Multi-year Aerosol Robotic Network (AERONET) and Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth (AOD) data are used to study AOD weekly variations at the global scale. A clear weekly cycle of AOD is observed in the United States (U.S.) and Central Europe. AOD during the weekday is larger than that during the weekend in 36 out of 43 AERONET sites in the U.S. The average U.S. weekend effect (the percent difference in AOD during the weekday and the weekend) is 3.8%. A weekly periodicity with lower AODs on Sunday and Monday and higher AODs from Wednesday until Saturday is revealed over Central Europe and the average weekend effect there is 4.0%. The weekly cycle in urban sites is greater than that in rural sites. AOD during the weekday is also significantly larger than that during the weekend in urban AERONET sites in South America and South Korea. However, a reversed AOD weekly cycle is observed in the Middle East and India. AODs on Thursday and Friday, the “weekend” for Middle East cultures, are relatively lower than AODs on other days. There is no clear weekly variation of AOD over eastern China. The striking feature in this region is the occurrence of much higher AOD on Sunday and this phenomenon is independent of season. The analysis of MODIS aerosol data is in good agreement with that of AERONET data.
 Atmospheric pollutants show several interesting periodic cycles due to nonanthropogenic or anthropogenic influences, including diurnal and seasonal cycles. Urban morning and afternoon rush hours lead to a marked increase in surface fine particle concentration [DeGaetano and Doherty, 2004; Yu et al., 2006]. A typical pattern of the aerosol optical depth (AOD) increase by 10–40% occurs during the day at most urban/industrial Aerosol Robotic Network (AEORNET) sites [Smirnov et al., 2002; Xia et al., 2006]. Industrial and urban emissions of aerosols in heavily populated regions are present year-round but due to several meteorological factors, AOD is often greater during the summer than during the winter [Holben et al., 2001; Chu et al., 2003; Eck et al., 2005; Li et al., 2007]. Weekly periodicities of atmospheric pollutants have been reported, although there are no known nonanthropogenic or meteorological mechanisms responsible for these weekly periodicities. Measurements indicate that NOX and O3 in Phoenix, Arizona exhibit a significant weekly cycle. NOX and O3 during the weekday are significantly larger/less than that during the weekend. These cycles are likely due to the weekly pattern of vehicle traffic related emissions [Shutters and Balling, 2006]. The effect of a 7-d cycle of human activity on tropospheric NO2 has been studied on the global scale using the Global Ozone Monitoring Experiment (GOME) data. A clear Sunday minimum of tropospheric NO2 can be seen in cities in the United States (U.S.), Europe and Japan. This weekly cycle of tropospheric NO2 is suggested to be associated with the reduced industrial activity and traffic in industrialized countries during the weekend [Beirle et al., 2003]. These clear weekly cycles of atmospheric pollutants are often called the “Weekend effect” [Cleveland et al., 1974].
 The concentration of near-surface aerosol properties is also a function of the day of the week, with significantly lower concentrations observed on Saturday and Sunday in urban regions in the U.S. [DeGaetano and Doherty, 2004; Shutters and Balling, 2006]. Surface PM10 (aerosol of diameter less than 10 μm) concentration during the weekday was higher than that during the weekend in an urban area in Taiwan and in eastern China [Tasi, 2005; Gong et al., 2007]. A clear weekly cycle of column-integrated aerosol properties, for example, AOD, has also been reported in some regions. Jin et al.  found a weekly cycle of AOD using ground-based remote sensing data in New York City, with maximum AOD on Wednesday and lower AOD on Saturday and Sunday, but only in summer and under some conditions. Statistical analyses of AOD data from 14 AERONET sites over Central Europe showed a clear weekly periodicity, with lowest values on Sunday and Monday, but greatest values from Wednesday until Saturday [Bäumer et al., 2008]. The weekly cycle of AOD was only reported on a limited scale, so the question still remains whether there is a weekly cycle of aerosol loading with a considerable vertical extent on larger horizontal scales than the urban scale [Bäumer et al., 2008]. This is the primary focus of the current study. To this end, two well-known aerosol data sets are used to study the weekly variation of AOD. One is the AERONET data set. AERONET is a federated network of a global ground-based remote sensing of aerosol optical properties. AODs at several wavelengths are derived from standardized instruments and procedures. Satellite measurements allow a new and independent approach to determine the columnar loading of airborne particles and therefore the AOD. Using, for example, the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor on the Terra and Aqua satellites, the entire globe is monitored with a single instrument under the same conditions and over long periods of time. Hence one would expect further insight on AOD weekly cycles from the analysis of satellite data. The MODIS/Terra aerosol product is used in this study due to its long-term data record and high accuracy [Remer et al., 2005; Levy et al., 2007].
 The paper is organized as follows. The introduction to data and methodology is presented in the following section. Weekly cycles of AOD derived from AERONET and MODIS data are presented in sections 3 and 4, respectively. Discussion and conclusions are presented in the last two sections.
2. Data and Methodology
 AERONET is a ground-based internationally federated, globally distributed network of automatic sun/sky radiometers dating back to 1993. The instrument, measurement sequences, accuracy (∼0.01 in visible and near infrared) and cloud screening procedure are described in several references [Holben et al., 1998; Eck et al., 1999; Smirnov et al., 2000]. The analysis in the current study uses the daily average quality-assured data (Version 2.0/level 2.0). The data were downloaded from the AERONET web-site (http://aeronet.gsfc.nasa.gov). There were approximately 650 temporal or permanent AERONET sites as of 2007, but only 169 sites were used that met the following requirements: the data are available for at least 1 year; the site is located over land.
 MODIS is one of the first passive satellite radiometers designed to retrieve aerosols over land and ocean. Kaufman et al.  introduced a strategy for retrieving aerosol over land from MODIS. The algorithm has been modified periodically and designated by different collections since the launch of MODIS [Remer et al., 2005]. A fixed relationship between the visible surface reflectance (470 nm and 660 nm) and the 2130 nm reflectance in previous versions of the algorithm has been replaced by a dynamic one in the latest version (Collection 005) that varies with the vegetation index and scattering angle [Levy et al., 2007]. Preliminary validations of this collection show much improved retrievals of AOD, although uncertainties remain because of highly simplified aerosol models used in some regions [Levy et al., 2007; Mi et al., 2007]. The MODIS Terra collection 005 level-3 daily AOD data from March 2000 to May 2007 are used in this study. The spatial resolution of MODIS data is 1° × 1°.
 All daily AODs for a week are expressed as a percentage departure (%) from the weekly average. This sampling procedure renders systematic weekly cycles more evident, similar to the diurnal analysis of Smirnov et al. . Only those weeks meeting the following requirements are included in the analysis of the weekly cycle of AOD: AOD should be observed in at least 2 d for a week; at least 1 d should be among the weekdays (Tuesday to Friday) and at least one other should be among weekend days (Saturday to Monday). We present the weekly cycle of AOD at the wavelength of 440 nm for AERONET data and at a wavelength of 550 nm for MODIS data, unless otherwise noted. We chose to define the weekend effect as the percent difference in the average AOD percent departure from the weekly average on Wednesday, Thursday, and Friday and that on Saturday, Sunday, and Monday. We averaged AOD over 3-d periods rather than two for more robust statistics. In the following figures and discussion, statistical significance is indicated only when it exceeds the 90% confidence level based on the t-test technique.
3. Weekly Cycle of AOD From the Perspective of AERONET Data
Figure 1 presents the AOD percent differences in the average AOD percent departure from the weekly average on Wednesday to Friday and that on weekend (Saturday to Monday) at 169 AERONET sites. The red and blue circles represent positive and negative values, respectively. The filled circles represent significant difference exceeding the 90% confidence level. The results show a clear and spatial homogenous weekend effect in the U.S. and Central Europe. AOD during the weekday is larger than that during weekend in 36 out of 43 AERONET sites in the U.S. and the mean weekend effect is about 3.8% (see Figure 2a). The weekly cycle in urban sites is generally more remarkable than that in rural sites. For example, the mean weekend effect is 6.5% for six urban sites and it is 3.0% for six rural sites in the U.S. (see Figure 3).
 A weekly periodicity with lower values on Sunday and Monday and higher values from Wednesday until Saturday is observed at most AERONET sites over Central Europe (45°–55° N; 0°–20° E) (see Figure 1). The mean weekend effect is 4.0% on average (see Figure 2b). Quite different magnitudes in the weekend effect are also observed in urban versus rural sites (see Figure 4). AODs during midweek are significantly larger than that on Monday and Sunday for six central European urban sites and the average weekend effect there is 8.0%. The weekend effect in central European rural sites is much lower and it is occasionally close to zero. The pattern in AOD weekly variations shown in this study is in good agreement with that of Bäumer et al. , in which the weekly AOD ranges (maximum minus minimum) are 10% in northern Italy and Switzerland and 20% in Germany and Greater Paris, respectively. These values are somewhat larger than those obtained in rural (8%) and urban (15%) sites in the current study. This is because different sites and different statistical methods are used in the calculation of the weekly AOD range. Another important feature supporting a clear AOD weekend cycle in the U.S. and Central Europe is that the minimum AOD occurs during weekend and the maximum AOD generally occurs during midweek (see Figure 5 for the maximum value and Figure 6 for the minimum value). The red circles indicate that the maximum AOD occurs during weekday or the minimum AOD occurs during weekend. The filled circles suggest that the values are significantly higher or lower than the weekly average. Both values are different from the weekly averages, exceeding the 90% confidence level in most sites, especially in urban sites. The weekly variability of AOD in the U.S. is somewhat different from that in Central Europe, although weekday AOD is generally larger than weekend AOD in the two regions. Saturday AOD values are slightly reduced in most U.S. stations (see Figures 2a and 3), but the maximum values occurred on this day in many Central Europe stations (see Figures 2b and 4). The minimum AOD in Central Europe occurs on Monday, however, it occurs on Sunday in the U.S. Urban AERONET stations in South America also exhibit a clear and significant weekend cycle. The average AOD during the weekday is larger than that during the weekend by 16% in Santiago and Sao Paulo (see Figure 1 the sites are represented by the largest red filled cycles in South America). However, a quite different weekly variation of AOD is observed in rural AERONET sites in this continent. The AODs from Monday to Thursday are lower than the weekly average and higher values occurred from Friday to Sunday, although the difference is not significant in most cases.
 A clear weekly cycle of surface PM10 concentration was observed over eastern China [Gong et al., 2007]. Maximum particle concentration occurred on Wednesday and lower values than the weekly mean occurred during the weekend. With regard to the column-integrated aerosol concentration, however, Sunday AODs are larger than the weekly mean and the minimum AOD occurs during the weekday at four Chinese mainland AERONET sites (see Table 1). Sunday AODs are the largest at Beijing (urban site) and Xianghe (rural site) and the values are significant larger than the weekly average by more than 13% and 17%, respectively. The maximum AOD occurs on Wednesday at Taihu, an urban site in the Yangtze Delta region, but Sunday AOD still exceeds the weekly mean by 10%. Significantly higher AOD also occurs on Sunday at Kanpur, a large city in India with a population of 2.8 million people, and at Bangkok, a mega city in Thailand with a population of nine million people (see Table 2). Sunday AODs exceed the weekly averages at Kanpur and Bangkok (Chulalongkorn) by 3.4% and 11.1%, respectively.
Table 1. AOD Percent Departure From Weekly Average AOD at 440 nm at Four AERONET Sites on the Chinese Mainland (%)a
The italic numbers indicates that the values are different from the weekly average at 90% significance level.
Table 2. Same as Table 1 but at Five AERONET Sites in India, Thailand, South Korean, and Japan
 Urban AERONET stations in South Korea exhibit a similar weekly cycle as that in the U.S. and Central Europe. For example, the AOD weekend effect is 32% in Seoul and 12% in Gwangju, respectively. In Osaka, Japan (population of 2.6 million) higher AOD occurs on Friday and Saturday and AODs on the other days are close to or lower than the weekly average (see Table 2). The AERONET stations in the Middle East show relatively lower AOD on Thursday and Friday and higher values on other days (see Table 3). In the Middle East the “weekend” is Thursday and Friday for Middle East cultures, therefore, the cultural lifestyles and industrial production schedules of this region are consistent with the observed AOD “Weekend effect”.
Table 3. Same as Table 1 but at Six AERONET Sites in Middle East
4. Weekly Cycle of AOD From the Perspective of MODIS/Terra Data
 The AOD weekend effect was observed in most AERONET sites in North America and Central Europe, as shown in section 3, however, a quite different weekly cycle was observed in AERONET sites in other regions, for example, in China, India and Middle East. The question that arises here is whether the AOD weekend effect exhibited by the ground-based data is a phenomenon on a larger scale. We cannot get a complete answer from the ground-based remote sensing data because the data are limited to the local scale. Therefore we turn to MODIS Collection 005 aerosol data and try to obtain a clear answer to this question in this section. We focus on four regions of the world with enhanced AOD values: (1) the U.S., (2) Central Europe, (3) eastern China and (4) India.
Figure 7 presents the AOD percent differences in the average AOD percent departure from the weekly average on Wednesday to Friday and that on weekend (Saturday to Monday) in North America. A distinct spatial pattern in AOD changes is found over much of the U.S., where AOD is generally higher on weekdays than during weekends in the eastern and western regions. A clear weekly periodicity of AOD is shown in Figure 8a (showing the weekly cycle of AOD in the eastern coastal region of U.S., the region surrounded by the box in Figure 7). The maximum AOD occurs on Thursday that is larger than the weekly averaged AOD by 4.0%. The minimum value that occurs on Sunday is less than the weekly averaged AOD by 3.8%. The geographical extent of the AOD weekend effect revealed by the MODIS AOD data suggests that it is unlikely to be only a phenomenon of the local scale in the U.S.
 The AOD during the weekday is significantly greater than that during the weekend in the industrialized countries such as France, Germany, and Netherlands (see Figure 9). The weekend effect in these countries is clearly shown in Figure 8b. Higher AODs than the weekly average AOD occur from Wednesday to Saturday. The minimum AOD, lower than the mean by 6%, occurs on Monday. The pattern of AOD weekly variation from MODIS is consistent with the results revealed from AERONET data.
 There is no clear weekly cycle of AOD over much of eastern China. AOD during the weekday in this region is generally less than that during the weekend, especially in the Yangtze delta region (Figure 10). The result is also consistent with that revealed from AERONET ground-based remote sensing data. This reversed weekend effect results from very large AOD that generally occurs on Sunday. As shown in Figure 8c, the maximum AOD occurs on Sunday that is larger than the weekly average AOD by 3% in the Yangtze delta region. AOD during the weekday is larger than that during the weekend in the South Korea, which is also in agreement with the result from AERONET data. A quite different weekly cycle of AOD is observed over much of India. AODs from Wednesday to Friday are generally less than the weekly average AOD and on the contrary, AODs during the weekend are generally larger than the weekly average AOD (Figure 8d).
 Several studies have reported clear weekly cycles of a few meteorological variables. Gordon  reported a significant but very small weekly temperature cycle for the northern hemisphere. Weekly periodicities of precipitation, tropical cyclone maximum wind speed as well as a difference between day and night wind speed in tropical cyclones over the northwest Atlantic region were reported [Cerveny and Balling, 1998]. Forster and Solomon  found a distinct large-scale weekend effect in the diurnal temperature range (DTR) and concluded that this short timescale and large-scale geophysical phenomenon was necessarily human in origin. Results showed that the DTR weekend effect depended on the season and opposite signals between winter and summer season were observed over much of eastern China [Gong et al., 2006]. The anthropogenic fraction of atmospheric aerosols was suggested to be responsible for those weekly cycles via its direct and indirect forcing, but this hypothesis has not yet been proven.
Gong et al.  reported that the PM10 concentrations over eastern China showed a significant weekly cycle with the largest values around midweek and smallest values in weekend in summertime. There was also a clear weekly cycle in meteorological variables. Surface wind speed and the horizontal wind speed below the 2-km level increase after Wednesday with higher values from Thursday to Friday. The temperature below the 850 hPa level shows considerably higher values from Monday to Wednesday, followed by lower values in the later part of the week. Furthermore, a weekly cycle of light rain frequency is evident, with maximum on Sunday and minimum on Wednesday. Gong et al.  suggested that the accumulation of aerosols likely triggered the changes in the atmospheric circulation through aerosol direct and indirect forcing, thereby suggesting human influence on weather. However, the analysis of AOD weekly variation on the basis of year-round AERONET and MODIS data in this study is not consistent with the pattern of surface particle concentration. In order to check whether the AOD weekly cycle has a seasonal dependence, similar analysis was repeated based on MODIS AOD data in summertime (June, July, and August) and in wintertime (December, January, and February), respectively. The results show that over the Yangtze Delta region, the average AOD during the weekday is still less than that during the weekend in summertime and wintertime (Figure 11 for summertime, by 0.5%). The larger value during the weekend is mainly because generally higher AOD occurs on Sunday. Sunday AOD over the Yangtze Delta region is larger than the weekly average by 0.7% and 1.1% in summertime and wintertime, respectively. The inconsistency between the weekly cycle of PM10 and AOD suggests that caution should be taken in using the surface measurement of aerosol to represent the column-integrated value in the analysis of aerosol loading on the weekly scale, although in summertime PM10 was positively correlated to AOD in Beijing [Xia et al., 2006]. The mismatch of the PM10 concentrations and MODIS AOD was observed in Beijing [Chu et al., 2003] and in Kanpur [Singh et al., 2004]. The surface particle concentration was suggested to be a good indicator of the column-integrated AOD only in those regions where pollution sources remained relatively invariable and air masses were stable. The absence of the AOD weekend effect in China suggests that the aerosol effect on the weekly cycle of meteorological variables and vice versa remains still open.
 The weekly cycle in AOD in the U.S. and Central Europe can be explained by the weekly cycle of individual transport that appears to be a major aerosol source in these two regions. The traffic source with its direct and indirect emissions turns out to be the most important source for PM10 pollution in Milan, accounting for approximately 60% of the PM10 mass [Marcazzan et al., 2003]. In Chicago, Illinois, motor vehicles are one of the three highest fine particle contributors, accounting for 31% of fine particle mass [Rizzo and Scheff, 2007]. The absence of the AOD weekend effect in China may indicate that Chinese aerosol emissions are dominated by power plants and heavy industry operating throughout the week rather than individual transport. This speculation is also supported by the fact that there is no NO2 weekend effect in China [Beirle et al., 2003]. In the early 1980s, traffic contributed to only 0.9% of total particle mass in Beijing [Wang, 1985]. The contribution by traffic to fine particle should have increased dramatically due to rapid growth in the number of automobiles from the mid 1980s to the present. More recently, observations show that vehicle exhaust contributes to 6% of PM2.5 mass concentration in Beijing [Zhang et al., 2004] and 3% in Nanjing, a city in the Yangtze delta region [Huang et al., 2006], respectively. These observations suggest an increasing contribution by vehicle exhaust to particle mass and AOD in eastern China during recent years, but the values are still much less than those in the U.S. and Central Europe. Therefore relative low traffic emissions in China as compared with that in the U.S. and Central Europe may account for the lack of a clear weekly cycle of AOD in eastern China. However, the question that arises here is: Why is the Sunday AOD higher than the weekly average AOD in eastern China? To understand these features in the weekly cycle of AOD in detail, further information about local emissions, wind speed and direction, and lifetime of aerosols is necessary. This is beyond the scope of the current study.
 The AOD weekend effect has been found from both the AERONET and MODIS data in the U.S. and Central Europe. The minimum AOD occurs on Sunday in the U.S., but in Central Europe, it occurs on Monday. Saturday AOD often exceeds the weekly average in Central Europe but in the U.S., Saturday AOD is close to the weekly average. Saturday is also a work free day for most professions in western countries (reduced industrial activity), but a preferred day for shopping or short trips (possible increase of individual transport) [Beirle et al., 2003]. This may account for higher AOD on Saturday in Central Europe, but clearly this cannot explain the normal level of AOD on Saturday in the U.S. region. Fuller examination of possible mechanisms is required for understanding the difference in the weekly variation of AOD between the U.S. and Central Europe.
 Multi-year AERONET and MODIS AOD data are used to study AOD weekly variations at the global scale. The main conclusions are as follows.
 The significant AOD weekend effect is observed over much of the U.S. and Central Europe. The AOD weekend effect is indicated by at least one of following features: (1) The AOD during the weekday is significantly greater than that during the weekend; (2) The maximum AOD occurs in the second half of the working week, usually from Wednesday to Friday or Saturday; (3) The minimum AOD occurs on Sunday and Monday. These characteristics are more evident in urban and industrial regions as compared with rural regions.
 Over much of eastern China, the AOD during the weekday is often less than that during the weekend. This is mainly because significantly larger AOD often occurs on Sunday. The phenomenon shows a weak seasonal dependence. At present we do not understand the causes of this AOD weekly variation in eastern China. The weekly cycle of column-integrated aerosol loading in the current study is inconsistent with that of surface particle concentrations [Gong et al., 2007]. This fact suggests that the surface measurement of aerosol is not a good indicator of the column-integrated value in the analysis of the weekly cycle. Care should be taken in the analysis of the weekly cycle of aerosol-meteorology interactions in China.
 A reversed weekly periodicity in AOD is observed over much of India and Middle East. The AOD during the weekday is often significantly lower than that during the weekend.
 Further research is required to more fully understand these interesting weekly variations of AOD and their effects on weather and climate.
 Our thanks are extended to the individual principal investigators and site managers who have maintained an active interest in the AERONET project and resulted in long-term observations. We also acknowledge the MODIS mission scientists and associated NASA personnel for the production of the MODIS data used in this research effort. The research is supported by the National Basic Research Program of China (2006CB403706), the National Science Foundation of China (40775009) and the Knowledge Innovation Program of the Chinese Academy of Sciences (IAP07115).