Corresponding author: Q. Min, Atmospheric Science Research Center, State University of New York, Albany NY 12203, USA. (firstname.lastname@example.org)
 The long-term multifilter rotating shadowband radiometer measurements at three Atmospheric Radiation Measurements Climate Research Facility sites of Darwin, Nauru, and Manus have been processed to develop the climatology of aerosols in the tropical warm pool region at the interannual, seasonal, and diurnal temporal scales. Due to their unique geolocations and associated large-scale circulation patterns, aerosols at the Nauru site exhibit background oceanic characteristics (strongly correlated with the sea surface wind), aerosols at the Darwin site show strong influences by biomass-burning aerosols, particularly in the dry season, and aerosols at the Manus site have climatologic characteristics in between the Darwin and Nauru sites. There are no obvious trends of aerosol loading for past decades at all three sites. El Niño/Southern Oscillation has its impacts on aerosol optical depth, as well as particle size and composition, at all three sites. Madden-Julian Oscillation modulates aerosol optical depth at the Manus and Nauru sites along the equator but has no apparent impact at the Darwin site. The annual or seasonal variation of aerosols is closely linked with Indo-Australian monsoons, exhibiting wet and dry season differences. The aerosol loading is significant lower with relatively larger particles in the wet season than in the dry season. There are significant diurnal cycles in both aerosol optical depth and Angstrom exponent at the Darwin site: low values of aerosol optical depth and Angstrom exponent in the midday and the two peaks in the early morning and late afternoon. There are noticeable changes between the dry and wet seasons. The amplitude of diurnal variation during La Niña periods is greater than that during El Niño periods. However, there are no significant diurnal variations of aerosol loading at the Manus and Nauru sites.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The tropical warm pool (TWP), the largest body of warm water on the planet, is a vital important source for the dynamic and thermodynamic forcing of the atmospheric general circulation. It also plays a central role in the El Niño/Southern Oscillation (ENSO) phenomena [Webster and Lukas, 1992]. Besides the ENSO phenomena, TWP is also an ideal region for the understanding of some other important climate drivers, such as Madden-Julian Oscillation (MJO), Indian Monsoon, Intertropical Convergence Zone, and South Pacific Convergence Zone. To solve these scientific issues and predict their impacts on global climate change, routine atmospheric and oceanic observations have been made for decades, including satellites and surface-based observation networks. Meanwhile, many global climate models have been developed to simulate and predict the atmospheric circulation and climate changes associated with TWP.
 Three Atmospheric Radiation Measurements Climate Research Facility (ACRF) sites, i.e., Manus, Nauru, and Darwin, shown in Figure 1, have been established in the TWP region. The Manus site, the first ACRF site, established on an island in the West Pacific Ocean (Los Negros Island in Manus, Papua New Guinea) in 1996, is located in the center of the TWP region. The Nauru site, established on a remote small island in the Pacific Ocean (Nauru Island) in 1998, is located in the east edge of TWP. The Darwin site, named after the closest Darwin city, established in 2002, is located in the coastal area of northern Australia and also in the south edge of the TWP region. The observations from these three ACRF sites cover the area roughly between 0°S and 12°S and from 130°E to 167°E, a region that plays a large role in the interannual variability observed in the global climate system. With sufficiently long term and high temporal resolution measurements at these sites, we can assess climate variability of many important atmospheric characteristics, such as aerosols, clouds, and precipitation, at various temporal scales. As the first paper of this series, we only focus our studies on the climatology of aerosol properties at these three ACRF TWP sites, not only including the general analysis of interannual variation, annual variation, and diurnal variation of aerosol properties, but also including the specific discussions about the effects of ENSO events and MJO events.
 Aerosols affect Earth's energy budget by scattering and absorbing radiation (the “direct effect”) and by modifying amounts and microphysical and radiative properties of clouds (the “indirect effects”). The climatology of aerosol properties is also affected by the global climate evolution; the climatology of observed aerosol properties can be used to evaluate and improve the climate models in the climate studies. With the establishment of the ACRF sites in TWP, the capability of accurate measurements of aerosol optical properties is substantially enhanced in this region particularly with long-term deployments of Cimel Sun-photometers (CSPHOTs) and multifilter rotating shadowband radiometers (MFRSRs). CSPHOT observation at the Nauru site, a remote small island and far from big islands and continents, has been studied as typical maritime background aerosols [Kambezidis and Kaskaoutis, 2008; Knobelspiesse et al., 2004]. In recent years, aerosol properties measured at the Darwin site, such as aerosol size distribution, aerosol types and sources, vertical aerosol distribution, the annual variation of aerosol optical depth (AOD) and Angstrom exponent, have been analyzed by many scientists [Allen et al., 2008; Qin and Mitchell, 2009; Bouya et al., 2010; Bouya and Box, 2011]. In this study, we mainly derive the aerosol optical properties, such as aerosol optical depth and Angstrom exponent, from spectral irradiance measurements of MFRSRs, which have been deployed at the ACRF Manus, Nauru, and Darwin sites since 1996, 1998, and 2002, respectively. We also use CSPHOT measurements from the Aerosol Robotic Network [Holben et al., 1998] to evaluate MFRSR aerosol retrievals.
2 Aerosol Observations
 The MFRSR uses an automated shadowbanding technique to make simultaneous measurements of the total-horizontal, diffuse-horizontal, and direct normal spectral irradiances at six passbands with 10 nm full width at half maximum centered near 415, 500, 610, 665, 862, and 940 nm [Harrison et al., 1994]. Because the separated spectral irradiance components of MFRSR measurements share the same passbands and calibration coefficients, the Langley regression of the direct-normal irradiance taken on clear stable days can be used to extrapolate the instrument's response to the top of the atmosphere, and then be applied to all components of irradiance [Harrison and Michalsky, 1994]. For the Manus, Nauru, and Darwin sites, MFRSRs have been operated continuously for more than a decade. There are 55, 96, and 154 Langley calibrations each year for the Manus, Nauru, and Darwin sites, respectively. By using the temporal and spectral analysis procedures of Forgan , the solar constants obtained from Langley regressions for the spectral bands are interpolated to any particular day. In general, the more Langley calibrations in the same time period, the higher the calibration accuracy. Among three sites, the accuracy of calibration at the Darwin site is the highest, with accuracy of 1% or 0.01 aerosol optical depth. To minimize the interference of gaseous absorption, the retrieval algorithm selects the 415 and 870 nm channels for aerosol retrievals. As outlined in Min et al. , aerosol optical depth and Angstrom coefficient can be retrieved accurately from simultaneous spectral measurements of direct beam irradiances. Furthermore, aerosol periods are separated from thin cloud periods, based on the temporal and spectral characteristics of direct and diffuse irradiances [Min et al., 2008]. An atmospheric stability test with a slide window of 30 min is carried out to further filter out potential thin cloud contaminations.
 To evaluate our retrievals, we compare MFRSR retrievals with the collocated CSPHOT products, which are processed by the standard Aerosol Robotic Network processing package [Dubovik and King, 2000; Dubovik et al., 2000, 2002]. For CSPHOT, although it does not have a 415 nm channel, the AOD at 415 nm can be interpolated from the AODs measured at 380 and 440 nm channels. As shown in Figure 2, the daily average aerosol optical depth and Angstrom exponent of MFRSR are consistent with the retrievals from CSPHOT at the Darwin site. The correlation coefficient for the AOD at 415 nm is 0.94, with a fitting slope of 0.98 and a bias of 0.01. Given the calibration accuracies of both CSPHOT and MFRSR, the bias of 0.01 is within the uncertainty of both instruments. The comparison of aerosol Angstrom exponent (derived from AOD at 415 and 870 nm) shows that the correlation coefficient, the fitting slope, and the bias are 0.85, 0.90, and 0.03, respectively. The difference in daily average values of Angstrom exponent between CSPHOT and MFRSR is mainly from the different cloud-screening algorithm, in addition to the issues associated with the calibration accuracy and instrument field of views. Due to relatively more cloudiness at both Manus and Nauru sites and lower aerosol loading, the comparisons at both sites have relatively lower correlation coefficients and fitting biases.
3 Interannual Variation of Aerosol Optical Properties
3.1 El Niño/Southern Oscillation Effects
 El Niño/Southern Oscillation, a quasiperiodic pattern, directly affects climate of TWP and many regions of the world through teleconnection. There are many indices to monitor the ENSO events, and the Southern Oscillation Index (SOI) is used in our paper. The SOI is derived from the monthly mean sea level atmospheric pressure difference between Tahiti and Darwin, indicating the variation of trade wind in this region. Figure 3a shows the interannual variation of SOI from 1997 through 2012: the red crosses indicate the El Niño periods (SOI < = −0.2), the blue crosses indicate the La Niña periods (SOI > = 0.2), and the green circles indicate the neutral time periods (−0.2 < SOI < 0.2). There are several significant El Niño periods (i.e., 1997–1998, 2002, 2004, and 2009–2010) and La Niña periods (i.e., 1999–2001, 2006, 2007–2008, and 2010–2011).
 The interannual variations of AOD and Angstrom exponent at the three ACRF TWP sites are also shown in the Figures 3b–3g. The color symbols used in Figures 3b–3g are consistent with the color symbols for the ENSO phases in Figure 3a. The smooth black lines indicate the 15 day smoothed daily mean values by a LOWESS smooth algorithm. There is no detectable long-term trend of aerosol properties for past decades at all three sites. Aerosol properties at the Darwin site vary with strong seasonal cycles, while aerosols at the two oceanic sites of Manus and Nauru have no significant seasonal cycles (more discussion later). At the Darwin site, shown in Figures 3b and 3c, annual maximum values of both AOD and Angstrom exponent are larger with a few exceptions in the El Niño periods (e.g., 2002, 2004, and 2010) than in the neutral and La Niña periods (e.g., 2003, 2008, July 2010 to April 2011).
 To further illustrate the impacts of ENSO, the monthly anomaly values of aerosol properties, derived by subtracting the long term mean monthly values from the specific monthly mean values, are plotted against the SOI index, shown in Figure 4. At the Darwin site, the monthly anomaly values of AOD and Angstrom exponent decrease with SOI. The correlation coefficients of −0.35 and −0.36 with slopes of −0.02 and −0.08, respectively, are statistically significant. It indicates that the effects of ENSO events on aerosol properties at the site are clearly distinguishable. During El Niño periods, most of Australia experiences less precipitation or a prolonged dry season. The prolonged droughts increase the frequency and intensity of the wildfires in Australia, producing more biomass-burning aerosols with fine size particles and a large Angstrom exponent. Because the local sources of other aerosols vary little in different years, with the enhanced contribution of biomass burning aerosols, both AOD and Angstrom exponent increase. In contrast, during La Niña periods, the north and northeast parts of Australia often experience more precipitation, restraining wildfires, and reducing biomass burning aerosols.
 The interannual variations of aerosol properties at the Manus and Nauru sites are not obvious in the time series as at the Darwin site. However, the impact of ENSO events is still detectable at the Manus and Nauru sites. At the Manus site, the monthly anomaly values of AOD and Angstrom exponent decrease with SOI, similar to but weaker than that at the Darwin site, as the correlation coefficients are 0.24 and 0.13 with the fitting slopes of −0.01 and −0.03, respectively. As stated previously, the Manus site is located on an island in the West Pacific and not far from the adjacent big islands and continent (such as New Guinea Island, Indonesia Islands, Australia). The contribution of transported aerosols from these big islands and continent is substantial and can be even dominant sometimes. During El Niño periods, weaker Indo-Australia Monsoons and drier weather in most of Indonesia, the Philippines, and eastern Australia result in more biomass-burning aerosols. The aerosol can be easily transported to the Manus site, resulting in larger AOD and Angstrom exponent. During La Niña periods, the adjacent big islands of Manus Island are wetter than normal years, both locally produced and transported aerosols are smaller. Compared to the Darwin site, the source of biomass burning aerosols at the Manus site is much farther away and smaller, thus the effects of ENSO events on aerosols at the Manus site are weaker.
 At the Nauru site, it is interesting that the monthly anomaly values of AOD are bifurcately correlated with SOI, depending on the phase of ENSO. As the Nauru is located on a remote small island and far from the big islands and continent, maritime aerosols are dominant throughout the year. Figure 5 shows that AOD is linearly correlated with the wind speed measured at the 10 meter observation tower. The fitting slope is about 0.01 AOD per 1 m/s wind speed and the intercept is about 0.04. The Angstrom exponents are small and weakly dependent on the wind speed. It suggests that aerosols at the Nauru site may comprise two distinct aerosol types: sea-salt from sea-spray (more dependent on wind speed) and secondary aerosol of sulphate and organic species (less dependent on wind speed) [O’Dowd et al., 1997]. As shown in Figure 5c, the wind speed anomaly is also bifurcated with the ENSO phases. The bifurcation is asymmetric: the wind speed anomaly per SOI is three times stronger in El Niño periods than in La Niña periods. It suggests that the wind speed anomaly depends on not only the sea surface temperature changes but also the circulation pattern changes associated with ENSO, as pointed out by Li et al. . Strong correlation between the monthly AOD anomaly (Figure 4c) and the wind speed anomaly (Figure 5c) in terms of SOI suggests that ENSO effects on maritime aerosols is mainly due to the changes of atmospheric circulation and wind speed through mechanical production of sea-salt aerosols.
3.2 Madden-Julian Oscillation Impacts
 The Madden-Julian Oscillation consists of large-scale coupled patterns in atmospheric circulation and deep convection, propagating eastward slowly (5 m s–1) through the portion of the Indian and Pacific oceans where the sea surface is warm. It is the dominant component of the intraseasonal variability in the tropical atmosphere with periods of 40–70 days. For the regions within the MJO path, the MJO events change the local meteorology and bring in additional precipitation. Large intraseasonal variations of aerosols related to the MJO over the whole Tropics were found by analyzing multiple satellite derived aerosol products [Tian et al., 2008, 2011]. More interestingly, rainfall anomalies were negatively correlated with the TOMS aerosol index but positively correlated with MODIS aerosol optical depth [Tian et al., 2008].
 To investigate the MJO impacts on aerosols at the sites, we analyze the power spectra of both precipitation and aerosol optical depth using fast Fourier transform. To minimize the strong ENSO signals, we choose the daily mean values of AOD and daily precipitation (rain rate) from 2003 through 2006, during which SOI is small and varies within a narrow range. To suppress the high frequency signal, a LOWESS smooth algorithm with the smooth window of 15 days is applied to the original observation data. It is worthy to notice that the missed value of daily mean AOD is interpolated by valid aerosol measurements on the closest days. As shown in Figure 6, there are significant signals with periods of 40–70 days in both rain rate and AOD spectra at the Manus and Nauru sites, and no apparent signals at the Darwin site. The meteorological changes associated with MJO modulate the wind speed and circulation pattern, and impact the precipitation. Over the sea, the increased wind speed can increase the sea salt production; over the land, the increased precipitation can restrain the intensity and the frequency of the wildfire events, which will influence the source of biomass burning aerosols. In addition, the wind speed and precipitation also have substantial impacts on the aerosol transport and wet and dry depositions. In general, the MJO event will decrease the aerosol loading when it passes over somewhere, especially in the biomass burning seasons. The observed signals of precipitation and AOD with 40–70 day periods are likely the consequence of MJO modulation. Unlike Manus and Nauru sites along the equator, the Darwin site is far away from the equator and outside the MJO path. Although MJO moves southward and closer to the Darwin site in the wet season, the combined effects of Indo-Australian Monsoon result in no obvious MJO signal at the Darwin site.
 As discussed above, MJO changes the circulation patterns and local meteorology, as well as clouds and precipitation. Those factors impact on aerosol processes of emission, growth, transport, and dissipation differently, resulting in a complex relationship between aerosols and precipitation at the scale of MJO. There is no significant correlation between aerosol optical depth and precipitation in our study at the two oceanic sites, but both have similar spectral powers with periods of 40–70 days. It is worthy to emphasize that our spectral and temporal cloud screening procedure in our aerosol retrievals limits potential cloud contamination, one of major issues for satellite retrievals [Tian et al., 2008].
4 Annual Variation of Aerosol Optical Properties
 Figure 7 shows the annual variations of aerosol properties at the Manus, Nauru, and Darwin sites. The aerosol optical depth at the Darwin site is the largest among the three sites with a distinct seasonality: decreasing from January to April, increasing after April and reaching the maximum about October, and then decreasing again. The Angstrom exponent is much larger than the other two sites and reaches its maximum in October. Aerosols at the remote and background Nauru site are the cleanest with the lowest Angstrom exponent. The seasonality of them is weak but detectable, which has substantial correlation with wind speed. Our results at the Nauru site are different from Kambezidis and Kaskaouti , in which no obvious seasonal cycle of AOD and surface wind speed was observed. At the Manus site, aerosol properties are in between the other two sites, with no significant seasonal cycle.
 In general, the meteorology in the region can be divided into three types of periods: the dry season (from May to September), the wet season (from December to March of next year), and the transitional season (from April to May and October to November). For simplicity and consistency among the three sites, we divide into two periods or seasons: the dry season (May–December) and the wet season (January–April), and use the scatter plots of AOD and Angstrom exponent, shown in Figure 8, to illustrate aerosol inherent properties.
 In the dry seasons, droughts cause large regions of wildfire on the Australian continent, especially in the northern and northeastern tropical savanna regions, and the continental wind from south or southeast bring the biomass burning aerosols into the region [Russell-Smith et al., 2007; Craig et al., 2002]. As stated by Qin and Mitchell , 93% of aerosols are aged smoke or fresh smoke at the Darwin site during the dry season, which suggests limited local pollution. The AOD-Angstrom exponent plots reveal the Angstrom exponent increases with AOD at the Darwin site. It suggests that the more biomass burning aerosols, the smaller particle size with higher Angstrom exponent. The ENSO effects slightly change the particle sizes of aerosols at the site, as found previously. There is a similarity of aerosol characteristics between Manus and Darwin sites, although aerosol loading is lower at the Manus site. At the Nauru site, most aerosols have the inherent characteristic of Angstrom exponent decreasing with AOD, a typical sea-salt characteristic. However, during the El Niño periods in the dry season, occasionally aerosols behavior like the biomass burning aerosols with Angstrom exponent increasing with AOD. It could be due to prevailing wind during the dry season that brings the biomass-burning aerosols into the Nauru site.
 In the wet seasons, the Indo-Australian monsoon dominates the meteorology in the region, bringing in maritime air mass and lots of precipitation into Australian continent and big islands. The wildfires in Australia (and big islands) are significantly restrained and biomass-burning aerosols are suppressed, while the contribution of maritime aerosol is enhanced and substantial. The aerosol loading is significantly lower in the wet season than in the dry season. The inherent AOD-Angstrom exponent relationship is more like maritime aerosols, shown in Figure 8. The ENSO has small impact on this inherent relationship.
5 Diurnal Cycle of Aerosol Optical Properties
 Because MFRSRs were continuously operating with a high time resolution of 20 s, it is possible to derive the diurnal cycle of aerosol properties. We first calculate the daily diurnal percentage change by normalizing with the daily mean value for each clear-sky day, and then average the daily diurnal percentage change over the study period to define the diurnal cycle, similar to Michalsky et al. . Figure 9 shows the diurnal cycles of AOD and Angstrom exponent at the three ACRF sites.
 At the Darwin site, there are significant diurnal cycles in both AOD and Angstrom exponent. The Darwin site is located close to Darwin city and adjacent to the Darwin international airport. Although there is no industrial pollution in the surrounding area, the impacts of local human activity, especially vehicular emission, cannot be ignored. Generally, the peak times of vehicular emission are in early morning and late afternoon, and the aerosols originated from vehicular emission are often small in size with large Angstrom exponent. Also the sea breeze in the middle of the day or early afternoon blows over fresh air and maritime aerosols from the ocean and influences the long-term aerosol transport from other regions. The combined effect results in the observed diurnal variation of aerosol properties: low values of AOD and Angstrom exponent in the midday and the two peaks in the early morning and late afternoon. Figure 10 shows the more details of the diurnal percentage variation of aerosol properties at the Darwin site, with further segregating by the wet and dry seasons and by the El Niño, La Niña, and neutral phases. There are noticeable changes between the dry and wet seasons. The prevailing wind from south and south-east during the dry season reduces the oceanic influence, resulting in less diurnal variation in Angstrom exponent. The amplitude of diurnal variation during La Niña periods is bigger than that during El periods.
 As discussed previously, aerosols at the Nauru site are dominated by maritime aerosols while at the Manus site comprise of mainly maritime aerosols and small portion of biomass-burning and anthropogenic aerosols. There are no significant diurnal variations of aerosol loading at both sites. The slight lower values of Angstrom exponent in the middle of the day are correlated with the diurnal variation of wind speed. Also, there is no significant difference between the dry and wet seasons and among the ENSO phases (not shown). The diurnal cycle of AOD at the Nauru site is consistent with the analysis of Smirnov et al. .
 Aerosols affect Earth's energy budget and climate directly by scattering and absorbing radiation and indirectly by modifying amounts and microphysical and radiative properties of clouds and precipitation. Due to the immense diversity of aerosols, not only in particle size, composition, and origin, but also in spatial and temporal distribution, it requires coordinated efforts at integrating data from multiple platforms and techniques to understand aerosol impacts on climate. Observed climatology of aerosols is also required to provide constraints and validation of the climate models. We used long-term MFRSR measurements at three ACRF sites of Darwin, Nauru, and Manus to develop the climatology of aerosols in tropical warm pool region at the interannual, seasonal, and diurnal temporal scales.
 Due to their unique geolocations and associated large-scale circulation patterns, aerosols at the Nauru site exhibit background oceanic characteristics (strongly correlated with the sea surface wind), aerosols at the Darwin site show strong influences by biomass-burning aerosols, particularly in the dry season, and aerosols at the Manus site have climatologic characteristics in between the Darwin and Nauru sites. There are no obvious trends of aerosol loading for past decades at all three sites. ENSO has its impacts on aerosol optical depth, as well as on particle size and composition, at all three sites, through the influence of precipitation and local circulation. MJO modulates aerosol optical depth at the Manus and Nauru sites along the equator but has no apparent impact at the Darwin site.
 The annual or seasonal variation of aerosols is closely linked with Indo-Australian monsoons, exhibiting wet and dry season differences. During the dry seasons, the prevailing continental wind from south or southeast bring the biomass burning aerosols into the region and overwhelming the local and maritime aerosols, resulting in a large amount of small size aerosols. The Indo-Australian monsoon brings maritime air mass/aerosols and precipitation, suppressing wildfire related biomass burning aerosols. The aerosol loading is significantly lower with relatively larger particles in the wet season than in the dry season.
 There are significant diurnal cycles in both AOD and Angstrom exponent at the Darwin site: low values of AOD and Angstrom exponent in the midday and the two peaks in the early morning and late afternoon. There are noticeable changes between the dry and wet seasons. The amplitude of diurnal variation during La Niña periods is bigger than that during El Niño periods. However, there are no significant diurnal variations of aerosol loading at the Manus and Nauru sites. The slightly lower values of Angstrom exponent in the middle of the day are correlated with the diurnal variation of wind speed. Also there is no significant difference between the dry and wet seasons and amongst the ENSO phases.
 A better understanding of aerosol-cloud interactions is one of the greatest challenges in studying climate changes, and it depends on the ability to improve representation of aerosols and clouds. With sufficiently long-term and high temporal resolution cloud retrievals of MFRSRs at these sites [Min and Harrison, 1996; Min et al., 2004], we will assess the climatology of clouds at various temporal scales in the same region in the near future, including the effects of ENSO events and MJO events on clouds in TWP.
 This work was supported by US DOE's Atmospheric System Research program (Office of Science, OBER) under contract DE-FG02-03ER63531, by the NSF under contract AGS-1138495, and by the NOAA Educational Partnership Program with Minority Serving Institutions (EPP/MSI) under cooperative agreements NA17AE1625 and NA17AE1623.