Quantifying wildﬁre drivers and predictability in boreal peatlands using a two-step error-correcting machine learning framework in TeFire v1.0

Aug 5, 2026

Upload by Aurora

Wildfires increasingly threaten boreal peatland (BP) ecosystem sustainability and can destabilize boreal carbon storage. Forecasting rare and extreme BP fire occurrence remains difficult, and the quantitative contribution of both natural and human drivers is still unclear. This study quantifies BP fire predictability and key controlling factors from 1997 to 2015 using a two-step error-correcting machine learning framework combining multiple classifiers, regression, and error-correction. Results show oversampling improves recall, nonparametric models—especially random forest—perform best, and simulations indicate temperature, air dryness, and climate extremes dominate over precipitation, wind, and human activities.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Abstract

Introduction

Boreal peatlands and fire relevance

Smoldering combustion and knowledge gaps

Drivers and modeling challenges

Geosci. Model Dev., 17, 1525–1542, 2024 [https://doi.org/10.5194/gmd-17-1525-2024](https://doi.org/10.5194/gmd-17-1525-2024)[ ](https://doi.org/10.5194/gmd-17-1525-2024)© Author(s) 2024 . This work is distributed under the Creative Commons Attribution 4 .0 License. Quantifying wildﬁre drivers and predictability in boreal peatlands using a two-step error-correcting machine learning framework in TeFire v1.0 Rongyun Tang 1,2 , Mingzhou Jin 1 , Jiafu Mao3 , Daniel M. Ricciuto3 , Anping Chen4 , and Yulong Zhang 1 1Institute for a Secure and Sustainable Environment and Department of Industrial and Systems Engineering, University of Tennessee, Knoxville, TN 37996, USA 2Department of Biological & Agricultural Engineering, University of Arkansas, Fayetteville, AR 72701, USA 3Environmental Sciences Division and Climate Change Science Institute, Oak Ridge National Laboratory, Oak Ridge, TN 37830, USA 4Department of Biology and Graduate Degree Program in Ecology, Colorado State University, Fort Collins, CO 80523, USA Correspondence: Mingzhou Jin ([jin@utk.edu](jin@utk.edu)) and Jiafu Mao ([maoj@ornl.gov](maoj@ornl.gov)) Received: 29 January 2023 – Discussion started: 21 February 2023 Revised: 6 November 2023 – Accepted: 7 November 2023 – Published: 21 February 2024 Abstract. Wildﬁres are becoming an increasing challenge to the sustainability of boreal peatland (BP) ecosystems and can alter the stability of boreal carbon storage. However, predicting the occurrence of rare and extreme BP ﬁres proves to be challenging, and gaining a quantitative understanding of the factors, both natural and anthropogenic, inducing BP ﬁres remains elusive. Here, we quantiﬁed the predictability of BP ﬁres and their primary controlling factors from 1997 to 2015 using a two-step correcting machine learning (ML) framework that combines multiple ML classiﬁers, regression models, and an error-correcting technique. We found that (1) the adopted oversampling algorithm effectively addressed the unbalanced data and improved the recall rate by 26.88 %–48.62 % when using multiple datasets, and the error-correcting technique tackled the overestimation of ﬁre sizes during ﬁre seasons; (2) nonparametric models outperformed parametric models in predicting ﬁre occurrences, and the random forest machine learning model performed the best, with the area under the receiver operating characteristic curve ranging from 0.83 to 0.93 across multiple ﬁre datasets; and (3) four sets offactor-control simulations consistently indicated the dominant role of temperature, air dryness, and climate extreme (i.e., frost) for boreal peatland ﬁres, overriding the effects of precipitation, wind speed, and human activities. Our ﬁndings demonstrate the efﬁciency and accuracy of ML techniques in predicting rare and extreme ﬁre events and dis- entangle the primary factors determining BP ﬁres, which are critical for predicting future ﬁre risks under climate change. 1 Introduction Carbon-rich boreal peatlands (BPs) cover only 􀀘 2 % of the Earth's surface (Gorham, 1991) but have accumulated 􀀘 20 %–40 %(450 􀀆 150 PgC) of the global soil carbon, historically having a net cooling effect on the global radiation balance (Hugelius et al., 2020; Page and Hooijer, 2016; Scharlemann et al., 2014) . This major land carbon pool, however, is highly vulnerable to current global warming, which tends to introduce carbon emissions into the atmosphere through increasing decomposition of peat soil organic matter and ﬁre combustion (Turetsky et al., 2014) . In particular, BP ﬁre regimes have been undergoing pronounced changes over recent decades in terms of ﬁre extent, frequency, and duration (Field and Raupach, 2004; Kelly et al., 2013) . In BPs, there are two types of wildﬁres – surface ﬂaming and underground smoldering – that can transition from one to the other at different phases. It is noteworthy that compared to ﬂaming combustion, smoldering combustion is easier to ignite, harder to suppress

FAQs

Why are boreal peatland fires difficult to predict quantitatively?

Rare and extreme BP fire occurrence is challenging to forecast, and the quantitative roles of both natural and anthropogenic drivers have remained elusive.

What does the two-step error-correcting machine learning framework do in this study?

It combines multiple machine learning classifiers and regression models with an error-correcting technique to address unbalanced data and to reduce overestimation of fire sizes during fire seasons.

Which factors are identified as dominant drivers of boreal peatland fires?

Factor-control simulations consistently highlight temperature, air dryness, and climate extremes (e.g., frost) as dominant drivers, overriding precipitation, wind speed, and human activities.

Aug 5, 2026

Upload by Aurora