Bacterial degradation of dissolved organic carbon in the water column: an experimental and modelling approach

Eichinger, E. 2008. Bacterial degradation of dissolved organic carbon in the water column: an experimental and modelling approach. PhD-thesis, Vrije Universiteit, Amsterdam (2008/01/23) & Université de la Méditerrannée (2007/12/21)
Nederlandse versie


This thesis deals with the growth of heterotrophic pelagic bacteria which use the dissolved organic carbon (DOC) as nutritive resource. It is widely recognised that heterotrophic bacteria play a predominant role in the carbon cycle. Indeed, they represent the most important living biomass in aquatic ecosystems and constitute the major DOC consumers. DOC is the second most important stock of bioreactive carbon in ocean and its dynamics is important for understanding the global carbon cycle and changes of atmospheric CO2. DOC may play an important role in the biogeochemistry of the oceanic carbon cycle as it contributes to the biological pump by the export of sinking biogenic particles. The carbon flow through the bacterial compartment is investigated by the bacterial growth efficiency (BGE). BGE provides an estimation of the DOC fraction that is used by bacteria for their growth, the rest being remineralised. Numerous studies investigated the influence of environmental factors on BGE values. These factors generally comprise temperature, season, distance from the shore and substrate quality.

This thesis aims to utilise both experimental and modelling approaches to investigate growth of heterotrophic bacteria. The experimental work used natural as well as artificial seawater. Various models for bacterial growth, comprising different levels of complexity, were investigated to represent mathematically the dynamics of the different experiments. Two main axes merge in this work: (1) the study of growth models, constructed from experimental results, with a view to implement them in ecosystem models, and (2) the investigation of the environmental factors influencing the BGE with these models. The main objective consists of the study of bacterial growth in different environmental contexts and to deduce a suitable mathematical formulation for describing the interaction between growth and DOC to include this in a biogeochemical model later on.

We first studied bacterial and DOC dynamics from in situ samples. Numerous biodegradation experiments, implying natural DOC and bacterial assemblages, were realised in stable conditions in the Northeast Atlantic Ocean according to several seasons and depths. We utilised the Monod model for representing the data acquired during these experiments. This model is empirical, constructed with Michaelis-Menten kinetics and is the most widely used for describing bacterial growth in ecosystem models. BGE was estimated experimentally, as generally done by most authors, and also by using the model. We demonstrated that BGE varies according to season and depth, the dynamics being the same with both methods of estimation. As BGE is one parameter of the Monod model, this result proved that it is inaccurate for representing the utilisation of DOC by bacteria in ecosystem models.

We then decided to carry out experiments in artificial conditions, with a monospecific bacterial strain and a single DOC substrate. This setup provides data sets that are easier to analyse and allows the application of more complex models. To test the performance of several models, comprising several levels of complexity, including the Monod model, 2 kinds of experiments were performed: 1 experiment was realised in constant conditions with a single substrate load at the start of the experiment, as the previous study, the other experiment was carried out by pulsing the substrate supply periodically. The total amount of substrate was the same in both experiments, the only difference consists of the input regime. The substrate pulses mimic the spatial and temporal variability of DOC distribution. We demonstrated that the Monod model is inaccurate to represent bacterial dynamics when they are in starved conditions, which may often occur in natural environments. We also utilised a model implementing the bacterial maintenance, the Marr-Pirt model, and another model, constructed from the dynamic energy budget (DEB) theory, including maintenance as well as a reserve compartment. Both models match the data very well. However, the DEB model, due to its mechanistic basis, is more flexible and is able to adapt to more situations. BGE was estimated experimentally and with the 3 models for both experiments. We demonstrated that BGE is higher in the pulse experiment than in the experiment carried out in stable conditions with all methods of BGE estimation. Consequently, the spatial and temporal variability of DOC distribution has a profound impact on BGE value.

Data of the pulse experiment were also used to formulate a mechanistic model, based on the DEB theory as stated previously. In a third section, we investigate this model more profoundly and its potential inclusion in ecosystem models. We adapted a bacterial growth model with the theory in order to account for the processes highlighted by the experiment. The model was first improved by considering 2 maintenance processes: when mobilised reserves are sufficient, maintenance is realised from the reserve pool, the remaining energy being used for growth; however, when the reserve flux is not enough to sustain maintenance, growth ceases and maintenance is realised from the reserve plus the structural volume and the cell shrinks. When maintenance is performed from the structure, the model permits the release of refractory material in the medium by bacteria. Maintenance was modelled in this way to account for the increasing non-used DOC in the culture. This model is quite complex to represent only a bacterial component and is thus difficult to implement in ecosystem models. The original model, comprising 4 state variables, was thus reduced to a system of 2 differential equations which may be easier implemented in global models. This result has a profound impact in the context of global modelling, as model simplification allows easier calibration, simulation and the understanding of model outputs.

The results highlighted by this thesis were obtained thanks to the coupled experimentation-modelling approach. The experiments revealed key processes and facilitated the construction of models on the basis of biological insights, and models allow highlighted gaps in the knowledge which is required for a better representation of the system. Consequently, models may suggest new experiments to be performed and the best strategy alternates repeatedly between experiment and modelling.

Full text in pdf format

This is the symposium that concludes my project

Marie's project page