1 Introduction

Ammonium (ammonium ion + ammonia) present in the surface layer of the ocean is important for biological productivity. Ammonium is the most preferred nitrogen source for uptake by phytoplankton (Dortch 1990), and ammonium uptake is 2–3 times higher than nitrate uptake (Eppley and Peterson 1979; Rees et al. 2006; Clark et al. 2007). Heterotrophic bacteria assimilate ammonium (Wheeler and Kirchman 1986), and a type of Archaea and other bacteria have the ability to oxidize ammonia and utilize ammonium as an energy source (Francis et al. 2005; Ward 2011). Although the ammonium concentration is sometimes elevated to several micromolar in the ocean (Karl et al. 1992; Nishino et al. 2011), it is generally low. Usually, the concentration is less than 0.5 µM because of active uptake by plankton (Saino et al. 1983; Kirchman et al. 1990), and in oligotrophic oceans it is depleted to less than several tenths of a nanomolar (Brzezinski 1988; Lipschultz 2001).

The most popular technique for ammonium determination during oceanic observations is colorimetry based on the indophenol reaction (Parsons et al. 1984; Li et al. 2005; McCarthy and Bronk 2008). Previously, the measurement of submicromolar ammonium concentrations was difficult because of high contamination levels and low sensitivity (Li et al. 2005; Parsons et al. 1984). Therefore, highly sensitive ammonium analyses with detection limits less than 10 nM were developed in the last quarter of a century (Ma et al. 2014; O’Connor Šraj et al. 2014). These methods are either colorimetric, based on the formation of indophenol blue (Brzezinski 1987; Li et al. 2005; Chen et al. 2011; Zhu et al. 2014) or fluorometric, based on the reaction of ammonium with o-phthaldialdehyde (Jones 1991; Kerouel and Aminot 1997; Amornthammarong and Zhang 2008). Fluorometric methods are sometimes applied to nanomolar level ammonium detection during oceanic observations (e.g., Lipschultz 2001; Matsumoto et al. 2004; Rees et al. 2006; Raimbault and Garcia 2008; Raimbault et al. 2008). However, indophenol blue methods for oligotrophic oceanic observations are quite limited (Brzezinski 1988; Harrison et al. 1996; Li et al. 2005; Zhu et al. 2014), although the methods have been further developed.

One of the problems with the indophenol blue method is finding an appropriate blank for the ammonium measurement. In highly sensitive analytical methods for nitrate, phosphate, and silicate, the blank is typically ultrapure water (Garside 1982), seawater with the target nutrient removed via coprecipitation (Karl and Tien 1992; Li and Hansell 2008; Patey et al. 2008; Hashihama et al. 2009; Hashihama and Kanda 2010), or water with a target nutrient concentration determined using ultrapure water as a blank (Patey et al. 2008; Hashihama et al. 2009). However, for ammonium determination using indophenol blue methods, seawater collected in an oligotrophic ocean (Brzezinski 1987, 1988) or seawater with added sodium hydroxide that has been heated at 60 °C (Li et al. 2005) is typically used as the blank. While the ammonium concentration in surface seawater, determined using the fluorescence method, is as low as ultrapure water (Jones 1991), it is doubtful that the ammonium concentration in this blank is uniform. This is because ammonium concentrations in surface seawater are not stable in the subtropical gyre (Lipschultz 2001). The treated seawater used by Li et al. (2005) was likely contaminated during handling (i.e., during filtration).

While ultrapure water is a suitable blank, it is not used because most indophenol blue methods are generally not free of the salt effect. However, Zhu et al. (2014) report that the salt effect was negligible in their system. In contrast, Li et al. (2005) showed that the ammonium concentration was increased by increases in salinity. Furthermore, salinity differences in the sample matrix cause spurious peaks, which make accurate analysis difficult (Coverly et al. 2012). Chen et al. (2011) reported a highly sensitive indophenol blue method, which is free from the salt effect and uses ultrapure water as the blank solution. However, this method requires complex equipment for online solid-phase extractions. Moreover, the measurement time (20 min per sample) is too long for application to routine oceanic analyses, and; hence, unsuitable, because the ammonium concentration of the samples should be determined within a few hours after collection (Hydes et al. 2010). Gas diffusion is a promising technique to eliminate the salt effect (Oliveira et al. 2009); however, this highly sensitive technique was only reported with a conductometric method with a detection limit of 10 nM (Plant et al. 2009).

A narrow detection range is another problem with the highly sensitive colorimetric and fluorometric methods. In the ocean, the ammonium concentration varies from several nanomolar to micromolar values (vide supra), but the maximum linear detection limits are less than 2 µM with highly sensitive methods (Brzezinski 1987; Jones 1991; Kerouel and Aminot 1997; Li et al. 2005; Amornthammarong and Zhang 2008; Chen et al. 2011). Therefore, we developed a quick and highly sensitive ammonium determination method, which is accurate over a wide range of concentrations and is largely independent of salinity effects. This method relies on a gas diffusion technique (Van Son et al. 1983), and only requires a liquid waveguide capillary cell (LWCC). The LWCC can be attached to an existing continuous flow autoanalyzer, which is a popular method for highly sensitive nutrient analyses (Li et al. 2005; Patey et al. 2008).

2 Experimental

2.1 Automated analytical system

A gas-segmented continuous flow colorimeter (AAII, Bran + Luebbe) was employed for the automated low-level determination of ammonium in seawater (Fig. 1). The manifold configuration and flow diagram were modified from the report by Van Son et al. (1983); the reagents and their flow rates were also changed. The manifold was separated into two sections: an outgassing section, and an ingassing and indophenol reaction section. In the outgassing section, the sample solution was mixed with an alkaline solution containing ethylenediaminetetraacetic acid (EDTA) and flows through glass coils. Elevation of the pH in the presence of the chelating agent EDTA resulted in the formation of ammonia (gas) from the ammonium ions in the sample without precipitation of hydroxide salt. The ammonia gas was absorbed by a sulfuric acid solution in a gas diffusion unit (Fig. 1). The gas diffusion unit consisted of a PTFE gas-permeable membrane filter (pore size = 0.5 μm, thickness = 0.075 mm, T050A090C, ADVANTEC) and a dialyzer. The pH was measured using an electronic pH meter (M-12, Horiba). The residence time for the ingassing and outgassing sections of the dialyzer was 7 s and 16 s, respectively. The ammonia absorbed in the acid solution reacted with alkali-phenol and hypochlorite solutions in a 45 °C bath to generate indophenol blue dye. After the segmented gas was removed from the waste line (Fig. 1), the intensity of light transmitted through the reacted solution was detected using a 1-m path length LWCC (LWCC-2100, World Precision Instruments), a fiber optic tungsten light source (FO-6000, World Precision Instruments), optical fibers (P-200-2-UU/VIS, Ocean Optics), and a fiber-optic spectrometer (USB4000, Ocean Optics). The spectrometer was connected to a personal computer using the SpectraSuite software (Ocean Optics).

Fig. 1
figure 1

Manifold configuration and flow diagram for the gas-segmented continuous flow analysis of trace ammonium using an LWCC

Full size image

According to the Lambert–Beer’s law, absorbance was calculated using the following equation:

$$A_{\lambda } = - \log_{10} \left( {\frac{{S_{\lambda } - D_{\lambda } }}{{R_{\lambda } - D_{\lambda } }}} \right).$$
(1)

In this equation, A λ and S λ are the absorbance and light intensity observed through the sample with the added reagents at a wavelength of λ nm, respectively. D λ and R λ are the dark intensity and reference intensity at a wavelength of λ nm, respectively. In the present study, the highest absorbance and the most sensitive measurements with indophenol blue were obtained at a wavelength of 630 nm. Low sensitivity values were obtained for a wavelength of 530 (Fig. 2). The values of D 630 and D 530 were set at zero in the present study because the SpectraSuite software can correct the electric dark to zero when a USB4000 spectrometer is used. The values of R 630 and R 530 equal the light intensity through ultrapure water containing reagents. In the present study, the value of R630 and R530 were adjusted by changing the integration time (the frequency of the measurement) because the light source intensity cannot be adjusted and the transmitted light intensity (S λ and R λ ) cannot be measured for values higher than 65,500 counts using a USB4000. S 630 and S530 were recorded and outputted at least once per second during the measurement.

Fig. 2
figure 2

Typical absorbance (black line) and transmitted light intensity (gray line) for ammonium at wavelengths between 400 nm and 1000 nm. The horizontal dashed lines denote the obtained wavelength of 530 and 630 nm

Full size image

2.2 Analytical reagents

All the chemicals used in this study were reagent grade and were dissolved in ultrapure water purified using an ultrapure water system (AQUARIUS, Advantec). The concentrations of the following reagents were prepared manually.

2.2.1 30 % Triton-X solution

Triton-X 100 (Sigma, 30 mL) was dissolved in ultrapure water (70 mL) and ultrasonicated for 20 min at room temperature.

2.2.2 EDTA (0.28 M EDTA/0.065 M H3BO3) reagent

EDTA·4Na (Dojindo 62.5 g) and boric acid (Wako 2.0 g) were dissolved in ultrapure water (ca. 400 mL) and then diluted to 500 ml in a measuring flask. The 30 % Triton-X solution (2.5 mL) was then added to the EDTA mixture. The mixing ratio of this reagent to the sample solution was 0.32:1.

2.2.3 EDTA/NaOH (0.18 M EDTA/1.25 M NaOH) reagent

EDTA·4Na (16 g) and sodium hydroxide (Wako 10 g) were dissolved in ultrapure water (ca. 150 mL), and then the solution was diluted with ultrapure water in a measuring flask to 200 mL. The mixing ratio of this reagent to the sample was 0.16:1.

2.2.4 Nitroprusside (0.67 mM nitroprusside sodium/0.01 M H2SO4) reagent

A 1 M H2SO4 stock reagent was first prepared by dissolving sulfuric acid (Wako 5.6 mL) in ultrapure water (95 mL). An 8.4 mM nitroprusside sodium reagent was then prepared by dissolving nitroprusside sodium (Wako 0.25 g) and 1 M H2SO4 (0.20 mL) in ultrapure water (100 mL). These stock reagents were stored in the dark at 4 °C. A working solution was prepared by adding 1 M H2SO4 (5.0 mL) and the stock nitroprusside sodium reagent (4.0 mL) to ultrapure water (500 mL) followed by the 30 % Triton-X solution (1.0 mL). The reagent suction rate in the flow line was 0.80 mL min−1 when the sample suction rate was 1.0 mL min−1, and the ratio of the flow volumes at the outgassing and ingassing sections in the dialyzer was 1:0.43.

2.2.5 Alkali-phenol (0.53 M phenol/0.039 M citrate/0.625 M NaOH) reagent

First, sodium hydroxide (Wako 5.0 g) was dissolved in ultrapure water (200 mL). After cooling, trisodium citrate dehydrate (Wako 2.0 g) and solid phenol (Wako 10.0 g) were dissolved in the sodium hydroxide solution. This solution could be stored for at least 1 week at 4 °C. The mixing ratio of this reagent to the nitroprusside reagent was 1:5.

2.2.6 Hypochlorous reagent

Liquid sodium hypochlorite (Dojindo 5 mL) with an available chlorine concentration greater than 5.0 % was mixed with ultrapure water (100 mL). This reagent was prepared daily. The mixing ratio of this reagent to the nitroprusside reagent was 1:8.

2.3 Standards and blanks

An ammonium stock standard solution (1 L, 10 mM) was prepared from ammonium sulfate (analytical reagent grade) that was pre-dried (110 °C for 2 h) and stored in the dark at 4 °C after adding one drop of chloroform. In accordance with Li et al. (2005), working standards from 5 nM to 10,000 nM were prepared in polypropylene measuring flasks (100 mL or 200 mL) using both ultrapure water and low-nutrient seawater (LNSW). The LNSW was collected at 30°N, 138°E of the oligotrophic western North Pacific in October 2009, and stored for more than 1 year in light at room temperature.

Blank solutions included ultrapure water with a conductivity less than 0.06 µS cm−1, ion-exchanged water with a conductivity less than 0.1 µS cm−1, a 0.625 M sodium hydroxide solution heated overnight at 65 °C (NaOH aq.), a 3.5 % (0.060 M) sodium chloride aqueous solution, and 0.2-µm filtered and unfiltered LNSW. The filtered LNSW was collected and stored in the same manner as the unfiltered LNSW. In addition, according to Li et al. (2005), a blank was also prepared by adding 0.625 M sodium hydroxide (1 mL) to the LNSW supernatant (9 mL) and heating overnight at 65 °C (alkaline-LNSW).

2.4 Cleaning and maintenance

Before and after daily use, the reaction path in the manifold and the LWCC were cleaned using a 1 % Triton-X 100 aqueous solution. In this study, the precipitation of hydroxide salt was prevented via the addition of EDTA. However, precipitates were deposited in the reaction path when EDTA concentrations were lower than normal because of issues with flow lines, such as when the pump tube was broken. These precipitates were readily removed by rinsing with 6 M HCl. In addition, introduction of 6 M HCl via suction through the hypochlorous reagent and alkali-phenol reagent lines was effective for cleaning the LWCC. The PTFE membrane filter generally remained in good condition for at least 2 weeks; however, it was changed every 2 weeks. The signal for filter deterioration was the length of the gas-segmented patterns, which was different before and after the dialyzer when the filter had deteriorated.

3 Results and discussion

3.1 Linear dynamic range and detection limit

A typical output signal for nanomolar concentrations of ammonium was derived from analyses using the gas-segmented flow injection analyzer equipped with the LWCC (Fig. 3). Coverly et al. (2012) reported that spurious peaks were formed when the matrix changed from freshwater to seawater because of the salinity difference; however, spurious peaks were not observed in this study (Fig. 3). Therefore, the hydrophobic PTFE membrane filter was considered able to prevent the transport of ions, which was consistent with Oliveira et al. (2009). To determine the optimum sample suction time for the 400 nM working standard, the time was varied from 30 s to 240 s. A sample suction time of 180 s was found to be sufficient for reaching 95 % of the maximum absorbance, which is the 10 s mean absorbance at the top of the 240 s suction using this system (Fig. 4). The peak shapes also indicated that a sample suction of 180 s was sufficient for reaching maximum absorbance. Therefore, sample and wash times greater than 180 s were considered to be effective for eliminating the influence of sample dispersion. An acceptable carryover coefficient of 0.2 % was obtained when the sample suction intervals were set to 360 s; the coefficient for well-designed and maintained channels is <0.3 % (Gordon et al. 1993). Therefore, the treatment capacity of the system (10 samples per hour) and the sample volume (3.0 mL) were suitable for achieving accurate analysis results. However, this treatment capacity was lower than the previous method (22 samples per hour; Zhu et al. 2014).

Fig. 3
figure 3

Typical output signals for ammonium from the flow analyzer equipped with an LWCC. Ultrapure water was injected for the first 450 s, and then working standards of various ammonium concentrations prepared using ultrapure water and seawater samples collected in the oligotrophic western Pacific were sequentially injected for 210 s, with ultrapure water injected for 150 s in the intervening periods

Full size image
Fig. 4
figure 4

Percentage of output signal for 400 nM ammonium standards as a function of sample suction duration. Error bars indicate the standard deviation (n = 3)

Full size image

A calibration curve for A 630 (the absorbance at 630 nm) was established using triplicate measurements of ten different concentrations of working standards from 0 nM to 2,000 nM, which were prepared using ultrapure water (Fig. 5a). A linear absorbance response to trace ammonium concentrations ≤2,000 nM were obtained with a significant correlation (r 2 = 0.9993). In addition, working standards up to 200 nM also provided a linear absorbance response (r 2 = 0.9956, Fig. 5b). The day-to-day slope variations ranged from 4.0 × 10−4 to 5.8 × 10−4 (mean ± SD 4.9 × 10−4 ± 0.7 × 10−4, n = 6). However, the slope variations were not stable, and the exact causes for these variations are not clear; the age-related degradation of the filter, and the differences in the working reagents may affect the variations. The minimum detection limits were 2.3–7.6 nM (5.5 ± 1.8 nM, n = 6), which was estimated to be three times the standard deviation of the measurement blank (ultrapure water; n = 5; Miller and Miller 1988). The variation of the detection limit in this study was due to variations in the day-to-day slope. Our lowest detection limit (2.7 nM) was the lowest of all reported for high sensitive colorimetric methods and higher than those for the fluorometric methods (Table 1). On the other hand, the highest detection limit (7.6 nM) was the highest of the previously reported highly sensitive analysis except for the conductometric method using a PTFE filter (Plant et al. 2009). The reproducibility of the 100 nM standards had a coefficient of variation of 3.8 % (n = 5), which is lower than the previous studies using the indophenol blue methods (5 %; Li et al. 2005 and 4.4 %; Zhu et al. 2014). Therefore, detection of trace ammonium concentrations with high precision was achieved using this LWCC system.

Fig. 5
figure 5

Linearity of ammonium concentration at A 630 a from 0 nM to 2000 nM and b from 0 nM to 200 nM using the highly sensitive continuous flow analysis

Full size image
Table 1 Comparison of detection ranges and blanks for other highly sensitive systems
Full size table

However, the maximum detection limit was not high enough for application to oceanic observations, because the ammonium concentration in the ocean may be several micromolar (Jones 1991; Karl et al. 1992; Nishino et al. 2011). Thus, the A 530 values, which have a lower sensitivity than the A 630 values were concurrently measured with A 630 values using the same settings. As a result, a significant linear absorbance response to ammonium concentrations ≤10,000 nM was obtained using the calibration curve for A 530 (r 2 = 0.9993, Fig. 6). The daily slope variation ranged from 1.21 × 10−4 to 1.48 × 10−4 (mean ± SD 1.42 × 10−4 ± 1.1 × 10−5, n = 6). The minimum detection limit of 13.0 ± 5.3 nM (n = 6) was too high to determine ammonium concentrations at nanomolar levels; however, this was not a critical problem because low-level ammonium concentrations were measured with A 630 values. The molar absorptivity values at 630 nm and 530 nm were 8.7 × 103 and 1.4 × 103 M−1cm−1, respectively.

Fig. 6
figure 6

Linearity of ammonium concentration at A 530 using the highly sensitive continuous flow analysis

Full size image

3.2 The salt effect and the blank

Ten different concentrations of working standards from 5 nM to 3,000 nM were prepared using ultrapure water and filtered LNSW (Fig. 7). The relationship between A 630 and the standards was nonlinear, because the A 630 value for the 3,000 nM standards was beyond the upper limit of the detection range. The values of both A 630 and A 530 were not generally different between freshwater and seawater (n = 3; Fig. 7), although significant differences were obtained for the 50 nM and 500 nM standards (t test, p < 0.05). The A 630 values for the 50 nM and 500 nM standards and A 530 values for the 500 nM standards were higher than those estimated by linear regression; thus, contamination may have occurred in these two standards. Plant et al. (2009) reported that salinity has an effect on ammonium diffusion, and the ammonium transit in the diffusion cell (dialyzer in this study) decreased by 12 % for the seawater compared with the ultrapure water because of the pH difference. The ammonium diffusive flux (J) through the PTFE membrane is derived from the following equation (Schulze et al. 1988):

Fig. 7
figure 7

Comparison of ammonium concentration in working standards prepared using ultrapure water (crosses) and unfiltered LNSW (open circles) at a A 630 and b A 530. Each working standard was measured three times, and the bars denote the standard deviation. The value of D λ was set as “0.”

Full size image
$$J = k \times D \times \left( {A/L} \right) \times \varDelta c,$$
(2)

where, k, D, A, L, and Δc are the lump constant, the contact area between the outgassing and ingassing sections, the membrane thickness, the diffusion coefficient of ammonia in air, and the ammonia concentration gradient, respectively. The membrane thickness (A) in this study was suitable for the gas diffusion method according to Plant et al. (2009). In Eq. 2, Δc is theoretically different between seawater and ultrapure water, because the pH difference causes the conversion of ammonium ion to ammonia. This difference in the pH of ultrapure water (theoretically 7, but usually less) and seawater (8.32) is significant. In the present system, the pHs of the seawater with added reagents and the ultrapure water with added reagents were 11.98 ± 0.012 and 11.77 ± 0.004 (both n = 5), respectively. The acid dissociation constant of ammonium/ammonia (K a) is 5.6 × 10−10 (Dean 1998); thus, the ammonium/ammonia in the samples existed as 99.7 % and 99.8 % ammonia in the seawater and ultrapure samples, respectively. This difference was rarely detected for the current method. In addition, the hydrophobic membrane filter generally prevents the transport of ions from samples containing EDTA to the nitroprusside reagent (Oliveira et al. 2009). Therefore, ultrapure water can be used as the blank solution and for preparation of the standards.

The ammonium concentration in ultrapure water was assumed to be “zero” in previous studies (Table 1). The conductivity of pure water is theoretically 0.055 µS cm−1 (Dean 1998), while the conductivity of ultrapure water used in this study was less than 0.06 µS cm−1. The specific electrical conductance of ammonium and hydroxide ions is 53.5 and 198 S cm2 equiv−1 at 25 °C, respectively, and K a is 5.6 × 10−10 (Dean 1998). When only ammonium contamination is assumed to increase the conductivity of pure water from the theoretical value to a practical value (<0.01 µS cm−1), the ammonium concentration for ultrapure water would be less than 0.05 nM, which is much lower than the detection limit and can be ignored. Therefore, the ammonium concentration in the ultrapure water was considered as 0 nM. In addition, the ammonium concentration in the ion-exchanged water (conductivity <0.1 µS cm−1) is theoretically less than 0.4 nM, and the measured ammonium concentration in ion-exchanged water was not significantly different from ultrapure water (Fig. 8).

Fig. 8
figure 8

Comparison of the ammonium concentration in the favored blank solutions. Error bars indicate the standard deviation (n = 4)

Full size image

The ammonium concentration in unfiltered LNSW was lower than the detection limit (<2.3 nM) and not significantly higher than ultrapure water (Fig. 8). On the other hand, the ammonium concentration in filtered LNSW differed between batches (LNSWs: a–c in Fig. 8), while the concentration in all batches was significantly higher than ultrapure water. The ammonium concentration in ASW was also higher than ultrapure water. The ammonium concentrations in the 0.625 M NaOH aq. solution and the alkaline-LNSW were 448 ± 15.8 nM and 99 ± 29.3 nM, respectively (Fig. 8). These results indicate that ultrapure water, ion-exchanged water, and unfiltered LNSW are suitable for use as blank solutions. Furthermore, ultrapure water is considered to be the best blank solution for use with this method because of its stability, economics, and ease of handling. Since the ammonium concentration in the 0.625 M NaOH aq. solution and the alkaline-LNSW (Fig. 8) were high, NaOH is the source of the ammonium contamination. The EDTA/NaOH and alkali-phenol reagents can also contaminate ammonium, which decreases the light intensity (S λ and R λ ) and sensitivity of the analysis. Therefore, the NaOH used for preparing reagents in this study was as new as possible in order to minimize contamination.

3.3 Interference study

To assess the possibility of interference with ammonium determination, solutions containing 10 mM urea ((NH2)2CO) and glycine (aminoacetic acid NH2CH2COOH) in ultrapure water were prepared, and their absorbance was measured. The absorbance of the aqueous 10 mM urea and glycine solutions was significantly higher than ultrapure water, with indicated ammonium concentrations of 2,277 ± 34.6 nM and 98 ± 10.0 nM, respectively (both n = 4). However, in open and coastal oceans the concentrations of urea and dissolved free amino acids are generally lower than 1 µM (Bronk 2002), and the absorbance of urea and glycine solutions at 1 µM are 0.23 and 0.098 nM of ammonium concentrations, respectively.

4 Application to observations

4.1 Vertical distribution of ammonium concentration

The nanomolar level ammonium concentrations were investigated in the vicinity of the Kuroshio Current along 138°E (the O-line) in summer (September 2012) during the R/V Soyo-maru cruise (Fig. 9) to confirm the highly sensitive analysis results for oceanic observations. The Kuroshio axis exists at 33°15′N based on the Quick Bulletin of the Ocean Conditions published by the Japanese Coast Guard (http://www1.kaiho.mlit.go.jp/KANKYO/KAIYO/qboc/index_E.html). Samples for ammonium concentration determination were collected using a Rosette Niskin attached to a CTD system (Seabird) at depths ranging from 5 m to 200 m, and immediately analyzed on board. The concentrations of other nutrients (nitrate, nitrite, silicate, and phosphate) were determined at an on land laboratory with a TrAAcs 2000 system (Bran + Luebbe) by conventional methods using stored (frozen) samples.

Fig. 9
figure 9

Stations where the highly sensitive ammonium determination system was applied in the south of Japan. The arrow indicates the position of the Kuroshio current during the observation period based on the Quick Bulletin of Ocean Conditions

Full size image

The minimum detection limit was 6 nM at a wavelength of 630 nm during the observations. The ammonium concentrations measured on board varied from <6 to 253 nM during the cruise (Fig. 10). Typical vertical profiles for ammonium concentration were obtained. The maximum subsurface ammonium concentration was just below the subsurface chlorophyll maximum (data not shown), and nitracline was the same as the oligotrophic Atlantic (Brzezinski 1988; Rees et al. 2006) (Fig.10). These results indicated that our system could be applied for ocean observations to understand the distributions of nanomolar level ammonium concentration in the oligotrophic ocean.

Fig. 10
figure 10

Vertical distributions of a ammonium concentration and b nitrate + nitrite concentration. The arrowheads above the panels denote the position of the Kuroshio axis

Full size image

4.2 Measurement of ammonium excretion rate by zooplankton

This highly sensitive analysis method has potential for evaluation of the zooplankton excretion rate by zooplankton because it allows short-term and low-density incubations, which are insulated from the influence of starvation and stress of crowds (Steinberg and Saba 2008). Ammonium excretion by two different species of female adult copepods, Calanus jashnovi [prosoma length (PL) 3.18 ± 0.07 mm (mean ± SD), n = 6] and Oncaea venusta (PL: 0.65 ± 0.08 mm, n = 12), was measured during an early spring cruise (March 2013) along the O-line. The zooplankton were gently collected by a NORPAC net at 31°N and 33°45′N, immediately and individually sorted to 13 mL filtered in situ surface seawater, and incubated for approximately 4 h in darkness at the in situ sea surface temperature (18 °C).

The ammonium concentration increased to 2,680 ± 70 nM and 184 ± 107 nM for the incubated bottles of C. jashnovi and O. venusta from the control bottles (with only filtered in situ surface seawater), respectively. The excretion rates were 116 ± 7 and 7.2 ± 4.8 ng N h−1 ind−1 by C. jashnovi and O. venusta, respectively. These values were the same order of magnitude as the reported values from the same temperature range but with different zooplankton species (Ikeda et al. 2001). Therefore, this method was used for the estimation of short-term ammonium excretion rates by copepods, which are approximately two-orders of magnitude different by species and/or sizes, because the present method can measure a four-order range of ammonium concentration.

5 Conclusions

The present method provides a highly sensitive technique for analyzing a wide range of ammonium concentrations in seawater. This system may be applied for different kinds of oceanic observations and experiments because it is based on existing and commercially available systems, which can detect nanomolar and micromolar concentrations of ammonium. In addition, deionized water was determined as an acceptable blank solution. Thus, unified blank and working standards can be utilized for ammonium analysis. Furthermore, a continuous flow autoanalyzer equipped with LWCCs was applied for the continuous determination of nanomolar concentrations of nitrate, phosphate, and silicate in the oceanic surface water. This method provided clear horizontal distributions and spatial variations of nutrient concentrations in oligotrophic oceans (Hashihama et al. 2009, 2010, 2014; Kodama et al. 2011). This new highly sensitive analytical method for ammonium determination may also be used to detect the horizontal distribution of ammonium concentrations along with other nutrients, thus, making major contributions to the understanding of nitrogen dynamics in oligotrophic oceans.