1 of 1

What controls global fire? Evaluating emergent relationships in satellite observations and global vegetation models using machine learning

Fire is a major disturbance agent in terrestrial ecosystems. The occurrence and spread of wildfires is controlled by the interplay of human activities, weather conditions, and the conditions of vegetation and litter fuels. Most state-of-the-art global ecosystem models represent such controls to simulate fire effects on vegetation dynamics and global carbon cycling. However, global fire models poorly reproduce the observed dynamics and variability of fire burned area. Here we aim to identify and evaluate functional responses of global burned area to environmental and human controls. We use several global satellite, climate, and socioeconomic datasets, and simulations from the Fire Model Inter-comparison Project (FireMIP) [1] to predict the observed or modelled burned area with the random forest machine-learning algorithm. We then derive from the trained random forests individual conditional expectation curves [2], which represent emergent functional responses of burned area to controlling factors. These functional responses allow us to compare data- and model-derived sensitivities. FireMIP models mostly represent the emergent responses to climate variables but show diverse responses to human population, land cover, and vegetation. The models especially underestimate the emergent strong increase of burned area with increasing precedent plant productivity in many semi-arid ecosystems. The results suggest that FireMIP models misrepresent the links between plant productivity, biomass allocation, litter turnover, and fuel production. Additionally, the good performance of data-driven modelling approaches [3] suggests to develop hybrid global fire models to better represent and predict the role of fire dynamics for ecosystem functioning and vegetation-climate interactions. REFERENCES: 1. Rabin, S.S., Melton, J.R., Lasslop, G., Bachelet, D., Forrest, M., Hantson, S., Kaplan, J.O., Li, F., Mangeon, S., Ward, D.S., Yue, C., Arora, V.K., Hickler, T., Kloster, S., Knorr, W., Nieradzik, L., Spessa, A., Folberth, G.A., Sheehan, T., Voulgarakis, A., Kelley, D.I., Prentice, I.C., Sitch, S., Harrison, S., Arneth, A., 2017. The Fire Modeling Intercomparison Project (FireMIP), phase 1: experimental and analytical protocols with detailed model descriptions. Geosci Model Dev 10, 1175–1197. https://doi.org/10.5194/gmd-10-1175-2017 2. Goldstein, A., Kapelner, A., Bleich, J., Pitkin, E., 2013. Peeking Inside the Black Box: Visualizing Statistical Learning with Plots of Individual Conditional Expectation. ArXiv13096392 Stat. 3. Forkel, M., Dorigo, W., Lasslop, G., Teubner, I., Chuvieco, E., Thonicke, K., 2017. A data-driven approach to identify controls on global fire activity from satellite and climate observations (SOFIA V1). Geosci Model Dev 10, 4443–4476. https://doi.org/10.5194/gmd-10-4443-2017


Citation style:
Could not load citation form.


Use and reproduction: