10.1.1.319.2Total dissolved nitrogen analysis: comparisons between the persulfate, UV and high temperature oxidation methods451

.Marine Chemistry 69 2000 163178www.elsevier.nlrlocatermarchem

Total dissolved nitrogen analysis: comparisons between thepersulfate, UV and high temperature oxidation methods

Deborah A. Bronk a,), Michael W. Lomas b, Patricia M. Glibert b,Karyn J. Schukert a, Marta P. Sanderson a

a Department of Marine Sciences, School of Marine Programs, Uniersity of Georgia, Athens, GA 30602-2206, USAb Uniersity of Maryland, Center for Enironmental Science, Horn Point Laboratory, P.O. Box 775, Cambridge, MD 21613, USA

Received 28 August 1998; accepted 29 October 1999

Abstract

. . .We compared the persulfate PO , ultraviolet UV , and high temperature oxidation HTO methods used to analyze total .dissolved nitrogen TDN concentrations in aquatic samples to determine whether the three methods differed in terms of

.standard parameters blanks, limits of detection and linearity, and precision or in oxidation efficiency of standardcompounds and field samples of varying salinity. The TDN concentrations of several N-containing standard compounds, aswell as a humic mixture and a suite of field samples collected from the Sargasso Sea, Chesapeake Bay and an aquaculturepond were determined with the three methods. The PO method had the highest percent recoveries for the range of labile and

.refractory standard compounds tested 93"13 . The HTO method yielded recoveries of 87"14; recoveries increased to91"10 under optimized conditions. The standard UV method, with 30% H O as the oxidant, was found to be highly2 2

.variable, producing the lowest percent recoveries 71"21 ; the oxidation efficiency of the UV method increased .substantially in subsequent trials 91"12 , when the PO reagent was used in place of H O . In the field sample2 2

comparison, the PO, UV with PO reagent, and HTO method produced similar results slopes of the Model II regression lines2 .comparing them ranged from 1.00 to 1.05 with r G0.99 . The standard UV method, however, produced concentrations 5%

to 40% lower than the other methods. Analysis of the spectra emitted by the UV lamp used in this study suggests thatvariations in the UV spectra reaching the sample may have caused the reduced efficiencies. The poor recovery of somestandard compounds with each of the methods suggests that concentrations of TDN, and subsequently DON, measured in thefield with any of the methods will likely be underestimated to some degree depending on the composition of the TDN poolat that time. With careful attention to detail, however, the PO and HTO methods can provide reproducible results consistentwith each other. The standard UV method, however, was found to be highly unpredictable in practice. q 2000 ElsevierScience B.V. All rights reserved.

Keywords: total dissolved organic N concentration methods

) Corresponding author. Tel.: q1-706-542-7671; Fax: q1-706-542-5888; E-mail: [email protected]

1. Introduction

.Dissolved organic nitrogen DON is the largest .pool of fixed nitrogen N in many marine and

0304-4203r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. .PII: S0304-4203 99 00103-6

( )D.A. Bronk et al.rMarine Chemistry 69 2000 163178164

.aquatic systems Sharp, 1983 . Even in oligotrophic,open-ocean environments where dissolved inorganicN concentrations are at or near the limit of detection,concentrations of DON are commonly in the range

y1 .of 410 mg-at N l Sharp, 1983 . Despite thequantitative dominance of the DON pool in manyenvironments, it is the N fraction about which theabundance and cycling are known the least. This lackof information persists because there are consider-able methodological difficulties associated with ana-lyzing the DON pool.

Over the past decade, beginning with the work of .Suzuki and Sugimura 1985 and Sugimura and

.Suzuki 1988 , there has been a renewed interest indissolved organics. Central to this interest has beenthe need to determine accurate concentrations ofdissolved organics with high precision. Most of theresearch, however, has been directed at the measure-

.ment of dissolved organic carbon DOC , and severalintercomparison studies have been conducted to as-sess the relative efficiency of the different oxidation

protocols used to measure DOC methods reviewed.in Wangersky, 1993 . Extensive discussion at meet-

ings and in the literature has led to the generalconclusion that the UV and HTO oxidation tech-

niques provide similar results Benner and Hedges,.1993; Sharp et al., 1995 , as do PO and HTO

.Peltzer et al., 1996 , provided that instrument blanksare carefully determined and HTO columns are suffi-ciently conditioned. By comparison, DON has re-ceived little attention.

To calculate DON concentrations, accurate total .dissolved N TDN concentrations must first be ob-

tained. TDN consists of two fractions: an inorganic q.fraction, composed of ammonium NH , nitrate4

y. y.NO and nitrite NO , and an organic fraction3 2 .i.e., DON , the composition of which is unknownbut which can include amino acids, proteins, urea,and humic and fulvic acids. To measure TDN con-centrations, there are three groups of methods com-monly used in aquatic systems: persulfate oxidationPO; Menzel and Vaccaro, 1964; Sharp, 1973;

. Valderrama, 1981 ; ultraviolet oxidation UV; Arm-.strong et al., 1966; Armstrong and Tibbitts, 1968 ;

and high temperature oxidation HTO, includinghigh-temperature catalytic oxidation; Sharp, 1973;

.Suzuki and Sugimura, 1985 . Another method, Kjel-dahl digestion, which uses sulfuric acid to convert

DON to NHq, has high blanks and resulting low4precision and is therefore inappropriate for analysis

. .of seawater SW samples DElia et al., 1977 .After a TDN concentration has been measured, thecombined concentration of NHq and NOyrNOy is4 3 2subtracted from it with the residual being defined asDON. This approach is problematic because DONconcentration estimates have the combined analyticalerror and uncertainty of three analyses: TDN, NHq4and NOyrNOy. In highly eutrophic environments or3 2open-ocean deep water, concentrations of inorganicNOy and NHq may be many times the concentra-3 4tion of the DON pool. Under these circumstances,DON concentrations often have high standard devia-

.tions SD and negative DON concentrations are .occasionally calculated Hansell, 1993 .

As research begins to focus increasingly on thedynamic nature of the DON pool, accurate measuresof DON concentrations are becoming more impor-tant. There is no consensus, however, as to whichmethod is best for making these measurements.

.Walsh 1989 compared the UV method to the HTOmethod and found that, for DON compounds be-

.lieved common in SW in a SW matrix the twomethods produce similar results. In this study, wecompared all three commonly used techniques, thePO, UV and HTO, to measure TDN concentration ofseveral standard N-containing compounds and a suiteof field samples. We evaluated each method on thebasis of their blanks, limits of detection and linearity,precision, and oxidation efficiency of standard com-pounds. We also quantified the effect of salinity onoxidation efficiency. The objective of this study wasto evaluate these methods in order to make practicalrecommendations based on the environment understudy.

2. Materials and Methods

Samples run with the PO and UV oxidation meth-ods were analyzed at the University of Georgias .UGA Department of Marine Sciences. Samples runwith the HTO method were analyzed on an AntekInstruments High Temperature Combustion Analyzer .7000 series at the University of Marylands HornPoint Laboratory.

( )D.A. Bronk et al.rMarine Chemistry 69 2000 163178 165

All glassware and plasticware used in the PO andUV analyses were washed in dilute LiquiNox soap,

.rinsed with deionized water DW , soaked overnight .in 10% hydrochloric acid HCl , and rinsed with

.copious amounts of DW Barnstead NANOPureUV .All glassware was dried in a drying oven, coveredwith aluminum foil, and baked for )2 h at 5008C.Samples for the HTO method were analyzed insterile autosampler vials; all glassware used in mak-ing daily standards was acid washed with 10% HClimmediately before use and rinsed with copious

.amounts of DW MilliQ .In this study, TDN is defined as all N species

which pass through a 0.2-mm Supor filter. The DONpool is defined as TDN minus the concentrations ofNHq and NOyrNOy measured independently. We4 3 2define the final analysis of the TDN sample after POand UV oxidation as a TDNNOy analysis to differ-3entiate it from a straight NOy measurement without3prior oxidation.

2.1. Standard compound preparation

Standard N solutions, at a range of concentra-tions, were prepared in DW; the N-containing com-

pounds included potassium nitrate Fisher aP263-. .500 , urea EM aUX0065-1 , ammonium sulfate

. .Baker a0792-01 , glycine Fisher aG46-500 , .EDTA Baker a8993-01 and 4-aminoantipyrine

. Sigma aA-4382 . A humic mixture Aldrich,.HI,675.2, 0.333 g into 500 ml DW also was ana-

lyzed; concentrations of the humic standards were5.5 and 11 mg-at N ly1 as measured by CHNanalysis of the dried humics. Standard solutions werepoured into acid-washed HDPE bottles and frozenimmediately afterwards to ensure that all analyses .PO, UV and HTO were performed on identicalstandard batches.

2.2. Field samples

Field samples were collected from three sites inChesapeake Bay, one site in the Sargasso Sea andfrom a freshwater aquaculture pond. Chesapeake Baysamples were collected between 29 April and 2 May

X X1996 from North Bay 0.5 m; 39826.4 N, 7681.53 W;. X0 ppt salinity , Mid Bay 21.5 m; 38829.97 N,

X . 76824.39 W; 10 ppt salinity , and South Bay 0.5 m;X X .36857.11 N, 75859.95 W; 15 ppt salinity . The Sar-

gasso Sea water had been previously collected andstored unfrozen at Horn Point Laboratory for use inalgal media preparation. The aquaculture pond sam-ples were collected on 11 May 1996 from a linedpond at Horn Point Laboratory. All natural water

.samples were filtered -50 mm Hg through 0.2-mmSupor filters and frozen immediately in acid-washedHDPE bottles. The concentration of NHq in field4samples was measured with the manual phenol

.hypochlorite technique Parsons et al., 1984 . FieldNOy concentrations were measured with the manual3

. .spongy cadmium SpCd method Jones, 1984 .

2.3. Persulfate oxidation

Analysis of TDN by the PO method was per-formed under alkaline conditions in duplicate follow-

. ing the protocol of Valderrama 1981 . Samples 15. ml were pipetted into glass culture tubes 20=125

.mm; Pyrex, Corning a9826-20 and 2 ml of theoxidizing reagent was added. To prepare the oxidiz-

ing reagent, 45 g potassium persulfate Fisher. .aP282-500 and 27 g boric acid Fisher a0084-01

were dissolved into 315 ml 1 M sodium hydroxide,and brought up to a final volume of 900 ml withDW. Tubes were capped tightly with Teflon-lined

.screw caps Baker a0084-01 and autoclaved for 0.52 h at 1218C and 15 lbrin pressure Market Forge

. ySterilmatic autoclave . Concentrations of TDNNO3were measured with the SpCd method.

In addition to the standard PO method, we ran aset of urea standards that were boiled rather thanautoclaved to determine whether an autoclave wasnecessary. Samples were boiled in a covered pot on ahot plate for 1 h. We also ran a duplicate set of ureasamples that were flame-sealed in glass ampoules,

which were reported to be superior over test tubes J..Bauer, personal communication . Pyrex ampoules

.were precombusted 5008C for 2 h prior to use.After the sample was added, the ampoule was sealedwith a torch.

2.4. UV oxidation

Analysis of TDN by UV oxidation was performedin duplicate using a UV oxidation machine manufac-


tured in the UGA machine shop following the design .of Armstrong and Tibbitts 1968 . The machine was

equipped with a 1200-W mercury vapor lamp . .Hanovia a189A0100 . Oxidation tubes 25 ml were

.manufactured with Amerseil quartz aCGQ-0800-68in the UGA glass shop.

.Samples 20 ml were pipetted into precombustedquartz tubes and 200 ml 30% hydrogen peroxide .H O ; Fisher aH323-500 was added; the quartz2 2tubes were later etched with a line at the 20 ml markto eliminate the pipetting step as a possible source ofcontamination. Tubes were stoppered with 20-mm

.Teflon-lined silicon septa Wheaton a224173 andsecured with aluminum crimp tops NSC aC4020-

.5A . Samples were oxidized in the UV chamber for24 h prior to TDNNOy analysis with the manual3SpCd method. Field samples were corrected for lossof linearity at concentrations over 100 mg-at N ly1.

Though we had been using the UV oxidationmethod for several years with high oxidation effi-

. qciencies G94% of cAMP and NH , we began4measuring low oxidation efficiencies during thismethod comparison. To check the availability ofsufficient oxidant, we ran a series of standards withthe PO oxidation reagent in place of H O . Analo-2 2gous to the PO procedure, 2 ml of PO reagent wasadded to 15 ml samples in quartz tubes. The sampleswere then treated as standard UV samples as de-scribed above.

2.5. Nitrate analysis with spongy cadmium

Following PO and UV oxidation, TDNNOy3samples were analyzed manually by the SpCd method .Jones, 1984 . The SpCd was prepared by placing

. .acid-washed HCl zinc sticks Baker a4274-01into glass culture tubes containing a 20% solution of

.cadmium sulfate Fisher aC19-500 for 812 h. TheSpCd aggregates that accumulated were scraped offthe zinc sticks and stored under DW; SpCd can beused repeatedly, but must be regenerated prior toeach NOy analysis by decanting the DW, covering3the cadmium with 6 N HCl, then immediately rinsing

the cadmium with copious amounts of DW until.pH)5 .

To measure TDNNOy concentrations, duplicate35 ml aliquots from each of the PO or UV oxidized

samples were analyzed in plastic 15 ml Corningcentrifuge tubes to which 1 ml of ammonium chlo-

.ride solution 4.7% , and 0.5 to 0.6 g SpCd wasadded. We increased the weight of SpCd from 0.2 g

.per 5 ml sample Jones, 1984 because the largeraddition decreased the SD of the measurements. Thecentrifuge tubes were capped and shaken at 200 rpmfor 1.5 h in the dark; tubes were oriented parallel tothe direction of shaking to ensure maximum contactbetween SpCd and sample.

After shaking, 5 ml of sample from each tube waspipetted out, held in the pipette tip while the SpCdand excess liquid was discarded, returned to the tube,and 250 ml of color reagent was added. The colorreagent, which we found must be prepared fresh

weekly, consisted of 5 g sulfanilimide Fisher. .a04525-100 , 0.5 g N- 1-napthyl ethylenediamine

.dihydrochloride Aldrich a22,248-8 dissolved in 50ml of 85% phosphoric acid, and diluted with DW toa final volume of 500 ml. Tubes were vortexed, leftundisturbed for 10 min in the dark to allow for fullcolor development, and then the absorbance wasmeasured spectrophotometrically at 540 nm. TheTDNNOy content of the Chesapeake Bay, aquacul-3ture pond and Sargasso Sea samples were analyzedwith both the SpCd method as described above .Jones, 1984 and with a standard autoanalyzer

method Technicon a158-71WrA; Whitledge et al.,.1981 to determine whether these two methods

yielded similar results.

2.6. High temperature oxidation

All HTO analyses were performed on an AntekInstruments High Temperature Combustion Analyzer .Model a7000B-N modified for the analysis of SW

with a SW specific pyrolysis tube Antek Product. .a71136 . Samples 7 ml were injected with a hori-

zontal autosampler Central Instruments aGC-311-H.as modified by Antek Instruments fitted with a

10-ml viscous syringe Central Instruments a501-.092 . The autosampler and sample combustionr

ozone generator used Zero Grade Ultra High Purity,.99.8% O and -3.5 ppm H O oxygen. After 1502 2

injections, the septa in both the autosampler Central. Instruments a501976 and in the pyrolysis tube An-

.tek a69091 were replaced.


Table 1Settings for the Antek Instruments High Temperature Combustion

.Analyzer Model a7000B-N used during the comparison ofstandard compounds in this studyVariable Setting

Primary autosampler gas flow 78 psiPrimary analysis gas flow 20 psiSecondary analysis gas flow 5 psiSample purge pressure 4.5 psiSystem back pressure 1.1 psiPyrolosis tube rotometer 1.8Ozone generator rotometer 1.31.4Photomultiplier tube voltage 775 VPyrolosis temperature 10008CSample purge time 15 sInjection dwell time 10 s

Prior to analysis, all standards and samples werepipetted into 1.9 ml autosampler vials Wheaton

.a225175 with 10 ml of 3 N HCl and capped .Wheaton a22411-01 . This volume was sufficientto allow for three analytical replicates per vial with aflush time of 15 s at a flow rate of 1 droprs from the

.waste tube Table 1 . The acid allows carbonates andbicarbonates to be removed as CO , thus reducing2the possibility of salt plugs forming in the ceramicinsert. An additional set of standards was run afterthe sample run each day to assess instrumental drift.

2.7. Measurement of standard parameters

Potassium nitrate in DW or SW was used as thestandard for analyses of all the methods that em-ployed oxidation, as per the recommendation of

.Hopkinson et al. 1993 . Blanks were estimated intwo ways. First, blank samples were run in duplicateor triplicate as part of every standard curve. Theresulting absorbance or counts of the blanks weremultiplied by the slope of the standard curve toobtain the blank concentration. Second, blanks wereestimated by taking the y-intercept of the standardModel I regression line. To calculate slope andy-intercept values for each day of analysis, Model I

linear regressions of the raw absorbance PO and. .UV or fluorescent count HTO data were used.

These data were then used to convert the sampleabsorbances or fluorescent counts to final TDN con-centrations. Standards and samples were run with the

same salinity for all methods; one exception was theanalysis of field samples by the HTO method wherea NOy standard in DW was used.3

For all standard compounds analyzed by PO andUV, a total of six replicate vials for each substratewere run over 4 to 6 days. Within-day variabilitywas measured by running two analytical replicatesfrom a single bottle on the same day. For HTO,within-day variability was measured on three repli-cate vials, with each being the average concentrationfrom three replicate injections from the same vial.

Field samples were also analyzed by all methods.Since the actual TDN concentrations of field samplesare unknown, the efficiency of each method wasevaluated on the basis of how measured concentra-tions compared with one another, and whether signif-icant differences in mean values were observed.Concentrations of DON were calculated by subtract-ing the concentration of NHq and NO-rNOy from4 3 2the TDN concentration. The SD of the DON concen-tration was calculated by propagating the error withthe following equation:

1r22 2 2q y yS s S qS qS .DON TDN NH NO r NO4 3 2

where S2 is the variance of the three measurements .Bevington, 1969 . All regressions comparing thevarious methods with each other are reported usingModel II geometric mean regression parameters .Sokal and Rohlf, 1995 .

The effect of salinity on the oxidation efficiencyy q of NO , NH and urea standards 25 and 75 mmol3 4

y1 .N l was examined for all analytical methods.Standards were prepared in DW and Sargasso Sea

. water HTO or artificial SW PO, UV and UV.wrPOR and analyzed as previously described.

2.8. Limit of linearity

To test the limit of linearity, urea standards inDW were prepared in concentrations from 1 to 200mg-at N ly1 and analyzed by the PO and UVmethods. In the case of HTO, there are six gainsettings on the Antek. The same gain setting used inthe analysis of the standard N substrates was used inthe limit of linearity test. Accordingly, the limit oflinearity was measured from 0 to 100 mg-at N ly1,which was the maximum range at this gain setting.


3. Results

We provide data on each method for standardparameters blanks, limits of detection and linearity,.and precision , as well as the oxidation efficiency of

standard compounds. We also demonstrate the effectof sample salinity on each method. Lastly, we com-pare the methods with respect to the analyses ofnatural water samples taken from an estuary, theopen ocean and an estuarine aquaculture pond.

3.1. Standard parameters

Blanks were measured directly as part of thestandard curve or were estimated by taking the y-in-

tercept of the standard Model I regression line Table.2 . The second y-intercept method is a more robust

measure because it is less susceptible to contamina-tion of a single blank sample; we use the y-interceptblanks throughout the remainder of the paper. The

UV oxidation method had the lowest blanks 1.01"y1 .0.51 mg-at N l and the PO method had the

.highest 1.54"0.49; Table 2 . We estimate the limitof detection as three times the SD of the blank .Taylor, 1987 . Using this measure, the PO method

y1 .had the lowest limit of detection 1.46 mg-at N land the UV wrPOR method had the highest limit of

y1 .detection 1.89 mg-at N l .We ran urea standard curves to determine whether

the absorbance or counts used to calculate measuredconcentrations increased linearly with actual concen-tration for each method. The PO and UV wrPORmethods produced a linear response with urea con-

y1 .centrations from 1 to 200 mg-at N l Fig. 1A . We

also found that boiling PO samples in either testtubes or sealed ampoules produced comparable ab-sorbances to the standard autoclave method, and thatthe resulting curves were linear up to 200 mg-at Ny1 .l Fig. 2 .

The standard UV method was only linear up to 50y1 .mg-at N l Fig. 1B . We ran curves with two

different UV lamps the one we had been using inthis comparison and a new lamp installed to deter-mine whether it would increase the linearity andoxidation efficiency of the method; both lamps re-sulted in comparable curves. We had previouslydetermined that the oxidation efficiency of the UVmethod increased when we increased the H O addi-2 2

tion from the standard 50 ml to 200 ml Fig. 1B;.Bronk, unpublished data . When we added 400 and

800 ml to 100 and 200 mg-at N ly1 urea samples,however, the absorbances actually decreased Fig.

.1B .In the case of the HTO method, the limit of

linearity can be controlled by the gain setting on themachine. Here we only present data up to 100 mg-atN ly1 because this was the linear range of themachine at the gain setting we chose to use through-

.out the standard compound comparison Fig. 1C .Higher concentrations will also yield a linear re-sponse if the gain setting is dropped as was done inthe analysis of field samples. The upper limit ofquantitation at the lowest gain setting is near 1 M.

Precision, the degree of agreement between re-peated measurements, is evaluated based on the SD

.and coefficient of variation CV at a given concen-tration level. The UV method had the lowest CV atthe lower concentration, both at the within-day and

Table 2 .Method blanks as estimated from blanks measured directly as part of the standard curve direct blanks and by the y-intercept of the Model I

y y1 .regression line of NO standards in DW from 0 to 50 or 100 mg-at N l . Each estimate n was determined on a different day. The mean3direct blanks were calculated from the individual blanks, run in duplicate or triplicate, as part of the standard curve used to determine the

.y-intercept data presented. The limit of detection LOD, Taylor, 1987 is defined as three times the SD of the blankMethod Mean Direct Mean y-intercept LOD n

y1 .direct blanks y-intercept SD mg-at N ly1 y1 . .blanks SD mg-at N l mg-at N l

PO 1.87 0.85 1.54 0.49 1.46 8UV 1.21 0.76 1.01 0.51 1.54 8UV wrPOR 1.46 0.52 1.48 0.63 1.89 6HTO 0.85 0.39 1.22 0.61 1.83 5


Fig. 1. The limit of linearity measured with standard urea solu- . 2 .tions in DW for the A UV wrPOR v, r s0.999 and PO

2 . ., r s0.999 methods; B the standard UV method with . . hydrogen peroxide H O additions of 50 ml , 200 ml v,2 2

2 y1 . .r up to 100 mg-at N l equals 0.997 , 400 ml B , and 800 ml . . 2 .I ; and C the HTO method , r s0.999 . The 400 and 800ml H O additions were made to 100 and 200 mg-at N ly12 2samples; the symbols are offset on the graph to provide a clearerdepiction of error bars. Error bars represent SD; when no errorbars are visible, SD were smaller than the size of the symbols.

between-day level, as well as the within-day level at .the higher concentration Table 3 .

3.2. Oxidation efficiency of standard solutions

Standard solutions of potassium nitrate, urea, am-monium sulfate, glycine, EDTA, antipyrine and hu-

mics at a number of different concentrations wereoxidized and compared to the calculated concentra-

.tion Table 4 . All of the methods had reducedoxidation efficiencies with at least one of the stan-

.dard compounds tested Table 4 .The PO method had the highest overall recovery

.of the compounds tested 93"13%; Table 4 ; POproduced nearly quantitative recovery of urea, NHq4and glycine in DW; EDTA and antipyrine were lessefficiently oxidized. The oxidation efficiency of ureawas lower in SW than in DW for all methods. Humic

.acids were oxidized with high efficiency )94%though the efficiency did decrease at the higherconcentration tested.

The standard UV method had the lowest overallrecovery of all the methods tested 71"21%; Table

.4 . Consistent with the PO method, urea oxidationefficiencies were lower in SW than in DW. Humicacids were oxidized with greater than 92% efficiency .Table 4 . The standard UV method showed twotrends that were not observed in the other methods.First, the oxidation efficiency of NHq was less than4for glycine; paradoxically, glycine is believed to be

q broken down to NH , which is then oxidized Manny4.et al., 1971 . Second, there was a consistent drop in

the oxidation efficiency when the concentration ofthe standard increased.

Fig. 2. Limit of linearity measured with standard urea solutions inDW for the PO method where boiling for 1 h was used rather than

autoclaving. Samples were boiled contained in test tubes v,2 . 2 .r s0.993 or precombusted glass ampoules , r s0.999 .

Error bars represent SD; when no error bars are visible, SD weresmaller than the size of the symbols.


Table 3 .Precision of the four methods, both within day and between day i.e., for all measurements made . The mean SD and the coefficient of

w x .variation CVs SDrmean =100 for the PO, UV, UV wrPOR, and the HTO methods. All standard compounds shown in Table 4 thaty1 y1 . y1 y1 .were run at 5 mg-at N l including 5.5 mg-at N l humic acid and at 40 mg-at N l including 50 mg-at N l urea were used in the

calculationsy1 y1Method 5 mg-at N l 40 mg-at N l

Within day Between day Within day Between day

Mean Mean Mean MeanSD CV SD CV SD CV SD CV

PO 0.37 7.3 0.44 8.8 1.73 5.5 2.78 6.9UV 0.17 3.3 0.27 5.4 0.68 1.9 1.82 4.6UV wrPOR 0.21 4.0 0.33 6.6 0.90 2.5 1.28 3.2HTO 0.46 9.0 0.63 12.6 1.29 2.9 1.60 4.0

The oxidation efficiency of the standard UVmethod was substantially improved by using the PO

.reagent as the oxidant Table 4 . As with the POmethod, the oxidation efficiencies were higher in

.DW than in SW Table 4 . The UV wrPOR methodhad higher oxidation efficiencies than PO for themore refractory compounds such as antipyrine. Un-like the standard UV method, oxidation efficienciesdid not consistently decrease at higher concentra-tions.

As was found with the PO and UV wrPORmethods, the HTO method oxidized urea and NHq,4in DW nearly quantitatively; oxidation efficiencies

.decreased slightly in SW Table 4 . When the size ofthe sample injected was decreased and the pyrolysistube flow was decreased, oxidation efficiencies in-

creased for some compounds such as glycine 79%.to 96% recovery, Table 4 . Of the methods tested,

HTO produced the lowest recoveries for humic com-pounds though it had comparatively high recoveriesfor the most refractory compound tested, antipyrine.

3.3. Salt effect

When NOy, NHq and urea standard concentra-3 4tions based on calibration standards in the same salt.matrix were compared between DW and SW, inter-

nally consistent recoveries were obtained data not.shown . If, however, SW samples were calculated

based on calibration standards in a DW matrix, asmight be done with samples from an estuarine tran-sect of increasing salinity, the PO, UV and UVwrPOR methods tended to overestimate the ex-

pected concentration by an average 12%. In contrast,the HTO method slightly underestimated the actualconcentration. Although in this study none of thesedifferences were significant, variations in the salinitybetween samples and calibration standards should beconsidered during analysis.

3.4. Comparison of field samples

In general, the PO, UV wrPOR and HTO meth-ods produced comparable TDN concentration esti-

.mates for field samples Table 5 . Based on our salteffect data, we note that the use of a DW standardcurve may have resulted in a small underestimate ofthe South Bay and Sargasso Sea water samplesdetermined by the HTO method. The UV methodproduced the lowest concentrations for each field

sample tested 5% to 40% lower than the other.methods . The mean CVs were lowest for the UV

.wrPOR samples 1.8"1.0 and highest for the UV .method 4.3"4.2 .

Model II regressions comparing the four methodsyielded the following equations:

POs3.54q 1.13=UV r 2 s0.994 .POsy0.75q 1.05=UV with POR r 2 s0.986 .POsy0.99q 1.05=HTO r 2 s0.996 .HTOs4.34q 1.08=UV r 2 s0.993 .HTOs0.24q 1.01=UV with POR r 2 s0.995 .UVsy5.24q 0.95=UV with POR r 2 s0.991 .


Table 4 .Percent recoveries of compounds in DW or SW obtained in this study I compared with published results. Data is not shown for the

potassium nitrate standard; recoveries of potassium nitrate were 100.0"1.6, 100.2"0.5, 100.4"0.6 and 100.4"3.7 for the PO, UV, UVwrPOR and HTO methods, respectively. HTO values in parentheses were obtained when sample volume was decreased to 5 ml and theflow was set to 0.3Compound Matrix STD % Recovery Reference

.mM PO UV UV wrPOR HTOUrea DW 2 100.0 Nydahl, 1978

DW 10 100.0 Collos and Mornet, 1993SW 10 76.0 Collos and Mornet, 1993SW 10 83.3 100.2 Walsh, 1989SW 40 72.5 102.1 Walsh, 1989DW 50 84.0 Le Poupon et al., 1997DW 5 98.7 101.1 81.1 96.7 This studyDW 25 105.6 96.2 96.7 96.3 This studyDW 75 100.1 75.7 97.0 97.3 This studySW 25 96.8 76.5 75.7 96.8 This study

aSW 75 87.5 63.6 75.5 87.0 This study

Ammonium DW 2 99.7100.1 Nydahl, 1978chloride or SW 40 99.5 100.2 Walsh, 1989

.sulfate DW 5 95.6 78.0 95.7 94.1 This studyDW 25 99.4 35.0 99.5 93.9 This studyDW 75 103.6 28.5 109.0 98.5 This studySW 25 104.2 67.4 99.5 93.8 This study

aSW 75 98.8 49.5 94.9 92.1 This study

Glycine DW 2 92.993.7 Nydahl, 1978DW 10 100.0 Collos and Mornet, 1993SW 10 63.0 Collos and Mornet, 1993SW 20 95.5 99.6 Walsh, 1989

.DW 5 98.5 84.6 96.9 93.7 104.3 This study .DW 40 97.2 55.4 95.2 78.6 95.6 This study

EDTA DW 2 99.1 Nydahl, 1978SW 10 102.9 101.6 Walsh, 1989SW 40 95.7 100.5 Walsh, 1989DW 50 96.6 Le Poupon et al., 1997

.DW 5 87.1 87.9 92.5 93.0 103.3 This study .DW 20 88.6 76.5 95.0 57.9 86.6 This study

Antipyrine DW 2 45.0 Nydahl, 1978SW 10 55.4 94.2 Walsh, 1989SW 40 50.9 101.1 Walsh, 1989DW 5 52.1 61.0 64.0 88.3 This studyDW 40 68.1 46.6 72.4 69.8 This study

Humic acid SW 20 8598 Koike and Tupas, 1993 .DW 5.5 99.7 96.4 98.1 62.1 85.9 This study .DW 11.0 94.8 91.6 104.8 59.9 63.3 This study

Mean"SD 93.1"13.4 70.6"21.4 91.3"12.2 87.0"14.2 This study only .for all 91.3"10.4

. . .Walsh 1989 used a 1200-W mercury lamp for 1822 h UV or an Antek 700 HTO with standards made to low nitrogen surface SW. .Collos and Mornet 1993 used UV oxidation after successive acid and alkaline conditions with an automated system. Le Poupon et al.

. .1997 used an Antek 7000. Nydahl 1978 oxidized at 1208C for 15 min in a pressure cooker.aSamples were 100 mM rather than 75 mM.


Table 5Concentrations and SD of TDN and DON from five fields sites determined with the PO, UV, UV wrPOR and HTO methods.

y1 .Concentrations are shown in mg-at N 1 . CVs STDrmean =100Chesapeake Chesapeake Chesapeake Sargasso sea Aquaculture CVBay north Bay mid Bay south pond Mean STD

TDN concentrationPO 101.7"2.3 98.5"3.4 18.3"0.7 8.2"0.7 129.6"1.6 3.9 2.8UV 91.8"1.4 79.3"1.0 11.3"1.0 6.0"0.5 110.9"1.2 4.3 4.2UV wrPOR 107.1"1.3 93.1"0.7 18.2"0.5 7.1"0.2 116.4"1.7 1.8 1.0HTO 102.0"2.1 97.0"1.5 17.9"1.8 8.1"0.4 120.4"2.2 4.1 3.6DIN concentration

qNH 2.30"0.11 26.21"0.15 1.81"0.09 1.30"0.05 18.64"0.184yNO 86.3"0.2 41.3"0.2 1.3"0.0 0.3"0.0 24.4"0.13

DON concentrationPO 13.1"2.3 31.0"3.4 15.2"0.7 6.5"0.7 86.6"1.6 9.2 6.1UV 3.2"1.4 11.8"1.0 8.2"1.0 4.4"0.5 67.9"1.2 15.7 16.2UV wrPOR 18.5"1.3 25.6"0.7 15.1"0.5 5.4"0.2 73.4"1.7 3.9 1.8HTO 13.4"2.1 29.5"1.5 14.8"1.8 6.4"0.4 77.4"2.2 8.4 5.4

When inorganic N concentrations were subtractedto calculate DON concentrations, the agreement be-tween methods decreased substantially, because re-moving inorganic N, particularly NOy, which does3not need to be oxidized, magnified differences in the

.oxidation efficiency of the four methods Table 5 .Overall, the PO method produced the highest con-centrations. The standard UV technique came in adistant fourth with significantly lower concentrationsmeasured on all samples analyzed. The mean CVsfor DON measurements increased by a factor of twoto four over those calculated for TDN.

When we compared the TDNNOy concentra-3tions of the field samples, the manual SpCd methodyielded comparable results to the standard autoana-

lyzer method for all field samples tested Model II2 .regression yielded an r s0.999; data not shown .

4. Discussion

Here we examine how the oxidation efficiency ofeach method varies with different standard com-pounds and with the analysis of field samples. Inaddition, we describe advantages and disadvantagesand troubleshooting advice for each method, theset-up and operation of the HTO instrument, thepotential salt effects associated with the different

methods, and the use of SpCd compared with theautoanalyzer for TDNNOy determinations. We3conclude with specific recommendations for differ-ent environments and situations.

4.1. Comparison of standard solutions and fieldsamples

The PO, UV wrPOR and HTO methods oxidizedmost N compounds with high efficiency, averaging93%, 91% and 87%, respectively; the UV method

had a much lower oxidation efficiency of 71% Ta-.ble 4 . Other more refractory compounds such as

antipyrine were not oxidized as readily; therefore,their detection is limited by the oxidation efficiencyof the method used. Urea in DW was more com-pletely oxidized than urea in SW with all the meth-ods; these data are consistent with the UV study of

.Collos and Mornet 1993 . In contrast, with the UVmethod, the oxidation efficiency for NHq was higher4

.in SW than in DW. Manny et al. 1971 also ob-served higher oxidation efficiencies of NHq in SW4with a similar UV method, and they attributed thedifference in efficiency to a pH effect; at acidic pHs,NHq is not oxidized efficiently.4

In general, the PO, UV wrPOR and HTO meth-ods gave comparable results for TDN in our natural

.water sample comparison Table 5 ; differences be-tween the methods, however, were magnified when


DON concentrations were calculated. Our data illus-trate a common problem in calculating DON concen-trations the inflation of the CV due to the multi-

.ple analyses required discussed in Hansell, 1993 .This is especially problematic with samples that havehigh concentrations of dissolved inorganic N, such asthe North Chesapeake Bay samples, where DONcomprises a relatively small percentage of the sam-

.ple Table 5 . The standard UV method produced thelowest concentrations of all field samples Table 5,

.see discussion below . The use of the PO reagent, inplace of H O , increased the concentrations consid-2 2erably up to 63% in the case of the South

.Chesapeake Bay sample Table 5 .Overall, the standard UV oxidation method had

the lowest oxidation efficiencies for standard com- .pounds and field samples Tables 4 and 5 . Our

results, however, are similar to those obtained by .Collos and Mornet 1993 who used an automated

UV technique; they recovered 52% to 100% of the Nin 14 standard organic compounds dissolved in DWand 37% to 76% when the compound was dissolvedin SW. Recovery of many of the compounds wasimproved by oxidizing the samples under both alka-

.line and acidic conditions Collos and Mornet, 1993 .Our results indicate that the oxidation efficiency ofthe UV method is also limited at high substrateconcentrations as seen by the decrease in percentrecovery at the higher concentrations measured. Thehigher recoveries at lower concentrations suggeststhat diluting the sample prior to oxidizing it could

.yield higher recoveries Henricksen, 1970 .We note that we used the UV technique with

good success with 30% H O as the oxidant for over2 26 years. Ammonium and cAMP standards were rou-tinely run with field samples and were found to be

oxidized with G94% efficiency Bronk, unpub-.lished data; Bronk and Ward, 1999 . During this

time, however, we occasionally have noted signifi-cant variability in the method which we believe wasdue to variations in the mercury vapor lamps used.For example, previously 18 h was determined to besufficient to effect an oxidation of NHq and cAMP4

.of G94% Bronk and Ward, 1999 . An identicalmachine, manufactured with identical parts same

.company and part numbers was found to require 24h for complete oxidation; both machines used 1200-

.W mercury vapor lamps from Hanovia a189A0100 .

Variations in lamps, both between different lampsor within the same lamp as it ages, is the most likelyexplanation for the variability in oxidation efficiencywe observed. We have tried several permutations forincreasing the oxidation efficiency with any given

lamp: increasing both the amount of oxidant dis-.cussed below and time of oxidation, as well as the

use of several different bottles of H O None of the2 2.variations were found to improve the oxidation effi-ciency. Though pH has caused problems in the past,we were unable to increase the oxidation efficiencyby altering the pH at the various stages in the

analysis in this comparison discussed in detail in.Solorzano and Sharp, 1980 . Another possible source

of error is conversion of organic N to NHq; Walsh4 . q y1989 routinely analyzed for both NH and NO at4 3the end of the oxidation. In our study, we did not seeappreciable production of NHq with the standard4compounds tested; with the NHq samples, however,4N that was not converted to NOy could still be3quantitatively measured as NHq.4

Temperature is another variable that could affectoxidation efficiency as well as a variable that couldchange between lamps and as a lamp ages. In the UVmachine used in this study, and in our past work .Bronk and Ward, 1999 , quartz tubes were placedon one of three levels top, middle or bottom.Tubes at the various levels of the machine weretherefore exposed to a temperature gradient withinthe machine. Average temperatures of triplicate wa-

ter samples, equally spaced on each level in the.12:00, 4:00 and 8:00 positions were 70"08C, 56"

18C and 41"18C for the top, middle and bottomlevels, respectively. Though theoretically tempera-ture could affect oxidation efficiencies, we did notsee a significant difference in oxidation efficiency in

q .the compounds tested NH , cAMP and glycine4with our present UV machine. In the past, our UVmachine was configured in a hood such that coolingwas not as effective and temperatures on the top,middle and bottom levels were ;948C, 818C and648C, respectively. We subsequently increased theventilation to prevent the tops of our sample tubesfrom occasionally coming off. When we comparedstandard samples oxidized under the two ventingconfigurations, we did not see a significant differ-ence in oxidation efficiency. Oxidation efficienciesof samples on different levels within a UV machine


were also measured during time-course experiments .18, 24 and 40 h using a UV machine at the

University of California at Santa Cruz Bronk and.Ward, unpublished data . The temperature of sam-

ples within that machine varied from 958C to 658Cfrom the top to the bottom levels, respectively. Again,no significant differences were observed in oxidationefficiencies between the different levels with the

qstandard compounds tested NH , cAMP and4.glycine .

Lastly we analyzed the spectrum emitted by theUV lamp itself. We used an Optronics spectrora-diometer Model 730A and a fiber optic probe with aTeflon diffuser to scan the emission of our UV lampwithin the UV range. We compare the measuredspectrum with the theoretical emission spectrum ofmercury and to the expected spectrum supplied by

.the manufacturer of the lamp Fig. 3 . To facilitatethe comparison of the spectra, we standardized theemission intensities at the different wavelengths topeaks at 300, 313 and 313 nm for the theoretical,

.expected and measured spectra, respectively Fig. 3 .Note that the largest peak in the theoretical mercuryspectrum, at 255 nm, does not appear as a large peakin the spectra of the UV lamps manufactured byHanovia and most commonly used for organic matter

.analysis Golimowski and Golimowska, 1996 . Thespectrum which we measured from our UV lamp hadthe expected peaks at 255 and 313 nm but lacked anumber of other expected peaks within the 260 to310 nm range, as well as a large peak at 336 nm.Although we can document the lack of some peaks,it is unclear why they are absent. Possible causes arethat some wavelengths are being absorbed by thequartz of the lamp or the quartz housing in ourmachine or that the peak is missing due to someinterference.

Though we cannot say conclusively what hascaused the decrease in the UV oxidation efficiency,the observation that the efficiency dropped dramati-cally when a new lamp was installed indicates thatsome variation in the lamp is the likely cause. Wehave used a total of seven lamps over the past 6years. As noted above, the first five produced high

.oxidation efficiencies consistent with Walsh 1989 .At the start of this method comparison, we installeda new lamp and observed reduced oxidation efficien-cies; unfortunately, the first of the two lamps was

.Fig. 3. Emission spectra for mercury in the low UV region: A .the theoretical spectrum of mercury vapor; B the expected

spectral output of our UV lamp according to the manufacturer; .and C the emission spectrum directly measured from our UV

lamp. To facilitate the comparison of the spectra, we standardizedthe emission intensities at the different wavelengths to peaks at300, 313 and 313 nm, expressed as a percentage, for the theoreti-cal, expected and measured spectra, respectively.

returned so we are unable to measure its spectrum.The last of these seven lamps also produced reduced

.oxidation efficiencies shown in Table 4 . The varia-tions between the expected and measured spectraloutput of the lamp is one potential reason.

4.2. Adantages and disadantages of each methodand troubleshooting adice

We describe the pros and cons of each method inreference to: blanks, ease of analysis, required sam-ple volume, instrumentation costs and number of

samples per run. We note that outliers )20% dif-.ference are occasionally observed with all methods


tested; we therefore suggest sample analysis be donein triplicate if possible.

The PO method has the disadvantage of highcontamination potential due to the large oxidizing

.reagent addition 2 mlr15 ml sample . We foundthat different batches of potassium persulfate yieldedvery different N blanks; for example, prior to thismethod comparison, a new batch of PO reagent hada blank of 7.81 mg-at N ly1. If high blanks areobtained, recrystallization of the potassium persulfateonce or twice can be done to exclude N contami-

nants in the crystals, thus lowering the blanks 0.8y1 .mmol N l ; this process, however, is time consum-

ing. The PO method is a simple method to doanalytically, requiring little more than competentpipetting. The required sample volume generallyranges from 5 to 20 ml depending on the volumerequired for the subsequent TDNNOy analysis. Of3all the methods tested, the PO technique has thelowest instrumentation cost. We found simply boil-ing samples in a pot resulted in comparable oxida-tion efficiencies to autoclaving. The number of sam-ples that can be run simultaneously is limited onlyby the size of the autoclave or pot.

The use of sealed ampoules is a recommendedalternative to PO in glass test tubes. In the past wehave also used Teflon and HDPE bottles. Teflonbottles are expensive, bottles are permanently mis-shapen during autoclaving, and the material getspliable when heated which often results in capsloosening when autoclaved. Bottles made of HDPEmust be autoclaved prior to first use as ;10% ofthem explode or leak during autoclaving. Even if thebottles survive the initial autoclaving, it is not un-common to lose ;3% to 5% per batch. Sealedampoules remain intact throughout autoclaving andboiling but the sealing step does add time in process-ing.

Ease of analysis and required sample volume arethe same for the UV methods as for the PO method.The advantage of the UV technique is the potentialfor low blanks; mean blanks for the other methodswere 21% to 53% higher. The small addition ofH O is far superior in this respect to the use of the2 2PO reagent. Based on the drastic changes in oxida-tion efficiencies observed over time, we stronglyrecommend that a NOy and an NHq, urea, or some3 4other organic standard be included in every UV run

to monitor the oxidation efficiency. If the oxidationefficiency starts to decline, we found that the POreagent can be substituted for H O with excellent2 2

results. Another study using a modified UV 500-W.mercury lamp method also found a potassium per-

sulfate reagent to be superior to H O in oxidizing2 2 .organic N Pizzicannella et al., 1996 .

Disadvantages of the UV methods include thecost and effort of building a UV machine, and the

.long oxidation time 18 to 24 h required to processthe limited number of samples that can be accommo-dated by a given machine; automation schemes havebeen proposed that could potentially circumvent this

problem Collos and Mornet, 1993; Le Poupon et al.,.1997 . The greatest disadvantage of the method,

however, is the inconsistency. We found inconsisten-cies with the UV method performed on different UVmachines and within the same machine with differ-ent UV lamps.

Higher recoveries were found with a 200-ml addi-tion of H O than with the more commonly used2 2

.50-ml addition Fig. 1 . Oddly, when we used an-other UV machine built with the same specifications,the 50 mlr20 ml sample addition was determined to

be optimal Bronk, unpublished data; Bronk and.Ward, 1999 . Based on our limited trials, we recom-

mend that H O additions of greater than 200 ml not2 2be used because our trials with 400 and 800 mlH O additions resulted in a decrease in absorbances2 2 .Fig. 1 . Another difference between our presentmachine and a machine used in earlier work was theoxidation time required for complete oxidation ofstandard compounds. We therefore recommend thateach analyst experimentally determine the optimumoxidant addition and oxidation time for their ma-chine initially and after each lamp change.

With the HTO method, samples are combusted athigh temperatures in the presence of pure oxygen.

The HTO method has a relatively low blank 1.22".0.61 ; this blank is similar to the 1.5 to 3.0 mg-at N

y1 .l Hansell et al. 1993 report for multiple injec-tions of MilliQ water in their HTO instrument andconsiderably lower than the 13.1 mg-at N ly1 re-ported for a modified Antek 7000N by Frankovich

.and Jones 1998 . Though chemically simple, HTOis an exacting method to achieve reproducible resultsfrom day to day. The flow rate of oxygen gas, boththrough the ozone generator and the pyrolysis tube,


are critical to obtaining precise and accurate mea-surements. The gas flow through the pyrolysis tubeneeds to be optimized to ensure that complete com-bustion of the sample is achieved before the finalsample is flushed from the combustion furnace. Thegas flow through the ozone generator is also impor-tant; at excessively high rates the sample will beflushed from the reaction chamber before the NOU2destabilizes resulting in fluorescence. In this analy-

sis, the flow rate as determined by the back pressure.within the system and all other parameters were

held constant while the ozone flow rate was varied.As the flow rate of the ozone and pyrolysis tube areindicated by rotometers in the standard instrumentpackage, it is difficult to quantify the flow rate, andeven more difficult to ensure that it is constant fromday to day. This variation in ozone flow rate caninfluence the accuracy of the results. Finally, elec-tronic noise in the instrument can lead to erroneously

.high values. The first photomultiplier tube PMTinstalled in the Antek 7000 series analyzer wasextremely sensitive, but prone to electronic noise.This resulted in precision values much greater than

.those published by Walsh 1989 , who used an An-tek 700 series instrument. When a new PMT wasinstalled, the sensitivity was only about a quarter ofthe original PMT, but the precision was much better

for both standards and samples within 13% of theprecision values of Walsh, 1989; Lomas and Glibert,

.unpublished data .Advantages of the HTO method are the tiny

sample volume required and the large number ofsamples can be processed in one day; Frankovich

.and Jones 1998 describe a modification of theAntek 7000N that could potentially increase samplethroughput significantly. Gain settings can also beadjusted for the HTO technique, allowing optimiza-tion of the sample range. The main disadvantagewith this method is the expense of the equipment,and the time and effort required for set-up andtroubleshooting. Of the three methods, the HTOtechnique requires the highest degree of mechanicalaptitude.

4.3. Choice of standard

It has been previously recommended that NOy,3the most oxidized form of N, be used as the standard

.in TDN analyses Hopkinson et al., 1993 , and wehave followed that protocol here. However, the useof an oxidized nitrogenous compound as a standarddoes not permit the analyst to determine if and whenoxidation efficiencies may vary from run to run orfrom day to day. Therefore, use of a reduced, ororganic, standard is recommended as a secondarystandard for all methods. The replication of thereduced or organic standard relative to NOy is3essential to understanding how well the reactionandror instrument is working on a regular basis, andto avoid misinterpreting variability due to oxidationefficiency as variability in sample concentration; thisis especially true for the UV oxidation method. Onecould also use large batches of preserved naturalwaters to monitor the oxidation efficiency of a givenmethod over time. This approach would be useful forthe UV method to monitor potential changes in thelamp over time.

4.4. Salt effects

The effects of varying salinity are usually notconsidered in the development of methods as most

are developed for either freshwater or saltwater e.g.,.Kalff and Bentzen, 1984 . However, in estuarine

fields samples, variations in salinity should be con-sidered. In this study, we did see a small salinityeffect though it was not statistically significant.

.Manny et al. 1971 suggested that the oxidationefficiency by UV radiation varied with salinity,

.whereas Walsh 1989 concluded that the HTOmethod was not impacted by varying salinity. As it isthe oxidation step which is likely to be effected byvariations in salinity, future research should focus onthe specific reaction chemistry. Alternatively, a saltfactor table similar to that described for NHq4 .Koroleff, 1983 could be generated.

4.5. Spongy cadmium ersus autoanalyzer

We recommend the use of SpCd as a usefulalternative to standard reduction columns for TDNNOy or straight NOy analysis under many circum-3 3stances. First, when an autoanalyzer is unavailable,SpCd is far superior because dozens of samples canbe run simultaneously without the need for multiplereduction columns. Second, the technique is useful


when there are too few samples to make an autoana-lyzer run worthwhile. Third, when sample concentra-tions vary over a wide concentration range, the SpCdmethod can be used without the need to rerun sam-ples using different calibrations to maximize peakheights as often must be done with autoanalyzers.Finally, the SpCd method is ideal for analysis ofsediment pore water samples where the presence ofresidual sulfide can deactivate cadmium reductioncolumns.

5. Conclusions and recommendations

In this study, we found that there are definiteadvantages and disadvantages associated with all themethods tested. The PO method had the highest

.overall recovery 93% of the standard compoundswe tested. The low oxidation efficiency observed forantipyrine, however, suggests that PO may not oxi-dize refractory compounds as efficiently as some ofthe other methods. The high oxidation efficiency ofhumic compounds makes the PO method well suitedto estuarine, riverine and coastal SW samples wherehumics can be a significant fraction of the DONpool. The low cost and potential for high samplethroughput makes the PO method ideal for largefield programs or surveys. The ability to processsamples with simple boiling allows this technique tobe easily transportable to ship or field. The potentialfor high blanks requires that this method be usedcautiously in low TDN open ocean waters.

Analysis of TDN by the UV method, which hastraditionally been considered to be both reliable andefficient, and which we have used successfully in thepast, yielded the lowest percent recoveries in ourstudy. Recovery of standard compounds can be in-creased with the use of PO reagent, in place ofH O , but then any advantage of potentially low2 2blanks is lost. The small number of samples that canbe analyzed simultaneously make this a problematictechnique for large field programs or survey work.

The HTO technique oxidized most compoundswith high efficiency, and it had the highest oxidationefficiency for the most refractory compound tested,antipyrine, which suggests it would likely be a goodmethod for use in oligotrophic systems where small,

labile, easily oxidized compounds would be a smallfraction of the pool. The low oxidation efficiency ofhumics, however, would likely be problematic whenanalyzing estuarine, river or near shore samples thathave high humic concentrations. The small samplesize and potential for high throughput makes theHTO method particularly attractive if the equipmentand trained personnel are available.

Clearly, the appropriateness of different methodsfor TDN analysis is environment specific. The poorrecovery of some standard compounds with each ofthe methods suggests that concentrations of TDN,and therefore DON, in the field will likely be under-estimated with any of the methods we tested to somedegree depending on the composition of the TDNpool. With careful attention to details, the PO andHTO methods can provide reproducible results com-parable to each other. The UV method should bemonitored closely with PO reagent substituted forH O when low oxidation efficiencies are observed.2 2

Acknowledgements

We thank G. Puckett for assistance with PO andUV sample processing, E. Peltzer for advice onstatistics, and S. Seitzinger for advice on HTO analy-sis with the Antek. We also thank O. Zafiriou and S.Opsahl for assistance with the scan of the UV lamp,and R. Zepp for use of his spectroradiometer. Thisresearch was supported by Sea Grant NA26RG-0-37301 and NSF grant OCE-9522617 to DAB andNSF grant OCE-9521254 to PMG. This is contribu-tion number 3241 from the Center for EnvironmentalScience, University of Maryland.

References

Armstrong, F.A.J., Tibbitts, S., 1968. Photochemical combustionof organic matter in seawater, for nitrogen, phosphorus, andcarbon determination. J. Mar. Biol. Assays 48, 143152.

Armstrong, F.A.J., Williams, P.M., Strickland, J.D.H., 1966.Photo-oxidation of organic matter in sea water by ultra-violetradiation, analytical and other applications. Nature 211, 481483.

Benner, R., Hedges, J.I., 1993. A test for the accuracy of freshwa-


ter DOC measurements by high-temperature catalytic oxida-tion and UV-promoted persulfate oxidation. Mar. Chem. 41,161165.

Bronk, D.A., Ward, B.B., 1999. Gross and net nitrogen uptakeand DON release in the euphotic zone of Monterey Bay. Calif.Limnol. Oceanogr. 44, 573585.

Collos, Y., Mornet, F., 1993. Automated procedure for determina-tion of dissolved organic nitrogen and phosphorus in aquaticenvironments. Mar. Biol. 116, 685688.

DElia, C.F., Steudler, P.A., Corwin, N., 1977. Determination oftotal nitrogen in aqueous samples using persulfate digestion.Limnol. Oceanogr. 22, 760764.

Frankovich, T.A., Jones, R.D., 1998. A rapid, precise and sensi-tive method for the determination of total nitrogen in naturalwaters. Mar. Chem. 60, 227234.

Golimowski, J., Golimowska, K., 1996. UV-photoproduction as apretreatment step in inorganic analysis of environmental sam-ples. Anal. Chim. Acta 325, 111133.

Hansell, D.A., 1993. Results and observations from the measure-ment of DOC and DON in seawater using a high-temperaturecatalytic oxidation technique. Mar. Chem. 41, 195202.

Hansell, D.A., Williams, P.M., Ward, B.B., 1993. Measurementsof DOC and DON in the Southern California Bight usingoxidation by high temperature combustion. Deep-Sea Res. 40,219234.

Henricksen, A., 1970. Determination of total nitrogen, phospho-rus, and iron in freshwater by photooxidation with ultravioletradiation. Analyst 9, 601608.

Hopkinson, C., Cifuentes, L., Burdige, D., Fitzwater, S., Hansell,D., Henrichs, S., Kahler, P., Koike, I., Walsh, T., Bergam-aschi, B., 1993. DON subgroup report. Mar. Chem. 41, 2336.

Jones, M.N., 1984. Nitrate reduction by shaking with cadmium.Water Res. 18, 643646.

Kalff, J., Bentzen, E., 1984. A method for the analysis of totalnitrogen in natural waters. Can. J. Fish. Aquat. Sci. 41,815819.

Koike, I., Tupas, L., 1993. Total dissolved nitrogen in the North-ern North Pacific assessed by a high-temperature combustionmethod. Mar. Chem. 41, 209214.

Koroleff, F., 1983. Determination of nutrients. In: Grasshoff, K., .Ehrhardt, M., Kremling, K. Eds. , Methods of Seawater Anal-

ysis. Verlag Chemie.Le Poupon, C., Lin, R.G., Benaim, J.Y., Merks, A.G.A., 1997.

Research on automation of a reductive dissolved organicnitrogen analyzer. Oceanol. Acta 20, 349358.

Manny, B.A., Miller, M.C., Wetzel, R.G., 1971. Ultraviolet com-bustion of dissolved organic nitrogen compounds in lakewater. Limnol. Oceanogr. 16, 7185.

Menzel, D.W., Vaccaro, R.F., 1964. The measurement of dis-solved organic and particulate carbon in seawater. Limnol.Oceanogr. 9, 138142.

Nydahl, F., 1978. On the peroxidisulfate oxidation of total nitro-gen in waters to nitrate. Water Res. 12, 11231130.

Parsons, T.R., Maita, Y., Lalli, C., 1984. A Manual of Chemicaland Biological Methods for Seawater Analysis. Pergamon.

Peltzer, E.T., Fry, B., Doering, P.H., McKenna, J.H., Norrman,B., Zweifel, U.L., 1996. A comparison of methods for themeasurement of dissolved organic carbon in natural waters.Mar. Chem. 54, 8596.

Pizzicannella, F., Nirel, P., Landry, J.-C., 1996. Mise au PointdUne Methode de Determination de lAzote Organique Dis-

.sous. Arch. Sci. Geneva 49, 5968.Sharp, J.H., 1973. Total organic carbon in seawater compari-

son of measurements using persulfate oxidation and hightemperature combustion. Mar. Chem. 1, 211229.

Sharp, J.H., 1983. The distribution of inorganic nitrogen anddissolved and particulate organic nitrogen in the sea. In:

.Carpenter, E.J., Capone, D.G. Eds. , Nitrogen in the MarineEnvironment. Plenum, pp. 124138.

Sharp, J.H., Benner, R., Bennett, L., Carlson, C.A., Fitzwater,S.E., Peltzer, E.T., Tupas, L.M., 1995. Analyses of dissolvedorganic carbon in seawater: the JGOFS EqPac methods com-parison. Mar. Chem. 48, 91108.

Sokal, R.R., Rohlf, F.J., 1995. Biometry, 3rd. edn. Freeman.Solorzano, L., Sharp, J.H., 1980. Determination of total dissolved

nitrogen in natural waters. Limnol. Oceanogr. 25, 751754.Sugimura, Y., Suzuki, Y., 1988. A high temperature catalytic

oxidation method on non- volatile dissolved organic carbon inseawater by direct injection of liquid samples. Mar. Chem. 24,105131.

Suzuki, Y., Sugimura, Y., 1985. A catalytic oxidation method forthe determination of total nitrogen dissolved in seawater. Mar.Chem. 16, 8397.

Taylor, J.K., 1987. Quality Assurance of Chemical Measurements.Lewis Publishers.

Valderrama, J.C., 1981. The simultaneous analysis of total nitro-gen and total phosphorus in natural waters. Mar. Chem. 10,109122.

Walsh, T.W., 1989. Total dissolved nitrogen in seawater: a newhigh temperature combustion method and a comparison withphoto-oxidation. Mar. Chem. 29, 295311.

Wangersky, P.J., 1993. Dissolved organic carbon methods: acritical review. Mar. Chem. 41, 6174.

Whitledge, T.E., Malloy, S.E., Patton, C.J., Wirick, C.E., 1981.Automated nutrient analysis in seawater. Brookhaven NationalLaboratory Technical Report a51398.

10.1.1.319.2Total dissolved nitrogen analysis: comparisons between the persulfate, UV and high temperature oxidation methods451

Documents