Integrating Experimental Findings and Process-Based Models
In EDGE, modelers and experimental ecologists are working in close collaboration to incorporate field data into an integrated experiment-modeling framework (Figure 6.4), where data are used to estimate parameters of the process-based model, TECO. In turn, this process identifies information gaps, for which additional sampling efforts can be conducted within the experiment. For example, some variables being measured in EDGE concerning C cycling processes are aboveground and belowground standing crop biomass, aboveground litter, soil respiration, and total soil C content. By mapping and comparing TECO output to these measured variables using a Markov chain Monte Carlo technique (Metropolis-Hastings algorithm; Hastings 1970), multiple model parameters are able to be estimated for each site. Additionally, model parameters not well constrained using the current level of data inform where additional sampling efforts should be applied.
FIGURE 6.4
Conceptual representation of the integrated experiment-modeling framework employed by EDGE to assess mechanisms behind differential ecosystem sensitivity to drought across grasslands spanning climatic gradients and having various ecosystem attributes. Data obtained from the multisite experiment are integrated into the process-based model, TECO, to identify scaling rules for model parameters. These, in turn, are incorporated into regional-scale projections that allow for broader testing of hypotheses related to future ecosystem status under altered climate regimes.
Once parameters are able to be properly estimated, they can then be linked with the climatic conditions across sites to provide insight into how ecosystem processes as represented by model parameters and climate should covary. Second, examination of parameter estimates in different sites, but within the same climate envelope, gives insight into how parameters should be assigned based on ecosystem attributes separate from climatic context. For example, despite both having mean annual precipitation close to 350 mm and mean annual temperature around 9°C, paired sites in northern Colorado and southern Wyoming have very different functional composition with one site being dominated by C4 warm-season grasses and the other by C3 cool-season grasses. These different functional groups have very different growth strategies and water use efficiencies, which will impact how these grasslands respond to alterations in rainfall. Variation of estimated parameters in these grasslands provides insight into how mechanisms driving ecosystem responses to drought may differ along with plant functional types and how model projections should be scaled across ecosystems varying in plant functional types. Using the information about how model parameters vary across space due to environmental context and ecosystem attributes allows for model projections to be scaled up from single-sites to the regional level. Once scaling rules are established and incorporated into process-based models, hypotheses are then able to be tested at larger spatial scales.
Both the design of the EDGE study and integration of experiment and process-based models are extremely important for this study's success. We suggest that, to provide much needed predictive power under altered climate regimes at regional scales, future studies examining ecosystem sensitivity to environmental drivers should consider implementing consistent methodology across study sites at least when experimental sites exist within the same biome. Assimilating data across those regional sites would improve model performance upscaling from distributed experiments.
