Analytica Chimica Acta 377 (1998) 167177

A low power ultra violet spectrophotometer for measurement of nitrate

in seawater: introduction, calibration and initial sea trials
Miles S. Fincha,*, David J. Hydesa, Charles H. Claysona Bernhard Weiglb,
John Dakinc, Pat Gwilliama
a
Southampton Oceanography Centre, Southampton SO14 3ZH, UK
b
Department of Bioengineering, University of Washington, St. Louis, MO 63130, USA
c
Department of Electronics, University of Southampton, Southampton SO17 1BJ, UK

Received 27 March 1998; received in revised form 20 August 1998; accepted 26 August 1998

Abstract

A low power UV spectroabsorptiometer has been developed to measure the concentration of dissolved nitrate in seawater, in
situ, with a rapid 1 Hz response. Measurements at a number of wavelengths in the region of 220 nm determine the
concentrations of dissolved nitrate, sea salt and, possibly, dissolved organics, whilst a third channel at 300 nm acts as a
reference channel to correct for light intensity changes related to lamp output and scattering by particles in seawater.
This paper describes the rst prototype UV nitrate sensor designed to evaluate and compare options for the light source,
spectral lter and detector. The choice of components was based on a compromise between performance, cost, power
consumption and convenience of use. The eventual instrument uses a xenon ashlamp light source, fused silica windows and
lenses, a sample cavity, a grating spectrometer and UV enhanced silicon photodiode detectors. The power consumption of the
sensor was typically from 3 to 4 W, depending on the selected repetition rate of the ashlamp source. Initial laboratory trials
have shown the sensor to be unresponsive to changes in pressure allowing for continuous use to depths of up to 5000 m. Initial
responses to changes in temperature have been eliminated by the adjustment of the sensor's power supply and lamp ash rate.
Calibration of the sensor has been performed under laboratory conditions and resulted in a second order linear t (r2>0.99,
p<0.05) with respect to nitrate concentration. Calibrations performed to assess the response of the instrument to changes in
salinity also resulted in second order linear ts (r2>0.99, p<0.05). Salinity calibrations indicated that a change of 1 psu resulted
in a change in absorbance equivalent to 0.52 mmol l1 NO3. The detection limit of the sensor is a function of the signal/noise
ratio. The minimum nitrate concentration change resolved by the sensor was found to be >0.21 mmol l1 NO3.
Measurements made in situ were compared to concentrations of nitrate determined in water samples using an AAII type
autoanalyser. These demonstrated the capability of the instrument to precisely determine nitrate concentrations in seawater.
# 1998 Elsevier Science B.V. All rights reserved.

Keywords: Nitrate; Dissolved nitrate; Ultra-violet; Sensor; Autoanalyser; Calibration; Spectroabsorptiometer

*Corresponding author. E-mail: [email protected]

0003-2670/98/$ see front matter # 1998 Elsevier Science B.V. All rights reserved.
PII: S0003-2670(98)00616-3
168 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177

1. Introduction centrations that the interference they produce is neg-

ligible [8]. Molar extinction coefcients () for halides
The main objectives of this sensor programme were are about 106 times less than nitrate but their con-
to produce a self contained monitoring system with a centration in seawater is typically 5104 times higher
low power consumption which would allow for towed resulting in relative absorbance 5102 times that of
or lowered oceanographic deployment in addition to nitrate. Under the conditions used in this study,
long term unattended deployment using a battery Cl 0.01 M1 cm1, at 220 nm at 220 nm, at sea-
power source. The sensor would have to have a rapid water concentrations with a path length of 15 cm and
response time (<1 s) to enable a resolution of 1 m NO3 644 M1 cm1, at 220 nm, at a concentration of
when lowered through seawater on vertical hydro- 10 mM with a path length of 15 cm.
graphic casts and a resolution of 5 m on an undulat- The ions HCO 2
3 and CO3 are present in the open
ing towed vehicle. ocean in a ratio of approximately from 97% to 2%.
Several other methods have been developed to Since HCO 3 has its main absorption band below
measure the concentration of nitrate in seawater. 210 nm and CO2 3 is present in such low concentra-
These include selective colour indicator tests, nitrate tions, they should not contribute signicantly to any
sensitive electrodes [1,2], and nitrate optrodes [35] as interference effect. However, in estuaries, particularly
well as the use of a bre optic coupled UV photometer fed by two chalk rivers, the effects of CO2 3 ; HCO3

[6]. However, none of these methods meet all the will become more signicant due to the higher con-
requirements for long term monitoring. Recently, a centrations of CO2 3 ions resulting from a shift in the
submersible osmotically pumped analyser has also carbonate equilibrium. These effects are currently
been reported [7] but this does not have a sufciently being investigated.
rapid response time for towed applications. Fig. 1 illustrates the UV absorption spectra of
The determination of nitrate using ultra violet nitrate, nitrite, sodium chloride and seawater solutions
spectroscopy is well known [8,9]. Several investiga- in distilled water. In seawater, chloride and bromide
tions have been carried out [1012] to assess the are present in a nearly constant ratio, both to one
absorption of UV light by natural waters including another and to the total salt content (salinity). Salinity
seawater. A number of problems have been found with is routinely determined to a high precision in marine
this method of analysis [10], but it has shown that a waters so a simple correction can be applied to
simple measurement of UV absorbance at 220 nm absorbance measurements made to determine nitrate.
could be used as an indicator of dissolved nitrate
distribution in open ocean waters [9,13]. Methods
for freshwater analysis of dissolved nitrate have been
reported [1416] but the application of a self-con-
tained electronic/optical instrument utilising UV mea-
surement has never been widely applied in the marine
environment due to the reliable chemical based ana-
lysis involving the coppercadmium reduction
method. Recent developments have combined UV
spectroscopy with ow injection analysis or have used
2nd derivative spectroscopic techniques [17].
Nitrate absorbs signicantly at wavelengths up to
230 nm. The main ionic species found in seawater
exhibiting similar characteristics include Cl and Br
plus low concentrations of SO2 2
4 ; CO3 ; HCO3 ;

HPO4 ; H2 PO4 ; NO3 ; and NO2 . Except for NO

2
2;
NO 3 and CO 2
3 these anions have their main absorp-
tion band well below 210 nm. All other ions, except Fig. 1. Absorption spectrum versus wavelength for nitrate, nitrite,
chloride and bromide, are present in such low con- sodium chloride and seawater measured in distilled water.
 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177 169

Signicant problems may be associated with UV The xenon ashlamp is operated at a ash rate
analysis of nitrate in seawater due to absorption of UV of 16 Hz, with a ash input energy of approximately
light by dissolved organic compounds as their con- 100 mJ; this gives a combination of adequate
centrations are not predictable. Aliphatic and aromatic light output energy and a lamp life of at least three
compounds, for example, mineral oils, petrol, alcohols months.
and biologically derived compounds (as well as sev-
eral anionic species in addition to Cl and Br) exhibit 2.1.1.2. UV prefilters. Most detectors are several
broad band absorption within the wavelength range orders of magnitude more sensitive to visible and
from 150 to 400 nm. However, previous observations IR light than they are to UV. Although the grating
suggest that the interference may be relatively small in spectrometer separates the visible spectrum from the
open ocean water [8]. UV spectrum of interest, stray scattered visible light
could reach the detectors. Any such visible light was
removed by two dichromatic filters before it arrived at
2. Experimental the grating spectrometer. The filters, which perform in
a 458 reflection mode in the UV waveband, transmit
2.1. Experimental development of in situ device any visible light to an absorber. The reflection
efficiency of the filters is only about 40% at
The design of a spectrophotometric instrument 220 nm and therefore they are the single greatest
for monitoring nitrate in seawater followed a three- loss factor in the optical system.
stage process. Firstly, an optical bench mounted
conguration was used to evaluate potential compo- 2.1.1.3. Lenses. Lenses are required for collimation of
nents such as light sources, lters and detectors. the light from the source and for imaging the spectrum
Secondly, after the preferred components had been upon the detectors. UV-grade fused silica lenses with a
selected, a boxed sensor was designed to perform transmission of roughly 90% at 200 nm were selected.
measurements at three wavelengths in the laboratory. Calcium and magnesium fluoride also offer low loss in
Thirdly, a fully submersible unit was manufactured to this region, but lenses made of these materials are less
allow tests to be carried out in the eld. This unit widely available and are slightly susceptible to
has subsequently been redesigned for commercial thermal and mechanical shock. Synthetic sapphire is
production, it is intended to report on this version very durable but is considerably more expensive and
in the technical literature in due course, but the current appreciably less transparent at 200 nm, it was also
paper is more concerned with the technique than the found to suffer from solarisation (darkening) problems
details. in high intensity UV fields unless a grade of the
highest purity is used.
2.1.1. Choice of components
2.1.1.4. Windows. Fused silica windows with a
2.1.1.1. Light sources. The need for low overall power thickness of 3 mm and a diameter of 25 mm
consumption placed a major constraint upon the (Heraeus Suprasil) were selected for the boxed
choice, as the instrument was required to be sensor prototype. The windows were bonded to the
capable of battery operation over extended periods. outer surface of the test unit housings. The losses
A practical limit on battery pack capacity is of order through each uncoated window were about 10% in air,
1 kWh and a typical deployment lasts 10 days, primarily due to Fresnel reflection losses which are
resulting in a mean power budget of about 4 W. reduced when the unit is operated in water. The fully
The ideal light source would also have spectral submersible unit uses a single window, combined with
stability, intensity stability and a long lifetime. Table 1 an external retro-reflecting porroprism, this window is
shows the properties of the light sources tested. The designed to withstand up to 500 bar differential
xenon ashlamp satised the selection criteria most pressure for deep-ocean use.
fully. It had an excellent UVoutput coupled with a low Degradation of the transparency of optical windows
mean power consumption. after long term exposure to seawater is a well-known
 170

Table 1
Summary of requisite lamp characteristics

Lamp type Peak radiance at 220 nm, Requirements for optical Typical power consumption Expected lamp lifetime
relative to flashlamp pre-filtering

Xenon flaslamp 1. This lamp used as the Maximum intensity, Imax is 230 nm, <2 W at a flash repetition 10005000 ha under the
reference level for comparison appreciable longer wave output, frequency of 100 Hz, using conditions specified at left
of the other lamps hence a need to provide visible a storage capacitance of
and near IR rejection 0.1 mF with a 600 V supply

Quartz halogen 105 Imax is at 550 nm, output at this 6W 200 h unfortunately lamp cannot
wavelength is 105 higher than at be under run to increase lifetime
220 nm. Thus, very high visible without giving very low UV output
and near IR rejection is required

Deuterium lamp 1. Similar to the output of Imax is below 200 nm, lamp has 30 W 1000 h
the flashlamp, however, a very little visible and IR output,
higher mean output is therefore no pre-filtering required
available due to continuous
operation

Mercury lamp Very low due to discrete nature Difficult to assess as no strong 9W 8000 h
of the emission spectrum and discrete lines are available at
the lack of strong lines in the the required wavelengths
M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177

required region
a
The lifetime of this type of flashlamp does not end abruptly. There is a slow decrease in light intensity. Therefore, the lamp can be exchanged when the measured light intensity
becomes less than a defined threshold.
 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177 171

problem in oceanographic instrumentation, this is detected energies in three spectral bands. In order
usually caused by algal growth or plankton deposition to achieve the desired spacing, each detector was
on the window surface. However, this effect should be connected to a short tail of optical fibre, to allow
somewhat reduced in this system due to the biocidal closer packing of the light receiving apertures in the
properties of UV light. focal plane of the grating spectrometer. For the fully
submersible version, a linear photodiode array was
2.1.1.5. Spectrometer. An 1800 line mm1 holo- used in conjunction with an imaging lens selected to
graphic grating spectrometer was used to disperse give a spectral intervals of 20 nm at the photodiode
the UV light onto several output plane apertures to spacing. Specially designed low-power electronic
select the operational wavelengths. This has circuits then integrated the photodiode output
advantages of low-loss, flexibility and cheapness in current pulses, converting them to voltages
comparison to multiple filter systems. In the boxed proportional to the detected energies in the three
system, selection of wavelengths was achieved simply spectral bands.
by adjusting the position of the detectors in the output Fig. 2 illustrates the arrangement of the boxed UV
focal plane of the spectrometer. In the fully monitor. It employed a xenon ashlamp light source,
submersible version, this was achieved by selecting fused silica windows and lenses, a holographic grating
elements of a multi-element photodiode array. and an array of three discrete UV-enhanced silicon
photodiodes coupled to transimpedance pre-ampli-
ers. Three wavelength operation was chosen to pro-
2.1.1.6. Detectors. A number of detector types,
vide:
including UV-enhanced silicon photodiodes, flame
sensors, and photomultipliers were initially tested. 1. a measurement channel at a wavelength where
Photomultipliers exhibit poor quantum efficiency in both nitrate and halide have signicant absorption,
comparison with silicon photodiodes and are 2. a channel close to wavelength (1) where absorption
excessively bulky for use in a portable multi- due to halide is very small, and
channel instrument. For the boxed prototype, 3. a reference channel which is primarily affected by
discrete photodiode detectors were employed in obscuring particles and changes in the intensity of
conjunction with FET transimpedance amplifiers. A the light source, allowing normalization of the
lock-in amplifier was then used to monitor the other two channels for these variations.

Fig. 2. The Mark I ultra violet nitrate sensor.

172 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177

The three detector channels 1, 2 and 3 were set at signal due to introducing seawater into the path. The
wavelengths of 220, 240 and 305 nm, respectively, absorption of nitrate solutions followed the Beer
with a typical grating spectrometer bandwidth from 5 Lambert law. The cross sensitivity to chloride was
to 10 nm, depending on the aperture size selected in tested with a xed 10 mmol l1 nitrate solution con-
the output plane of the spectrometer. taining different concentrations of chloride. The pure
nitrate solutions were measured using 1 cm path-
2.1.1.7. Electronic systems. The boxed version was length, vitreous silica cuvettes, whereas the solutions
operated using standard laboratory units for power containing sodium chloride were contained in a 10 cm
supply, control of flash rate and for lock-in detection pathlength silica cuvette. For easier comparison, how-
of the transimpedance amplifier output signals. For the ever, all concentrations given were normalised to
fully submersible version, compact low-power correspond to the same absorption over the eventual
electronic circuits were designed to take the place expected measurement pathlength of 230 mm.
of these units. ACDC converters were used for power
supply, providing complete electrical isolation. The
flash rate was selectable, but a rate of 16 Hz was found 3. Methods
to be optimum as discussed below. The integrators
employed a dark current correction phase which was To examine the feasibility of using ultra violet
effective in further reducing stray light. The output spectroscopy to determine nitrate concentrations in
signal for each wavelength was filtered, with a cut-off the marine environment, a study was carried out at sea
frequency of 2 Hz, resulting in the quoted noise level during the Charles Darwin 84 cruise. A continuous
equivalent to 0.21 mmol l1 of nitrate. system was set up using an AAII type autoanalyser
[18,19]. In addition to the normal NED/sulphanil-
2.1.1.8. Mechanical arrangement. The boxed version amide method for nitrate analysis, a second parallel
was constructed primarily to allow investigations of channel was set up on the autoanalyser using a deu-
relative absorbance at the three wavelengths using terium light source and an appropriate lter
samples in a cuvette introduced into the external (2205 nm) to measure ultra violet absorption
light path between the windows. directly in unmodied samples of seawater taken from
The fully submersible prototype version employs a the same sample cup as the colorimetric samples. The
rugged aluminium alloy housing, rated to 500 bar results illustrated in Fig. 3, indicate an acceptable
external pressure, with separate electrical bulkhead correspondence (p<0.05) between measured nitrate
connectors for power and for signals. The internal and the pak height derived from the UV absorbance
electronic units (light source, detector/signal proces- signal. From Fig. 3, the offset in the UV derived data
sing unit and power supply unit) are contained in (peak height) is thought to be due to absorbance by
separate electrically screened containers, mounted other seawater ions. A gradient of 0.835 as opposed to
on a rigid optical chassis together with the spectro- 1 is probably a result of the equipment used not being
meter unit. The external light path is folded via a fused specically designed for accurate UV measurements.
silica porroprism and its length can be adjusted by the A number of reasons such as the quality of the UV
use of appropriate spacers. Long spacers can be used lter and the quality of the output from the deuterium
to give enhanced sensitivity. A total sample path light source could all have contributed to cause a
length of 66.7 mm is commonly used for open ocean gradient in the data of less than 1.
use. However, this initial investigation was only ever
intended as a test to determine whether further studies
2.2. Nitrate measurements were warranted. From the data displayed in Fig. 3 it
was decided that there was enough evidence to indi-
Using a lock-in amplier with 300 ms integration cate that there were worthwhile grounds for further
time to obtain a DC output signal from the pre- research and the development of a fully working
amplier pulse output, the noise level of the prototype engineered sensor that could be tested in the marine
nitrate sensor was equivalent to less than 0.1% of the environment.
 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177 173

nitrate solution were added. After each addition of

nitrate, the solution in the end cap was stirred to ensure
complete homogeneity. The sensor was allowed to
equilibrate for a period of 2 min between additions.
The concentration of nitrate ranged from 0 to
400 mmol l1 nitrate. The results of these additions
were converted to (220 nm) absorbance values, using
Beers law. The equation of the line was calculated
using the method of least squares. Fig. 4(a) illustrates
the results of nitrate calibrations carried out in both
distilled water and seawater. The detection limit of the
sensor is a function of the signal/noise ratio (see
Fig. 5). The detection limit of the sensor is dened
as the mean of the smallest difference between absor-
Fig. 3. Relationship between AAII UV measured and AAII
chemically measured dissolved nitrate using data collected during
Charles Darwin Cruise CD84 at salinity 35.330.05 psu.

3.1. Calibration of the engineered nitrate sensor

Prior to using the nitrate sensor during oceano-

graphic surveys to measure nitrate concentration in
seawater the response of the sensor to changing con-
centrations of nitrate, salinity and to changing tem-
perature were assessed in the laboratory.

3.1.1. Nitrate calibrations

Following the tests on the prototype UV monitor, a
fully engineered version was manufactured and tted
in a pressure housing to enable measurements to be
made at sea. The three detected wavelengths were
such that the 300 nm detector output (the reference
channel), showed negligible change with a nitrate
solution of 10 mmol l1, relative to distilled water. This
solution resulted in changes of around 60% and 20% in
the 220 and 240 nm detector outputs, respectively.
Referencing or normalisation was achieved by dividing
the output of the measurement channel (220 nm) by the
output of the reference channel.
To facilitate calibration, an end cap was fabricated
which tted over the end of the sensor and allowed the
use of a xed volume of water to be used to calculate
the exact concentrations of added nitrate. The end cap
also provided a convenient method to eliminate all
external uctuations in light intensity.
The end cap was lled with either distilled water or Fig. 4. (a) Nitrate sensor calibrations in low nutrient seawater and
seawater and aliquots of 50 and 100 ml of a 10 mM freshwater, (b) absorbance at 220 nm versus salinity.
174 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177

Fig. 5. Typical output from the Mark I Sensor indicating limit of detection as a function of the signal noise ratio.

bance values that can be resolved by the sensor, above The temperatures chosen were: (i) room tempera-
the signal noise and was calculated to be equivalent to ture, (ii) 208C, and (iii) 58C. Following these tests the
>0.21 mmol l1 NO3. Fig. 5 illustrates a typical output temperature inside the cabinet was varied over a
from the sensor indicating the detection limit of the period of hours to and from 58C to 258C using active
sensor. warming and active cooling. The 300 nm channel
was used to measure the response of the source to
the changes in temperature. Fig. 6 illustrates a typical
3.1.2. Salinity calibrations
temperature stability trace before and after adjust-
To determine the response of the UV sensor to
ment of the ash rate. Before adjustment the
changes in salinity a series of tests was carried out.
300 nm output signal shows greater noise and jumps
The calibration end cap was lled with ``standard''
in output induced by changes in the ambient tempera-
North Atlantic Seawater (nominal salinity 35 psu)
ture. The greatest effect of temperature can be seen
obtained from Ocean Scientic International (Peters-
during the periods of warming and cooling. After
eld, Surrey). The resulting salinity was determined
adjustment the signal noise is not only reduced, but
using a eld salinometer (Model WTW LF325 Con-
the ``step'' features of the signal have also been
ductivity Meter, Weilheim, Germany). This method of
reduced.
determining salinity has been used throughout this
investigation. The solution was then progressively
3.3. Interferences
diluted using distilled water to achieve a range of
salinity (see Fig. 4(b)). After each addition of distilled
The problem of interferences is not fully addressed
water the solution in the end cap was stirred to ensure
in this paper as ongoing research into the viability of
homogeneity. The system was allowed to equilibrate
this system for the measurement of dissolved nitrate is
for a period of two minutes. The resulting salinity and
still proceeding. A number of steps have been taken in
absorbance were then measured.
these initial investigations to optimise the data so that
the problem of interference is reduced, but not com-
3.2. Temperature stability tests pletely eliminated. These include the selection of the
normalization channel. The latest device currently
These tests were achieved by placing the sensor in a being tested allows a 250 nm normalization channel
temperature controlled cabinet and varying the tem- to be used. This has the effect of reducing interference
perature whilst recording the output. of the 220 nm channel from the effects of salinity and
 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177 175

Fig. 6. Effect of temperature on the output voltage of the Mark I Sensor: before and after adjustment of flash rate.

to a lesser extent, the interference caused by organic ing absorbance. The autoanalyser data was then
components in the water column. A future paper plotted against the sensor absorbance values to pro-
devoted to the problem of interferences will be sub- vide a calibration for the sensor.
mitted in due course.

3.4. Initial sea trials 4. Results and discussion

During Charles Darwin cruise CD94 the rst pro- 4.1. Laboratory calibration of the UV nitrate sensor
totype UV sensor was used to measure nitrate con-
centrations throughout the water column. The sensor 4.1.1. Nitrate
was mounted on the CTD rig and lowered down Following the elimination of the effects of tempera-
through the water column. Tests in the laboratory, ture, calibrations were carried out to produce a series
prior to the cruise indicated no effect on the output of equations to derive nitrate concentration from out-
voltage of the sensor from changes in pressure. The put voltage. The output voltage of the sensor was rst
sensor was mounted on a CTD rig and lowered converted to absorbance using the BeerLambert law.
through the water column. Data was recorded on Nitrate concentration was then derived from absor-
the upcast and water samples for nutrient analysis bance by measuring the changes in absorbance pro-
were collected in clean 5l Niskin bottles red at the duced by known concentrations of nitrate and using
appropriate depths. These water samples were ana- the method of least squares to derive a line of best t.
lysed immediately after the sub-samples were taken The line of best t was found to be a second order
from the Niskin bottles using a Chemlab AAII auto- equation. Fig. 4(a) shows a typical plot of absorbance
analyser following the accepted protocol [18,19]. The versus nitrate concentration. Repeated calibration runs
output of the sensor was converted to absorbance by produced similar lines of t, but in each case the slopes
reference to a measurement made with the sensor of the equations were always different. A typical value
immersed in distilled water prior to the cast. The for the between run variation in initial output voltage
output recorded was then inspected to match the time is 3.530.11 V (n15) or a variation of 3% leading to
of collection of the water sample with the correspond- a similar variation in the gradient. This means that
176 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177

during normal use, the sensor would require calibra- wise increase or decrease in the signal output depend-
tion prior to each new run. This difference in calibra- ing on whether the temperature is increasing or
tion slopes does not appear to be related to any decreasing. The temperature effect was found to be
temperature effect as calibrations, whether carried caused by a high ash rate coupled with a low ash
out at a constant temperature or with the temperature energy. This temperature effect was corrected by
changing, produced similar differences. These differ- increasing the ash energy. Increasing the ash energy
ences in the slopes of the calibration lines appear to be resulted in the lamp running at closer to its optimum
related to the ``zero voltage value'' the output rating. Overall power consumption of the unit was a
voltage value when the sensor at the start of each primary consideration in its development and so by
run. Absorbance is calculated using the formula reducing the ash rate from 64 to 16 Hz, the overall
ABS ln V0 =V, where V0 is the initial voltage and power consumption of the unit was kept at a similar
V is the voltage after each addition of nitrate. There- level.
fore, the calculation of absorbance is directly reliant
on this initial voltage value and so if this varies, so 4.3. Initial sea trials of the UV sensor
does the slope of the calibration line.
During Charles Darwin Cruise CD94, data was
4.1.2. Salinity collected from nine geographic positions. Fig. 7
As a result of the sensitivity of the 220 nm channel shows data from station 42. The resulting plot of
to changes in salinity, it was necessary to carry out a measured nitrate concentration versus concentration
series of tests to determine what the relationship was of nitrate calculated from the absorbance measure-
between salinity and absorbance. As with calibrations ments at the corresponding depth. Nitrate was mea-
for nitrate, the lines of best t follow a second order sured directly from bottle samples using an
relationship over the salinity range from 0 to 35 psu. autoanalyser AAII [18,19]. From Fig. 7 it can be seen
More concentrated saline solutions were not tested as that the calculated nitrate concentration corresponds
most oceanic and coastal waters fall in the above range closely to the directly measured nitrate concentration.
which is where the sensor is most likely to be used. The equation of the line relating the two parameters:
This salinity effect means that when calibrating the
UVNO3 0:974  0:04  AANO3
absorbance output from the sensor, additional correc-
tion must be made to the calibration procedure for 0:151  0:17 r 2 0:99
changes in salinity. The resulting change in absor-
bance produced by changing the salinity by 1 psu is
equivalent to 0.52 mmol l1 NO3. Hence, for open
ocean work where changes in salinity are small, often
varying by less than one unit over large spatial scales,
corrections for salinity will be relatively small. How-
ever, if work were to be carried out in rivers and
particularly estuaries where larger changes in the
salinity can occur (from 0 to 35 psu) particular atten-
tion would have to be paid to the relationship between
salinity and absorbance.

4.2. Temperature

The series of calibrations carried out in the labora-

tory using the rst fully engineered prototype of the
nitrate sensor indicated a residual temperature effect
(see Fig. 6). This effect was typically seen in the Fig. 7. AAII measured nitrate versus sensor derived nitrate from
reference channel (300 nm) as an incremental step- data obtained at station 42 during Charles Darwin Cruise CD94.
 M.S. Finch et al. / Analytica Chimica Acta 377 (1998) 167177 177

has a gradient close to 1 and a small y-intercept References

suggesting minimal interference from components
that are not changing in parallel to the changes in [1] J. Alonsochamarro, J. Bartroli, C. Jimenez, Anal. Chim. Acta.
the concentration of nitrate. This provides an indica- 261 (1992) 419423.
[2] M.J. Rocheleau, W.C. Purdy, Analyst 117 (1992)
tion of the potential for a UV based, rapid response
177179.
time sensor for the measurement of dissolved nitrate in [3] R. Lumpp, J. Reichert, H.I. Ache, Sensors and Actuators B 7
the marine environment. (1992) 473475.
[4] P.C. Hauser, S.S.S. Tan, Analyst 118 (1993) 991995.
[5] G.J. Mohr, B. Kovacs, O.S. Wolfbeis, in: Proceedings of the
5. Conclusions Eurotrope II Conference, Florence, Sensor and Actuators B
(1994).
[6] M.A. Stanley, J. Maxwell, M. Forrestal, A.P. Doherty, B.D.
This investigation into the feasibility of producing a MacCraith, D. Diamond, G.V. Johannes, Anal. Chim. Acta
self-contained, low power, high resolution instrument 299 (1994) 8190.
has shown that it is possible to measure dissolved nitrate [7] H.W. Jannasch, K.S. Johnson, C.M. Sakamoto, Anal. Chem.
in situ in seawater using the ultra violet absorbing 66 (1994) 33523361.
[8] N. Ogura, T. Hanya, Nature 212 (1966) 758.
properties of the nitrate ion. Measurements made at
[9] N. Ogura, T. Hanya, Int. J. Oceanology and Limnology 1
sea during Charles Darwin cruise CD94 using the fully (1967) 91102.
engineered sensor showed good agreement with those [10] F.A.J. Armstrong, G.T. Boalch, J. Mar. Biol. Assoc. UK 41
obtainedfromchemicalanalysis, thusproving the ability (1961) 591597.
of the instrument in the open ocean environment. How- [11] P. Foster, A.W. Morris, Water Res. 5 (1971) 1927.
[12] O. Thomas, S. Gallot, Fres. J. Anal. Chem. 338 (1990) 234
ever, there remain a number of problem areas which still
237.
need to be addressed particularly potential problems [13] J. Mertens, D.L. Massart, Bull. Soc. Chim. Belges 80 (1971)
from interferences by the higher concentrations of dis- 151158.
solved organic matter and carbonate in coastal and [14] M. Yaqoob, M. Massoom, J. Chem. Soc. Pakistan 14 (1992)
estuarine waters. Further studies will concentrate on 241243.
[15] O. Thomas, F. Theraulaz, M. Domeizel, C. Massiani, Environ.
determining the extent of these the interferences. A
Technol. 14 (1992) 11871192.
second generation sensor has already been constructed [16] K. Puttanna, E.S.V.P. Rao, Comm. Soil Sci. Plant Anal. 24
with the aim of resolving the above questions allowing (1993) 737743.
the use of this sensor in open ocean environments, which [17] N. Suzuki, R. Koruda, Analyst 112 (1987) 10771079.
wasitisoriginalremit.This unit hasthe potential tomake [18] K. Grasshoff, Methods of Seawater Analysis, Verlag Chemie,
Weinheim, 1976, p. 355.
measurements at six different wavelengths (205, 220,
[19] D.J. Hydes, A manual method for the continuous
235, 250, 265 and 280 nm) which should allow better determination of ammonia, nitratenitrite, phosphate
denition of the spectral changes associated with and silicate in seawater, IOS Report No. 177, 1984,
changes of potential interfering species. p. 40.