Causes of uncertainty in simulated burnt area by fire-enabled DGVMs
- 1Senckenberg Gesellshaft für Naturforschung, Biodiversität und Klima Forschungszentrum (BiK-F), Frankfurt am Main, Germany (matthew.forrest@senckenberg.de)
- 2Met Office Hadley Centre, Exeter, UK
- 3Potsdam Institute for Climate Impact Research, Member of the Leibniz Association, Potsdam, Germany
- 4Earth System Science Program, Faculty of Natural Sciences, Universidad del Rosario, Bogota, Colombia
- 5International Center for Climate and Environment Sciences, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China
- 6Canadian Centre for Climate Modelling and Analysis, Climate Research Division, Environment and Climate Change Canada, Victoria, BC, Canada
- 7Department of Physical Geography and Ecosystem Sciences, Lund University, Lund, Sweden
- 8Center for Environmental Prediction, School of Environmental & Biological Sciences, Rutgers University, New Brunswick, NJ, USA
- 9Department of Geography, College of Life and Environmental Sciences, University of Exeter, Exeter, UK
- 10College of Natural Resources and Environment, Northwest A & F University, Yangling, China
- 11State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau, Northwest A&F University, Yangling, China
- 12Department of Physical Geography, Goethe University, Frankfurt am Main, Germany
Fire-enabled dynamic global vegetation models (DGVMs) can be used to study how fire activity responds to its main drivers, including climate/weather, vegetation and human activities, at coarse spatial scales. Such models can also be used to examine the effects of fire on vegetation, and, when embedded in Earth system models, investigate the feedback of fire on the climate system. Thus they are valuable tools for studying wildfires. Accordingly, the Fire Model Intercomparison Project (FireMIP) was established to evaluate and utilise these models using consistent protocols.
Here we present the second round of FireMIP simulations to focus historic wildfire drivers (1901 to present). A six-member ensemble of simulations from fire-enabled DGVMs was compared to remotely-sensed burnt area observations and to the previous round of historical FireMIP simulations. We found that the model skill when simulating spatial patterns of burnt area shows modest improvements compared to the previous FireMIP round, and that the simulations mostly reproduce the decreasing trend in global burnt area found over the last two decades. However, whilst the broad global patterns are reasonable, there are considerable discrepancies with regards to regional agreement and timing of burnt area. Furthermore, the models show diverging trends in the pre-satellite era.
To investigate further and inform future model development, we explored the residuals between simulated burnt area from the FireMIP models and remotely-sensed burnt area as a function of climate, vegetation, anthropogenic and topographic variables using generalised additive models (GAMs). We found some common responses across the models, with many over-predicting fire activity in arid/low productivity areas and all models under-predicting at low road density. However, with respect to other variables, such as wind speed and cropland fraction, the models residuals showed divergent responses. It is anticipated that these results should aid further development of global fire models in terms of driving variables, process representations and model structure.
How to cite: Forrest, M., Burton, C., Drüke, M., Hantson, S., Li, F., Melton, J., Nieradzik, L., Rabin, S., Sitch, S., Yue, C., and Hickler, T.: Causes of uncertainty in simulated burnt area by fire-enabled DGVMs, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-12604, https://doi.org/10.5194/egusphere-egu23-12604, 2023.
