Columnar aerosol properties from sun-and-star photometry: statistical comparisons and day-to-night dynamic

This work presents the first analysis of longterm correlative day-to-night columnar aerosol optical properties. The aim is to better understand columnar aerosol dynamic from ground-based observations, which are poorly studied until now. To this end we have used a combination of sun-and-star photometry measurements acquired in the city of Granada (37.16 ◦ N, 3.60 W, 680 m a.s.l.; South-East of Spain) from 2007 to 2010. For the whole study period, mean aerosol optical depth (AOD) around 440 nm (±standard deviation) is 0.18 ± 0.10 and 0.19 ± 0.11 for daytime and nighttime, respectively, while the mean Angstr öm exponent ( α) is 1.0 ± 0.4 and 0.9 ± 0.4 for daytime and nighttime. The ANOVA statistical tests reveal that there are no significant differences between AOD and α obtained at daytime and those at nighttime. Additionally, the mean daytime values of AOD andα obtained during this study period are coherent with the values obtained in the surrounding AERONET stations. On the other hand, AOD around 440 nm present evident seasonal patterns characterised by large values in summer (mean value of 0.20 ± 0.10 both at daytime and nighttime) and low values in winter (mean value of 0.15 ± 0.09 at daytime and 0.17 ± 0.10 at nighttime). The Angstr öm exponents also present seasonal patterns, but with low values in summer (mean values of 0.8 ± 0.4 and 0.9 ± 0.4 at dayand night-time) and relatively large values in winter (mean values of 1.2 ± 0.4 and 1.0 ± 0.3 at daytime and nighttime). These seasonal patterns are explained by the differences in the meteorological conditions and by the differences in the strength of the aerosol sources. To take more insight about the changes in aerosol particles between day and night, the spectral differences of the Angstr öm exponent as function of the Angstr̈ om exponent are also studied. These analyses reveal increases of the fine mode radius and of the fine mode contribution to AOD during nighttime, being more remarkable in the summer seasons. These variations are explained by the changes of the local aerosol sources and by the meteorological conditions between daytime and nighttime, as well as aerosol aging processes. Case studies during summer and winter for different aerosol loads and types are also presented to clearly illustrate these findings.


Introduction
Atmospheric aerosol is noted by the Fourth Intergovernmental Panel for Climate Change (IPCC 2007) as a key component on the climate (Forster et al., 2007). Atmospheric aerosol particles directly affect Earth's radiation budget by scattering short-wavelengths radiation and absorbing shortwave and longwave radiation (e.g., Haywood and Shine, 1997;Forster et al., 2007). Furthermore, atmospheric aerosol particles can act as cloud condensation nuclei and, thus, they can modify cloud droplet size and cloud albedo (Forster et al., 2007). In addition, they have effects on air quality and, thus, on the human health (e.g., Pope et al., 2002;Brunekreef and Forsberg, 2005;Miller et al., 2007). The IPCC 2007 also reported that due to the lack of adequate information on aerosol temporal and spatial variability the uncertainties associated with each aerosol effect are much larger than those associated with greenhouse gases. Therefore, it is really 9720 D. Pérez-Ramírez et al.: Columnar aerosol properties from sun-and-star photometry important to measure and characterise the aerosol properties in different sites for better understanding the aerosol impact at least at regional scale.
Research on the atmospheric effects of the aerosol particles has become a top priority. In this sense, several satellite programmes have been developed to study long-term spectral aerosol optical depth (AOD) on a global scale (e.g., Kaufman et al., 1997Kaufman et al., , 2002Kahn et al., 2005;Remer et al., 2005). However, satellite measurements present low temporal resolution. Surface-based passive measurements allow the study of columnar aerosol properties, and the global network AERONET (Holben et al., 1998) has been developed. But all these instrumentations acquire measurements at daytime. Currently, Raman lidar systems allow retrieving vertical profiles of aerosol optical properties at nighttime, but the high cost of this technique make these measurements sparse in time (Bösenberg et al., 2001;Pappalardo et al., 2009). Other simpler configurations, such as elastic lidar systems, can also acquire nighttime measurements, but they need constraints as the extinction-to-backscatter ratio (Sassano et al., 1989;Welton et al., 2001).
Until now, the knowledge of columnar aerosol properties at nighttime is quite limited due to the absence of continuous measurements. Although the role of columnar aerosol at nighttime in radiative forcing is weak because only small aerosol absorption is expected in the thermal infrared, the knowledge of nighttime columnar aerosol properties is very important to evaluate aerosol dynamic. These studies will allow us to have a whole picture of the daily behaviour of the atmospheric aerosol, covering the different stages in the evolution of the planetary boundary layer and pre-convection and pre-photochemistry processes that affect the atmospheric aerosol. The knowledge of AOD at nighttime will also contribute to aerosol transport and chemistry models validation efforts, being important within the air-quality schemes. In addition, the AOD at nighttime can be also used as constraints for correlative lidar measurements (e.g., , both for ground-based and space-borne missions. Therefore, research studies about columnar aerosol properties at nighttime are calling for and currently some research groups are working with ground-based irradiance measurements from stars (e.g., Herber et al., 2002;Perez-Ramirez et al., 2008a;Baibakov et al., 2009) or from the moon (e.g., Esposito et al., 1998;Herber et al., 2002;Berkoff et al., 2011) to obtain AOD at nighttime.
First attempts in star-photometry were made by Leiterer et al. (1995) who developed an instrument based on a photodetector as measurement device, acquiring very valuable measurements during 10 days in April 1994 at the Lindenberg observatory (52.14 • N; 14.12 • E; Germany). This design was also used by Herber et al. (2002) to acquire measurements during the Arctic winter from 1996 to 1999 at the Koldeway station in Ny-Ålesund (78.95 • N; 11.93 • E; Norway). However, these last studies did not deal with correlative day-to-night measurements due to the characteristics of the place. Moreover, evaluations of daytime and nighttime differences in particle sizes (types) were not performed.
To address the problem of the lack of continuous measurements of spectral AOD at nighttime, this work uses the measurements acquired by the star photometer EXCALIBUR based on a CCD camera as detector device (Perez-Ramirez et al., 2008a, b). This instrument, together with a sun photometer CIMEL, operates in the Andalusian Center for Environmental Research in the city of Granada (37.16 • N,3.60 • W, 680 m a.s.l.; South-East of Spain). The star photometer EX-CALIBUR is also a versatile instrument because it is able to obtain precipitable water vapour (Perez-Ramirez et al., 2012a) and is also used to estimate the sky quality in Astronomical Centers (Sanchez et al., 2007).
Moreover, atmospheric aerosol particles are constantly affected by physical and chemical processes in the atmosphere that induce changes in the optical and radiative properties of these particles. The spectral dependence of AOD is related to the sizes (types) of the predominant particles. Thus, using sun/star photometry, the possible changes in aerosol particle sizes (types) can be analysed by means of studying the spectral dependence of AOD (e.g., O'Neill et al., 2001O'Neill et al., , 2003Schuster, 2006;Gobbi et al., 2007) or by studying aerosol size distributions retrieved by inversion methods using sky radiance measurements (e.g., Dubovik and King, 2000;Dubovik et al., 2006;Olmo et al., 2006Olmo et al., , 2008. In this work, in order to study the possible change in the aerosol particle sizes between daytime and nighttime, and due to the lack of sky radiance measurements by star photometry, we use the simple graphical method proposed by Gobbi et al. (2007).
The scope of this work is to study day-to-night columnar aerosol properties which, to our knowledge, have not been already done. To this end, we use consecutive day-to-night sun/star photometry measurements acquired between 2007 and 2010 in the city of Granada (Spain). The instruments used and the experimental site are described in Sect. 2. The descriptions of the methodologies are in Sect. 3. Later, in Sect. 4 we present the main results, with an in depth analysis of intra-annual and seasonal evolutions of columnar aerosol properties, as well as the spectral analysis of the Angström exponent, both at daytime and nighttime. We also present two case studies related to the air-masses origin in order to show day-to-night columnar aerosol dynamic. Concluding remarks are given in Sect. 5. D. Pérez-Ramírez et al.: Columnar aerosol properties from sun-and-star photometry 9721 at 340,380,440,670,870,940 and 1020 nm. The fullwidth at half-maximum (FWHM) of the interference filters are 2 nm at 340 nm, 4 nm at 380 nm and 10 nm for the other wavelengths. More details about the CIMEL CE-318-4 can be found in e.g., Holben et al. (1998); Alados-Arboledas et al. (2008). On the other hand, for nighttime the star photometer EXCALIBUR acquires direct star irradiance measurements at 380, 436, 500, 670, 880, 940 and 1020 nm (nominal wavelengths). The FWHM range between 7.7 and 11.2 nm for the different filters. EXCALIBUR star photometer presents the difference to those previously developed by Leiter et al. (1995) that uses a CCD camera as detector device. Further details about this instrument can be found in Perez-Ramirez et al. (2008a, b). Both instruments operated in the Andalusian Center for Environmental Research (CEAMA) located in the city of Granada (37.16 • N, 3.60 • W, 680 m a.s.l.; South-East of Spain). Granada is a non-industrialised and medium-sized city, with a municipal population around 250 000 inhabitants and twice including its metropolitan area. The city is situated in a natural basin surrounded by mountains, with the highest hills over 3000 m a.s.l. located at the Southeast of the basin. The Mediterranean-Continental climate conditions prevailing at this site are responsible for large seasonal temperature differences, providing mild winters and hot summers. On the other hand, most of the rainfall is registered in spring-winter. The summers are usually very dry, with very few rainfall events. For the past 50 yr, according to Spanish Meteorological Agency (AEMET; http://www.aemet.es), the mean annual rainfall in this area is 370 mm. The relative humidity (RH) is larger in winter (with average values of 60 % and 70 % for daytime and nighttime, respectively) than in summer (38 % and 49 % for daytime and nighttime, respectively). Finally, the RH shows a clear diurnal cycle in all the seasons with the largest values at night and the lowest values at noon.
Due to its location in the Iberian Peninsula, the study area is usually affected by air-masses with different origins. By one hand, air-masses with origin in the Sahara desert (North Africa) are usually associated with large amounts of mineral dust particles (e.g., Lyamani et al., 2005Lyamani et al., , 2006aGuerrero-Rascado et al., 2009;Valenzuela et al., 2012a, b). These airmasses are more frequent during the summer season (Valenzuela et al., 2012b). On the other hand, the air-masses coming from Europe or from the Mediterranean basin can transport large loads of anthropogenic particles to the study area (e.g., Lyamani et al., 2006a, b). Atlantic air-masses affecting the study area are usually associated with low aerosol loads (e.g., Lyamani et al., 2010), being more frequent in the winter season. Furthermore, the main local anthropogenic source of aerosol particles is traffic, and also domestic heating (based on fuel oil combustion) in winter (e.g., Lyamani et al., 2008Lyamani et al., , 2010Lyamani et al., , 2011Titos et al., 2012).

Methodology
Attenuation of sun or star irradiance through the Earth's atmosphere follows the Beer-Bouger-Lambert law that is given by (for an average Sun/star-Earth distance): where V (λ) is the signal measured by the photometer, V 0 (λ) is the extraterrestrial signal (what is known as calibration constant), m r is the optical relative air mass and δ atm (λ) is the total atmospheric optical depth.  (Holben et al., 1998). Using equation 1 and sun/star photometer measurements, the aerosol optical depth (AOD(λ)) at the selected spectral channels have been computed following the methods described in the works of Alados-Arboledas et al. (2003 and Perez-Ramirez et al. (2008a), for sun-and star-photometers, respectively. Uncertainties in AOD(λ) for the star-photometer EXCALIBUR are 0.02 for λ < 800 nm and 0.01 for λ > 800 nm , and for sun-photometer CIMEL are 0.02 for λ < 400 nm and 0.01 for λ > 400 nm (Holben et al., 1998). From the Angström turbidity formula AOD(λ) = βλ −α , least-squares fits (in a log-log scale) were applied to determine the coefficients α and β. The β parameter is related to particle concentration and represents the aerosol optical depth at 1 µm. In the solar spectrum, the Angström exponent α characterises the spectral features of aerosol particles and is related to the size of the particles; α > 1.5 are mainly determined by the fine mode (submicron aerosol particles), while α < 0.5 are largely determined by the coarse mode (e.g., Dubovik et al., 2002;Gobbi et al., 2007). In this work, the Angström exponent, α(436-880 nm), obtained at nighttime is computed from AOD at 436, 667, 880 nm and the Angström exponent, α(440-870 nm), obtained at daytime is calculated from AOD at 440, 670 and 870 nm.
Several authors have discussed how the spectral curvature of the Angström exponent α can provide further information about the aerosol size distribution (e.g., Eck et al., 1999;O'Neill et al., 2001O'Neill et al., , 2003Schuster, 2006;Gobbi et al., 2007). In this work, we use the simple graphical method proposed by Gobbi et al. (2007). In this method, for a bimodal size distribution with different fine (r f ) and coarse (r c ) modal radius and fixed widths of fine and coarse modes of σ f = 1.5 µm and σ c = 1.8 µm, respectively, the difference δα = α(440-670 nm) −α(670-870 nm) was computed and represented versus α(440-870 nm). This computation was also made taking into account different contributions of the fine mode to AOD at 670 nm (η). The computations were 9722 D. Pérez-Ramírez et al.: Columnar aerosol properties from sun-and-star photometry done using the Mie theory with a fixed aerosol refractive index m = 1.4-0.001i (Gobbi et al., 2007). Moreover, this method assumes that the particles are spherical, which has no significant impact on the results (Gobbi et al., 2007). It is important to note that due to the large uncertainties in the Angström exponent for low AOD(λ), this method is only applicable for AOD(670 nm) > 0.15 (Gobbi et al., 2007).
Finally, five-day backward-trajectories of the air-masses affecting the study area are calculated by the HYSPLIT-4 model (Draxler and Rolph, 2003). The meteorological data used to run the model are 6-hourly GDAS (Global Data Assimilation System, ftp://www.arl.noaa.gov/pub/archives/ gdas1/). The HYSPLIT model has been run twice a day at 00:00 and 12:00 UTC at 1500 m a.g.l.; this altitude was selected as representative of the whole atmospheric column (e.g., Toledano et al., 2009). To classify the air-masses that affected the study area we follow the classification scheme developed by Toledano et al. (2009). This method is based on the residence time of the trajectory within geographic sectors. We defined five major air-mass sectors: North Atlantic Ocean (maritime-polar), South Atlantic Ocean (maritimetropical), European Continent and Iberian Peninsula (continental), Mediterranean Sea (Eastern), North Africa (Southern) and the Local one.

Results
The data used in this work were acquired from 2007 to 2010. Aerosol optical depths obtained by the sun-photometer were cloud-screened using the algorithm proposed by Smirnov et al. (2000) (used in AERONET network). For nighttime, the star photometer cloud-free data were obtained applying the algorithm proposed by Perez-Ramirez et al. (2012b). Additionally, the days and nights that present less than 2 h of measurements were eliminated from the database. Monthly values with less than 8 days/nights of measurements were also eliminated. Monthly values are also plotted. There are some gaps in both AOD(λ) data series which are due to instruments maintenance and calibration as well as to bad meteorological conditions. From Fig. 1, both at daytime and nighttime, there are variations in AOD(λ) which are generally of random nature. However, these random variations are modulated by more regular longer period variations. Low values of AOD(λ) are more frequently found during the winter months, while large values are frequently obtained in summer months, both at daytime and nighttime. Figure 2 shows the temporal evolutions of daytime mean values of α(440-870 nm) and nighttime mean values of α(436-880 nm). Monthly values are also plotted. As for AOD(λ), there is an important variability in the Angström exponent values between day-to-day and night-to-night. During the daytime, α(440-870 nm) shows large values in winter and low values in summer months. However, during the nighttime these seasonal differences are not so remarkable. Table 1 presents a statistical summary of daytime and nighttime mean values of AOD(λ) and α for the whole study period; particularly the mean value, standard deviation (STD), median, maximum and minimum values, as well as the corresponding percentiles at 10, 25, 75 and 90 % (P10, P25, P75 and P90). ANOVA statistical tests were used to compare daytime and nighttime datasets. These tests are based on the analysis of variances. The null hypothesis indicates that day and night data series are not different. The ANOVA statistical tests reveal that at 95 % level the datasets of AOD(λ) and α obtained at daytime are statistically equal to those obtained at nighttime. During daytime, AOD(440 nm) ranges from 0.02 to 0.95 with a mean value of 0.18 ± 0.10, while α(440-870 nm) varies between 0.01 and 1.8 with a mean value of 1.0 ± 0.4. During nighttime, AOD(436 nm) varies from 0.02 to 0.68 with a mean value of 0.19 ± 0.11, while α(436-880 nm), varies from 0.1 D. Pérez-Ramírez et al.: Columnar aerosol properties from sun-and-star photometry Table 1. Daytime and nighttime spectral aerosol optical depth AOD(λ) and Angström exponent α statistics for the 4-year data series in the city of Granada; STD is the standard deviation; P10, P25, P75 and P90 are the corresponding percentiles at 10, 25, 75 and 90 %.

Parameter
Aerosol Optical Depth (AOD) at different wavelenghts (   up to 2.1 with a mean value of 0.9 ± 0.4. Thus, there are no statistical significant differences in the sets of daytime and nighttime data of AOD(λ) and α.
The standard deviations and percentiles of AOD(λ) and α are large, both at daytime and nighttime, indicating the large variability of the atmospheric aerosol loads and types, and associated with the variability in the synoptic conditions. In fact, as mentioned before, the different air-masses affecting the study area are responsible of aerosol variability (e.g., Atlantic air-masses are very clean while North African air-masses can transport large loads of mineral dust; Continental and Mediterranean air-masses usually transport anthropogenic particles). Another factor that can explain the large variability is the changeable meteorological conditions.  Rainfall events favour the aerosol wet deposition, leading to the reduction of the aerosol load. Additionally, very high values of AOD(λ), both at daytime and nighttime, can be also associated with extreme events such as pollution or biomass burning (e.g., Alados-Arboledas et al., 2011). For the whole study period, Fig. 3 shows the frequency distributions of AOD(440 nm) and α(440-870 nm) obtained at daytime, and AOD(436 nm) and α(436-880 nm) obtained at nighttime. Both AOD(λ) frequency distributions are unimodal with strong skewness at low values of AOD(λ). The AOD(λ) modal value is 0.13 both at daytime and nighttime. On the other hand, the Angström exponent shows bimodal distributions both at daytime and nighttime. The first mode is centred at 0.55 and 0.45 with 6 % and 7 % frequencies of occurrence at daytime and nighttime, respectively, which reflects the contribution of large particles associated with longrange transport of dust particles and with local re-suspended soil dust. The second mode is located at 1.25 and 1.05 with approximately 10 % frequencies of occurrence at daytime and nighttime, respectively, and evidences cases associated with a mixture of fine (mainly from anthropogenic origin) and coarse particles.
However, analysing only α values does not provide clear information about the changes of fine/coarse mode to the aerosol load. To take more insight about the day-to-night changes of α, we use the simple graphical method proposed by Gobbi et al. (2007). Figure 4 shows both daytime and nighttime δα versus α. The data presented in Fig. 4 correspond to daily or nightly mean values. The data that fall out of the diagrams in the upper region can be explained because they have fine mode radius (r f ) lower than 0.05 µm, while those in the bottom region can be explained by the use of a fixed refractive index (Gobbi et al., 2007).
From Fig. 4 changes in the fine mode fraction and radius between daytime and nighttime are observed, with increases of η and r f at nighttime. For AOD(670 nm) ranging between 0.15 and 0.3, at daytime most of the data present r f values ranging from 0.10 to 0.15 µm, while the fine mode fraction (η) varies a lot with values up to 70 % (Fig. 4a). The variability of r f and η, and also of α(440-870 nm) (from 0.07 to 1.63) indicate large variability in the aerosol types and sizes. At nighttime a clockwise rotation toward larger r f (up to 0.3 µm) and η (up to 99 %) is observed. For other ranges of AOD(670 nm), α is lower than 0.7 both at nighttime and daytime, and the clockwise rotation at nighttime is observed as well. Changes in η can be associated with differences in the aerosol supplies either by natural or anthropogenic emissions, and must be studied detailed for every season (see Sect. 4.3). The increase of r f at nighttime might be associated with aerosol aging (e.g., Reid et al., 1998Reid et al., , 1999Dubovik et al., 2002;Eck et al., 2001Eck et al., , 2003a.

Inter-comparison with surrounding AERONET stations
For the study period, Table 2 shows the mean AOD(440 nm) and α(440-870 nm) values obtained at 14 AERONET stations located in the Iberian Peninsula, Western Mediterranean Basin, North West Africa and Canary Island. For small urban areas in the Iberian Peninsula like Evora, Cáceres and Palencia, lower values of AOD(440 nm) than in the city of Granada are obtained. The α(440-870 nm) values at Granada are lower than those at Cáceres and Palencia. Atlantic flow advections are much more frequent in the West than in the East of the Iberian Peninsula (e.g., Querol et al., 2009), and the low aerosol load associated with these airmasses can explain the low values of AOD(440 nm) at Evora, Cáceres and Palencia. Additionally, the larger impact of Saharan dust outbreaks over the South of Spain (e.g., Querol et al., 2009;Toledano et al., 2007aToledano et al., , 2009) with large aerosol loads and low values of the Angström exponent (e.g., Lyamani et al., 2006a, b;Toledano et al., 2007a;Valenzuela et al., 2012a, b) can also explain the results obtained at Granada. On the other hand, although El Arenosillo is a remote station, the AOD in this site is quite similar to that obtained in Granada. This can be explained by the effects of anthropogenic industrial emissions in the South-West of Spain and by the Saharan dust intrusions over this remote station (e.g., Toledano et al., 2007b;Prats et al., 2008;Córdoba-Jabanero et al., 2011;Bennouna et al., 2011). The stations of Valencia and Barcelona present larger values of AOD(440 nm) and α(440-870 nm) than those at the station of Granada. These two sites correspond to bigger cities in Spain with considerable levels of local anthropogenic emissions. These sites are also affected by Saharan dust intrusions, but with less frequency compared to that at the station of Granada because Barcelona and Valencia are further to the dust sources in North Africa (e.g., Rodriguez et al., 2001;Estelles et al., 2007;Querol et al., 2009). In other sites in the Western Mediterranean like Avignon, Ispra, Rome, Toulon and Lecce, AOD(440 nm) and α(440-870 nm) are larger than those obtained in Granada and in other more polluted sites in the Iberian Peninsula such as Valencia and Barcelona. This is because these cities (Avignon, Ispra, Rome, Toulon and Lecce) are urban areas with high local anthropogenic emissions and quite affected by highly polluted air-masses from Europe (e.g., Pace et al., 2006;Santese et al., 2008;Mazzola et al., 2010). In addition, they are also affected by Saharan dust intrusions (e.g., Perrone et al., 2005;Santese et al., 2008;Meloni et al., 2007;Pavese et al., 2009). It is worth noting that the station of Blida is  For the nighttime, to our knowledge currently there are no systematic measurements of AOD(λ) in the surrounding areas to our station and, thus, it is not possible to make any comparison. However, AOD(λ) and α mean values obtained in this work at nighttime are quite similar to those obtained during daytime (Table 1). But it is important to note that these findings are only relevant to sites with similar characteristics to that in the station of Granada.

Seasonal evolution of aerosol optical properties
To analyse the seasonal variations of aerosol optical properties, the data are grouped in four seasons: winter (January, February and December of the previous year), spring (March, April and May), summer (June, July and August) and autumn (September, October and November). For the study period, Table 3 shows the seasonal mean values of AOD(440 nm) and α(440-870 nm) obtained at daytime, and AOD(436 nm) and α(436-880 nm) obtained nighttime. Figure 5 shows the seasonal Box-Whisker diagrams of AOD(440 nm) and α(440-870 nm) obtained at daytime, and AOD(436 nm) and α(436-880 nm) obtained at nighttime during the study period. Figure 5a reveals an evident seasonal pattern in AOD(λ), characterised by large values in springsummer and low values in autumn-winter. For α Fig. 5b also reveals a seasonal pattern characterised by low values in summer and in early-autumn and large values in winter and early spring, both at daytime and nighttime. Although there are no statistical differences between day and night seasonal patterns of AOD(λ) and α, according to the mean values of α (Table 3) this parameter shows slightly more remarkable pattern at daytime than at nighttime. The seasonal patterns of AOD(λ) and α can be explained by several reasons. In late-spring, summer and early autumn, the higher frequency of Saharan dust intrusions in the study area and the low ventilation rates of air-masses in the Western-Mediterranean basin (e.g., Millan et al., 1997;Rodriguez et al., 2001;Lyamani et al., 2006a, b) can explain the large values of AOD(λ) and the relatively low values of α. Moreover, the intense atmospheric convective activity prevailing in this area in these months, together with the aridity of the soil during this particular period, provides a mineral dust (coarse particles) loading to the atmosphere from local soil. Another reason of these high AOD(λ) is the low rainfall rates, which are responsible of the aerosol load reduction. On the other hand, in late-autumn, winter and early-spring the aridity of the soil is quite reduced mainly by the rainfall and there are also less Saharan dust intrusions in the study area. In addition, the rain and clean Atlantic air-masses are more frequent in this period. All these reasons can explain the low values of AOD(λ) and large α values.
In the hot months the convective activity is more intense at daytime than at nighttime. As the convective activity ceases during the nighttime large particles can deposit, which can induce an increase in α (explaining their slightly larger values at nighttime). During the late-autumn and winter, local anthropogenic emissions are more relevant, which mainly supplies fine particles in the atmosphere. These emissions are more active during daytime, which can explain the larger values of α obtained during daytime compared to nighttime. Figure 6 shows the frequency distributions of daytime and nighttime AOD(λ) and α for the different seasons. For AOD(440 nm) and AOD(436 nm) obtained at daytime and nighttime, respectively, all the distributions are unimodal with skewness at low AOD(λ) values. This skewness changes to slightly larger AOD(λ) values in summer. In all the seasons, the differences of the AOD(λ) modal values between daytime and nighttime are negligible. On the other hand, for α both at daytime and nighttime they present similar shape, and can be adjusted to a bimodal distribution. The first mode is ∼ 0.1-0.5 and can be associated with the presence of dust particles either from long range transport or local soil. Actually, this mode presents larger frequencies in summer when the influence of dust particles is more relevant. The second mode is observed at larger values of α (∼ 1.1-1.4) and is associated with the presence of anthropogenic particles. Additionally, differences in the modal values of α can be observed between daytime and nighttime. But these differences are very low (∼ 0.1) to obtain any conclusion about the changes in particle types between daytime and nighttime. Actually, only in autumn relevant differences are observed with the lack of the first mode at nighttime (corresponding to coarse particles), which can be associated with the presence of a strong Saharan dust outbreak on 6 September 2007 at daytime (Guerrero-Rascado et al., 2009). However, the study of δα can reveal further differences. In this sense, Fig. 7 and  and winter season, respectively. We focus on these seasons to avoid the changeable synoptic and meteorological conditions in spring and autumn and, thus, study the differences in particle types between day and night for hot and cold months, respectively. Every point presented in the figures corresponds to a single measurement by the sun-photometer or 30 min average measurements by the star-photometer. For the data plotted in Fig. 7, taking into account that only data with AOD(670 nm) > 0.15 are used, there are no important differences between daytime and nighttime AOD(λ). Actually, mean values of AOD(440 nm) at daytime are 0.34 ± 0.11, 0.33 ± 0.12, 0.30 ± 0.06, 0.32 ± 0.  Fig. 7 during the daytime, although there are differences among the different summers, two patterns of δα are clearly differentiated. The first one presents α(440-870 nm) < 0.5 and corresponds to η < 30 %, r f < 0.2 µm and δα > 0. Backward-trajectories analysis and images of MODIS sensor on board of TERRA and AQUA satellites (http://modis.gsfc.nasa.gov) (graphs not shown) revealed that these data are mainly associated with Saharan dust intrusions. The work of Basart et al. (2009) showed that for sunphotometry measurements at the Sahara-Sahel deserts, mineral dust particles presented δα approximately between −0.5 and 0.1, α < 0.3, η < 40 % and r f ∼ 0.3. The differences between these values and those obtained in this study can be explained by the mixture of dust with local anthropogenic particles, and by the deposition of the larger particles during their transport to the study area. This pattern is more remarked during the years 2007 and 2008 which was affected by more intense and frequent Saharan dust outbreaks. The second pattern is characterised by α(440-870 nm) > 0.75, and most of the data present δα > 0 and η of ∼ 30-70 %, and r f of ∼ 0.10-0.15 µm. These data are usually associated with predominance of fine particles (e.g., Basart et al., 2009). Particularly, the influence of sporadic biomass-burning events in this season and anthropogenic particles transported from polluted Mediterranean and European areas can explain these values of δα, α(440-870 nm) and η (e.g., Lyamani et al., 2006a, b;Alados-Arboledas et al., 2011). In fact, in summers 2009 and 2010 the larger influence of continental and Mediterranean air-masses explains the larger presence of this pattern. The clustering is toward a r f line and the extension to higher AOD happens perpendicular the r f line, into larger fine mode fraction. This indicates growth of the fine mode by aging processes (Gobbi et al., 2007). The study of Gobbi et al. (2007) also showed patterns like that one for highly polluted places like Beijing (China), Kanpur (India), Ispra (Italy), Mexico DC (Mexico) or Goddard Space Flight Centre (USA).

D. Pérez-Ramírez et al.: Columnar aerosol properties from sun-and-star photometry
During nighttime, the two patterns of daytime data in Fig. 7 are not observed, being found a clockwise rotation of the data (η ranging from 30 to 99 %, r f ranging from 0.1 to 0.3 µm and δα < 0). This implies changes in the fine mode particle characteristics between daytime and nighttime. Although quick changes of the synoptic conditions between day and night can be found, no systematic differences of day-tonight air-masses affecting the study area are expected and, thus, differences of air-masses can not explain these day-tonight changes. The supply of local mineral particle from the ground during daytime by the strong convective activity can explain the lower values of η. The increase of r f at nighttime can be explained by aerosol aging processes such as coagulation and condensation (Seinfeld and Pandis, 1998). The combined increase in η and r f can be associated with hygroscopic growth (Gobbi et al., 2007). Although the relative humidity at the surface is low (∼ 50 %), larger values can be obtained at higher altitudes and favour hygroscopic growth processes. However, we must be careful about this statement because of the absence of relative humidity vertical profiles, information about mixing conditions and aerosol  Fig. 8b, mean AOD(440 nm) is 0.26 ± 0.14 at daytime and mean AOD(436 nm) is 0.28 ± 0.10 at nighttime. However, the patterns of δα versus α between both years are very different.
The inter-annual variation of the air-masses affecting the study area explains the differences of the aerosol particles between 2008 and 2009. According to the five-day backwardtrajectories analysis and MODIS satellite images, during the late winter of 2008 some Saharan dust intrusions affected the study area (approximately 35 % of the days and nights used in Fig. 8a). From Fig. 8a, at daytime two patterns can be observed. The first one presents δα > 0, α < 0.5, η < 30 % and r f < 0.2 µm, and is mainly associated with dust particles transported from North Africa. The second one is characterised by δα > −0.5, α > 0.75, η up to 90 % and most data with r f below 0.15 µm. This last pattern is associated with a mixture of different aerosol types, both mineral dust and anthropogenic particles. For nighttime, the two patterns previously mentioned are also observed and in both cases η values are lower than those observed at daytime. As commented before, anthropogenic emissions are more active during the daytime than during the nighttime, and can explain the predominance of the fine mode particles at daytime. This result is in good agreement with the comment before about the slightly larger values of α observed during daytime in winter (Fig. 5b).
In winter 2009 (Fig. 8b) the synoptic conditions were different to those in 2008, with a lot of rain periods. The AOD(670 nm) > 0.15 were obtained only during the period from 18 to 26 February. For this short period, the air-masses affecting the study area were mainly of continental origin. Particularly, larger values of η and r f are obtained at nighttime. During the daytime, η ranges between 30 % and 70 %, r f between 0.1 µm and 0.2 µm, and δα between −0.25 and 0.25, while at nighttime η ranges between 50 % and 99 %, r f between 0.15 µm and 0.25 µm, and δα between −0.75 and 0.1. Aging processes can again explain this increase of the fine mode radius. Particularly the hygroscopic growth is an important factor at high relative humidity for anthropogenic particles (e.g., Kotchenruther et al., 1999;Raut and Chazette, 2007;Randriamiarisoa et al., 2006;Veselovskii et al., 2009). But again we have to insist on the limitations of the methodology proposed that does not allow the separation between any processes. On the other hand, from 24 to 25 February (stars symbol in Fig. 8b) there were quick and intense pollution plumes associated with air-masses originated in the European continent and in the Mediterranean Sea, which can explain the differences of r f and η to those observed for the other days and nights during this winter.
The differences in the results between years 2008 and 2009 do not allow for the obtaining of any conclusion about the day-to-night aerosol types in winter. It is associated with the complexity of the atmosphere during this season at the study area, which includes changes in the synoptic conditions, in the planetary boundary layer and in aerosol emission sources.

Columnar aerosol properties dynamic: study cases
The comment above showed the similarity in AOD(λ) and α between daytime and nighttime for aerosol climatology. In spite of the continuity in the mean values, shorter time scales are needed to study "day-to-night" aerosol dynamic  and also to know the effects of the changes of the air-masses. Two studies are presented focussing on summer and winter seasons with very different atmosphere characteristics. Figure 9 shows, for the period between 18 June and 2 July 2008, the time evolutions of AOD(440 nm) at daytime and AOD(436 nm) at nighttime (Fig. 9a), and also time evolutions of α(436-880 nm) at daytime and α(436-880 nm) at nighttime (Fig. 9b). From Fig. 9 can be observed a good continuity in day-tonight AOD(λ) values. Similar continuity is also observed in α (Fig. 9b). Mean values of AOD(λ) and α and the origin of the air-masses for the days and nights of Fig. 9 are showed in Table 4. As can be observed mean values of AOD(440 nm) and α(440-870 nm) at daytime are similar to those obtained at nighttime, with smooth variations in day-to-night evolutions. Day-to-night differences between mean values are remarkable only when there are changes in the origins of the air-masses between day and night, although these variations of AOD(λ) and α are coherent with the smooth time evolutions showed in Fig. 9. However, the variability in the airmass affecting the study area induces important differences in the day-to-day and night-to-night evolutions of AOD(λ) and α. Atlantic air-masses present low AOD(λ) while those from Sahara desert present large AOD(λ) and low values of α. Continental and Mediterranean air-masses are usually responsible of large values of AOD(λ) and α.

(a) Summer season: 18 June-2 July 2008
On 18 June the study area was affected by Atlantic airmasses while on the following days and nights, the airmasses came from the North of Morocco and later from the South of the Iberian Peninsula with implies changes in the time-evolutions (Fig. 9), increasing AOD(λ) (maximum values up to 0.25 on 20 June) and α (maximum values up to 1.5 on 20 June). On 22 and 23 June the air-masses affecting the area were originated in the Sahara desert transporting mineral dust particles and inducing a sharp increase of AOD(436 nm) on 22-23 June night (maximum values up to 0.7), while the posterior changes to Atlantic air-masses cleaned the atmosphere explaining the sharp decrease of AOD(λ) (minimum values down to 0.14). From 25 June afternoon the air-masses over the study area changed to those with origin in the Iberian Peninsula increasing the aerosol load. However, α reaches very low values (∼ 0.45) that in the study area are usually associated with long-range transport of Saharan dust particles (Lyamani et al., , 2006Valenzuela et al., 2012a, b). The more in depth analysis of backward-trajectories for this night revealed that the air-masses reaching the study area at 4000 m a.s.l. on 25, 26 and 27 June came from the Sahara desert. In this sense, the work of Guerrero-Rascado et al. (2009) showed that long-range transport of mineral Saharan dust particles can reach altitudes above 3000 m a.s.l. in the study area. For this particular case, range corrected signals of correlative lidar measurements during these days and nights showed strong backscattered signal at these high altitudes (graphs not shown, but can be consulted at http: //atmosferera.ugr.es). Therefore, presence of Saharan dust particles at these altitudes can explain the low values of α for these days and nights.
On 27-28 June night, the air-mass precedence moves to the East, with origin between the South of the Iberian Peninsula and the North African countries. No transport of Saharan dust particles is observed (again by checking correlative lidar measurements), and α(436-880 nm) increases (∼ 1.2-1.5). These values of α can be explained by the increase in the anthropogenic emissions in the North Africa countries (Rodríguez et al., 2011). From 28 June to 1 July the study area was affected by Mediterranean air-masses which are associated with very variable aerosol loads and types  because this sea can be a reservoir of pollutants from the surrounding countries (Lyamani et al., 2006;Rodríguez et al., 2011) and also of dust particles transported from the Sahara desert (Meloni et al., 2007;Rodríguez et al., 2011). During this short period, large AOD(λ) remain. However, there is variability of α. Until 29 June large values of α (values up to 1.6) suggest predominance of fine particles associated with anthropogenic particles, while the decrease of α from late 29 to early 30 June (values ∼ 0.75) suggest some influence of Saharan dust particles over the Mediterranean. This last has been checked by MODIS images and also explains the sharp increase of AOD(λ) (maximum values up to 0.55 on 30 June). Finally, the decrease of AOD(λ) Table 4. Mean values of day-and night-time aerosol optical depth (AOD) and Angström exponent as well as the associated air-mass types that affected the study area for every day and night between 18 June and 1 July 2008. Air-masses backward-trajectories were computed by HYSPLIT model at 1500 m a.g.l.  (values down to 0.1) during the following days and nights is explained by the changes towards North Atlantic air-masses. In spite of the continuity of α between daytime and nighttime, the changes in the day-to-night evolutions make us study the behaviour of α in detail. In this sense, Fig. 10 shows δα versus α for the data showed in Fig. 9, both at daytime and nighttime.
From 19 to 20 June (Fig. 10a, b, c) with relatively low aerosol load, α is larger than 1 both at daytime and nighttime, but there are remarkable differences in δα. Daytime values present r f < 0.1 µm and η < 50 %, being around one half of the data with η < 30 %. However, during nighttime the pattern is very different, presenting all the data r f > 0.1 µm and η > 50 %, and around one half of data with r f > 0.15 µm and η > 75 %. On 22 and 23 June (Fig. 10d, e), with presence of Saharan dust particles, at daytime r f is lower than 0.15 µm and η lower than 30 %, while at nighttime the opposite occurs (r f > 0.15 µm and η > 30 %). On 24 June (Fig. 10f) the Atlantic air-masses affecting the study area induced lower aerosol load and larger α values, being the values of r f and η similar to those just discussed for Fig. 10a, b, c.
From 25 afternoon to 27 morning, (with dust particles above 3000 m a.s.l.) Fig. 10g, h shows that at daytime r f is lower than 0.15 µm, and η is below than 30 % on 25 and 26 June (except some data at the beginning of 25 June associated with the changes in the air-masses), and below 50 % on 27 June. During nighttimes there are again differences with larger values of r f (values up to 0.3 µm) and η (up to 90 %). On the night of 27-28, there were changes of the air-masses that induces larger values of r f and η as well. For the following period from 28 to 29 June affected by transport of fine particles from the Mediterranean sea, Fig. 10j at daytime r f < 0.1 µm and η < 70 %. The more variability of η on these days might be associated with the contribution of coarse particles from the ground as the convective activity increase. On the night of 28-29, η does not change significantly, but r f present important changes toward larger values (between 0.15 and 0.20 µm). Finally, on the night of 29-30 June and on 30 June (daytime) the Mediterranean airmasses transported Saharan dust particles. In this case, the pattern observed is again of larger values of r f and η during the nighttime.
The findings of Fig. 10 reveal increases of r f and of η during nighttime for the different aerosol loads and types. This has been found for different types of air-masses, and also when there are changes in the air-masses between day and night. These results also agree with the general results obtained in Fig. 7 for the summer season, and can be explained by the same reasons. Again, with the methodology used here we can not differentiate between the different processes that induce changes in particle types between day and night.  Figure 11 shows for the period between 21 January and 1 February 2008 the temporal evolutions of AOD(440 nm) at daytime and AOD(436 nm) at nighttime (Fig. 11a), and also temporal evolutions of α(436-880 nm) at daytime and α(436-880 nm) at nighttime (Fig. 11b). Table 5 shows mean values of AOD(λ) and α, and also the origin of the air-masses for these days and nights. Again, from Fig. 11, we can observe a good continuity in day-to-night evolutions of AOD(λ) and α. Particularly, α is usually slightly lower at nighttime and agrees with the comment in Fig. 5 for the winter season. The larger errors in AOD(λ) at nighttime can explain the fluctuations of α in the nights with low aerosol load (error in α are larger for low AOD(λ)).

(b) Winter season: 21 January-1 February 2008
As for the previous study case during the summer, mean values of AOD and α at daytime are similar to those obtained at nighttime and only the differences are notable when changes in the air-masses origin are presented (although the differences are smoothed as shown in Fig. 11). However, there are again large day-to-day and night-to-night variations associated with the changes of the air-masses that reach the study area. On 21 January the air-masses that affected the study area were from European origin. On 22-25 January the study area was under the influence of clean Atlantic airmasses, which explain the low aerosol loads (AOD(λ) < 0.2 for these days and nights) with large variability of α.
From 26 to 29 January the situation changed and the airmasses came from the European and Mediterranean sectors which transported anthropogenic particles to the study area. These air-masses produced an increase of AOD(λ) and α (Ta-ble 5). The larger aerosol loads during these days and nights allow the study of δα versus α (Fig. 12). On 26 January, at day time, it is observed that δα > 0, r f < 0.1 µm and η between 30 % and 50 % (Fig. 12a). However, at nighttime we observe δα > 0 and most data present r f < 0.1 µm and η < 30 %. On 27 January very similar results to those obtained in the previous day and night are observed (Fig. 12b). The larger contribution of fine mode particles at daytime can be explained by the larger anthropogenic emissions during daytime. It is worth noting that although r f is very small (< 0.1 µm) both at daytime and nighttime (Fig. 12), daytime values are generally larger than nighttime ones, which can be explained by the contribution of larger particles by road traffic or by secondary aerosol originated from primary anthropogenic emissions (Lyamani et al., 2010).
On 28 and 29 January the study area was under the influence of Mediterranean air-masses influence, with large values of α (between 1.2 and 1.5 at daytime and from 0.8 to 1.35 at night-time) indicating the predominance of fine particles. Furthermore, on 28-29 January r f and η patterns are similar to those obtained on 26 and 27 January (Fig. 12c). Finally, from 30 January to 1 February the air-masses precedence changed to Atlantic which made AOD(λ) and α decrease.
This study case remarked the continuity of day-to-night columnar aerosol properties in winter, but do not allow conclusions about the δα. Indeed, together with Fig. 8, remarked the complexity of any general conclusion for winter due to the presence of different aerosol types, different intensity of aerosol aerosol sources (natural and anthropogenic) and differences in the atmosphere's temperature in planetary boundary layer and the changeable synoptic conditions in terms of rain and wind velocity. Thus, more efforts are needed using high technology instrumentations to characterise vertical atmosphere thermodynamics and vertical aerosol properties.

Conclusions
An analysis of correlative day-to-night columnar aerosol properties at Granada (South-East of Spain) has been presented. This study was possible thanks to the recent developments in star photometry combined with the well-known sun photometry technique (particularly using the CIMEL instrument). The combination of correlative measurements with both techniques has allowed the study of day-to-night aerosol optical depth (AOD), Angström exponent α, and spectral dependence of the Angström exponent (δα).
Time evolutions of AOD and α have revealed good continuity and coherence between daytime and nighttime values, both for the whole database and for special study cases. Moreover, the statistical analyses have shown no significant differences between daytime and nighttime for both parameters. For daytime values, the comparison carried out with other AERONET surrounded stations has revealed that the values obtained at Granada do not by differentiate remarkably with those obtained in the surrounded areas, in spite of the differences in synoptic conditions and aerosol sources. However, at nighttime no comparisons have been possible due to the lack of continuous measurements and, thus, the continuity observed in the long-term columnar aerosol properties at Granada cannot be extrapolated to other aerosol sites because of differences in meteorological conditions and aerosol sources. In this sense, the development of continuous nighttime measurements worldwide is needed to better understand aerosol dynamics globally. A seasonal pattern for AOD has been obtained, both at daytime and nighttime, characterised by lower values in winter and larger values in summer. The Angström exponent has also presented a seasonal pattern but with lower values in summer and larger values in winter. No statistical differences in AOD and α seasonal pattern between day and night have been found out. These patterns have been associated with the different soil conditions at each season, with changes in the synoptic conditions and rainfall, and with the intensity of the local aerosol sources.
The spectral difference of the Angström exponent (δα) as function of α has been studied, both at daytime and nighttime. Particular studies cases have showed the evolution of δα detailed. In summer, an increase of the fine mode radius (r f ) and of the fine mode contribution to AOD (η) has been observed at nighttime. This has been obtained for many different air-masses and aerosol loads and types, and has been explained by the changes of the local aerosol source emissions and meteorological conditions between daytime and nighttime. Moreover, the increase of r f during nighttime has been also associated with aerosol aging processes. In the winter season, for δα the situation is more complex due to the more variability of the synoptic conditions and aerosol sources. For low-polluted aerosol load, the special cases analysed have shown larger r f and η during daytime. It has been associated with the more intensive local anthropogenic sources, aerosol secondary aerosol formation and groundbased particles supplies by traffic. However, for highly pol-luted air-masses it is observed just the opposite, with an increase of r f and η at nighttime. This has been explained by aging processes or deposition of coarse particles during airmass transport. However, the reduced data points obtained under these conditions make us be careful. We would like to remark that the study of spectral variation of α alone does not allow distinguishing between different aerosol aging processes. Many of such processes compete, including changes in the aerosol load due to meteorological conditions, dry/wet deposition, coagulation/condensation processes, hygroscopic growth or the supply of particles from the ground or from anthropogenic activities.
The results presented allow the scientific community to advance in the knowledge of day-to-night columnar aerosol dynamics, both for large database and shorter temporal scales. However, we would like to point out that the differences between daytime and nighttime aerosol particles are only referred to a particular site with particular conditions, and these results cannot be extrapolated to any other place. In this sense, more efforts of the scientific community should be done to improve the knowledge of columnar aerosol properties at nighttime.