Measurements of UV Aerosol Optical Depth in the French Southern Alps

Measurements of UV Aerosol Optical Depth in the French Southern Alps J. Lenoble, C. Brogniez, A. de La Casinière, T. Cabot, V. Buchard, and F. Guirado Laboratoire d’Optique Atmosphérique (LOA), Université des Sciences et Technologies de Lille, 59655 Villeneuve d’Ascq Cedex, France Interactions Rayonnement Solaire Atmosphère (IRSA), Université Joseph Fourier de Grenoble, 17 Quai Claude Bernard, 38000 Grenoble, France Centre Européen Médical et Bioclimatologique de Recherche et d’Enseignement Supérieur (CEMBREU), 35, rue Croix de Bretagne, Villars St Pancrace, 05100 Briançon, France Received: 29 November 2007 – Accepted: 4 December 2007 – Published: 8 January 2008 Correspondence to: J. Lenoble (jacqueline.lenoble@wanadoo.fr)


Introduction
10 Spectral ultraviolet (UV) global and diffuse irradiances on an horizontal plane have been routinely measured during several years at the Briançon station in the French Southern Alps. The direct irradiance is therefore obtained by difference, and is used for retrieving AODs. The objectives are: i) to build a climatology of aerosols in a clean unpolluted area, ii) to detect possible cases of turbidity and to analyse their causes. 15 We present the measurement site and the instruments in Sect. 2, the data and their analysis in Sect. 3. Section 4 shows the results, mainly the AOD annual variation. Summary and conclusion are in Sect. 5.

Station and instruments
The measurement site is located in Villars Saint Pancrace, a small village, close to the 20 town of Briançon (44.90 • N, 6.65 • E), in a sunny and dry Alpine valley, at the altitude of 1310 m a.s.l.; the site belongs to the CEMBREU. The instruments are operated under the supervision of LOA and IRSA groups. Figure 1 shows a photograph of the instruments on the CEMBREU platform. 162 ery 30 min (Pachart,1997;Masserot, 2001); it was equipped with a shadow disc in July 2003, and thereafter measured a spectrum every 10 min, alternately global and diffuse irradiances. The characteristics of the two instruments are very similar (Lenoble et al., 2005). The largest wavelength recorded is 400 nm for the BE and 450 nm for the JY.
Both instruments are calibrated about every 3 months with a standard lamp traceable 10 to NIST. They have taken part into instrument intercomparisons (Gröbner et al., 2000;de La Casinière et al., 2005); their uncertainty is assumed to be around 5%. The data of the two instruments are processed completely independently; several comparisons of the measured irradiances confirm an agreement between the two instruments within ±5%. 15 The direct solar irradiance DIR(t) at time t, is retrieved from the measured global GLO and diffuse DIF irradiances, as where ∆t=10 min, for the JY and ∆t=5 min for the BE. The retrieval is limited to clear stable conditions, when the solar zenith angle (SZA) does not vary too rapidly, prefer-20 ably around local noon.

Method
The direct irradiance can be written, as

EGU
where F 0 is the extraterrestrial solar flux, µ 0 the cosine of the SZA, τ Ray the molecular Rayleigh optical depth, τ Oz the ozone absorption optical depth, τ aer the AOD. The AOD is the only unknown in (Eq. 2). The direct irradiance that would be observed for a perfectly clear atmosphere, without aerosols, DIR cl , is computed as at step 0.05 nm, using the ATLAS3 high resolution spectrum (Van Hoosier, 1996), and it is convoluted using the instrument slit function. The Rayleigh optical depth is computed for a standard mid latitude atmosphere winter or summer (McClatchey et al., 1972), at the altitude of Briançon. The ozone optical depth uses the Paur and Bass (1985) ozone cross sections and the ozone total column amount given by the Total Ozone Mapping A similar approach consists in first deriving the total optical depth τ tot = τ Ray + τ Oz + τ aer (5) 15 from (Eq. 2), as where F 0 is preliminary convoluted to the instrument slit function; then τ Ray and τ Oz are substracted from τ tot . This second method has been compared satisfactorily (Fig. 2) to the first one, used throughout the paper.   , 2004), and its results are comforted by the present work.

Uncertainties
For this analysis, we have found more convenient to use the formalism of the second method, although both methods are rigourously equivalent. From (Eq. 6), the uncertainty on τ tot is due to the independent uncertainties on the 10 measured irradiance DIR and on the extraterrestrial flux F 0 .
On τ aer retrieved from Eq. (5), further uncertainties are added by the uncertainties on τ Ray , and τ Oz . We will try briefly to analyze these various causes of error. The uncertainties on the measurements of either GLO or DIF have been discussed 15 by several authors (Bernhard and Seckmeyer, 1999;Houët, 2003;Webb et al., 1998). Roughly half of the uncertainty is due to the calibration itself, both to the lamp calibration uncertainty, and to the uncertainties during the calibration procedure. Another half of the error comes from the measure, as instrument noise, cosine error of the diffusor, wavelength shift non perfectly corrected. Houët (2003) estimates the measurement 20 uncertainty to about 4-5% for a confidence interval of 95%. The error becomes larger when the signal is small, i.e. at short wavelengths, and when the sun is low on the horizon. On the direct irradiance DIR, a further error could come from the averaging process in (Eq. 1); however, it is negligible for wavelengths larger than 310 nm, and SZA smaller than 60 • (Houët, 2003). We have limited our analysis to these conditions.

25
The different extraterrestrial spectra measured by various instruments (Cebula et al., 1999), differ by about 2-3%, which suggest a possible systematic error of about 2% on 165 Introduction varying between 0.050 for a high sun to 0.025 for a sun at 60 • from the zenith; actualy, the influence of µ 0 in (Eq. 8), can be partly compensated, because the measurement 5 uncertainty increases for low sun, as mentionned above.
On τ Ray , there are two causes of uncertainty, one on the Rayleigh cross-sections, due to the uncertainties on the depolarization factor, and on the air refractive index (Bodhaine et al., 1999), and another one due to the uncertainty on the surface pressure; these independent errors can be estimated respectively to about 2% and 1.5%, 10 leading to For 320 nm, ∆τ Ray =0.020; it decreases toward large wavelengths to 0.008 at 400 nm. Similarly on τ Oz , the uncertainty is due to the uncertainty on the ozone cross-sections (about 2%), and on the ozone total column amount evaluated to about 4% for TOMS 15 data; this leads to This error increases sharply with τ Oz towards short wavelengths, but it becomes completely negligible above 330 nm. This is a further reason to limit our analysis to a wavelength range with a lower limit 20 between 310 and 330 nm. Roughly, we estimate the uncertainty on AOD to be around 0.05, knowing that it slightly varies with wavelength and with SZA. For the small AOD observed in Briançon (0.1-0.3), this means a large relative uncertainty, reaching 50% for the smallest values.
166 remain (Fig. 2); definitively, they are artefacts, and not real spectral variations. We did not find a satisfactory explanation for these oscillations, which are of the order of the expected error. The AOD slowly increases toward short wavelengths between 440 nm and 320 nm; as said before, the uncertainty becomes too large below 320 nm to allow analysis. When results are obtained from the two instruments on the same day, they 10 always agree wihin ±3%, that is better than the estimated uncertainty of ±5%.
We have therefore decided to focus on analyzing time variations, more precisely annual variations at 12:00 TU, which is not far (within about half an hour) from local noon, at two specific wavelengths. The wavelengths chosen are 340 nm and 380 nm, which correspond to the UV channels of the CIMEL sunphotometer of the AErosol 15 RObotic NETwork (AERONET) (Holben et al., 1998). Figure 3 presents the AOD annual variations at wavelength 340 nm obtained from both instruments for the three years 2003, 2004, and 2005. A general behaviour appears for the three years, with low values, smaller than 0.1, 20 in winter, and higher values, generally around 0.2, in summer. Some much higher values, around 0.3-0.4, appear between mid-July to mid-September; they are much more numerous, and still appear late October, in 2003, which was characterized by a very hot summer. For all the three years, the increase after the low winter values begins in mid-February, and the decrease in mid-November. Even considering the large relative uncertainty mentionned above, the annual variation is significant. EGU rms clearly show the general annual pattern. The rather large std is due partly to the uncertainty and partly to real day to day variations. As expected a similar behaviour is observed for 380 nm (Fig. 4), with values slightly smaller than at 340 nm.

Annual variations
We had previously analyzed the Bentham data for 2001and 2002(Lenoble et al., 2004, and presented an average AOD in the UV-A, whereas we present in this paper AODs at 340 and 380 nm. We have already stated that AOD varies only slowly in the wavelength range 320-400 nm. Therefore the previous results can be compared qualitatively to the present ones. In Fig. 5, we have reported the results from Fig. 2 in Lenoble et al. (2004); it confirms a similar pattern, with low winter values, and higher summer values in 2001 and 2002, as in 2003, 2004, and 2005. This behaviour seems typical of Briançon site over five years.

Complementary analysis
The satellite instrument TOMS (McPeters et al., 1998) provides regularly an aerosol index (AI), which is related to the aerosol optical depth and absorption in a rather 15 complex way. The retrieval of AI from TOMS observations, and its analysis in term of aerosol characteristics, have been described in several publications (Herman et al., 1997;Torres et al., 2002). We consider it here only as a qualitative parameter, and we present in Fig. 6, TOMS AI for the Briançon site on the clear days of 2003, 2004, and 2005. Surprisingly or not, the AI annual variation shows some similarity with the AOD 20 behaviour; positive AIs in summer could point to absorbing aerosols. Trying to push forward the comparison would be risky. As we have seen (Fig. 2), the spectral variation of τ aer is weak, with oscillations, and we can look only for an approximate information on the general slope. We have chosen, as reference, an Angström parameter α defined between 340 and 380 nm. 25 Figure 7 shows α plotted versus τ aer (380) for all the BE data. For small AODs, the average value of α is 1.7, with a very large dispersion, mostly due to the uncertainties on AODs; when τ aer becomes larger than 0.2, α tends to stabilize around 1.3, with a 168 with an average of 0.2, at 340 nm. This behaviour is observed at all wavelengths between 320-400 nm, the spectral variation being weak over this limited interval. It is observed for the five year period from 2001 to 2005, with the largest summer values in 2003, and it is definitively significant despite the large relative uncertainty. The increase of turbidity in spring can be due to the vegetation, pollen and other 10 vegetal particles in summer. Another contribution can be dust particles, originating from the dry soil in summer, whereas the ground is generally covered with snow in winter. The few cases with the highest turbidity, could have other more specific explanations, as an arrival of desert dust particles, which sometimes reach the Alps, after crossing 15 the Mediterranean Sea. Another source could be biomass burning particles, from the frequent forest fires in the South of France. An analysis of these events and of the air mass trajectories would be necessary to check the above hypothesis.  Optics, 43, 3133-3139, 2004. Lenoble, J., Brogniez, C., Houët, M., Legrand, M., de La Casinière, A., Cabot, T., and Guirado,