Deep Learning‐Based Super‐Resolution Climate Simulator‐Emulator Framework for Urban Heat Studies

Abstract

Climate change is projected to increase the frequency and intensity of extreme weather events, including heat waves (IPCC, 2013). These projected changes in heat waves coupled with the urban heat island (UHI) effect, as manifested by elevated near-surface air temperatures in urban areas compared to their non-urban surroundings, exposes urban dwellers to additional heat stress. The higher urban temperatures are largely related to thermal and radiative properties of built surfaces, substantially different from its surrounding natural environment, and to a lesser vegetation coverage with limited evaporative cooling. UHI is further enhanced by heat emitted from transportation, heating and air conditioning systems (Oke, 1982). Adaptation and mitigation strategies considered to reduce UHI, which could reduce the impacts during heat waves, include increasing the reflectivity of urban regions and increasing vegetation coverage (Alexandri & Jones, 2008;Costanzo et al., 2016;Touchaei & Akbari, 2013). Development of effective adaptation and mitigation strategies will require information of mean and extreme temperature changes at super-resolution (<250 m). Climate change information at such super-resolution are not available for cities, primarily due to the inadequate representation of urban regions in climate models due to their coarse resolution. Climate processes are complex-more so in urban regions (Bai et al., 2018). Super-resolution climate modeling that includes good representation of urban regions is required to capture the urban-climate feedbacks. With significant developments in high performance computing, it has now become possible to execute regional climate simulations at 4 to 1 km resolutions, but for shorter periods. High computational cost continues to be

Self‐Supervised Classification of Weather Systems Based on Spatiotemporal Contrastive Learning

“…In recent years, deep learning has excelled in tackling high‐dimensional data by virtue of its inherent highly nonlinear transformation characteristics. In order to fully utilize massive‐scale labeled data and capture multi‐scale synoptic features, research has begun to focus on how to integrate deep learning into the meteorological field (Wu et al., 2021; Xing et al., 2021). Traditional supervised learning approaches are heavily reliant on the amount of annotated training data available, while labeling could cost tremendous time and manpower (Camps‐Valls, 2021).…”

Refining time-space traffic diagrams: A multiple linear regression model

“…Convolutional neural networks (CNNs) are the most popular foundation of various deep learning-based supersolution approaches due to their powerful capability of processing images (Wu et al, 2021). Machine learning-based supersolution approaches do not require solving complex equations, and the drawbacks of these approaches are twofold.…”

“…Machine learning-based supersolution approaches do not require solving complex equations, and the drawbacks of these approaches are twofold. First, the computational cost is usually high, hindering the real-time applications of supersolution approaches (Wu et al, 2021). Second, a large number of images are needed for training deep-learning models (Zhang et al, 2022).…”

Passive Tracer Modelling at Super-Resolution with WRF-ARW to Assess Mass-Balance Schemes

“…Generating model simulations of atmospheric processes at high spatial and temporal resolutions (super-resolution) have numerous applications including hybrid physical-model and machine-learning applications (Onishi et al, 2019), the dynamic downscaling of coarse resolution climate and weather information (Watson et al, 2020), and urban-climate feedback studies (Wu et al, 2021). Super-resolution modelling products (∆x < 100 m, ∆t < 1 s) can also provide desirable information at the scale of measurements during top-down campaigns which can be analyzed in conjunction with measurement data to: interpret observations, quantify uncertainty in the measurements, test the validity of assumptions in the employed top-down methodologies, and help fill the information gap in measurements.…”

“…To study these effects through model simulations, the model resolutions should be chosen to resolve dynamical processes (turbulence) at the spatio-temporal scales at which aircraft in-situ measurements are made. For instance, to simulate (and evaluate) in-situ measurements at a flying/sampling speed of 100 m/s (e.g., Conley et al, 2017;Gordon et al, 2015), the model should be able to simulate (and output) atmospheric fields at length and time scales of ∆x ≤ 100 m and ∆t ≤ 1 s. Recent real-case LES-modelling studies have commonly referred to such resolutions (∆x ≤ 250 m) as "super-resolution" (e.g., Wu et al, 2021;Onishi et al, 2019;Watson et al, 2020), herein we use the same terminology to describe our WRF model simulations.…”