Customarily I analyze field data using standard statistical approaches. Last fall, however, I showed a data set on habitat occupancy to a mathematical modeler. Over the next few months we worked out a predictive model tested this past spring. We discovered that our relatively simple mathematical model predicted system behavior with remarkable accuracy. The behavior of this system could have been analyzed statistically, but a successful predictive model gives us more confidence that we understand the interplay of forces driving the system.

As a field ecologist, three of the most important things I learned during this collaborative process were as follows:

1. Joining with a mathematical modeler is a very useful approach in carrying out some ecological field projects. Testing the behavior of a system against a successful predictive model is a more satisfying and powerful approach than describing it only statistically.

2. Mathematicians like to simplify––to reduce systems to their lowest terms. Field ecologists, by contrast, see subtleties and complexities in ecological systems that would make predictive models useless or unwieldy. A successful collaboration is one that is able and willing to hammer out a model that represents a workable compromise between these two tendencies.

3. Frequent sampling over significant periods of time may be a prerequisite to successful modeling. Some ecological patterns can be perceived only for systems that have been "densely" sampled.

In short, I believe that carefully developed mathematical models can and do work for some field ecological systems. Mathematical modeling is what makes physics "hard science". Field ecology will become progressively more "hardened" as it incorporates mathematical modeling techniques.