Widespread Dieback of Forests in North America under Rapid Global Warming: Response to the VINCERA Future Climate Scenarios Simulated by the MC1 DGVM
The VINCERA project is an intercomparison among three dynamic general vegetation models (DGVMs) simulating the response of North American ecosystems to six new future climate scenarios. The scenarios were produced by three general circulation models, each using two different future trace gas emissions scenarios. All of the scenarios are near the warmer end of the Intergovernmental Panel on Climate Change's projected future temperature range. Here we present results from the MC1 DGVM. All major forested ecosystems in North America exhibit carbon sequestration until the late 20th or early 21st century, followed by a drought induced decline and loss of carbon to levels below those at 1900 in the absence of fire suppression. By the end of the 21st century, the entire continent will have lost from 10 to 30 Pg of carbon, depending on the scenario. However, fire suppression can significantly mitigate carbon losses and ecosystem declines, producing a net change in carbon from a loss of about 5 Pg to a gain of about 8 Pg under the different scenarios. Most of the suppression benefits are obtained by forests in the western U.S. Suppression also mitigates carbon losses and conversions to savanna or grassland in the eastern U.S., but forest decline still occurs in the east under all scenarios. Dieback is triggered by two mechanisms. Reduced regional precipitation, variable among the scenarios, is one. The second more pervasive mechanism is the influence of rising temperatures on evapotranspiration. Even with the benefits of enhanced water use efficiency from elevated CO2 and slight increases in precipitation, dramatic increases in temperature can produce widespread forest dieback, and increases in fire severity. The eastern United States appear to be particularly vulnerable, as does the central Canadian boreal forest because of the relative flatness of climate gradients near ecotones. Under some scenarios, dieback is also driven by both increasing temperatures and decreasing precipitation, most notably the southeastern and northwestern United States. Following a period of gradual carbon sequestration, the enhanced evapotranspiration appears to overtake the 'greening' processes producing a rapid dieback. The point of turnaround from greenup to dieback occurs about now for the temperate forests and about a decade from now in the boreal forests, initiating an extended period of rapid losses of ecosystem carbon. These results underscore the critical importance of addressing uncertainties with respect to ecosystem water balance and the direct effects of elevated CO2 concentrations.
Possible impacts of 21s century climate on vegetation in West Africa.
We use the simulated climate from two IPCC SRES scenarios and two climate models to force the dynamic vegetation model IBIS over West Africa. We use the simulations for 2070-2099 together with the present day simulation and the CRUO5 20th century observed data to force IBIS and evaluate the equilibrium changes in vegetation distribution at the end of the 21st century. The simulated changes in climate affect mostly countries along the Guinean Golf (Ghana, Benin, Togo) and coastal areas of Congo-Brazzavile and Congo. The tropical forest in Central Africa (Gabon, Cameroon) is less affected. Because the simulated vegetation distribution is strongly dependent on the fire regime, we perform a sensitivity study to the parameters of the fire module in IBIS. We also use existing and new paleodata from the region to asses the likelihood of the simulated changes.
VEMAP vs VINCERA: A DGVM sensitivity to climate, CO2, nitrogen, and fire
The MC1 DGVM has been used in two large model comparison projects, VEMAP (Vegetation Ecosystem Modeling and Analysis Project) and VINCERA (Vulnerability and Impacts of North American forests to Climate Change: Ecosystem Responses and Adaptation). This talk will highlight the challenges faced with each new climate dataset: we will compare future climate scenarios from both VEMAP and VINCERA projects and how the different inputs have affected the carbon budget and vegetation distribution projections. In those two projects, MC1 has been compared with other models which included different assumptions about plant responses to climate change and atmospheric CO2 enhancement. We will compare observations with MC1's simulated plant response to similar environmental conditions to those of the FACE experiments and discuss the merits of the 'greenworld' vs. the 'brown-down' theories. We will show the impacts of including N limitation on model results and the uncertainty associated with the available N input information. Finally we will discuss the importance of human impacts such as fire suppression on simulation results.
Direct Physical Effects of CO2-Fertilization on Global Climate
CO2-fertilization affects plant growth, which modifies surface physical properties, altering the surface albedo and latent heat fluxes. Here we investigate how such changes to surface properties via CO2-fertilization, including changes in vegetation distribution, would directly affect the physical climate system. We know of no previous study that has investigated this question. Using a global three-dimensional climate-carbon model that simulates vegetation dynamics, we compare two multi-century simulations: a "Control" simulation with no emissions, and a "Fertilization-NoGHG" simulation where the land biosphere is fertilized as a result of prescribed CO2 emissions, but where the climate model sees no additional greenhouse gas forcing. Our simulations indicate that the direct physical effect of CO2-fertilization could be warming over a timescale of a few centuries; we obtain an annual- and global-mean warming of about 0.65 K over 430 years in our model. The average land warming is 1.4 K. We find that this warming is mostly due to the albedo decrease in the Northern Hemisphere boreal forest regions. This albedo-based warming could partially offset the century-scale cooling effect of additional CO2 uptake due to CO2-fertilization. Further study is needed to confirm and better quantify our results.
Modelling the Terrestrial Biosphere Under Long Term Climate Change Scenarios With a Complex Earth System Model
A complex coupled earth system model, consisting of an atmosphere and ocean general circulation model, an ocean biogeochemistry model, a 3-dimensional thermodynamical ice sheet model and a dynamic global vegetation model, was used to study the long term behaviour of climate and carbon cycle under increased atmospheric CO2 levels. A set of experiments was performed, forced with CO2 emissions from historical reconstructions (1750-2000), the SRES IPCC scenarios B1, A1B and A2 (2001-2100) and an exponential decay of the emissions for the period 2101-3000. The experiments give a reasonable reconstruction of the measured CO2 concentrations up to 2000. After stabilization of the CO2 concentration (between 2200 and 2500, depending on the scenario), the terrestrial biosphere equilibrates in a few hundred years. The ocean stays a sink for carbon all up to the end of the experiments. Carbon uptake by the terrestrial biosphere is fastest and largest in the tropical regions. The northern hemisphere high latitudes tend to store large amounts of carbon as well due to a northward shift of the boreal forests. This process is slower, but in the end of the same size as in the tropical regions. Offline experiments with only climate change or CO2 concentration as forcing are used to distinguish between climate change and CO2 change effects. CO2 increase alone causes an uptake of carbon by the terrestrial biosphere, the accompanying climate change reduces the uptake of carbon almost everywhere, mainly due to higher respiration rates caused by higher temperatures.
Simulations of site-level ecosystem responses to SRES climate scenarios using a DGVM
An extensively modified version of the IBIS dynamic global vegetation model was validated at several eddy covariance sites located in Canada and the USA. Each site was considered to be dominated by a single Plant Functional Type (PFT) as parameterised for IBIS. After extensive testing and adjustments, the finalized version of the model was run using site-level climate observations to generate stable soil, litter and vegetation carbon pools for the year 2000. These stabilized pools were then perturbed by running the model for an additional 100 years under each of six GCM climate scenarios (Canadian CGCM2, UK Hadley Centre HadCM3 and Australian CSIRO Mark 2), for each of the IPCC SRES A2 and B2 emissions scenarios. In each simulation, atmospheric CO2 concentration was assumed to increase as projected under the appropriate SRES scenario. Local climate observations were adjusted by combining them with decadal means of the GCM change fields previously interpolated to the site coordinates. The effects of changes in climate alone were also investigated by repeating the runs with atmospheric CO2 concentration fixed at the 2000 level. The general trends in PFT responses to the simulated climate at each site showed significant differences among GCMs, but aside from the greater climatic influences resulting from the warmer A2 scenarios, trends were consistent between the two emissions scenarios. Differences among the trends in key indicators including biomass and litter accumulation, and soil carbon decomposition, reflected regional differences in the scenarios of future climate as projected by each GCM. Changes in net primary productivity production and net ecosystem exchange also varied among the different scenarios but were generally positive. However, much of this positive response could be attributed to the projected changes in CO2 concentration, raising questions about the significance of elevated CO2 concentrations in large-scale simulations of vegetation change.
Estimation of Diurnal to Seasonal Ecosystem Parameters Using an Ensemble Kalman Filter
An important part of the creating complex numerical ecosystem models is to determine parameter values. Parameter estimation is typically carried out with non-sequential strategies such as least-squares fitting. However, a potential advantage of sequential methods such as ensemble Kalman Filter (EnKF) is that parameter values can drift through time in response to observations. This research explores how to use an EnKF to generate posterior distributions and seasonality of the model parameter values of a simple carbon cycle model using observed fluxes of carbon (C), weather, hydrology, energy, and remote sensing data at three forest sites: Howland (Maine, USA), Boreas (Alberta, Canada) and Niwot Ridge Forest (Colorado, USA). The analyses demonstrate that the model parameters, such as light use efficiency, respiration rates, minimum and optimum temperature and so on, are most highly constrained by eddy flux data at daily to seasonal time scales. Light use efficiency of the ecosystems demonstrates a strong seasonality and better constrained by C fluxes in the growing season. Results show that, via data assimilation and simultaneous estimation of parameter values, the prediction of GPP, respiration and NEE improved significantly compared with those predicted by the original model without data assimilation. However, a significant portion of the variances cannot be explained by the model (in forecasting mode) probably due to the simplicity of the model structure and errors in daily flux data (caused by measurements and estimation). Nevertheless, EnKF can be very useful in evaluating and developing ecosystem models, improving understanding and quantification of the C cycle parameters and processes, and aiding selection of eddy flux tower sites and measurement frequency.
Identifying critical parameters in a terrestrial ecosystem model for accuracy improvement through integration of data and/or assimilation
Previous studies on model inter-comparison among existing terrestrial ecosystem models agreed on basic features of the biospheres, but showing considerable differences in total estimates due to differences in model assumptions about vegetation structure, model parameterization and input datasets. In order to enhance the confidence in the estimates, the first step can be assimilation of critical parameters or integration of satellite-derived datasets in the ecosystem model employing most recent parameterizations and reliable vegetation structures. This research aimed at identifying such critical parameters for the purpose of integration of satellite derived data products and/or assimilation in the terrestrial ecosystem model SimCYCLE. SimCYCLE is a process-based model having five compartments and estimates monthly and yearly global terrestrial NPP (Net Primary Productivity) within reasonable limits. The process of identifying the critical parameters as well the integration of data and assimilation in essentially involved three steps; first, sensitivity analysis of selected internal and input parameters at different steps of a simulation; second, selection of input parameters based on the sensitivity analysis, integration of validated satellite derived datasets and validation of estimated NPP with ground truth GPPDI (Global Primary Productivity Data Initiative); third, selection of parameters for assimilation based on sensitivity analysis and assimilation based on validated NPP in the second step. With the constraint on the availability of reliable satellite derived datasets, Leaf Area Index (LAI) and fractional Absorbed Photosynthetically Active Radiation (fAPAR) were selected as the input parameters for integration of satellite-derived datasets. Specific Leaf Area (SLA) and the extinction coefficient (Kd) were the two internal parameters selected for assimilation.