Articles | Volume 16, issue 3
Model description paper
03 Feb 2023
Model description paper |  | 03 Feb 2023

AttentionFire_v1.0: interpretable machine learning fire model for burned-area predictions over tropics

Fa Li, Qing Zhu, William J. Riley, Lei Zhao, Li Xu, Kunxiaojia Yuan, Min Chen, Huayi Wu, Zhipeng Gui, Jianya Gong, and James T. Randerson

Data sets

Global Fire Emissions Database J. T. Randerson, G. R. van der Werf, L. Giglio, G. J. Collatz, and P. S. Kasibhatla

NCEP--DOE AMIP-II Reanalysis (R-2) ( Masao Kanamitsu, Wesley Ebisuzaki, Jack Woollen, Shi-Keng Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter

Harmonization of global land use change and management for the period 850–2100 (LUH2) for CMIP6 ( George C. Hurtt, Louise Chini, Ritvik Sahajpal, et al.

Global patterns of current and future road infrastructure ( Johan R Meijer, Mark A. J. Huijbregts, Kees C. G. J. Schotten, and Aafke M. Schipper

Mapping the global distribution of livestock ( Timothy P. Robinson, G. R. William Wint, Giulia Conchedda, Thomas P. Van Boeckel, Valentina Ercoli, Elisa Palamara, Giuseppina Cinardi, Laura D'Aietti, Simon I. Hay, and Marius Gilbert

Climate Indices: Monthly Atmospheric and Ocean Time Series NOAA

Model code and software

AttentionFire (1.0) F. Li, Q. Zhu, W. J. Riley, L. Zhao, L. Xu, K. Yuan, M. Chen, H. Wu, Z. Gui, J. Gong, and J. T. Randerson

Short summary
We developed an interpretable machine learning model to predict sub-seasonal and near-future wildfire-burned area over African and South American regions. We found strong time-lagged controls (up to 6–8 months) of local climate wetness on burned areas. A skillful use of such time-lagged controls in machine learning models results in highly accurate predictions of wildfire-burned areas; this will also help develop relevant early-warning and management systems for tropical wildfires.