High concentrations of dissolved nutrients in stormwater have been identified as contributing to eutrophication of receiving waterways near urban areas. To reduce dissolved nutrient concentrations in stormwater a range of devices such as wetlands and bioretention systems are used. Bioretention systems are increasingly employed for their supposedly high nutrient removal capacity, however very little is known about their treatment efficiency or the chemical and biological mechanisms controlling their function. This research aimed firstly to test and compare the efficiency of different bioretention system designs for the removal of dissolved nutrients from stormwater, and secondly to investigate the chemical and biological mechanisms responsible for the nutrient removal (sorption, microbial uptake, and plant uptake). Bioretention mesocosms were built in plastic containers (1 m x 0.5 m x 0.5 m). Three different media treatments were built, representing those most commonly used: gravel, fine sand and loamy-sand. To assess the nutrient removal capacity of plants, vegetated and unvegetated examples of each media type were made. The mesocosms were regularly irrigated with tap water for six months, and then regularly irrigated with synthetic stormwater for a further six months to ensure that the treatment performance assessed would represent fully established systems. The synthetic stormwater solution was based on field measurements of stormwater, and was made using a combination of inorganic chemicals and organic fertilisers. By incorporating organic carbon and major cations (Ca, Mg, Na, K), the measured treatment performance of the biofilters would be more realistic than previous studies that did not corporate these compounds. Some mesocosms were watered only with tap water so that the effect of frequent fertilisation (enrichment) could be compared. It was expected that vegetated media would enhance nutrient removal directly through plant uptake, and indirectly by stimulating microbial productivity and microbial uptake in the rhizosphere. Nutrient removal was evaluated by comparing the influent to the effluent. Detention times of 24 and 72 hours were compared to test if longer contact periods resulted in greater nutrient removal. The mesocosms were also flushed with tap water (no nutrients) to determine the proportion of entrained nutrients that might subsequently leach from the media. Vegetated bioretention mesocosms were much more efficient than unvegetated systems at removing total nitrogen (63 – 77 % removal compared to -12 – 25 %) and total phosphorus (85 – 94 % removal compared to 31 – 90 %). The vegetation effect did not improve dissolved organic carbon removal but there was a difference between soil types, with smaller particle size media removing more organic carbon. Enriched mesocosms removed similar quantities of nutrients to non-enriched mesocosms. Extending the detention time from 24 hours to 72 hours slightly increased the removal of total nitrogen from the vegetated mesocosms, but reduced total nitrogen removal from unvegetated mesocosms. When flushed with tap water, inorganic and organic forms of nitrogen and phosphorus leached from the unvegetated mesocosms, but were mostly retained within the vegetated mesocosms...