Managed aquifer recharge (MAR) can enhance the use of urban stormwater and treated wastewater. In addition to providing storage, aquifers can provide water quality treatment and therefore can be used as a treatment barrier within a multiple barrier approach to recycle urban waste waters. However, the treatment performance of a MAR scheme can vary considerably due to the heterogeneity of aquifers, the type of MAR scheme used and how it is operated. It is essential to improve the understanding of water quality treatment in aquifers, firstly to ensure the water quality recovered from a MAR scheme is fit for its intended use, and secondly to ensure the groundwater quality beyond the ‘treatment zone’ is protected from contamination. This study utilises a probabilistic method to evaluate aquifer treatment performance based on the removal of total organic carbon (TOC), total nitrogen (TN) and total phosphorus (TP) from urban stormwater, during aquifer storage and recovery (ASR) in a confined, limestone aquifer. Four long-term stormwater ASR schemes were used to assess nutrient removal. Operational water quality data from several injection and recovery events were combined and fitted to lognormal probability density function (PDFs) to represent the ‘injectant’ and ‘recovery’; these injectant and recovery PDFs were then used to derive a theoretical ASR removal efficiency (RE) PDF. TOC and TN removal was dominated by redox processes. Median removal of TOC was 50 to 60% at all four sites, while for TN removal was 40 to 50% at three sites with no change at the fourth. Median TP removal of 29 to 53% was attributed to filtration and sorption. It was concluded that the probabilistic method was able to quantify TOC, TN and TP removal capacity of the anoxic carbonate aquifer treatment barrier, which demonstrates that aquifers can provide effective natural treatment of urban stormwater.