Super-Resolution Simulation for Real-Time Prediction of Urban Micrometeorology

Abstract

We propose a super-resolution (SR) simulation system that consists of a physics-based meteorological simulation and an SR method based on a deep convolutional neural network (CNN). The CNN is trained using pairs of high-resolution (HR) and low-resolution (LR) images created from meteorological simulation results for different resolutions so that it can map LR simulation images to HR ones. The proposed SR simulation system, which performs LR simulations, can provide HR prediction results in much shorter operating cycles than those required for corresponding HR simulation prediction system. We apply the SR simulation system to urban micrometeorology, which is strongly affected by buildings and human activity. Urban micrometeorology simulations that need to resolve urban buildings are computationally costly and thus cannot be used for operational real-time predictions even when run on supercomputers. We performed HR micrometeorology simulations on a supercomputer to obtain datasets for training the CNN in the SR method. It is shown that the proposed SR method can be used with a spatial scaling factor of 4 and that it outperforms conventional interpolation methods by a large margin. It is also shown that the proposed SR simulation system has the potential to be used for operational urban micrometeorology predictions.

Adversarial super-resolution of climatological wind and solar data

“…Deep learning-based SR has been performed on various remote sensing datasets such as satellite imagery (32) and sea surface temperature measurements (33). Additionally, these techniques have been used to enhance simulation outputs such as data from models for heat prediction in urban areas (34) and simulated turbulence data (35,36). For climatological data fields, deep CNNs have successfully increased the resolution of short-term regional precipitation forecasts by 5× (37) and 8× (38).…”

Machine-learning-based spatio-temporal super resolution reconstruction of turbulent flows

“…We demonstrated that high-resolution two-dimensional turbulent flow fields of a grid can be reconstructed from the input data on a coarse grid via machine learning methods. Applications and extensions of SR reconstruction can be considered for not only computational (Onishi, Sugiyama & Matsuda 2019; Liu et al. 2020) but also experimental fluid dynamics (Deng et al.…”

“…We demonstrated that high-resolution two-dimensional turbulent flow fields of a 128 × 128 grid can be reconstructed from the input data on a coarse 4 × 4 grid via machine learning methods. Applications and extensions of SR reconstruction can be considered for not only computational (Onishi, Sugiyama & Matsuda 2019;Liu et al 2020) but also experimental fluid dynamics (Deng et al 2019;Morimoto, Fukami & Fukagata 2020). Although these attempts showed great potential of machine-learning-based SR methods to handle high-resolved fluid big data efficiently, their applicability has been so far limited only to two-dimensional spatial reconstruction.…”

SATREPS Project for Sri Lanka with Regard to “Development of Early Warning Technology of Rain-Induced Rapid and Long-Travelling Landslides”

“…6 requires huge computer resources, and thus was conducted on a super-computer at JAMSTEC with the upper-level computation nest covering a 200 km by 200 km area of the island and the lower-level nest at resolution of 500 m by 500 m. To make this simulation possible even on a workstation, parameters for the MSSG climate model are to be adjusted for better interpolation of weather at resolution of 2 km by 2 km. The obtained prediction at 2 km resolution will be spatially interpolated into 500 m resolution prediction with the aid of the super-resolution mapping technology based on the deep convolutional neural network (Onishi et al 2019 ). This will realize the 24 h-in-advance reliable prediction of heavy rainfalls on a common workstation.…”