PACER: Physics Informed Uncertainty Aware Climate Emulator

2024-10-29

arXiv.org

AI Summary

Traditional numerical climate models suffer from high computational cost, while deep learning approaches often lack long-term stability in climate simulation. To address these challenges, this paper proposes a lightweight (684K-parameter) physics-informed climate simulator. Our method is the first to embed the advection–diffusion equation directly into a neural network architecture, jointly modeling boundary conditions and empirically estimating key physical parameters from greenhouse gas emission data. We further integrate uncertainty quantification via DropPath, multi-source ensemble training using 15 CMIP models (ClimateSet), and boundary-constrained optimization. The resulting model generates stable 86-year temperature and precipitation projections solely from emission inputs. It outperforms all baselines on climate diagnostic tasks, achieves new state-of-the-art performance on most training models, and accelerates inference by over 1000×—significantly lowering the computational barrier for long-term climate scenario analysis.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Learning to Search

Application Category

• 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
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Climate models serve as critical tools for evaluating the effects of climate change and projecting future climate scenarios. However, the reliance on numerical simulations of physical equations renders them computationally intensive and inefficient. While deep learning methodologies have made significant progress in weather forecasting, they are still unstable for climate emulation tasks. Here, we propose PACER, a lightweight 684K parameter Physics Informed Uncertainty Aware Climate Emulator. PACER emulates temperature and precipitation stably for 86 years while only being trained on greenhouse gas emissions data. We incorporate a fundamental physical law of advection-diffusion in PACER accounting for boundary conditions and empirically estimating the diffusion co-efficient and flow velocities from emissions data. PACER has been trained on 15 climate models provided by ClimateSet outperforming baselines across most of the climate models and advancing a new state of the art in a climate diagnostic task.

Problem

• Emulates climate variables using minimal parameters and emissions data
• Incorporates physical laws to ensure stable long-term climate projections
• Addresses computational inefficiency of traditional climate modeling approaches

Innovation

• Lightweight emulator with 684K parameters
• Incorporates advection-diffusion physical law
• Estimates diffusion coefficients from emissions data