Diffusion models for probabilistic precipitation generation from atmospheric variables

2025-04-01

AI Summary

Earth system models (ESMs) suffer from systematic biases in precipitation parameterization, particularly in representing extreme events, largely due to computationally expensive and spatially agnostic column-wise approximations. Method: We propose the first conditional diffusion model framework tailored for ESM precipitation parameterization, integrated with a UNet architecture that generates global 0.25° daily precipitation fields using only a small set of large-scale atmospheric variables. Contribution/Results: The method enables probabilistic, hyperparameter-free, and computationally efficient generation, supporting uncertainty quantification and multi-member ensemble prediction. It preserves interannual variability while significantly improving fidelity in reproducing extreme precipitation. With rapid inference, it facilitates high-throughput climate scenario generation and probabilistic forecasting—achieving a balance between local process realism and global spatial consistency.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Diffusion Models for Vision
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Improving the representation of precipitation in Earth system models (ESMs) is critical for assessing the impacts of climate change and especially of extreme events like floods and droughts. In existing ESMs, precipitation is not resolved explicitly, but represented by parameterizations. These typically rely on resolving approximated but computationally expensive column-based physics, not accounting for interactions between locations. They struggle to capture fine-scale precipitation processes and introduce significant biases. We present a novel approach, based on generative machine learning, which integrates a conditional diffusion model with a UNet architecture to generate accurate, high-resolution (0.25{deg}) global daily precipitation fields from a small set of prognostic atmospheric variables. Unlike traditional parameterizations, our framework efficiently produces ensemble predictions, capturing uncertainties in precipitation, and does not require fine-tuning by hand. We train our model on the ERA5 reanalysis and present a method that allows us to apply it to arbitrary ESM data, enabling fast generation of probabilistic forecasts and climate scenarios. By leveraging interactions between global prognostic variables, our approach provides an alternative parameterization scheme that mitigates biases present in the ESM precipitation while maintaining consistency with its large-scale (annual) trends. This work demonstrates that complex precipitation patterns can be learned directly from large-scale atmospheric variables, offering a computationally efficient alternative to conventional schemes.

Problem

• Generate high-resolution precipitation fields from atmospheric variables
• Mitigate biases in Earth system models' precipitation representation
• Provide computationally efficient alternative to traditional parameterizations

Innovation

• Generative diffusion model for precipitation generation
• UNet architecture for high-resolution global fields
• Ensemble predictions without manual fine-tuning