Use of column experiments to investigate the fate of organic micropollutants – a review

Although column experiments are frequently used to investigate the transport of organic micropollutants, little guidance is available on what they can be used for, how they should be set up, and how the experiments should be carried out. This review covers the use of column experiments to investigate the fate of organic micropollutants. Alternative setups are discussed together with their respective advantages and limitations. An overview is presented of published column experiments 10 investigating the transport of organic micropollutants, and suggestions are offered on how to improve the comparability of future results from different experiments. The main purpose of column experiments is to investigate the transport and attenuation of a specific compound within a specific sediment or substrate. The transport of (organic) solutes in groundwater is influenced by the chemical and physical properties of the compounds, the solvent (i.e. the groundwater, including all solutes), and the substrate (the aquifer material). By adjusting these boundary conditions a multitude of different processes and related 15 research questions can be investigated using a variety of experimental setups. Apart from the ability to effectively control the individual boundary conditions, the main advantage of column experiments compared to other experimental setups (such as those used in field experiments, or in batch microcosm experiments) is that conservative and reactive solute breakthrough curves can be derived, which represent the sum of the transport processes. There are well-established methods for analyzing these curves. The effects observed in column studies are often a result of dynamic, non-equilibrium processes. Time (or flow 20 velocity) is an important factor, in contrast to batch experiments where all processes are observed until equilibrium is reached in the substrate-solution system. Slight variations in the boundary conditions of different experiments can have a marked influence on the transport and degradation of organic micropollutants. This is of critical importance when comparing general results from different column experiments investigating the transport behavior of a specific organic compound. Such variations unfortunately mean that the results from most column experiments are not transferable to other hydrogeochemical 25 environments but are only valid for the specific experimental setup used Column experiments are fast, flexible, and easy to manage; their boundary conditions can be controlled and they are cheap compared to extensive field experiments. They can provide good estimates of all relevant transport parameters. However, the obtained results will almost always be limited to the scale of the experiment and not directly transferrable to field scales as too many parameters are exclusive to the column setup. The challenge for the future is to develop standardized column experiments 30 on organic micropollutants in order to overcome these issues. Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c © Author(s) 2016. CC-BY 3.0 License.

micropollutants show redox-sensitive behavior under groundwater conditions (e.g. Banzhaf et al., 2012;Barbieri et al., 2011;Burke et al., 2014;Heberer et al., 2008;Massmann et al., 2008) it may be possible to use such compounds or groups of compounds to determine redox zones in complex anthropogenically-influenced settings. Greskowiak et al. (2006) were able to demonstrate a correlation between temperature-dependent redox zonation in groundwater and the fate of phenazone at an infiltration site in Berlin. 5 In view of the above-mentioned concerns regarding the presence of organic micropollutants in aquatic environments, and especially in groundwater, there is a clear need to develop a sound understanding of how they are transported and behave in groundwater. Laboratory experiments on the transport and eventual fate of organic micropollutants under defined boundary conditions will always be important because the boundary conditions for field studies are poorly known, which affects the transferability of their results to other systems. This paper therefore provides a review of published column experiments 10 investigating the properties and transport behavior of organic micropollutants, since such experiments provide a suitable setup for this task. The relevant transport properties of organic micropollutants are first presented, followed by a discussion of which compounds and which of their properties can be investigated using the experimental setup of a column experiment. The weaknesses and problems, as well as the advantages, of different experimental setups will be discussed and finally, other laboratory methods that can be used to investigate organic micropollutants are compared with column experiments. 15

Factors affecting the transport of micropollutants in groundwater and in column experiments
The transport of (organic) solutes in groundwater depends on the chemical and physical properties of the compounds, the solvent (i.e. the groundwater, including all solutes), and the substrate (the aquifer material). The main processes of solute transport are advection and hydrodynamic dispersion. The movement of solutes can be retarded compared to that of the containing groundwater, mainly as a result of sorption. Oxidation-reduction reactions, precipitation-dissolution reactions, and 20 mechanical filtering are other mechanisms that can also reduce the velocity of solutes during their transport in groundwater, which is only driven by advection and hydrodynamic dispersion. Substances that behave in a very similar way to groundwater are known as conservative tracers. Such substances (e.g. bromide, chloride, uranine, eosine, etc.) are used, not only in laboratory experiments but also in field tracer tests, to identify differences between the transport behavior of reactive substances and the groundwater movement . 25 The effect that different factors have on solute transport in groundwater is shown as a schematic breakthrough curve in Fig. 1. A conservative tracer or substance does not react with the soil and/or groundwater, nor does it undergo biological or radioactive decay (Fetter, 1988). It is only influenced by advection and hydrodynamic dispersion (the red breakthrough curve in Fig. 1).
In contrast, a non-conservative tracer or substance is likely to react with the soil and/or groundwater and its transport will be influenced by many (or all) of the controlling factors for solute transport described below . 30 Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c Author(s) 2016. CC-BY 3.0 License.
can dissociate. The charge of such molecules can therefore change from positive (cationic) to neutral or negative (anionic) depending on the degree of dissociation. It is even possible for molecules to bear differently charged (positive and negative) functional groups at the same time and to then behave as both anions and cations (zwitter ions).
The species of dissociative organic compounds in groundwater is controlled by the dissociation constant of organic acids, Ka, and (if present) organic bases, Kb (Schaffer and Licha, 2014). The relationship between the inverse common logarithms of the 5 acidic/basic dissociation constants pKa and pKb (i.e. the inverse common logarithm of the concentration of dissociated protons) and the pH defines the charge state of the compound. Where the pH << pKa (which is generally the case for pKa > 10 at normal groundwater pH values of between 6 and 8) the substance will exist in its undissociated (neutral) form, and where the pH >> pKa (which is generally the case for pKa values between 0 and 3 at normal groundwater pH values) the substance will exist in its dissociated (anionic/acidic) form. The relationship between pKb and pH is the opposite of that between pKa and pH: where 10 the pH << pKb the substance will exist in its dissociated (cationic/basic) form and where the pH >> pKb it will exist in its undissociated (neutral) form. Substances with pKa or pKb values close to the pH of the solvent will exist in both non-polar and polar forms.
The pH of groundwater therefore defines the polar character of organic compounds and consequently has an important influence on the sorption behavior of the compounds (Schwarzenbach et al., 2003). 15 The distribution coefficient (K OW or P) between water and octanol is used to describe the hydrophilic or hydrophobic character of a non-ionizable or neutral organic compound. The distribution coefficient for ionizable or non-neutral compounds is D, which is pH-dependent.
A compound with a distribution coefficient of log K OW (= log P or log D) > 0 is more lipophilic (i.e. migrates into the organic 20 phase), and < 0 is more hydrophilic (i.e. migrates into the aqueous phase). Predicted values for P and D at specific pH values can be obtained from databases (e.g. Chemicalize.org; SciFinder) or from chemical drawing programs (e.g. ChemDraw; MarvinSketch; Smyx Draw).
In order to obtain the distribution coefficient between a solid and liquid phase (K d ) the K OC must be corrected because log K OC describes the partitioning between water and a 100% organic carbon phase: 25 with f OC as the fraction of organic carbon (Appelo and Postma, 2005).
If the distribution coefficient between a solid and liquid phase (K d ) of a specific compound is known, together with the bulk density ρ and the porosity Θ of the substrate, the retardation factor of this compound can be approximated as follows (Stumm and Morgan, 1996): 30 Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c Author(s) 2016. CC-BY 3.0 License.

Degradation
Degradation is the mineralization of complex molecules to form inorganic molecules or elements. A reactive substance can undergo chemical, biological, or radioactive degradation during transport, which will reduce both the concentration and the total content of that particular substance (Fetter, 1988 and Fig. 1), and hence lower the breakthrough curve (green curve in Fig.   1). Radioactive decay is not relevant for organic compounds but both microbial degradation and chemical degradation (e.g. 5 hydrolysis, oxidation-reduction reactions, UV degradation) are potentially significant. Redox reactions are electron transfer reactions that induce metabolism of organic molecules. This can lead to total mineralization and to the formation of metabolites with properties that may differ from those of the original molecules. In most cases these processes are catalyzed by microbes (Schwarzenbach et al., 2003). Redox reactions during groundwater transport as a result of changes in thermodynamic conditions along the flow path can also affect the solubility of a compound. Redox processes control the natural concentrations 10 of O_2, Fe^(2+)/Fe^(3+), SO_4^(2-), H_2 S, CH_4, and other compounds in groundwater (Appelo and Postma, 2005), and can therefore induce zones with different oxygen concentrations, zones of nitrate-reduction, iron-reduction, sulfate-reduction, or methanogenesis, and zones with different pH values. This chemical zoning may then produce precipitation-dissolution reactions in the transported compounds, which can result in mass loss for a particular substance during transport and therefore appear as degradation. 15

Transport behavior of organic micropollutants
Organic compounds include a very wide range of the compound-specific properties described above. Different species and metabolites of a compound can exist in the same sample. Transport behavior is furthermore dependent on the pH of the fluid and on the properties of the substrate (e.g. the organic carbon content, the presence of sorption sites on clay minerals, or the presence of Fe/Mg oxides or hydroxides). It is therefore not possible to describe the general behavior of individual compounds 20 or groups of compounds in groundwater. It is even difficult to predict the behavior of particular compounds in specific settings because of the large number of defining variables. The future objective should therefore be to progress from individual case studies to a general process understanding, with the ultimate aim being to be able to make general predictions concerning the behavior and eventual fate of organic compounds in natural aqueous environments.

Setup and evaluation of laboratory column experiments
Laboratory column experiments can be used for many different applications. The boundary conditions and experimental setup can be varied to best address particular research questions or compounds. Column experiments are generally used to investigate the transport behavior, sorption, and degradation of a specific compound, or group of substances. However, more specific issues can also be addressed such as the effect of entrapped air on soil permeability (Christiansen, 1944), the effect of a 30 Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 (Sinke et al., 1998), the effect of preferential flow on solute transport (Schoen et al., 1999), the effect of a sterilized soil on sorption (Lotrario et al., 1995), or the influence of methanol on the retention of hydrophobic organic chemicals (Nay et al., 1999). An overview of the types of investigations that can be carried out with column experiments is provided below, the methods involved presented, and practical issues discussed. Additional information on column experiments for specific compounds is provided in Section 3.2. 5

Technical setup
The basic principle of a column experiment is to pump water with a specific composition, including solutes of interest, through a column filled with a specific substrate. Since the aim of column experiments is to investigate (dynamic) processes it is common to start with a "neutral" fluid that is in hydrochemical equilibrium with the substrate, and then switch to the actual test fluid containing the solutes of interest and a mandatory conservative tracer. Figure 2 shows three basic setups for laboratory 10 column experiments. The normal practice is to operate columns in an upright position. This allows percolation through the unsaturated zone to be simulated, as shown in Fig. 2(a). Different degrees of saturation can thus be achieved within the column and the effects of a fluctuating water table can also be simulated. In order to simulate saturated groundwater conditions the column can be operated with an upward flow direction, as shown in Fig. 2(b). This results in saturated conditions throughout the column and may help to avoid any entrapment of gas bubbles. Different parameters or boundary conditions and their 15 evolution along a flow path can be investigated by coupling several columns together, as shown in Fig. 2(c).
Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci.   Scheytt et al., 2004), (c) saturated columns connected in series in order to measure the evolution of hydrochemical parameters (such as the oxidation-reduction potential) along the flow path (e.g. Nay et al., 1999). 5 Columns are typically made of stainless steel (Alotaibi et al., 2015;Banzhaf et al., 2012;Burke et al., 2013;Unold et al., 2009;Xu et al., 2010) in order to prevent interactions with solutes, or from acrylic glass (Gruenheid et al., 2008;Hebig et al., submitted;Rauch-Williams et al., 2010;Yao et al., 2012) for improved visual control of saturation levels and tracer transport (where a dye tracer is used). Although acrylic glass is a synthetic plastic it is reported to be inert to many 10 Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c Author(s) 2016. CC-BY 3.0 License. organic solutes (Hebig et al., 2014) and therefore suitable for column experiments with organic micropollutants. Glass columns are also often used (e.g. Estrella et al., 1993;Fan et al., 2011;Nay et al., 1999;Persson et al., 2008;Simon et al., 2000) since glass is assumed to be inert with respect to organic compounds. Other materials used for columns include PVC (e.g. Bertelkamp et al., 2012;Greenhagen et al., 2014;Salvia et al., 2014;Sinke et al., 1998), polyethylene (Bertelkamp et al., 2012, and aluminum (Burke et al., 2014). 5 The physical dimensions of columns are not yet standardized. Short columns enable fast experiments and many repetitions, while longer columns need more time for equilibration before the actual start of the experiment but allow longer reaction times (which means that more reactions can be distinguished) and more complex settings (e.g. the establishment of redox zonings).
The dimensions of columns vary considerably due to the broad range of research applications; for example, lengths have been reported ranging from 5 cm (Estrella et al., 1993;Teijón et al., 2014) to 2.4 m (Cordy et al., 2004, and inner diameters have 10 been reported ranging from 2 cm (Teijón et al., 2014) to 36 cm (Bertelkamp et al., 2012;Bertelkamp et al., 2014). The choice of a reasonable length-to-diameter ratio is important if scaling effects are to be avoided. In long columns with small diameters the main flow (and hence the main transport) can occur by preferential flow along the boundary between the inner column surface and the sediment grains. If the diameter is too large transversal dispersivity can become significant for solute transport, which makes analysis even more complex (column experiments are generally designed to exclude transversal dispersivity from 15 the transport model). Using a short column with too large a diameter relative to its length may prevent uniform, homogeneous flow within the sediment. Lewis and Sjöstrom (2010), in their extensive review of the design of column experiments, recommended a diameter-to-length ratio of 1:4 in order to avoid such effects. However, this ratio is rarely reported in published literature.
The tubing and other materials reported are commonly made from Teflon/PTFE (e.g. De Wilde et al., 2009;Fan et al., 2011;20 Ke et al., 2012;Strauss et al., 2011;Teijón et al., 2014) or stainless steel (e.g. Ke et al., 2012;Nay et al., 1999;Teijón et al., 2014), which are known to be inert materials. Other materials that have been used in column experiments are Pharmed tubing (Strauss et al., 2011), (dark) polyethylene (Bertelkamp et al., 2012;Bertelkamp et al., 2014), PVC, (black) Tygon tubing, vinyl, polysulfone, brass (Greenhagen et al., 2014), polypropylene, and silicone (Srivastava et al., 2009). However, the influence of many of these reported materials on organic compounds is unknown and may be problematic, as is the case for Pharmed tubing, 25 silicone, and Tygon tubing (Hebig et al., 2014). Unfortunately, there is often little information provided concerning the materials used (in particular the filter materials used), even though they can have a significant influence on the mass recoveries of organic micropollutants. Only a few investigations have specifically addressed the issue of possible interactions between the materials used in experiments and the compounds and fluids under investigation, or included preliminary investigations to allow such interactions to be avoided (Greenhagen et al., 2014;Gruenheid et al., 2008;Hebig et al., 2014;Srivastava et al., 30 2009). This often untested impact of laboratory materials on compound concentrations may therefore have a significant (but unknown) influence on the results of column experiments.
To prevent inhomogeneous flow and mass transport through the column a layer of clean, well sorted filter quartz sand is often 5 included at both the inlet and the outlet ends of the column (e.g. Banzhaf et al., 2012;Gruenheid et al., 2008;Persson et al., 2008;Unold et al., 2009). Other materials reported to have been used as filters are glass beads or globes (e.g. Salvia et al., 2014;Scheytt et al., 2004;Scheytt et al., 2006), and glass wool (Nay et al., 1999;Salem Attia et al., 2013).  (Lorphensri et al., 2007), Teflon gauze nets (Scheytt et al., 2004, personal communication), porous glass (Siemens et al., 2010), porous ceramic plates (Unold et al., 2009), or paper filters (Srivastava et al., 2009). In this way the hydraulic contrast between the tube/inlet and the substrate can be reduced and the incoming water and solutes spread over the entire width of the column. Any intrusion or wash-out of finer sand particles through the inlet or outlet of the column should 15 also be avoided through the use of such filters.
A wide range of substrates have been used as filling for columns in published research, depending on the specific objectives.
These include natural (site-specific) aquifer sediment (e.g. Alotaibi et al., 2015;Burke et al., 2013;Lopez-Blanco et al., 2005;Mersmann et al., 2002;Preuss et al., 2001;Teijón et al., 2014), natural soil (e.g. Aga et al., 2003;Cordy et al., 2004;Kamra et al., 2001;Murillo-Torres et al., 2012;Rodriguez-Cruz et al., 2007;Xu et al., 2010), artificial soil (De Wilde et al., 2009), 20 artificial (model) sediments such as well sorted filter sands or technical quartz sands (e.g. Baumgarten et al., 2011;Bertelkamp et al., 2014;Greenhagen et al., 2014;Nay et al., 1999), and other artificial materials such as iron coated sand (Hebig et al., submitted), magnetic nanoparticle-coated zeolite (Salem Attia et al., 2013), alumina, silica gel (Lorphensri et al., 2007), and biochar (Yao et al., 2012). The substrate can be installed wet or saturated (e.g. Alotaibi et al., 2015;Nay et al., 1999;Simon et al., 2000), or dry (e.g. Banzhaf et al., 2012;Fan et al., 2011;Scheytt et al., 2007;Sinke et al., 1998;Teijón et al., 2014), ideally 25 in small (1-2 cm) layers that are individually compacted using a tool such as a stamp (Banzhaf et al., 2012), plunger (Scheytt et al., 2007)  , of which only the lower limit is representative of effective porosities found in natural aquifers. It appears that effective porosities lower than 0.30 are only achieved when the column is filled with undisturbed sediments. The high effective porosities in most experiments may lead to lower flow velocities and Hydrol. Earth Syst. Sci. Discuss., doi: 10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c Author(s) 2016. CC-BY 3.0 License. lower reactive surface areas than would be expected in a natural environment and may therefore be responsible for the often noted differences between laboratory results and results from field tests. However, many published investigations either do not specify the effective porosities or only report the total porosities (from the ratio of the weight of the column filled with dry sediment to the weight of the column with fully saturated sediment). Pore water velocities vary according to the experimental conditions. The flow velocity should ideally reproduce natural groundwater flow velocities, which one would normally expect 5 to be between 1 cm d -1 and 1 m d -1 . Using higher velocities allows experiments to be completed more quickly and hence many repetitions, but slow velocities are more likely to provide a realistic representation of natural processes, involving equilibration of solute and solid phases. Flow velocities in columns experiments are typically between 4 cm d -1 (Strauss et al., 2011) and 348 cm d -1 (Fan et al., 2011). There are similar uncertainties in reported flow velocities to those previously mentioned for porosities, as authors do not always state clearly whether they are referring to pore water velocity or Darcy velocity. Various 10 types of fluid have been used as solvents in column experiments depending on their objectives, including natural (site-specific) groundwater (Greenhagen et al., 2014;Mersmann et al., 2002;Preuss et al., 2001;Scheytt et al., 2004) or surface water (e.g. Banzhaf et al., 2012;Baumgarten et al., 2011;Burke et al., 2014;Schaffer et al., 2012), model or artificial (ground)water (De Wilde et al., 2009;Estrella et al., 1993;Murillo-Torres et al., 2012;Xu et al., 2010), treated wastewater (e.g. Alotaibi et al., 2015;Cordy et al., 2004;Ke et al., 2012), tap water (Bertelkamp et al., 2012;Burke et al., 2013), distilled water (Aga et al., 15 2003), ultrapure water (Simon et al., 2000), and even liquid manure (Strauss et al., 2011).
The reported solute concentrations vary considerably depending on the objectives of the individual experiments, ranging between 60 ng L -1 (Rauch-Williams et al., 2010) and 2 mg L -1 (Siemens et al., 2010;Yao et al., 2012). Some investigations therefore use concentrations that are found in natural environments or in waste waters, while others may aim to identify specific processes and therefore use higher concentrations to enhance the effect of the processes. Higher concentrations are also easier 20 to measure.
An important basic parameter is the pH of the fluid used in column experiments as the polar character of many organic micropollutants varies according to the relationship between the pKa value and the pH. An organic compound may therefore be persistent and highly mobile under the specific conditions of one experiment but be strongly retarded in another experiment, due to the different pH values of the fluid used. 25 Field parameters and/or tracers are commonly measured using flowthrough cells fitted with probes (e.g. Mersmann et al., 2002;Müller et al., 2013). Sampling for solutes can be performed in a number of different ways including "by hand" (bottle, beaker, e.g. Burke et al., 2014;Cordy et al., 2004;Hebig et al., submitted), using an automated fraction collector (e.g. Banzhaf et al., 2012;Rodriguez-Cruz et al., 2007;Scheytt et al., 2004;Srivastava et al., 2009;Unold et al., 2009), using sampling ports attached laterally to the column (Alotaibi et al., 2015;Baumgarten et al., 2011;Burke et al., 2014), or online and in real-time 30 using a spectrometer at the outlet from the column (Teijón et al., 2014). Sampling ports alongside the column risk altering the hydraulics within the column and should only be used with great caution, using small volumes and low sampling rates. A further means of assessing the fate of organic compounds in a column experiment is to extract the substrate from the column after completing the experiment and to analyze the solid phase for irreversibly sorbed solutes (Banzhaf et al., 2012). This Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c Author(s) 2016. CC-BY 3.0 License. enables all parts of the mass balance to be determined since the mass of solute degraded can be determined by deducting the recovered mass and the irreversibly sorbed mass from the injected mass.

Significance and influence of boundary conditions and heterogeneity
The boundary conditions of column experiments need to be known in order to ensure correct interpretation of the results obtained. Van Genuchten and Parker (1984) discussed the physical and mathematical significance of the boundary conditions 5 that apply to one-dimensional solute transport in laboratory column experiments. They presented solutions of the convectivedispersive transport equation that can be used to analyze column effluent data. Leij et al. (1993) investigated analytical solutions for non-equilibrium solute transport in 3-dimensional porous media. They found that the effect of non-equilibrium on 1D transport is similar to that on 3D transport. The effect of non-uniform boundary conditions on steady flow in saturated homogeneous cylindrical soil columns was investigated by Barry (2009). He reported that uniform flow in the column could 10 be achieved if a baffle zone was established at each end of the column. Sentenac et al. (2001) used fiber-optic sensing to measure side-wall boundary effects in soil columns. They detected flow velocity differences between the center of the column and the boundary wall, which surface roughness was varied. Seyfried and Rao (1987) described preferential flow effects in columns containing undisturbed substrates. They assumed that flow occurred through series of large pores or pore sequences.  (1989) found that the dispersivity of chloride and tritium tracers used in unsaturated column experiments increased 20 significantly with column length, while retardation factors remained essentially the same. However, Ribeiro et al. (2011) showed that the retardation factor for potassium ions investigated during leaching experiments decreased with increasing column length, and that the dispersive-diffusive coefficient and dispersivity both increased with increasing pore-water velocity and increasing soil column length. Rühle et al. (2013) described changes over time to the water flow through the porous medium in a column, from uniform to non-uniform. They concluded that flow path changes occurred due to clogging of small 25 pores near the column inlet as a result of microbial growth and calcite precipitation, which then caused non-uniform water flow and solute transport. Bromly et al. (2007) reviewed published data on experiments on almost 300 repacked saturated homogeneous column experiments. They related the dispersivity to the length of the column used, i.e. short columns had greater dispersivities. However, clay content was identified as the most important factor controlling dispersivity, followed by the diameter of the column. However, they pointed out that the individual experimental design (e.g. the column geometry, inlet 30 dead volume, and soil packing) needs to be taken into account in order to be able to relate dispersivity to soil properties.
hydraulic conductivity of a sandy soil with a low water content following compaction by various methods and found that conductivities varied by almost three orders of magnitude depending on the compaction.

Evaluation of column experiments
It should be pointed out that the main issues when evaluating and interpreting results from column experiments are the same for different organic compounds, these being the interactions with surfaces and the ionic or neutral character of the molecules. 5 In order to describe the transport behavior of organic solutes in column experiments their (reactive) breakthrough curves are often compared with the breakthrough curves of a conservative tracer (which represent the flow velocity of the fluid). This is sometimes performed by graphical analysis (e.g. Scheytt et al., 2004). In both reactive and conservative breakthrough curves the instant is identified at which the observed concentration reaches 50% of the injected concentration (i.e. c/c0 = 0.5). The temporal offset between those two points in time represents the retardation of the reactive compound compared to the tracer 10 or fluid. In this approach the mass recovery (as a measure of the degradation) is calculated from the area below the breakthrough curve and the pumped volume.
Time (or temporal) moment analysis has also frequently been used to evaluate column experiments (Kamra et al., 2001;Murillo-Torres et al., 2012). This method allows the mean breakthrough time, spreading, and asymmetry of a breakthrough curve to be characterized by integration of the breakthrough curve (Appelo and Postma, 2005;Hebig et al., 2015). It is, 15 however, more common to model (or "fit") the solute concentrations to the convective-dispersive equation (CDE): in which the term on the left-hand side describes the retardation of a solute, and the terms on the right-hand side describe the dispersion, the average pore-water velocity, and the first-order decay (i.e. degradation) of the solute, (described in detail in, e.g. Parker and Vangenuchten, 1984;Toride et al., 1999). Since the retardation of a conservative tracer is defined as 1 (i.e. no 20 retardation compared to the pore-water velocity) and both the pore-water velocity and the dispersion are aquifer-specific parameters, the breakthrough curve of a conservative tracer can be used to iteratively vary both parameters until the "fit" matches the observed concentrations sufficiently well. Knowing the dispersion and velocity in the experiment makes it possible to reduce the number of variables for the reactive compounds and to then repeat the fitting procedure in order to estimate the retardation and degradation. 25 Computer software commonly used to fit the CDE to observed breakthrough curves is the CXTFIT software (Toride et al., 1999) which is available in various forms, for example through the public domain STANMOD software (Šimůnek et al., 1999),

Column experiments on different groups of organic micropollutants
Column experiments on different groups of organic micropollutants are described in greater detail below, distinguishing between investigations into pharmaceuticals, pesticides, and other organic micropollutants. Selected column experiments for 5 non-organic compounds are also presented for comparison. The presented examples of column experiment are then summarized and discussed.

Pharmaceuticals
Column experiments are often used to investigate the transport of pharmaceutical compounds under both saturated and unsaturated conditions. These are presented separately in this section, beginning with experiments under saturated conditions. 10 Mersmann et al. (2002) investigated the transport of carbamazepine, clofibric acid, diclofenac, ibuprofen, and propyphenazone under saturated conditions and found the transport of all of these compounds to be unaffected by changes in the pH, temperature, dissolved oxygen content, or ion concentration of the water used. Moreover, they found that each of these compounds except for clofibric acid was retarded in the column. In contrast, Gruenheid et al. (2008) found that temperature affected the degradation of sulfamethoxazole under saturated conditions, with higher temperatures resulting in increased 15 mineralization. However, the biodegradation of iopromide was high at all investigated temperatures (5,15, and 25°C) while the naphthalenedisulfonic acids showed no biodegradation at any of these temperatures. Müller et al. (2013) investigated the influence of different total organic carbon contents and the resulting changes in redox conditions on the transport of primidone, carbamazepine, and sulfamethoxazole. Carbamazepine and primidone were found to be retarded in the presence of organic matter while sulfamethoxazole remained unaffected. However, none of the three compounds was degraded under any of the 20 investigated conditions. Banzhaf et al. (2012) also carried out saturated column experiments under varying redox conditions. They found that the degradation of sulfamethoxazole was clearly redox dependent, exhibiting strong degradation associated with denitrification. Both retardation and degradation were observed for carbamazepine. Retardation in the column was observed for diclofenac but not for ibuprofen. Scheytt et al. (2004) performed experiments under saturated conditions and found neither degradation nor retardation for clofibric acid, whereas diclofenac and propyphenazone were both retarded. 25 Patterson et al. (2010) observed no rapid degradation of carbamazepine and oxazepam under saturated experimental conditions. Teijón et al. (2014) investigated naproxen and found that the sorption of this compound was independent of the flow rate during the experiment; they noted a generally low sorption affinity for naproxen. Bertelkamp et al. (2014) investigated sorption and biodegradation for ibuprofen, ketoprofen, gemfibrozil, acetaminophen, trimethoprim, propranolol, metoprolol, carbamazepine, and phenytoin under oxic conditions, in order to simulate bank filtration. They found that the biodegradation 30 of these compounds could be predicted on the basis of their functional groups. Moreover, they observed no retardation for any of the investigated compounds. Hebig et al. (submitted) showed a strong correlation between retardation and organic carbon Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. bezafibrate, diclofenac, gemfibrozil, clarithromycin, trimethoprim, clindamycin, erythromycin, and metoprolol; they found that the investigated clay soil had significant potential to retain these pharmaceuticals. The retention capacity was, however, 10 limited and all compounds were leached to some extent. Wu et al. (2010) found low mobility for carbamazepine, diphenhydramine, fluoxetine, diltiazem, and clindamycin, and also for two metabolites (carbamazepine-10,11-epoxide and norfluoxetine), in unsaturated leaching experiments. Moreover, carbamazepine, diphenhydramine, and fluoxetine were persistent throughout the experiment. Leaching experiments carried out by Salvia et al. (2014) indicated that the transfer and degradation of the investigated pharmaceuticals, these being sulfonamides (sulfanilamide, sulfadiazine, sulfathiazole, 15 sulfameter, sulfadimidine, sulfabenzamide, sulfadimethoxine, and sulfamethoxazole), macrolides (erythromycin, tylosin, roxithromycin), trimethoprim, dicyclanil, penicillin G, carbamazepine, fluvoxamine, and paracetamol, were dependent on the soil characteristics, i.e. on the amount of clay in the soil and its pH. All of the investigated compounds were found to degrade both substantially and rapidly except for roxithromycin and carbamazepine, which were relatively persistent. Kay et al. (2005) investigated the leaching of oxytetracycline, sulfachloropyridazine, and tylosin from clay soils after slurry application. 20 Although the pH was significantly affected by the slurry this had no effect on oxytetracycline leaching. Scheytt et al. (2006) found that diclofenac, ibuprofen, and propyphenazone showed similar mobilities in both saturated and unsaturated column experiments, but carbamazepine showed lower sorption and elimination under unsaturated conditions than under saturated conditions.

Pesticides 25
In contrast to column experiments on pharmaceuticals, most column experiments on pesticide leaching have been carried out under unsaturated conditions in order to reflect the main input path of pesticides into groundwater, which is through agricultural use. Nkedi-Kizza et al. (1987) carried out leaching experiments on atrazine and diuron using various mixtures of water and methanol. They found a significant reduction in the retardation factor as the volumetric fraction of the organic cosolvent 30 methanol increased. Persson et al. (2008) reported leaching of 30% of the investigated chlorophenols from contaminated soils, of which 1-3% was associated with colloids. Increasing the pore water velocity had no influence on their mobility. Lopez-Blanco et al. (2005) investigated the transport of endosulfan under unsaturated conditions. They found that high soil moisture Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci.  Persson et al. (2008) investigated the leaching of polychlorinated diphenyl ethers (PCDEs), polychlorinated dibenzofurans (PCDFs), and polychlorinated dibenzo-p-dioxins (PCDDs) from contaminated soil. Less than 0.2% of PCDEs, PCDFs and PCDDs were mobilized during the experiments and the compounds were found to be preferably associated with the particulate fraction of the leachate. Sinke et al. (1998) investigated the degradation of 4-nitrobenzonate and toluene within a column under a fluctuating water table and reported that both compounds were degraded. They also found that microbial processes induced chemical and physical heterogeneity in the column and that the fluctuating water table introduced additional heterogeneity. Patterson et al. (2010) conducted column experiments on a number of organic micropollutants under saturated conditions. 15

Other organic compounds
They found rapid degradation for bisphenol A, 17b-estradiol, 17a-ethynylestradiol, and iodipamide, but only relatively gradual degradation for nitrosodimethylamine, N-nitrosomorpholine and iohexol. Alotaibi et al. (2015) investigated the transport of benzotriazole and 5-methylbenzotriazole under anaerobic saturated conditions and observed biodegradation in both compounds. Liu et al. (2008) simulated the use of bacteria for in situ remediation of biodiesel contamination, using a saturated column experiment and were able to demonstrate successful degradation of the biodiesel. 20

Non-organic compounds
Although not the primary subject of this review, column experiments are also used to investigate non-organic compounds and selected investigations are presented in this section. Smith et al. (1985) investigated the transport of Escherichia coli through both disturbed and undisturbed soil columns. They found mixed and repacked soils to be much more effective in filtering the bacteria than undisturbed soils, which allowed up to 96% of the Escherichia coli to pass through the columns. Jin et al. (2000)  25 investigated virus removal and transport in both saturated and unsaturated sand columns and found significantly higher removal under unsaturated flow conditions. Pang et al. (2002) investigated the effect of pore-water velocity on the transport of Cd, Zn, and Pb under non-equilibrium chemical conditions in alluvial gravel columns; they found the proportion of exchange sites available to be independent of the pore-water velocity. Amos et al. (2004) investigated the remediation of acidic mine drainage using column experiments and found Fe removal during long term operation of column experiments to be a good indicator of 30 the column's ability to remediate acidic mine drainage. Ilg et al. (2007) investigated colloid transport in unsaturated soil columns. They found that colloid transport could be overestimated, depending on the sampling system used.

Comparison of column experiments with other available methods
As described and discussed in Sections 3.1 and 3.2, laboratory column experiments are suitable (and widely used) for investigations into the fate of organic micropollutants. There are, however, alternative methods available that can also be used for this purpose, depending on the objectives and the facilities available:  Incubation experiments (such as batch experiments, microcosms, etc.) used, for example, to determine sorption 5 coefficients (Scheytt et al., 2005) or to investigate the persistence of pharmaceuticals (Lam et al., 2004), the elimination of pharmaceuticals (Radke and Maier, 2014), the microbial degradation potential of pollutants (Barra Caracciolo et al., 2013), or the biotransformation of micro-contaminants (Nödler et al., 2014).
 Reactive field tracer tests, e.g. on organic micropollutants and pharmaceuticals (Hillebrand et al., 2015;Kunkel and Radke, 2011;Riml et al., 2013). 10 By far the largest number of comparisons have been between batch experiments and column experiments on organic micropollutants and we therefore focus on comparisons between these two laboratory methods. However, selected alternative methods are also discussed briefly in this section. Maeng et al. (2011) (estrone, 17β-estradiol, estriol, 17α-ethynylestradiol, 4-tert-octylphenol, and bisphenol A) and two pharmaceuticals (ibuprofen and naproxen).
Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c Author(s) 2016. CC-BY 3.0 License. Benker et al. (1998) compared the retardation coefficients for trichloroethene estimated from batch and column experiments with field data. They were able to confirm the sorption behavior observed in column experiments from the field data and, provided the sorptive properties of the sediment were correctly determined, batch experiments then allowed the retardation of trichloroethene in the field to be reliably predicted. Scheytt et al. (2007) compared the results obtained from unsaturated column experiments on clofibric acid, diclofenac, ibuprofen, and propyphenazone to field measurements from former sewage farms. The equilibrium established within a column may differ from that in the inflowing water during the experiment. Batch experiments are often carried out using unrealistic sediment-to-water ratios (e.g. 1:5) that do not reflect realistic aquifer conditions. While the theoretical maximum sorption capacity (under ideal conditions and equilibrium) can be reasonable well 20 determined from batch experiments, neither advection nor the (dynamic) sorption-desorption behavior (for example) can be determined with this method. Batch experiments are therefore an established method and are suitable for determining equilibrium parameters for interactions between a specific organic and a specific sediment, but they are less able to reproduce the dynamic (i.e. non-equilibrium) groundwater conditions of an aquifer than column experiments. Batch sorption experiments are therefore suitable for determining the sorption behavior of specific compound-sediment combinations under equilibrium 25 conditions, while column experiments are more suitable for determining the transport behavior of specific compound-fluidsediment combinations, under either equilibrium or non-equilibrium conditions.

Summary and discussion
The practical objectives of column experiments on organic micropollutants relate to possibilities for their removal, either by natural processes during passage through soil and groundwater, or by technical processes such as those used in WWTPs. They 30 can also improve our understanding of the relationship between the properties of specific compounds and the properties of fluids and aquifers, with the objective of using organic micropollutants as indicators of aquifer conditions and groundwater history. Different boundary conditions (such as the redox conditions and the degree of saturation) clearly have a strong Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c Author(s) 2016. CC-BY 3.0 License.
influence on the transport and degradation of organic micropollutants, and are therefore critical to defining the transport behavior of a specific organic compound in a way that is applicable to any given hydrogeochemical environment.
Unfortunately the results of most column experiments therefore remain restricted to the specific boundary conditions of each column experiment, since variations in just one of these boundary conditions (e.g. the redox conditions) can have a major impact on the results. For example, the degradation of a compound such as sulfamethoxazole can range from rather low levels 5 (e.g. Suarez et al., 2010) to high levels (e.g. Banzhaf et al., 2012), depending mainly on the redox conditions. The degradation of sulfamethoxazole has been shown to be strongly dependent on the organic carbon content of the sediment (Hebig et al., submitted). The behavior of such compounds can also depend on the level of saturation. Diclofenac, for example, shows high levels of degradation under unsaturated conditions but very low levels of degradation under saturated conditions (Scheytt et al., 2004;Scheytt et al., 2007). 10 Another problem faced in designing technical processes to remove organic compounds is that different organic compounds can be sensitive to different hydrochemical conditions, which makes it difficult to find conditions that will allow all such compounds to be removed at the same time. However, column experiments can assist in finding such conditions as it is rather easy with this experimental setup to vary the boundary conditions. Batch sorption experiments are suitable if only the equilibrium sorption behavior of a specific compound-sediment combination is to be determined. It should be noted that 15 column experiments always remain limited in their transferability to real world conditions because of experimental restrictions (such as those imposed by limitations of scale), which mean that processes that might occur simultaneously in nature cannot be fully reproduced in a laboratory. Column experiments therefore sometimes represent field conditions quite well and sometimes do not. However, the main objective of column experiments should not be specifically to achieve laboratory results that are transferable to real world conditions but to achieve an improved general understanding of the behavior of organic 20 compounds, i.e. how different boundary conditions affect the behavior of the investigated compounds in natural environments.

Conclusion
Laboratory column experiments are a valuable and appropriate method for investigating and characterizing the transport behavior of organic micropollutants. They have been widely used in recent decades for numerous investigations into a great variety of organic compounds. This has led to an enormous increase in our understanding of this ever growing group of 25 compounds. While the experimental method in general can now be considered a standard method, many different setups have been used which is a major issue when it comes to comparing results. A standardized setup for column experiments would yield results that are fully comparable and transferable between different column experiments. It is of course not surprising that such a standardized setup does not exist as the setup used invariably depends very much on the specific research question being investigated. Steps towards the standardization of column experiments could include following the suggestion by 30 Mackay and Seremet (2008) that model substances be used for investigations into cation exchange, and the use of reference soils to characterize different compounds as suggested by (Bi et al., 2006). These suggestions have to date only been applied Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-196, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 2 May 2016 c Author(s) 2016. CC-BY 3.0 License.
to batch sorption experiments and not to column experiments. It would however be of great benefit if the research community could agree on specific reference compounds and substrates to use in column experiments, in order to help overcome the issue of comparability. This would not only facilitate comparisons between different experiments but also eventually bring us closer to achieving a more universal understanding of the transport and eventual fate of organic micropollutants in groundwater.