An assessment of HgII to preserve carbonate system parameters in organic‐rich estuarine waters

This work assesses the effectiveness of sample preservation techniques for measurements of pHT (total scale), total dissolved inorganic carbon (CT), and total alkalinity (AT) in organic‐rich estuarine waters as well as the internal consistency of measurements and calculations (e.g., AT, pHT, and CT) in these waters. Using mercuric chloride (HgCl2)‐treated and untreated water samples, measurements of these carbonate system parameters were examined over a period of 3 months. Respiration of dissolved organic matter in untreated samples created large discrepancies in CT concentrations (~37 μmol kg−1 increase, p < 0.0001), while CT was effectively constant in treated samples (3095.0 ± 1.14 μmol kg−1). AT changes were observed for both treated and untreated samples, with HgCl2‐treated samples showing the greatest variation (~ 26 μmol kg−1 decrease, p < 0.001). In response to changing AT/CT ratios, pHT changes occurred in both treated and untreated samples but were relatively small in treated samples. Results in organic‐rich estuarine waters that reflect the in situ carbonate system characteristics of the samples at the time of collection can be improved when samples obtained for CT and AT analysis are collected and stored separately. Accurate analyses of CT can be obtained by filtration and preservation with HgCl2. Accuracy of AT analyses can be improved by filtration and storage without adding HgCl2. The quality of pHT measurements can be improved by prompt analysis in the field and, if this cannot be accomplished, then samples can be preserved with HgCl2 and measured in the laboratory within 1 week.

Accurate measurements of carbonate system parameters, including pH T (total scale), total dissolved inorganic carbon (C T ), total alkalinity (A T ), and carbon dioxide fugacity ƒ(CO 2 ), in coastal and estuarine environments are essential for understanding the mechanisms controlling ecosystem health and its resources (Doney, 2006;Cai et al., 2021).Characterization of the carbonate system is fundamental for a quantitative understanding of buffer intensity, indicating the extent to which pH is maintained within the limited range required for optimal functioning of biological processes (Butler, 1991;Pilson, 2012;Stumm and Morgan, 2012).Measurements of in situ pH T along with either C T or A T are essential, as well, for calculating in situ saturation states (Byrne, 2014) relevant to calcareous organisms.Due to the substantial complexity of low salinity (S < 20) environments with high organic matter content, and their susceptibility to rapid changes from freshwater sources, these ecosystems are poorly understood relative to pelagic environments (Hunt et al., 2011;Song et al., 2020).
Carbonate system parameters are typically preserved in the field for subsequent analysis in a suitable laboratory.Following a geochemical investigation of four rivers that discharge into Tampa Bay, Florida USA, where large seasonal variations were observed (Moore, 2022), a quantitative assessment of sample preservation techniques was conducted over a 3-month period.Time series measurements of key carbonate system parameters were assessed using both mercuric chloride (HgCl 2 )-treated and untreated water samples.Although sample preservation techniques have been extensively evaluated in some environments (Wong, 1970;Dickson et al., 2007), studies in organic rich waters are rare (Mos et al., 2021).In the investigation of Mos et al. (2021), the effectiveness of diverse preservation and storage methods on the total alkalinity (A T ) and pH of aqueous environmental samples (seawater, estuarine water, freshwater, and groundwater) were examined over a period of 0, 1, and 6 months using potentiometric techniques.Our investigation extends the work of Mos et al. (2021) by documenting changes in three carbonate system parameters pH T (i.e., Àlog[H + ] T ), C T , and A T for low salinity organic-rich estuarine samples preserved with and without HgCl 2 using state-of-the-art spectrophotometric techniques (pH T and A T ) and coulometric methods (C T ) over a highly resolved 3 month timescale (daily to weekly).
In addition to addressing sample stability issues, these measurements allow assessment of how sample-storage and preservation influence the final concentrations of directly measured parameters (e.g., A T ), and parameters that are calculated from, for example, C T and pH T .Investigations of the internal consistency of carbonate system measurements in high salinity marine waters (direct measurements vs. calculations) have been conducted over several decades.The internal consistency between measured and calculated carbonate system parameters has improved over time due to improvements in both measurement procedures and refinements of the models that are used in carbonate system calculations.For high salinity marine data sets, a variety of investigations have shown that differences between A T that is directly measured and A T that is calculated from pH T and C T are currently on the order of 3-4 μmol kg À1 .This difference between measured and calculated A T is similar to uncertainties typically observed for A T measurements (Millero et al., 1993;Patsavas et al., 2015;Fong and Dickson, 2019).Prior assessments of carbonate system parameters showed that internal consistency evaluations (i.e., measurements and calculations of A T ) have been limited to the range of salinity (S) within which sulfonephthalein pH indicators, such as meta-cresol purple (mCP), have been carefully calibrated (20 ≤ S ≤ 40).Until the recent publications of Douglas and Byrne (2017) and Müller and Rehder (2018), carbonate system internal consistency evaluations using modern spectrophotometric pH measurement techniques in estuarine and fresh waters have not been possible.In addition to providing a general assessment of the use of HgCl 2 as a preservative for measurements of carbonate system parameters in organicrich estuarine waters, our investigation provides perspectives on the influence of Hg II on carbonate system internal consistency calculations under low salinity conditions.

Sample collection
Hillsborough River water samples were collected from the boardwalk of the Lowry Park Riverside Trail (28.0138N, 82.4646 W) (Fig. 1) on 26 January 2021, during an incoming tide of a mixed semi-diurnal tidal cycle.Water temperature (21.47 C) and salinity (3.439) were promptly measured using a Sea-Bird Scientific Micro Thermosalinograph, and then two 30-L Niskin bottles were used simultaneously to collect surface river water.Water from each Niskin was filtered, promptly analyzed spectrophotometrically to establish initial pH T , and then preserved for analysis of major inorganic nutrients.The filtration was performed under positive pressure (Bockman and Dickson, 2014) using a 12 V peristaltic pump (GeoTech) and peristaltic tubing (Masterflex).The tubing was connected to a 142 mm diameter acrylic filter housing (GeoTech) that contained a 0.45 μm cellulose nitrate filter (GeoTech).Water collected for analysis of major inorganic nutrients was additionally filtered using a sterile 0.22 μm pressure filter (Sterivex) attached by Luer lock to a 60 mL syringe and then stored in a 30 mL acid-washed high density polyethylene bottle (Nalgene).Once each nutrient sample bottle was sealed, the bottle was stored on ice, transported to the laboratory, and then frozen until analysis (Dore et al. 1996).Next, water samples from each Niskin were filtered into a total of 70 borosilicate glass bottles (300 mL total volume) using the methods described by Dickson et al. (2007).Water samples from the first Niskin (35 samples) were treated with 100 μL of 6.5% HgCl 2 (LabChem Cat # LC166201 Lot # G102-16).Water samples from the second Niskin (35 samples) were not treated.Both sets of bottles were promptly sealed with Apiezon M grease and capped with a BOD bottle cap (Wheaton) for transport.After these samples were collected, additional water from each Niskin was filtered and analyzed spectrophotometrically to determine the pH T of the water in each Niskin immediately after the final water samples were drawn for carbonate system analyses.Both Niskin bottles were then sampled for analysis of major inorganic nutrients.All water samples were then transported to the U.S. Geological Survey Carbon Laboratory in St Petersburg, Florida where the borosilicate bottles were stored in the dark at room temperature ($ 25 C) and sequentially analyzed, beginning hours after collection, over a period of 3 months for pH T (total scale), total dissolved inorganic carbon (C T ), and total alkalinity (A T ).

Sample analysis
Field analysis (pH T ) Until recently, pH measurements in low ionic strength waters (S < 20) were limited to potentiometric methods.Recent advances in the characterization of spectrophotometric sulfonephthalein indicator dyes and pK2e2 models for spectrophotometric pH T and A T analysis have enabled analyses in waters over a full river-to-sea range of salinity (0 ≤ S ≤ 40) and temperature (0 ≤ T ≤ 40) (pH: Douglas and Byrne, 2017;Müller and Rehder, 2018;A T : Hudson-Heck et al., 2021).Sulfonephthalein indicator dyes for spectrophotometric pH measurements are molecularly characterized using calibration buffers and eliminate bias caused by salinity-induced variations in the liquid junction potentials of potentiometric pH systems.Spectrophotometric pH T measurements have since been directly compared to potentiometric methods by repeated analyses at low ionic strengths and are shown to be more accurate (French et al., 2002;Young et al., 2022).Notably, spectrophotometric pH T observations are used to calibrate glass pH electrodes and ion sensitive field effect transistor (ISFET) pH electrodes in seawater (Easley and Byrne, 2012) and estuarine waters (Martell-Bonet and Byrne, 2020).Spectrophotometric pH T field analyses were performed using the methods of Clayton and Byrne (1993).Each filtered pH T sample was collected in a 10 cm pathlength cylindrical glass cell and placed within an Ocean Optics CUV-UV-10 cuvette holder connected to an Ocean Optics LS-1 Tungsten light source and an Ocean Optics USB2000 spectrometer.Two 10 μL additions of purified mCP indicator dye (Liu et al., 2011) were added with a Gilmont GS-1100 Micrometer Syringe, and pH T was corrected for any indicator dye perturbation by back correction from the second dye addition (Clayton and Byrne, 1993).Absorbance measurements were made using the Ocean Optics software package OOIBase 32.The measurement temperature for these analyses ($ 26.5 C) was determined using a Fluke 51 II Handheld Digital Probe Thermometer.The temperatures provided by this probe were in good agreement with those obtained with a Hewlett Packard 2804A quartz thermometer (ΔT ≈ 0.125 AE 0.024 C).Sample pH T was calculated on the total scale using the algorithm of Müller and Rehder (2018).Each pH T was temperature corrected to 20 C in CO2sys.m (van Heuven et al., 2011) using the K 1 K 2 constants of Waters et al. (2014), the KSO 4 constants of Dickson (1990), and the total boron to salinity ratio (B T /S) of Lee et al. (2010).

Laboratory analysis of pH T , C T , and A T and nutrients
HgCl 2 -treated and untreated water samples were analyzed over a period of 3 months.Samples were first analyzed within a few hours of collection and then subsequently analyzed every other day for 7 d.After the first week, samples were analyzed weekly for the duration of the experiment.The measurement methods used over this 3-month period were identical for each type of analysis, and samples were analyzed in the order they were initially collected.On each day of analysis, four sample bottles, two HgCl 2 -treated samples and two untreated samples, were analyzed in the following order: pH T , C T , and A T .
Spectrophotometric pH T analysis in the laboratory was performed using the methods of Clayton and Byrne (1993).Sample water was poured from the borosilicate bottle into two 10 cm pathlength cylindrical glass cells.Each cylindrical cell was placed in a temperature-controlled Ocean Optics CUV-UV-10 cuvette holder connected to an Ocean Optics LS-1 Tungsten light source and Ocean Optics USB2000 spectrometer.Methods for pH T analyses in the laboratory were identical to those used for pH T measurements in the field with the exception that pH T field samples were collected directly from each Niskin and analyzed at ambient temperature.
Immediately after the pH T measurements, water samples for duplicate total dissolved inorganic carbon (C T ) analyses were withdrawn from the sample bottle.Using a 60 mL syringe connected to a 3-way Luer lock valve and stopper to minimize gas exchange, two 20 mL same samples were drawn from each bottle.Analyses were performed with a CM5017 CO 2 carbon coulometer coupled to a CM5330 acidification module (UIC) following the methods described by Dickson and Goyet (1994).Sample weights were determined using a Denver Instruments PI-214 analytical balance (AE 0.1 mg).Injected sample mass was determined as the difference in syringe weights before and after each sample was injected through a septum into the stripping chamber of the acidification module.Samples were acidified with 10 mL of 8.5% H 3 PO 4 .Analytical grade N 2 was used as the carrier gas for CO 2 from the acidified samples to the coulometer.The titration endpoint was determined by the coulometer, and C T was calculated using software from UIC. Accuracy and precision was determined from analysis of certified reference materials (CRM) obtained from Andrew Dickson at the Scripps Institution of Oceanography (Dickson, 2010).Repeat measurements (n = 51) of CRM Batch 189 and CRM Batch 183 yielded precisions of AE1.09 and AE 0.71 μmol kg À1 respectively (standard deviation of each average daily CRM analysis) and an accuracy of 1.43 AE 1.04 μmol kg À1 (average of the absolute value of the difference between each measured CRM and the reported batch concentration).
After completion of the C T analyses, measurements of total alkalinity (A T ) were performed using the spectrophotometric methods of Yao and Byrne (1998) and Liu et al. (2015).Water sample mass ($ 100 g) and masses of added acid were determined gravimetrically using a Denver Instruments PI-214 analytical balance (AE 0.1 mg).Water samples were analyzed in an open top square glass cell (Hellma Cells) placed within a custom plastic frame connected to an Ocean Optics LS-1 Tungsten light source and Ocean Optics USB2000 spectrometer.Bromocresol purple (BCP) indicator dye (4 mM stock solution) was administered by pipette (100 μL), and the sample was titrated using 0.100 AE 0.0001 N standardized HCl (Lab Chem) added with a plastic 10 mL syringe fitted with a Teflon syringe needle.Absorbance measurements were made using the Ocean Optics software package OOIBase 32 (Landis, 2005).The pH T of each solution was measured throughout the titration to an endpoint near pH T 4.3.The total weight of added acid was determined from the difference in the syringe weight before and after acid addition.At the end of each titration, the solution was purged of CO 2 with a stream of N 2 gas that had been pre saturated with Milli-Q H 2 O.After purging, final absorbance measurements were made, and solution temperature was determined using a Fluke 51 II Handheld Digital Probe Thermometer.A T was calculated using the BCP equations of Hudson-Heck et al. (2021).Accuracy and precision were determined from analysis of CRMs from Scripps Institution of Oceanography (Dickson, 2010).Repeat measurements (n = 33) of CRM Batches 183, 186, and 189 yielded precisions of AE 0.93 μmol kg À1 , AE 0.76 μmol kg À1 , and AE 1.39 μmol kg À1 respectively, and an accuracy of 0.77 AE 0.93 μmol kg À1 .
Calculations of A T internal consistency from measured pH T and C T were performed in CO2sys.m (van Heuven et al., 2011) using the K 1 K 2 constants of Waters et al. (2014), the KSO 4 constants of Dickson (1990), and the total boron to salinity ratio (B T /S) of Lee et al. (2010).Significance tests with 95% confidence levels were performed in Microsoft Excel using the Data Analysis Regression tool.Inorganic nutrient analyses were performed at the University of Tampa.Samples were analyzed for inorganic nitrogen (NO 3 À + NO 2 À ), inorganic phosphate (H 2 PO 4 À + HPO 4 2À + PO 4 3À ), ammonia (NH 3 + NH 4 + ), and silica (SiO(OH) 3 À + Si (OH) 4 )) on a Seal Analytical Autoanalyzer 3 using the methods of Gordon et al. (1993).

Assessment and discussion
Clear distinctions were observed over 3 months between HgCl 2 -treated and untreated water samples for all three measured carbonate system parameters: pH T , C T , and A T .From the first analyses, C T (untreated) was greater than C T (treated), and the difference between C T (untreated) and C T (treated) increased throughout the 91 d a period of observation (Fig. 2 The Fig. 2 results suggest that respiration of dissolved organic matter (DOM) dramatically increased C T in the untreated samples (Fig. 2).In contrast, results for the treated C T samples show that the addition of HgCl 2 was effective at minimizing variation over a period of 3 months.It is noted that, at the inception of the experiment, there was a small C T difference between the two sample types (HgCl 2 -treated and untreated) (Fig. 2); the calculated C T intercepts differ by about 5.2 AE 2.4 μmol kg À1 .This 5.2 μmol kg À1 difference may be a real sampling artifact (i.e., a small difference between samples collected simultaneously at the same location); the difference is similar to the 1 σ uncertainty (AE 2.4 μmol kg À1 ) in the calculated difference between the two intercepts.
A T results also show a large difference between HgCl 2 -treated and untreated samples (Fig. 3).Unlike C T observations, however, a large difference between the HgCl 2 -treated and untreated samples was observed at the inception of the measurement period.Furthermore, in contrast to the near constancy of C T for the HgCl 2 -treated samples, the A T of the untreated samples was relatively constant, decreasing by approximately 7.1 μmol kg À1 over the course of the experiment (p < 0.01), while the A T of HgCl 2 -treated samples decreased by approximately 26 μmol kg À1 over the course of the experiment (p < 0.001).Overall, at the beginning of the measurement period the A T of the HgCl 2 -treated samples was lower than the A T of the untreated samples by more than 26 μmol kg À1 , and this difference increased to more than 45 μmol kg À1 at the end of the 3-month period of measurements.
Respiration of DOM likely caused A T to moderately decrease through time in the untreated samples (Fig. 3).Consistent with the modern revised Redfield equation (P T : N T : C T = 1 : 16 : 124) (Broecker et al., 1985;Takahashi et al., 1985;Anderson and Sarmiento, 1994;Körtzinger et al., 2001), respiration of DOM produces an increase in C T that is about 7.3 times larger than the accompanying decrease in A T .Based on the 37 μmol kg À1 increase in C T over 91 d, and the expected factor of 7.3 difference between changes in C T and changes in A T (i.e., ΔC T /ΔA T = 7.3), the expected decrease in A T is somewhat smaller than the observed decrease.However, considering the uncertainties in slope in Figs. 2 and 3 for the untreated samples, the expected change in A T calculated as ΔC T /7.3 (i.e., 5.1 AE 0.5 μmol kg À1 ) is reasonably consistent with the observed change in A T (7.1 AE 2.4 μmol kg À1 ).
Our results indicating that HgCl 2 -treated A T samples (Fig. 3) show a larger decrease in A T than untreated samples are consistent with observations of Mos et al. (2021) and suggest that this decrease is due to complexation of DOM by Hg II .Hg II has a strong affinity for organic ligands such as those found in the DOM of natural waters (Andren and Harriss, 1975;Mierle and Ingram, 1991;Varshal et al., 1996;Ravichandran, 2004).The increasing A T difference between treated and untreated samples over time is plausibly attributable to a slowly increasing extent of DOM complexation by Hg II .After the addition of a comparable concentration (0.03%) of HgCl 2 to fresh, estuarine, and groundwater water samples stored in borosilicate glass bottles, Mos et al. ( 2021) observed a rapid decrease in A T followed by a subsequent decrease over a period of 1 or 6 months.A decrease in the alkalinity of treated DOM rich samples would be expected if organic bases are complexed by Hg II and a portion of these complexes do not dissociate (i.e., remain untitrated) at the lowest pH in an alkalinity titration.In this study, the extent of DOM complexation by Hg II could not be directly assessed.However, it is shown for the first time, at high resolution, that Hg II addition, which is the standard method of sample preservation for carbonate system water samples, rapidly affects the pH T and A T of estuarine water samples after collection.Toward the objective of understanding A T characteristics in coastal waters, this work also suggests that organic alkalinity titrations (i.e., characterization of the fraction of DOM that actively exchanges hydrogen ions) may be more informative than measurements of DOM.
The results for prompt pH (20 C) T measurements in the field (shown as triangle and diamond markers near time zero) and measurements made in the laboratory (shown as circle  For untreated samples, Fig. 4 shows a temporal decrease in pH T consistent with increasing C T and decreasing A T .For the HgCl 2 -treated samples, the observed decrease in pH T can be attributed solely to decreasing A T through time, which is relatively small.The initial offset in pH T , whereby pH T (untreated) was greater than pH T (HgCl 2 -treated), was consistent with the substantial initial alkalinity difference between the treated and untreated samples and is unlikely a result of Hg II hydrolysis.The water samples salinity of 3.439 implies a sufficiently large chloride concentration to preclude substantial hydrolysis of Hg II (Hudson-Heck and Byrne, 2022).
Nutrient results collected from each Niskin before and after borosilicate glass bottle sampling suggest that the waters sampled by the two Niskin bottles were somewhat different (Table 1).For inorganic nitrogen, phosphate, and ammonia, the nutrient concentrations in Niskin 1 were 1.4-2.5 times greater in than in Niskin 2, while silica in Niskin 2 was 1.3 times greater than in Niskin 1.Although this is substantial evidence for real distinctions between the initial conditions of the HgCl 2 -treated and untreated samples (rather than differences created during laboratory analyses), it is the large changes in C T , A T , and pH T through time, and the increasingly large differences between HgCl 2 -treated and untreated samples that is the critically important focus of this investigation (i.e., Figs.2-4).
DOM measurements were not made in this study; however, the sample site has well known and consistent DOM sources upstreamterrestrial inputs from the Green Swamp region, an area of 560,000 acres that drain to create the headwaters of four major rivers including the Hillsborough River, and ground water inputs from numerous springs of varying magnitude (Lewis and Estevez, 1988;Southwest Florida Water Management District, 2018).These two water types are known to be significant sources of dissolved organic groups with substantial and wide-ranging affinities for H + ions (Song et al., 2020).As such, these organics are expected to have Note that the first field measurements, denoted as "a," occurred several minutes before sample bottle preservation (i.e., the beginning of elapsed time axis), while the second and final field measurements, denoted as "b," occurred after sample bottle preservation.The time between the first and second field measurements was $ 45 min.The first laboratory measurements occurred $ 2.25 h after the final field pH T measurements.Dotted lines represent best fit linear trendlines only for data points obtained from stored samples that were subsequently measured in the laboratory over time and not data from water samples that were collected and immediately measured in the field.substantial affinities for cations such as Hg II that exhibit strong covalent behavior (Martell and Hancock, 1996).Such associations are much weaker in seawater because of strong complexation of Hg II by chloride ions at higher salinities.
The initial pH T and A T differences (HgCl 2 -treated vs. untreated) in the first weeks of the measurements were unexpected.However, as shown in Table 1, initial differences in A T , pH T , and C T for the treated and untreated water samples were consistent with well-defined differences in observed inorganic nutrient concentrations.The small initial C T differences in the treated and untreated water samples can be attributed to differences in the chemical composition of the two collected water samples.The most plausible explanation for the large initial A T and pH T differences between treated and untreated samples is initial differences between the collected water samples plus rapid complexation of organic bases by added Hg II that decreased A T and released H + ions, which decreased pH T .Consistent with the observations of Mos et al. (2021), the extent of this complexation then continued over a period of months (Fig. 3).
The relatively recent development of carbonate system models appropriate to low ionic strength waters allows comparisons between direct measurements of A T (Fig. 3) and values that are calculated from observed values of C T (Fig. 2) and pH T (Fig. 4). Figure 5 shows a comparison between values of A T (measured) and A T (calculated) that reveal the length of time that A T and C T samples can be stored before changes in the samples begin to compromise use of the samples to accurately estimate the characteristics of the samples at the time of collection.For the purpose of this analysis, the initial pH T of the sample water was taken as the average of prompt measurements obtained in the field prior to sample sequestration in one of the 70 collection bottles.These values are shown in Fig. 4 as those that were first analyzed after t = 0 d.These pH T measurements were paired with C T measurements obtained over a period of 91 d for HgCl 2 -treated samples in order to obtain A T (calculated).The A T (measured) values shown in Fig. 5 are those for the untreated samples in Fig. 2 because it is these values that are clearly closely identified with the carbonate system characteristics of the samples at their time of collection (i.e., prior to complexation of organics by Hg II ). Figure 5 shows that A T (measured) results are larger than calculated values, signifying the presence of titratable organic alkalinity (Note, however, that small contributions from river-derived boron are also possible; Lee et al., 2019).The average difference between measured and calculated A T (A T (measured) À A T (calculated) was 8.52 AE 4.6 μmol kg À1 .While this difference is comparable to results obtained by Patsavas et al. (2015) for open ocean waters, there is a clear difference between results obtained prior to day 49 (11.6 AE 2.2 μmol kg À1 ) and results obtained for days 49-91 d (3.49AE 2.3 μmol kg À1 ).This change appears to be attributable to respiration-induced changes in untreated samples stored for A T analysis.Unless storage at lower temperature mitigates this respiration problem, this indicates that there are limits on the time that A T samples can be stored satisfactorily.The results shown in Figs. 3 and  Hg II to preserve organic-rich waters satisfactory basis for predicting the C T characteristics of the sample at the time of collection.
Improved approach for sample preservation in organic rich waters The overarching objective of sound storage and preservation methods is to maintain samples so that they reflect the in situ carbonate system characteristics of the samples at the time of collection.The results obtained in this work demonstrate that, for C T measurements, preserving (treating) samples with HgCl 2 is effective in achieving this objective (Fig. 2).Over a period of 3 months, C T measurements for the treated samples were essentially invariant while the filtered samples that were not treated (untreated) exhibited substantial changes.Although it is possible that sample storage at low temperatures, as advocated by Mos et al. (2021), would reduce or even eliminate respiration-derived C T in untreated samples, our results clearly indicate the effectiveness of HgCl 2 additions in reducing variability in treated C T samples.While Mos et al. (2021) demonstrated that their procedures (filtration, storage in polypropylene bottles, and refrigeration at 4 C) effectively eliminated changes in A T , their work did not measure the effects of these methods on eliminating changes in C T .It is much easier to detect changes in C T than for A T (i.e., ΔC T / ΔA T = 7.3); and, unless further work demonstrates that filtration in conjunction with low temperature storage eliminates changes in C T , it appears prudent to preserve samples for C T analysis through addition of HgCl 2 .
Our results indicate that addition of HgCl 2 and subsequent long-term storage of organic-rich samples for A T analysis is ineffective (Fig. 3).Interactions of organic alkalinity (the fraction of DOM that actively exchanges hydrogen ions) and Hg II in organic-rich low salinity samples appear to increase through time, and this confounds interpretation of A T results obtained using treated samples.In contrast, for untreated samples, although decreases in A T are observed that are broadly consistent with the increases in C T (Fig. 2) the changes are small.As such, as shown in the study of Mos et al. (2021), it is possible that storage at lower temperature would further reduce changes in A T during storage.Therefore, consistent with the results of Mos et al. (2021), improved accuracy of measurements for A T analysis can be achieved when samples are filtered and stored without adding HgCl 2 .Minimizing the time between sample collection and analysis improves accuracy of results (Fig. 3).Notably, preserving C T samples with HgCl 2 , but not A T samples, requires that the two types of samples are stored separately.
The results shown in Fig. 4 indicate that long-term storage of samples for pH T analyses affects accuracy of results.Sample pH T is sensitive to changes in C T and A T .In the absence of Hg II additions, respiration of DOM increases C T (Fig. 2) thereby decreasing the pH T .For treated samples, changes in A T can alter pH T measurements.Therefore, prompt pH T measurements at the time of collection can improve accuracy of results.In contrast to measurements of C T and A T , spectrophotometric pH T measurements are uniquely amenable to analysis in the field.As an alternative, if measurements of pH T in the laboratory are preferred, average Hg II -treated measurements of pH T over 1 week of storage were shown to be similar (7.304AE 0.013) (Fig. 4) to average field pH T measurements (7.310 AE 0.019) obtained immediately before the samples were bottled.

Comments and recommendations
Although DOM measurements can provide context surrounding the potential significance of interactions between organics and Hg II , titrimetric measurements of organic acid/ base concentrations and pK values in the presence and absence of Hg II would be especially useful for improving interpretations of A T in organic-rich systems.While further investigation of sample preservation and storage is desirable, the results obtained in this study support the following improvements for carbonate system sample collection and analysis of organic-rich waters: • Accurate analyses of C T can be achieved by filtration and preservation with HgCl 2 .• A T analyses can be improved by filtration and separate storage.Sample storage in polypropylene bottles at 4 C as advocated by Mos et al. (2021) may be advantageous.Preservation with HgCl 2 is not advisable.• Good quality measurements of pH T can be obtained by prompt analyses in the field.If this is not feasible, then samples can be preserved with HgCl 2 and measured in the laboratory within $ 1 week.

Fig. 1 .
Fig. 1.A map showing the location of the sample collection site on the Hillsborough River in relation to the upper fringe of Hillsborough Bay in Tampa Bay (upper) and to Lowry Park (lower).
).The C T of HgCl 2 -treated samples were essentially constant throughout the duration of the experiment (3095.0AE 1.14) while the C T of untreated samples increased by $ 37 μmol kg À1 over 91 d.The regression of C T vs. time for the untreated samples (C T = 3100.2AE 1.78 + [0.4088 AE 0.04] x, where x = days), with a significance test result of p < 0.0001 (Fig. 2), indicated that C T increased by $ 0.41 μmol kg À1 per day over the course of the experiment (note that anomalous results obtained on day 2 for both HgCl 2 -treated and untreated samples are not included in the figure but are shown in the associated data release, Moore et al. (2022), doi.org/10.5066/P9J9IYFD).

Fig. 2 .
Fig. 2. Average (AVG) measured C T for treated (HgCl 2 ) and untreated samples over time where error bars are the standard deviation of the average from each day of analysis.Dotted lines represent a best fit linear trendline for the data.

Fig. 3 .
Fig. 3. Average (AVG) measured A T for treated (HgCl 2 ) and untreated samples over time where error bars are the standard deviation of the average from each day of analysis.Dotted lines represent a best fit linear trendline for the data.
and square markers) after field-collection in borosilicate bottles are shown in Fig.4.As was observed for C T and A T , a clear difference was observed between HgCl 2 -treated and untreated pH T samples.Consistent with the decreasing A T and the substantially increasing C T for the untreated samples, pH T (untreated) decreased from an average of 7.377 AE 0.022 for days 1 through 21 to 7.317 AE 0.024 for days 28 through 91.The pH T changes of the HgCl 2 -treated samples were smaller than those of the untreated samples, averaging 7.302 AE 0.018 for days 1 through 21 and 7.275 AE 0.020 for days 28 through 91.

Fig. 4 .
Fig. 4. Average (AVG) measured pH T made in the field, notated as Niskin Field 1 and Niskin Field 2, and average measured pH T made in the laboratory, notated as HgCl 2 AVG (Niskin 1) and Untreated AVG (Niskin 2), where error bars are the standard deviation of the average from each day of analysis.Note that the first field measurements, denoted as "a," occurred several minutes before sample bottle preservation (i.e., the beginning of elapsed time axis), while the second and final field measurements, denoted as "b," occurred after sample bottle preservation.The time between the first and second field measurements was $ 45 min.The first laboratory measurements occurred $ 2.25 h after the final field pH T measurements.Dotted lines represent best fit linear trendlines only for data points obtained from stored samples that were subsequently measured in the laboratory over time and not data from water samples that were collected and immediately measured in the field.
Fig. 5. Untreated average (AVG) measured A T vs. calculated A T from HgCl 2 -treated C T measurements over time and the average of pH T measurements made promptly in the field from both Niskin bottles.A T error bars for untreated samples are the standard deviation of the measured average from each day of analysis.Calculated A T error bars are the standard deviation of the average calculated A T values from HgCl 2 -treated C T replicates and the average of field pH measurements from each Niskin bottle before and after BOD sample bottle collection.Calculated A T measurements were performed with CO2sys.m using the ancillary salinity, temperature, average phosphate, and average silica values.

Table 1 .
Mean and standard deviation (Mean AE SD) of measured inorganic nutrients sampled from each Niskin bottle before and after borosilicate glass bottle sampling.