Ecosystem Network Shifts As Indicators of Climate Response

Monday, 15 December 2014
Allison Eva Goodwell, University of Illinois at Urbana Champaign, Urbana, IL, United States and Praveen Kumar, University of Illinois, Urbana, IL, United States
Ecosystem states evolve due to complex interactions over various space and time scales. Process networks, in which nodes are time series variables and directional links are measures of information transfer, provide a method to analyze an ecosystem in terms of feedbacks, information transfer, and synchronization. It has been shown using FLUXNET data and ecohydrological modeling that variables such as precipitation, soil temperature, soil moisture, and heat fluxes exhibit forcings and feedbacks that are altered during periods of climate extremes such as drought. In this study, we use methods to deal with short datasets to observe shifts in network behavior over hourly to daily timescales. We compute network properties including transfer entropy, mutual information, and net system transport. To test our methods, we first generate chaotic test networks of various sizes and connectivity structures. It is found that a single feedback between two nodes causes a “self-feedback” to be detected at both nodes, which propagates throughout the network causing complete connectivity at predictable timescales. Depending on the symmetry of feedbacks and overall connectivity, a network may partially or completely synchronize. We then apply our methods to evaluate short-term ecosystem responses to climate extremes in agricultural landscapes in Illinois. We use 30- minute flux tower data from Bondville, IL, and 1 to15-minute data from recently installed weather stations and flux tower in the Sangamon River watershed to analyze network structure before, during, and after rainfall events or dry periods. Simulations in MLCan, a plant-atmosphere-canopy model, are performed to incorporate unmeasured nodes involving photosynthesis and soil hydrology. We compare the structure of feedbacks, forcings, and synchronization to vegetation response as measured by LAI or NDVI, in addition to a comparison with our test networks. This type of analysis can identify the feedbacks and links critical for ecosystem resilience or vulnerability to climate extremes.