Improving prediction and assessment of global fires using multilayer neural networks

Abstract

Fires determine vegetation patterns, impact human societies, and are a part of complex feedbacks into the global climate system. Empirical and process-based models differ in their scale and mechanistic assumptions, giving divergent predictions of fire drivers and extent. Although humans have historically used and managed fires, the current role of anthropogenic drivers of fires remains less quantified. Whereas patterns in fire–climate interactions are consistent across the globe, fire–human–vegetation relationships vary strongly by region. Taking a data-driven approach, we use an artificial neural network to learn region-specific relationships between fire and its socio-environmental drivers across the globe. As a result, our models achieve higher predictability as compared to many state-of-the-art fire models, with global spatial correlation of 0.92, monthly temporal correlation of 0.76, interannual correlation of 0.69, and grid-cell level correlation of 0.60, between predicted and observed burned area. Given the current socio-anthropogenic conditions, Equatorial Asia, southern Africa, and Australia show a strong sensitivity of burned area to temperature whereas northern Africa shows a strong negative sensitivity. Overall, forests and shrublands show a stronger sensitivity of burned area to temperature compared to savannas, potentially weakening their status as carbon sinks under future climate-change scenarios.

Machine Learning Methods and Synthetic Data Generation to Predict Large Wildfires

“…Earlier research, as [ 111 ], propose the need to establish a threshold to delimit a large fire event [ 112 ]; however, this threshold is debatable [ 113 ]. On the other hand, recent ML-based models at continental or global scales for predicting burn areas offer good results in general term but fail to distinguish large wildfires [ 114 ]. This imbalance of data justifies the use of synthetic data as proposed in this project.…”

Winter and spring climate explains a large portion of interannual variability and trend in western U.S. summer fire burned area

“…Inclusion of summer conditions results in only a slight bias reduction for the extreme 2020 fire season which is underestimated by GAMs (figure 2). Failure to capture the extreme 2020 burned area extent is likely due to the rarity of the event which is not reflected in the training data [20,56]. Recent research suggests the extreme 2020 fire season was enabled by anomalously dry conditions [5]; however, burned area underestimation from our climate-based statistical models motivate future research to further understand the respective contributions of large-scale climate and factors such as local fire weather, fuel availability and ignitions to this extreme fire year.…”

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

“…Five other widely used Machine learning (ML) models are used as baseline models to compare with AttentionFire model: random forest (RF) [Coffield et al, 2019;Gray et al, 2018b], decision tree (DT) [Amatulli et al, 2006], gradient boosting decision tree (GBDT) [Coffield et al, 2019], artificial neuro network (ANN) [Joshi and Sukumar, 2021;Zhu et al, 2021], and naive LSTM. The inputs of climate and fuel-related variables for the first four models (non-sequence models) are variables of the latest three month available for prediction [Yu et al, 2020] while the corresponding inputs of naive LSTM and AttentionFire models are whole-year historical time sequences which cover dynamics from wet to dry seasons to capture short-and long-term dependencies underlying the input sequence [Guo et al, 2019;Li et al, 2020;Qin et al, 2017;Vaswani et al, 2017].…”