Spatial pattern formation in semi-arid shrubland: a priori predicted versus observed pattern characteristics Frank M. Schurr 1,2, * , Oliver Bossdorf 1,3 , Suzanne J. Milton 4 and Jens Schumacher 1,5 1 Institute of Ecology, University of Jena, Dornburger Str. 159, 07743 Jena, Germany; 2 Current address: UFZ Centre for Environmental Research Leipzig-Halle, Department of Ecological Modelling, Permoserstr. 15, 04318 Leipzig, Germany; 3 UFZ Centre for Environmental Research Leipzig-Halle, Department of Community Ecology, Theodor-Lieser-Str. 4, 06120 Halle, Germany; 4 Department of Conservation Ecology, University of Stellenbosch, Private Bag X1, 7602 Matieland, South Africa; 5 Max-Planck Institute for Biogeochemistry, Carl-Zeiss-Promenade 10, 07745 Jena, Germany; * Author for correspondence (tel.: +49 341 2352491; fax +49 341 2353500, e-mail schurr@oesa.ufz.de) Received 20 September 2002; accepted in revised form 27 June 2003 Key words: Community structure, Functional groups, Null models, Root competition, Seed dispersal, Spatial sta- tistics Abstract Ecologists increasingly use spatial statistics to study vegetation patterns. Mostly, however, these techniques are applied in a purely descriptive fashion without a priori statements on the pattern characteristics expected. We formulated such a priori predictions in a study of spatial pattern in a semi-arid Karoo shrubland, South Africa. Both seed dispersal and root competition have been discussed as processes shaping the spatial structure of this community. If either of the two processes dominates pattern formation, patterns within and between shrub func- tional groups are expected to show distinct deviations from null models. We predicted the type and scale of these deviations and compared predicted to observed pattern characteristics. As predicted by the seed dispersal hypo- thesis, small-scale co-occurrence within and between groups of colonisers and successors was increased as com- pared to complete spatially random arrangement of shrubs. The root competition predictions, however, were not met as shrubs of similar rooting depth co-occurred more frequently than expected under random shrub arrange- ment. Since the distribution of rooting groups to the given shrub locations also failed to match the root compe- tition predictions, there was little evidence for dominance of root competition in pattern formation. Although other processes may contribute to small-scale plant co-occurrence, the sufficient and most parsimonious expla- nation for the observed pattern is that its formation was dominated by seed dispersal. To characterise point pat- terns we applied both cumulative uni- and bivariate K-function and local pair- and mark-correlation function techniques. Based on our results we recommend that future studies of vegetation patterns include local charac- teristics as they independently describe a pattern at different scales and can be easily related to processes chang- ing with interplant distance in a predictable fashion. Introduction The distribution of plants within a community is rarely random Hutchings 1997; Freckleton and Wat- kinson 2000. Spatial structuring of plant populations and communities is a consequence of restricted dis- persal and the action of mortality factors such as competition, herbivory and abiotic stress that often have a local and species-specific impact. Since spa- tial patterns reflect underlying ecological processes, the statistical analysis of these patterns has become increasingly popular among plant ecologists e.g. 271 © 2004 Kluwer Academic Publishers. Printed in the Netherlands. Plant Ecology 173: 271–282, 2004.