Abstract:Addressing the challenges of climate change requires accurate and high-resolution mapping of geospatial data, especially climate and weather variables. However, many existing geospatial datasets, such as the gridded outputs of the state-of-the-art numerical climate models (e.g., general circulation models), are only available at very coarse spatial resolutions due to the model complexity and extremely high computational demand. Deep-learning-based methods, particularly generative adversarial networks (GANs) and their variants, have proved effective for refining natural images and have shown great promise in improving geospatial datasets. This paper describes a conditional GAN-based stochastic geospatial downscaling method that can accommodates very high scaling factors. Compared to most existing methods, the method can generate high-resolution accurate climate datasets from very low-resolution inputs. More importantly, the method explicitly considers the uncertainty inherent to the downscaling process that tends to be ignored in existing methods. Given an input, the method can produce a multitude of plausible high-resolution samples instead of one single deterministic result. These samples allow for an empirical exploration and inferences of model uncertainty and robustness. With a case study of gridded climate datasets (wind velocity and solar irradiance), we demonstrate the performances of the framework in downscaling tasks with large scaling factors (up to $64\times$) and highlight the advantages of the framework with a comprehensive comparison with commonly used and most recent downscaling methods, including area-to-point (ATP) kriging, deep image prior (DIP), enhanced super-resolution generative adversarial networks (ESRGAN), physics-informed resolution-enhancing GAN (PhIRE GAN), and an efficient diffusion model for remote sensing image super-resolution (EDiffSR).

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to improve the resolution of geospatial data, especially for high - resolution mapping of climate and weather variables. Many existing geospatial data sets, such as the gridded outputs of state - of - the - art numerical climate models (e.g., general circulation models), can only be provided at very coarse spatial resolutions due to model complexity and extremely high computational requirements. Therefore, effective downscaling methods are needed to improve these gridded climate data sets and other scientific data sets. Specifically, this study proposes a conditional generative adversarial network (CGAN) framework based on the generative adversarial network (GAN) for extreme geospatial downscaling tasks. This framework can handle very high downscaling factors (e.g., up to 64 times) and explicitly considers uncertainty during the downscaling process. Unlike most existing methods, this method can generate multiple possible high - resolution samples from low - resolution inputs, rather than a single deterministic result. This allows for empirical exploration and inference of the model's uncertainty and robustness. In addition, this study also demonstrates its performance advantages in handling tasks with large downscaling factors through a comprehensive comparison with a series of commonly used downscaling methods, including area - to - point kriging (ATP kriging), deep image prior (DIP), enhanced super - resolution generative adversarial network (ESRGAN), physics - informed resolution - enhanced GAN (PhIRE GAN), and efficient diffusion model for super - resolution (EDiffSR).

Generative Adversarial Models for Extreme Geospatial Downscaling

SpatialGAN: Progressive Image Generation Based on Spatial Recursive Adversarial Expansion

Capturing Climatic Variability: Using Deep Learning for Stochastic Downscaling

Deep generative model super-resolves spatially correlated multiregional climate data

A Generative Deep Learning Approach to Stochastic Downscaling of Precipitation Forecasts

Generative Diffusion-based Downscaling for Climate

On the Extrapolation of Generative Adversarial Networks for downscaling precipitation extremes in warmer climates

A Spatial Weather Generator Based on Conditional Deep Convolution Generative Adversarial Nets (Cdcgan)

Downscaling Extreme Rainfall Using Physical-Statistical Generative Adversarial Learning

Stochastic Super-Resolution for Downscaling Time-Evolving Atmospheric Fields With a Generative Adversarial Network

DeepSD: Generating High Resolution Climate Change Projections through Single Image Super-Resolution

Generating High-Resolution Regional Precipitation Using Conditional Diffusion Model

A Fast Generative Adversarial Network Combined With Transformer for Downscaling GRACE Terrestrial Water Storage Data in Southwestern China

A High-Resolution Land Surface Temperature Downscaling Method Based on Geographically Weighted Neural Network Regression

Statistical Downscaling via High-Dimensional Distribution Matching with Generative Models

Dynamical-generative downscaling of climate model ensembles

High-resolution meteorology with climate change impacts from global climate model data using generative machine learning

Algorithmic Hallucinations of Near-Surface Winds: Statistical Downscaling with Generative Adversarial Networks to Convection-Permitting Scales

Unpaired Satellite Precipitation Downscaling via Multiple Generative Constrains

Hard-Constrained Deep Learning for Climate Downscaling

Global spatio-temporal downscaling of ERA5 precipitation through generative AI