Digitala Vetenskapliga Arkivet

Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Artificial intelligence in weather and climate prediction: Learning atmospheric dynamics
Stockholm University, Faculty of Science, Department of Meteorology .ORCID iD: 0000-0002-6314-8833
2020 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Weather and climate prediction is dominated by high dimensionality, interactions on many different spatial and temporal scales and chaotic dynamics. This makes many problems in the field quite complex ones, and also state-of-the-art numerical models are - despite their immense computational costs - not sufficient for many applications. Therefore, it is appealing to use emerging new technologies such as artificial intelligence to tackle these problems.

We show that it is possible to use deep neural networks to emulate the full dynamics of a strongly simplified general circulation model, providing both good forecasts of the model state several days ahead as well as stable long-term climate timeseries. This method partly also works on more complex and realistic models, but only for forecasting the model's weather several days ahead, not for creating climate runs. It is sufficient to use 50-100 years of data for training the networks. The same neural network method can be combined with singular value decomposition from numerical ensemble weather forecasting in order to generate probabilistic ensemble forecasts with the neural networks.

On a more fundamental level, we show that in a simple dynamical systems setting there seem to be limitations in the ability of feed-forward neural networks to generalize to new regions of the system. This is caused by different parts of the network learning to model different parts of the system. Contradictory, for another simple dynamical system this is shown not to be an issue, raising doubts on the usefulness of results from simple models in the context of more complex ones. Additionally, we show that neural networks are to some extent able to “learn” the influence of slowly changing external forcings on the dynamics of the system, but only given broad enough forcing regimes.

Finally, we present a method to complement operational weather forecasts. Given the initial fields and the error of past weather forecasts, a neural network is used to predict the uncertainty in new forecasts, given only the initial field of the new forecast.

Place, publisher, year, edition, pages
Stockholm: Department of Meteorology, Stockholm University , 2020. , p. 30
National Category
Meteorology and Atmospheric Sciences
Research subject
Atmospheric Sciences and Oceanography
Identifiers
URN: urn:nbn:se:su:diva-180877ISBN: 978-91-7911-128-1 (print)ISBN: 978-91-7911-129-8 (electronic)OAI: oai:DiVA.org:su-180877DiVA, id: diva2:1425352
Public defence
2020-06-12, Vivi Täckholmsalen (Q-salen), Svante Arrhenius väg 20, Stockholm, 10:00 (English)
Opponent
Supervisors
Funder
Swedish Research Council, 2016-03724
Note

At the time of the doctoral defense, the following paper was unpublished and had a status as follows: Paper 4: Manuscript.

Available from: 2020-05-18 Created: 2020-04-20 Last updated: 2020-05-26Bibliographically approved
List of papers
1. Toward Data-Driven Weather and Climate Forecasting: Approximating a Simple General Circulation Model With Deep Learning
Open this publication in new window or tab >>Toward Data-Driven Weather and Climate Forecasting: Approximating a Simple General Circulation Model With Deep Learning
2018 (English)In: Geophysical Research Letters, ISSN 0094-8276, E-ISSN 1944-8007, Vol. 45, no 22, p. 12616-12622Article in journal (Refereed) Published
Abstract [en]

It is shown that it is possible to emulate the dynamics of a simple general circulation model with a deep neural network. After being trained on the model, the network can predict the complete model state several time steps aheadwhich conceptually is making weather forecasts in the model world. Additionally, after being initialized with an arbitrary model state, the network can through repeatedly feeding back its predictions into its inputs create a climate run, which has similar climate statistics to the climate of the general circulation model. This network climate run shows no long-term drift, even though no conservation properties were explicitly designed into the network. Plain Language Summary Numerical weather prediction and climate models are complex computer programs that represent the physics of the atmosphere. They are essential tools for predicting the weather and for studying the Earth's climate. Recently, a lot of progress has been made in machine learning methods. These are data-driven algorithms that learn from existing data. We show that it is possible that such an algorithm learns the dynamics of a simple climate model. After being presented with enough data from the climate model, the network can successfully predict the time evolution of the model's state, thus replacing the dynamics of the model. This finding is an important step toward purely data-driven weather forecastingthus weather forecasting without the use of traditional numerical models and also opens up new possibilities for climate modeling.

Keywords
machine learning, weather prediction, neural networks, deep learning, climate models
National Category
Earth and Related Environmental Sciences
Research subject
Atmospheric Sciences and Oceanography
Identifiers
urn:nbn:se:su:diva-163610 (URN)10.1029/2018GL080704 (DOI)000453250000058 ()
Available from: 2019-01-09 Created: 2019-01-09 Last updated: 2020-04-27Bibliographically approved
2. Weather and climate forecasting with neural networks: using general circulation models (GCMs) with different complexity as a study ground
Open this publication in new window or tab >>Weather and climate forecasting with neural networks: using general circulation models (GCMs) with different complexity as a study ground
2019 (English)In: Geoscientific Model Development, ISSN 1991-959X, E-ISSN 1991-9603, Vol. 12, no 7, p. 2797-2809Article in journal (Refereed) Published
Abstract [en]

Recently, there has been growing interest in the possibility of using neural networks for both weather forecasting and the generation of climate datasets. We use a bottom-up approach for assessing whether it should, in principle, be possible to do this. We use the relatively simple general circulation models (GCMs) PUMA and PLASIM as a simplified reality on which we train deep neural networks, which we then use for predicting the model weather at lead times of a few days. We specifically assess how the complexity of the climate model affects the neural network's forecast skill and how dependent the skill is on the length of the provided training period. Additionally, we show that using the neural networks to reproduce the climate of general circulation models including a seasonal cycle remains challenging - in contrast to earlier promising results on a model without seasonal cycle.

National Category
Earth and Related Environmental Sciences
Research subject
Atmospheric Sciences and Oceanography
Identifiers
urn:nbn:se:su:diva-171750 (URN)10.5194/gmd-12-2797-2019 (DOI)000474740000001 ()
Available from: 2019-09-04 Created: 2019-09-04 Last updated: 2020-04-27Bibliographically approved
3. Generalization properties of feed-forward neural networks trained on Lorenz systems
Open this publication in new window or tab >>Generalization properties of feed-forward neural networks trained on Lorenz systems
2019 (English)In: Nonlinear processes in geophysics, ISSN 1023-5809, E-ISSN 1607-7946, Vol. 26, no 4, p. 381-399Article in journal (Refereed) Published
Abstract [en]

Neural networks are able to approximate chaotic dynamical systems when provided with training data that cover all relevant regions of the system's phase space. However, many practical applications diverge from this idealized scenario. Here, we investigate the ability of feed-forward neural networks to (1) learn the behavior of dynamical systems from incomplete training data and (2) learn the influence of an external forcing on the dynamics. Climate science is a real-world example where these questions may be relevant: it is concerned with a non-stationary chaotic system subject to external forcing and whose behavior is known only through comparatively short data series. Our analysis is performed on the Lorenz63 and Lorenz95 models. We show that for the Lorenz63 system, neural networks trained on data covering only part of the system's phase space struggle to make skillful short-term forecasts in the regions excluded from the training. Additionally, when making long series of consecutive forecasts, the networks struggle to reproduce trajectories exploring regions beyond those seen in the training data, except for cases where only small parts are left out during training. We find this is due to the neural network learning a localized mapping for each region of phase space in the training data rather than a global mapping. This manifests itself in that parts of the networks learn only particular parts of the phase space. In contrast, for the Lorenz95 system the networks succeed in generalizing to new parts of the phase space not seen in the training data. We also find that the networks are able to learn the influence of an external forcing, but only when given relatively large ranges of the forcing in the training. These results point to potential limitations of feed-forward neural networks in generalizing a system's behavior given limited initial information. Much attention must therefore be given to designing appropriate train-test splits for real-world applications.

National Category
Earth and Related Environmental Sciences
Research subject
Atmospheric Sciences and Oceanography
Identifiers
urn:nbn:se:su:diva-176560 (URN)10.5194/npg-26-381-2019 (DOI)000495389400001 ()
Available from: 2019-12-17 Created: 2019-12-17 Last updated: 2020-04-27Bibliographically approved
4. Ensemble neural network forecasts with singular value decomposition
Open this publication in new window or tab >>Ensemble neural network forecasts with singular value decomposition
(English)Manuscript (preprint) (Other academic)
Abstract [en]

Ensemble weather forecasts enable a measure of uncertainty to be attached to each forecast by computing the ensemble's spread. However, generating an ensemble with a good error-spread relationship is far from trivial, and a wide range of approaches to achieve this have been explored. Random perturbations of the initial model state typically provide unsatisfactory results when applied to numerical weather prediction models. Singular value decomposition has proved more successful in this context, and as a result has been widely used for creating perturbed initial states of weather prediction models. We demonstrate how to apply the technique of singular value decomposition to purely neural-network based forecasts. Additionally, we explore the use of random initial perturbations for neural network ensembles, and the creation of neural network ensembles via retraining the network. We find that the singular value decomposition results in ensemble forecasts that have some probabilistic skill, but are inferior to the ensemble created by retraining the neural network several times. Compared to random initial perturbations, the singular value technique performs better when forecasting a simple general circulation model, comparably when forecasting atmospheric reanalysis data, and worse when forecasting the lorenz95 system - a highly idealized model designed to mimic certain aspects of the mid-latitude atmosphere.

National Category
Meteorology and Atmospheric Sciences
Research subject
Atmospheric Sciences and Oceanography
Identifiers
urn:nbn:se:su:diva-180872 (URN)
Available from: 2020-04-17 Created: 2020-04-17 Last updated: 2020-04-27
5. Predicting weather forecast uncertainty with machine learning
Open this publication in new window or tab >>Predicting weather forecast uncertainty with machine learning
2018 (English)In: Quarterly Journal of the Royal Meteorological Society, ISSN 0035-9009, E-ISSN 1477-870X, Vol. 144, no 717, p. 2830-2841Article in journal (Refereed) Published
Abstract [en]

Weather forecasts are inherently uncertain. Therefore, for many applications forecasts are only considered valuable if an uncertainty estimate can be assigned to them. Currently, the best method to provide a confidence estimate for individual forecasts is to produce an ensemble of numerical weather simulations, which is computationally very expensive. Here, we assess whether machine learning techniques can provide an alternative approach to predict the uncertainty of a weather forecast given the large-scale atmospheric state at initialization. We propose a method based on deep learning with artificial convolutional neural networks that is trained on past weather forecasts. Given a new weather situation, it assigns a scalar value of confidence to medium-range forecasts initialized from the said atmospheric state, indicating whether the predictability is higher or lower than usual for the time of the year. While our method has a lower skill than ensemble weather forecast models in predicting forecast uncertainty, it is computationally very efficient and outperforms a range of alternative methods that do not involve performing numerical forecasts. This shows that it is possible to use machine learning in order to estimate future forecast uncertainty from past forecasts. The main constraint in the performance of our method seems to be the number of past forecasts available for training the machine learning algorithm.

Keywords
ensembles, machine learning, statistical methods, weather forecasts
National Category
Earth and Related Environmental Sciences
Research subject
Atmospheric Sciences and Oceanography
Identifiers
urn:nbn:se:su:diva-165737 (URN)10.1002/qj.3410 (DOI)000455586500029 ()
Available from: 2019-02-06 Created: 2019-02-06 Last updated: 2020-04-27Bibliographically approved

Open Access in DiVA

Artificial intelligence in weather and climate prediction: Learning atmospheric dynamics(1000 kB)33 downloads
File information
File name FULLTEXT01.pdfFile size 1000 kBChecksum SHA-512
4050ee5f9c784034202991934ba89a20d3fdfefce4955bb2ac910b59fa5c52ec511319612d10b7b1002c6eb2ac5a8bc25a906b6219d1fe5b7fc46f3c5aa3c6de
Type fulltextMimetype application/pdf

Search in DiVA

By author/editor
Scher, Sebastian
By organisation
Department of Meteorology
Meteorology and Atmospheric Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 33 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

isbn
urn-nbn

Altmetric score

isbn
urn-nbn
Total: 62 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf