Emerging AI-based weather prediction models as downscaling tools

Nikolay Koldunov,Thomas Rackow,Christian Lessig,Sergey Danilov,Suvarchal K. Cheedela,Dmitry Sidorenko,Irina Sandu,Thomas Jung

2024-06-26

Abstract:The demand for high-resolution information on climate change is critical for accurate projections and decision-making. Presently, this need is addressed through high-resolution climate models or downscaling. High-resolution models are computationally demanding and creating ensemble simulations with them is typically prohibitively expensive. Downscaling methods are more affordable but are typically limited to small regions. This study proposes the use of existing AI-based numerical weather prediction systems (AI-NWP) to perform global downscaling of climate information from low-resolution climate models. Our results demonstrate that AI-NWP initalized from low-resolution initial conditions can develop detailed forecasts closely resembling the resolution of the training data using a one day lead time. We constructed year-long atmospheric fields using AI-NWP forecasts initialized from smoothed ERA5 and low-resolution CMIP6 models. Our analysis for 2-metre temperature indicates that AI-NWP can generate high-quality, long-term datasets and potentially perform bias correction, bringing climate model outputs closer to observed data. The study highlights the potential for off-the-shelf AI-NWP to enhance climate data downscaling, offering a simple and computationally efficient alternative to traditional downscaling techniques. The downscaled data can be used either directly for localized climate information or as boundary conditions for further dynamical downscaling.

Atmospheric and Oceanic Physics

What problem does this paper attempt to address?

The problem this paper attempts to address is the issue of climate data downscaling. Specifically, the researchers propose a method for global climate data downscaling using an existing AI-based numerical weather prediction system (AI-NWP). Traditional high-resolution climate models, while accurate, are computationally expensive; traditional downscaling methods are usually limited to small regions. Therefore, the research team hypothesizes that by using AI-NWP models initialized with low-resolution initial conditions, it is possible to generate detailed forecasts similar in resolution to the training data, thereby achieving automatic downscaling. The main objectives of the paper include: 1. **Evaluate whether AI-NWP can generate outputs with higher levels of detail under low-resolution initial conditions**: By using smoothed ERA5 data as initial conditions, verify whether AI-NWP can generate high-resolution outputs similar to the original ERA5 data. 2. **Construct high-quality, long-term atmospheric field records**: By combining a series of AI-NWP forecast results initialized from low-resolution data, form a year-long atmospheric field record and assess the quality of these records. 3. **Use low-resolution CMIP6 model data for historical and future scenario simulations**: Verify the performance of AI-NWP when processing low-resolution CMIP6 model data and evaluate its applicability across different time scales. Through these experiments, the researchers hope to demonstrate that AI-NWP can serve as a simple and computationally efficient alternative to traditional downscaling techniques, thereby providing more accurate data support for regional climate assessments.

Emerging AI-based weather prediction models as downscaling tools

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

A two-step downscaling method for high-scale super-resolution of daily temperature — a case study of Wei River Basin, China

OMG-HD: A High-Resolution AI Weather Model for End-to-End Forecasts from Observations

MetaSD: A Unified Framework for Scalable Downscaling of Meteorological Variables in Diverse Situations

Generative Diffusion-based Downscaling for Climate

Investigating transformer‐based models for spatial downscaling and correcting biases of near‐surface temperature and wind‐speed forecasts

Climate Variable Downscaling with Conditional Normalizing Flows

Accelerating Community-Wide Evaluation of AI Models for Global Weather Prediction by Facilitating Access to Model Output

A Climate Downscaling Deep Learning Model Considering the Multiscale Spatial Correlations and Chaos of Meteorological Events

Using a Two-Step Downscaling Method to Assess the Impact of Climate Change on Total Nitrogen Load in a Small Basin

Downscaling long lead time daily rainfall ensemble forecasts through deep learning

DSAF: A Dual-Stage Adaptive Framework for Numerical Weather Prediction Downscaling

Repeatable high-resolution statistical downscaling through deep learning

Interpolation-Free Deep Learning for Meteorological Downscaling on Unaligned Grids Across Multiple Domains with Application to Wind Power

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

Data-Driven Weather Forecasting and Climate Modeling from the Perspective of Development

Dynamical-generative downscaling of climate model ensembles

Machine learning-based spatial downscaling and bias-correction framework for high-resolution temperature forecasting

Advancing Data-driven Weather Forecasting: Time-Sliding Data Augmentation of ERA5

Self Supervised Vision for Climate Downscaling