Delay-SDE-net: A deep learning approach for time series modelling with memory and uncertainty estimates

Eggen, Mari Dahl; Midtfjord, Alise Danielle

doi:10.48550/arxiv.2303.08587

Cited by 2 publications

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition to the nowcasting methodology applied in this work, a forecasting training methodology is available through the Delay-SDE-net framework (Eggen & Midtfjord, 2023), and these capabilities might be utilized to forecast the stratospheric circulation in future studies. Further numerical experiments can also be conducted to assess to what extent stratospheric wind forecasting benefits from being informed with infrasound data in addition to the present and past stratospheric wind and temperature.…”

Section: Discussionmentioning

confidence: 99%

“…A novel methodology for elaborating such frameworks using neural networks is derived by Eggen and Midtfjord (2023), where Delay-SDE-net is presented. The Delay-SDE-net model can be considered as a discrete-time SDDE where the model coefficients are independent neural networks.…”

Section: Arraymentioning

confidence: 99%

“…In the current study, a first attempt is made to provide a robust stratospheric polar vortex diagnostic using high-latitude infrasound data. These data are used to train a stochastic-based machine-learning model called Delay-SDE-net (Eggen & Midtfjord, 2023). This is a novel neural network model using a stochastic differential equation (SDE) in a time-series modeling framework.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Estimating stratospheric polar vortex strength using ambient ocean‐generated infrasound and stochastics‐based machine learning

Vorobeva,

Eggen,

Midtfjord

et al. 2024

Quart J Royal Meteoro Soc

View full text Add to dashboard Cite

There are sparse opportunities for direct measurement of upper stratospheric winds, yet improving their representation in subseasonal‐to‐seasonal prediction models can have significant benefits. There is solid evidence from previous research that global atmospheric infrasound waves are sensitive to stratospheric dynamics. However, there is a lack of results providing a direct mapping between infrasound recordings and polar‐cap upper stratospheric winds. The global International Monitoring System (IMS), which monitors compliance with the Comprehensive Nuclear‐Test‐Ban Treaty, includes ground‐based stations that can be used to characterize the infrasound soundscape continuously. In this study, multi‐station IMS infrasound data were utilized along with a machine‐learning supported stochastic model, Delay‐SDE‐net, to demonstrate how a near‐real‐time estimate of the polar‐cap averaged zonal wind at 1‐hPa pressure level can be found from infrasound data. The infrasound was filtered to a temporal low‐frequency regime dominated by microbaroms, which are ambient‐noise infrasonic waves continuously radiated into the atmosphere from nonlinear interaction between counter‐propagating ocean surface waves. Delay‐SDE‐net was trained on 5 years (2014–2018) of infrasound data from three stations and the ERA5 reanalysis 1‐hPa polar‐cap averaged zonal wind. Using infrasound in 2019–2020 for validation, we demonstrate a prediction of the polar‐cap averaged zonal wind, with an error standard deviation of around 12 m·s compared with ERA5. These findings highlight the potential of using infrasound data for near‐real‐time measurements of upper stratospheric dynamics. A long‐term goal is to improve high‐top atmospheric model accuracy, which can have significant implications for weather and climate prediction.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Arraymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Estimating stratospheric polar vortex strength using ambient ocean‐generated infrasound and stochastics‐based machine learning

Vorobeva,

Eggen,

Midtfjord

et al. 2024

Quart J Royal Meteoro Soc

View full text Add to dashboard Cite

show abstract

“…Although traditional inversion procedures (e.g., gradient-based optimization) have not been used to map microbarom recordings to atmospheric winds, Vorobeva et al [28] developed an innovative Machine Learning (ML) based approach to determine zonal polar-cap stratospheric winds based on microbarom infrasound data [28]. This model trains a neural network [29] to estimate the coefficients of Stochastic Delay Differential Equations along with estimates of the aleatoric and epistemic uncertainties of the model output. Quantifying uncertainties is key not only for inversion but also for interpreting "black-box" ML outputs.…”

Section: Stratospheric Inversionmentioning

confidence: 99%

Probing a planet from the subsurface to the atmosphere with infrasound data

Froment,

Brissaud,

Näsholm

et al. 2024

Preprint

View full text Add to dashboard Cite

The analysis of infrasound waves has a significant potential for retrieving a range of geophysical parameters across various scales, such as atmospheric structures, characteristics of surface and buried sources, and seismic velocity structures. This potential was recently showcased in infrasound studies that illustrated the ability to retrieve Earth’s shallow velocities and Mars’ near-surface atmospheric winds from remote acoustic observations. Consequently, infrasound is becoming an essential component of planetary missions, providing insights into the interiors and atmospheres of other terrestrial planets in the solar system. However, utilizing infrasound data requires efficient forward wave simulation techniques and an accurate description of waveform sensitivity to source and medium parameters. Even under ideal conditions, many inverse problems associated with infrasound-based probing are inherently ill-posed, necessitating regularization or other approaches to ensure reliable results. In this contribution, we highlight recent seismo-acoustic research findings, addressing atmospheric probing at both smaller and global scales, as well as innovative methods to accelerate infrasonic wave propagation modeling.

show abstract

Delay-SDE-net: A deep learning approach for time series modelling with memory and uncertainty estimates

Cited by 2 publications

References 0 publications

Estimating stratospheric polar vortex strength using ambient ocean‐generated infrasound and stochastics‐based machine learning

Estimating stratospheric polar vortex strength using ambient ocean‐generated infrasound and stochastics‐based machine learning

Probing a planet from the subsurface to the atmosphere with infrasound data

Contact Info

Product

Resources

About