Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Graph Convolutional Neural Networks as Surrogate Models for Climate Simulation

Kevin Potter, Carianne Martinez, Reina Pradhan, Samantha Brozak, Steven Sleder, Lauren Wheeler

TL;DR

The authors' surrogate simulated 80 years in approximately 310 seconds on a single A100 GPU, compared to weeks for the ESM model while having mean temperature errors below $0.1^{\circ}C$ and maximum errors below $2^{\circ}C$.

Abstract

Many climate processes are characterized using large systems of nonlinear differential equations; this, along with the immense amount of data required to parameterize complex interactions, means that Earth-System Model (ESM) simulations may take weeks to run on large clusters. Uncertainty quantification may require thousands of runs, making ESM simulations impractical for preliminary assessment. Alternatives may include simplifying the processes in the model, but recent efforts have focused on using machine learning to complement these models or even act as full surrogates. \textit{We leverage machine learning, specifically fully-connected neural networks (FCNNs) and graph convolutional neural networks (GCNNs), to enable rapid simulation and uncertainty quantification in order to inform more extensive ESM simulations.} Our surrogate simulated 80 years in approximately 310 seconds on a single A100 GPU, compared to weeks for the ESM model while having mean temperature errors below $0.1^{\circ}C$ and maximum errors below $2^{\circ}C$.

Graph Convolutional Neural Networks as Surrogate Models for Climate Simulation

TL;DR

The authors' surrogate simulated 80 years in approximately 310 seconds on a single A100 GPU, compared to weeks for the ESM model while having mean temperature errors below and maximum errors below .

Abstract

Many climate processes are characterized using large systems of nonlinear differential equations; this, along with the immense amount of data required to parameterize complex interactions, means that Earth-System Model (ESM) simulations may take weeks to run on large clusters. Uncertainty quantification may require thousands of runs, making ESM simulations impractical for preliminary assessment. Alternatives may include simplifying the processes in the model, but recent efforts have focused on using machine learning to complement these models or even act as full surrogates. \textit{We leverage machine learning, specifically fully-connected neural networks (FCNNs) and graph convolutional neural networks (GCNNs), to enable rapid simulation and uncertainty quantification in order to inform more extensive ESM simulations.} Our surrogate simulated 80 years in approximately 310 seconds on a single A100 GPU, compared to weeks for the ESM model while having mean temperature errors below and maximum errors below .
Paper Structure (11 sections, 8 figures, 2 tables)

This paper contains 11 sections, 8 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Model architecture
  3. Fully-connected Neural Network (FCNN)
  4. FCNN architecture details
  5. Graph Convolutional Neural Network (GCNN)
  6. GCNN architecture details
  7. Data and training
  8. Results
  9. FCNN
  10. GCNN
  11. Discussion and future work

Figures (8)

  • Figure 1: Diagram illustrating the architecture of an FCNN, showing the full connections between layers.
  • Figure 2: Left: 2-D map projection of altitudes. Right: graph-structured downsampled map (land only).
  • Figure 3: Diagram illustrating the architecture of a UNet for GCNNs.
  • Figure 4: Left: FCNN prediction for the difference in TSA between a control and feedback run for one month (January 2080). Center: uncalibrated uncertainty estimates for FCNN prediction. Right: FCNN error with respect to GLENS simulation data.
  • Figure 5: Layer tests for the GCNN in terms of mean absolute error (MAE) and maximum absolute error (MaxAE). Charts are normalized by the minimum MAE and MaxAE for comparison.
  • ...and 3 more figures