Fossil Creek Software

About POP-II
Using POP-II
On modeling
POP-II / III Users
License & contact
About the author
Return to the FCS home page

On modeling with POP-II

Modeling with any simulation program is an art. It is impossible in these few pages to convey the range of methods one might use—or difficulties one may encounter. However, Fossil Creek Software offers some general words of wisdom and encouragement for potential users of POP-II and POP-III.

Principle One: Procrastination should be avoided. It is easy to postpone trying to simulate a population because you think that you do not have enough good data. You will not know what enough data really means until you try a simulation. Similarly, good is a relative term (see Assessing the Model). Part of the modeling game is to test your data – to use the model as a "lie detector" for your data. So, put together a test case; see what happens. If you wait until your data are TRUE, you will be waiting forever. Besides, if your data were flawless, you would not need a model. Models help most in cases where your data is only fair.

Principal Two: Arrange your data in order of reliability. In other words, list the types of data you are working with first; then, order that list by how much you trust the data. Typically, harvest figures will be at the top of the list and natural mortality or wounding loss will be at the bottom. However, each set of data is different.

Principal Three: Test the model against what you know has happened. Choose your best observations as the input data and try to simulate the population's history from them. Initially, the model probably will not behave as you anticipate. The appropriate parameters to examine are population size trend, sex ratios, percent yearlings in the harvest, and age structure of the population. Obviously, you will want to compare the simulation's output with any observations you may have.

Principal Four: Vary the test data in sequence. The appropriate next step is to alter the least reliable data in your list. Vary it through what you consider a reasonable range. If you are able to make the model better mimic what you believe accurately reflects the history of the population, then you are on the right track. Often, you will need to retreat to the "next level of unreliability." Vary the next worst data throughout its reasonable range. Then ask yourself if you have a satisfactory simulation. This is the process of alignment by parameter calibration.

Principal Five: Pay close attention to your starting point. You should keep in mind that events early in the simulation's chronology have a more profound effect on the population than do events in later years. You should pay particular attention to the initial population size. One piece of wisdom learned through many years of population modeling is that most people underestimate their population's size by at least 10%. Other things like a surge in reproductive rate or a bad winter early in the simulation can have similar "momentum."