Beware of effects on isotopes of dissolved oxygen during storage of natural iron-rich water samples: A technical note

Rationale: Investigations of the isotope ratios of dissolved oxygen ( δ 18 O DO ) provide valuable information about the oxygen cycle in aquatic systems. However, oxidation of Fe(II) may change pristine δ 18 O DO values during storage and can lead to a misinterpretation. We sampled an Fe(II)-rich spring system and measured δ 18 O DO values at various time intervals in order to determine influences of Fe-oxidation. Methods: Water samples were collected from an Fe-rich spring and related stream and the δ 18 O DO values were measured in fresh, 4- and 13-day-old samples with an isotope ratio mass spectrometer. Three replicates were measured for each sample with a 1 σ of ± 0.2 ‰ . On-site parameters and Fe(II) contents were also measured over the course of the spring system by multi-parameter probes and spectrophotometry. Results: The δ 18 O DO values over the course of the spring system in fresh, 4- and 13-day-old samples revealed differences of up to 8 ‰ . We explain this increase by the consumption of DO by Fe(II)-oxidation. After a flow length of 85 m the differences in δ 18


| INTRODUCTION
Dissolved oxygen (DO) is one of the most important ecological parameters for the quality of water as it is essential for the survival of aerobic aquatic organisms. Additional measurements of its 18  On the other hand, addition of DO due to ongoing photosynthesis after sampling or its removal due to consumption by respiration is efficiently prevented by preserving samples with mercuric chloride (HgCl 2 ). This is often used as a standard technique to inhibit further biological activity. [4][5][6] However, little attention has so far been paid to the possibility of changes to δ 18 O DO values by oxidation of reduced dissolved metal ions between sampling and measurement. One of the most common forms of such reduced mineral phases is Fe(II). In order to close this knowledge gap, our aims were to find out: (1) how quickly δ 18 O DO values may change after sampling when a critical mass of reduced metals is present in solution and (2) how strong such effects might be. For this purpose, we chose the Fe(II)-rich Espan spring and stream system in the city of Fürth, Germany ( Figure 1). This is an ideal environment to test these questions, because it showed dissolved Fe(II) concentrations between 0 and 6.6 mg L −1 and DO concentrations between 2.3 and 11.0 mg L −1 . This study is also timely because to date only a few investigations exist about effects of metal oxidation on δ 18 O DO values, and most of them were carried out in either acidic or suboxic environments. [7][8][9] Also, their pH values were not specifically mentioned, 10 or they focused on experiments under controlled laboratory conditions. 11,12 In particular, the latter two studies showed that closed system iron oxidation leads to a consumption of dissolved oxygen and an associated increase in

| Sampling procedures
In a field campaign in February 2020 water was collected at 14 sampling points along the stream and divided into 3 sets of samples. One set was measured within less than 3 h, one was stored in the dark at 4 C for 4 days before measurement, and one was stored under the same conditions for 13 days before measurement.
Samples that had experienced any longer time periods between sampling and measurement were not tested because we assumed that most of the geochemical alterations would occur in the first 2 weeks after sampling. We were also not able to test any shorter time periods and, due to field and transport logistics, isotope analyses within 3 h were the best possible option.
Samples for δ 18 O DO measurements were collected in 12-mL Exetainers® (Labco Ltd, Ceredigion, UK) that were prepared with 10 μL of a saturated HgCl 2 solution to prevent biological activity after sampling. The Exetainers were filled with water that was syringefiltered using 0.45 μm pore size nylon filters until they were entirely full and free of air bubbles. They were then carefully closed with screw caps with a butyl septum in order to avoid atmospheric contamination. Test series in the field and in the laboratory showed that the degree of atmospheric contamination during this filling procedure was negligible. 19 Onsite parameters (pH, temperature, and DO) were measured with a HACH HQ40D multimeter i (Hach Lange GmbH, Düsseldorf, Germany).
The iron content was measured with an iron(II/III) cuvette test set (HACH) in combination with a portable HACH spectrophotometer (model DR 2800). These samples were also filtered with 0.45 μm pore size nylon filters to minimize iron precipitation and turbidity.

| Laboratory methods
The stable isotope ratios of DO were measured on a Delta Advantage isotope ratio mass spectrometer that was linked to a Gasbench II autosampler and an extraction unit (all from Thermo Fisher Scientific).
The procedure was modified from a method by Barth  and major ion measurements was alway better than 1%.

| Calculations of saturation states
The saturation states were calculated as a function of pH, pE, ion concentrations as well as the alkalinity and temperature with the programme PhreeqC (version 3). 22

| RESULTS AND DISCUSSION
Our data showed that:   Downstream of this sampling point no differences between fresh and stored samples could be found ( Figure 2C). With the obvious Fe-oxide precipitations the oxygen consumption must have resulted to a large part from this process. Calculations with PhreeqC also showed that Fe-oxides can form directly in the stream and also in the vials after sampling (see Tables S1 and S2,  In the following we discuss specific differences between the sampling points over time in more detail.

| Differences between fresh and 4-day-old samples
Compared with sampling points E1b to E3, E1a showed a relatively small increase in δ 18 O DO values between fresh and 4-day-old samples.
This indicates that oxygen was consumed through Fe-oxidation at sampling points E1b to E3. One possible reason for this observation could be a higher pH value of 6.5 (Table 2) at these sampling points.
Under neutral to circumneutral conditions, Fe-oxidation occurs more rapidly, while acidic pH values can slow down Fe(II)-oxidation.
Specifically, the initial pH of 6.  We also used a Rayleigh model to calculate how much oxygen would still be remaining in the vials after 4 and 13 days. This approach outlines the change in isotope ratios in a diminishing reservoir with known fractionation factors. 25  We therefore assume that these values are close to the real situation in the system.

| Differences between 4-and 13-day-old samples
If samples have to be stored for longer time periods before measurement one can also consider stabilizing them by acidification as Fe(II)-oxidation is known to considerably slow down under acidic conditions. 26 We have undertaken first attempts in this direction. However, other metal oxidations such as those of manganese or aluminium may also have to be taken into account. So far, first attempts to preserve water against metal oxidation after sampling by acidification were not successful. We therefore recommend rapid analyses after sampling as the best approach for iron-rich solutions.
As an alternative we recommend extraction of the DO into a headspace and its transfer into helium-flushed vials directly on site.
The latter would offer rapid isolation of the extracted O 2 and would