For more information about the SunPhotoSpectrometer and data, please contact  Dr. Tom McElroy

The SunPhotoSpectrometer (SPS) has participated in campaigns at Eureka since 2004.  The last campaign was in 2020 before the pandemic arrived.  Figure 1 shows the instrument, and the tracker which points it at the sun and the sky, being mounted on the roof at the Ridge Lab in February, 2008.

Figure 1: The SPS Tracker being installed on the roof of the Ridge Lab. From left to right: Clive Midwinter, Mareile Wolff and Tom McElroy.  Photo: P. Fogal.

The SPS is a very simple spectrophotometer designed for rugged use at a wide temperature range.  As will be discussed later, the instrument was designed and constructed at Environment Canada and was flown on the Space Shuttle Columbia and the NASA ER-2 high-altitude research aircraft to investigate ozone depletion chemistry.  When the instrument was installed at PEARL, the primary focus was twilight sky measurements of ozone (O3) and nitrogen dioxide (NO2).  Later, other measurement types were incorporated.  The twilight sky measurements are the most interesting and will be illustrated here.  The instrument measures in both the first and second order of a diffraction grating.  The first order covers the wavelength range from approximately 370 nm to 770 nm, while the second order measures from 185 to 385 nm.  The range is determined by the factor of two between the two orders.  The two orders are separated by the use of two different band-limiting filters mounted in a motor-driven filter wheel within the instrument that can be automatically selected.

Figure 2: Twilight visible light spectra.  Two of four integration times used to match the dynamic range of the light signals to the detector and electronics are shown.

Figure 2 shows a set of visible, zenith sky spectra taken throughout the morning of March 27, 2020.  The red spectrum could be used as a reference to analyze the others.  The blue and purple spectra were recorded at different integration times in order to maintain a high-sensitivity response and yet not overfill the detector elements.  In addition to the spectra shown, there are two other integration times that are used, up to 10 seconds maximum for the dimmest parts of the day near sunrise and sunset.  One reference spectrum is selected from a day when the sky is brightest, usually the last day or two of the campaign, and used to analyze all days.  In fact the results here were based on the use of a reference from an earlier year, which allows near real time processing in the field.

            Solar line features and the oxygen absorption lines are used as references to fine-tune the wavelength assignment of the reference to agree with a very high resolution reference spectrum used in the data analysis model.  It is important to have accurate wavelength assignment so that the absorption cross-sections used to analyze the data are appropriately aligned with the observed optical depths calculated from the spectral observations.  The large oxygen A band can be seen at approximately 762 nm.  The oxygen B band is around 685 nm.  The obvious feature at 430 nm is the Fraunhofer G line, a solar feature.

Figure 3: The spectrum of NO2 smoothed, differential and total in the region with maximum differential information.

The amount of absorber in the solar path through the atmosphere is determined by spectral fitting.  A model of the absorption process is created using a theoretical, high-resolution extraterrestrial spectrum and the model passed light through a model atmosphere that calculates the intensity of the light.  The spectrum is then smoothed to the instrument resolution.  The effective optical depth is then compared to the observed optical depth calculated for each observed spectrum by dividing by the observed reference spectrum.  A number of fitting vectors are used to achieve an agreement between the model and the observed optical depth spectra.  These include the absorbers to be measured or accounted for, wavelength shift and stretch between the reference and the individual observed spectra and several vectors to account for differences in absolute intensity between the reference and the observed spectra.  These differences could be due to atmospheric scattering due to aerosol and air, or the presence of cloud.  With the slow, wavelength dependent changes accounted for in this way, the retrieved amounts of gases in the atmospheric path then depend mostly on the structure of the absorption features.  In the SPS analysis, ozone is retrieved in just this way since it is a robust feature, however, the retrieval of NO2 which has a very much smaller optical depth, is done by fitting the differential component of the spectrum so the any changes in the spectra as a function of wavelength over a longer wavelength range will not be accounted for by spurious amounts of NO2 absorption.
The differential component of the NO2 spectrum is calculated by smoothing the spectrum strongly and then subtracting the smoothed spectrum from the full cross-section.  When the spectral fitting is done the amount of NO2 is determined using the differential cross-section and the absorption due to the smoothed component is added in by multiplying the smoothed cross-section by the amount of NO2 determined by the differential fit.  The smoothed, total and differential cross sections are depicted in Figure 3.

Figure 4: Apparent column ozone amount from twilight sky measurements in the morning and afternoon.

Figure 4 shows a plot of the apparent ozone column as a function of the solar zenith angle (SZA) on March 27, 2020.  The morning and afternoon curves are plotted separately.  It can be seen that the amount of absorption from ozone increases up until about 92o in both cases but diverges after that.  This is most likely due to either the difference in the ozone gradient toward the east and the west or temporal changes.  Figure 5 shows a similar plot for NO2.  Again the curves are very similar for the morning and evening.  This is somewhat unusual because by the time the end of March is reached a distinct difference between morning and evening NO2 columns is usually seen.  This is due to the photolysis of N2O5 during the daylight hours and the loss of NO2 overnight.  The presence of reactive nitrogen depends on the release of nitrogen from nitric acid by short wavelength photolysis in the spring.

Most twilight sky measurements are now analyzed by the use of an algorithm supported by the Network for the Detection of Atmospheric Composition Change (NDACC).  The algorithm is quite complicated and includes modelling as well as the input of other sources of data.  To some extent this makes it difficult to understand what the actual twilight measurements are contributing and how sensitive the reduced data are to the input of the twilight data.

Figure 5: As in Figure 3 but for NO2.

In the 1927 paper on differential optical absorption measurements, Brewer et al. suggested a simple interpretation of the slope of the absorption curves at 90o SZA, as in Figures 4 and 5, as directly related to the absorption of the gas over a path of 111 km at an altitude between 20 and 30 km, depending on wavelength and the pressure profile.  Figures 6 and 7 indicate the change in this quantity for ozone and NO2 over a number of days in March, 2020.  More study will be needed to determine how useful this can be, but at least the retrieved quantities are of the correct order and the morning and evening values appear to be coherent.


[1 ]  McElroy, C.T., J.B. Kerr, D.I. Wardle, L.J.B. McArthur, G.M. Shah, M. Garneau, S.G. MacLean, R. Thirsk, J.A. Davies, W.F.J. Evans, R.W. Nicholls, J.C. McConnell, and M. Cann, SPEAM-I Observations of High-Altitude Ozone from STS 4l-G, Can. J. Phys., 69, 1123-1127, 1991.

[ 2 ]  McElroy, C.T. , A spectroradiometer for the measurement of direct and scattered solar spectral irradiance from on-board the NASA ER-2 high-altitude research aircraft, Geophys. Res. Lett., 22, 1361-1364, 1995.

[ 3 ]  McElroy, C.T. , C. Midwinter, D.V. Barton, and R.B. Hall,  A Comparison of J-values estimated by the composition and photodissociative flux measurement with model calculations, Geophys. Res.Lett., 22, 1365-1368, 1995.