LD2-1: Past changes in climate-vegetation-disturbance interactions triggered changes in landscape processes that caused threshold changes to new stable states.

1.1. Interpretation of the Paleo-record

The sediment proxy data will be interpreted explicitly as climate, ecosystem, and fire states that can be used directly by ALFRESCO in calibration/validation and experimentation (Fig. 1).

1.1.1. Ecosystem history. ALFRESCO ecosystem types will be assigned to sediment samples based on pollen, stomate and macrofossil data. Ecosystem assignments using pollen data will follow rules developed as part of PAIN and CAPE (NSF-supported projects to describe circumarctic vegetation history: see http://www.ngdc.noaa.gov/paleo/parcs/atlas/beringia/polldata.htm). Stomate and macrofossil data will be used to confirm the local presence of P. glauca and Betula species. Ecosystem assignments will be used to initialize model runs (9.5 ka and 8.5 ka) and to assess the realism of simulation results (9.5-8.5 ka and 8.5-6 ka) (Fig. 1).

Fig. 1. Conceptual integration of paleo-data and ALFRESCO. The input data sets will be derived and extrapolated from the climate and ecosystem interpretations of the proxy data. The modeling component has two parts: (a) calibration and validation for 9.5-8.5 ka, and (b) experimentation to test ecological understanding for 8.5-6 ka by comparing simulated and reconstructed ecosystem dynamics. Light arrows indicates data used for model parameterization; dark arrows indicate independent data used to assess simulations.

1.1.2. Fire history. Of the parameters used to characterize fire regimes (frequency, intensity, severity and extent) only fire frequency will be estimated from the sediment charcoal record. Fire frequency information will be used in two ways (see below): 1) to calibrate the fire cycle subroutine in the calibration/validation period (9.5-8.5 ka), and 2) to assess the realism of simulated ALFRESCO fire history output for the experimentation period (8.5-6 ka) (Fig. 1).

1.1.3. Climatic history. The climatic interpretation of geochemical data will be aided by knowledge of: 1) the hydrologic and limnologic characteristics at each lake, described on the basis of field reconnaissance, topographic maps, and satellite image; 2) d¹⁸O and trace element content of lake water and carbonates (both abiotically-precipitated carbonate and ostracodes) in surface sediments; and 3) calibration of geochemical records from ²¹⁰Pb-dated cores, based on comparisons of geochemical trends at each site with instrumental weather data from the nearest weather station.We will use the quantitative proxy-temperature data (d¹⁸O) to assign four temperature classes for ALFRESCO. The coupling of Sr/Ca, Mg/Ca, and d¹⁸O data will help differentiate the signals of moisture vs temperature so that relative moisture classes can be assigned. Since the temperature and precipitation classes of ALFRESCO have defined effects on vegetation and disturbance, a complete climatic record (i.e., explicit growing-season temperature and precipitation values) is not required. Similarly, because ALFRESCO uses climate to drive ecosystem changes, only climatic trends (i.e., departures from the norm) are important.

1.2. Modeling and Integration with Paleo-records

We will use ALFRESCO to test whether our understanding of the causes of the 8.5-6 ka. P. glauca fluctuation is both internally consistent and consistent with the paleo-data. First, model development and validation will be done for the 9.5-8.5 ka period. Second, the improved model will be used in a series of experiments that examine the role of causal factors for the 8.5-6 ka period. Specifically, we will evaluate the relative effects of different factors (from migration rates to fire frequency to system feedbacks) by comparing paleo-records of ecosystem change with results of model experiments and by performing sensitivity tests.

1.2.1. Model input. ALFRESCO requires two types of input derived from the paleodata (Fig. 1). The first consists of maps defining the initial distribution of ecosystems (tundra, P. glauca forest, and deciduous forest) across the study area at 9.5 ka and 8.5 ka. These will be generated in a GIS, based upon point data from the paleorecords and our best estimate as to the spatial arrangement and extent of each ecosystem type in the study landscape. The 9.5 ka ecosystem distribution will be used for initial model calibration and validation (see below). The 8.5 ka distribution (which should be similar to the endpoint ecosystem distribution simulated during calibration and validation runs) will be the initial conditions for tests of causal factors. The second input is a temporal record of paleoclimatic classifications derived from the geochemical analysis, beginning 9.5 ka. We will use an interpolation procedure to express the geochemical reconstruction of climate in decadal time steps for ALFRESCO. For each decade, growing-season temperature and precipitation classes will be mapped, using the same methods as for ecosystem distributions. Thus the reconstructed climate will explicitly define the climate scenarios used in model simulations (i.e., climate input will be “hardwired” rather than stochastically simulated).

1.2.2. Model calibration and validation (9.5-8.5 ka). Model calibration involves the development of logical and consistent rules and assumptions. During calibration we will use the 9.5-8.5 ka paleo-record to parameterize the model, so that our assumptions and rules reflect the empirical data. True validation of ALFRESCO is impossible due to the stochastic nature of the model (i.e., fire ignition and spread) and the strong influence of site history on stand and landscape-level dynamics. However, we can show that the simulation results are reasonable and within the range of reconstructed observations. Once validated in this sense, we will have confidence to test our understanding of the causal mechanism(s) responsible for the fluctuation in P. glauca.

We will address two model-development tasks. The first issue is to develop flammability algorithms for each ecosystem type, using fire-frequency information from the sediment record to inform the fire routine. The current model was calibrated to provide realistic values of the number of fires and area burned under currently observed vegetation and climatic conditions. Changes are currently being made to the fire routine to better incorporate fire-history information and to provide more realistic simulations of the interactions among climate, vegetation, and fire. We will use the charcoal analysis to develop a coarse resolution map of fire history for the study region for the period 9.5-8.5 ka. Specifically, fire-frequency data will be used to calculate fire cycles and to develop flammability algorithms for each ecosystem type.

The second model-development issue involves the simulation of the migration/recruitment process. The long-standing disagreement between the rates of tree migrations observed in the paleo-record and those predicted from life history considerations has resulted in new insights into dispersal theory. Current research identifies the need to represent both local and long-distance dispersal dynamics in models of landscape-level vegetation change. We propose to address this question by developing and testing several dispersal routines representing the current state of dispersal theory (i.e., long-tail versus maximum-threshold approach). For example, we will conduct simulations using the long-tail dispersal theory and compare the results to a maximum-threshold approach (ALFRESCO’s current approach). From those results we can then develop alternative approaches for representing the dispersal process.

Model validation will demonstrate that ALFRESCO is reasonably able to simulate the ecosystem patterns and migration trends indicated by the paleo-data from 9.5-8.5 ka. Although, true validation will be impossible (see above), we will assess the ability of the model to: 1) realistically simulate an endpoint vegetation distribution similar to the reconstructed distribution at 8.5 ka, and 2) simulate reasonable migration rates for P. glauca. A direct assessment of the simulated and reconstructed fire histories will not be conducted because the reconstructed fire history will be used to parameterize the model. However, we will conduct a cursory model assessment for this period, focusing on the realism of the simulated spatial variability of fire across the study area and the sensitivity of fire to differences in regional climate and vegetational distribution.

1.2.3. Model experiments (8.5-6 ka). We will use the newly calibrated and validated version of ALFRESCO to test alternative explanations for the patterns and trends observed in Picea distribution on the landscape during the 8.5-6 ka period. Our first task will be to simulate vegetational trends, given the initial (8.5 ka) ecosystem distribution and the climate scenario reconstructed from the geochemical analysis. If we have properly calibrated and successfully validated the model, the simulation results should reflect the spatial and temporal changes in ecosystem distribution and fire frequency identified independently by the paleo-record. In this assessment, the ecosystem assignments of pollen samples will be interpolated to decadal or multi-decadal intervals for comparison with the decadal time steps of model simulations. In addition, because the sediment fire-frequency data will not be used internally by ALFRESCO during this period, these data can be used for comparison with the simulated fire history, providing an independent check on the model’s ability to realistically simulate past fire regimes.

Our second task will be to explore whether there is more than one plausible fit to the paleo-record, and, if so, to examine each as an alternative scenario. This task will allow us to assess what processes and interactions were potentially responsible for the early-to-mid Holocene treeline changes and to evaluate the sensitivity of each identified mechanism. For example, we may find that altering fire frequency (within the range of reconstructed observations) and/or using different dispersal methods (i.e., kernal versus long-tail) produce results similar to the historic observations. Specifically, we will determine the relative importance of seed availability, migration rate, climate constraints on establishment, and fire spread in determining the spatial and temporal dynamics across the landscape. As an example, a sensitivity test for the rate of species migration would systematically alter the migration rate (kilometers per decade), while keeping all other parameters constant.

In summary, the integration of paleo-data and ALFRESCO will allow us to explore how climate, vegetation, and disturbance regime may have interacted in the past. The paleo-data are a critical reality check on the validity of ALFRESCO, because they are the only data with a long enough temporal perspective to accurately evaluate changes in vegetation in response to climate. Insights gained and improvements to the model will increase our ability to realistically simulate the future. Real-world questions that can be better addressed using the improved model include: How strongly does climatic warming affect a forest patch within tundra (i.e., a refugium) relative to a forest patch within forest? Will large treeless patches, such as might be created by large fires respond differently to climate than small patches? Does the arrangement of patches matter to treeline advance (e.g., a uniform forest margin vs. scattered islands of trees or expansion along river corridors)? And ultimately, under what circumstances can climate-vegetation-disturbance interactions result in nonlinear responses of boreal ecosystems to climate change?