Conundrums in mixed woody–herbaceous plant systems

Aims To identify approaches to improve our understanding of, and predictive capability for, mixed tree–grass systems. Elucidation of the interactions, dynamics and determinants, and identification of robust generalizations that can be broadly applied to tree–grass systems would benefit ecological theory, modelling and land management. Methods A series of workshops brought together scientific expertise to review theory, data availability, modelling approaches and key questions. Location Ecosystems characterized by mixtures of herbaceous and woody plant lifeforms, often termed savannas , range from open grasslands with few woody plants, to woodlands or forests with a grass layer. These ecosystems represent a substantial portion of the terrestrial biosphere, an important wildlife habitat, and a major resource for provision of livestock, fuel wood and other products. Results Although many concepts and principles developed for grassland and forest systems are relevant to these dual life-form communities, the novel, complex, nonlinear behaviour of mixed tree–grass systems cannot be accounted for by simply studying or modelling woody and herbaceous components independently. A more robust understanding requires addressing three fundamental conundrums: (1) The treeness conundrum. What controls the relative abundance of woody and herbaceous plants for a given set of conditions at given site? (2) The coexistence conundrum. How do the life-forms interact with each other? Is a given woody–herbaceous ratio dynamically stable and persistent under a particular set of conditions? (3) The net primary productivity (NPP) conundrum. How does NPP of the woody vegetation, the herbaceous vegetation, and the total ecosystem (woody þ herbaceous) change with changes in the tree–grass ratio? Tests of the theory and conceptual models of determinants of mixed woody–herbaceous systems have been largely siteor region-specific and have seldom been broadly or quantitatively evaluated. Cross-site syntheses based on data and modelling are required to address the conundrums and identify emerging patterns, yet, there are very few data sets for which either biomass or NPP have been quantified for both the woody and the herbaceous components of tree–grass systems. Furthermore, there are few cross-site comparisons spanning the diverse array of woody–herbaceous mixtures. Hence, initial synthesis studies should focus on compiling and standardizing a global data base which could be (1) explored to ascertain if robust generalizations and consistent patterns exist; and (2) used to evaluate the performance of savanna simulation models over a range of woody–herbaceous mixtures. Savanna structure and productivity are the result of complex and dynamic interactions between climate, soils and disturbances, notably fire and herbivory. Such factors are difficult to isolate or experimentally manipulate in order *Correspondence: Joanna I. House, Max Planck Institute of Biogeochemie, Winzerlaer Strasse 10, 07745 Jena, Germany. E-mail: [email protected] **NCEAS (National Center for Ecological Analysis and Synthesis) work group on Tree–Grass Interactions: M.B. Coughenour, M.B. Dodd, J. Gignoux, D.O. Hall, N.P. Hanan, R. Joffre, X. Le Roux, J.A. Ludwig, J.-C. Menaut, R. Montes, W.J. Parton, J.J. San Jose, J.C. Scanlan, J.M.O. Scurlock, G. Simioni, B. Thorrold. Journal of Biogeography, 30, 1763–1777 2003 Blackwell Publishing Ltd to evaluate their impacts at spatial and temporal scales appropriate for assessing ecosystem dynamics. These factors can, however, be evaluated with simulation models. Existing savanna models vary markedly with respect to their conceptual approach, their data requirements and the extent to which they incorporate mechanistic processes. Model intercomparisons can elucidate those approaches most suitable for various research questions and management applications. Conclusion Theoretical and conceptual advances could be achieved by considering a broad continuum of grass–shrub–tree combinations using data meta-analysis techniques and modelling.