UNDERSTANDING ECOSYSTEM CARBON DYNAMICS BY MODELING APPROACHES

Abstract

Ecosystem models are a useful tool to explore ecological processes and their responses to climate change. The basic structures of current ecosystem C cycle models are similar and robust, but their uncertainties are high, especially when coupled with water and nutrient cycles and disturbance effects. In this dissertation, I studied three issues in ecosystem C cycle modeling: interactions between water and C processes, information contribution of theoretical basis (model structure) vs. observations (data), and ecosystem C storage capacity at disequilibrium state due to effects of disturbances. These three issues represent the basic theoretical problems in the development and application of ecosystem models: 1) how the representations of interactions among ecological processes affect the simulation of ecosystem C cycle? 2) Once a model is built up, how much information can be brought in by model calibration? 3) For large spatial C cycle modeling, how will the paradigm of ecosystem states affect our C cycle modeling?

In the first study, we evaluated the effects of soil hydrological properties on the interactions of water and carbon dynamics of a grassland ecosystem in response to altered precipitation amount and frequency, increased temperature, elevated atmospheric CO2 with changes in soil available water capacity (AWC). A process-based terrestrial ecosystem (TECO) model was used to simulate responses of soil moisture, evaporation, transpiration, runoff, net primary production (NPP), ecosystem respiration (Rh), and net ecosystem production (NEP) to changes in precipitation amounts and intensity, temperature, and CO2 concentration along a soil texture gradient. Simulation results showed that soil AWC altered partitioning of precipitation among runoff, evaporation, and transpiration, and consequently regulated ecosystem responses to global environmental changes. Fractions of precipitation that were used for evaporation and transpiration increased with soil AWC but decreased for runoff. High AWC could greatly buffer water stress during long drought periods, particularly after a large rainfall event. NPP, Rh, and NEP usually increased with AWC under ambient and 50% increased precipitation scenarios but increased from 7% to 7.5% of AWC followed by declines under the halved precipitation amount. Warming and CO2 effects on soil moisture, evapotranspiration, and runoff were magnified by soil AWC. CO2 effect on NPP, Rh, and NEP increased with soil AWC. Our results indicate that variations in soil texture may be one of the major causes underlying variable responses of ecosystems to global changes observed from different experiments. These results also imply that the interactions between C and water processes can be some soil texture.

In the second study, I evaluated the information contribution of model and observations to model predictions by a data assimilation approach. Eight sets of ten-year data (foliage, woody, and fine root biomass, litter fall, forest floor carbon (C), microbial C, soil C, and soil respiration) collected from Duke Forest were assimilated into a Terrestrial ECOsystem model (TECO) using a Monte Carlo Markov Chain approach. The relative information contribution was measured by the Shannon information index calculated from probability density functions (PDF) of carbon pool sizes. Our results showed that the information contribution of the model to constrain carbon dynamics increased with time whereas the data contribution declined. The eight data sets contributed more than the model to constrain C dynamics in foliage and fine root pools over the 100-year forecasts. The model, however, contributed more than the data sets to constrain the litter, fast soil organic matter (SOM), and passive SOM pools. For the two major C pools, woody biomass and slow SOM, the model contributed less information in the first few decades and then more in the following decades than the data. The knowledge on relative information contributions of model vs. data is useful for model development, uncertainty analysis, future data collection, and evaluation of ecological forecasting.

In the third study, I integrated the temporal patterns of C storage and spatial patterns of ecosystem states to develope a model to analytically describe relationships between ecosystem carbon storage and NPP, C residence time, and disturbance intervals and severity. The model represents a disequilibrium perspective for examining C storage dynamics in light of the impacts of disturbances and improves our predictive understanding of regional C dynamics. The carbon cycle at the scale of the ecosystem is almost always in dynamic disequilibrium with most ecosystems accumulating carbon at various stages of recovery with intermittent disturbances that release large amounts of carbon. This disequilibrium perspective is critical for scaling of site-level observations to estimate regional and global carbon sinks, for modeling studies on carbon-climate feedbacks, and for design of field experiments and observation networks.

These studies showed that current ecosystem C modeling protocols, i.e., a Farquhar model based canopy model simulating C input to the system and a compartmentalized C pool model simulating C allocation, transfer, and decomposition, work well in simulating the short-term patterns of ecosystem C dynamics, but have high uncertainties in simulating the interactions of multiple processes and are very sensitive to some parameters and boundary conditions. Data assimilation is an effective method to combine information from models and data and improve model parameterization and accuracy of predictions and reduce model uncertainties. However, once a model structure is given, optimizing parameters by data assimilation approaches can only find out the best agreement with observations within the space defined by the given model. The theoretical understanding of ecosystem dynamics is central to ecosystem modeling studies. As illustrated by our disturbance model (the third study), new theories and paradigms can fundamentally changes the way in which ecosystems are represented in models.