TemperatureGAN: Generative Modeling of Regional Atmospheric Temperatures

TL;DR

TemperatureGAN is a novel generative adversarial network designed to produce high-resolution, realistic regional atmospheric temperature data conditioned on time and location, aiding climate impact assessments.

Contribution

The paper introduces TemperatureGAN, a new GAN model that accurately generates hourly regional temperature data conditioned on various factors, with specialized evaluation metrics.

Findings

Produces high-fidelity temperature samples

Captures spatial and temporal temperature patterns

Maintains realistic diurnal cycles

Abstract

Stochastic generators are useful for estimating climate impacts on various sectors. Projecting climate risk in various sectors, e.g. energy systems, requires generators that are accurate (statistical resemblance to ground-truth), reliable (do not produce erroneous examples), and efficient. Leveraging data from the North American Land Data Assimilation System, we introduce TemperatureGAN, a Generative Adversarial Network conditioned on months, locations, and time periods, to generate 2m above ground atmospheric temperatures at an hourly resolution. We propose evaluation methods and metrics to measure the quality of generated samples. We show that TemperatureGAN produces high-fidelity examples with good spatial representation and temporal dynamics consistent with known diurnal cycles.