Brewer spectrometer total ozone column measurements in Sodankylä

Brewer total ozone column measurements started in Sodankylä in May 1988, 9 months after the signing of The Montreal Protocol. The Brewer instrument has been well maintained and frequently calibrated since then to produce a high-quality ozone time series now spanning more than 25 years. The data have now been uniformly reprocessed between 1988 and 2014. The quality of the data has been assured by automatic data rejection rules as well as by manual checking. Daily mean values calculated from the highestquality direct sun measurements are available 77 % of time with up to 75 measurements per day on clear days. Zenith sky measurements fill another 14 % of the time series and winter months are sparsely covered by moon measurements. The time series provides information to survey the evolution of Arctic ozone layer and can be used as a reference point for assessing other total ozone column measurement practices.


Introduction
Ozone is a molecule consisting of three oxygen atoms (O 3 ). It is produced from oxygen atoms and molecules by photochemistry in the equatorial stratosphere, where the UV radiation is at its strongest. It is then transported to higher latitudes by atmospheric circulation pattern (e.g. Müller, 2012). Ozone is a strong absorber of ultra violet (UV) radiation and thus an important constituent of 15 atmosphere enabling life on earth as we know it. Extensive amount of UV radiation is known to have harmful biological effects. It is known to cause diseases on eyes and skin for example (Lucas et al., 2006) as well as having bad influence on vegetation (Teramura and Ziska, 1996).
In 1984 it was found out that total ozone above Antarctica diminishes during antarctic spring (Farman et al., 1985) and the discovery was later confirmed by analysing satellite data (Bhartia et al.,20 velopement of total ozone amounts have been under close scrutiny since and the state of the ozone layer is revised regularly (WMO, 1985(WMO, , 1988(WMO, , 1989(WMO, , 1991(WMO, , 1995(WMO, , 1998(WMO, , 2003(WMO, , 2007(WMO, , 2011(WMO, , 2014. In the wake of The Montreal Protocol ozone measurements were introduced at the Finnish Meteorologial Institute Arctic Research Center (FMI-ARC) in Sodankylä as both the Brewer measurements and the regular ozone soundings started in 1988. Even though the ozone depletion is more pro-30 nounced and regular in the Antarctis the Arctic has witnessed some really low total ozone amounts in the cold winters and springs of 1990's and 2000's (Rex et al., 1997(Rex et al., , 2002Manney et al., 2011).
The location of FMI-ARC (67.368 • N, 26.633 • E) is well suited for Arctic ozone studies as it often lies within or on the edge of the polar vortex during the time of ozone depletion in the spring.
Sodankylä is one of only seven Brewer stations north of the Arctic Circle (Kipp & Zonen, 2015). 35 As the Arctic ozone is highly variable these few measurement sites are extremely valuable for monitoring the developement of ozone layer. Also the measurement techniques face challenges at high latitudes where the solar elevation angle is very low for long periods. The FMI-ARC has been active in evaluating the effects of low solar elevation angle and Brewer #037 has been serving as a reference instument for satellite validation as well as validation for other ground based measurement methods. 40 As a manned station north of the Arctic circle, FMI-ARC is a rarity, and can produce data of high quality and with infrequent data gaps. In this paper the uniformly processed Brewer dataset from 1988 to 2014 is presented. In section 2 the instrument itself is described and the measurement procedure is explained. The means of maintaining the quality and continuity of the measurements are presented. In section 3 the data flow is shortly described and the data processing algorithm is 45 presented with the rules applied to disregard inaccurate or erroneus data. In section 4 the dataset obtained is demonstrated and some features are highlighted. Some useful diagnostics and statistics of the availability of the data are presented.

50
Total ozone column over Sodankylä has been measured with Brewer spectrophotometer, serial number 037, since May 1988. In September 2012 a new Brewer, serial number 214, was acquired but is still in comparison phase rather than in operational use.
The Brewer instrument stands on a tripod which can be leveled by adjusting the height of all three legs separately. On top of the tripod sits a tracker which maintains the azimuthal orientation of the 55 entrance optics towards the Sun. Accurate tracking of the Sun requires the rotational axis of the tracker to be exactly vertical. The spectrometer itself is mounted on top of the tracker.
The light source is selected with a rotating prism inside the spectrometer. The input can be global radiation through the diffuser on top of the instrument, direct radiation from a point source or radi-the wanted wavelength band.
In place of an exit slit there is a slit mask with 8 different positions. The mask can be rotated fast to go through a set of predefined wavelengths without moving the grating. One of the positions has the exit slit completely shut for determining the dark current of the photomultiplier and one position allows light from two slits to enter the photomultiplier for characterisation purposes. The wavelength 70 resolution for the Brewers is roughly 0.6 nm (FWHM of the instrument function), slightly depending on the individual instrument. The wavelength range is 290 to 325 nm for the Brewer #037 and 286.5 to 365 nm for the Brewer #214.
The photons are counted by a photomultiplier and the counts are recorded. Due to high variability in the intensity along the day there is a filter wheel with set of wavelength neutral filters with different 75 attenuations to choose from depending on the initial intensity check. Another filter wheel is deployed to choose a diffuser, a polarizer or a clear slit depending on the measurement procedure.

Measurement procedure
There are four measurement procedures to measure ozone with Brewer instrument each to suit different conditions. Direct sun (ds) measurement suits clear sky conditions with solar elevation angle 80 higher than 15 degrees. Focused sun (fz) measurements are used for direct sunlight when the elevation angle is below 15 degrees. Zenith sky (zs) measurements are used when the direct sunlight is blocked due to cloudy weather. During the polar winter when the sun elevation stays low for a long period focused moon (fm) measurements are possible when the phase is close to full moon.
In direct sun measurement the azimuth tracker rotates towards the Sun and the zenith prism turns 85 so that the direct light from the Sun is guided to the entrance slit. A diffuser is used to smooth the input light so there is less problems due to slight misalignment. A further attenuation filter might be selected depending on the intensity in a pretest. The grating is kept stationary at a predefined ozone measurement position and the slit mask is rotated to rapidly to select the output wavelengths.
The wavelengths are roughly 306.3, 310.1, 313.5, 316.8 and 320.1 nanometers. The wavelengths are 90 more accurately defined for each individual instrument in the yearly or biennial calibration.
For low solar elevation angles the procedure called focused sun is performed. To enhance the input light intensity the diffuser is omitted but a neutral density filter may be applied to protect the photomultiplier. The forward scattered sky light entering the photomultiplier is estimated, by measuring the intensity just beside the Sun, and subtracted form the intensities measured from the 95 direct sunlight.
When the direct sunlight is blocked by clouds the total ozone column is estimated by measuring scattered light from the zenith. The prism is directed at the zenith angle of 0 degrees and the same wavelengths are measured as in the direct sun measurements. No diffuser is applied but neutral filters are applied when needed. To be less affected by the clouds light goes through a film polarizer that 100 is fixed so that the polarizing axis is parallel with respect to the reference horizontal plane of the instrument (Muthama et al., 1995).
For high latitudes there is a time when the Sun does not rise high enough for any of above mentioned measurement procedures. For that time of the year the only possibility is to measure from the moon. The routine is called focused moon (fm). There is no diffuser used. The intensities are rather 105 low and usually no neutral density filter is needed.
The instrument operation is fully automated. The commands can be given manually but normally the instrument operates on a schedule written by the operator. The schedule consists of command strings set to start when a certain solar zenith angle is reached. In addition to the ozone measurement modes mentioned above, there are several commands to check the status of the motors and follow 110 the changes of instrument characteristics as well as measurement mode for measuring global UV -radiation.

Calibration
The instruments need to be calibrated regularly to keep track of possible changes in their characteristics and the effect of these changes in the ozone measurements. Calibration can be done at the 115 instruments home institute against a well calibrated reference instrument or the instrument can be moved to take part in calibration campaign where the instrument can be calibrated against a reference instrument or by using a Langley extrapolation method. The calibration history of the Brewer #037 is presented in table 1. The main characteristics to be redefined are the extraterrestrial constant (ETC), the ratio of measurements from the internal calibration lamp at ozone measurement wave-120 lengths, the wavelength setting in ozone measurements, the instrument function and the absorption coefficient (α).
The ETC tells what the instrument would measure if there was no atmosphere between the instrument and the Sun. It is dependent on the spectral response of the instrument and the change in ETC is monitored by measuring the intensities at the same wavelengths from an internal calibration  The optimal grating position for ozone measurements is searched by measuring ozone with several positions close to the expected one. The wavelength scale for the micrometer positions is determined by measuring a set of known emission lines from different discharge lamps, for example mercury and cadmium lamps. From these lamp measurements the instrument function is also retrieved. Absorption coefficient is calculated for the exact operational wavelengths as a convolution between the 135 instrument function and the ozone absorption cross section.
In addition to these annual or biennial checks the dead time of the photomultiplier is constantly monitored. The temperature dependence has been characterized already at the factory by the manufacturer. The neutral filter wavelength dependence has also been measured more frequently lately but no corrections have been applied this far.

Quality control
Quality control includes procedures to ensure the instrument is measuring in the ideal manner, the measurements are not stopped or obstructed and all the changes in the instrument are recorded.
To ensure the view is not obstructed and the sunlight is transmitted ideally inside the instrument the window is regularly manually cleaned using isopropyl alcohol. For the snow and rain a blower Because the Brewer is also used for spectral UV measurements the changes in the response are followed very carefully by measuring 1000 W calibration lamp in a dark room every 6 weeks. After each calibration before the instrument returns to the measurement site the tracker is opened and the friction wheel rotating the tracker is cleaned to allow the maximum friction and to ensure there is no 165 slipping and losing steps which would result in poor azimuthal pointing accuracy.
To ensure the maximum intensity and thus the maximum contrast in the measurements the orientation of the tracker and the zenith prism need to be checked regularly. This procedure is called sighting. The instrument is oriented towards the Sun and the entrance slit is viewed from a view port on top of the instrument. If the instrument orientation is perfect the picture of the Sun is centered 170 around the entrance slit. If this is not the case the azimuth tracker and zenith prism motors are commanded from buttons on the side of the instrument casing to turn the instrument to achieve perfect orientation.
For an accurate orientation check a human eye is needed but for a rough check an automatic routine measurement has been developed. In this routine the intensity of the light through the entrance Since the beginning of 2015 a database system called IDEAS has been in use to enhance the continuity and quality of the data. The system parses the raw data files and updates in approximately 10 180 minute intervals. Warnings are raised and sent by e-mail if there are signs in the data for misbehavior of the instrument, for example if the wavelength calibration is too much off of the reference value.
This warning system ensures the erroneus behavior of the instrument is noticed as soon as possible and corrective action can be taken.

Data flow and data storage
Data is collected by the operating computer controlling the Brewer. At this point the data consists of raw counts and preliminary data reduction including not only the measurements of ozone but also the wavelength calibration and stability check information. This data is retrieved to a local server once every 5 minutes. Another server hosting a database system, IDEAS, collects the data from 190 the primary local server every 5 minutes. IDEAS parses the data and is able to send out warning messages by e-mail in case of suspected malfunction of the instrument. This way the any problems in the measurements or in the instrument are noticed within 10 minutes of the start of the problem.
The processing of the final products that are sent to World Ozone and UV Data Center (WOUDC) and to the FMI-ARC database litdb.fmi.fi still require human interaction. The data is collected from 195 the primary local server to a personal computer and processed with a software described in section 3.2. There are rules to assure the quality of the data (section 3.3) but some erroneus measurements must still be filtered out by human eye. At the moment the data is available in WOUDC database until May 2010 and the update for Sodankylä data will be made when the quality assurance is finished.
Finland is participating in Eubrewnet COST action (COST-ES1207, www.eubrewnet.org) and the 200 Brewer data is sent from primary local server to Eubrewnet database every 20 minutes. The database is under developement and not fully operational at this moment. When operational the database will receive Brewer raw data from all the participating countries and the final products will be processed by a central computer. This way way the data will be coherent through the whole network.

205
For all ozone measurement procedures the raw data includes total counts from the photomultiplier for seven positions of the slit mask, F0 to F6. To retrieve vertical ozone column O3Brewer -software, written in Delphi language by Martin Stanek, is used (Stanek, 2015). The retrieval process is described below. Before the physical modeling of the absorption itself the raw counts are first turned into count 210 rates C (counts/second) (SCI-TEC Instruments Inc., 1999). First the dark counts are removed and the integration time for each slit exposure is taken into account in term 2/IT and CY is the number of slit exposure cycles. Dark counts, F 1 , are measured at slit mask position 1, where the exit slit is closed. The count rate, C i , for slit mask position i, becomes IT × CY , for position i = 2..6. (1)

215
The photomultiplier has a dead time (DT) , a time it takes it to recover from a hit by a photon to be able to count another photon. The effect is assumed to follow the Poisson statistics and hence the equation where C i,m is the measured count rate and C i,t is the true count rate. The first estimate for the true 220 count rate is given by After this a better estimate is calculated by plugging the first estimate in the original equation This is repeated 8 times and the result is expected to converge until then, making C i,t,9 the final 225 estimate for C i,t .
Later will be shown that, when calculating ozone we are dealing with logarithms of irradiance ratios. For this reason the count rates are converted to logarithmic scale, C i,l C i,l = 10 4 × log 10 (C i,t ) The response of the instrument may change according to the change in temperature inside the instru-230 ment. The effect is assumed linear and has been characterized at the factory. Corrected count rate, where T is temperature and TC i is the wavelength dependent temperature correction coefficient.
The final instrumental correction will be neutral density correction. This actually does not affect 235 the ozone calculation as the attenuation is assumed to be wavelength independent. The effect of filter non-neutrality has been found very small in recent calibrations. The final count, rate C i,f , will be where NF is the 10000 times the ten based logarithm of attenuation ratio of the filter.
The physical model of retrieval from final count rates to total ozone column, well described by 240 Savastiouk (2006), is based on Beer-Lambert -law considering Rayleigh scattering and scattering by aerosols as well as absorption by sulphur dioxide and ozone. The data from slit mask position 0 is not taken into account so using the other five (positions 2 to 6) wavelength a set of wavelength dependent weighting coefficients are found such that the ozone calculus is in theory not affected by sulphur dioxide or aerosols. The final count rates are corrected for Rayleigh scattering according to where BE i is Rayleigh correction coefficient for slit mask position i. The BE i used are the original ones calculated by the instrument designer (citation). P is the pressure of the instrument location, set to 1000 mbar for Sodankylä. µ r is the air mass factor for Rayleigh scattering, calculated from where A is the solar zenith angle at the time of the measurement. The ratio, R6, affected only by ozone is calculated from Rayleigh corrected count rates, Ci, as Total ozone column, X O3 , is retrieved from the ratio as where extraterrestrial constant, ETC, and absorption coefficient, α, are constants determined in calibration. SL is the change in R6 value measured from a calibration lamp. For this a reference value R6 sl,ref is determined during calibration campaign and the changes are followed by measuring this ratio, R6 sl , several times per day. The lamp is assumed not to change and so the changes should 260 reflect changes in the ETC. The assumption is then checked at the next calibration when also the new reference value is determined. The correction factor, SL, is based on daily mean value R6 sl as Ozone airmass factor, µ O3 , describes the path length that the light has to travel compared to a thickness of the ozone layer right above us. For the calculation ozone is considered to lie in a thin layer where A is the solar zenith angle. For the focused sun measurements, fz, the algorithm is the same. Before calculating the R6 however, the count rate measured 0.5 degrees off of the Sun is deducted from the count rate measured 270 from the direct sun. This is made to counter the effect of multiply scattered light reaching the instrument from the same direction than the direct light (Josefsson, 1992). For moon measurements the algorithm is identical to direct sun algorithm.
The processing of the zenith sky measurements is based on the comparison of zenith sky (zs) and direct sun measurements (ds) (Muthama et al., 1995). A polynomial will be fitted between a number 275 of near simultaneous zs and ds measurements so that using the X O3 from ds measurements a best possible set of coefficients a to i are found for relation R6 zs −ET C = a+b×µ+c×µ 2 +d×X +e×Xµ+f ×Xµ 2 +g ×X 2 +h×X 2 µ+i×X 2 µ 2 . (14) These coefficients depend on the instrument and on the measurement location. For Brewer #037 the fitting of the coefficients, the so called sky chart, has been done by Karhu (1995).

280
The daily value is calculated as an average of measurements that meet the filtering criteria (see 3. 3). Only one type of measurement is used for one daily value. The measurement types are ranked by their expected quality so that the order is ds, fz, zs, fm. The measurements of lower rank are only used if there is no measurements at the higher level. The ranking is justified by the assumptions that focused sun and moon measurements are more prone to errors due to small deviations in pointing 285 accuracy and zs is more prone to errors due to retrieval algorithm based on a non physical model.

Quality assurance
To assure the quality of the data some automatic data filtering rules have been obeyed. The measurement routine for one ozone value has five consecutive measurements. The standard deviation among these five measurements is not allowed to be more than 3 Dobson units (DU) in ds-, zs-or 290 fz-measurements. For moon measurements the deviation is allowed to be 5 Dobson units. The air mass factor has been limited to 3.8 for ds-and to 3.0 for zs-measurements to avoid the effects of stray light but to allow measurements to start as early as possible in the spring. For focused sun measurements the air mass limit is set to 6 and for moon measurements to 3.5. Before submitting the data to database it is also inspected by eye for obvious spikes or other suspicious behavior. The random 295 errors of individual observations are within ±1% in about 90% of all measurements (Fioletov et al., 2005).

Demonstration of the dataset
The daily values of total ozone column over Sodankylä are presented as a time series in figure 1.  Sodankylä Brewer data has been used in several satellite comparisons (e.g. Kyrö, 1993;Balis et al., 2007;Damiani et al., 2012) and in comparison to other ground based instruments (e.g. Vigouroux et al., 2008). Sodankylä also hosted two total ozone column intercomparison cam-

Conclusions
There is a dataset of 26 years of uniformly processed Brewer ozone data from Sodankylä and the processing algorithm with the data acceptance rules have been described in this paper. The instrument has been well maintained and regularly calibrated during its time of operation. An automatic warning system has been built to allow fast response to any unexpected stops in the measurements   and to follow that the instrument is operating in an ideal way. The quality assured data will be available in WOUDC database as well as in litdb.fmi.fi. Data is also sent to Eubrewnet database where the Brewer community will set up best possible data processing algorithm to achieve coherent total ozone data across the Eubrewnet network. Sodankylä data is valuable for Arctic ozone studies and for validating other ozone measurements at high latitudes.  Figure 6. Comparison of Brewer #037 direct sun measurements and OMI satellite data using TOMS retrieval algorithm. The Brewer reference value was the closest direct sun measurement with a limit of maximum of one hour difference in the measurement times.
Acknowledgements. Authors thank Johanna Tamminen for the valuable comments during the writing process.
Authors also thank Martin Stanek for the review and comments on the ozone retrieval algorithm as well as for clarification on several data handling issues of Brewer data. Authors are grateful to Volodya Savastiouk for setting up the IDEAS database software and for the help on clarifying any questions we have had about Brewer software or hardware issues.