My research program aims to develop novel modeling techniques for elucidating ecological patterns in a watershed context and to assist environmental management. During the last ten years, I have developed several mechanistic and statistical (both Bayesian and frequentist methods) models to address eutrophication problems, quantify land-sea interactions, identify climate change effects on aquatic ecosystem phenology, and study food web dynamics. My research plans/interests can be classified into five main categories:
One of my main research interests is the development of eutrophication models for water quality management purposes. I have recently developed a complex eutrophication model for Lake Washington that simulates multiple elemental cycles (org. C, N, P, Si, O) and multiple functional phytoplankton (diatoms, green algae and cyanobacteria) and zooplankton (copepods and cladocerans) groups (University of Washington, Professor M.T. Brett). This model will be used for testing alternative watershed management schemes and assess their effects on the lake's dynamics. In addition, I conducted a meta-analysis of 153 modeling studies, which represent almost everything that was published relevant to the topic of aquatic mechanistic biogeochemical modeling from 1990 to 2002. This study provides a critical review of the current modeling practice and discusses several guidelines towards a more rational model development. In this direction, my main research interest is the application of methods that refocus from single optimal parameter selection to a distribution of parameter sets that provide acceptable (behavioral) model performance. This calibration form will address the well-known equifinality problem (i.e., several distinct choices of model inputs that lead to the same model outputs) and provide the basis for estimating model prediction error associated with model parameters. My plan is test several methods of uncertainty analysis and Bayesian parameter estimation (e.g., Generalized Likelihood Uncertainty Estimation, Markov Chain Monte Carlo) and provide a comparative description of the strengths and weaknesses, advantages and disadvantages regarding their efficiency as surrogates of the posterior parameter distributions. This research aims to provide a much needed analysis/demonstration of the benefits and the practical difficulties of several recently developed parameter estimation techniques in a way that can assist the common water quality modeling practice, and thus provide new methodological tools for natural resource management and sampling design for water quality monitoring programs.
My research interest in watershed-aquatic ecosystem interactions has its roots in my Ph.D. dissertation, where I developed an integrated methodology for assessing the contribution of non-point pollution sources (agricultural runoff and soil erosion) to coastal marine eutrophication (University of the Aegean, Professor M. Karydis). I studied the dynamic processes of both terrestrial and marine ecosystems through an extensive sampling network and developed watershed, hydrodynamic and eutrophication models. I have also assessed the urbanization effects on stream water quality dynamics in the greater Seattle region, based on both short (daily) and long (decade) time scales. My current research interest mainly involves the effects and propagation of uncertainty in integrated environmental modeling systems (coupled watershed-receiving waterbody models). The application of these coupled modeling constructs involves substantial uncertainty contributed by both model structure (e.g., spatiotemporal resolution mismatch) and parameters. My interest is to show how Bayesian techniques can allow existing data to refine our knowledge of model input parameters, obtain insight into the degree of information the data contain about model inputs, and obtain predictions and uncertainty bounds for modeled output variables.
Lakes and streams are particularly sensitive to the ecological impacts of climate forcing, and several long time-series have shown a close coupling between climate, lake thermal properties and individual organism physiology, population abundance, community structure, and food-web dynamics. Thus, understanding the complex interplay between meteorological forcing, hydrological variability, and ecosystem functioning is essential basic knowledge for assisting risk assessment and water resource/fisheries management. During the last three years, several colleagues and I have shown the effects of climate variability on the Lake Washington thermal structure, timing of the spring bloom, coupling of the trophic interactions between phytoplankton and zooplankton, interspecific niche differentiation, and the sockeye salmon (Oncorhynchus nerka) behavioral patterns. I plan to continue this research and investigate the effects of climate variability on the North American lake ecosystem phenology. I hope to use for this study several of the most renowned Canadian freshwater monitoring programs with uniquely detailed records for weather conditions, hydrological flows and physical, chemical and biological variables. My plan is to develop novel modeling techniques (Bayesian Hierarchical/Dynamic Linear Models, Structural Equation Modeling) to detect spatiotemporal trends in the physical structure, water chemistry and the food-web dynamics of North American lake ecosystems. This combination of mechanistic and empirical approaches is often highlighted as the optimal modeling framework to study how climate signals cascade through natural ecosystems, and how they shape abiotic variability and/or biotic responses. Plankton communities will be a central component of this research due to the socioeconomic impacts of their potential structural shifts (e.g., bottom-up forcing on commercially exploited fish stocks) and their complex interactions with the atmospheric CO2. My research acknowledges the importance of two complementary directions of research: firstly, the need to elucidate the wide array of in-lake processes that are likely to be affected by the climate change; and, secondly, the need to examine the heterogeneity in responses between different waterbodies. The rational of this approach and its importance for dealing with uncertainty in ecological forecasts is advocated by several recent review papers.
I also collaborate with colleagues (Professor S.K. Golfinopoulos and Professor A.D. Nikolaou) from the University of the Aegean, Greece, to develop quantitative tools (statistical and dynamic models) for predicting the formation of chlorination by-products in water treatment plants. Currently, our focus is the improvement of an earlier dynamic modeling approach, where the water treatment plant is represented as a mixed flow reactor and the formation of total trihalomethanes (TTHM) is predicated on a generalized reaction of total halogens with an organic precursor. The combination of a steady-state approach with Bayesian updating seems to be a promising framework.