The important role of fire in regulating vegetation community composition and contributions to emissions of greenhouse gases and aerosols make it a critical component of dynamic global vegetation models and Earth system models. Over 2 decades of development, a wide variety of model structures and mechanisms have been designed and incorporated into global fire models, which have been linked to different vegetation models. However, there has not yet been a systematic examination of how these different strategies contribute to model performance. Here we describe the structure of the first phase of the Fire Model Intercomparison Project (FireMIP), which for the first time seeks to systematically compare a number of models. By combining a standardized set of input data and model experiments with a rigorous comparison of model outputs to each other and to observations, we will improve the understanding of what drives vegetation fire, how it can best be simulated, and what new or improved observational data could allow better constraints on model behavior. In this paper, we introduce the fire models used in the first phase of FireMIP, the simulation protocols applied, and the benchmarking system used to evaluate the models. We have also created supplementary tables that describe, in thorough mathematical detail, the structure of each model

Abstract. Global fire-vegetation models are widely used to assess impacts of environmental change on fire regimes and the carbon cycle and to infer relationships between climate, land use and fire. However, differences in model structure and parameterizations, in both the vegetation and fire components of these models, could influence overall model performance, and to date there has been limited evaluation of how well different models represent various aspects of fire regimes. The Fire Model Intercomparison Project (FireMIP) is coordinating the evaluation of state-of-the-art global fire models, in order to improve projections of fire characteristics and fire impacts on ecosystems and human societies in the context of global environmental change. Here we perform a systematic evaluation of historical simulations made by nine FireMIP models to quantify their ability to reproduce a range of fire and vegetation benchmarks. The FireMIP models simulate a wide range in global annual total burnt area (39–536 Mha) and global annual fire carbon emission (0.91–4.75 Pg C yr−1) for modern conditions (2002–2012), but most of the range in burnt area is within observational uncertainty (345–468 Mha). Benchmarking scores indicate that seven out of nine FireMIP models are able to represent the spatial pattern in burnt area. The models also reproduce the seasonality in burnt area reasonably well but struggle to simulate fire season length and are largely unable to represent interannual variations in burnt area. However, models that represent cropland fires see improved simulation of fire seasonality in the Northern Hemisphere. The three FireMIP models which explicitly simulate individual fires are able to reproduce the spatial pattern in number of fires, but fire sizes are too small in key regions, and this results in an underestimation of burnt area. The correct representation of spatial and seasonal patterns in vegetation appears to correlate with a better representation of burnt area. The two older fire models included in the FireMIP ensemble (LPJ–GUESS–GlobFIRM, MC2) clearly perform less well globally than other models, but it is difficult to distinguish between the remaining ensemble members; some of these models are better at representing certain aspects of the fire regime; none clearly outperforms all other models across the full range of variables assessed.

Climate change adaptation and mitigation require understanding of vegetation response to climate change. Using the MC2 dynamic global vegetation model (DGVM) we simulate vegetation for the Northwest United States using results from 20 different Climate Model Intercomparison Project Phase 5 (CMIP5) models downscaled using the MACA algorithm. Results were generated for representative concentration pathways (RCP) 4.5 and 8.5 under vegetation modeling scenarios with and without fire suppression for a total of 80 model runs for future projections. For analysis, results were aggregated by three subregions: the western Northwest (WNW), from the crest of the Cascade Mountains west; Northwest plains and plateau (NWPP), the non-mountainous areas east of the Cascade Mountains; and eastern Northwest mountains (ENWM), the mountainous areas east of the Cascade Mountains. In the WNW, mean fire interval (MFI) averaged over all climate projections decreases by up to 48%, and potential vegetation shifts from conifer to mixed forest under RCP 4.5 and 8.5 with and without fire suppression. In the NWPP MFI averaged over all climate projections decreases by up to 82% without fire suppression and increases by up to 14% with fire suppression resulting in woodier vegetation cover. In the ENWM, MFI averaged across all climate projections decreases by up to81%, subalpine communities are lost, but conifer forests continue to dominate the subregionin the future.

The dynamic global vegetation model (DGVM) MC2 was run over the conterminous USA at 30 arc sec (~800 m) to simulate the impacts of nine climate futures generated by 3GCMs (CSIRO, MIROC and CGCM3) using 3 emission scenarios (A2, A1B and B1) in the context of the LandCarbon national carbon sequestration assessment. It first simulated potential vegetation dynamics from coast to coast assuming no human impacts and naturally occurring wildfires. A moderate effect of increased atmospheric CO2 on water use efficiency and growth enhanced carbon sequestration but did not greatly influence woody encroachment. The wildfires maintained prairie-forest ecotones in the Great Plains. With simulated fire suppression, the number and impacts of wildfires was reduced as only catastrophic fires were allowed to escape. This greatly increased the expansion of forests and woodlands across the western USA and some of the ecotones disappeared. However, when fires did occur, their impacts (both extent and biomass consumed) were very large. We also evaluated the relative influence of human land use including forest and crop harvest by running the DGVM with land use (and fire suppression) and simple land management rules. From 2041 through 2060, carbon stocks (live biomass, soil and dead biomass) of US terrestrial ecosystems varied between 155 and 162 Pg C across the three emission scenarios when potential natural vegetation was simulated. With land use, periodic harvest of croplands and timberlands as well as the prevention of woody expansion across the West reduced carbon stocks to a range of 122-126 Pg C, while effective fire suppression reduced fire emissions by about 50%. Despite the simplicity of our approach, the differences between the size of the carbon stocks confirm other reports of the importance of land use on the carbon cycle over climate change.

Climate change has already affected southern California where regional increases in temperature and vegetation shifts have been observed. While all the CMIP5 temperature projections agree on a substantial level of warming throughout the year, there is fair bit of divergence in the magnitude and seasonality of projected changes in rainfall. While desert plants and animals are generally adapted to extreme conditions, some species may be approaching their physiological threshold. We calculated the climate velocity of both temperature and aridity (PPT/PET) in SE California to illustrate the spatial variability of climate projections and reported on the probable expansion of barren lands reducing current species survivorship. We used a vegetation model to illustrate both temporal and spatial shifts in land cover in response to changes in environmental conditions. Such information is useful to plan land use for renewable energy siting in the region.

No abstract

Understanding where and how fire patterns may change is critical for management and policy decision-making. To map future fire patterns, statistical correlative models are typically developed, which associate observed fire locations with recent climate maps, and are then applied to maps of future climate projections. A potential source of uncertainty is the common omission of static or dynamic vegetation as predictor variables. We therefore assessed the sensitivity of future fire projections to different combinations of vegetation maps used as explanatory variables in a statistically based fire modeling framework. We compared models without vegetation to models that incorporated static vegetation maps and that included output from a dynamic vegetation model that imposed three scenarios of fire and one scenario of land use change. We mapped projected future probability of all and large fires (> = 40 ha) under two climate scenarios in a heterogeneous study area spanning a large elevational gradient in the Sierra Nevada, California, USA. Results showed high model sensitivity to the treatment of vegetation as a predictor variable, particularly for models of large fire probability and for models accounting for wildfire effects on vegetation, which lowered future fire probability. Some scenarios resulted in opposite directional trends in the extent and probability of future fire, which could have serious implications for policy and management resource allocation. Model sensitivity resulted from high relative importance of vegetation variables in the baseline models and from large predicted changes in vegetation, particularly when simulating wildfire. Although statistical fire models often omit vegetation due to uncertainty, model sensitivity demonstrated here suggests a need to account for that uncertainty. Coupling statistical and processed based models may be a promising approach to reflect a more plausible range of scenarios.

Abstract. Global fire-vegetation models are widely used to assess impacts of environmental change on fire regimes and the carbon cycle, and to infer relationships between climate, land use, and fire. However, differences in model structure and parameterizations, in both the vegetation and fire components of these models, could influence overall model performance, and to date there has been limited evaluation of how well different models represent various aspects of fire regimes. The Fire Model Intercomparison Project (FireMIP) is coordinating the evaluation of state-of-the-art global fire models, with the aim of improving projections of fire regime characteristic and fire impacts on ecosystems and human societies under the context of global environmental change. Here we perform a systematic evaluation of historical simulations made by nine FireMIP models in order to quantify their ability to reproduce a range of fire and vegetation benchmarks. The FireMIP models simulate a wide range in global annual total burnt area (39–536 Mha), and global annual fire carbon emission (0.91–4.75 Pg C a−1) for modern conditions (2002–2012), but most of the range in burnt area is within observational uncertainty (345–468 Mha). Benchmarking scores indicate that seven out of nine FireMIP models are able to represent the spatial pattern in burnt area. The models also reproduce the seasonality in burnt area reasonably well but struggle to simulate fire season length and are largely unable to represent inter-annual variations in burnt area. However, models that represent cropland fires see improved simulation of fire seasonality in the northern hemisphere. The three FireMIP models which explicitly simulate individual fires are able to reproduce the spatial pattern in number of fires, but fire sizes are too small in key regions and this results in an underestimation of burnt area. The correct representation of spatial and seasonal patterns in vegetation appears to correlate with a better representation of burnt area. While some FireMIP models are better at representing certain aspects of the fire regime, no model clearly outperforms all other models across the full range of variables assessed.