THE DYNAMICS OF INFECTION BY THE TAKE-ALL FUNGUS ON SEMINAL ROOTS OF WHEAT - SENSITIVITY ANALYSIS OF A STOCHASTIC SIMULATION-MODEL

The effects and interactions of inoculum density (I rho), rate of fung al growth (kf) and the maximum distances for primary (w(p)) and second ary (cross) (w(g)) infection on the temporal progress of infection of the take-all fungus, Gaeumannomyces graminis, in small populations of contiguous wheat plants are examined in factorial combination using a stochastic simulation model. The effects are analyzed in relation to i nfection progress curves for the percentage infected roots, the total length of infected root, the density of separate infection and the mea n length of separate infections as well as the densities of primary an d cross infections. Linear models are used to analyze the effects of c hanging parameter values on components of the infection progress curve s. Non-linear, logistic models are used to summarize infection progres s curves and to map the effects of changing simulation model parameter s onto the parameters of the simpler, growth curve function. The propo rtion of infected roots was most influenced by I rho and w(p) with a n egligible effect due to w(s). The density and length of infections on roots was principally controlled by kf. Changes in infection length we re non-monotonic. The average length of infections increased towards a temporary maximum, following the first wave of primary infections, di pped as infections overlapped and then increased rapidly as root colon ization progressed. The first wave of cross infection occurred 7 d aft er the initiation of primary infection. The density of primary infecti ons was controlled principally by I rho and w(p). The density of cross infections was controlled not only by kf and w(s), which are directly involved in cross infection, but also by I rho and w(p) which affect the amount of infected and susceptible tissue. The asymptotic paramete r of the logistic model for the infection progress curves was the most frequently affected parameter, followed by the delay parameter, with relatively few changes affecting the rate parameter. Some problems in the use of complex, stochastic simulation models to simulate experimen tal conditions are discussed.