We analyse the role played by two different approaches of spatial heterogeneity in theoretical models of annual plants dynamics. The first approach is called quasi-continuous gradient in which one type of resource is changing gradually along the gradient line. In the second one, called the patches approach, part of the habitat is covered by patches and the resource has a different value in each patch. We show that when the spatial heterogeneity of the habitat is small, the two approaches yield the same average number of surviving species, even if a small number of patches is used. In a strong heterogeneity it takes many patches to get similar results as in the gradient case. The difference between the gradient and patchy description of the spatial heterogeneity increases with the number of species present in the system. We have also shown that even when the average number of surviving species is the same, the abundances of species are ordered in a different way, like different species are the dominant ones. The conclusion of this paper is that modelling spatial heterogeneity in a system of plants is not a simple task. Special care is needed when the heterogeneity of the habitat is large, since then depending of the choice of a method, some predictions may differ significantly, making the model non-robust. Therefore the type of theoretical approach must closely match the modelled ecosystem.