A systematic comparison of pore water sampling methods in flooded soils: Element-specific biases in inorganic contaminants due to oxidation and filtration artefacts

Flooding of contaminated soils affects the mobility and bioavailability of inorganic contaminants. Accurately identifying these changes requires pore water sampling methods with minimal artefacts due to oxidation or exclusion of mobile colloids. This study was set up to identify such artefacts by comparison of the solution composition among different three pore water sampling methods as a function of the soil redox potential E_h, i.e., Diffusion Equilibrium of Thin Films (DET), rhizons (suction cups), and soil centrifugates, followed by filtration. The first two methods were applied to intact flooded soil, whereas the latter was applied to destructively sampled soil aliquots. The soil (pH 6.7, 8.9 % organic carbon), a candidate for wetland construction, was sampled in a natural grassland as intact 35 cm cores of unsaturated soil. The soil columns were flooded, and pore water was sampled at various depths (6–20 cm) and times after flooding (3–100 days), representing samples with E_h between +450 mV and −240 mV. The pore waters sampled by the rhizons confirmed well-known contrasting trends between elements that are mobilized (Fe, U) or immobilised (Cd, Zn) upon reduction. Pore water compositions were similar among the three methods for redox-insensitive elements, mainly occurring as free ions such as Na, K, Ca, Mg and Ba. In contrast, strong deviations were found for redox-sensitive elements. The sampling bias, expressed as the ratio of pore water concentrations in the centrifuged samples to those in the rhizon, drifted from close to 1 at three days of waterlogging (no bias) to 0.01 (Fe), 0.17 (U) and 0.08 in (P) after 100 days. Conversely, opposite trends were found with ratios of Cd (82), Zn (20) and SO₄²⁻ (18). All of this suggests oxidation artefacts during centrifugation. The DET oversampled Fe compared to rhizons due to oxidation and precipitation of Fe in the DET, which created an adsorptive sink, particularly for post-transition metals and P and, hence, oversampled several mobile elements in reduced conditions. The sampling bias (DET/rhizon) was factor 4 for Fe (geomean ratio) but peaked up to 390 for P and 1300 for Pb, which may be related to higher exclusion of colloids by the rhizons, that had the smallest pore size of the filters. Overall, rhizon samplers showed superior performance in capturing redox-induced changes, but may underestimate in-situ mobility of Pb. No single pore water sampling method can unequivocally detect the mobility changes for all elements or species in reduced conditions: a weight of evidence of different methods is generally recommended.