Optimization and quality control of suspended particulate matter concentration measurement using turbidity measurements

The dry weight concentration of suspended particulate material, [SPM] (units: mg L–1), is measured by passing a known volume of seawater through a preweighed filter and reweighing the filter after drying. This is apparently a simple procedure, but accuracy and precision of [SPM] measurements vary widely depending on the measurement protocol and experience and skills of the person filtering. We show that measurements of turbidity, T (units: FNU), which are low cost, simple, and fast, can be used to optimally set the filtration volume, to detect problems with the mixing of the sample during subsampling, and to quality control [SPM]. A relationship between T and ‘optimal filtration volume’, Vopt, is established where Vopt is the volume at which enough matter is retained by the filter for precise measurement, but not so much that the filter clogs. This relationship is based on an assessment of procedural uncertainties in the [SPM] measurement protocol, including salt retention, filter preparation, weighing, and handling, and on a value for minimum relative precision for replicates. The effect of filtration volume on the precision of [SPM] measurement is investigated by filtering volumes of seawater ranging between one fifth and twice Vopt. It is shown that filtrations at Vopt maximize precision and cost effectiveness of [SPM]. Finally, the 90% prediction bounds of the T versus [SPM] regression allow the quality control of [SPM] determinations. In conclusion it is recommended that existing [SPM] gravimetric measurements be refined to include measurement of turbidity to improve their precision and quality control. *Corresponding author: E-mail: [email protected]. Present address: Scripps Institution of Oceanography, Marine Physical Laboratory, University of California–San Diego, La Jolla, CA, USA. Acknowledgments Daniel Saudemont, Marc Knockaert, Gijs Coulier, and Edwige Devreker of the Chemistry Laboratory of the Management Unit of the North Sea Mathematical Models (MUMM) are acknowledged for suspended matter and phytoplankton pigment data lab analysis. The crew of the Belgica research vessel is thanked for kind help during sea campaigns. We thank Lucie Courcot of the Laboratoire d’Océanologie et Géosciences (LOG) for electron microscopy analysis. The lead author was supported by the Belcolour-2 and GEOCOLOUR projects funded by the STEREO (Support to the Exploitation and Research in Earth Observation data) programme of the Belgian Science Policy Office under contracts SR/00/104 and SR/00/139, respectively. DOI 10.4319/lom.2012.10.1011 Limnol. Oceanogr.: Methods 10, 2012, 1011–1023 © 2012, by the American Society of Limnology and Oceanography, Inc. LIMNOLOGY and OCEANOGRAPHY: METHODS eral particles (Stramski et al. 2004). The term ‘total’ may be misleading, however, since very small particles pass through the filter and their dry weight is not included. We therefore adopted the symbol SPM. The dry weight concentration of SPM, [SPM] in units of mg L–1 or g m–3, is determined gravimetrically by passing a known volume of seawater through a preweighed filter. The filter is then reweighed after drying and [SPM] is calculated from the ratio of the difference in filter weight by the volume of the filtrate. Protocols for [SPM] measurement vary widely in procedures for filter preparation and treatment, including washing, drying, and ashing, and washing of sea salt after filtration. Also, while the measurement of [SPM] is apparently a simple procedure, accuracy and precision of the measurements vary widely depending on the measurement protocol (materials used, filter preparation and treatment, laboratory conditions, etc.) and the experience and skills of the person filtering. Because [SPM] is defined operationally, many measurement protocol specifications have been evaluated previously. The retention of salts by glass fiber filters leading to overestimation of [SPM] has gained considerable attention, and washing of filters and filter edges with deionized water (or MilliQ water) after filtration have been proposed to remove sea salt (Strickland and Parsons 1968; van der Linde 1998). Different wash volumes have been recommended, varying between 30 mL (Pearlman et al. 1995) and 250 mL (Sheldon 1972). Despite a MilliQ wash of 300 mL, Stavn et al. (2009) found salt retention by 47 mm diameter GF/F filters to vary between 0.6 mg and 1.1 mg with increasing salinity from 15 to 34 PSU (Practical Salinity Units, see their Fig. 1) and irrespective of filtration volume. Organic material may be lost from living cells through cell-wall rupture by osmotic gradient after rinsing with MilliQ (Goldman and Dennett 1985) and/or by air suction (Goldman and Dennett 1985; Kiene and Linn 1999). Such material losses are dependent on species (Booth 1987; Kirst 1990) and filter type (Kiene and Linn 1999) and are considered to be less important on GF/F filters (van der Linde 1998), which work by adsorption. Some protocols state that the rinsing should be done with 10-20 mL of isotonic ammonium formate solution to minimize osmotic shock (ICES 2004; PML and ICES 2004). Drying time and temperature affect final dry weight (Lovegrove 1966). The vacuum pressure under which filtration takes place was not found to affect the mass retention by the filters (Sheldon 1972), even though delicate particles might break when the pressure is too high. A pressure of 300-400 mmHg is recommended (Stavn et al. 2009). The effective pore diameter of glass fiber or polycarbonate filters is known to decrease from the nominal value with increasing filtration volume until the filter is clogged (Sheldon 1972; Sheldon and Sutcliffe 1969). The filtration volume should be such that the mass retained by the filter is sufficient to be precisely measured, but not so much that the filter clogs. Despite its importance, the estimation of filtration volume is somewhat arbitrary and depends on the experience of the person carrying out the filtration. Typically, the person carrying out the filtration determines the filtration volume from visual inspection of the seawater sample. In this study, we investigate how low cost, simple, and fast measurements of turbidity, which is a good proxy for [SPM] (Boss et al. 2009; Neukermans et al. 2012), can be used to estimate filtration volume objectively and hence improve reproducibility of measurements. We further investigate [SPM] measurement uncertainties associated with filter preparation and treatment, salt retention, and filtration volume. Whereas the approach described in this paper is specific to the measurement of mass concentration of SPM, the concept of preand post-filtration turbidity measurements may be applicable to the improvement of the quality and the quality control of the measurement of many other physical or chemical properties of SPM (e.g., Groundwater et al. 2012). Materials and methods Measurement of [SPM] Measurement protocol [SPM] is determined gravimetrically following the protocol of Tilstone et al. (2002), based on van der Linde (1998), by filtration of a known volume of sea water onto 47 mm Whatman GFF glass fiber filters with a nominal pore size of 0.7 μm. The filters were pre-ashed at 450°C for 1 h (see step 1 in the flowchart in Fig. 1), gently washed in 0.5 L of MilliQ water (2) to remove friable fractions that can be dislodged during filtration, dried at 75°C for 1 h (3), pre-weighed on a Sartorius LE 2445 analytical balance with an accuracy of 0.1 mg, denoted wb (4), stored in a desiccator for use within 2 weeks (5), and transferred to clean 50 mm diameter Petri plates for transport. Seawater samples were filtered immediately after collection on triplicate ashed and preweighed filters using a 250 mL Millipore apparatus with an applied vacuum of 300-400 mmHg. Filter supports were washed before filtration with MilliQ to remove any particles that had adhered to the glass. After placement on the fritted glass filter supports (6), filters Neukermans et al. Optimizing [SPM] Measurement 1012 Fig. 1. Procedural flow for the measurement of [SPM] of seawater. were wetted with MilliQ (7), and a known volume of seawater, V, was passed through the filter (8). The measuring cylinder was rinsed with 3 ¥ 30 mL aliquots of MilliQ water to flush any remaining particles (9). To remove salt, filters were washed with 250 mL of MilliQ water after filtration (10). The filter funnel was also rinsed with 3 ¥ 30 mL aliquots of MilliQ water (11). After removal of the funnel, the filter edge was carefully washed with MilliQ to flush possible diffused salt (13, Strickland and Parsons 1968). The total MilliQ wash volume per filter is thus 400-450 mL, much larger than recommended by Sheldon 1972, 300 mL; Trees 1978, 50 mL; and Pearlman et al. 1995, 30 mL. The samples were stored at –20°C until further analysis in MUMM’s Marine Chemistry Laboratory (14), usually within a few months after sampling. Filters were dried for 24 h at 75°C (15) and reweighed on the same balance (16), giving weight wa, from which [SPM] is obtained as (wa – wb): V. Filter blanks At the start and the end of each sampling campaign, a series of filter blanks, also termed procedural control filters, were included, to assess uncertainties associated with filter operations in the laboratory and during filtrations. Three different types of blank measurements have been made with filtration of a) no water (“dry blank”), b) synthetic seawater (SSW), prepared by dissolving 34 g of NaCl in 10 L of MilliQ water, and c) MilliQ water. An overview of these filter blanks and their operations is given in Table 1. The MilliQ and SSW filter blanks were treated exactly as the sample filters (steps 1-16 in Fig. 1) except that 250 mL of MilliQ or 500 mL of SSW was passed through the filter instead of a volume V of sampled seawater (step 8). No liquid was passed through the dry filter blanks, which were subjected only to freezing (step 14) before further analysis in the lab. A one-way analysis of variance (ANOVA) was carried out to test for differences between blanks (details of statistical tests are described further in the text). Salt retention tests A laboratory experiment was carried out to test whether salts diffused onto the rim of the filter were properly flushed by the rim rinsing procedure (step 13 in Fig. 1). First, all steps of the procedure as described in Fig. 1 were carried out filtering a volume of 250 mL SSW onto 10 replicate filters. Next, all steps except the rim-rinsing (step 13) were carried out filtering a volume of 250 mL SSW onto another set of 10 filters. Differences between groups were then tested using an ANOVA test. To test the dependence of salt retention on sampling volume, different volumes of SSW ranging between 150 and 2000 mL were filtered according to the procedure in Fig. 1. Differences in wa – wb between SSW volume groups were tested with an ANOVA. To check whether salts were properly flushed with the MilliQ wash of 400-450 mL, one unrinsed and one rinsed filter through which 500 mL of SSW was passed were analyzed using scanning electron microscopy (SEM, LEO 438VP tungsten filament SEM) with electron dispersive spectral analysis (EDS). Samples were sputter-coated with Au/Pd (Polaron SC7620). Turbidity measurements Turbidity, T, defined by ISO 1999 as ‘the reduction of transparency of a liquid caused by the presence of undissolved matter’, can be quantified in various ways (e.g., Secchi disk, light attenuation, side scatter). The Hach 2100P portable turbidity instrument measures the ratio of Light Emitting Diode (LED) light scattered at an angle of 90° ± 2.5° at a wavelength of 860 nm ± 60 nm to forward transmitted light, as compared with the same ratio for a standard suspension of Formazine. This optical technique for measurement of T from the side-scattering coefficient is in accordance with ISO 1999 and has significant advantages over alternative measurements of turbidity: Secchi depth measurements are obviously highly subjective and the use of instruments with a broadband light source such as the tungsten lamp suggested by EPA 1993 may be much more sensitive to spectral variations of lamp output and particle absorption properties than for the monochromatic near infrared source used here. T is expressed in Formazine Nephelometric Units (FNU) and instruments are calibrated using a set of Formazine Turbidity Standards. At the start of each sea campaign, instrument stability is ensured by recording turbidity of Hach STABLCAL Formazine standards of 0.1, 20, 100, and 800 FNU and an instrument recalibration is made if necessary. Side scattering signals are averaged over 10 measurements at 1.2 s intervals. Glass sample cells of 10 mL are used to record seawater T. The glass cell is rinsed with sampled seawater before filling. The exterior of the sample cell is rinsed with MilliQ water, dried with paper tissue, swiped with a soft microfiber lint-free cloth treated with silicon oil, and swiped again with a dry cloth. Prior to turbidity measurement, the sample cell is visually inspected for dust particles, condensation droplets, or air bubbles. T was recorded in triplicate, gently tumbling the sample cell between each measurement. Neukermans et al. Optimizing [SPM] Measurement 1013 Table 1. Overview of types of [SPM] procedural control filters and treatments. Filter Filter operations (nrs as in Fig. 1) Volume filtered n Sampled Dry blank 1-5, 14, 15-16 0 mL 87 2008-2010 MilliQ blank 1-16, replacing samples with MilliQ water in step 8 250 mL 96 2008-2010 SSW blank 1-16, replacing samples with SSW water in step 8 500 mL 126 2007-2010 Sample 1-16 variable 366 2007-2010 T is recorded before and after [SPM] filtrations to check adequate mixing of the water sample during subsampling for filtration. T measurement typically takes about 4 minutes to complete and portable turbidity meters can be purchased for less than 1500 US dollars. The 2100P model is no longer manufactured and has been superseded by the Hach 2100Q portable turbidimeter with improved measurements for rapidly settling samples. Calibration standards are available in sealed containers and are stable for at least 1 year, facilitating use of the method for scientists worldwide with minimal resources and/or in remote areas. Optimal filtration volume T as proxy for [SPM] T and [SPM] measurements were carried out in surface waters in coastal and offshore waters around Europe and French Guyana between 2007 and 2010. Sampling sites are described and mapped in Neukermans et al. 2012. A ‘least squares cubic’ type II regression (York 1966) is applied to the log transformed T and [SPM] data. The least squares cubic regression, which takes into account measurement uncertainties, is applied after removal of outliers identified by the MATLAB robustfit.m routine. Correlation coefficients are given with their 95% confidence intervals, obtained from bootstrapping. Details of these statistical procedures are described in the web appendix of Neukermans et al. 2012. Based on the [SPM]-T regression, an estimate of [SPM] can be derived from measurements of T prior to filtration. From this estimate of [SPM], the volume of seawater to be filtered can then be estimated so that an optimal mass is retained by the filter as described below. Determining optimal filtration volume The filtration volume, V, should be high enough so that the dry mass of the particles retained by the filter, wa – wb, is sufficient to be precisely measured, but not so much that the filter clogs. Its estimation requires a quantification of minimum measurement uncertainty on wa – wb, assessed from procedural control filters, and a maximum value for the relative uncertainty on [SPM]. For measurements of weight, the detection limit (DL) of the balance gives the minimum measurement uncertainty. The minimum uncertainty on the difference between filter weights before and after filtration, wa – wb, is then given by (ISO 1995):