Skillful Twelve Hour Precipitation Forecasts using Large Context Neural Networks

Espeholt, Lasse; Agrawal, Shreya; Sønderby, Casper Kaae; Kumar, M.; Jonathan, Heek,; Bromberg, Carla; Gazen, Cenk; Hickey, Jason; Bell, Aaron; Kalchbrenner, Nal

doi:10.48550/arxiv.2111.07470

Cited by 26 publications

(36 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Hybrid Learning (HyL), or gray box modelling as called in its early days in the 90's (Psichogios & Ungar, 1992;Rico-Martinez et al, 1994;Thompson & Kramer, 1994;Rivera-Sampayo & Vélez-Reyes, 2001;Braun & Chaturvedi, 2002), has been an appropriate method to learn models that are both expressive and interpretable, while also allowing them to be learnt on fewer data. The interest for HyL (Mehta et al, 2020;Lei & Mirams, 2021;Reichstein et al, 2019;Saha et al, 2020;Guen & Thome, 2020;Levine & Stuart, 2021;Espeholt et al, 2021) has greatly renewed since the outbreak of recent neural network architectures that simplify the combination of physical equations within ML models. As an example, Neural ODE (Chen et al, 2018) and convolutional neural networks (LeCun et al, 1995, CNN) are privileged architectures to work with dynamical systems described by ODEs or PDEs.…”

Section: Hybrid Modellingmentioning

confidence: 99%

Robust Hybrid Learning With Expert Augmentation

Wehenkel¹,

Behrmann²,

Hsu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Hybrid modelling reduces the misspecification of expert models by combining them with machine learning (ML) components learned from data. Like for many ML algorithms, hybrid model performance guarantees are limited to the training distribution. Leveraging the insight that the expert model is usually valid even outside the training domain, we overcome this limitation by introducing a hybrid data augmentation strategy termed expert augmentation. Based on a probabilistic formalization of hybrid modelling, we show why expert augmentation improves generalization. Finally, we validate the practical benefits of augmented hybrid models on a set of controlled experiments, modelling dynamical systems described by ordinary and partial differential equations.

show abstract

Section: Hybrid Modellingmentioning

confidence: 99%

Robust Hybrid Learning With Expert Augmentation

Wehenkel¹,

Behrmann²,

Hsu³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…This is especially relevant in our application, where we use global gridded data and try to directly incorporate and model existing global teleconnections with the potential of uncovering new ones. The importance of having a wider field-of-view was found critical in Espeholt et al 13 using dilated convolutions, but deformable convolutions are actually a generalization of dilated convolutions 19 .…”

Section: Deformable Convolutional Neural Networkmentioning

confidence: 99%

“…We use BO for each lead time independently, meaning we will potentially have a unique set of hyperparameters for each lead time. We could also have trained an aggregate model for all lead times with appropriate lead time encoding similar to other studies 13 . However, this tends to yield similar performance 13 and is mostly relevant when training time is a bottleneck.…”

Section: /12mentioning

confidence: 99%

“…We could also have trained an aggregate model for all lead times with appropriate lead time encoding similar to other studies 13 . However, this tends to yield similar performance 13 and is mostly relevant when training time is a bottleneck. This is not the case in our framework, as considerable parallelization is possible, and training time is around 30 minutes per lead time, including pretraining, on two Nvidia RTX 2080-RTI GPUs.…”

Section: /12mentioning

confidence: 99%

“…This success is largely attributed to deep learning models being able to effectively learn patterns in highdimensional data rather than being explicitly formulated by domain experts. This has enabled advances in, for example, weather forecasting 11,13 , parameterization of convection 14 , and detection of tropical cyclones and synoptic-scale fronts 15,16 .…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Forecasting large-scale circulation regimes using deformable convolutional neural networks and global spatiotemporal climate data

Nielsen¹,

Iosifidis²,

Karstoft³

2022

Preprint

View full text Add to dashboard Cite

Classifying the state of the atmosphere into a finite number of large-scale circulation regimes is a popular way of investigating teleconnections, the predictability of severe weather events, and climate change. Here, we investigate a supervised machine learning approach based on deformable convolutional neural networks (deCNNs) and transfer learning to forecast the North Atlantic-European weather regimes during extended boreal winter for 1 to 15 days into the future. We apply state-ofthe-art interpretation techniques from the machine learning literature to attribute particular regions of interest or potential teleconnections relevant for any given weather cluster prediction or regime transition. We demonstrate superior forecasting performance relative to several classical meteorological benchmarks, as well as logistic regression and random forests. Due to its wider field of view, we also observe deCNN achieving considerably better performance than regular convolutional neural networks at lead times beyond 5-6 days. Finally, we find transfer learning to be of paramount importance, similar to previous data-driven atmospheric forecasting studies.

show abstract

Correcting rainfall forecasts of a numerical weather prediction model using generative adversarial networks

Jeong

2022

J Supercomput

View full text Add to dashboard Cite

Skillful Twelve Hour Precipitation Forecasts using Large Context Neural Networks

Cited by 26 publications

References 0 publications

Robust Hybrid Learning With Expert Augmentation

Robust Hybrid Learning With Expert Augmentation

Forecasting large-scale circulation regimes using deformable convolutional neural networks and global spatiotemporal climate data

Correcting rainfall forecasts of a numerical weather prediction model using generative adversarial networks

Contact Info

Product

Resources

About