fv3gfs-wrapper: a Python wrapper of the FV3GFS atmospheric model

McGibbon, Jeremy; Brenowitz, Noah D.; Cheeseman, Mark; Clark, Spencer K.; Dahm, Johann; Davis, Eddie C.; Elbert, Oliver D.; George, Rhea C.; Harris, Lucas; Henn, Brian; Kwa, Anna; Perkins, W. A.; Watt‐Meyer, Oliver; Wicky, Tobias; Bretherton, Christopher S.; Fuhrer, Oliver

doi:10.5194/gmd-14-4401-2021

Cited by 13 publications

(15 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We use a Python wrapper of the FV3GFS Fortran model (McGibbon et al, 2021) in order to execute Python code during the model simulation. Briefly, the wrapper allows viewing and modifying the model state from a Python script at certain checkpoints in the main Fortran time loop.…”

Section: Coupling Of Machine Learning To Gcmmentioning

confidence: 99%

Correcting Weather and Climate Models by Machine Learning Nudged Historical Simulations

Watt‐Meyer

Brenowitz

Clark

et al. 2021

Geophysical Research Letters

Self Cite

View full text Add to dashboard Cite

Despite steady improvements in the skill of numerical weather and climate models over the last decades, a longstanding issue is the development of biases after initialization. These biases (systematic forecast errors) cause degradation of performance for both medium range weather forecasting and subseasonal to decadal climate predictions. They arise from issues like limited resolution, inaccurate physical parameterizations, and imperfect initial conditions. Typically, postprocessing steps are developed to handle these biases such as model output statistics for weather forecasting (Glahn & Lowry, 1972) or ensemble bias correction for seasonal prediction (Arribas et al., 2011;Stockdale et al., 1988). In this study, we propose an online bias correction method using machine learning (ML). We apply a corrective tendency to the prognostic state of the atmospheric model at each time step in order to reduce atmospheric model error growth. The necessary corrective tendencies are estimated from a hindcast simulation which is linearly nudged towards an observational analysis. An ML model is trained to predict the nudging tendencies using only the state of the model as inputs. This ML model can then be used in a forecast to keep the model evolution on a more realistic manifold.

show abstract

Section: Coupling Of Machine Learning To Gcmmentioning

confidence: 99%

Correcting Weather and Climate Models by Machine Learning Nudged Historical Simulations

Watt‐Meyer

Brenowitz

Clark

et al. 2021

Geophysical Research Letters

Self Cite

View full text Add to dashboard Cite

show abstract

“…The results presented here use a C48 horizontal grid with approximately 200 km horizontal resolution, a 15 min physics timestep, and 6 dynamics sub steps per physics time step. Our version of FV3GFS, described in McGibbon et al (2021), is built from portions of NOAA's Unified Forecast System (https://ufscommunity.org; code repository at https://doi.org/10.5281/zenodo.4460292). We disable microphysical updates within the dynamical core of the coarse-grid model to cleanly separate tendencies due to model dynamics and physics.…”

Section: Coarse-grid Model Python Wrapper and Cloud-based ML Workflowmentioning

confidence: 99%

“…Because of the wealth of powerful machine-learning packages available in Python, major units of the FV3GFS Fortran code were wrapped in Python (McGibbon et al, 2021). The ML and FV3GFS workflows were executed as containerized steps on Google Cloud Platform, similar to WM21.…”

Section: Coarse-grid Model Python Wrapper and Cloud-based ML Workflowmentioning

confidence: 99%

Correcting Coarse‐Grid Weather and Climate Models by Machine Learning From Global Storm‐Resolving Simulations

Bretherton

Henn

Kwa

et al. 2022

J Adv Model Earth Syst

Self Cite

View full text Add to dashboard Cite

Global atmospheric “storm‐resolving” models with horizontal grid spacing of less than 5 km resolve deep cumulus convection and flow in complex terrain. They promise to be reference models that could be used to improve computationally affordable coarse‐grid global climate models across a range of climates, reducing uncertainties in regional precipitation and temperature trends. Here, machine learning of nudging tendencies as functions of column state is used to correct the physical parameterization tendencies of temperature, humidity, and optionally winds, in a real‐geography coarse‐grid model (FV3GFS with a 200 km grid) to be closer to those of a 40‐day reference simulation using X‐SHiELD, a modified version of FV3GFS with a 3 km grid. Both simulations specify the same historical sea‐surface temperature fields. This methodology builds on a prior study using a global observational analysis as the reference. The coarse‐grid model without machine learning corrections has too few clouds, causing too much daytime heating of land surfaces that creates excessive surface latent heat flux and rainfall. This bias is avoided by learning downwelling radiative flux from the fine‐grid model. The best configuration uses learned nudging tendencies for temperature and humidity but not winds. Neural nets slightly outperform random forests. Forecasts of 850 hPa temperature gain 18 hr of skill at 3–7 days leads and time‐mean precipitation patterns are improved 30% by applying the ML correction. Adding machine‐learned wind tendencies improves 500 hPa height skill for the first five days of forecasts but degrades time‐mean upper tropospheric temperature and zonal wind patterns thereafter.

show abstract

“…Pace is a GT4Py implementation of the nonhydrostatic FV3 dynamical core (Putman and Lin, 2007;Harris et al, 2021). It is based on the same version of the National Oceanic and Atmospheric Administration (NOAA) Unified Forecast System (UFS) model as McGibbon et al (2021), forked from the UFS respository v1 in December 2019 (Zhou et al, 2019), and is nearly identical to the dynamical core used in SHiELD (Harris et al, 2020b). At present Pace only supports nonhydrostatic, uniformresolution simulations, with a restricted set of subgrid reconstruction schemes (hord and kord values).…”

Section: Pacementioning

confidence: 99%

“…Our "standard" test case is generated from NCEP reanalysis data from 0Z on August 1, 2016 (as in McGibbon et al, 2021), and the other is the baroclinic instability test case described in Jablonowski and Williamson (2006). Initial versions of the DSL code were tested on the standard case run on 6 MPI ranks (one per tile) and a 12 by 12 horizontal grid on each tile face with 79 vertical levels.…”

Section: Model Validationmentioning

confidence: 99%

Pace v0.1: A Python-based Performance-Portable Implementation of the FV3 Dynamical Core

Dahm

Davis²,

Deconinck³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. Progress in leveraging current and emerging high-performance computing infrastructures using traditional weather and climate models has been slow. This has become known more broadly as the software productivity gap. With the end of Moore's Law driving forward rapid specialization of hardware architectures, building simulation codes on a low-level language with hardware specific optimizations is a significant risk. As a solution, we present Pace, an implementation of the nonhydrostatic FV3 dynamical core which is entirely Python-based. In order to achieve high performance on a diverse set of hardware architectures, Pace is written using the GT4Py domain-specific language. We demonstrate that with this approach we can achieve portability and performance, while significantly improving the readability and maintainability of the code as compared to the Fortran reference implementation. We show that Pace can run at scale on leadership-class supercomputers and achieve performance speeds 3.5–4 times faster than the Fortran code on GPU-accelerated supercomputers. Furthermore, we demonstrate how a Python-based simulation code facilitates existing or enables entirely new use-cases and workflows. Pace demonstrates how a high-level language can insulate us from disruptive changes, provide a more productive development environment, and facilitate the integration with new technologies such as machine learning.

show abstract

fv3gfs-wrapper: a Python wrapper of the FV3GFS atmospheric model

Cited by 13 publications

References 17 publications

Correcting Weather and Climate Models by Machine Learning Nudged Historical Simulations

Correcting Weather and Climate Models by Machine Learning Nudged Historical Simulations

Correcting Coarse‐Grid Weather and Climate Models by Machine Learning From Global Storm‐Resolving Simulations

Pace v0.1: A Python-based Performance-Portable Implementation of the FV3 Dynamical Core

Contact Info

Product

Resources

About