Application of Deep Learning to Estimate Atmospheric Gravity Wave Parameters in Reanalysis Data Sets

• Published in: Geophysical research letters Vol. 47; no. 19
• Main Authors: Matsuoka, D., Watanabe, S., Sato, K., Kawazoe, S., Yu, W., Easterbrook, S.
• Format: Journal Article
• Language: English
• Published: Washington John Wiley & Sons, Inc 16.10.2020
• Subjects: Artificial neural networks; Atmosphere; Atmospheric circulation; Atmospheric data; Atmospheric flows; Atmospheric gravity waves; Atmospheric sciences; Climate prediction; Computational efficiency; Computer applications; Computing costs; convolutional neural network; Data; Datasets; Deep learning; Distribution; Fluxes; gravity wave; Gravity waves; Humidity; Lower stratosphere; machine learning; Meridional wind; Middle atmosphere; Momentum; Momentum flux; Momentum transfer; Neural networks; Orographic gravity waves; Parameter estimation; Parameters; Physical properties; Specific humidity; Stratosphere; Temperature; Training; Wave parameters; Wavelengths; Weather; Winds
• ISSN: 0094-8276; 1944-8007
• DOI: 10.1029/2020GL089436

Gravity waves play an essential role in driving and maintaining global circulation. To understand their contribution in the atmosphere, the accurate reproduction of their distribution is important. Thus, a deep learning approach for the estimation of gravity wave momentum fluxes was proposed, and its performance at 100 hPa was tested using data from low‐resolution zonal and meridional winds, temperature, and specific humidity at 300, 700, and 850 hPa in the Hokkaido region (Japan). To this end, a deep convolutional neural network was trained on 29‐year reanalysis data sets (JRA‐55 and DSJRA‐55), and the final 5‐year data were reserved for evaluation. The results showed that the fine‐scale momentum flux distribution of the gravity waves could be estimated at a reasonable computational cost. Particularly, in winter, when gravity waves are stronger, the median root means square errors (RMSEs) of the maximum momentum flux and the characteristic zonal wavenumber were 0.06–0.13 mPa and 1.0 × 10−5, respectively. Plain Language Summary Deep learning has been proven to be a powerful tool in the atmospheric sciences and in weather and climate prediction applications. In this study, deep learning was used to obtain the physical parameters of fine‐scale orographic gravity waves in the lower stratosphere (~18 km), which propagate significant momentum in the middle atmosphere (10–100 km), based on large‐scale low‐level (1–9 km) atmospheric flows, temperature, and humidity. By training a convolutional neural network using a 29‐year atmospheric reanalysis data set, the large‐scale inputs were well down‐scaled into fine‐scale gravity wave parameters at a reasonable computational cost. Key Points A deep learning approach was proposed to estimate orographic gravity waves using 29‐year reanalysis data Gravity wave momentum fluxes at 100 hPa were directly converted from lower atmospheric data with a spatial resolution of 60 km Using the proposed method, wave structures of the strong momentum flux in the target area could be estimated quite well