Abstract
Spatial disposition of plants in intercrops, and differences in sowing time between species, can strongly affect their ecological interactions and, in consequence, the system's viability and performance. Empirical exploration of a wide range of spatial and temporal plant arrangements is costly and time-consuming. Modelling the growth of mixed crops is a tool which, combined with empirical tests, can greatly reduce the time and investment required for this task. Spatially explicit, individual-based dynamic models seem well suited for this purpose; their exploration and experimental validation for the case of simple, two-species, artificial plant communities, can also provide further insight as to how the spatial and temporal scales of a plant's multispecific neighbourhood affect its growth and performance. The aim of this investigation was to further develop a published spatially explicit individual-based mixed crop growth model [Vandermeer, J. H. (1989). The Ecology of Intercropping, Cambridge, U.K.: Cambridge University Press, p. 237], and to validate it experimentally. With this purpose in mind: (1) computer programs to simulate individual plant growth and to perform statistical analysis of both deterministic and stochastic versions of the model were developed; (2) the model was parametrized using a complex experimental diculture with several cohorts and spatial arrangements; (3) the predictive capacity of the model was tested using independent spatio-temporal experimental arrangements; (4) a modified version of the model was written, which abandons the assumption of linearity of the neighbourhood index at the cost of increasing the number of parameters; (5) The performance of stochastic versions of both Vandermeer's and our modified model were compared, employing a non-parametric measure of goodness of fit. We conclude that this approach to modelling plant growth subject to intra and interspecific competition is a remarkably efficient, general, conceptually elegant, heuristic tool whose predictive power can be further improved when nonlinear terms are introduced into the neighbourhood competition index, as done in our modified version of Vandermeer's model.