signature=74535180b485cb7510c60696b3da015a,Declining pre-monsoon dust loading over South Asia: Signa...

The atmospheric aerosol loading over the Indian region peaks during the pre-monsoon (March to May) and the monsoon (June to August) seasons depending on location of interest and the spatial distribution of rainfall

Changes in column aerosol optical depth

Ground based Measurements

Recent studies have shown that aerosol loading over the Indian region is increasing on an annual mean basis

Our analysis reveals that almost all the ground based sites (see Fig. 1), with sufficiently long time series data of at least a decade, show a decreasing trend in the column aerosol loading during pre-monsoon period. Also, the decrease is most pronounced over the sites located to the west of the Indo-Gangetic Plains with a clear East-West gradient. Lahore and Jaipur, to the west of the IGP/North India show highest decrease (0.025 and 0.014 per year respectively, which is ~3% per year with respect to year 2000) in the column loading. On the other hand, Kanpur, a site at center of the IGP shows comparatively lesser decrease (0.003 per year, ~0.5%). Moreover, the decrease in AOD is even prominent over the site Gandhi College, the eastern part of the IGP. This shows that the decrease is widespread during the pre-monsoon season over the whole greater IGP belt. However, the largest decreases were observed over Jaipur and Lahore. It may however be noted, that annually, each of these surface sites showed an increasing trend (mostly attributable to anthropogenic activities) in aerosol loading

Figure 1

(a) The trend (year−1) of seasonal mean aerosol optical depth over the AERONET sites in the IGP for pre-monsoon season. The bold text indicates significance at 90% confidence level. (b) to (e) shows the spatial pattern of trends (year−1) in aerosol optical depth observed using different sensor/satellite platforms (b) MODIS Terra & (c) MODIS Aqua combined Deep Blue Dark Target, (d) MISR and (e) OMI-UV Aerosol Index. The black dots (circle) represent statistical significance at 90% (95%) confidence level. The map was generated using MATLAB 2015b, www.mathworks.com.

Satellite based measurements

The trend analysis were carried out using satellite based measurements from different platforms (MODIS Terra, MODIS Aqua, MISR and OMI) and different retrieval methodologies (Dark Target (DT) approach1b to e). The black dots (circle) represent statistical significance at 90% (95%) confidence level. The East-West gradient in ground based measurements is found to be quite evident here. Nevertheless, it may be noted that the Eastern regions still shows a clear increasing trend in the aerosol loading. However, the largest decreasing trend is found over the Thar Desert and the arid regions to the west.

Consistent results obtained using both ground (five sites) and satellite (eight retrieval methods) based measurements provide us confidence that these changes are not due to retrieval errors (differing methodologies) or platform specific drifts (in each of the sensors) that may lead to erroneous trends

Figure 2a summarizes, changes observed using the ground based AERONET sites over the IGP along with similar analysis based on different satellite measurements (based on different retrievals and platforms). All show similar results, albeit slight changes in their magnitudes. A clear east-west gradient is also observed in the trends with larger decreases in AOD to the West. Though ground based measurements showed a decreasing trend over Gandhi College (the eastern most site), none of the MODIS retrievals showed a decreasing trend for Gandhi College, which is the East most site among the sites used in our study. However, both ground based AERONET and MISR (0.5 × 0.5 degree) showed a consistent decreasing trend. It is not clear, what could have caused this, but we speculate it to be a bias created by the use of coarser resolution (1 × 1 degree) MODIS that is representative of a larger region around the ground site. In addition, Fig. 2b shows the mean trend from all satellite sensors and ground stations (normalized with respect to 2003), which clearly exhibits the decreasing trend. It may be noted that even the 95th percentile of the data for all sites and satellites lie below the base year mean.

Figure 2

Inter annual trend (year−1) in (a) AOD from satellite and ground based data (2001 to 2015) for Stations over IGP, the dots represent statistical significance at the 90% confidence level (b) mean AOD (normalized with initial value) obtained from all the satellite data used in the present study over five sites, error bars represents the 95th percentile of inter-site and inter-sensor differences and (c) angstrom exponent (α440–870), bold font indicates statistical significance at the 90% confidence level. The map was generated using MATLAB 2015b, www.mathworks.com.

The possibility for the east-west gradient in trends could be the decrease in dust contribution away from dust sources (to the West), which renders anthropogenic aerosols to dominate the fractional contribution to the total loading and hence trends (smaller decreases) in the Eastern part of IGP. It may also be noted that different retrieval methodologies have different assumptions, complexities and differing levels of sophistication. It is therefore difficult to narrow down the exact reason for slight differences in trends between sensors and retrieval methodologies. Overall majority of the sensors and methodologies agree to the declining aerosol loading over the different sites with a clear east-west gradient.

It may also be mentioned that there are no indications showing that other types of aerosols such as from anthropogenic sources (such as Black Carbon) are decreasing. It may be noted that if BC were decreasing (increasing), it will strengthen (weaken) the trends. However in this study, we focus our attention on dust as its contribution to total AOD is several fold larger than other aerosols over this region. On a cursory note, the increasing economic activities and resultant anthropogenic emissions render the possibility of a decrease in BC a remote possibility. This is also evident from the trend in wavelength dependence of spectral AOD (i.e. in Angstrom Exponent (α440–870, AE) observed in the AERONET data showing a clear increase over the years during the pre-monsoon season (see Fig. 2c). Similar results were observed for spatial variability in AE (see Fig. S1). However, increasing AE could be due to either, increasing smaller particles (perhaps of anthropogenic origin) or decreasing larger particles (natural dust).

Having determined that the decrease in AOD is tracked by both surface and satellite based sensors and the spatial pattern indicating that the decrease is mostly to the west of IGP and over the Thar Desert. We therefore use a simple methodology to confirm whether the changes are due to dust. It may be noted that the observed spatial (west to east) gradient is only an indication and is not an evidence that dust is decreasing during the study period.

Observed change in dust from ground based measurements

In order to determine what type of aerosol have changed and to confirm our hypothesis on dust change, we used a decision tree based methodology developed by Lee et al., 2010S2). This observed seasonality is in agreement with previous studies e.g.

Based on the longest available data source, we have chosen Karachi, Kanpur and Lahore AERONET sites for the change analysis. Incidentally these sites are also located at two extreme sides of the IGP. Our analysis (Fig. 3a) reveals that there has been significant reduction in dust loading over the period of our study. The decreases are between 10 and 20% over all the sites. The percentage decreases are more pronounced over Kanpur, in the Eastern part, than over Lahore, which is closer to the dust source regions. This may be due to the difference in lower frequency of occurrence and/or dust loading over Kanpur (in comparison to cities close to desert source) thereby showing a higher % value of change over the East than over the West, where frequency of occurrence of dust events are higher. This analysis provides us the confidence/evidence that the decrease observed in column loading based on ground and satellite based measurements is due to changes in dust aerosols. We also used MERRA2 retrieved dust AOD to derive the trends during the study period (Fig. 3b). It is found that there exists a clear decreasing trend in dust AOD with a stronger gradient towards the west (Fig. 3b). This finding further provides us an additional indication that the observed change in AOD’s were due to dust and the maximum trends are over the dust source regions to the North-western part of India.

Figure 3

The changes in dust observed over the Northern part of the sub-continent (a) the percentage change in dust loading during pre-monsoon season (between the period 2011–2014 and 2006–2010) using an independent analysis to identify dust loading using AERONET retrieved aerosol size information and absorption. (b) MERRA2 reanalysis for the period 2002 to 2015 (year−1). The black dots (circle) represent statistical significance at 90% (95%) level. The map was generated using MATLAB 2015b, www.mathworks.com.

We carried out several analyses to confirm that aerosol type classification used in this study is robust. Our investigation (see Supplementary material) shows that errors (as per AERONET retrieval) associated with either absorption or extinction (is <5%) does not affect our finding with respect to aerosol type identification (Fig. 3a). MERRA2 dust AOD also shows similar results albeit slight differences in magnitude. Having confirmed that the change in column loading is due to change in dust using both ground based measurements and MERRA2 reanalysis dust AOD, we proceed to investigate the potential cause for the observed change.

Potential cause for decrease in dust loading

The aerosol loading over any location is a function of the strength of local source, sink and advection (long range transport). Therefore, changes to both rainfall and circulation patterns could have played a significant role in the observed changes to dust over the Indian region. Increased rainfall over the source region will alter dust concentration/loading in two ways, one by increasing wet removal and other by altering the emission strength due to changes in soil moisture, which in turn alters the erodibility. Therefore, even minor changes to rainfall over dust source regions could have large effect on column dust loading. We therefore used Tropical Rainfall Measuring Mission (TRMM), Global Precipitation Climatology Project (GPCP), University of Delaware gridded Precipitation (UDel) and India Meteorology Department (IMD) gridded rainfall datasets to analyse the precipitation changes over the study domain.

Our analysis (Fig. 4) reveals that the pre monsoon rainfall has increased during this period with maximum increase observed over the Pakistan region and also over the Thar Desert. The rainfall increase over Thar Desert is not significant for datasets using station based datasets. This could be a result of issues related to number of stations over the desert region going in to the production of these gridded data products such as IMD, UDel and GPCP. However, TRMM which directly obtains the spatial pattern (at each grid point) shows the increases in rainfall over the Thar Desert region. Overall, it is found that the rainfall is increasing over the arid and semi-arid regions to the NW part of the sub-continent.

Figure 4

Spatial pattern of rainfall trends (year-1) from different datasets (a) TRMM (b) GPCP (c) UDel and (d) IMD. The black dots (large dots) indicate 90% (95%) confidence level. The map was generated using MATLAB 2015b, www.mathworks.com.

In addition, any change in circulation that may alter the winds over the source region will also alter the strength of dust emission. We therefore calculated the trend in 10 m wind speed obtained from ECMWF–ERA reanalysis. There is a clear indication of a slowdown of winds (see Fig. 5a) especially in the vicinity of Thar Desert. This in turn may reduce local emissions of dust loading. Similar observations were also made over African region5b). Based on the above observations, we hypothesise that major factors that contributed to pre-monsoon decreasing trends in aerosols over Indian region are, (i) Increased rainfall in the western part of Indian subcontinent that removed aerosol loading through wet scavenging including the dust aerosols. (ii) Reduced emission of dust due to increased precipitation which decreased the erodibility and hence dust emission. (iii) The slowing winds further reduced the efficiency of long range transport thereby decreasing loading over regions away from the dust sources. Even the dust that were generated will not be transported longer distances if the winds are slowing.

Figure 5

The spatial pattern of trends (year−1) in (a) 10 m wind speed (ms−1) (ECMWF-ERA-Interim) (b) Dust wet deposition (kg m−2s−1) (MERRA-2). The black dots (circle) represent statistical significance at 90% (95%) confidence level. (c) Trend in Extinction Coefficient (km−1) observed from the CALIPSO data over Jaipur (close to the dust source regions) during the period 2006 to 2015. The plots were generated using MATLAB 2015b, www.mathworks.com

Therefore, our analysis reveals that there is a clear decrease in AOD observed during pre-monsoon period over the Northern part of the Indian region and specifically over the North western part of Indian subcontinent due to declining dust emissions. This is due to the increased rainfall, which inhibited dust emission and at the same time increased wet scavenging. It may be mentioned that the role of rainfall on aerosol loading will be more evident if high resolution vertical distribution of aerosols are available. We therefore used the layer mean extinction coefficient at 532 nm from CALIPSO profiles nearest to Jaipur station located close to the Desert region to assess the change. As expected, our analysis (Fig. 5c) reveals that the reduction in aerosol loading is most concentrated below 2.5 km altitude. This is a clear evidence for the role of wet scavenging/removal and/or reduced emissions and therefore lower loading conditions in these altitudes. In addition, the slowing down of winds over this region also further plays a role in reducing the generation and thereafter long-range transport of dust aerosols, which were eventually picked up by sites far from the dust sources. However, this appears to be of minor importance compared to the role played by rainfall. It may be mentioned that factors contributing to change in observed rainfall and winds over this region are beyond the scope of this work and are not explored further. We have thus restricted out investigation to only the changes to dust and their immediate meteorological causes for declining trends during pre-monsoon season. Summarizing, our analysis using long term ground, multiple satellite and model based reanalysis products reveal that dust aerosol loading over Northern India has decreased by as much as 10 to 20% since year 2000, depending on location in the past 15 years due to increasing pre-monsoon rainfall and slowing atmospheric circulation pattern. This decrease in dust loading may also provide an additional feedback leading to increased irradiance over the region thereby strengthening the Pakistan low that influences the monsoon flow. These competing effects along with anthropogenic aerosols and changing climate may modify monsoon rainfall in different ways and need to be addressed so as to understand the effect of aerosols, and specifically dust, on monsoon rainfall. It may be mentioned that the above analysis and hence the conclusions are based primarily on observations arising out of multiple ground and satellite based measurements. Therefore, it primarily shows that the changing regional climate is modifying dust dynamics over the south Asian region. Further research is necessary based on regional climate modelling initiatives to understand the feedback induced by changes to dust cycle on regional climate over South Asia.