One of the most pressing challenges of the 21st century will be to feed a projected global population of 9 billion people whilst at the same time reducing humanity's agricultural and environmental footprint. With more than 50% of the human population now living in cities there is growing interest in the possibility of using urban agriculture (UA) to provide some of the food required by an expanding population. This thesis addresses two underexplored aspects of UA, namely its productivity and the role that biologically mediated ecosystem services play within it, using UA systems in Sydney, Australia, as case studies.<br/> I monitored the inputs and outputs of urban food gardens in Sydney for one year to assess their productivity and resource use efficiency through economic and emergetic analyses and used spatial analysis to model the productivity of scenarios in which UA production in Sydney was substantially expanded. I then conducted surveys and experimental trials to identify pollinators and natural enemies in these gardens and measured predation and pollination service provision using sentinel plants and insects.<br/> Urban agriculture, as performed by amateur gardeners, was found to be highly productive, yielding nearly twice as much produce per unit area as typical rural fruit and vegetable farms. However the inputs of labour and materials used in the process were substantial, calling into question the economic viability and sustainability of UA. Despite this, many of the inputs required could potentially be substituted for recycled materials common in urban areas, raising the possibility that alternative sourcing of inputs could significantly improve sustainability.<br/> Spatial modelling concluded that Sydney could produce around 15% of its total food supply, or its entire vegetable supply, using street verges, unused spaces (e.g. vacant land, park margins) and 25% of domestic yard space. Sufficient water and organic matter is available within the waste stream of the city to provide the irrigation and soil amendments required. However human labour would be a major limiting factor, with the amount of labour available under a scenario where the entire population worked at the same rate as typical amateur gardeners falling an order of magnitude short of what is required to manage the available land. A small professional workforce may be able to adequately manage the available land, however this would create social challenges related to access to private property and ownership of the produce derived from it.<br/> Pollinators within the UA systems studied were found to be diverse and highly active, resulting in sentinel plants not being pollen limited. The abundance and/or richness of floral resources within gardens was positively related to both visitation rate and richness of insect pollinators, however these same factors, surprisingly, exerted a negative influence on plant reproductive output. We did not find any relationship between pollinator communities and yield of food crops, but found these yields were highly influenced by the surrounding environment, likely mediated by abiotic factors.<br/> Natural enemies, namely vertebrates and flying insects, were found to remove a substantial proportion of sentinel prey. Investigations of the trap nesting fauna showed it to be dominated by wasps that attacked spiders, implying that spiders were abundant in the studied sites and raising the possibility that these could also play a major role in sentinel prey removal had they not been excluded by our experimental methodology. Relationships with individual environmental variables were found to be relatively weak, likely due the heterogeneous nature of the gardens and their surrounds.<br/> We conclude that UA can be highly productive and that urban areas contain many of the resources required for it to make a non-trivial contribution to overall food supply, including land, materials and ecosystem services. However social factors likely serve as a major barrier to the significant expansion of the practice.