Global Lightning-Ignited Wildfires Prediction and Climate Change Projections based on Explainable Machine Learning Models

TL;DR

This study presents machine learning models designed to characterize and predict lightning-ignited wildfires on a global scale and demonstrates that climate changes have steadily increased the global risk of lightning-ignited wildfires over the past decade.

Abstract

Wildfires pose a significant natural disaster risk to populations and contribute to accelerated climate change. As wildfires are also affected by climate change, extreme wildfires are becoming increasingly frequent. Although they occur less frequently globally than those sparked by human activities, lightning-ignited wildfires play a substantial role in carbon emissions and account for the majority of burned areas in certain regions. While existing computational models, especially those based on machine learning, aim to predict lightning-ignited wildfires, they are typically tailored to specific regions with unique characteristics, limiting their global applicability. In this study, we present machine learning models designed to characterize and predict lightning-ignited wildfires on a global scale. Our approach involves classifying lightning-ignited versus anthropogenic wildfires, and estimating with high accuracy the probability of lightning to ignite a fire based on a wide spectrum of factors such as meteorological conditions and vegetation. Utilizing these models, we analyze seasonal and spatial trends in lightning-ignited wildfires shedding light on the impact of climate change on this phenomenon. We analyze the influence of various features on the models using eXplainable Artificial Intelligence (XAI) frameworks. Our findings highlight significant global differences between anthropogenic and lightning-ignited wildfires. Moreover, we demonstrate that, even over a short time span of less than a decade, climate changes have steadily increased the global risk of lightning-ignited wildfires. This distinction underscores the imperative need for dedicated predictive models and fire weather indices tailored specifically to each type of wildfire.

Figures (5)

• Figure 1: A Pearson correlation matrix between the source features and themselves and between the source features and the target feature.
• Figure 2: (a) Descriptive statistics of lightning with or without ignition (b) Descriptive statistics of anthropogenic versus lightning-ignited wildfires
• Figure 3: Model accuracy and predicted seasonal ignition risk based on the full XGBoost model. (a) Average accuracy of the XGBoost model (b) Average accuracy of the logistic regression (c) Predicted annual ignition risk based on the full XGBoost model (d-g) Predicted seasonal ignition risk based on the full XGBoost model: (d) December-February (e) March-May (f) June-August (g) September-November. Similar figures for additional models are presented in the Supplementary Information file (Figures S5-S7). The figure was generated using the Cartopy library cartopy.
• Figure 4: (a) Mean annual difference in the model's prediction of ignition risk between 2014-2021. (b) Increase in wildfire risk between present day and 2100. The figure was generated using the Cartopy library cartopy.
• Figure 5: A schematic view of the methodology used in this study.