulations to crash within a few days. Analyzing the linearized response of such an NN reveals that it learns to exploit a strong synchrony between precipitation and the atmospheric state above 10 km. Removing these variables from the NN's inputs stabilizes the coupled simulations, which predict the future state more accurately than a coarse‐resolution simulation without any parametrizations of subgrid‐scale variability, although the mean state slowly drifts.

show abstract

“…For example, Rasp and Lerch () used neural networks (NNs) to successfully improve postprocessing of GCM forecasts to surface stations, while Rodrigues et al () demonstrated the ability of deep NNs to downscale GCM output to higher horizontal resolution. Deep NNs have also been used to identify extreme weather and climate patterns in observed and modeled atmospheric states (Kurth et al, ; Lagerquist et al, ; Liu et al, ), improve parameterizations in GCMs (e.g., Brenowitz & Bretherton, ; Rasp et al, ), and predict uncertainty in weather forecasts (Scher & Messori, ). Larraondo et al () demonstrated the ability of deep NNs to extract spatial patterns in precipitation from gridded atmospheric fields.…”

Section: Introductionmentioning

confidence: 99%

Can Machines Learn to Predict Weather? Using Deep Learning to Predict Gridded 500‐hPa Geopotential Height From Historical Weather Data

Weyn

Durran

Caruana

2019

J Adv Model Earth Syst

226

166

View full text Add to dashboard Cite

We develop elementary weather prediction models using deep convolutional neural networks (CNNs) trained on past weather data to forecast one or two fundamental meteorological fields on a Northern Hemisphere grid with no explicit knowledge about physical processes. At forecast lead times up to 3 days, CNNs trained to predict only 500‐hPa geopotential height easily outperform persistence, climatology, and the dynamics‐based barotropic vorticity model, but do not beat an operational full‐physics weather prediction model. These CNNs are capable of forecasting significant changes in the intensity of weather systems, which is notable because this is beyond the capability of the fundamental dynamical equation that relies solely on 500‐hPa data, the barotropic vorticity equation. Modest improvements to the CNN forecasts can be made by adding 700‐ to 300‐hPa thickness to the input data. Our best performing CNN does a good job of capturing the climatology and annual variability of 500‐hPa heights and is capable of forecasting realistic atmospheric states at lead times of 14 days. Although our simple models do not perform better than an operational weather model, machine learning warrants further exploration as a weather forecasting tool; in particular, the potential efficiency of CNNs might make them attractive for ensemble forecasting.

show abstract

Deep learning to represent subgrid processes in climate models

Cited by 649 publications

References 36 publications

A Deep Learning Parameterization for Ozone Dry Deposition Velocities

A Deep Learning Parameterization for Ozone Dry Deposition Velocities

Spatially Extended Tests of a Neural Network Parametrization Trained by Coarse‐Graining

Can Machines Learn to Predict Weather? Using Deep Learning to Predict Gridded 500‐hPa Geopotential Height From Historical Weather Data

Contact Info

Product

Resources

About