The determination of aerosol properties, especially the
aerosol optical depth (AOD) in the ultraviolet (UV) wavelength region, is of
great importance for understanding the climatological variability of UV
radiation. However, operational retrievals of AOD at the biologically most
harmful wavelengths in the UVB are currently only made at very few places.
This paper reports on the UVPFR (UV precision filter radiometer)
sunphotometer, a stable and robust instrument that can be used for AOD
retrievals at four UV wavelengths. Instrument characteristics and results of
Langley calibrations at a high-altitude site were presented. It was shown
that due to the relatively wide spectral response functions of the UVPFR,
the calibration constants (
One of the most important atmospheric processes related to solar ultraviolet (UV) attenuation is the absorption and scattering of solar radiation by aerosols (IPCC, 2013; Madronich et al., 2015; UNEP, 2010). The effect of aerosols on solar UV radiation is important as it is linked with the impact on UV radiation on human health (Rieder et al., 2008; Cordero et al., 2009), atmospheric chemistry (e.g., Gerasopoulos et al., 2012) and the biosphere (Diffey, 1991). Especially in heavily polluted areas, analysis of past data series shows that the decrease of UVB (wavelength range 280–315 nm) radiation due to aerosol attenuation can become larger than the expected increase of UVB radiation due to the declining ozone levels (e.g., Meleti et al., 2009; Zerefos et al., 2012; De Bock et al., 2014). Thus, the determination of aerosol properties, especially the aerosol optical depth (AOD) in both the UVA (315–400 nm) and UVB wavelength region, is of great importance in order to understand the climatological variability of UV radiation. However, even though the aerosol attenuation on the solar UVB wavelength range is higher than the one at longer wavelengths, most of the available surface-based and satellite AOD measurements are related to the UVA, visible (VIS) and near-infrared (NIR) ranges because they represent the part of the spectrum associated with the higher solar irradiance levels reaching the Earth's surface.
Concerning AOD measurements at the UV range, the largest surface-based aerosol sunphotometric network, the Aerosol Robotic Network (AERONET) (Holben et al., 1998), includes a number of instruments that are able to measure AOD at 340 and 380 nm. In addition, the Global Atmospheric Watch precision filter radiometer network (GAW-PFR) provides AOD at 368 nm (Wehrli, 2008). In order to extrapolate the UVA and VIS AOD to the UVB the spectral dependence and the aerosol type is needed. This is because the simple Ångström power law includes a wavelength dependence that is related to the different aerosol types, potentially leading to very poor accuracy of AOD in the UVB determined from extrapolation of accurate AOD values in the VIS to NIR range of the spectrum (Li et al., 2012).
Only a few instruments such as the UV multifilter radiometer (UVMFR) (Krotkov et al., 2005; Corr et al., 2009; Kazadzis et al., 2016) can be used to provide AOD retrievals in the UVB wavelength range. The Brewer spectrophotometer is an instrument initially designed for providing total column ozone (TCO) measurements based on the use of direct sun (DS) irradiance measured at specific wavelengths in the short UVA and in the UVB range (e.g., Kerr et al., 1985). During the past years, several attempts have been presented in the literature, which showed the use of the abovementioned Brewer measurements in order to retrieve AOD in the UVB (e.g., Marenco et al., 1997, 2002; Cheymol and De Backer, 2003; Cheymol et al., 2006; Gröbner and Meleti, 2004; Kazadzis et al., 2005, 2007; Meleti et al., 2009; De Bock et al., 2010, 2014; Kumharn et al., 2012). In addition, Arola and Koskela (2004) have discussed the uncertainties and possible systematic errors linked with the Brewer related DS retrieval for AOD.
Recently, the European COST project EUBREWNET (European Brewer network,
The instrument in focus of this study is the UVPFR sunphotometer, which is a
modified version of the precision filter radiometer (PFR) designed and built
in the late 1990s at PMOD/WRC in Davos, Switzerland. It measures direct
solar irradiance at the four nominal wavelengths 305, 311, 318 and 332 nm at
bandwidths of approximately 1.0–1.3 nm at full width at half maximum (FWHM).
The detectors are operated in a controlled environment and are exposed to
solar radiation only during actual measurements. A Peltier thermostat
maintains the ion-assisted deposition filters and silicon detectors at a
constant (
A recent improvement of the instrument was the addition of an UG11 low-pass filter at all four channels to remove out-of-band leakage that had been observed in the original version of the UVPFR.
The spectral response functions of the UVPFR no. 1001, used in this study,
were measured in the laboratory at PMOD/WRC in February 2016, using an
EKSPLA NT 200 tuneable laser (
Wavelength characteristics of UVPFR no. 1001 based on laboratory measurements in February 2016. The third column shows effective central wavelength resulting from convolving the spectral response function with an extraterrestrial solar spectrum.
In order to perform DS measurements, the UVPFR is mounted on a sun tracker so that it is continuously pointing to the Sun. The four photometric channels are measured simultaneously by a commercial data logger system (Campbell Scientific CR10X) with 13 bit resolution. Automatic signal ranging within the PFR and logger system is used to increase the dynamic range to 16 bit. The logger clock is frequently updated to be accurate within 1 s. Signal measurements made at full minutes are averages of 10 samples for each channel made over a total duration of 1.25 s and can be considered as instantaneous values.
The full field of view of the instrument is 2.5
The standard PFR has the same specifications as the UVPFR except that the PFR measures at the nominal wavelengths 368, 412, 500 and 862 nm with a 5 nm FWHM bandwidth. The PFR, together with an evaluation of different calibration methods, has been described in detail by Wehrli (2000).
The Brewer spectrophotometer (Kerr et al., 1985) is an instrument designed
for automated measurements of solar UV irradiance and through them for the
retrieval of atmospheric ozone (total column and vertical profile) and
sulfur dioxide (SO
AOD can be retrieved from the standard DS measurements (e.g., Cheymol and De Backer, 2003) or spectral DS measurements (Kazadzis et al., 2007). In the current study AOD retrievals from the double monochromator Brewer MkIII no. 163 at the wavelengths 306.3, 310.1, 313.5, 316.8 and 320.0 have been used.
The UVPFR was calibrated at the Izaña Atmospheric Observatory (IZO) on
the island of Tenerife (28.31
The home site of the UVPFR is at PMOD/WRC, which is located in Davos in the
Swiss Alps (46.81
Calibration of reference sunphotometers with the Langley technique is preferably performed at high-altitude stations since it requires low and stable aerosol load (e.g., Shaw, 1983). Difficulties with Langley calibration at a low-altitude and urban site, when calibration at a high altitude is not possible, have been discussed by Arola and Koskela (2004) and were recently demonstrated by Diémoz et al. (2016). For instruments measuring at wavelengths affected by absorption in ozone, an ideally stable total ozone amount is needed during the Langley related period of measurements. These requirements can be relatively frequently fulfilled at IZO.
During May to August 2015 the UVPFR no. 1001 was operated at the IZO station, with the exception of the time period from the 20 May to 10 June. In September 2016 the next Langley calibration at IZO was performed. In addition to the favorable measurement conditions an advantage of the IZO station is the co-location with other instruments, such as Brewer spectrophotometers and standard PFR sunphotometers. These instruments measure among others TCO and AOD in the 368–862 nm range, respectively. These additional variables are highly valuable and help to determine whether measurement conditions during half days (mornings or afternoons) have been suitable for the so-called Langley plot calibrations.
The classic Langley method to determine the calibration constant
where the wavelength-dependent quantities ln(
Using a single, common air mass
Results of all the Langley plot calibrations at IZO during
May–August 2015. The final
From the quality of the linear fit of the Langley plot and using TCO and AOD
data from the other instruments, the selection of exact air mass range
(within 1.2–2.9) and validity of the Langley plot events were mainly based
on subjective judging by the analyst. During the periods when the UVPFR
no. 1001 was at IZO, 27 accepted Langley plot occasions were found in 2015
and 11 were found in 2016. The resulting
In principle, TCO can also be estimated by the UVPFR itself. It is, however,
believed that Brewer spectrophotometers are superior to the UVPFR in TCO
determination. At the same time, it is important to remember that the
Langley plot calibration of the UVPFR becomes dependent on the ozone
measurements when these are used to correct for ozone changes during Langley
events. In case there is a small air-mass-dependent error in the Brewer
(triad) measurements, there will also be an error in the UVPFR
Langley calibration results for UVPFR no. 1001 at Izaña 2015 and
2016, together with calculated
As is clear from Fig. 1, taking just a single Langley plot event is not
enough, if high accuracy accompanied with uncertainty estimation is aimed
for. The (experimental) standard deviation of the
The final calibration values are shown in Table 2. Over the period of slightly more
than 1 year between the calibrations at IZO, the decrease in
sensitivity was as small as
Also, the PFR-N24 used in this study was calibrated by the refined Langley method in 2015. This was done at the high-altitude station at the Mauna Loa Observatory, Hawaii. After this calibration, the PFR-N24 was included as a new member in the WORCC PFR triad operated at PMOD/WRC.
Both Brewer no. 163 and the UVPFR no. 1001 participated in the 10th
RBCC-E campaign 27 May–4 June 2015 at the INTA (Instituto de Técnica
Aerospacial) El Arenosillo station in southern Spain (37.10
Due to the large variation with wavelength of ozone absorption in the UV, spectral transmission measurements need to be performed at well-defined and narrow passbands in this wavelength region. The bandwidth of the UVPFR filters, on the order of 1 nm, is significantly narrower than for standard VIS–NIR sunphotometers, but about twice as wide as the slit functions of Brewer spectrophotometers. Therefore, the effect of finite bandwidths was investigated for the UVPFR. Effective central wavelengths and FWHM are given in Table 1.
Due to the very strong increase in ozone absorption with decreasing wavelength, and hence its stronger change with air mass at the shorter wavelength side of the filter band passes, this leads to an increase in the effective wavelengths seen by the UVPFR when the air mass increases. This in turn leads to errors in the extrapolation to zero air mass during a Langley calibration. The FWHM effect has been quantified with simple but high resolution modeling with the Bouguer–Lambert–Beer law.
Using an extraterrestrial solar spectrum of 0.05 nm resolution with a 0.01 nm increment (Egli et al., 2013), together with ozone absorption coefficients for 223 K from Molecular Spectroscopy Lab, Institute of Environmental Physics (IUP), University of Bremen (Serdyuchenko et al., 2011), direct solar irradiance spectra at the surface were calculated for different air masses and TCO amounts. The IUP ozone cross sections were chosen by convenience since they matched the 0.01 nm resolution of the used extraterrestrial solar spectrum. This was not the case for the cross sections by Bass and Paur (1985) which are used in the operational TCO determinations by the Brewers, as well as for the AOD determinations with both the UVPFR and the Brewer, discussed later in this study. It is assumed that the choice of ozone cross sections does not significantly affect the modeled FWHM effects within their estimated uncertainty.
The aerosol extinction was modeled using the Ångström law,
AOD
In the calculations a station pressure of 770 hPa was used, which is close
to the average value at the IZO station during the evaluated Langley plot
events. Effective ozone altitude was set to 25 and 22 km for calculations
corresponding to measurements at Izaña and Davos, respectively. These
values on ozone altitude were also used for the Langley calibrations at IZO
(Sect. 2.2) and for the AOD determinations in Davos (Sect. 5). For the
relative optical air mass for ozone absorption the algorithm/formula by
Komhyr et al. (1989) was used. Rayleigh optical depth,
Results of Langley plots of the simulated UVPFR direct irradiances,
Calculated change in effective ozone optical depth with air mass due to the UVPFR filter bandwidths.
Not only the derived
The apparent change in ozone optical depth is not a perfect linear function
with air mass. With little loss in accuracy, the ozone optical depth
correction is still estimated as a linear function of
All in all, at an air mass of 2 and TCO amount of 350 DU the
effect of the FWHM corrections on derived AOD at 305 nm is about
A more detailed form of the Bouguer–Lambert–Beer law in Eq. (1), valid at a
(monochromatic) UVPFR wavelength
The ozone optical depth,
Indeed, using different datasets on ozone cross sections would result in different AOD values, especially at the shortest wavelengths. The effect of different cross sections is not further investigated here. In any case the same cross sections should be used for both TCO and AOD determinations.
The ozone amounts taken from a collocated Brewer are calculated with Rayleigh scattering coefficients according to Nicolet (1984), instead of the standard ones used in the operational Brewer program. As an example, for Brewer no. 163 in Davos the corrected TCO values are 2.7 DU lower than the operational ones. Using Rayleigh scattering coefficients calculated according to Bodhaine et al. (1999) gives similar results, within 0.1 DU, as with the coefficients according to Nicolet (1984).
The other parameters on the right-hand side of Eq. (8) are calculated mainly
from position and time and the applied air mass formulas were given in Sect. 2.3 above. As above, Rayleigh optical
depth,
AOD values calculated by Eq. (8) are only valid for times when there are no
clouds in front of the Sun. The cloud screening applied in this study is
based on the method by Alexandrov et al. (2004) with modifications to fit
the UVPFR measurements. The Alexandrov et al. (2004) cloud screening
algorithm was developed for optical depth measurements at 870 nm wavelength
and for a sampling interval of 20 s. Stability tests were performed
with a 15-measurement window, which consequently spanned over 5 min.
For the cloud screening, optical depth at the longest UVPFR wavelength (332 nm) was used. Since the UVPFR only takes measurements once every minute, only
five measurements were used for the stability check. Also, the threshold for
the inhomogeneity parameter
An uncertainty analysis according to GUM (GUM, 2008) has been made for the
AOD values retrieved from a UVPFR sunphotometer. Assume we have an arbitrary
measurand with its estimated value,
At the shortest UVPFR wavelengths the most dominant source of uncertainty in
AOD determinations originates from the uncertainty in ozone optical depth. In
the
For the ozone amount 1 % (1
The estimated uncertainty in effective ozone temperature is a function of
latitude and time of the year. At low latitudes the day-to-day variation in
effective ozone temperature is low. From 30
The total standard uncertainty connected ozone optical depth is calculated as
In the
Contributing to the
Frequency distribution of
Uncertainties in the
As mentioned above, in the calculations of
As mentioned earlier, there is a large uncertainty in ozone optical depth at
wavelengths with high ozone absorption. While this adds some uncertainty to
Finally, a drift term of
The standard uncertainty of the Rayleigh optical depth,
Uncertainty in voltage readings,
This way, the estimated additional diffuse light entering the instrument
does not result in a bias of calibration through Langley plots, since it is
not dependent on air mass. In reality, as suggested by Arola and Koskela (2004), diffuse light could introduce a significant negative bias in Langley
plot results at UVB wavelengths under high AOD conditions. In our case, the
average UVB AOD during the Langley calibrations of the UVPFR at Izaña
was only about 0.05. At the same time, the average of Ångström's
wavelength exponent calculated from AOD in the 368–862 nm range was about
1.5 during the Langley plot events, which indicates that the aerosol forward
scattering was not particularly high. In addition, the maximum air mass
during Langley plots never exceeded 3. It is therefore assumed that the
diffuse light influence was very small on the UVPFR calibrations. Hence,
this source of uncertainty was not specifically taken into account in the
already conservative
Absorption in NO
At polluted sites with NO
For the calculation of the standard uncertainty due to neglecting absorption
in SO
A SO
Not taking NO
Based on comparison between
The actual vertical distribution of gases and aerosol particles in the
atmosphere is not known. This introduces uncertainties in the relative
optical air masses used for AOD calculation. As necessary input to the
air mass algorithms the true or apparent solar zenith angle,
SZA
The air mass term thought to be the least uncertain is the air mass for
Rayleigh scattering,
Estimated expanded uncertainty,
In this study the vertical aerosol particle distribution is assumed to be
more concentrated near the ground than the vertical distribution of the
molecules of the air, leading to
In Fig. 4 the estimated expanded uncertainty (
Clearly, the dominant part of the AOD uncertainty is caused by the
uncertainty in the ozone optical depth at the three shortest wavelengths. As
the absorption by ozone decreases with wavelength the size of the
Major contributions to these uncertainties come from (unknown) systematic
effects. Therefore, the uncertainty of average AOD values based on a number
of measurements,
It is believed that the most dominant uncertainties have been included in the
current analysis. However, in addition to neglecting the effect of correlated
variables, there are still some uncertainty sources which have not been taken
into account when calculating the total uncertainty. For example, no
information on potential nonlinearity in the voltage output from the UVPFR
has been found. This source of uncertainty is assumed to be small and has
therefore been neglected. The pointing accuracy is monitored with the UVPFR.
Normally, the pointing error is
For the two shortest wavelengths the estimated AOD uncertainties are very
high, which of course is not very encouraging. At the same time, the
estimated 2
After the calibration at Izaña in summer 2015 the UVPFR has been operated about 2 months during autumn 2015 and spring 2016, respectively, at PMOD/WRC in Davos. These measurements were analyzed to show an example of AOD determination with the UVPFR. The calibration results from 2015 have been used for the whole period in Davos.
As a first result cloud-screened 1 min AOD values from UVPFR no. 1001 during the day on 12 October 2015 in Davos are shown in Fig. 5. AODs from PFR-N24 (wavelengths 368, 412, 500 and 862 nm) are also shown in the figure. During this day the turbidity in Davos was very low, which rather frequently occurs at high-altitude stations. Under these conditions the effect of the FWHM corrections of the UVPFR data becomes extra important. From around 09:30 UTC, the NIR to the UVB range AOD increases with decreasing wavelength, according to the results in Fig. 5. Without the FWHM corrections this would not have been the case in the UV. For the whole day, AOD at 305 nm would have been lower than at 332 nm and often even lower than at 368 nm. AOD at 311 nm would also have been lower than at 332 nm part of the day. Based on these results for low-turbidity conditions it is assumed that AODs from the UVPFR really do become more realistic when the proposed FWHM corrections are applied.
The 1 min AOD determined by UVPFR no. 1001 (dots) and PFR-N24 (lines) on the 12 October 2015 in Davos. Data points disturbed by clouds have been removed.
Daily mean AOD at 305 and 332 nm in Davos (left) and mean of daily means of AOD during the whole study from the UVPFR and a standard PFR (right).
Daily means of cloud-screened 1 min AOD values at the 305 and 332 nm wavelengths are shown to the left in Fig. 6. The averages of the logarithm of daily mean AODs at all the UVPFR wavelengths, as well as from the PFR-N24, are shown to the right. Clearly, very low values of AOD are often experienced over Davos, even at UVB wavelengths. Especially during autumn 2015 this was the case. At the end of October and in November AOD at 305 nm were mostly measured lower than at 332 nm by up to 0.02 units of optical depth. In spring 2016, the turbidity conditions were higher and more variable. The average AOD values for the whole period were measured lower than 0.1 at all four UVPFR wavelengths. While the average of daily AOD was highest at the shortest UVPFR wavelength, the average of the logarithm of daily values was actually smallest at the 305 nm wavelength due to the many very low values in autumn of 2015, which get more weight when using the logarithm of the AOD.
During the measurements in Davos the average AOD values in the UVB are not very well estimated by extrapolating AOD values at the UVA–NIR wavelengths using the common Ångström relation, represented by the (red) full line in the right panel of Fig. 6. To be more specific, extrapolated AOD at UV wavelengths is overestimated. Using a second-order fit in the log–log space, earlier introduced by Eck et al. (1999), leads to better results, at least for the two longest UVPFR wavelengths. As shown above, the uncertainties of the UV AOD values are, however, considerable and the AOD values measured by the UVPFR are not significantly different from any of the extrapolated values in this low-turbidity case.
Differences in AOD, UVPFR–Brewer, at Brewer wavelengths for measurements during autumn 2015 and spring 2016 in Davos. Percentage of differences within WMO traceability limits is given in each graph.
In the calibration section (Sect. 2.2) the UVPFR sensitivity was shown to be satisfactorily stable over 1 year. As an additional stability and consistency check, AOD from the UVPFR has been compared to AOD derived from a Brewer spectrophotometer. At PMOD/WRC, the Brewer MkIII no. 163 is operated. This instrument provided the ozone values used in the AOD calculations based on spectral transmission data from the UVPFR in Davos.
Using the calibration against the UVPFR during the RBCC-E campaign in 2015, UV AOD has been determined from Brewer no. 163 during its measurements in Davos. Also, a small temperature correction was applied to the Brewer direct irradiance readings as well as a polarization correction suggested by Cede et al. (2006).
The comparison of AOD from Brewer no. 163 and the UVPFR no. 1001 in Davos is
shown in Fig. 7. Since the UVPFR has the highest sampling rate (1 measurement min
Individual AOD differences (UVPFR–Brewer) for cloud-screened and near-simultaneous measurements are shown in Fig. 7. In the graphs, the suggested WMO traceability limits for absolute AOD differences (that have been defined for AOD at wavelengths without gaseous absorption in the UVA–NIR wavelengths range) (WMO/GAW, 2005) are also shown. Obviously, the agreement is very good between the Brewer and the UVPFR for these measurements taken 4–11 months after the calibration. During the calibration of Brewer no. 163, more than 98 % of the AOD residuals, AOD(Brewer)–AOD(UVPFR), were within the WMO limits at all wavelengths. During the comparison in Davos, at four out of the five Brewer wavelengths, more than 95 % of the differences fall within the WMO limits. Only at the shortest wavelength, with 85.6 % of the differences within the limits, was the traceability requirement of 95 % not fulfilled. This could indicate a small change in any of the instruments at the shortest wavelength(s). The root mean squared difference is still low at all wavelengths, amounting to [0.008, 0.006, 0.006, 0.005, 0.005] for the 306–320 nm wavelengths.
During the low AOD period from the end of October until November, AOD from
the Brewer also showed the unexpected behavior of giving decreasing AOD values
with decreasing wavelength. Therefore, the AOD differences between the UVPFR
and the Brewer also remained small during this period. There are several
possible explanations for the low AOD values at the shortest wavelengths.
The most plausible reason is that the used ozone absorption
coefficients and/or ozone amount were too high. Also, the use of too low
calibration values could be a possible contributor. Based on the relatively
stable differences over the day between AOD at, e.g., 305 and 368 nm, in
addition to the fact that
This paper reports on the UVPFR sunphotometer, an instrument that can be
used for AOD measurements at four UV wavelengths. The standard PFRs were
designed with emphasis on precision and stability, while also being robust
instruments. These goals have been reached by the PFRs (Wehrli, 2000;
Gröbner et al., 2015). The UVPFR is of similar design and, based on the
results of this first study, including suggested corrections, the UVPFR
appears to be a stable high-quality radiometer for AOD determination in the
UV. According to Langley plot calibrations at a high-altitude station the
sensitivity of the UVPFR changed by
It was shown that due to the relative wide FWHM of the UVPFR the calibration
constants (
Even with the suggested corrections applied, the expanded uncertainty of AOD derived from UVPFR measurements, as well as from other UVB instruments, remains relatively high at the shortest wavelengths. The major source of uncertainty is the ozone optical depth uncertainty, resulting from uncertainties in ozone cross section, ozone temperature and TCO amount. The second largest source of uncertainty at the three shortest wavelengths, and the largest source of uncertainty at 332 nm, is the calibration uncertainty, especially at high sun/low air mass conditions.
Despite the relatively high AOD uncertainties at the short wavelengths, it is still considered worthwhile to continue working with the AOD at, e.g., 305–306 nm to learn more on AOD retrievals in the UVB. Most probably, better input information connected to ozone will be available in the future which will reduce the AOD uncertainty. Also, if the same ozone cross section data and effective ozone temperature data are used by different instruments/groups/sites, as will be the case within EUBREWNET for example, the AOD results will be consistent and much more comparable.
An example of very good agreement of UV AOD retrievals was shown by a comparison between the UVPFR no. 1001 and Brewer no. 163 for several months of measurements in Davos. Since Brewer no. 163 and UVPFR no. 1001 calibrations were partly linked at an earlier date, the comparison was not performed by fully independent instruments and therefore we should expect a relatively good agreement. The comparison indeed confirms good agreement for the measurements taken 4–11 months after the Brewer calibration. The root mean squared AOD differences were < 0.01 at all the 306–320 nm Brewer wavelengths. This can be considered a very good result for an AOD comparison at UVB wavelengths. An additional very likely reason for the good agreement is the fact that both instrument types measure at close wavelengths in the UVB. In earlier studies in which AOD was determined from Brewer DS measurements the validation has so far only been done against measurements at UVA or even VIS wavelengths (Marenco et al., 2002; Cheymol and De Backer, 2003; Cheymol et al., 2006; Gröbner and Meleti, 2004; Kazadzis et al., 2005, 2007; De Bock et al., 2010; Kumharn et al., 2012). Also, earlier comparisons of AOD from Brewers of different type have shown larger differences than between the UVPFR and the MkIII Brewer in this study (Kazadzis et al., 2005; Kumharn et al., 2012).
In addition to a low-turbidity case showing AOD values from the UVPFR
consistent with a standard PFR, average UV AOD values of the UVPFR during
the measurements in Davos were compared with highly accurate AOD values,
2
Despite the fact that the total uncertainty of AOD in the UVB is relatively
high, based on the comparison between the UVPFR and a Brewer it is estimated
that calibrated and well maintained UVPFR sunphotometers and Brewer
spectrophotometers can measure AOD at a precision of 0.01 (1
The total column ozone data used in this study can be
downloaded from the EUBREWNET website:
The authors declare that they have no conflict of interest.
Thomas Carlund was supported through grant no. C14.0025 from the Swiss
Staatssektretariat für Bildung, Forschung und Innovation (SBFI) within
COST ES1207. Part of the work was supported by a STSM grant from COST Action
ES1207 (EUBREWNET – A European Brewer Network). The slit function
measurements were done on the tuneable laser facility ATLAS, funded through
contract number IDEAS