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

Abstract. African and South American (ASA) wildfires account for more than 70 % of global burned areas have strong connection to local climate sub-seasonal seasonal wildfire dynamics. However, representation the wildfire–climate relationship remains challenging due spatiotemporally heterogenous responses variability human influences. Here, we developed an interpretable machine learning (ML) fire model (AttentionFire_v1.0) resolve complex controls activities on better predict over ASA regions. Our ML substantially improved predictability both spatial temporal dynamics compared with five commonly used models. More importantly, revealed time-lagged control from wetness areas. The also predicted that, under a high-emission future scenario, recently observed declines in area will reverse America near changes. study provides reliable highlights importance lagged relationships historical predictions.