Percolation experiment

In every SIS, one intact soil column was collected using a stainless steel cylinder (diameter: 20 cm, length: 30 cm) that was knocked into the soil. The cylinder was excavated and the embedded soil column was pushed into a second steel cylinder (Figure ^S3). The bottom of the soil column was straightened with a knife and approx. 1–2 cm sand was added to ensure connection to a 0.45μm nylon membrane. The height of soil column F.3 was lower (21 cm) than of W.10 and V.18 (25 cm) because the soil layer of F.3 was shallower. During sampling, the vegetation was left on the soil column as far as possible. Where it was too long, it was superficially shortened. Plants were not removed so as not to disturb the soil structure.

The top of the soil column was connected to two storage vessels (Figures ^S5, ^S6). The first storage vessel was connected via a tube with a small liquid layer on the top of the soil (approximately 2 cm). The second storage vessel was connected to the first one with two tubes that kept the water level between the storage vessels and the soil column constant. At the bottom of the soil column, the solution was placed under tension with a 45 cm water head to quicken flow rates and thus reduce anaerobic conditions in the soil column.The bottom of the column was enclosed with a 0.45μm nylon membrane. The percolate was collected in 1 L-glass bottles that were placed on weighing devices (Figure ^S5) to calculate the flow rate. To avoid photodegradation, all vessels, tubes and bottles were wrapped with aluminum foil. The soil columns were saturated with deionized water from bottom to top at 7cmd-1. Thereafter, deionized water was percolated to reach constant flow conditions and to decrease the DOC load of the percolate. Pre-tests with Br- (NaBr, Carl Roth GmbH & Co KG, Karlsruhe, Germany) were conducted to set up a sampling protocol for each soil column. Before starting the main experiment, Br- was washed out by deionized water.

The main percolation experiment started with a 1.5 h flushing of the soil columns with deionized water. Subsequently, all water in the storage vessels and on the soil surface was replaced by a tracer/biocide solution. For each soil column, 15 L of initial solution consisting of deionized water, tracers and biocides were prepared. Target tracer concentrations were 50mgL-1Br-, 25mgL-1Cl- (CaCl2, VWR International GmbH, Darmstadt, Germany), 10mgL-1 UR (Simon & Werner GmbH, Flörsheim, Germany), 400mgL-1 SRB (Chroma GmbH & Co KG, Münster, Germany). Target biocide concentrations are based on commonly measured concentrations in facade runoff⁹³ and were 50mgL-1 each of diuron, terbutryn (NEOCHEMA GmbH, Bodenheim, Germany) and OIT (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany). The tracer stock solutions were prepared with deionized water and solid substances, while the biocides were already dissolved in acetonitrile by the manufacturer. UR and SRB solutions were stored in amber glass bottles and wrapped with aluminum foil to prevent photolytic decay.

The measured initial concentrations only slightly deviated from the intended concentrations (Table ^S5). The experiment lasted for 28 h (F.3) and 12.7 h (W.10, V.18) depending on the flow velocity. We excluded biodegradation of biocides since their half-times in soils are much higher than the duration of the experiment²⁹. Furthermore, terbutryn, diuron and OIT are assumed to be stable to aqueous hydrolysis^{28, 94}. In the beginning of the experiment, samples of the percolate were taken every 15 min, later every 20, 30, 45 or 60 min. An aliquot of the collected percolate was filled into 100 mL amber glass bottles for UR, SRB and biocide measurements, and 100 mL polyethylene bottles for Br-, Cl- and pH measurements. Samples were stored at approximately 6∘C for measurement for a maximum of ten days and frozen for longer storage. No changes in concentrations were observed in preliminary laboratory tests measuring biocide concentrations before and after storage (freezing of samples for multiple weeks).

The estimated total pore volume PVtot,est (L) of each soil column (Table 2) was calculated by the following:

The porosity (–) was estimated according to⁵⁰, who provide average porosities for soils in dependence of their texture and OM content. The estimated water volume that flowed through the column at time t (PVtLL-1) was calculated by dividing the outflow (L) at a certain time step by PVtot,est. PVt was used to normalize the percolated amount of water and make solute transport comparable. Maxima of BTCs of the solutes were estimated by calculating the mean breakthrough (%) between PVs of three and four. The saturated hydraulic conductivity Ks (cmh-1) was calculated according to Darcy’s law:

where q (cmh-1) is the water flow through the soil column, ΔH (cm) is the hydraulic head difference between upper and lower boundary of the soil column, and L (cm) is the length of the column. Data analysis was performed with R statistics (version 3.3.4)⁹⁵.