Uncertainty Analysis of Phytoplankton Dynamics in Coastal Waters

Abstract

There is an increasing concern about the interactions between phytoplankton and coastal ecosystems, especially on the negative effects from coastal eutrophication and phytoplankton blooms. As the key indicator of the coastal ecosystem, phytoplankton plays an important role in the whole impact-effect chain. Primary production by phytoplankton forms the basic link in the food-chain. A lot of effort has been paid to the investigation of phytoplankton dynamics on the basis of literature surveys, field observations, and model predictions, providing a better understanding of the coastal ecosystem. In this thesis, the significance of phytoplankton is stressed and no discussion is given to zooplankton. Phytoplankton dynamics (i.e. growth, loss, grazing, biomass, bloom) is closely related to environmental variables, such as light intensity, temperature, nutrients, suspended matter, wind profiles, and tidal currents. In chapter 2, factor analysis is developed to characterize the contributions of the environmental variables to the phytoplankton biomass (in terms of chlorophyll a), determined by the 10-year’s historical record from 2000 through 2009 in the case study of the Frisian Inlet. In this thesis we focus on three elements of phytoplankton dynamics: phytoplankton growth, phytoplankton biomass, and phytoplankton blooms. Based on the specific properties of the case zones, the Frisian Inlet and the Jiangsu coast, different focuses are taken. Field measurement of phytoplankton dynamics is expensive, thus we use mathematical models as the useful and convenient tool to perform the investigation. The BLOOM II model and the phytoplankton model are introduced to investigate the annual variation of the phytoplankton biomass in coastal waters (chapter 3, chapter 4, and chapter 5). The reliability of the parameter estimation largely determines the confidence of the model output. The estimate function of the phytoplankton growth rate is controlled by the variables of temperature, light intensity and nutrients, separately or comprehensively. The phytoplankton needs light to grow through the photosynthesis process, whereas the light intensity is attenuated due to the absorption by chlorophyll a, salinity, organic matter, turbid water, and background extinction. Phytoplankton consumes nutrients, in turn, phytoplankton releases nutrients back to the water bodies through its death and the subsequent decay. In this research the growth rate is estimated with the effects of light intensity and ambient water temperature. The loss rate and the grazing rate are simplified as constants in the models, but actually are varied with the environmental variables. Moreover, the role of the vertical mixing process on the phytoplankton is significant, controlling the vertical distributions of the phytoplankton biomass and affecting the availability of light intensity and nutrients. Although a vertical phytoplankton model is discussed in chapter 4 and chapter 5, reducing the three-dimensional model to a one-dimensional model, the vertical mixing rate involved in both cases is processed with the Delft3D model. In this context, the estimation of the vertical mixing rate increases the applicability of the phytoplankton model. Chapter 4 discusses the effect of the vertical turbulent diffusivity on the variation of the phytoplankton biomass, driven by the physical and chemical conditions. Chapter 5 performs a similar study of the vertical mixing rate as described in chapter 4, but now only driven by the physical condition, as well as one driver (vertical stability threshold) of the occurrence of the phytoplankton blooms. The model prediction is always accompanied with the simplification, overestimating or underestimating the actual status. Thus, uncertainty analysis is required to be integrated with the model output. The uncertainty arising from the model output is focused, only a short discussion is given to the uncertainty arising from the input. The Bootstrap method and the Bayesian Markov Chain Monte Carlo simulation are approached to give insight in the model prediction with a characterization of uncertainty analysis.