Research

The Workbench as a tool for Research

A new approach to biological oceanography. Bridging the gap between observations and modelling. Build your hypothis into the specifification for a Virtual Ecosystem. Apply the Ecological Turing Test, using field observations collected for that purpose.

Throughout the 20th century biological oceanographers have pursued a strategy best described as scientific exploration. They did not consider their science was ready for the kind of experimental approach pursued by physicists. Their goal was to find out what was in the plankton ecosystem and how it changed with geographical location and time. The subject followed the doctrine of Francis Bacon. Progress came from making discoveries and proposing theories to explain them. That was a long way from the method of conjecture and refutation advocated by Karl Popper. There was simply no way formally to test their theories. Virtually Ecology solves that problem. It offers a new approach to biological oceanography.

The Ecological Turing Test provides a formal procedure for testing conjectures about the plankton ecosystem. The conjecture is embedded in the specification for a virtual ecosystem. The test compares emergent properties of that simulation with observations of the same property. As in experimental physics the comparison takes account of the uncertainties in both the simulation and the observation. The test succeeds only if the difference between observation and simulation is statistically significant, given those uncertainties. If not, the simulation cannot be shown to be wrong, and the conjecture survives until a sharper test can be devised. That might involve targetting a different emergent property, or collecting a better data set, or both.

In the 21st century, field operations will still be needed to discover new aspects of the plankton ecosystem. We are still living in the era of discovery, and what we know is poorly documented in many regions. However some field operations should be designed for ETT, i.e. to collect data to try to refute our conjectures about the way that the plankton ecosystem works. These ETT field operations must be preceeded by extensive research with virtual ecosystems, aimed at establishing the robustness of each conjecture before embarking on the much higher cost of fieldwork.

Given that research strategy, the challenge is to identify conjectures about the plankton ecosystem that are scientifically interesting and which can be tested given the inevitable uncertainties in simulations and observations. This will involve a deep knowledge of the biological oceanography and familiarity with the techniques of virtual ecology. It will also need the skill of an information scientist to estimate the uncertainties in observations, including sampling errors in a patchy ecosystem. One way to start down this path is to re-examine some of the well-established theories of biological oceanography. Here are some examples.

Sverdrup's theory of the spring bloom

Woods & Barkmann (WB'93) re-examined Sverdrup's classical theory for the light-limited growth of phytoplankton in the spring bloom. They showed that while the basic mechanism is correct, the emergent properties of a virtual ecosystem are rather different from those derived from Sverdrup's simplified model. In particular, the inclusion of diurnal variation starts the growing season at the winter solstice, i.e. three months before the Sverdrup criterion is met. Continuous Plankton recorder monitoring provides some evidence to support that result. However, there has been no attempt to challenge Sverdrup's theory with the kind of rigor provided by the the ETT. To do so the virtual ecosystem must include all relevant phenomena, including diurnal variation. One of the results of WB'93 was that there is no clear convergence of the depths of the mixed layer and the photosynthetic compensation depth,which is the key emergent property of Sverdrup's model. So the ETT must be based on some other emergent property that can be measured with sufficiently low uncertainty.

Incubation

One of the classic methods to measure primary production is to incubate phytoplankton in a bottle held all day at a fixed depth in the mixing layer. That procedure does not take account of rapidly changes in ambient phytoplankters as turbulence advects them randomly up and down. Barkmann & Woods (1996) made a virtual ecosystem which included a sample of plankton held at a fixed depth, as in a virtual incubation bottle. They compared primary production in the bottle with that of the population of phytoplankton moving freely in the mixing layer. The difference a factor of two (suprisingly large!). This numerical expriment showed how it is possible to use a virtual ecosystem to calibrate incubation measurements performed in a bottle in the sea.

The f-ratio of new and regenerated production

The classical way to measure the f-ratio is to compare the uptake of nitrate and ammonium by phytoplankton. We sought to predict f-ratio as an emergent property of a virtual ecosystem. To do so we designed a model that discriminates between new and regenerated production by labelling nitrogen in five categories. The first is “preformed” at the start of the growing season. The second has been taken up a phytoplankter since the start of the growing season. The third and fourth have been remineralized by bacteria acting on the corpses of dead phytoplankton and on faecal pellets of herbivorous zooplankton. The fifth is the result of remineralization of faecal pellets from carnivorous zooplankton. It takes only an hour to code such a model by editing an existing model with an appropriate plankton community. The emergent pattern of f-ratio based on these variables suggest targets for testing by ETT.

Seasonal succession

Nogueira et al. (2004) created a virtual ecosystem with three species of phytoplankton. They were all in the same functional group, so they shared the same biological equations. But they had different sizes: 20, 40 & 60 µm ESD respectively, which gave them different values for the parameters in those equations. The values followed allometric equations that had been established experimentally by others. The three species showed a seasonal succession as a stationary emergent property of the virtual ecosystem. That provides a target for ETT with observations based on [an instrument] that samples the phytoplankton by size.

Biodiversity

Competition between species depending on shared resources is the quintessence of community ecology and natural selection. Numerical experiments by Woods & Wiley (WW'01) have measured the time to extinction of competing varieties of diatoms that differ in just one model parameter, namely the amount of energy stored in a diatom immediately after cell division. The unexpected result was that the extinction time was typically a few decades. And that varieties close to extinction could rise up the rankings to become dominant if the ambient climate changed to make them better fitted to the upper ocean environment. Such changes occur on time scales shorter than the extinction times of many plankton. So the ocean supports a rich community, most of which are on their way towards extinction at any place, but are likely to be saved by ocean circualtion before they become extinct. That is a solution to Hutchinson's famous paradox of the plankton. The numerical experiment opens the way to prognostic biodiversity.