Climate change, population growth and urbanisation contribute to an increasing demand for freshwater resources. In order to face this challenge, stormwater harvesting for aquifer storage and recovery has gained interest over the past decades. However, pollutants from various surfaces are transported with the runoff, posing a threat for receiving groundwater upon aquifer infiltration. Biofiltration is a low-cost and low-energy technology that uses natural processes to improve the stormwater quality. Heavy metals are a contaminant of concern, because of their lack of degradability and toxicity at low levels. So far, removal potential has been shown in lab and column tests, but field research is limited. Furthermore, most biofiltration studies focused on removal of total metals, whereas dissolved metals are known to have higher bioavailibilities and thus toxicities. This research aims to gain more insight in the processes in field scale biofilters, by monitoring and analysing the concentrations of metals (Fe, Mn, As, Co, Cr, Cu, Ni, Pb, Zn) in the inlet and the outlet of a stormwater biofilter in Spangen, Rotterdam. In the analysis, a distinction was made between the total metal concentrations, consisting of dissolved and suspended metals, and dissolved metal concentrations. The main chemical water composition was also monitored, to get a better picture of the conditions in which the biofilter was operated. This information was used to identify how the design and operation of the system can be improved with regards to metal removal. Additionally, a transport model in PHREEQC was used to estimate the lifetime of the biofilter before breakthrough occurs, and how operational choices affect this lifetime. Preferential flows and short-circuiting as a result of design- and operational choices were uncovered by analysis of the electrical conductivity. Heterogeneous distribution of water on the filter and the inlet located closely to the outlet contributed to this. A great variation in hydraulic conductivity supported these observations. The hydraulic conductivity was generally much higher than is recommended for biofilters. Additionally, the system was irrigated with recovered water from the aquifer to avoid a foul smell of extracted water and keep the biofilter wet during longer dry periods. This means that the system was not only fed with stormwater, but also with water from the aquifer. Feeding of two different water sources with different compositions lead to the increase of total and dissolved concentrations of As, Co, Cu, Ni \& Pb in the outlet, as well as higher dissolved Zn concentrations. Phoshate, ammonium, and total and dissolved concentrations of Fe, Mn, and Cr on the other hand decreased. The effect of mixing of two different waters was accounted for, by estimating the expected concentration of each pollutant if mixing was the only mechanism effecting their concentration. This was done using mixing fractions based on the electrical conductivity. Differences in estimated and measured concentrations showed that the release and removal of various pollutants was not a mere result of mixing only. The change in water composition and subsequent competition for sorption spots was found to be the main mechanism involved. The transport model showed that preferential flows resulted in C/C0 = 0.8 breakthrough of metals occurring much faster than in a plug flow. These observations show that the current design and operation of the biofilter do not provide adequate removal of metals in the biofilter.