Primary Research ArticleAdapting a dynamic vegetation model for regional biomass, plant biogeography, and fire modeling in the Greater Yellowstone Ecosystem: Evaluating LPJ-GUESS-LMfireCF
Introduction
Changes in regional climate and disturbance characteristics threaten the resiliency and function of forested ecosystems (Turner et al., 2019; Walker et al., 2019), increasing the need for regional modeling of forest and disturbance dynamics (U.S. DOE, 2018). Forest resiliency is defined as the capacity of a forest to absorb change and disturbance and have structure and function persist (Holling, 1973). With increases in fire season length, area burned, and frequency of large fires in the boreal and temperate forests of North America (Dennison et al., 2014; Kasischke and Turetsky, 2006; Westerling, 2016, 2006), there is uncertainty about the resiliency of these forests and their carbon storage (Turner et al., 2019; Walker et al., 2019). In particular, the extent of high-severity fires that kill most or all of the trees across entire stands, referred to here as stand-replacing fires, is increasing (Flannigan et al., 2009; Stephens et al., 2014). The ability to predict future forest resiliency is dependent on the development of process-based simulation models that can represent complex interactions between forest demography, disturbance, climatic factors, and increasing atmospheric carbon dioxide (U.S. DOE, 2018). Therefore, appropriately modeling plant geography, disturbance-driven biomass turnover, and forest regrowth under a changing climate is critical to predicting future changes to forests’ persistence and function at a regional scale.
Current common approaches to regional forest modeling are empirical models, individual-based simulation models that are extended to landscape applications, or Dynamic Global Vegetation Models (DGVMs) applied regionally. Empirical modeling, including correlation, regression, and principal component analyses can reveal important climatic and disturbance relationships affecting forest productivity (Emmett et al., 2019; Notaro et al., 2019; Potter, 2019). Yet empirical models are often constrained by the limited number of variables included by the modeler that may fail to capture complicated feedbacks between ecological processes. Also, since empirical models are based on correlative relationships between variables from field or remotely sensed data, they rely on statistical extrapolation to make inferences about conditions or areas not explicitly measured. Individual-based forest landscape models offer high-resolution (e.g. 2 m to 30 m spatial resolution) simulations of forest dynamics (Mladenoff, 2004; Seidl et al., 2012). Individual-based simulation models trade their high-spatial resolution for limited spatial extent, making regional-scale simulations computationally impractical.
In contrast, DGVMs were developed for understanding feedbacks between vegetation dynamics, biogeography, and biogeochemistry (Bachelet et al., 2001; Moorcroft et al., 2001; Sitch et al., 2003) at the regional to global scale. DGVMs are process-based simulation models that represent vegetation dynamics including plant establishment, growth, competition, and mortality. The formulations within process-based models are often more closely based on principles of vegetation dynamics (e.g. canopy scaling based on optimum leaf nitrogen distribution) than on empirical relationships, and therefore are less dependent on statistical extrapolation for novel conditions. They also incorporate physical processes (e.g. soil hydrology) and physiological processes (e.g. photosynthesis, respiration, and carbon allocation) important for representing ecological function. The benefit of adapting a DGVM to regional applications is the inclusion of these vegetation dynamics and ecological processes that interact to determine forest resiliency.
Further development of fire dynamics and forest demography in DGVMs is needed to more accurately represent these interactions (Pugh et al., 2019a; U.S. DOE, 2018; Zhu et al., 2016). To capture the ecosystem responses and interactions between vegetation, climate, and disturbance, DGVMs must include comprehensive and realistic fire modules (Keane et al., 2015). The necessity of fire module development in DGVMs led to the formation of the Fire Modeling Intercomparison Project and remains an area of active research (Hantson et al., 2020, 2016; Li et al., 2019, Rabin et al., 2017). While there have been many advances, improvement is needed in the representation of forest fires that burn the crowns of trees or shrubs killing most or all of the overstory (Pugh et al., 2019a, 2019b), hereafter referred to as stand-replacing crown fires (Agee, 1996; Scott and Reinhardt, 2001). Currently, the prominent DGVMs rely upon surface fire models and empirical fire behavior models which do not include the transition of a surface fire into the canopy, limiting their ability to reproduce and predict stand-replacing crown fires (Chaste et al., 2018; Gavin et al., 2014; Hantson et al., 2016; Lehsten et al., 2016; Li et al., 2012; Rabin et al., 2017).
Advances are also needed in the modeling of forest demography within dynamic vegetation models. DGVMs typically simulate plant functional types (PFTs) representing multiple plant species grouped by their physical, phenological, and phylogenetic characteristics. Process-based simulation models are often initiated and constrained by field or remotely sensed data, prescribing the distribution of PFTs or forest productivity. For example, in the Carnegie Ames Stanford Approach ecosystem model simulations, land cover type was prescribed from satellite imagery, and Moderate Resolution Imaging Spectroradiometer (MODIS) Enhanced Vegetation Index and airborne remotely sensed coarse woody debris were used as inputs to estimate net primary productivity (Potter et al., 2011). While initializing and constraining process-based models with observational data can lead to more realistic simulation results, such approaches fail to demonstrate the emergent properties of the model and thereby reduce confidence in prognostic simulations. Calibrating model parameters for simulated PFTs and represented processes is a necessary step to improve model performance without relying on initialization of plant cover type. For example, improving the representation of high-latitude vegetation in another DGVM, Organizing Carbon and Hydrology in Dynamic Ecosystems, required modification of photosynthesis parameters for each simulated PFT, adjustment of temperature limits to tree distributions, and revised tree mortality calculations (Zhu et al., 2015). These DGVM developments have not yet included crown fire simulations, nor have they been applied to all DGVMs.
The dynamic vegetation model LPJ-GUESS is well suited for regional application because it incorporates the physiological and biophysical processes of a DGVM while simulating cohorts of PFTs to represent forest demography. LPJ-GUESS was developed from the DGVM LPJ (Lund-Potsdam-Jena) (Sitch et al., 2003) which represented PFTs as average populations. The General Ecosystem Simulator (GUESS) version represents vegetation as age-based cohorts, allowing the representation of stand structure and mixed plant composition. Simulating age-based cohorts is an intermediate approach between individual-based and population-based models, adding structural complexity while minimizing computational costs to enable regional to global scale applications (Fisher et al., 2018). In LPJ-GUESS vegetation is simulated in independent patches for each grid cell with plant establishment, growth, and competition represented as mechanistic processes based on first principles and empirical relationships (for detailed model description see [Smith et al., 2001]). Previous regional applications of LPJ-GUESS explored vegetation dynamics and plant biogeography in forests of the Northeastern U.S. (Hickler et al., 2004; Tang et al., 2012) and subregions across Europe (Hickler et al., 2012; Koca et al., 2006; Morales et al., 2007; Smith et al., 2008, 2001). However, to adapt LPJ-GUESS for regional application to western U.S. forest biomes several modifications were needed. Minor model modifications included adjustments to pedotransfer functions and the inclusion of soil parent material data (Section 2.2). Major developments included parameterization of regional PFTs to represent the composition of western forests (Section 2.3). In turn, the carbon allocation scheme was modified to produce more realistic tree heights (Section 2.4) and calculations of PFT-specific tree crown length were added (Section 2.5). Most importantly, the fire module LMfireCF was developed to simulate the stand-replacing crown fires characteristic of many western forests (Sections 2.6 and 2.7). The development history of the DGVMs and fire modules relevant to this study is shown in Fig. 1.
GlobFIRM, the original fire module commonly used within LPJ-GUESS, represents fire occurrence, size, and effects using semi-empirical approaches for non-ignition limited ecosystems (Thonicke et al., 2001). Fire occurrence is determined daily by whether air temperature, fuel moisture content, and abundance of aboveground litter are above minimum thresholds. Fire size is represented as the fraction of a grid cell burned as a function of fire season length, based on an empirical relationship between length of fire season and area burned. While this approach is intuitive, the simplifications overlook potential feedbacks that could be represented by a more mechanistic approach. For example, fire occurrence does not consider the availability of an ignition source or fuel conditions that may be limiting in some ecosystems. Fire size does not consider any mechanisms for fire spread, thereby all fires for a season will be of the same size regardless of fuel availability and fuel moisture. Fire effects are represented as the fraction of biomass burnt by fire is based on a PFT-specific fire resistance parameter. All burned live and dead biomass is considered entirely combusted and added to the annual carbon flux to the atmosphere. This representation of fire effects ignores any varying resistance to fire based on age class and fails to distinguish partially burned (e.g. standing dead, woody debris) and combusted fuels (emitted to the atmosphere).
To better represent fire occurrence, size, and effects in LPJ-GUESS simulations, we integrated LMfire, a mechanistic fire module originally developed for use in LPJ (Pfeiffer et al., 2013). LMfire simulates fire occurrence, behavior, and impact from a more mechanistic perspective by incorporating fire danger indices with fire spread modeling (Pfeiffer et al., 2013; Thonicke et al., 2010). Fire occurrence is based on calculated probabilities of natural ignitions and anthropogenic burning habits. Instead of having fire size based on an empirical relationship with fire season length as in GlobFIRM, LMfire represents fire behavior by calculating fire rate of spread based on weather and topography, following Rothermel's equations (Rothermel, 1972). It also allows multi-day burning and coalescence of fires within patches, more realistically representing fire behavior. However, it must be noted that patches are simulated independently, so fire does not spread between patches or grid cells. Tree mortality is a function of crown scorch and cambial damage based on the current tree height and bark thickness for each PFT.
Simulating stand-replacing crown fires necessitated new developments to the LMfire module, now termed LMfireCF, to represent crown fire (CF), resulting in the ecosystem model variant LPJ-GUESS-LMfireCF. LMfireCF assesses if critical conditions are met for crown fire initiation and spread. Crown fire initiation is dependent on surface fire intensity resulting in a scorch height that reaches the canopy base height. Crown fire spread depends on average canopy bulk density and canopy foliar moisture content meeting critical thresholds.
The purpose of this paper is to introduce LPJ-GUESS-LMfireCF designed for regional application for western forests of the U.S. and to evaluate the model's performance based on regional PFTs and the new fire module. Simulations were run to compare two different fire modules: GlobFIRM and the newly developed LMfireCF. To compare the different potential PFTs, simulations were run with the global PFTs and the newly parameterized regional PFTs. LPJ-GUESS-LMfireCF performance was evaluated for simulated 1) landscape biomass distribution, 2) dominant plant cover distributions, 3) fire activity, and 4) postfire forest regeneration by comparing simulated results to field and satellite-based metrics. Simulations of Yellowstone National Park (YNP) vegetation and wildfire dynamics are used here to demonstrate the utility of LPJ-GUESS-LMfireCF for regional applications in U.S. western forests. YNP serves as a model forested landscape to study the interactions of vegetation, climate, and disturbance dynamics because large stand-replacing crown fires have played an important role in dictating vegetation patterns since the Holocene (Whitlock et al., 2003). The 1988 Yellowstone fires burned about one third of the area of YNP, serving as a natural experiment and validation for fire models.
Section snippets
Study area
Yellowstone National Park is part of the U.S. Rocky Mountains (Fig. 2). YNP extends 8983 km2 with a mean elevation of ~2400 m and an elevation range from 1610 m to 3462 m. Long, cold winters and cool summers characterize the climate of YNP (boreal cool summer under the Koppen-Geiger climate classifications, (Kottek et al., 2006)). The northern half of the park experiences warmer mean annual temperatures and lower annual precipitation relative to the southern half (Fig. 3). Vegetation
Landscape biomass
Fire model development in combination with the newly parameterized regional PFTs greatly improved modeled live carbon in vegetation relative to LPJ-GUESS-GlobFIRM (Fig. 5). Field data from FIA plots estimated mean total live aboveground carbon to be 50.8 ± 2.04 Mg C ha−1 (range 0–201 Mg C ha−1), with measurements limited to forested areas. Estimates from field data were higher than the satellite-based GlobBiomass estimates and LPJ-GUESS-LMfireCF simulation results, which included non-forested
Landscape biomass
Developments included in LPJ-GUESS-LMfireCF improved model performance in simulating carbon in vegetation in YNP compared to LPJ-GUESS with the GlobFIRM fire module (Fig. 5). The parameterization of regional PFTs (blue dash-dot line) compared to global PFTs (purple dotted line) also improved model performance in simulating carbon. Total carbon in vegetation in YNP was overestimated by 12% by LPJ-GUESS-LMfireCF with regional PFTs (green dashed line) compared to GlobBiomass estimations (solid
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
Thank you to Robert Keane, Monica Turner, Dave Roberts, and Andrew Hansen for their helpful feedback. Special thanks to Frances Ambrose for field data collection. Computations were performed on the Hyalite High Performance Computing System, operated and supported by University Information Technology Research Cyberinfrastructure at Montana State University; thank you to the staff administrators Pol Llovet, Erik Bryer, and Jonathan Hilmer, among others. KDE acknowledges funding from an NSF
References (134)
- et al.
Which tree species and biome types are most vulnerable to climate change in the US Northern Rocky Mountains?
For. Ecol. Manage.
(2015) - et al.
Guess the impact of Ips typographus—An ecosystem modelling approach for simulating spruce bark beetle outbreaks
Agric For Meteorol
(2012) - et al.
Representing climate, disturbance, and vegetation interactions in landscape models
Ecological Modelling
(2015) LANDIS and forest landscape models
Ecol Modell
(2004)- et al.
Root biomass studies in forest ecosystems
Pedobiologia (Jena)
(1977) - et al.
An individual-based process model to simulate landscape-scale forest ecosystem dynamics
Ecol Modell
(2012) Fire Ecology of Pacific Northwest Forests
(1996)Calculating and interpreting forest fire intensities
Can. J. Bot.
(1982)- et al.
Historical carbon dioxide emissions caused by land-use changes are possibly larger than assumed
Nat. Geosci
(2017) - et al.
Silvics of Whitebark Pine (Pinus albicaulis) (No. INT-GTR-253)
(1989)
MC1: A Dynamic Vegetation Model For Estimating the Distribution of Vegetation and Associated Ecosystem Fluxes of Carbon, Nutrients, and Water (General Technical Report), PNW-GTR-508
Lightning, Fuels, Topography, Climate, and Fire Behavior
Fire ecology in Rocky Mountain landscapes
Fire regimes on andesitic mountain Terrain in Northeastern Yellowstone National Park, Wyoming
Int. J. Wildland Fire
The relative importance of fuels and weather on fire behavior in subalpine forests
Ecology
rgdal: Bindings for the “Geospatial”
Predicting Duff and Woody fuel consumption in northern idaho prescribed fires
For. Sci.
Root biomass allocation in the world’s upland forests
Oecologia
Historic & projected climate change in the greater yellowstone ecosystem
Yellowstone Sci.
The pyrogeography of eastern boreal Canada from 1901 to 2012 simulated with the LPJ-LMfire model
Biogeosciences
Climate changes and wildfire alter vegetation of Yellowstone National Park, but forest cover persists
Ecosphere
A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soils
Water Resour Res
Spatial autocorrelation and statistical tests in ecology
Écoscience
Large wildfire trends in the western United States, 1984-2011
Geophys. Res. Lett.
Nonpyrogenous climax lodgepole pine communities in Yellowstone National Park
Ecology
Disentangling climate and disturbance effects on regional vegetation greening trends
Ecosystems
Global Agro-Ecological Zones Assessment for Agriculture (GAEZ 2008)
Vegetation demographics in Earth System Models: a review of progress and priorities
Glob Change Biol
Implications of changing climate for global wildland fire
Int. J. Wildland Fire
Global Carbon Budget 2019
Earth Syst. Sci. Data
Climate refugia: joint inference from fossil records, species distribution models and phylogeography
New Phytol
The Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2)
J. Climate
Rationale for a national annual forest inventory program
J. For.
Complex challenges of maintaining whitebark pine in greater Yellowstone under climate change: a call for innovative research, management, and policy approaches
Forests
It takes a few to tango: changing climate and fire regimes can cause regeneration failure of two subalpine conifers
Ecology
Origins of abrupt change? Postfire subalpine conifer regeneration declines nonlinearly with warming and drying
Ecol Monogr
The status and challenge of global fire modelling
Biogeosciences
Quantitative assessment of fire and vegetation properties in simulations with fire-enabled vegetation models from the Fire Model Intercomparison Project
Geosci. Model Dev.
High and dry: post-fire tree seedling establishment in subalpine forests decreases with post-fire drought and large stand-replacing burn patches: drought and post-fire tree seedlings
Global Ecol. Biogeogr.
BIOME3: an equilibrium terrestrial biosphere model based on ecophysiological constraints, resource availability, and competition among plant functional types
Global Biogeochem Cycles
Using a generalized vegetation model to simulate vegetation dynamics in Northeastern USA
Ecology
Projecting the future distribution of European potential natural vegetation zones with a generalized, tree species-based dynamic vegetation model: future changes in European vegetation zones
Glob. Ecol. Biogeography
Linking tree-ring and sediment-charcoal records to reconstruct fire occurrence and area burned in subalpine forests of Yellowstone National Park, USA
Holocene
Resilience and Stability of Ecological Systems
Annu Rev Ecol Syst
Complex response of white pines to past environmental variability increases understanding of future vulnerability
PLoS ONE
Postfire changes in forest carbon storage over a 300-year chronosequence of Pinus contorta -dominated forests
Ecol Monogr
Carbon storage on landscapes with stand-replacing Fires
Bioscience
Recent changes in the fire regime across the North American boreal region—Spatial and temporal patterns of burning across Canada and Alaska
Geophys. Res. Lett.
Cited by (7)
Modelling changes in vegetation productivity and carbon balance under future climate scenarios in southeastern Australia
2024, Science of the Total EnvironmentClimate Change: Anticipating and Adapting to the Impacts on Terrestrial Species
2024, Encyclopedia of Biodiversity, Third Edition: Volume 1-7Terrestrial carbon dynamics in an era of increasing wildfire
2023, Nature Climate ChangeModeling wildfire risk in western Iran based on the integration of AHP and GIS
2022, Environmental Monitoring and Assessment