Principal Investigator Ian Noble Project s58

Research School of Biological Sciences Machine VP

Co-Investigators Sandra Lavorel and Mark Stafford-Smith,

C.N.R.S., Montpellier, France and CSIRO, Division of Wildlife and Ecology, Alice Springs


Dynamics of Mistletoe Invasion of Fragmented Woodlands

The model investigates the effects of the fragmentation of woodlands on the spread of mistletoe populations. The model specifically examines the interactions between the biology of mistletoe dispersal and woodland pattern in terms of total tree cover and degree of tree agregation. Hierarchically patterned maps of tree pattern are generated. The population dynamics is simulated for each tree by following mistletoe cohorts. Dispersal by mistletoe birds is modelled as a function of distance between trees and attractiveness of trees depending on size and resident mistletoe fruit production. Parameter values are derived from field studies of the demography of mistletoe in New England and Central Australia.

The results are used to design field investigation of demography and bird behaviour. They are also used to suggest ways in which mistletoe spread could be controlled by manipulating tree pattern and disinfecting appropriate trees.


What are the basic questions addressed?

Theoretical questions concern the interaction of landscape patterns and the dispersal biology of organisms. We analyse which aspects of landscape heterogeneity (total fraction of trees and scale of fragmentation) affect dispersal most strongly. We also examine which parameters of the dispersal strategy are most important in determining population success in the case of a bird dispersed species.

Practical questions concern disinfection of mistletoe invaded woodlands. The aim of the model is to use theoretical results to generate and compare disinfection scenarios.

What are the results to date and future of the work?

A sensitivity analysis was carried out using a factorial design on landscape type and dispersal parameters. General linear modelling of the variation in final mistletoe population size showed that seed survival is the most important determinant of mistletoe population size on a given landscape.

Mistletoe population size was fitted to an exponential function of tree density. Population size increased less rapidly than tree density thus number of mistletoes per tree decreased for increasing tree densities. Degree of tree clumping for a given density level also influenced mistletoe population size. Larger populations were sustained by woodlands with a clumped tree pattern than by woodlands with scattered trees.

For a given level of survival, relative attractiveness of canopy volume vs standing fruit crop on a tree accounted for about one third of the variance in population size. Population size increased with increasing relative attractiveness of canopy volume. For lower seed survival, fraction of bird dispersal was nearly as important as the attractiveness ratio while tree density was less important. For higher seed survival, tree density became very important while fraction of bird dispersal played little role. Population size was negatively correlated with the fraction bird dispersal. The fraction of seed dispersal out of a bird's territory had a strong positive effect on mistletoe population size only for higher values (10%) which seem unlikely in the field.

The results support the hypothesis that woodland fragmentation may promote invasion by mistletoes. Although, with many more young individuals, simulated mistletoe populations for estimated parameter values were not a good fit to our natural population, the sensitivity analysis is valuable in focussing further efforts in field data collection. This study also illustrates how using a simulation model of population dynamics can help in determining a strategy of control of an invading organism. A reduction of seed survival and disinfection of larger trees would appear to be the most efficient strategy.

Further work will analyse in detail the spatial distribution and the age structure of the mistletoe population in relation to host tree distribution. This will allow further comparison with field data. An adaptation of the algorithm to simulate spread of clonal plants (eg. reeds) is in progress.

What computational techniques are used and why is a supercomputer required?

The model simulates the dynamics of mistletoes in a landscape consisting of host trees. First, a landscape pattern is generated using a hierarchical dichotomy for tree location. The landscape consists in a number of bird territories, each containing a certain number of trees. The procedure allows controlled experiments on landscape structure. The demography of mistletoes is modelled at the scale of each host tree using cohorts of mistletoes of equal age. The survival, growth and seed production are calculated for each cohort on each tree. Then seeds are dispersed between trees within a territory according to the fraction of bird dispersal, distance between trees and attractiveness of individual trees calculted in relation to size and mistletoe fruit crop. Dispersal has a stochastic dimension which requires that seeds be dispersed individually. This in particular, and the whole demographical dynamics which is organized in a spatially and demographically hierachical manner; yields a number of convoluted loops which are particularly appropriate for vectorisation. The present version of the program achieves about 90% vectorisation.

Publications

Spread of mistletoes (Amyema preissii) in fragmented woodlands: a simulation study. S. Lavorel, M. Stafford-Smith and N. Reid, Ecological Applications (in review).

Spread of mistletoes in fragmented woodlands, S. Lavorel, M. Stafford-Smith and N. Reid, Annual Meeting of the Ecological Society of America, Snowbird, USA (1995).