PS 15-125 - A test of demographic matrix models based on deviations from stable distribution

Monday, August 3, 2009
Exhibit Hall NE & SE, Albuquerque Convention Center
Martha M. Ellis, Wildlife Biology Program, University of Montana, Missoula, MT and Elizabeth E. Crone, Biology, Tufts University, Medford, MA
Background/Question/Methods

Demographic popualtion models are widely used in basic and applied ecology to predict long-term population dynamics from shorter term (e.g., 3-7 year) studeis.  Since long-term monitoring is inherently difficult, study length is going to be a major factor limiting inferences on population dynamics for most species.  Long-term studies, in combination with simulations, provide an opportunity to look at how study length affects our ability to capture population dynamics from shorter-term studies.  In this project, we use a 20-year study of plant population dynamics to evaluate the ability of different models and study lengths to describe features of population dynamics.  We expected these data to accurately estimate mean vital rates, but we hypothesized that long time series would be necessary to describe variation in population dynamics.  Specifically, we compared simulation models based on subsets of years to statistics based on the full time series.  For example, we tested model performance by comparing the amount of variation in population demographic structure  (distance between stage structure in each year/iteration and predicted stable stage structure) in actual plant populations with the amount of variation produced in simulated populations.

Results/Conclusions

Preliminary results suggest that simple matrix-selection simulations do not capture the full range of variation in a population structure, even in when parameterized  with the full long-term data set.  Simulations underestimated the distance that demographic structures get from the theoretical stable structure of the model.  While using fewer years leads to more variable results, as the number of years of data included in simulations increases, the difference in average distances from stable for the simulated versus observed population data stabilizes.  These results suggest that the lack of fit between simulated and observed population data for long-term data sets is due to limitations in model design, and not necessarily study length.  This work lays the groundword for a larger project evaluating the impact of deviations from a stable stage structure for applications of demographic models, and the structure of demographic variance in natural populations.

Copyright © . All rights reserved.
Banner photo by Flickr user greg westfall.