Jenn Nelson
Interactions among organisms are of primary importance in theoretical ecology, particularly in community ecology. A community is basically a group of interacting populations of organisms. How populations interact determines how they evolve. A given population interacts with many others. The relative strength of these interactions therefore determines the relative influence on the population. However, populations need not interact directly to affect each other.
We may consider an indirect interaction between two organisms is as a chain of direct interactions. It is clear that in order for there to be an indirect interaction, groups of species (communities) must contain at least 3 species. For example, given two competing species of plants, a grass and a wildflower. The plants affect each other directly by diminishing the growth of the other. Assume now there is a cow that feeds only on the grass, and does not interact with the wildflower in any way. The cow can still affect the growth of the wildflower by feeding on the grass. By decreasing the amount of grass, the cow can indirectly help the wildflower. Thus the cow indirectly interacts with the wildflower through the direct interaction of eating the grass, and the competition between the grass and the wildflower.
There are several ways to quantify the direct interactions among groups of organisms, but it is much harder to identify and quantify the indirect interactions. Direct interaction strengths are summarized in what are called community matrices. A community matrix is matrix in which the entries represent the effect of the column species on the growth of the row species. Using the premise that an indirect interaction is a chain of linked direct interactions, we used these matrices to quantify indirect interaction strengths.
The relative importance of indirect interactions as compared to direct interactions in community dynamics is currently unknown. The purpose of this project is to analyze how the strength of the interaction chains varies among groups of species. It is possible for these strengths of these indirect interaction chains to vary with the size of the community considered and the length of chain considered. The length of the interaction chain is the number of direct interactions within the chain. Thus this project looks at how the strength of the indirect interactions vary.
This project is an extension of research done at Kansas State University during the summer of 1998, currently in preparation for publication. The previous work included the creation of an extremely crude program to process the matrices. The program sufficed to generate the data, however, in order make it easier to use (for anyone other than me) it will be necessary to redesign the interface.
In the original implementation of the program the chain strength was found by determining the minimal direct interaction strength within the interaction chain. Community matrices from many different literature studies were used as the data set for this analysis. Thus to facilitate the comparisons among the different studies, it was necessary to standardize the matrices prior to analysis. The current method was a simple re-centering of the data. As this is a controversial procedure, the method used to standardize the matrices is currently being reevaluated. An additional component of the investigation looked at the normality of the distribution of the interaction strengths. The analysis involved testing the strengths using Kolmogorov-Smirnov Goodness of Fit test.
The first part of this project involves a simple alteration of the program to alter the method of computing the chain strength. As mentioned, the program currently determines the minimal chain strength. This is a very conservative estimate, as the strength of the influence most likely becomes diluted through the direct interactions of the chain. Thus the proposed change is to compute not the minimum of the direct interaction strengths, but rather the product. The second change to the underlying theory of the project is a possible change in the method of standardizing the matrices.
The original program was written very basically and crudely. It will be necessary to design and implement a workable user interface. The program needs to be able to output the data to a text file in order to move the information more easily to a spreadsheet. The program also needs an easier method to input the matrices to the program. This may involve a console window and most likely some form of a Graphical User Interface.
Another facet of this project is visualization of data. The output of the program is currently a very, very long list of numbers, followed by a few calculation results, such as the mean and variance, repeated several times. This obviously can be improved to facilitate analyzing the data, and navigating through the vast quantities of numerical output.
One of the questions looked at in the original manuscript/investigation of the project is the normality of the direct interaction strengths. To do this "boiler-plate" Goodness of Fit tests of normality were used. An alternate approach, which may be utilized on an experimental basis, is instead performing a transformation on the data and then graphing it, to visually check the normality of the strengths. As part of this project, this alternate method will be tested.
The current weak point of the project is the possible lack of an empirically testable aspect of the project. It may be possible to compare what happens as species levels change in a very small community (perhaps only 3 species). The change in species levels could be determined by considering only direct interactions and then considering indirect interactions in conjunction with direct interactions. Then comparing the differences in the predictions of population level changes, however it may be hard to ascertain which model is more accurate due to lack of experimental data.
Thus, despite the potential difficulties of empirical testing of the model, this project has definite potential for software development, in both the reprogramming the methods of calculating the chain strengths, and designing a workable interface and file output system. The project also includes questions about how to manipulate a very large data set, and ways to display the vast quantities of data generated by the program.