Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Total ozone column (TOC) measurements are traditionally obtained from ground-based instruments since the early 1920s by Dobson and the early 1980s by Brewer spectrophotometers. Recently diode-array or CCD equipped spectrographs are being tested for their ability to provide measurements of total ozone of accuracy comparable to that of the conventional instruments, such as the Pandora system (Herman et al. 2015; Tzortziou et al. 2012) and the Phaethon system (Kouremeti et al. 2013). TOC can be derived also from ultraviolet spectral measurements of the direct solar irradiance using the well-known Lambert-Beer law, which requires calibrated absolute irradiance measurements and knowledge of the extraterrestrial solar spectrum. The differential optical absorption spectroscopy (DOAS) technique (Platt and Stutz 2008) has been used extensively during the last 2–3 decades to derive total, stratospheric and tropospheric columns of various trace gases, such as NO2, O3, SO2, BrO, HCHO, etc. In this study, a variant of the DOAS method to derive TOC from direct radiance spectral measurements acquired by the Phaethon system, without being necessary to determine and use an extraterrestrial solar spectrum, is presented and evaluated with data of a collocated Brewer spectrophotometer.

2 Instrumentation and Data

Phaethon is a mini DOAS/max-DOAS system which consists of a cooled, miniature CCD spectrograph (AvaSpec-ULS2048LTEC) and a 2-axes tracker with precision of 0.125° (Kouremeti et al. 2013). It performs spectrally resolved measurements of direct solar irradiance and sky radiance at several elevation angles, including the zenith, in the UV-visible region (300–450 nm), with a resolution of less than 0.25 nm. The DOAS method is applied on these spectral measurements to derive total and tropospheric columns of trace gases (O3, NO2, HCHO, SO2, etc.). Phaethon operates regularly on the roof of the Physics Department building of the Aristotle University of Thessaloniki, Greece (40.634°N, 22.956°E) in the center of the city of Thessaloniki. At the same location a suite of others instruments are also in operation, including a Brewer spectrophotometer.

In this study, the TOC data were derived from direct irradiance spectra acquired by Phaethon at Thessaloniki in the periods January 2014–October 2014 and November 2015–February 2016 under cloud-free conditions. Measurements were performed during daylight hours every 20 min. Each set of data comprises 20 scans acquired within about 3 min, depending on the level of signal. The integration time is adjusted automatically to achieve the highest possible signal without saturating the detector.

3 Methodology of TOC Retrieval

The recorded spectra are corrected for the dark signal and stray light and are compared to a reference spectrum that has been recorded with the same instrument, to remove the absorption Fraunhofer signatures of solar irradiance and unmask the absorption signatures that are caused by atmospheric constituents. The analysis is performed with the QDOAS software (Danckaert et al. 2013) to derive the differential slant column density (DSCD) of ozone:

$$ DSCD_{i} = SCD_{i} - SCD_{REF} = TOC_{i} \cdot AMF_{i} - SCD_{REF} $$
(1)

where SCD i is the slant column density of ozone for the ith radiance spectrum, SCD REF is the slant column density of ozone for reference spectrum, and AMF i the airmass factor, which for solar zenith angles (SZA) of less than about 75° it is approximated with the secant of the SZA at the altitude of ~22 km, for mid-latitude locations.

The SCD REF , can be derived from (1) by Langley extrapolation of the DSCDs versus AMF, during periods with fairly constant TOC so that (1) is linear. Alternatively, the SCD REF can be derived from a collocated instrument measuring the TOCREF, e.g. by a Brewer spectrophotometer. Note that for the same instrument the reference spectrum can be measured once and used repeatedly, irrespective of location. Once the SCD REF is determined, (1) can be rewritten to derive the TOC from each measured direct radiance spectrum:

$$ TOC_{i} = \frac{{DSCD_{i} + SCD_{REF} }}{{AMF_{i} }} $$
(2)

The overall error in the TOC estimation can be derived by propagating the errors of the individual terms of (2), yielding:

$$ \sigma_{{TOC_{i} }} = \frac{{\sqrt {\sigma_{{DSCD_{i} }}^{2} + \sigma_{{SCD_{REF}}}^{2}}}}{{AMF_{i}}} $$
(3)

In (3) the error in the calculation of the AMF is assumed negligible, being mainly a geometrical factor. The error of the DSCD is provided by the QDOAS analysis software, while the error of the SCD REF depends on the method that is used. For the Langley extrapolation, the error can be represented by the uncertainty of the regression. When the TOC measured by the collocated Brewer is used, the uncertainty associated to this measurement can be considered. Usually it is less than 2.5 DU.

The DOAS analysis for the retrieval of TOC is applied in the spectral window 315–337 nm, and the broadband attenuation features of the spectral radiance data are removed by fitting a 4th degree polynomial. The cross-section datasets and temperatures used for different species are: (Paur and Bass 1984) at 228 K for ozone, (Vandaele et al. 1996) at 294 K for NO2, (Vandaele et al. 1994) at 294 K for SO2, and (Meller and Moortgat 2000) at 293 K for HCHO.

4 Results

The TOC derived by Phaethon is compared with that of the Brewer spectrophotometer #005, which has recently confirmed its calibration in an intercomparison campaign against the standard Brewer of the European Reginal Calibration Center of WMO (RBCC-E) in El Arenosillio, Spain, in May 2015. As the measurement protocols of the two systems are not synchronized, in the comparison data that were recorded within 30 min have been used, with 90 % of them taken within 10 min.

The distribution of errors in the TOC retrieved from Phaethon (Fig. 1, right) reveals that the majority of the data have errors less than 9 DU, i.e. roughly 3 % of the TOC, while only 30 % of the data have errors smaller than 3 DU. These errors which increase with solar zenith angle (SZA) are still high and further at-tempts were made to reduce the error to less than 1 % (about 3 DU). In addition, while generally the frequency of errors decreases with their size, a secondary peak is apparent at around 10 DU. This has been attributed to a period of 2–3 months when the system was not positioned perfectly towards the Sun and the data were contaminated by scattered radiation from the sky in the vicinity of the solar disk.

Fig. 1
figure 1

Left Frequency distribution of errors in TOC retrieval as reported by the DOAS analysis for Phaethon. The red line is a smoothing function fit and the vertical lines mark the percentage of data with error smaller than 3, 6, 9 and 15 DU. Right Scatter plot of TOC derived by Phaethon and Brewer #005

Full size image

Despite the large errors the overall agreement of Phaethon with the Brewer has not been affected significantly, leading to average difference of 0.96 ± 3.2 %, and an overall small offset of about 7 DU (Fig. 1, right). However, it is evident that part of the data has large discrepancies as expected from the large standard deviation (3.1 %) of the differences. During this 2-year period several attempts have been made to improve the accuracy of the derived TOC both in the DOAS retrieval methodology and the hardware.

In December 2015 a short-pass filter has been installed in Phaethon fore optics to suppress the signal at wavelengths longer than 370 nm, in order to increase the signal to noise ratio in the DOAS retrieval range of ozone (315–337 nm). Eventually this filter suppresses also part of the stray-light which contaminates the spectral measurements in single monochromator spectrometers, like Phaethon. This modification has led to a substantial reduction of the DOAS analysis error and improved the comparison with the Brewer. From the frequency distribution of the retrieval error reported by QDOAS (Fig. 2, left) it appears that about 80 % of the measurements have errors below 3 DU, compared to 30 % for the data of the period before December 2015. The scatter-plot of Fig. 2 (right) suggests a very good agreement between the two systems. The correlation coefficient of the regression between Phaethon-and Brewer-derived TOC is high (0.98), the slope is very close to unit (0.98), while the mean absolute difference is 0.76 ± 1.5 %.

Fig. 2
figure 2

Left Frequency distribution of errors in TOC retrieval as reported by the DOAS analysis of Phaethon spectra recorded through a short-pass filter. The red line is a smoothing function fit and the vertical lines mark the percentage of data with error smaller than 3, 6, 9 and 15 DU. Right Scatter plot of TOC derived by Phaethon spectra recorded through the short-pass filter and Brewer #005

Full size image

5 Conclusions

The Phaethon system which is used to derive total and tropospheric columns of various atmospheric gases has been tested for the retrieval of the total ozone column using the DOAS technique. Comparisons of data obtained by Phaethon with those of a collocated Brewer spectrophotometer showed overall a satisfactory agreement to within 2 %, but with retrieval errors of the order of 2–3 %. By modifying the system to record spectral measurements with suppressed signal at wavelengths longer than 370 nm the signal to noise ratio was improved by at least 4 times, reducing the retrieval error to less than 1 %. The agreement with the Brewer data has been also improved so that the average difference during the 3-month operation of Phaethon with the new configuration has been increased with relative differences of 0.76 ± 1.5 % and with retrieval errors of less than 1 %. Since all these data were recorded in the winter period when the SZA is large (the smallest SZA was 58°), it is expected that the agreement will be further improved as new data are collected at smaller SZAs.