1. Introduction
Traditional scanning spectrometers and Brewer spectrophotometers have become the gold standard for measuring spectral solar irradiance and in turn deriving total column ozone at ground level (Gröbner et al. 2006; Fioletov et al. 2005). However, they suffer some disadvantages, such as relatively high cost, susceptibility to movement, and speed of operation. In turn these issues lead to a limited network distribution, potential complications when being used as a campaign instrument, and limited sampling of solar radiation—a typical monitoring instrument will record a full spectrum only once every 30 min, taking several minutes to acquire each spectrum. In contrast solid-state array instruments can overcome these difficulties, being relatively economical to purchase, stable when transported, and being able to acquire full spectra in a period of seconds. But array instruments have their own drawbacks: increased stray light due to their intrinsic single-monochromator design, limited dynamic range, and sensitivity to the ambient temperature. In this study we describe a dual-channel array spectrometer, its characterization, data processing, and deployment for a period of several months to measure simultaneous global and direct solar irradiance. The particular aim is to mitigate against the inherent limitations of using array spectrometers when measuring solar UV and to assess what role such an instrument has in future ground-based monitoring. In this paper we focus on the capabilities of the instrument to derive total column ozone using a collocated Brewer spectrophotometer as a reference.
Total column ozone measurements were first made in the mid-1920s when Dobson attempted spectroscopic measurements in Oxford, United Kingdom (Dobson and Harrison 1926). A network of instruments using Dobson’s design was built up in the following years, many of which are still operating (Dobson 1931; Smedley et al. 2012; Rieder et al. 2011). The ozone network was subsequently expanded after the development of the Brewer spectrophotometer (Brewer 1973) and particular sites now possess long records (e.g., Staehelin et al. 1998). With the discovery of the ozone hole (Farman et al. 1985), the need for adequate network coverage became more pressing, both for ozone monitoring per se, including ground truthing of satellite measurements, and for monitoring of solar UV reaching ground level due to its importance for human exposure and impacts on the biosphere (Ennis 2011; Stocker et al. 2013, and references therein). The addition of a double-monochromator Brewer spectrometer (MkIII) assisted in this regard by extending the global irradiance spectral range up to 363 nm and thus covering most of the critical UV wave band; however, this is necessarily at the expense of deriving total column ozone from direct solar measurements. Monitoring of ozone remains important today due to its influence on the wider climate system by way of radiative forcing, while the local and regional variability of solar UV is still not adequately represented by the network as it stands.
Array instruments have been increasingly used in a number of atmospheric and solar radiation studies in recent years (Bais et al. 2005; Smedley et al. 2007; Kouremeti et al. 2008; Kreuter et al. 2014). Advances in technology mean that this new generation of instruments is becoming more portable and offers certain advantages over traditional scanning instruments, particularly for sites where cost or power can otherwise be restrictive.
One of the inherent limitations of array spectrometer design and layout is the use of a single monochromator, leading to the need to accurately determine the stray light contribution, especially at UV-B wavelengths, where the stray light signal can be greater than the detected signal. This problem was addressed in a theoretical sense by Zong et al. (2006), but it necessitated the use of a tunable laser system to measure the full stray light distribution function. Later Kreuter and Blumthaler (2009) applied this methodology to solar UV measurements and simplified the practical requirements to a single measurement of a laser line.
Array-based instruments have also been successfully applied to trace gas retrievals, which rely on direct solar measurements (notably in Herman et al. 2009; Tzortziou et al. 2012; Gröbner et al. 2014) or by viewing the zenith sky (e.g., Pastel et al. 2014; Constantin et al. 2013), but these studies do not measure the global irradiance incident at ground level. The instrument described here, however, attempts to accommodate these two measurement aims of high-frequency calibrated solar UV irradiance spectra and additionally to derive total column ozone. With a view to practical deployment and extending global networks, the approach here has been to rely on commercially available parts for the major components and therefore simplify the instrument build and setup.
The instrument used in this study was a self-contained dual-channel METCON instrument, temperature stabilized and weatherproofed. The direct solar optics were mounted upon a commercially available sun tracker, while the global irradiance input optics were tripod mounted alongside. The complete instrument setup was deployed at a rooftop city center location at the Manchester surface radiation monitoring site (53.47°N, 2.23°W; 76 m MSL). Measurements of direct solar and global irradiance spectra were made every 1 min from May 2013 to January 2014. Comparisons are made against Brewer spectrophotometer 172, which has been located at the site since 2000.
2. Methods
a. General instrument description
The METCON spectrometer system used in this study consists of a pair of nominally identical, monolithic, parallel-mounted, fixed grating spectrometers, using the same 15-bit data acquisition electronics in a 19-in. rack-mountable unit, itself located within a temperature-stabilized container. Each spectrometer channel is connected to its respective entrance optics via separate 5-m-long 600-μm-diameter optical fibers. The individual monochromators use a 512-pixel diode array detector covering the majority of the UV and visible spectral range (280–700 nm) with a spectral resolution of 2.2 nm.
One spectrometer channel receives its input from a global irradiance (cosine response) entrance optic (Schreder, model UV-J1002), mounted horizontally at a height of 1.3 m. The second channel receives its input from a direct sun optic mounted upon a sun tracker (Eko Instruments, model STR-21; Fig. 1). The direct sun optic is formed from a baffled tube, the front entrance of which is protected by a fused quartz window and an angled cover to prevent water ingress. At the rear of the tube, to reduce the incoming signal, a ground quartz disc is mounted behind the final baffle and in front of the entrance to the optical fiber. The two baffles at either end of the direct sun optic define the field of view (FOV) of the entrance optic, which was also independently measured on a rotating stage in the laboratory to be 2.0° ± 0.1°, independent of wavelength. The sun tracker upon which the direct sun optic is mounted has a pointing accuracy of <0.01°, and it makes use of a four-quadrant photodiode detector to ensure the sun remains in the FOV of the detection optics without the requirement of any additional computer control.
(a) Image of system in operation at Manchester city center monitoring site, showing temperature-stabilized housing, global entrance optics mounted on tripod, and direct solar optics mounted on sun tracker. Also shown is Brewer spectrophotometer 172. (b) Close-up image of direct solar optics on sun tracker; the four-quadrant photodiode detector used to maintain solar tracking can be seen below the entrance optics.
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
The core instrument is housed within a weatherproof aluminum container (0.55 m × 0.55 m × 0.38 m), lined internally with 10-mm thickness of polyethylene insulation (thermal conductivity = 0.044 W m−1 K−1). Sealed ports maintain a weatherproof seal while permitting data transfer cables and optical fibers to pass through the housing skin. The temperature of the system is maintained by way of an air-to-air active thermoelectric cooling system under proportional-integral-derivative (PID) control. The PID parameters 
With the PID temperature control settings chosen in this way, the instrument temperature reached a value within 0.01° of the 25°C set point in <10 min of startup. Laboratory tests for ambient temperatures in the range of −20° to +34°C confirmed that the instrument set point can be maintained to within ±0.01°C during stable conditions and within <0.05°C for rapid changes in ambient temperature (~1°C min−1). The test temperature range covers the full range of expected temperatures in Manchester, although during particularly hot days with direct sun additional shading was used to reduce the heat load.
One disadvantage of array spectrometers is their limited dynamic range in comparison to photomultiplier-tube-based scanning systems, so there is a need to make maximum use of the dynamic range by selecting an integration time where the array is nearly saturated (Blumthaler et al. 2013). Usually there is no prior knowledge of the signal to be measured, so that a two-stage methodology is required. First, a measurement at a short integration time is taken; this is then used to make an estimate of the ideal integration time in the second measurement. For a dual-channel instrument as in this study where a single data acquisition command results in spectra being recorded on both arrays, the situation is more complicated, as the integration time must be chosen so that neither channel is saturated. Rather than continually adjust the integration time for changing atmospheric signal, here a short fixed integration time of 25 ms is used for field measurements. This has the advantage that many repeated unsaturated spectra can be averaged to increase the signal-to-noise ratio, that it requires no settling time, and that it is operationally simple. For the METCON spectrometer system, the signal averaging is split between onboard averaging carried out by the data acquisition electronics and software averaging. This methodology permits us to average approximately 400 individual spectra over a period of 55 s and save a single averaged spectrum for each channel every 1 min. The spectra recorded therefore achieve almost complete temporal coverage and are each representative of the mean conditions during the minute within which they are recorded.
b. Instrument characterization and calibration
Before field deployment the METCON spectrometer system was characterized and calibrated in the laboratory; the various aspects of this are discussed in turn below.
Once the subpixel peak locations had been determined, a third-order polynomial fit was applied using the least squares method. The emission spectra used were taken from a simultaneous measurement of a Hg(Ar) and Ne gas discharge pencil lamps. This combination of lamps provided five single lines across the instrument’s wavelength range, and an additional three multiple lines were included by weighting individual line intensities. All emission line wavelengths were taken from Kramida et al. (2012). Both direct and global irradiance channels showed no pattern in the residuals and the rms difference was at the level of 0.05 nm. Further refinement to the wavelength calibration is carried out by reference to Fraunhofer line structures observable in solar spectra measured in the field (more details are provided at the end of section 2c).
To determine the instrument’s slit function (required as input to the wavelength refinement algorithm), we use a singlet emission line at 296.728 nm, located in the more critical UV wavelength range. The measured slit function at each pixel center is interpolated onto a wavelength grid with a resolution of 0.01 nm over a wavelength range of ±5 nm of the peak. In addition to removal of the dark counts spectrum, the values at the extreme of this wavelength range are used to calculate a linearly varying estimate of the Hg continuum, which is also subtracted.
Measured LSF (dark gray); portion of measured LSF used for SDF correction, not including in-band wavelength range (blue); model-based parameterization fit (red); and analytic power-law fit (green).
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
Example midday global irradiance solar spectrum (4 Jun 2013) before SLC (red trace) and after SLC (black trace).
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
A useful method of describing the capabilities of array spectrometers is the noise-equivalent irradiance (NEI), defined as the standard deviation (SD) of the dark spectra divided by the instrument’s responsivity (Bernhard and Seckmeyer 1999). For the direct sun channel, operating with repeated measurements at an integration time of 25 ms, this is found to be <0.1 mW m−2 nm−1 for wavelengths above 300 nm and <0.04 mW m−2 nm−1 for wavelengths above 315 nm (Fig. 4). For the global irradiance channel, the values are approximately an order of magnitude greater due to the reduced throughput. Applying the stray light correction as described to sample atmospheric spectra, we find also that stray light is removed as expected at wavelengths <290 nm, leaving only noise at the level of the NEI. Turning to the absolute calibration for both channels, this was done with respect to a National Institute of Standards and Technology (NIST)-traceable 1-kW FEL tungsten quartz halogen lamp. The output of this lamp was ensured by maintaining its current to within 0.1% of that specified by NIST and positioning the global entrance optic at the required distance (0.500 m). For the direct optic, the laboratory calibration distance was increased to 5.000 m so that the filament was fully within the FOV and that the NIST-specified irradiance was scaled appropriately. Before the instrument was transferred to the monitoring site, a reference measurement was performed for each channel with a 200-W transfer standard lamp housed within a field-transportable calibration unit. While carrying out the procedure, both global and direct optics were located in machined mountings to ensure a reproduceable positioning w.r.t. the 200-W lamp filament. During the monitoring period, the responsivity of each channel was checked by repeating measurements of the 200-W transfer standard at intervals of 2–5 weeks.
NEI for METCON direct solar channel.
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
Although the instrument housing is temperature stabilized to ±0.01°C under normal operating conditions, the effect of temperature on the instrument characteristics was also investigated. In turn the temperature was set at intervals of 1°C over a range of ±2° around its operating set point, and after a period of stabilization, the following spectra were recorded: dark counts, measurements of Hg(Ar) and Ne emission lamps, and measurements of a quartz halogen lamp. From these spectra the mean thermal wavelength shift was found to 0.002 nm °C−1 and over most of the spectral range the relative change in responsivity is <0.1% °C−1, although this increases to 3% °C−1 at 290 nm. The dark counts increase by 15 counts per second per degree of temperature rise (equivalent to 1.5 mW m−2 nm−1 at 290 nm and 0.2 mW m−2 nm−1 at visible wavelengths). In practice these thermal effects are approximately two orders of magnitude smaller than the change per degree due to the temperature stabilization and therefore do not unduly affect the data produced.
c. Data processing
The raw counts data produced by the instrument are denoted as level 0 data and have had no processing applied, except for averaging. To produce calibrated spectra for both channels, a number of steps are necessary. The first of these is to remove the dark counts spectra (the mean is 167 counts, the range is 5 counts). As the instrument has no internal shutters, datum dark count spectra at the same integration time as used for solar spectra are recorded as part of the calibration process, both initially in the laboratory and during field calibration checks of the responsivity.
After dark subtraction, solar spectra from both channels are scaled to units of counts per second before correcting them for stray light using a stray light distribution matrix based on an analytic power function as discussed in section 2a.
However, when overnight level 0 data are dark corrected, a fingerprint of the spectral structure in the datum dark spectra remains. The mean residual of this fingerprint shows no obvious trend over time scales of minutes and occasional step changes of approximately 10 counts were also observed at intervals of days to weeks. Therefore, a refinement to the dark subtraction process was included as a further processing step whereby the mean residual dark offset in the wavelength range 280–290 nm is used to scale the datum dark spectrum, which is then subtracted. For daytime atmospheric spectra, this second dark subtraction can only be carried out once the stray light correction has been applied. Including this additional dark subtraction stage reduced the residual offset to <0.07 ± 0.18 counts for the chosen integration time. The residual slope contributes a maximum of 0.06 counts to the apparent signal over the UV-A and UV-B wave bands (280–400 nm).
The dark- and stray-light-corrected spectra are then converted to level 1 calibrated spectra by applying the responsivity (calculated from measurements of the quartz halogen lamps, and also corrected for dark and stray light effects). Throughout, the wavelength grid used is that defined by the pixel locations.
Finally, level 1 spectra for both direct and global irradiance channels are passed through a wavelength, bandwidth, and wavelength-resolution homogenization function based on the methodology of Slaper et al. (1995) but extended to include longer wavelengths. This compares ground-level spectra to a high-resolution extraterrestrial spectrum and uses the solar Fraunhofer structure to make fine corrections to the wavelength calibration determined in the laboratory. The extraterrestrial spectrum used by this algorithm (hereinafter referred to as MHP_COKITH) was Kurucz (1994) for wavelengths < 600 nm and calculated from a physically based solar irradiance code [code for solar irradiance (COSI); Shapiro et al. 2010] for wavelengths > 600 nm. The combined spectrum was then adjusted w.r.t. the low-uncertainty satellite-based spectrum of Thuillier et al. (2003) before validation by four ground-based spectroradiometers. Further details can be found in Egli et al. (2012).
d. METCON ozone retrieval
The next stage is to analyze the direct sun channel spectral irradiances in order to retrieve total ozone column amounts. For an instrument with sufficient spectral resolution to resolve finescale spectral features of the trace gas absorption spectrum, differential optical absorption spectroscopy (DOAS) is often employed for this task (Platt and Stutz 2008). In this method the measured spectrum is normalized w.r.t. a reference spectrum (to remove features originating outside the atmosphere), and the resulting transmission spectrum is then split into a continuum and a differential component. A similar process is also applied to spectral cross sections of the species of interest, before fitting them to the differential transmission component to retrieve a slant path column density. This, in turn, can be used to calculate the vertical column density.
Here, however, the instrument’s spectral bandwidth is sufficiently large to mask much of the finescale structure of the ozone absorption spectrum and so a different approach must be used. This restriction also rules out weighting functions as used in the collocated Brewer spectrophotometer retrieval (see details in the following section). For that instrument the weightings of the five spectral channels are selected to remove the influence of SO2, as well as linear affects and aerosol absorption, but for the METCON spectrometer system, the similarities between the SO2 and O3 absorption spectra when convolved with the instrument bandwidth result in a suppression of the ozone signal.
e. Brewer ozone retrieval
The Brewer spectrophotometer used as a comparison instrument in this study has been operational at the Manchester surface radiation monitoring site since 2002. Details of its operation and design can be found in Brewer (1973) and Fioletov et al. (2005). In short Brewer 172 is a double monochromator instrument with inputs that can either measure global irradiance or direct solar radiation over its wavelength range of 286.5–363.5 nm. The normal operating schedule is set to measure global irradiance at the start of each half-hour period, followed by diagnostic tests to ensure wavelength stability, followed by direct sun and zenith sun measurements to calculate TCO. While global irradiance spectra necessitate scanning each wavelength in turn at intervals of 0.5 nm (taking 7 min), ozone measurements are carried out by setting the instrument to a specific wavelength and rapidly rotating a slit mask to acquire counts at five carefully selected wavelengths—306.3, 310.1, 313.5, 316.8, and 320.1 nm—for 20 cycles, taking a total of approximately 3 min.
In contrast to the METCON system, the Brewer spectrophotometer is calibrated for ozone measurements by intercomparison to a traveling standard, itself calibrated using the Langley method at a site with clean and stable atmosphere, typically Mauna Loa in Hawaii. The intercomparison between Brewer 172 and the traveling standard is scheduled every 2 years and is either carried out at the home site, or more recently by transferring Brewer 172 to a prearranged intercomparison campaign at El Arenosillo, Spain. Agreement between Brewer instruments calibrated in this way is typically 1% for direct solar ozone retrievals (Fioletov et al. 2005).
3. Results and discussion
a. Comparison on selected days
The first set of comparative results between the METCON spectrometer system and Brewer 172 presented are example days during the period May 2013–January 2014 when both instruments were located at the Manchester city center monitoring site (Figs. 5a–h). The METCON data are selected to be only those with a running 3-min standard deviation < 2.5 DU, so that both sets of data are filtered with the same methodology (as noted above, Brewer direct sun measurements take approximately 3 min and are selected as reliable using the same 2.5-DU criterion). Days have been chosen to illustrate data produced over a broad range of solar zenith angles (26 May 2013, 31 May 2014), retrieval of TCO data during periods of variable conditions (26 May 2013, 2 August 2013), and the applicability of the instrumental setup during less favorable conditions and solar zenith angles (25 October 2013). The rms difference values quoted in the legend are calculated by linearly interpolating Brewer spectrometer retrievals onto a time grid spacing of 1 min to match that from the METCON system and by rejecting observations more than 20 min before or after valid Brewer measurement times. In addition to the retrieved TCOs for the selected days shown in Figs. 5a,c,e,g, a time series for a range of weighted spectrally integrated quantities (erythemal, vitamin D, UV-A, and UV-B) are shown in the adjoining plots. Action spectra for erythema and vitamin D were taken from CIE (1998, 2006, respectively); the wave band limits used for UV-B were 280–315 nm and for UV-A, 315–400 nm.
Example days showing retrieved ozone from METCON against TCO retrieved via standard Brewer algorithm; additionally weighted global irradiances from METCON channel measured simultaneously. Wavelength limits for UV-A and UV-B bands are 315–400 and 280–315 nm, respectively. Vitamin D action spectrum used is from CIE (2006), and erythemal action spectrum is from CIE (1998). Example days shown are (a),(b) 26 May 2013 [day of year (DOY) 146] rms difference in TCO = 0.62%; (c),(d) 31 May 2013 (DOY 151) rms difference in TCO = 0.78%; (e),(f) 2 Aug 2013 (DOY 214) rms difference in TCO = 1.48%, plot also shows TCO from Brewer 172 for SD < 5; and (g),(h): 25 Oct 2013 (DOY 298) rms difference in TCO = 0.57%, plot also shows TCO from Brewer 172 for SD < 5.
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
A number of points are evident from these data. First is the ability of the METCON system to retrieve high-frequency measurements of ozone in addition to the calibrated irradiance spectra for both direct and global irradiance channels, thus allowing offline processing to calculate spectrally integrated quantities of the user’s choosing. The data collection rates of these quantities are comparable to that used for broadband instruments and therefore array instruments offer a potential replacement in the future. With adequate stray light rejection or correction, array instruments would not suffer from observations being limited to a single weighted quantity, nor are they restricted by a manufacturer’s ability to accurately reproduce a given action spectrum.
Second, and more relevant here, is the higher-frequency TCO data that the METCON instrument can provide. For the four example days, the agreement between the METCON and the Brewer TCO shows rms differences of <1.5% in all cases and <0.8% on three of those occasions, with increased discrepancies at the start and end of the day (see the following section). The METCON system also captures the variation in ozone from hour to hour (from 1200 to 1600 UTC on 31 May 2013, also from 1000 to 1600 UTC on 2 August 2013) that may not necessarily be observed by a Brewer spectrometer running a combined schedule consisting of both ozone and UV spectral measurements. For instance also included in the TCO observations from Brewer 172 on 2 August and 25 October 2013 are those with standard deviations up to 5 DU to show the validity of the temporal change seen in the METCON results. On 2 August the three data points between 1000 and 1100 UTC retrieved by Brewer 172 did not meet the usual 2.5-DU criterion, and on 25 October only the observations close to noon would normally be considered valid by the same test. As such the rapid increase in TCO from 305 DU at 1000 UTC to 325 DU at 1100 UTC on 2 August would not be considered a valid part of the time series; likewise, on 25 October the time series from Brewer 172 would constitute only the pair of measurements at noon and not include the steady increase from 290 DU at 1000 UTC to 305 DU by 1400 UTC.
In part this is due to the METCON’s method of operation and its dual-channel nature. By measuring quasi continuously, the instrument has the potential to retrieve valid TCO observations for any given minute (e.g., during variable cloud conditions after 1200 UTC on 26 May; in contrast there is a lack of valid METCON observations before 1000 UTC on 2 August 2013 during cloudy conditions). Although both instruments have the same 2.5-DU/3-min standard deviation validity criterion applied here, for Brewer 172 direct sun ozone observations are interleaved with spectral scans and diagnostic tests limiting measurement opportunities under some atmospheric conditions.
b. Individual measurement comparison
A comparison between the METCON system and the reference Brewer 172 is shown for all valid individual measurements during the monitoring period (Fig. 6). Again, we apply the same 2.5 DU over the 3-min standard deviation criterion and only select METCON measurements within 20 min of a valid Brewer observation. Considering all valid TCO data points, we find reasonable agreement between the two instruments over a range of ozone column values from 250 to 390 DU, but there is noticeable overestimation of the lowest TCOs and underestimation of the highest column amounts (slope of least squares fit = 0.901, R2 = 0.954, N = 11 589). Vertical trails of points in Fig. 6a result from varying METCON retrievals all ascribed to single Brewer observations. There is a marked dependence on air mass of 6.3 DU (2.0%) per airmass unit (Fig. 6b) with an overall bias of 1.1 DU (0.4%) and scatter of ±6.7 DU (2.2%) for retrievals obtained at 
(a) Comparison between individual TCO measurements retrieved from METCON and Brewer 172 at Manchester city center monitoring site (May 2013–Jan 2014). (b) TCO residuals for same period as a function of air mass. (c) TCO residuals as a function of slant column ozone.
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
c. Daily averages
For purposes of longer-term monitoring, the daily average value of TCO is often required. For both instruments we include all valid measurements in the calculation of the daily mean TCO, except to additionally reject outliers in the METCON retrievals for each day (those with standard deviations > 3.5). For days with valid data from both instruments (Fig. 7a), we find a better agreement but lower correlation than considering all individual data points (slope = 0.939, R2 = 0.908, N = 87). The discrepancy of the slope from a 1:1 line is greater than found between Aura Ozone Monitoring Instrument (OMI) overpass data and Brewer 172 during the monitoring period (Fig. 7b), with an improved correlation in the latter case (slope = 1.023, R2 = 0.954, N = 119). The lower R2 value for the METCON–Brewer is primarily due to a small number of days (~5) exhibiting larger discrepancies that occur during the winter months. This is perhaps to be expected in light of the poorer agreement between the METCON system and Brewer 172 at increased airmass values (section 3b), and the reduced opportunity for measurements near the winter solstice (on 21 December at the site, the lowest airmass achieved is 4.14).
(a) Comparison between daily mean TCO measurements retrieved from METCON and Brewer 172 at Manchester city center monitoring site (May 2013–Jan 2014). (b) Comparison between daily mean TCO measurements retrieved from Brewer 172 at Manchester city center monitoring site against daily mean Aura OMI overpass data for the same period.
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
Some of these days exhibiting higher discrepancies can be identified in Fig. 8 close to the end of the year. While the overall daily and monthly variation of total column ozone from the METCON matches fairly well that from OMI and Brewer 172 from May until November with an rmse of 2.0%, the airmass restriction means that little valid data are obtained during the winter months and that the rmse increases to 9.5% from mid-November to the end of the monitoring period.
(a) Time series of daily mean TCO measurements from METCON system, Brewer spectrophotometer 172, and overpass data from Aura OMI for Manchester city center during monitoring period. (b) Time series of daily mean TCO discrepancies for METCON system and Aura OMI overpass data w.r.t. Brewer 172.
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
d. Residuals
While we find reasonable agreement between the TCO from METCON and Brewer 172 for lower values of airmass and solar zenith angles, retrievals from the METCON system are less accurate during the winter months and for 
In an effort to trace the cause of this residual structure, a subset of data has been reprocessed using a slit function derived from a single laser line as opposed to a Hg(Ar) emission line and also using an alternative formulation for the stray light correction matrix. The mean and spread of the residual structure after removal of the slowly varying aerosol and Rayleigh components, and after subtraction of the retrieved TCO, are shown in Fig. 9. Three alternative processing routes are shown: first, as described in section 2, using a stray light correction based upon a single 405-nm laser line and a slit function derived from the 297-nm Hg(Ar) emission line (denoted “original” in Fig. 9); second, applying a stray light correction (SLC) matrix as calculated from measurements of a tunable laser setup at Physikalisch-Technische Bundesanstalt (Nevas et al. 2009), rather than relying on a single laser line (denoted “PTB stray light correction” in Fig. 9); and third, using a slit function derived from the 405-nm laser line but retaining the original stray light correction matrix (denoted “laser slit function” in Fig. 9). Either of these two changes in principle could improve the TCO retrieval. The advantage of tunable laser stray light correction is due to the inclusion of structure in the SLC matrix < 10−4 in normalized units not seen in the original stray light distribution, which additionally shows a relatively strong variation with peak wavelength. Processing data with the laser-line-derived slit function also has the potential for removing the residual structure, as there is no continuum background to remove as in the Hg(Ar) version, although this has to be balanced by the fact that the peak wavelength lies farther from the spectral region of interest.
Residual structure in transmission spectrum after removal of aerosol, Rayleigh, and TCO effects for data recorded on 4 Jun 2013. Shown are results from three processing routes as described in the text. Mean residual structure is shown as solid line. Dotted–dashed lines indicate ±1SD range.
Citation: Journal of Atmospheric and Oceanic Technology 32, 8; 10.1175/JTECH-D-14-00200.1
These potential advantages notwithstanding, the two alternative processing routes do little to remove the residual structure over the spectral window used (314–327 nm). The only salient differences occur at shorter wavelengths in the tunable laser SLC case. Finally, reprocessing using the MHP-COKITH solar spectrum as the reference in the DOAS procedure (matching that used in the wavelength and bandwidth homogenization function) also did not reduce the residuals. While these results show the robustness of the retrieval processing method, it then leaves the cause as an open question, perhaps related to the subpixel shape of the slit function.
4. Summary
High-frequency global and direct irradiance measurements have been made in Manchester, United Kingdom, over the period May 2013–January 2014 with a dual-channel array spectroradiometer system. The system is based upon a temperature-stabilized 15-bit METCON diode array spectrometer covering the spectral range 280–700 nm, and with the direct solar channel input optics mounted on a commercially available sun tracker. Although it is an unshuttered system, the method of subtracting datum dark spectra from solar measurements, with a second-stage subtraction after stray light correction, enables adequate dark subtraction to be carried out (<0.07 counts residual). A single integration time, chosen such that the instrument is never saturated, allows quasi-continuous measurements to be made. Together with a correction matrix based upon a single 405-nm laser measurement, stray light is reduced below the noise-equivalent irradiance level (found to be <0.1 mW m−2 nm−1 for wavelengths above 300 nm and <0.04 mW m−2 nm−1 for wavelengths above 315 nm).
Retrievals of TCO were made using a simple spectral fitting technique and were found to agree with Brewer measurements on example days to <1.5%. Considering all individual measurements together (range of ozone column values: 250–390 DU), there is reasonable correlation to Brewer TCOs (R2 = 0.954) but noticeable overestimation of the lowest TCOs and underestimation of the highest column amounts (slope of least squares fit = 0.901, N = 11589). There is a marked dependence on air mass of 6.3 DU (2.0%) per airmass unit with an overall bias of 1.1 DU (0.4%) and scatter of ±6.7 DU (2.2%) for retrievals obtained at 
Such an instrumental setup offers a useful alternative to Brewer spectrophotometers and traditional scanning instruments for ground-based measurements of solar irradiance, providing quasi-continuous data coverage and without the temporal-wavelength restriction inherent with scanning instruments, as all wavelengths can be measured simultaneously. Although the TCO retrievals are limited to days with periods of clear sky and lower air mass, for campaigns where these conditions are met, similar instruments can certainly provide a useful role. Incorporating array spectrometers with a smaller bandpass (closer to 1 nm) and higher wavelength sampling into the design would enable finescale structure of ozone and other trace gas absorption spectra to be resolved. With this improvement such an instrument could then make traditional DOAS retrievals over a wider range of solar zenith angles, in addition to simultaneous calibrated global and direct solar irradiance measurements, and thus offer a realistic alternative for long-term monitoring.
Acknowledgments
This study was funded as part of the EMRP ENV03 SolarUV project. The EMRP is jointly funded by the EMRP-participating countries within EURAMET and the European Union.
REFERENCES
Bais, A., and Coauthors, 2005: Intercomparison of solar UV direct irradiance spectral measurements at Izana in June 2005. Ultraviolet Ground- and Space-Based Measurements, Models, and Effects V, G. Bernhard et al., Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 5886), 588609, doi:10.1117/12.619925.
Bass, A. M., and Paur R. J. , 1985: The ultraviolet cross-sections of ozone: I. The measurements. Atmospheric Ozone: Proceedings of the Quadrennial Ozone Symposium, C. Z. Zerefos and A. Ghazi, Eds., Springer, 606–610.
Bernhard, G., and Seckmeyer G. , 1999: Uncertainty of measurements of spectral solar UV irradiance. J. Geophys. Res., 104, 14 321–14 345, doi:10.1029/1999JD900180.
Blumthaler, M., Gröbner J. , Egli L. , and Nevas S. , 2013: A guide to measuring solar UV spectra using array spectroradiometers. AIP Conf. Proc., 1531, 805, doi:10.1063/1.4804892.
Brewer, A. W., 1973: A replacement for the Dobson spectrophotometer? Pure Appl. Geophys., 106–108, 919–927, doi:10.1007/BF00881042.
CIE, 1998: Erythema reference action spectrum and standard erythema dose. Joint ISO/CIE Standard, ISO 17166:1999(E)/CIE S 007-1998, Commission Internationale de l’Eclairage, 8 pp.
CIE, 2006: Action spectrum for the production of previtamin D3 in human skin. CIE Tech. Rep. 174:2006, 12 pp.
Constantin, D.-E., Merlaud A. , Van Roozendael M. , Voiculescu M. , Fayt C. , Hendrick F. , Pinardi G. , and Georgescu L. , 2013: Measurements of tropospheric NO2 in Romania using a zenith-sky mobile DOAS system and comparisons with satellite observations. Sensors, 13, 3922–3940, doi:10.3390/s130303922.
Dobson, G. M. B., 1931: A photoelectric spectrophotometer for measuring the amount of atmospheric ozone. Proc. Phys. Soc., 43, 324–339, doi:10.1088/0959-5309/43/3/308.
Dobson, G. M. B., and Harrison D. N. , 1926: Measurements of the amount of ozone in the earth’s atmosphere and its relation to other geophysical conditions. Proc. Roy. Soc. London, A110, 660–693, doi:10.1098/rspa.1926.0040.
Egli, L., Gröbner J. , and Shapiro A. , 2012: Development of a new high resolution extraterrestrial spectrum. Physikalisch-Meteorologische Observatorium Davos und Weltstrahlungszentrum Annual Rep. 2012, 35.
Ennis, C. A., Ed., 2011: Scientific assessment of ozone depletion: 2010; Pursuant to article 6 of the Montreal Protocol on Substances that Deplete the Ozone Layer. WMO Global Ozone Research and Monitoring Project Rep. 52, 516 pp.
Farman, J. C., Gardiner B. G. , and Shanklin J. D. , 1985: Large losses of total ozone in Antarctica reveal seasonal ClOx/NOx interaction. Nature, 315, 207–210, doi:10.1038/315207a0.
Fioletov, V. E., Kerr J. B. , McElroy C. T. , Wardle D. I. , Savastiouk V. , and Grajnar T. S. , 2005: The Brewer reference triad. Geophys. Res. Lett., 32, L20805, doi:10.1029/2005GL024244.
Gröbner, J., Blumthaler M. , Kazadzis S. , Bais A. , Webb A. , Schreder J. , Seckmeyer G. , and Rembges D. , 2006: Quality assurance of spectral solar UV measurements: Results from 25 UV monitoring sites in Europe, 2002 to 2004. Metrologia, 43, S66–S71, doi:10.1088/0026-1394/43/2/S14.
Gröbner, J., Kouremeti N. , Soder R. , Wasser D. , de Coulon E. , Gyo M. , Dürig F. , and De Coulon E. , 2014: A precision solar spectroradiometer for spectral aerosol optical depth and solar irradiance measurements. UVnet Workshop 2014, Davos, Switzerland, PMOD/WRC. [Available online at http://projects.pmodwrc.ch/env03/images/documents_workshop/Groebner_et_al.pdf.]
Herman, J., Cede A. , Spinei E. , Mount G. , Tzortziou M. , and Abuhassan N. , 2009: NO2 column amounts from ground-based Pandora and MFDOAS spectrometers using the direct-sun DOAS technique: Intercomparisons and application to OMI validation. J. Geophys. Res., 114, D13307, doi:10.1029/2009JD011848.
Kouremeti, N., Bais A. , Kazadzis S. , Blumthaler M. , and Schmitt R. , 2008: Charge-coupled device spectrograph for direct solar irradiance and sky radiance measurements. Appl. Opt., 47, 1594–1607, doi:10.1364/AO.47.001594.
Kramida, A., Ralchenko Yu. , Reader J. , and NIST ASD Team, 2012: NIST atomic spectra database, version 5.0.0. National Institute of Standards and Technology, accessed 23 July 2012. [Available online at http://www.nist.gov/pml/data/asd.cfm.]
Kreuter, A., and Blumthaler M. , 2009: Stray light correction for solar measurements using array spectrometers. Rev. Sci. Instrum., 80, 096108, doi:10.1063/1.3233897.
Kreuter, A., Buras R. , Mayer B. , Webb A. , Kift R. , Bais A. , Kouremeti N. , and Blumthaler M. , 2014: Solar irradiance in the heterogeneous albedo environment of the Arctic coast: Measurements and a 3-D model study. Atmos. Chem. Phys., 14, 5989–6002, doi:10.5194/acp-14-5989-2014.
Kurucz, R. L., 1994: Synthetic infrared spectra. Infrared Solar Physics: Proceedings of the 154th Symposium of the International Astronomical Union, D. M. Rabin, J. T. Jefferies, and C. Lindsey, Eds., International Astronomical Union Symposia, Vol. 154, Kluwer, 523–531.
Mayer, B., and Kylling A. , 2005: Technical note: The libRadtran software package for radiative transfer calculations—Description and examples of use. Atmos. Chem. Phys. Discuss., 5, 1319–1381, doi:10.5194/acpd-5-1319-2005.
Nevas, S., Lindemann M. , Sperling A. , Teuber A. , and Maass R. , 2009: Colorimetry of LEDs with array spectroradiometers. MAPAN: J. Metrol. Soc. India, 24, 153–162, doi:10.1007/s12647-009-0019-5.
Nicolet, M., 1984: On the molecular scattering in the terrestrial atmosphere: An empirical formula for its calculation in the homosphere. Planet. Space Sci., 32, 1467–1468, doi:10.1016/0032-0633(84)90089-8.
Pastel, M., Pommereau J.-P. , Goutail F. , Richter A. , Pazmiño A. , Ionov D. , and Portafaix T. , 2014: Construction of merged satellite total O3 and NO2 time series in the tropics for trend studies and evaluation by comparison to NDACC SAOZ measurements. Atmos. Meas. Tech., 7, 3337–3354, doi:10.5194/amt-7-3337-2014.
Platt, U., and Stutz J. , 2008: Differential Optical Absorption Spectroscopy: Principles and Applications.Physics of Earth and Space Environments, Springer-Verlag, 598 pp.
Reda, I., and Andreas A. , 2003: Solar position algorithm for solar radiation applications. National Renewable Energy Laboratory Tech. Rep. NREL/TP-560-34302, 55 pp.
Rieder, H. E., and Coauthors, 2011: Extreme events in total ozone over the Northern mid-latitudes: An analysis based on long-term data sets from five European ground-based stations. Tellus B, 63, 860–874, doi:10.1111/j.1600-0889.2011.00575.x.
Savastiouk, V., and McElroy C. T. , 2005: Brewer spectrophotometer total ozone measurements made during the 1998 Middle Atmosphere Nitrogen Trend Assessment (MANTRA) campaign. Atmos.–Ocean, 43, 315–324, doi:10.3137/ao.430403.
Shapiro, A. I., Schmutz W. , Schoell M. , Haberreiter M. , and Rozanov E. , 2010: NLTE solar irradiance modeling with the COSI code. Astron. Astrophys., 517, A48, doi:10.1051/0004-6361/200913987.
Sharpe, M. R., and Irish D. , 1978: Stray light in diffraction grating monochromators. Opt. Acta, 25, 861–893, doi:10.1080/713819858.
Shortis, M. R., Clarke T. A. , and Short T. , 1994: Comparison of some techniques for the subpixel location of discrete target images. Videometrics III, S. F. El-Hakim, Ed., International Society for Optical Engineering (SPIE Proceedings, Vol. 2350), 239–250, doi:10.1117/12.189136.
Slaper, H., Blumthaler M. , Huber M. , and Kuik E. , 1995: Comparing ground-level spectrally resolved solar UV measurements using various instruments: A technique resolving effects of wavelength shift and slit width. Geophys. Res. Lett., 22, 2721–2724, doi:10.1029/95GL02824.
Smedley, A. R. D., Webb A. R. , and Saunders C. P. R. , 2007: Application of a diode array spectroradiometer to measuring the spectral scattering properties of cloud types in a laboratory. Atmos. Chem. Phys., 7, 5803–5813, doi:10.5194/acp-7-5803-2007.
Smedley, A. R. D., Rimmer J. S. , Moore D. , Toumi R. , and Webb A. R. , 2012: Total ozone and surface UV trends in the United Kingdom: 1979–2008. Int. J. Climatol., 32, 338–346, doi:10.1002/joc.2275.
Staehelin, J., Kegel R. , and Harris N. R. P. , 1998: Trend analysis of the homogenized total ozone series of Arosa (Switzerland), 1926–1996. J. Geophys. Res., 103, 8389–8399, doi:10.1029/97JD03650.
Stocker, T. F., and Coauthors, 2013: Climate Change 2013: The Physical Science Basis. Cambridge University Press, 1535 pp. [Available online at www.climatechange2013.org/images/report/WG1AR5_ALL_FINAL.pdf.]
Thuillier, G., Hersé M. , Labs D. , Foujols T. , Peetermans W. , Gillotay D. , Simon P. C. , and Mandel H. , 2003: The solar spectral irradiance from 200 to 2400 nm as measured by the SOLSPEC spectrometer from the Atlas and EURECA missions. Sol. Phys., 214, 1–22, doi:10.1023/A:1024048429145.
Tzortziou, M., Herman J. R. , Cede A. , and Abuhassan N. , 2012: High precision, absolute total column ozone measurements from the Pandora spectrometer system: Comparisons with data from a Brewer double monochromator and Aura OMI. J. Geophys. Res., 117, D16303, doi:10.1029/2012JD017814.
Ziegler, J. G., and Nichols N. B. , 1942: Optimum settings for automatic controllers. Trans. ASME, 64, 759–768.
Zong, Y., Brown S. W. , Johnson B. C. , Lykke K. R. , and Ohno Y. , 2006: Simple spectral stray light correction method for array spectroradiometers. Appl. Opt., 45, 1111–1119, doi:10.1364/AO.45.001111.
