MODELLING SIZE SPECTRA OF MICRO-ORGANISMS IN PELAGIC ECOSYSTEMS

Abstract

Microbial size spectra reveal the structure and function of planktonic communities, which vary temporally and spatially for different pelagic ecosystems. It is necessary to develop models to be used to investigate factors affecting community dynamics, and analyse the response to environmental disturbances. In this research, models are built to study the temporal and vertical variabilities in plankton size spectra, as well as bulk environmental parameters (primary production, nitrate and ammonium), and the relationships between characteristics of size spectra and bulk environmental parameters in different ecosystems. Theoretical and empirical approaches have been used in model construction. The models are calibrated using field data from BATS (Bermuda Atlantic Time Series) station off Bermuda and Massachusetts Bay collected in 1992 and 1993.

As a first step, a I-Dimensional theoretical model is constructed to simulate the changes in phytoplankton size structure and primary production in response to changes in the structure of the water column. By introducing cell size of phytoplankton, the rates of transfer processes including metabolic, grazing and sinking rates are able to be calculated from published allometric equations. The effect from light conditions, grazing stress and nutrient availability are taken into consideration.

The model reproduced fairly well the general overall features of the studied oceanic ecosystem. The fit of simulated primary production to the field observation at BATS station is good, with correlation coefficient R 2 greater than 0.8. The model depicts a late winter/ early spring bloom and low production during the summer and fall, and a deep phytoplankton biomass maximum structure in summer as well. The logarithmically transformed phytoplankton biomass size spectra in the model are compared with field data, showing a stable, almost flat spectrum for the oceanic environment.

Owing to the fluctuations in most ecosystems, efforts have also been made to develop empirical "grey box" and "black box" models. The results from the stepwise regression and multiple linear regression models ("grey box") show that the size spectra characteristics of phytoplankton size spectra may be correlated with bulk environmental parameters. Neural network ("black box") models are also applied in this study to model characteristics of microbial size spectra. The results show high values of correlation coefficients (close to unity) for training data sets, meaning the models reproducing the previous observations are good. The correlation coefficients for verification data sets are much higher than those of the multiple linear regression, especially for those outputs with much fluctuations, e.g. characteristics of size spectra from coastal areas.