Climate change has great significance globally in general and South Asia in particular. Here we have used data from a network of 35 aerosol observatories over the Indian region to generate the first time regional synthesis using primary data and estimate the aerosol trends. On an average, aerosol optical depth (AOD) was found increasing at a rate of 2.3% (of its value in 1985) per year and more rapidly (~4%) during the last decade. If the trends continue so, AOD at several locations would nearly double and approach unity in the next few decades leading to an enhancement in aerosol-induced lower atmospheric warming by a factor of two. However, a regionally averaged scenario can be ascertained only in the coming years, when longer and denser data would become available. The regional and global climate implications of such trends in the forcing elements need to be better assessed using GCMs.
 With its diverse and contrasting geographical features, high population density, rapid economic development, and the associated urbanization and industrialization, coupled with the contrasting synoptic meteorological conditions associated with the Asian monsoon, South Asia is quite susceptible to the adverse impacts of climate change [e.g., Bollasina et al., 2011]. Decisive role of Asian aerosols on regional and global climate is well recognized by the scientific community [Lau et al., 2006]. Although the global aerosol abundance is dominantly natural, regionally the anthropogenic species dominate in areas of high population density, industrialization and urbanization, or regions of extensive biomass burning [Intergovernmental Panel on Climate Change (IPCC), 2007; Moorthy and Satheesh, 2011]. Amongst the climate forcing parameters of aerosols, the most important is the spectral aerosol optical depth (AOD), which is the vertical integral, through the entire height of the atmosphere, of the fraction of incident light as a function of wavelength, scattered and/or absorbed by aerosols. The large spatial and temporal heterogeneity of the aerosol properties makes the climate-impact assessment of aerosols a challenging task, despite the concerted efforts of the global scientific community over the past decades.
 Long-term changes (trends) in aerosols are gaining increased interest recently due to its importance to global climate change [Dey and DeGirolamo, 2011; Kaskoutis et al., 2012; Ramachandran et al., 2012; Xia, 2011]. Even though several satellites provide aerosol optical depth regularly, its use for quantifying regional and global trends has limitations arising from the large uncertainties involved in the satellite-retrieved AODs, especially over the landmass due to the heterogeneous surface reflectance, uncertainties in cloud contamination, and inaccuracies in aerosol models used in AOD retrieval algorithms. Studies by Zhang et al. , Remer et al. , Kahn et al. , Levy et al. , and several others have suggested that caution should be exercised while using satellite data to avoid mistakenly interpreting noises and biases in the products as legitimate signals in long-term trend analysis. Such studies are also overwhelmed by calibration issues where the calibration drifts can be mistakenly interpreted as trends. As such, data from ground-based sun photometers (preferably network) are the best choice for trend analysis.
 In this paper, we have used the multi-year measurements of spectral AOD (UV through near IR) from a network of 35 aerosol observatories over the Indian region to generate the first time regional synthesis using primary data, estimate the regional trends, and discuss the potential climate implications.
2 Background, Instrument, and Data
 Although Indian research to characterize regional aerosol properties dates back to 1950, systematic investigations of the physicochemical properties of aerosols, their temporal and spatial heterogeneities, spectral characteristics, size distribution, and modulation of their properties by regional mesoscale and synoptic meteorological processes commenced only in the early 1980s under the Indian Middle Atmosphere Programme (I-MAP) [Moorthy et al., 2009]. With this modest beginning, networking a few select locations was fortified with a regional focus under the Indian Space Research Organization's Geosphere Biosphere Program (ISRO-GBP) [Moorthy and Satheesh, 2011] since the 1990s. With the increased recognition of the role of aerosols in regional and global climate forcing, the activity was consolidated under the Aerosol Radiative Forcing over India (ARFI) project [Moorthy et al., 2009; Moorthy and Satheesh, 2011]. It comprised a regional network (ARFINET) of 35 aerosol observatories spread over India. It covers urban, remote, island, coastal, inland, semi-arid, arid, and remote mountain regions over the mainland and the adjoining oceanic regions, and established over a period of time in a phased manner (see Table S1 and Figure S1 in the Supporting Information for details).
 The spectral AODs were estimated over all these stations using the multi wavelength radiometers (MWRs) designed and developed by Space Physics Laboratory (SPL) [see Moorthy et al., 1997, 1999 for details]. The instrument utilizes the filter wheel technique to make continuous spectral extinction measurements of ground reaching, direct solar radiation at 10 discrete wavelengths centered at 380, 400, 450, 500, 600, 650, 750, 850, 935, and 1025 nm, selected using narrow band interference filters with full width half maximum (FWHM) bandwidth of 5 nm. The instrument has a field-of-view of 2°. The AODs estimated using the MWRs have been intercompared on several occasions with the estimates made using other radiometers (such as multi-filter rotating shadow band radiometer (MFRSR), microtops Sun photometer (calibrated) and EKO Sun photometer) resulting in very good agreement with correlation coefficients of 0.99, 0.88, and 0.92, respectively [Kompalli et al., 2010]. The retrieved AODs have an uncertainty ranging from 0.03 to 0.05 over the various stations
3 Trends in Aerosols
 The simplest and statistically significant means to quantify trends in any environmental variable is to consider the changes to be varying linearly with time. Accordingly we have followed the widely used trend analysis techniques as described in Tiao et al. , Weatherhead et al. , Reinsel et al. , and Newchurch et al.  (see the Supporting Information for more details). Figure 1a depicts the long-term trends in the AOD at the mid-visible wavelength of 500 nm, derived from the ARFINET measurements where different colors and symbols differentiate the stations. Included in Figure 1 are stations that have data for 5 years or more and consists of stations (like Trivandrum (TVM), an unindustrialized coastal site at the southern tip of Indian mainland, see Figure S1 in the Supporting Information) with the longest (since 1985) record of AOD over India. The gap in data from June 1991 to December 1993 is because of the significant perturbations to the AOD caused by the Mount Pinatubo Volcanic eruption [Moorthy et al., 1996]. Notwithstanding the clustering of points (due to increased number of stations with time, due to phased expansion of the network) a statistically significant increasing trend is seen, which appears to differ marginally between different stations. At some of the stations, a tendency of leveling off or a very weak decrease is also seen. While the rural areas (with lower AOD) showed stronger increasing trends, the urban sites nearly leveled off, especially in the recent years (see Table S2 in the Supporting Information for detailed information on the trends at different stations with 8 years or longer data and the related statistics). These differences could be the consequence of stringent pollution control measures implemented in urban areas compared to the rural sites and the dispersion of the aerosol burden from region of high loading to the cleaner environments by the atmospheric motions. Using this primary data, regional mean AOD is estimated and its temporal variation is shown in Figure 1b. Here each point represents the regional monthly mean AOD (averaged over the available measurements for each month and would have more regional representation in later years due to increased number of stations), and the vertical bars represent the standard deviations. On an average, an increasing trend is discernible (as shown by the red line) indicating an increase in the regional mean AOD at the rate of 2.3% (of its value of 0.28 at 500 nm in 1985) per year. This trend is statistically significant at significance level at p < 0.0001 level. Even the lowest AOD values have gone up, indicating an overall decrease in the atmospheric transparency to solar radiation. Trend analysis on a regional scale is presented in Table 1 where we have divided the data set into three regions: (a) southern Peninsular India, (b) central peninsular India, and (c) northern India. Central peninsular India depicted higher increasing trends in AOD. If we restrict only to the last decade, the annual increase nearly doubles, to ~4% per year (Figure 2). If the trends continue so, aerosol optical depth at several locations would double in next few decades; when the AOD at mid-visible wavelengths would approach unity leading to an ~70% reduction in ground-reaching direct solar radiation at mid-visible.
Table 1. Regional Variation of Annual Trend in Aerosol Optical Depth (AOD)
Mean AOD at Starting Year
Regression Coefficient (R)
Slope (ω, AOD yr−1)
Standard Deviation of ω (σω)
Trend (% yr−1)
 It is important to note that there are only a few stations with data spanning more than a decade and it would need several years of continued effort (at least for 5 years) for more stations to attain that status and some stations to have longer data. However, our analysis indicates that at some stations the increasing AOD trends are stronger in the recent years while at some other stations there is a tendency to leveling off. Examining this feature, it appears that the increasing trend is more conspicuous at less industrialized stations, where the base level was low or in other words aerosol loading has been low in the initial years. In urban-like locations where the base level was quite high, there is a leveling off. One reason could be due to near saturation of the human activities at the urban location and shifting of the focus of development to less urbanized location. This can be ascertained only in the coming years, when longer and denser data would become available.
 This rate of increase in aerosol loading at several locations in India in just about two decades, although a major concern, is not totally unexpected. Over the last five decades, India has experienced a large increase in its population (0.3 billion in 1950 to 1.04 billion in the year 2002), accompanied by rapid urbanization. The population of India is expected to exceed 1.6 billion by the year 2050 [State of Environment (SoE), 2009], the increase being higher in urban areas. Further analysis indicates that a major fraction of increased aerosol loading is contributed by anthropogenic aerosols as seen from Figure 3, in which we show the time variation of the Angstrom wavelength exponent (derived from the individual AOD spectra by performing a regression fit to the Angstrom relation) again as regional monthly averages. The data from June 1991 to December 1993 are removed because of the significant perturbations to the AOD caused by the Mount Pinatubo Volcanic eruption. The higher values of the exponent indicate increased abundance of fine, submicron particles in the aerosol size spectrum, which are produced mostly by anthropogenic activities such as industrial, urbanization, and fuel use (in transport sector), besides other combustion processes. Despite the scatter, the mean (least squares fitted) line shows an increasing trend of 0.02 per year. The large scatter of the points indicates the different aerosol types that co-exist over this region due to distributed sources and long-range transport. It is interesting to note that during the last decade, Angstrom coefficient does not show any statistically significant trends.
 In order to examine the consequent radiative impact of an increase in aerosol loading, we have estimated the shortwave clear-sky aerosol radiative forcing at the surface and top of the atmosphere (TOA) using the regional AOD, and Angstrom wavelength exponent from the ARFINET data and the single scattering albedo are incorporated in a Discrete Ordinate Radiative Transfer model (Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART) model) [Ricchiazzi et al., 1998] (see the Supporting Information for details). We used eight streams in the radiative transfer calculation and computations were made for solar zenith angles at every 1°. The regionally and annually averaged surface-reaching solar radiation is estimated to decrease by 57 W m−2 (which is about 20% of aerosol-free surface irradiance) for an optical depth of unity. In such a scenario, the aerosol-induced warming of the lower atmosphere is estimated as 1.1 K d−1. This may be compared with the corresponding values at present, which are 26 W m−2 and 0.53 K d−1, respectively. In this exercise, we have used aerosol single scattering albedo (ratio of scattering to extinction) values from OMI data onboard AURA satellite (2005–2008 average) [Torres et al., 2007]. The regional and global climate implications of such trends in the forcing elements need to be better assessed using GCMs.
 Our paper presents the first time estimate of regional trends in aerosols over India from the primary data derived from a network of observatories. The main findings are as follows:
On an average, aerosol optical depth depicts a near-linear increasing trend at a rate of 2.3% (of its value of 0.28 at 500 nm in 1985) per year and more rapidly (~4%) during the last decade.
This increase is due to a corresponding increase in anthropogenic fraction, as is evidenced by the trend in the AOD spectral index. However, it is interesting to note that AOD spectral index does not show statistically significant trends during the last decade.
If this trend continues unabated, the AOD at 500 nm in several locations would nearly double and approach unity in next few decades, leading to an enhancement in aerosol-induced lower atmospheric warming by a factor of two.
However, a regionally averaged scenario can be ascertained only in the coming years, when longer and denser data would become available.
The regional and global climate implications of such trends in the forcing elements need to be better assessed using GCMs.
 This work is carried out as part of ARFI project of ISRO-Geosphere Biosphere Programme (ISRO-GBP). We thank all the ARFINET investigators for the sustained support and data. One of the authors (SKS) thanks the Department of Science and Technology (DST), New Delhi, for Swarna Jayanti Fellowship award.