Graphical Abstract

• Tweet

Kudo, A., 2022: Statistical post-processing for gridded temperature prediction using encoder‒decoder-based deep convolutional neural networks. J. Meteor. Soc. Japan, 100, 219-232.
https://doi.org/10.2151/jmsj.2022-011
Graphical Abstract Published

Plain Language Summary: In this study, an encoder‒decoder-based convolutional neural network (CNN) has been proposed to predict gridded temperatures at the surface around the Kanto region in Japan. Verification results showed that the proposed model greatly improves the Japan Meteorological Agency's (JMA's) operational gridded temperature guidance and can correct NWP model biases, such as a positional error of coastal fronts (Fig. 1) and extreme temperatures.

• Original PDF

Highlights:

• Seven-layer deep CNN model, which inputs seven MSM output variables, was used to predict 2- dimensional temperature field around the Kanto region.
• Verificationresultsusinganindependentdatasetfromtrainingandvalidationdatasetsshowedthatthe proposed CNN model greatly improved the JMA's operational gridded temperature guidance.
• The CNN model can predict temperatures associated with radiative cooling, coastal fronts, and heatwaves, which have been difficult to correct using the operational temperature guidance.