Total phosphorus analyses: is the convenience of automated inline-UV digestion robust enough for freshwater nutrient monitoring?

Abstract

Child AW, Moore BC. 2019. Total phosphorus analyses: is the convenience of automated inline-UV digestion robust enough for freshwater nutrient monitoring? Lake Reserve Manage. 35:140–147

Phosphorus (P) is an essential and often growth-limiting nutrient in aquatic ecosystems. Excessive P in aquatic ecosystems can result in accelerated eutrophication and is a prime cause of “impaired” water quality listing under the Clean Water Act. Total phosphorus (TP) is a commonly used metric for predicting productivity and assessing nutrient enrichment in aquatic ecosystems. Accurate and precise TP data are essential for aquatic resource managers and scientists. Inline-ultraviolet (inline-UV) and persulfate digestions are both approved methods for estimating TP within water samples, and literature on their relative digestion efficiencies is available. However, the available literature to date has only evaluated relatively high TP concentrations, >90 µg P/L, considerably higher than critical trophic concentrations for natural lakes and streams. We compared inline-UV digestion with the more common autoclave persulfate digestion, analyzing TP from 32 freshwater sites in northern Idaho and eastern Washington. Sites were chosen to represent common ranges of TP (5 to 350 µg P/L) found in surface waters throughout the world. Both digestions performed adequately on third-party reference materials used for laboratory accreditation requirements. The linear model comparing logarithmic transformations of TP estimates from both digestion methods were correlated (p < 0.001, R² = 0.8110), and suggested inline-UV digestion may not accurately estimate low-level TP concentrations in natural waters. Our results indicate that inline-UV digestion results are overall not comparable to autoclave digestion (absolute mean difference 16 µg/L or 22%, p < 0.001); lake managers should be aware that TP method selection can significantly alter data interpretation for nutrient monitoring plans.

Keywords:

• Pacific Northwest
• persulfate
• TP analysis
• UV digestion
• water quality

Phosphorus (P) is an essential element for all living organisms. Since pioneering works of the 1960s and 1970s firmly linked phosphorus to eutrophication (Vollenweider 1968, Vallentyne 1974, Schindler 1977), the “phosphorus paradigm” has been a central tenet in managing and controlling water quality in freshwater ecosystems (Wetzel 2001, Cooke et al. 2005). These and an abundance of subsequent studies have shown that increases in internal and external P loading drive increased algae and/or macrophyte growth, in turn affecting water clarity, dissolved oxygen (DO), specific conductivity, alkalinity, pH, and aquatic food webs (Nürnberg 1984, Hargeby et al. 1994, Correll 1998, Lathrop et al. 1999). Long-term studies have consistently affirmed that P is the most common growth-limiting factor for algae and other primary producers and must be the focus for successful management of eutrophic waters (Schindler 1977, Guildford and Hecky 2000, Wetzel 2001, Schindler et al. 2008, Søndergaard et al. 2017).

Organisms use P to build life-sustaining molecules (e.g., cell membranes, DNA, and RNA), and compounds necessary for the electron and energy transport systems of metabolic reactions (e.g., ADP, ATP) (Kortmann and Rich 1994). Phosphates in nucleic acids and membranes have slow rates of turnover in cells, but those in the electron transport systems have very rapid turnover (Wetzel 2001). Because of the relative scarcity of phosphates, and because algae have developed numerous methods to efficiently extract phosphates from water, concentrations of dissolved phosphates are typically quite low. Therefore, total phosphorus (TP) is a common metric used to evaluate productivity potential within aquatic ecosystems. Numerous studies have addressed the relationship between TP and chlorophyll-a (Chl-a) within lake ecosystems (Vollenweider 1968, Dillon and Rigler 1974, Jones and Bachmann 1976, Canfield and Bachmann 1981, Prepas and Trew 1983, Havens and Nürnberg 2004). TP seasonal or yearly averages are often employed to classify trophic status of lakes and reservoirs (e.g., Carlson Trophic State Index) (Carlson 1977, Correll 1998). Regulators of waterbodies listed as “impaired” under the Clean Water Act (i.e., 303(d)) due to excessive nutrient pollution are required to calculate total maximum daily load (TMDL) for the constituents, such as P, that have caused the impairment. TMDL models depend on accurate and precise estimations of chemical constituents (e.g., TP) to estimate target management goals (e.g., Havens and Schelske 2001, Effler et al. 2002, Havens and Walker 2002, Brett et al. 2016). Thus, accurate TP measurements are an integral tool for freshwater scientists and lake managers.

Automated inline-ultraviolet (inline-UV) digestion for TP analysis is becoming readily available and offers streamlined production that reduces sample processing time and costs. The goal of this study is to evaluate the precision and accuracy of automated inline-UV digestion using samples from a variety of freshwater ecosystems in comparison to the “benchmark” persulfate digestion method that is recommended by previous reviews (Maher and Woo 1998). Both methods are approved “Standard Methods” (i.e., 4500-P I, 4500-P J) for the detection of TP within freshwater (American Public Health Association [APHA] 2005), and multiple studies have addressed the comparability of these methods (reviewed by Maher and Woo 1998). However, most of the comparisons involved recovery of synthetic forms of dissolved organic material (e.g., phenyl-phosphate, trimethyl-phosphate, adenosine-5-mono-phosphate disodium salt, etc.), or of water and wastewater samples with very high TP concentrations representative of highly eutrophic aquatic ecosystems (i.e., >90 µg P/L; Goossen and Kloosterboer 1978, McKelvie et al. 1989, Benson et al. 1996, APHA 2005). We analyzed samples from lakes, reservoirs, ponds, and streams of varying sizes and trophic states to test how inline-UV digestion compares to conventional autoclave digestions at lower TP concentrations representing a wider variety of freshwater ecosystems.

Materials and methods

Study site

Samples were taken from 32 sites in freshwaters from northern Idaho and eastern Washington (Table 1). Sites included streams, reservoirs, ponds, and lakes, chosen to represent a gradient of TP concentrations in freshwater ecosystems (i.e., 5–350 µg P/L).

Table 1. Study site classifications, locations, and average TP estimations with associated error for both digestion methods.

Name	Type	Latitude	Longitude	Manual digestion		Inline-UV digestion
				Mean TP (µg/L)	Standard deviation (µg/L)	Mean TP (µg/L)	Standard deviation (µg/L)
Buffalo	Lake	48.06231	−118.89899	33	2.1	22	2.9
Roosevelt	Reservoir	47.94769	−118.98659	33	2.5	15	1.2
Banks	Reservoir	47.93501	−119.03041	32	1.5	6	0.6
Rebecca	Lake	48.05531	−118.93730	141	1.5	164	8.1
North Twin	Lake	48.28915	−118.36769	72	6.7	70	2.1
South Twin	Lake	48.27314	−118.38106	35	7.6	30	1.2
Owhi	Lake	48.22753	−118.90479	104	9.0	86	7.0
Round	Lake	48.29527	−118.32225	71	2.9	64	1.0
Sanpoil	River	48.22838	−118.69953	73	1.7	92	1.0
Snake	River	46.16237	−116.92293	90	2.0	29	1.7
Rock	Lake	47.13940	−117.72504	49	7.2	33	2.9
McGinnis	Lake	48.03570	−118.89880	38	1.5	30	9.2
Sunnyside	Pond	46.72924	−117.19215	328	14.2	333	3.5
Newman	Lake	47.77293	−117.08523	22	5.3	13	0.6
Fernan	Lake	47.67021	−116.74794	42	1.0	42	0.6
Hauser	Lake	47.77044	−117.01806	65	0.6	51	1.7
Spokane	River	47.68353	−117.11181	33	2.5	44	4.4
Hayden	Lake	47.75415	−116.75491	34	0.6	45	1.2
Liberty	Lake	47.64830	−117.09170	54	13.3	42	11.5
Coeur d'Alene	Lake	47.67307	−116.78757	30	1.5	40	1.0
North Fork Clearwater	River	46.50569	−116.31662	30	2.9	34	1.5
Main Stem Clearwater	River	46.50041	−116.31809	45	3.6	44	0.6
Garfield	Pond	46.99791	−117.19166	228	2.0	271	1.5
Horddeman's	Pond	46.73871	−116.97430	361	7.0	270	26.2
Winchester	Lake	46.23436	−116.62347	177	17.3	171	3.8
Moose Creek	Reservoir	46.87442	−116.42051	85	4.9	83	1.5
Potlach	River	46.62387	−116.63339	83	6.1	95	0.6
Paradise Creek	River	46.73129	−117.08627	207	6.2	233	5.9
Palouse	River	46.89080	−117.37478	335	2.5	341	0.6
Spring Valley	Reservoir	46.78553	−116.75516	38	2.1	27	2.6
Wawawai	Reservoir	46.63473	−117.37800	100	2.3	97	4.2
Lower Granite	Reservoir	46.62903	−117.38083	87	15.3	95	1.2

Sample collection and analysis

Surface water grab samples (i.e., 500–1000 mL) were collected in triplicate in acid-washed (i.e., 3× with 1 M HCl, 5× with 18.2 MΩ water) polypropylene bottles. Bottles were triple rinsed with sample water prior to final sample collection (Jarvie et al. 2002). Bottles were submerged directly below the air–water interface and filled to minimize head space. Samples were transported on ice from the field and stored in the dark at 4 C in accordance with suggested storage limits for TP analysis (Maher and Woo 1998). All samples (n = 96) were split and analyzed using both digestion methods.

Both manual autoclave and inline-UV digestion methods employ a strong oxidant, potassium persulfate, to convert all forms of P to orthophosphate (OP). Subsequently, OP was measured using the automated ascorbic acid reduction method with a continuous flow analyzer (APHA 2005; SEAL Analytical AA3, Mequon, WI). Detection limits of both methods were determined using standard methods (i.e., minimum detection limit 2 µg P/L, top calibrant 300 µg P/L) (APHA 2005). ACS reagent-grade chemicals and ultrapure water (i.e., 18.2 MΩ) were used to prepare all reagents and analytical matrices. Samples were shaken to evenly distribute particulates, and subsamples were then immediately removed for digestion and/or analysis. Manual TP digestion used alkaline persulfate reagent (15 g sodium hydroxide, 30 g boric acid, 50 g potassium persulfate per L of water), added to sample in a 1:10 ratio (i.e., 4 mL digestion solution:40 mL sample) prior to autoclaving at 120 C for 2 h. Inline-UV digestion incorporated 0.63 mL persulfate reagent (28.5 g potassium persulfate per 1 L of water) and 1.5 mL sample within the automated flow analyzer (SEAL Analytical AA3, Mequon, WI) prior to passing through an inline-UV lamp and 90 C heat bath. Analytical sample duplicates and blanks were incorporated during sample processing and required less than 10% error to pass quality assurance/quality control (QA/QC) protocols. Samples with error larger than 10% were reanalyzed until QA/QC requirements were satisfied; results outside of control limits were excluded from analytical comparisons. Third-party reference materials were digested using both methods to test acceptability for laboratory accreditation with Washington State Department of Ecology (Complex Nutrients-WP, RTC PEI-028-2, Sigma-Aldrich RTC, Laramie, WY).

Statistical analysis

Statistical analyses were conducted with R (version 3.5.1; R Project for Statistical Computing, Vienna, Austria). Logarithmic transformations of TP concentrations for each method were plotted in comparison to a 1:1 line, and linear regression analysis was used to compare accuracy and precision between methods. Logarithmic transformations were calculated to better represent the large data spread. We also compared the absolute differences between digestion methods using the Wilcoxon signed rank test.

Results and discussion

Both methods provided acceptable results for third-party quality control reference materials (Table 2), suggesting that both methods could pass “stringent” QA/QC protocols required by government agencies for water quality monitoring accreditation. However, third-party QA/QC tests may not adequately provide evidence that both methods are comparable, or efficient at performing accurate estimations on natural waters in which phosphorus is contained in more complex molecular fractions.

Table 2. Comparison of TP recovery from third-party QC reference material.

Method	n	Mean TP (µg/L)	Standard deviation (µg/L)	Certified value (µg/L)	Acceptable limits (µg/L)
Autoclave	7	205	1.7	207 ± 3.64	179 to 234
UV	5	205	11.4

TP concentrations ranged from 16 to 338 µg P/L with manual autoclave digestion, and from 5 to 335 µg P/L for inline-UV digestion. TP concentrations were representative of common nutrient levels in surface waterbodies. Logarithmic transformations of both digestion methods were linearly correlated (p < 0.001, R² = 0.8110). For inline-UV digestion, 95% confidence intervals surrounding the best fit linear model deviated from a 1:1 comparison line (Fig. 1), clearly demonstrating performance limitations at lower TP concentrations in natural waters. For higher TP concentrations (i.e., >90 µg P/L), results for both methods converged with the 1:1 linear model, similar to previously reported comparisons (Goossen and Kloosterboer 1978, McKelvie et al. 1989, Benson et al. 1996, APHA 2005). The divergence of our data from previous method comparisons is due to the use of water samples from natural waterbodies rather than synthetic organic phosphorus mixtures and suggests that inline-UV digestion may not yield acceptable results in water samples with low TP concentrations.

Figure 1. Comparison of log-log transformations of inline-UV and autoclave TP digestion methods. Points represent mean of triplicate samples. Solid black line represents best fit linear model (p < 0.001, R² = 0.8110) and gray polygon represents 95% confidence intervals. Dotted line represents a 1:1 relationship between methods.

Comparison of absolute differences in TP concentrations demonstrates a significant difference between methods (p < 0.001, Wilcoxon signed rank test). Absolute mean difference between methods was 16 µg P/L, with 90% confidence intervals of 31 µg P/L, and maximum differences up to 120 µg P/L. Large differences between methods were observed at high and low TP concentrations, but inline-UV digestion consistently yielded lower concentrations than manual digestion at low-level TP concentrations (Fig. 1). Although 34 of the samples (n = 96) yielded <10% difference between methods, generally acceptable for QA/QC protocols on duplicate samples, 64 sample results exceeded the 10% criterium. Mean percent difference between method was 22%, with a median of 19%, and a maximum difference of 84% between methods (Fig. 2).

Figure 2. Histogram of percent difference between TP digestion methods for all samples (n = 96). Grey bar represents samples that displayed <10% difference between methods (n = 34), and white bars outlined in black represent samples that displayed >10% difference (n = 62).

TMDL TP limits and/or target reductions in lakes are frequently set below 20 µg P/L (Wagner and Corbin 2003, Fulton and Smith 2008). Conclusions about lake trophic classification would be significantly altered at the observed degree of variation detected between methods (Auer et al. 1986). For example, a ±31 µg P/L margin of error, which encompasses the 90% confidence intervals surrounding our data, could lead to misinterpretation of lake impairment, trophic status, or compliance with mandated regulatory goals, and could grossly misrepresent the biochemical forces driving lake productivity. Lake and reservoir managers should be aware of the potential errors that may be associated with switching methods or laboratories simply to achieve cost savings or decreased sample processing time using automated instruments. Even if methods and laboratories are accredited, there is a potential that switching methods could lead to misinterpretations of long-term nutrient data trends. For example, previous work demonstrated that long-term nutrient monitoring on Long Lake (Spokane, WA) may have been biased as a consequence of adjusted methodology (Brett et al. 2016). Interestingly, the method in question was also inline-UV digestion, which resulted in 4 to 20 µg P/L (∼19%) higher TP estimations compared with more robust methods (Cusimano 2003, Hallock 2012, Brett et al. 2016). Our data generally contrast with the Long Lake results and indicate that inline-UV digestions consistently yield unacceptably low TP at low concentrations compared to manual persulfate digestion. However, results from several lake and stream samples in our data set yielded higher TP estimations by inline-UV digestion (Fig. 1). Ultimately, we suggest that both studies demonstrate the inability of inline-UV digestion to accurately estimate TP concentrations at low-levels in natural waters.

We believe that differences between TP digestion methods are influenced by physical and chemical properties of the phosphorus-containing materials in natural waters. Quality control samples are typically comprised of well-mixed, dissolved organic forms of P. Natural surface waters most often contain a more complex mixture of particulate, as well as dissolved organic and inorganic P compounds. Autoclave digestion likely produces more representative concentrations as particulate and dissolved forms of P are more evenly distributed within samples prior to digestion, and the larger volumes required from the autoclave method likely decrease bias from differing forms of P, especially in particulates. A review of P analytical methods in natural waters suggests persulfate digestions by autoclave or microwave provide the highest degree of precision (Maher and Woo 1998).

In the case of inline-UV digestion, samples may sit in a quiescent condition in the sample tray prior to analysis, allowing settling of particulates prior to sample loading into the analytical stream (Fig. 3). Therefore, there is likely sufficient time for negatively (e.g., diatoms) and/or positively (e.g., cyanobacteria) buoyant particulates to spatially separate, and be excluded from analysis, as the sampler extracts only a portion of the resulting heterogeneous sample (Fig. 3b) (Lund et al. 1963, Henriksen 1970, Reynolds 1973, 1984, Reynolds and Wiseman 1982). Selective bias of a heterogeneous sample by the automated sample probe could drastically alter TP estimations, and intrasample replicate variance could remain low as the physical properties of particulates between sample cups are similar. Spatial separation of particulates is one mechanism by which inline-UV digestion may underestimate (Fig. 3b) or overestimate (Fig. 3c) TP in natural waters, as plankton cells are a primary pool of standing-crop P in freshwater ecosystems (Henriksen 1970, Gulati et al. 1992, Havens and Schelske 2001). For example, Clark et al. (2010) demonstrated that a single large-bodied zooplankton (i.e., Daphnia sp.) could contaminate a 35 mL water sample with up to 25 μg P/L. Therefore, depending on particle density and size attributes, the automated sampler could entrain more or less biomass than is naturally present in a homogeneous sample that is more representative of natural water conditions (Fig. 3). TP variation is likely further exacerbated in automated inline-UV digestion because small volumes (i.e., 1.5 mL) do not capture the complete distribution of P fractions.

Figure 3. Graphical representation of (a) homogenized sample within a sample cup, (b) possible separation of positively and negatively buoyant particles within sample cup leading to underestimation of TP, and (c) possible congregation of neutrally buoyant particles within sample cup leading to overestimation of TP concentration.

Our results do not provide absolute conclusive evidence that automated inline-UV analysis is less precise than manual digestion as variance within triplicate samples is often less than the variance observed with manual digestions (Table 1). However, our data do suggest that particulate or other component(s) within the automated UV method could lead to over- or underestimation of TP concentrations in natural samples. Autosampler extraction of small volumes of heterogeneous samples containing particulate matter is one mechanism that could result in these errors (Fig. 3).

There is also the possibility that inline-UV digestion may not provide adequate time needed for UV exposure and heat treatment to completely oxidize strong organic-P bonds in compounds found in many natural waters (Maher and Woo 1998). For example, high concentrations of organic C have been shown to interfere with persulfate oxidation of P bonds (Williams et al. 1995, Maher and Woo 1998), although high organic C levels needed to interfere with persulfate oxidation (i.e., >200 mg C/L) are not commonly observed in natural waters (Thurman 1985). However, this does not negate the possibility that chemical constituents in natural waters may result in inaccurate and imprecise estimation of TP by both methods. We suggest that further work should be conducted before accepting automated inline-UV TP digestion for analyses of natural water samples.

Inline methods show great promise for streamlining analytical workload for TP, but may require significant modifications to improve reliability and comparability with autoclave digestions. For example, longer residence times in digestion and heat-bath loops and sampler agitators might overcome the issues identified in this and other studies. Sample mixing or agitation might be used to address heterogeneous distribution of particulates. Our results suggest that current QC standards may not truly evaluate the ability of analyses to provide accurate and precise TP concentrations in natural waters. We suggest that QC standards should be comprised of more complex mixtures of compounds, similar to the mixtures found in natural waters, to truly evaluate accuracy of analyses involving TP digestions in natural waters.