Articles | Volume 28, issue 3
https://doi.org/10.5194/npg-28-347-2021
© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/npg-28-347-2021
© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Research article
| 
30 Jul 2021
Research article |
| 30 Jul 2021
| 30 Jul 2021
Producing realistic climate data with generative adversarial networks
CERFACS, Toulouse, France
Institut National Polytechnique de Toulouse, Toulouse, France
Olivier Pannekoucke
CORRESPONDING AUTHOR
CNRM, Université de Toulouse, Météo-France, CNRS, Toulouse, France
Corentin Lapeyre
CORRESPONDING AUTHOR
CERFACS, Toulouse, France
CERFACS, Toulouse, France
Olivier Thual
CORRESPONDING AUTHOR
CERFACS, Toulouse, France
Institut de Mécanique des Fluides de Toulouse (IMFT), Université de Toulouse, CNRS, Toulouse, France
Related authors
No articles found.
Rachid El Montassir, Olivier Pannekoucke, and Corentin Lapeyre
EGUsphere, https://doi.org/10.5194/egusphere-2023-3078, https://doi.org/10.5194/egusphere-2023-3078, 2024
Short summary
Short summary
This study introduces a novel approach, combining Physics and Artificial Intelligence (AI) for improved cloud cover forecasting. This approach outperforms traditional Deep Learning (DL) methods in producing realistic and physically consistent results while requiring less training data. This architecture provides a promising solution to overcome the limitations of classical AI methods, and contributes to open up new possibilities for combining physical knowledge with deep learning models.
Théo Defontaine, Sophie Ricci, Corentin J. Lapeyre, Arthur Marchandise, and Etienne Le Pape
EGUsphere, https://doi.org/10.5194/egusphere-2023-2621, https://doi.org/10.5194/egusphere-2023-2621, 2024
Short summary
Short summary
This work presents how machine learning models predict discharge and outperforms the 6 h lead-time empirical model used operationally in Toulouse. The 40 k points database includes discharge and rainfall data, for 36 flood events. The approach also provides a reliable solution for extended 8 h lead-time. The scarcity and the heterogeneity of the data, especially in presence of outliers, heavily weigh on the learning strategy. Rainfall data processing increases the predictive performances.
Benjamin M. Sanderson, Angeline G. Pendergrass, Charles D. Koven, Florent Brient, Ben B. B. Booth, Rosie A. Fisher, and Reto Knutti
Earth Syst. Dynam., 12, 899–918, https://doi.org/10.5194/esd-12-899-2021, https://doi.org/10.5194/esd-12-899-2021, 2021
Short summary
Short summary
Emergent constraints promise a pathway to the reduction in climate projection uncertainties by exploiting ensemble relationships between observable quantities and unknown climate response parameters. This study considers the robustness of these relationships in light of biases and common simplifications that may be present in the original ensemble of climate simulations. We propose a classification scheme for constraints and a number of practical case studies.
Claudia Tebaldi, Kevin Debeire, Veronika Eyring, Erich Fischer, John Fyfe, Pierre Friedlingstein, Reto Knutti, Jason Lowe, Brian O'Neill, Benjamin Sanderson, Detlef van Vuuren, Keywan Riahi, Malte Meinshausen, Zebedee Nicholls, Katarzyna B. Tokarska, George Hurtt, Elmar Kriegler, Jean-Francois Lamarque, Gerald Meehl, Richard Moss, Susanne E. Bauer, Olivier Boucher, Victor Brovkin, Young-Hwa Byun, Martin Dix, Silvio Gualdi, Huan Guo, Jasmin G. John, Slava Kharin, YoungHo Kim, Tsuyoshi Koshiro, Libin Ma, Dirk Olivié, Swapna Panickal, Fangli Qiao, Xinyao Rong, Nan Rosenbloom, Martin Schupfner, Roland Séférian, Alistair Sellar, Tido Semmler, Xiaoying Shi, Zhenya Song, Christian Steger, Ronald Stouffer, Neil Swart, Kaoru Tachiiri, Qi Tang, Hiroaki Tatebe, Aurore Voldoire, Evgeny Volodin, Klaus Wyser, Xiaoge Xin, Shuting Yang, Yongqiang Yu, and Tilo Ziehn
Earth Syst. Dynam., 12, 253–293, https://doi.org/10.5194/esd-12-253-2021, https://doi.org/10.5194/esd-12-253-2021, 2021
Short summary
Short summary
We present an overview of CMIP6 ScenarioMIP outcomes from up to 38 participating ESMs according to the new SSP-based scenarios. Average temperature and precipitation projections according to a wide range of forcings, spanning a wider range than the CMIP5 projections, are documented as global averages and geographic patterns. Times of crossing various warming levels are computed, together with benefits of mitigation for selected pairs of scenarios. Comparisons with CMIP5 are also discussed.
Katherine Dagon, Benjamin M. Sanderson, Rosie A. Fisher, and David M. Lawrence
Adv. Stat. Clim. Meteorol. Oceanogr., 6, 223–244, https://doi.org/10.5194/ascmo-6-223-2020, https://doi.org/10.5194/ascmo-6-223-2020, 2020
Short summary
Short summary
Uncertainties in land model projections are important to understand in order to build confidence in Earth system modeling. In this paper, we introduce a framework for estimating uncertain land model parameters with machine learning. This method increases the computational efficiency of this process relative to traditional hand tuning approaches and provides objective methods to assess the results. We further identify key processes and parameters that are important for accurate land modeling.
Benjamin Sanderson
Earth Syst. Dynam., 11, 721–735, https://doi.org/10.5194/esd-11-721-2020, https://doi.org/10.5194/esd-11-721-2020, 2020
Short summary
Short summary
Here, we assess the degree to which the idealized responses to transient forcing increase and step change forcing increase relate to warming under future scenarios. We find a possible explanation for the poor performance of transient metrics (relative to equilibrium response) as a metric of high-emission future warming in terms of their sensitivity to non-equilibrated initial conditions, and propose alternative metrics which better describe warming under high mitigation scenarios.
Benjamin Sanderson
Earth Syst. Dynam., 11, 563–577, https://doi.org/10.5194/esd-11-563-2020, https://doi.org/10.5194/esd-11-563-2020, 2020
Short summary
Short summary
Levels of future temperature change are often used interchangeably with carbon budget allowances in climate policy, a relatively robust relationship on the timescale of this century. However, recent advances in understanding underline that continued warming after net-zero emissions have been achieved cannot be ruled out by observations of warming to date. We consider here how such behavior could be constrained and how policy can be framed in the context of these uncertainties.
Michael Wehner, Dáithí Stone, Dann Mitchell, Hideo Shiogama, Erich Fischer, Lise S. Graff, Viatcheslav V. Kharin, Ludwig Lierhammer, Benjamin Sanderson, and Harinarayan Krishnan
Earth Syst. Dynam., 9, 299–311, https://doi.org/10.5194/esd-9-299-2018, https://doi.org/10.5194/esd-9-299-2018, 2018
Short summary
Short summary
The United Nations Framework Convention on Climate Change challenged the scientific community to describe the impacts of stabilizing the global temperature at its 21st Conference of Parties. A specific target of 1.5 °C above preindustrial levels had not been seriously considered by the climate modeling community prior to the Paris Agreement. This paper analyzes heat waves in simulations designed for this target. We find there are reductions in extreme temperature compared to a 2 °C target.
Nadja Herger, Gab Abramowitz, Reto Knutti, Oliver Angélil, Karsten Lehmann, and Benjamin M. Sanderson
Earth Syst. Dynam., 9, 135–151, https://doi.org/10.5194/esd-9-135-2018, https://doi.org/10.5194/esd-9-135-2018, 2018
Short summary
Short summary
Users presented with large multi-model ensembles commonly use the equally weighted model mean as a best estimate, ignoring the issue of near replication of some climate models. We present an efficient and flexible tool that finds a subset of models with improved mean performance compared to the multi-model mean while at the same time maintaining the spread and addressing the problem of model interdependence. Out-of-sample skill and reliability are demonstrated using model-as-truth experiments.
Benjamin M. Sanderson, Yangyang Xu, Claudia Tebaldi, Michael Wehner, Brian O'Neill, Alexandra Jahn, Angeline G. Pendergrass, Flavio Lehner, Warren G. Strand, Lei Lin, Reto Knutti, and Jean Francois Lamarque
Earth Syst. Dynam., 8, 827–847, https://doi.org/10.5194/esd-8-827-2017, https://doi.org/10.5194/esd-8-827-2017, 2017
Short summary
Short summary
We present the results of a set of climate simulations designed to simulate futures in which the Earth's temperature is stabilized at the levels referred to in the 2015 Paris Agreement. We consider the necessary future emissions reductions and the aspects of extreme weather which differ significantly between the 2 and 1.5 °C climate in the simulations.
Benjamin M. Sanderson, Michael Wehner, and Reto Knutti
Geosci. Model Dev., 10, 2379–2395, https://doi.org/10.5194/gmd-10-2379-2017, https://doi.org/10.5194/gmd-10-2379-2017, 2017
Short summary
Short summary
How should climate model simulations be combined to produce an overall assessment that reflects both their performance and their interdependencies? This paper presents a strategy for weighting climate model output such that models that are replicated or models that perform poorly in a chosen set of metrics are appropriately weighted. We perform sensitivity tests to show how the method results depend on variables and parameter values.
Allison H. Baker, Dorit M. Hammerling, Sheri A. Mickelson, Haiying Xu, Martin B. Stolpe, Phillipe Naveau, Ben Sanderson, Imme Ebert-Uphoff, Savini Samarasinghe, Francesco De Simone, Francesco Carbone, Christian N. Gencarelli, John M. Dennis, Jennifer E. Kay, and Peter Lindstrom
Geosci. Model Dev., 9, 4381–4403, https://doi.org/10.5194/gmd-9-4381-2016, https://doi.org/10.5194/gmd-9-4381-2016, 2016
Short summary
Short summary
We apply lossy data compression to output from the Community Earth System Model Large Ensemble Community Project. We challenge climate scientists to examine features of the data relevant to their interests and identify which of the ensemble members have been compressed, and we perform direct comparisons on features critical to climate science. We find that applying lossy data compression to climate model data effectively reduces data volumes with minimal effect on scientific results.
Brian C. O'Neill, Claudia Tebaldi, Detlef P. van Vuuren, Veronika Eyring, Pierre Friedlingstein, George Hurtt, Reto Knutti, Elmar Kriegler, Jean-Francois Lamarque, Jason Lowe, Gerald A. Meehl, Richard Moss, Keywan Riahi, and Benjamin M. Sanderson
Geosci. Model Dev., 9, 3461–3482, https://doi.org/10.5194/gmd-9-3461-2016, https://doi.org/10.5194/gmd-9-3461-2016, 2016
Short summary
Short summary
The Scenario Model Intercomparison Project (ScenarioMIP) will provide multi-model climate projections based on alternative scenarios of future emissions and land use changes produced with integrated assessment models. The design consists of eight alternative 21st century scenarios plus one large initial condition ensemble and a set of long-term extensions. Climate model projections will facilitate integrated studies of climate change as well as address targeted scientific questions.
Related subject area
Subject: Time series, machine learning, networks, stochastic processes, extreme events | Topic: Climate, atmosphere, ocean, hydrology, cryosphere, biosphere | Techniques: Big data and artificial intelligence
A comparison of two causal methods in the context of climate analyses
A two-fold deep-learning strategy to correct and downscale winds over mountains
Downscaling of surface wind forecasts using convolutional neural networks
The Sampling Method for Optimal Precursors of ENSO Event
Data-driven methods to estimate the committor function in conceptual ocean models
Exploring meteorological droughts' spatial patterns across Europe through complex network theory
Integrated hydrodynamic and machine learning models for compound flooding prediction in a data-scarce estuarine delta
Predicting sea surface temperatures with coupled reservoir computers
Using neural networks to improve simulations in the gray zone
The blessing of dimensionality for the analysis of climate data
Identification of droughts and heatwaves in Germany with regional climate networks
Extracting statistically significant eddy signals from large Lagrangian datasets using wavelet ridge analysis, with application to the Gulf of Mexico
Ensemble-based statistical interpolation with Gaussian anamorphosis for the spatial analysis of precipitation
Applications of matrix factorization methods to climate data
Detecting dynamical anomalies in time series from different palaeoclimate proxy archives using windowed recurrence network analysis
Remember the past: a comparison of time-adaptive training schemes for non-homogeneous regression
David Docquier, Giorgia Di Capua, Reik V. Donner, Carlos A. L. Pires, Amélie Simon, and Stéphane Vannitsem
Nonlin. Processes Geophys., 31, 115–136, https://doi.org/10.5194/npg-31-115-2024, https://doi.org/10.5194/npg-31-115-2024, 2024
Short summary
Short summary
Identifying causes of specific processes is crucial in order to better understand our climate system. Traditionally, correlation analyses have been used to identify cause–effect relationships in climate studies. However, correlation does not imply causation, which justifies the need to use causal methods. We compare two independent causal methods and show that these are superior to classical correlation analyses. We also find some interesting differences between the two methods.
Louis Le Toumelin, Isabelle Gouttevin, Clovis Galiez, and Nora Helbig
Nonlin. Processes Geophys., 31, 75–97, https://doi.org/10.5194/npg-31-75-2024, https://doi.org/10.5194/npg-31-75-2024, 2024
Short summary
Short summary
Forecasting wind fields over mountains is of high importance for several applications and particularly for understanding how wind erodes and disperses snow. Forecasters rely on operational wind forecasts over mountains, which are currently only available on kilometric scales. These forecasts can also be affected by errors of diverse origins. Here we introduce a new strategy based on artificial intelligence to correct large-scale wind forecasts in mountains and increase their spatial resolution.
Florian Dupuy, Pierre Durand, and Thierry Hedde
Nonlin. Processes Geophys., 30, 553–570, https://doi.org/10.5194/npg-30-553-2023, https://doi.org/10.5194/npg-30-553-2023, 2023
Short summary
Short summary
Forecasting near-surface winds over complex terrain requires high-resolution numerical weather prediction models, which drastically increase the duration of simulations and hinder them in running on a routine basis. A faster alternative is statistical downscaling. We explore different ways of calculating near-surface wind speed and direction using artificial intelligence algorithms based on various convolutional neural networks in order to find the best approach for wind downscaling.
Bin Shi and Junjie Ma
EGUsphere, https://doi.org/10.5194/egusphere-2023-2092, https://doi.org/10.5194/egusphere-2023-2092, 2023
Short summary
Short summary
In this study, different from traditional deterministic optimization algorithms, we implement the sampling method to compute the CNOPs in the realistic and predictive coupled ocean-atmosphere model, which reduces the first-order information to the zeroth-order one avoiding the high-cost computation of the gradient. The numerical performance highlights the importance of stochastic optimization algorithms to compute CNOPs and capture initial optimal precursors.
Valérian Jacques-Dumas, René M. van Westen, Freddy Bouchet, and Henk A. Dijkstra
Nonlin. Processes Geophys., 30, 195–216, https://doi.org/10.5194/npg-30-195-2023, https://doi.org/10.5194/npg-30-195-2023, 2023
Short summary
Short summary
Computing the probability of occurrence of rare events is relevant because of their high impact but also difficult due to the lack of data. Rare event algorithms are designed for that task, but their efficiency relies on a score function that is hard to compute. We compare four methods that compute this function from data and measure their performance to assess which one would be best suited to be applied to a climate model. We find neural networks to be most robust and flexible for this task.
Domenico Giaquinto, Warner Marzocchi, and Jürgen Kurths
Nonlin. Processes Geophys., 30, 167–181, https://doi.org/10.5194/npg-30-167-2023, https://doi.org/10.5194/npg-30-167-2023, 2023
Short summary
Short summary
Despite being among the most severe climate extremes, it is still challenging to assess droughts’ features for specific regions. In this paper we study meteorological droughts in Europe using concepts derived from climate network theory. By exploring the synchronization in droughts occurrences across the continent we unveil regional clusters which are individually examined to identify droughts’ geographical propagation and source–sink systems, which could potentially support droughts’ forecast.
Joko Sampurno, Valentin Vallaeys, Randy Ardianto, and Emmanuel Hanert
Nonlin. Processes Geophys., 29, 301–315, https://doi.org/10.5194/npg-29-301-2022, https://doi.org/10.5194/npg-29-301-2022, 2022
Short summary
Short summary
In this study, we successfully built and evaluated machine learning models for predicting water level dynamics as a proxy for compound flooding hazards in a data-scarce delta. The issues that we tackled here are data scarcity and low computational resources for building flood forecasting models. The proposed approach is suitable for use by local water management agencies in developing countries that encounter these issues.
Benjamin Walleshauser and Erik Bollt
Nonlin. Processes Geophys., 29, 255–264, https://doi.org/10.5194/npg-29-255-2022, https://doi.org/10.5194/npg-29-255-2022, 2022
Short summary
Short summary
As sea surface temperature (SST) is vital for understanding the greater climate of the Earth and is also an important variable in weather prediction, we propose a model that effectively capitalizes on the reduced complexity of machine learning models while still being able to efficiently predict over a large spatial domain. We find that it is proficient at predicting the SST at specific locations as well as over the greater domain of the Earth’s oceans.
Raphael Kriegmair, Yvonne Ruckstuhl, Stephan Rasp, and George Craig
Nonlin. Processes Geophys., 29, 171–181, https://doi.org/10.5194/npg-29-171-2022, https://doi.org/10.5194/npg-29-171-2022, 2022
Short summary
Short summary
Our regional numerical weather prediction models run at kilometer-scale resolutions. Processes that occur at smaller scales not yet resolved contribute significantly to the atmospheric flow. We use a neural network (NN) to represent the unresolved part of physical process such as cumulus clouds. We test this approach on a simplified, yet representative, 1D model and find that the NN corrections vastly improve the model forecast up to a couple of days.
Bo Christiansen
Nonlin. Processes Geophys., 28, 409–422, https://doi.org/10.5194/npg-28-409-2021, https://doi.org/10.5194/npg-28-409-2021, 2021
Short summary
Short summary
In geophysics we often need to analyse large samples of high-dimensional fields. Fortunately but counterintuitively, such high dimensionality can be a blessing, and we demonstrate how this allows simple analytical results to be derived. These results include estimates of correlations between sample members and how the sample mean depends on the sample size. We show that the properties of high dimensionality with success can be applied to climate fields, such as those from ensemble modelling.
Gerd Schädler and Marcus Breil
Nonlin. Processes Geophys., 28, 231–245, https://doi.org/10.5194/npg-28-231-2021, https://doi.org/10.5194/npg-28-231-2021, 2021
Short summary
Short summary
We used regional climate networks (RCNs) to identify past heatwaves and droughts in Germany. RCNs provide information for whole areas and can provide many details of extreme events. The RCNs were constructed on the grid of the E-OBS data set. Time series correlation was used to construct the networks. Network metrics were compared to standard extreme indices and differed considerably between normal and extreme years. The results show that RCNs can identify severe and moderate extremes.
Jonathan M. Lilly and Paula Pérez-Brunius
Nonlin. Processes Geophys., 28, 181–212, https://doi.org/10.5194/npg-28-181-2021, https://doi.org/10.5194/npg-28-181-2021, 2021
Short summary
Short summary
Long-lived eddies are an important part of the ocean circulation. Here a dataset for studying eddies in the Gulf of Mexico is created through the analysis of trajectories of drifting instruments. The method involves the identification of quasi-periodic signals, characteristic of particles trapped in eddies, from the displacement records, followed by the creation of a measure of statistical significance. It is expected that this dataset will be of use to other authors studying this region.
Cristian Lussana, Thomas N. Nipen, Ivar A. Seierstad, and Christoffer A. Elo
Nonlin. Processes Geophys., 28, 61–91, https://doi.org/10.5194/npg-28-61-2021, https://doi.org/10.5194/npg-28-61-2021, 2021
Short summary
Short summary
An unprecedented amount of rainfall data is available nowadays, such as ensemble model output, weather radar estimates, and in situ observations from networks of both traditional and opportunistic sensors. Nevertheless, the exact amount of precipitation, to some extent, eludes our knowledge. The objective of our study is precipitation reconstruction through the combination of numerical model outputs with observations from multiple data sources.
Dylan Harries and Terence J. O'Kane
Nonlin. Processes Geophys., 27, 453–471, https://doi.org/10.5194/npg-27-453-2020, https://doi.org/10.5194/npg-27-453-2020, 2020
Short summary
Short summary
Different dimension reduction methods may produce profoundly different low-dimensional representations of multiscale systems. We perform a set of case studies to investigate these differences. When a clear scale separation is present, similar bases are obtained using all methods, but when this is not the case some methods may produce representations that are poorly suited for describing features of interest, highlighting the importance of a careful choice of method when designing analyses.
Jaqueline Lekscha and Reik V. Donner
Nonlin. Processes Geophys., 27, 261–275, https://doi.org/10.5194/npg-27-261-2020, https://doi.org/10.5194/npg-27-261-2020, 2020
Moritz N. Lang, Sebastian Lerch, Georg J. Mayr, Thorsten Simon, Reto Stauffer, and Achim Zeileis
Nonlin. Processes Geophys., 27, 23–34, https://doi.org/10.5194/npg-27-23-2020, https://doi.org/10.5194/npg-27-23-2020, 2020
Short summary
Short summary
Statistical post-processing aims to increase the predictive skill of probabilistic ensemble weather forecasts by learning the statistical relation between historical pairs of observations and ensemble forecasts within a given training data set. This study compares four different training schemes and shows that including multiple years of data in the training set typically yields a more stable post-processing while it loses the ability to quickly adjust to temporal changes in the underlying data.
Cited articles
Besombes, C.: Producing realistic climate data with GANs, Zenodo [data set], https://doi.org/10.5281/zenodo.4442450, 2021 (data available at: https://github.com/Cam-B04/Producing-realistic-climate-data-with-GANs.git, last access: January 2021). a
Beusch, L., Gudmundsson, L., and Seneviratne, S. I.: Emulating Earth system model temperatures with MESMER: from global mean temperature trajectories to grid-point-level realizations on land, Earth Syst. Dynam., 11, 139–159, https://doi.org/10.5194/esd-11-139-2020, 2020. a
Boukabara, S.-A., Krasnopolsky, V., Stewart, J. Q., Maddy, E. S., Shahroudi, N., and Hoffman, R. N.: Leveraging modern artificial intelligence for remote sensing and NWP: Benefits and challenges, B. Am. Meteorol. Soc.,
100, ES473–ES491, 2019. a
Chan, S. and Elsheikh, A. H.:
Parametric generation of conditional geological realizations using generative neural networks,
Computat. Geosci.,
23, 925–952, https://doi.org/10.1007/s10596-019-09850-7, 2019. a
Davies, T.:
Lateral boundary conditions for limited area models,
Q. J. Roy. Meteor. Soc.,
140, 185–196, 2014. a
Dramsch, J. S.: 70 years of machine learning in geoscience in review,
Adv. Geophys., 61, 1–55, 2020. a
Fraedrich, K., Kirk, E., Lunkeit, F., Luksch, U., and Lunkeit, F.: The portable university model of the atmosphere (PUMA): Storm track dynamics and low-frequency variability, Meteorol. Z, 14, 735–746, 2005b. a
Gagne, D. J., Christensen, H. M., Subramanian, A. C., and Monahan, A. H.:
Machine learning for stochastic parameterization: Generative adversarial networks in the Lorenz'96 model, J. Adv. Model. Earth Sy., 12, e2019MS001896, https://doi.org/10.1029/2019MS001896, 2020. a
Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y.: Generative adversarial networks, Communications of the ACM, 63, 139–144, 2020. a
He, K., Zhang, X., Ren, S., and Sun, J.:
Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, CVPR 2016, 770–778, 2016. a
Hergenrother, E., Bleile, A., Middleton, D., and Trembilski, A.:
The abalone interpolation: A visual interpolation procedure for the calculation of cloud movement,
in: Proceedings. XV Brazilian Symposium on Computer Graphics and Image Processing, Fortaleza, Brazil, 10 October 2002, 381–387, IEEE, 2002. a
Houtekamer, P. L. and Zhang, F.:
Review of the ensemble Kalman filter for atmospheric data assimilation,
Mon. Weather Rev.,
144, 4489–4532, 2016. a
Isola, P., Zhu, J.-Y., Zhou, T., and Efros, A. A.:
Image-to-image translation with conditional adversarial networks,
in: Proceedings of the IEEE conference on computer vision and pattern recognition, Honolulu, HI, USA, 21–26 July 2017, 1125–1134, 2017. a
Kantorovich, L. V. and Rubinshtein, S.: On a space of totally additive functions, Vestnik of the St. Petersburg University: Mathematics,
13, 52–59, 1958. a
Kingma, D. P. and Ba, J.:
Adam: A Method for Stochastic Optimization, arXiv [preprint],
arXiv:1412.6980, 22 December
2014. a
Lagerquist, R., McGovern, A., and Gagne II, D. J.:
Deep learning for spatially explicit prediction of synoptic-scale fronts,
Weather Forecast.,
34, 1137–1160, 2019. a
Ledig, C., Theis, L., Huszár, F., Caballero, J., Cunningham, A., Acosta, A., Aitken, A., Tejani, A., Totz, J., Wang, Z., and Shi, W.:
Photo-realistic single image super-resolution using a generative adversarial network, in: Proceedings of the IEEE conference on computer vision and pattern recognition, Honolulu, HI, USA, 21–26 July 2017,
4681–4690, 2017. a
Leinonen, J., Guillaume, A., and Yuan, T.:
Reconstruction of cloud vertical structure with a generative adversarial network,
Geophys. Res. Lett.,
46, 7035–7044, 2019. a
Li, J. and Heap, A. D.:
Spatial interpolation methods applied in the environmental sciences: A review,
Environ. Modell. Softw.,
53, 173–189, 2014. a
Lorenc, A. C.:
The potential of the ensemble Kalman filter for NWP–a comparison with 4D-Var,
Q. J. Roy. Meteor. Soc.,
129, 3183–3203, 2003. a
Nagarajan, V. and Kolter, J. Z.: Gradient descent GAN optimization is locally stable, arXiv [preprint], arXiv:1706.04156, 13 June 2017. a
Pannekoucke, O., Cebron, P., Oger, N., and Arbogast, P.:
From the Kalman Filter to the Particle Filter: A geometrical perspective of the curse of dimensionality, Adv. Meteorol.,
2016, 9372786, https://doi.org/10.1155/2016/9372786, 2016. a
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E.:
Scikit-learn: Machine Learning in Python,
J. Mach. Learn. Res.,
12, 2825–2830, 2011. a
Peleg, N., Fatichi, S., Paschalis, A., Molnar, P., and Burlando, P.: An advanced stochastic weather generator for simulating 2-D high-resolution climate variables, J. Adv. Model. Earth Sy., 9, 1595–1627, 2017. a
Reichstein, M., Camps-Valls, G., Stevens, B., Jung, M., Denzler, J., Carvalhais, N., and Prabhat:
Deep learning and process understanding for data-driven Earth system science,
Nature,
566, 195–204, 2019. a
Requena-Mesa, C., Reichstein, M., Mahecha, M., Kraft, B., and Denzler, J.:
Predicting landscapes as seen from space from environmental conditions,
in: IGARSS 2018 – 2018 IEEE International Geoscience and Remote Sensing Symposium, Valencia, Spain, 22–27 July 2018, 1768–1771, IEEE, 2018. a
Scher, S.: Toward data-driven weather and climate forecasting: Approximating a simple general circulation model with deep learning, Geophys. Res. Lett., 45, 12616–12622, 2018. a
Springenberg, J. T., Dosovitskiy, A., Brox, T., and Riedmiller, M.:
Striving for simplicity: The all convolutional net, arXiv [preprint], arXiv:1412.6806, 21 December 2014. a
Vallis, G. K.: Atmospheric and Oceanic Fluid Dynamics, Cambridge University Press, Cambridge, UK, https://doi.org/10.2277/0521849691, 2006. a, b, c
Watson-Parris, D.: Machine learning for weather and climate are worlds apart,
Philos. T. R. Soc. A, 379, 20200098, https://doi.org/10.1098/rsta.2020.0098, 2021. a
Weyn, J. A., Durran, D. R., and Caruana, R.:
Can machines learn to predict weather? Using deep learning to predict gridded 500-hPa geopotential height from historical weather data,
J. Adv. Model. Earth Sy.,
11, 2680–2693, 2019. a
Weyn, J. A., Durran, D. R., and Caruana, R.:
Improving Data-Driven Global Weather Prediction Using Deep Convolutional Neural Networks on a Cubed Sphere,
J. Adv. Model. Earth Sy.,
12, e2020MS002109, https://doi.org/10.1029/2020MS002109, 2020. a
White, T.: Sampling Generative Networks, arXiv [preprint], arXiv:1609.04468, 14 September 2016. a
Wilks, D. S. and Wilby, R. L.:
The weather generation game: a review of stochastic weather models,
Prog. Phys. Geog.,
23, 329–357, 1999. a
Wu, J.-L., Kashinath, K., Albert, A., Chirila, D., Prabhat, and Xiao, H.:
Enforcing statistical constraints in generative adversarial networks for modeling chaotic dynamical systems, J. Comput. Phys., 406, 109209, https://doi.org/10.1016/j.jcp.2019.109209, 2020. a
Yeh, R. A., Chen, C., Yian Lim, T., Schwing, A. G., Hasegawa-Johnson, M., and Do, M. N.:
Semantic image inpainting with deep generative models,
in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017, 5485–5493, 2017. a
Zhang, X.-C., Chen, J., Garbrecht, J., and Brissette, F.:
Evaluation of a weather generator-based method for statistically downscaling non-stationary climate scenarios for impact assessment at a point scale,
T. ASABE,
55, 1745–1756, 2012. a
Short summary
This paper investigates the potential of a type of deep generative neural network to produce realistic weather situations when trained from the climate of a general circulation model. The generator represents the climate in a compact latent space. It is able to reproduce many aspects of the targeted multivariate distribution. Some properties of our method open new perspectives such as the exploration of the extremes close to a given state or how to connect two realistic weather states.
This paper investigates the potential of a type of deep generative neural network to produce...
