Woody plant encroachment, that is, a marked increase in relative woody cover, is converting many semiarid savannas into woodlands. Most previous studies have simply reported the observed change in relative cover, implicitly assuming that woody cover increases linearly (i.e., by the addition the same amount of new cover in each time interval).

The objectives of this study were to evaluate whether encroachment would be better modeled by (1) incorporating density dependence and (2) incorporating the spatial configuration of existing woody patches in the model. To do this, a series of models was generated, each representing a particular hypothesis about the mechanism underlying encroachment (e.g. a constant rate of outward expansion of woody patch edge).

Chronological sequences of scanned aerial photographs, from 53 to 65 years long years, were constructed for three sites on the eastern Edwards Plateau of central Texas. After conversion to binary (woody/herbaceous) images, woody cover, number of patches, and total patch edge were measured in non-overlapping circular 56m radius plots. These data were used to estimate model parameters and to determine how well each model predicted the observed change in woody cover in each plot over each time interval.

Results/Conclusions

With only a few exceptions, the inclusion of density dependence substantially improved model performance. In a few instances, relative woody cover was apparently increasing linearly (i.e., addition of the same increment to cover in each plot regardless of initial cover).

In most instances, models that incorporated spatial configuration fit the data better than those that did not. The total length of patch edges was usually a better predictor of the increase in woody cover than was the number of woody patches. This is consistent with field observations of how woody patches expand: from the edges outwards. These results demonstrate that both density-dependence and spatial configuration have strong effects on the rate of woody plant encroachment.

We suggest that both factors be considered in studies of woody plant encroachment in other ecosystems. Our results identify specific hypotheses and the associated algebraic models for doing so.