[Truncated abstract] In the crop growing areas of Western Australia, two economically important weed species, Lolium rigidum Gaud. (annual ryegrass) and Raphanus raphanistrum L. (wild radish), have evolved widespread herbicide resistance to multiple chemistry groups. Consequently, grain growers in the region have adopted an integrated approach to weed management that includes many non herbicide tools, however many more are needed as these weed species become increasingly more difficult to control. This thesis examines, in a series of field trials carried out in the Western Australian crop growing area, the potential for weed seed predation of annual ryegrass and wild radish by naturally occurring granivores as a new weed management tool for grain growers . . . The study discusses the implications of these results with the view to manipulating predation of weed seed through agricultural management practices. Ants were shown to be the dominant seed predator in this environment, especially in the centre of fields. The study has identified that the ant species Melophorus turneri (Forel), Monomorium rothsteini (Forel), Pheidole hartmeyeri (Forel) and Rhytidoponera metallica (Smith) are potential biological control agents for annual ryegrass seeds while P. hartmeyeri was identified as the only species suitable for biological control of wild radish seed pods. Ants were found to be sensitive to disturbance and some to crop residue type and these effects are discussed in relation to seed removal. This study of weed seed predation in agricultural fields is the most complete in this environment and can be used to inform further work in this area. It has identified that naturally occurring granivores can be used as a weed management tool.

Over the past thirty years, price relativities and technological development have motivated an increase in the area of land allocated to cropping, as opposed to pasture production, throughout the central wheat belt of Western Australia. Nevertheless, reducing the proportion of pasture in these rotations has challenged the future productivity of farming systems in this area. First, the frequent application of selective herbicides for weed control in extended cropping rotations has promoted the development of herbicide resistance in a number of major agricultural weeds. Second, the primary use of annual plants has promoted the development of soil salinisation by allowing a significant proportion of rainfall to recharge saline water tables. The inclusion of perennial pasture phases between extended periods of cropping may mitigate or delay these constraints to production through (a) allowing the use of costeffective forms of non-selective weed control, and (b) through creating a buffer of dry soil that absorbs leakage occurring beneath subsequent crops. This study consequently explores the value of including perennial pasture phases in dryland agricultural systems in the eastern-central wheat belt of Western Australia, accounting for benefits related to herbicide resistance and water table management. A novel computational algorithm for the solution of multiple-phase optimal control problems is developed and used to conduct a conceptual analysis of the value of lucerne (Medicago sativa L.) pasture for managing annual ryegrass (Lolium rigidum Gaudin), the primary weed in wheat belt cropping systems. The competitiveness and fecundity of annual ryegrass provide strong economic incentives to maintain a low weed population, irrespective of herbicide-resistance status. Consequently, the ineffectiveness of selective herbicides primarily reduces the profitability of cropping by motivating the adoption of more costly non-selective forms of weed control. The inclusion of lucerne in land-use rotations is only optimal in the presence of severe herbicide resistance given (a) the low efficiency of alternative weed-management practices available during the pasture phase, relative to selective-herbicide application; (b) the significant cost of establishing this perennial pasture; and (c) the high relative profitability of cereal production in the absence of resistance. The value of lucerne, relative to annual pastures, for weed management is explored in greater detail through the use of compressed annealing to optimise a sophisticated simulation model. The profitability of candidate rotations is also manipulated to account for the long-term production losses accruing to the recharge of saline groundwaters that occurs beneath them. Sequences incorporating lucerne are only more profitable than those that include annual pasture at the standard set of parameter values if (a) annual ryegrass is resistant to all selective herbicides, (b) the water table is so shallow (approximately less than 3.5 m deep) that frequent rotation with perennials is required to avert soil salinisation, or (c) sheep production is highly profitable. The value of perennial pasture is sufficient under these circumstances to overcome its high establishment cost. Consistent with intuition, these benefits are reinforced by lower discount rates and higher rates of leakage occurring beneath annual-based systems. Formulation of an effective communication strategy to report these results to producers is justified given the complexity involved in determining the true magnitude of these intertemporal benefits through alternative means, such as field trials.