Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep learning surrogate models of JULES-INFERNO for wildfire prediction on a global scale

Sibo Cheng, Hector Chassagnon, Matthew Kasoar, Yike Guo, Rossella Arcucci

TL;DR

This work tackles the computational bottleneck of global wildfire forecasting by proposing two deep learning surrogates for JULES-INFERNO: CAE-LSTM and ConvLSTM. Trained on five $30$-year JULES-INFERNO runs using fields X, V, M, and T, the surrogates deliver near real-time predictions with high fidelity, achieving $<$ $0.3\%$ AEP and $>$ $98\%$ SSIM on unseen data after fine-tuning. The ConvLSTM-based joint model generally offers superior generalization, while the CAE-LSTM provides strong performance with efficient latent-space forecasting; both approaches achieve substantial speedups over full physics-based runs. This enables rapid, global-scale wildfire forecasting and can be extended to additional inputs and data-assimilation strategies, potentially informing decision-making in wildfire risk management.

Abstract

Global wildfire models play a crucial role in anticipating and responding to changing wildfire regimes. JULES-INFERNO is a global vegetation and fire model simulating wildfire emissions and area burnt on a global scale. However, because of the high data dimensionality and system complexity, JULES-INFERNO's computational costs make it challenging to apply to fire risk forecasting with unseen initial conditions. Typically, running JULES-INFERNO for 30 years of prediction will take several hours on High Performance Computing (HPC) clusters. To tackle this bottleneck, two data-driven models are built in this work based on Deep Learning techniques to surrogate the JULES-INFERNO model and speed up global wildfire forecasting. More precisely, these machine learning models take global temperature, vegetation density, soil moisture and previous forecasts as inputs to predict the subsequent global area burnt on an iterative basis. Average Error per Pixel (AEP) and Structural Similarity Index Measure (SSIM) are used as metrics to evaluate the performance of the proposed surrogate models. A fine tuning strategy is also proposed in this work to improve the algorithm performance for unseen scenarios. Numerical results show a strong performance of the proposed models, in terms of both computational efficiency (less than 20 seconds for 30 years of prediction on a laptop CPU) and prediction accuracy (with AEP under 0.3\% and SSIM over 98\% compared to the outputs of JULES-INFERNO).

Deep learning surrogate models of JULES-INFERNO for wildfire prediction on a global scale

TL;DR

This work tackles the computational bottleneck of global wildfire forecasting by proposing two deep learning surrogates for JULES-INFERNO: CAE-LSTM and ConvLSTM. Trained on five -year JULES-INFERNO runs using fields X, V, M, and T, the surrogates deliver near real-time predictions with high fidelity, achieving AEP and SSIM on unseen data after fine-tuning. The ConvLSTM-based joint model generally offers superior generalization, while the CAE-LSTM provides strong performance with efficient latent-space forecasting; both approaches achieve substantial speedups over full physics-based runs. This enables rapid, global-scale wildfire forecasting and can be extended to additional inputs and data-assimilation strategies, potentially informing decision-making in wildfire risk management.

Abstract

Global wildfire models play a crucial role in anticipating and responding to changing wildfire regimes. JULES-INFERNO is a global vegetation and fire model simulating wildfire emissions and area burnt on a global scale. However, because of the high data dimensionality and system complexity, JULES-INFERNO's computational costs make it challenging to apply to fire risk forecasting with unseen initial conditions. Typically, running JULES-INFERNO for 30 years of prediction will take several hours on High Performance Computing (HPC) clusters. To tackle this bottleneck, two data-driven models are built in this work based on Deep Learning techniques to surrogate the JULES-INFERNO model and speed up global wildfire forecasting. More precisely, these machine learning models take global temperature, vegetation density, soil moisture and previous forecasts as inputs to predict the subsequent global area burnt on an iterative basis. Average Error per Pixel (AEP) and Structural Similarity Index Measure (SSIM) are used as metrics to evaluate the performance of the proposed surrogate models. A fine tuning strategy is also proposed in this work to improve the algorithm performance for unseen scenarios. Numerical results show a strong performance of the proposed models, in terms of both computational efficiency (less than 20 seconds for 30 years of prediction on a laptop CPU) and prediction accuracy (with AEP under 0.3\% and SSIM over 98\% compared to the outputs of JULES-INFERNO).
Paper Structure (17 sections, 10 equations, 8 figures, 6 tables)

This paper contains 17 sections, 10 equations, 8 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Model and dataset
  3. JULES - INFERNO model
  4. Data generation
  5. Data preprocessing
  6. Methodology
  7. CAE - LSTM
  8. CAE
  9. LSTM
  10. ConvLSTM
  11. Model Fine tuning
  12. Numerical results and discussions
  13. Metrics
  14. Evaluation of data compression
  15. Evaluation of predictive models
  16. ...and 2 more sections

Figures (8)

  • Figure 1: Four spatially distributed variables in Logarithmic scale at $t = 300$
  • Figure 2: Workflow of Joint CAE-LSTM
  • Figure 3: Workflow of Joint ConvLSTM
  • Figure 4: Cumulative sum of AEP on $P_{4}$ for CAE-LSTM and ConvLSTM as Joint and Single models
  • Figure 5: Cumulative sum of AEP on $P_{4}$ for M2MJoint CAE - LSTM and Joint ConvLSTM with different values of $p$ and $n$
  • ...and 3 more figures