1 We developed a spatially explicit and individual-based simulation model describing the dynamics of tree populations across treeline ecotones. Our aims were to identify minimal factors and processes able to generate treeline types with abrupt vs. smooth transitions in different variables (tree height, age, density), to investigate the role of positive feedback in pattern generation, and to find out why krummholz appears at some treelines, but not at all. We hypothesised that a different balance between smooth growth and mortality gradients across the treeline ecotone could account for differences between commonly observed treeline types.

2 The model contained only processes and factors we regarded as essential for producing a treeline ecotone and was parameterized with an extensive individual-based data set from Pinus uncinata treelines in the Spanish Pyrenees. However, parameters expected to influence treeline type were systematically varied.

3 The simple model was able to generate major treeline types differing in abruptness and krummholz abundance. The most important factors determining treeline types were the relative strength of the growth and mortality gradients, followed by facilitation strength, whereas demographic parameters accounted for more subtle differences.

4 We developed a tentative classification scheme to locate treelines types within a two-dimensional state space spanned by two parameters defining the growth and mortality gradients. Only certain combinations of growth inhibition, mortality, and facilitation allowed emergence of certain treeline types characterised by contrasting abruptness values and the presence or absence of krummholz.

5 High krummholz densities emerged only under positive feedback and strongly increasing growth inhibition across the ecotone. Abruptness in adult tree density was positively correlated with facilitation strength and growth inhibition. In contrast, treelines with abrupt height transitions occurred only, where both low growth inhibition and a strongly increasing mortality occurred across the ecotone.

6 Our analysis suggests that treeline features are not arbitrary but there is a clear signal in the pattern which allows for inference of the underlying processes. Our approach of a systematic comparison of model predictions and various observed patterns can be widely applied for testing hypotheses on the functioning of ecological systems and for deriving specific explanations to be verified in the field.