A protocol towards a general theory of community ecology

Definitive answers are scarce in biology ^{1 2} but if we want the advance of ecological knowledge we should reach a point where we have a theory we can trust, and can be used for prediction.

Besides the use of experimental micro/mesocosmos has been criticized I advocate for its use in testing and generating ecological theory, using the following steps:

1. Do a experimental microcosm with a temporal and spatial homogeneous environment

2. Extend the experiment until the community reach steady state

3. Fit or compare with any community model of your taste

These were the 3 basic steps, but the fourth is the critical one:

1. Publish all the raw data in an open repository

If we have, different views, different theories, different models, all we need to reach such a theory is contrast it with data, data is the fundamental piece here, like genetic variability is to evolutionary theory, natural selection can not operate if there is nothing to select. The same happens to an ecologist, we can not select the model/theory if we don’t have data and it’s not enough with all the data one investigator can obtain in all its lifetime.

I emphasize another point: steady state, I think it’s very important because most models/theories assume a steady state, and much of the experimental microcosms studies don’t give enough information to assume that the systems is in steady state or not.

I agree that steady state should not be the most important thing if we want to emulate a natural system, but it is very important to test a model/theory. You can’t compare steady versus transient states, so a basis for comparison between experimental systems is that they reach a steady state. A criteria for steady state in ecological communities can be that the Shannon entropy of the system remain bounded to very small fluctuations around a fixed value ³.

So the next step is:

1. Contrast all data with a several models/theories using a rich set of predictions

I mean not only test eg species abundance distributions (SAD)⁴, but use more than one pattern like in pattern oriented modeling framework. It’s not bad to test only the SAD but a general theory should make more predictions than that.

The next step would be to translate the selected model/theory to natural systems, in which I expect the data will come from mostly observational studies, like LTER sites or NEON in the future.

This sounds almost like my research plan, but it will only work if other people take similar steps.

References