Document: DAV-3-38-23

Alternative yield-density models for the study of plant competition.

HUMPHREY, L.D.* and E.W.SCHUPP

Utah State University, Logan, UT 84322-5230 ¹

Abstract:
Analysis of yield per plant as a functon of neighbor densities in planned species mixtures is often used to study plant competition under more controlled conditions than is possible in natural communities. The reciprocal yield regression model is commonly used to analyze such data. We compared the reciprocal yield model and two related models, one linear and one nonlinear, using data from five field experiments that we conducted to give an indication of the relative efficacy of these three models. Three experiments assessed effects of competition on two perennial grasses and one annual grass with each of the three species as neighbors. Two other experiments assessed responses of only perennial grasses. In all experiments, plots consisted of a plant from which responses were measured surrounded by neighbor plants, with a range of densities of each neighbor species. Yield was aboveground biomass produced over a growing season. All three models usually provided significant fits with acceptable residuals distributions. We compared fits based on back-transformed residual sum of squares because different transformations on yield and nonlinear regressions were used. The nonlinear model provided the best fit in 11 cases. The other linear model fit as well as the nonlinear model in six cases but was clearly better only once. The reciprocal yield model was best or similarly good in only three cases. However, conditions causing problems with interpretation sometimes existed with the nonlinear model.

Keywords: density, plant competition, reciprocal yield

This abstract is being presented at: 10:30 AM in session:
PLANT DEMOGRAPHY