Five-year Ground-based Observation Trend of PM 2.5 and PM 10 , and Comparison with MERRA-2 Data over India

A comprehensive analysis of pollutant’s trend and their measurement techniques are crucial for evaluating the air quality, and thereby helpful in formulating better control policies. In this report, we summarise ground based PM 2.5 and PM 10 data report in India (2015 - 2019). The important points discussed here are: (i) review of the ground-based data of PM 2.5 and PM 10 and the techniques used, (ii) mapping of the data over India with spatial and temporal distribution so that better understanding on PM pollution level can be made, (iii) identifying the technological gaps in measurement of PM concentration in India, and (iv) evaluation of MERRA-2’s (Modern-Era Retrospective Analysis for Research and Applications, Version 2) simulation of PM 2.5 against surface measurements in India to better understand biases for spatial and seasonal distribution, and then (v) suggestions for better PM measurement protocols, policies and metrological aspects for both measurement and control policies. It is observed that the amount of ground data on ambient monitoring of fine PM is insufficient and has several inconsistencies which require adequate attention. In India, not much work has been done on developing certified reference materials, traceable standards and calibration facility for particulate matter measurement which is a crucial step to ensure quality checks. Further, the comparison of MERRA-2 and ground PM 2.5 concentrations revealed huge discrepancies with underestimating PM 2.5 measurements in highly polluted regions like the Indo-Gangetic plain, especially during winter when pollution load was high. Better PM 2.5 agreement was found in summer and monsoon season, based on performance statistics explained in this paper. Inconsistencies between MERRA-2 and ground PM 2.5 are partly due to few limitations in MERRA-2 reanalysis method which are discussed in this paper, apart from several issues in ground-based observation. The aim of this review and comparison is to highlight such issues and give more attention to the importance of data quality assurance for effective air quality management. The present study may be helpful for the researchers in evaluating and choosing appropriate reanalysis products for their future studies.


INTRODUCTION
Atmospheric pollution is one of the major environmental concerns worldwide, especially in urban and sub-urban areas.Particulate Matter (PM) is known as one of Five-year Ground-based Observation Trend of PM 2.5 and PM 10 , and Comparison with MERRA-2 Data over India the significant contributors to poor air quality which is defined as solid, liquid and/or mixed particles suspended in air.They emit from both natural and anthropogenic sources.The particles having aerodynamic diameter ≤2.5 μm (PM 2.5 ) have adverse impact on human health since they can penetrate deep into the respiratory system causing cardiovascular and respiratory diseases.Thereby, PM 2.5 mass concentration was adopted as the primary parameter for ambient air particulate pollution and generally used as indicator of adverse health effects first in the United States in 1997, and later in different countries of the world.In India, CPCB (Central Pollution Control Board) notified National Ambient Air Quality Standards (NAAQS, revised 2009) stating standards of ambient air quality under the Air (prevention and control of pollution) Act 1981.As per the guidelines, cities are required to attain PM 2.5 concentrations below 40 μg/m 3 (annual average) and below 60 μg/m 3 (24-hour average).For PM 10 , mass concentration standards below 60 μg/m 3 (annual average) and below 100 μg/m 3 (24-hour average) are permissible.
Despite being a key contributor to air quality, not much progress has been done to ensure the data quality of PM ambient mass measurements in India.Lack of uniform and continuous PM monitoring is one of the important issues hindering effective air quality management.These are the possible reasons for which several Indian cities are scaled as most polluted cities in Asia and world in the listing by WHO (Air quality database, WHO, April, 2022).To ensure accurate and precise PM measurements, regular quality checks and calibration of instruments/methods deployed are of utmost importance which in turn helps policy makers to effectively combat and control emissions.Various central and state regulatory bodies along with research institutes are involved in monitoring of PM at various stations throughout the country.However, the present literature shows inconsistent data and limited air monitoring network.
So, considering these issues, this paper aims to review and map ground-based PM studies published in recent years (2015)(2016)(2017)(2018)(2019) across India and the techniques used for PM monitoring, highlight limitations and variations in measurements and to suggest better PM measurement strategy.Literature shows high loading of PM 2.5 in most of the Indian region, especially in the Ingo-Gangetic Plain (IGP), almost all year round with values exceeding national standards several folds.The IGP is regarded as the hotspot for aerosol source since it is densely populat-ed, and meteorological as well as topographical conditions which favour the accumulation loading of PM with lower planetary boundary layer (PBL) height, especially in the winter season (Jain et al., 2021).Likewise, trends in PM mass concentrations in different regions of India, their sources along with measurement techniques used are discussed in this paper.Furthermore, it is observed that although ground air quality monitoring stations exist in various regions in India, but due to their limited and non-uniform distribution with almost no station in rural areas, the air quality data is not sufficient to analyse spatial and temporal distribution of PM.Henceforth, to study better spatiotemporal PM 2.5 variation in India with the available data, satellite data is used to compare with ground-based PM monitoring and thereby calculate differences in PM 2.5 mass concentration on spatial and seasonal basis.
This review is organised into six sections.Section 2 reviews published studies on ground-based mass concentration of PM 2.5 and PM 10 in five years (2015)(2016)(2017)(2018)(2019) in India along with measurement techniques.Section 3 introduces satellite PM 2.5 data and reanalysis method.In Section 4 results and discussions are given, comparing ground PM 2.5 mass concentration with that of satellite assimilated data on seasonal and spatial basis using statistical parameters (correlation coefficient, mean bias, standard deviation of bias, mean fraction, FAC2) to study spatio-temporal PM distribution over India.In addition, factors resulting in underestimation of MERRA-2 PM 2.5 are discussed.In Section 5, the limitations in ground air quality monitoring network are discussed.Conclusion and suggestions for better and accurate PM measurement techniques are given in Section 6.

GROUND-BASED MEASUREMENT TRENDS 1 Ambient PM 2.5 Mass Concentrations in Various Cities during the Years 2015-2019
We have reviewed available literature on PM 2.5 mass concentration ground-based measurements in Indian cities during 2015-2019.Fig. 1 displays the average PM 2.5 concentration distribution in different cities in India.The concentration values for each region along with the monitoring techniques are given in Table 1.The Indian Himalaya Range (IHR) is a great mountain terrain which encompasses an area of about 2,500 km 2 and forms a bar-   Table 1.
While, low levels of mean PM 2.5 were reported in Kashmir and Darjeeling, i.e. 20±13 μg/m 3 and 24±13.5 μg/ m 3 , respectively during January-March 2015 (Sen et al., 2017).Similarly, in Kullu the average PM 2.5 concentration levels have been reported as 31±17 μg/m 3 (during January-March 2015) (Sen et al., 2017).As per the available literature, it reveals that the average PM 2.5 concentration in IHR (32 μg/m 3 ) is around 4.5 times lower as compared to IGP urban sites (143 μg/m 3 ) such as Delhi, Agra and Varanasi.Low particulate concentration levels in the IHR may be attributed to the pristine conditions of Himalayas and its sparse population resulting in fewer anthropogenic emissions.
The Indo-Gangetic Plain (IGP) is a vast fertile flat land covering an area of about 700,000 km 2 , situated in the foothills of Himalayas, flanked by Bay of Bengal on its east, the Thar desert on its west and semi-arid Deccan plateau to its south (Sen et al., 2016;Beig and Ali, 2006).Based on the literature surveyed, we observed that in the IGP, the average PM 2.5 concentration levels are higher than other regions throughout the study period.
In Delhi, the mean PM 2.5 mass concentration ranged from 91.5 μg/m 3 (May-June 2017) (Devi et al., 2020) to 135±64 μg/m 3 ( January 2015-December 2016) ( Jain et al., 2021) showing seasonal trend with high concentration in winter sampling months as compared to summer.The major reasons for high concentration are emissions from biomass burning, fossil fuel combustion and vehicular activities (Jain et al., 2021).Wood/waste/biomass combustion is a significant contributor to PM 2.5 loading during winter while crustal material was found to be prominent source in summer (Pant et al., 2015).Further, meteorological conditions like calm weather conditions and low inversion layer in winter favours higher loading of PM.
Patiala, located in the upper IGP, recorded an increase in PM 2.5 concentration, i.e. from 55±13.5 μg/m 3 in May-June 2014 (Sen et al., 2014) to 66.6±41 μg/m 3 in May-June 2015 (Sen et al., 2017), due to increase in open burning activities during 2015 compared to 2014.Relatively high PM concentration was observed during winter sampling months in Patiala i.e. 93±35 μg/m 3 (January-March 2015) (Sen et al., 2017).Due to fossil fuel emissions from industries located in the vicinity of Patiala, post-harvest stubble burning during October-November and wood/biomass burning in winters are the important sources of PM 2.5 in Patiala (Rastogi et al., 2016).
As for Varanasi, an increasing trend in the mean concentration of PM 2.5 can be observed from 99±33 μg/m 3 ( January 2015-December 2016) ( Jain et al., 2021) to 134±48 μg/m 3 (November 2016-February 2017) (Pratap et al., 2020) followed by a slight decline in mass concentration i.e. 124±50 μg/m 3 (October 2016-May 2018) (Kumar et al., 2020).Varanasi located in the middle IGP region observed very high pollution load due to combined effect of accumulation of particulates from upper IGP and emissions from local sources (Sen et al., 2016).Kolkata, being the third most densely inhabited city in India (Census, 2011), recorded an increasing trend in PM 2.5 concentration from 48±9.3 μg/m 3 (May-June 2014) (Sen et al., 2016) to 93 μg/m 3 (May-June 2017) (Devi et al., 2020).High population, vehicular and industrial emissions are predominant sources of PM 2.5 in addition to the emissions from biomass burning and anthropogenic activities predominately from upper and middle IGP cities ( Jain et al., 2020).Jamshedpur, an urban site in east India, also recorded a high PM 2.5 annual average (119±43 μg/m 3 ) from December 2017 to November 2018 (Ambade et al., 2021).Potential sources contributing to high PM concentration are the anthropogenic activities such as coal burning, fossil fuel burning wood burning etc.
Among the north-eastern Indian sites, Jorhat reported substantially high loading of PM 2.5 showing seasonal variability with highest average mass concentration observed during post-monsoon season as 159±77.6μg/ m 3 (October-November 2018) (Islam et al., 2020a), followed by pre-monsoon, i.e. 127±80.4μg/m 3 (February-March 2018) and monsoon season i.e, 121±49 μg/m 3 (July-August 2020) (Islam and Saikia, 2020b).Poor air quality in this region is predominantly due to emissions from small and medium scale industries involved in tea processing, oil and brick kilns and coal-based (Islam et al., 2020a;Islam and Saikia, 2020b;Saika et al., 2018).During wintertime in northeast India, due to festive biomass burning episode, traditionally known as meji burning, copious amounts of particulate matter is released into the atmosphere, worsening air quality in this region (Deka and Hoque, 2014).Some other important sources of PM 2.5 include coal combustion, crustal emissions induced by traffic, wood burning and vehicular emissions (Khare and Baruah, 2010).
Over western India, Mumbai recorded high mass concentration of PM 2.5 .A study by Masih et al. (2019) estimated the average PM 2.5 concentration in Mumbai to be 98.45 μg/m 3 during September-December 2016 followed by a drop in mass concentration to 60 μg/m 3 observed during January-December 2017 (Anand et al., 2019).The direction of wind flow majorly influences the air quality of Mumbai, being a coastal city.When clean coastal winds flow towards the city, it causes dispersion of pollutants thus improving the air quality, however the reversal of winds along with high humid conditions favour accumulation of pollutants in the region (Anand et al., 2019).
Pune, on the other hand, observed quite low PM 2.5 concentration as it is located at a relatively high altitude and experiences moderate weather conditions in both summer and winter seasons.Annual PM 2.5 mass concentration in Pune fluctuated from 46±23 μg/m 3 (January-December 2015) (Krishna and Beig, 2018) to drop in concentration to 37.3 μg/m 3 (during April 2015-April 2016) (Gawhane et al., 2017), followed by slight increased concentration to 40±14 ( January-December 2016) (Gawhane et al., 2019) which finally peaked at 47 μg/m 3 ( January-December 2017) (Anand et al., 2019).Pune observed a boom in housing industry since past few years along with the increased vehicular emissions and industrial setups in Pune (Pipal and Satsangi, 2015) which are probable sources of PM.High altitude sites in western India like Mahabaleshwar and Sinhagad recorded quite low PM 2.5 concentrations well within the NAAQS, i.e. 26±14 μg/m 3 (March 2015-February 2017) (Buchende et al., 2019) and 27±8 μg/m 3 ( January-December 2017) (Budhavant et al., 2020), respectively.
As for Varanasi, an uptrend in annual average mass concentration is observed from 238±82 ( January-March 2015) (Sen et al., 2017) to 257±90 μg/m 3 ( January 2015-December 2016) ( Jain et al., 2021) followed by a decrease in concentration to 165 μg/m 3 observed during summer sampling months, i.e.May-June 2017 (Devi et al., 2020).Major sources of high PM 10 levels in Varanasi region are emissions from traffic, construction and development activities, resuspension of road dust and longrange transport of PM from upper IGP (Jain et al., 2021).
In Agra, a uniform trend of PM 10 mass concentration can be observed which is nearly three times higher than the annual average standards of PM 10 (60 μg/m 3 ) (NAAQS, 2009).Predominant sources of PM 10 include emissions from factories and vehicles, construction and development activities, road and soil dust.
In Kolkata, Jain et al. (2021) calculated average PM 10 mass concentration to be 179±77 μg/m 3 ( January 2015-December 2016) and Devi et al. (2020) reported mass concentration of about 151 μg/m 3 (May-June 2017).Kolkata being a coastal city, experiences high loading of coarse mode marine aerosols in addition to local anthropogenic emissions, resulting in elevated PM 10 mass concentration levels especially in the winter sampling months (Sen et al., 2014).
In western India, the annual average PM 10 mass concentration for Pune slightly decreased from 98±45.5 μg/ m 3 (January-December 2015) (Krishna and Beig, 2018) to 63±25 μg/m 3 (January-December 2016) (Gawhane et al., 2019) and then increased to 87 μg/m 3 ( January-December 2017) (Anand et al., 2019).Anthropogenic activities, sea-salt and dust aerosols are the major sources contributing to PM 10 concentration (Gawhane et al., 2019).Ajmer, a semi-arid site, observed low PM 10 concentration (80±44 μg/m 3 during January-March 2015, Sen et al., 2017) as it is surrounded by Aravalli range resulting in lower transport of windblown dust aerosols from Thar desert in its vicinity.Ahmedabad, which is large city in western India, recorded average PM 10 concentration as 126 μg/m 3 during January-December 2017 (Anand et al., 2019).Even in summer, the PM concentration remained equally high i.e., 130±31 μg/m 3 (during May-June 2017) due to sea salt and mineral aerosols, frequent dust storms and heat wave conditions due to proximity to Thar Desert and Arabian Sea (Patel et al., 2021).
Over north-eastern India, sampling of PM 10 size fraction was carried out in Jorhat (Islam et al., 2020a;Islam and Saikia, 2020b) for which the mean concentration values were 261±89 μg/m 3 , 274±79 μg/m 3 , and 153± 45 μg/m 3 during February-March 2018, October-November 2018 and July-August 2019, respectively exhibiting clear seasonal trend.High particulate concentrations at these sites could be due to the influx of air masses loaded with particulate matter from within IGP region into eastern region (BoB) due to favourable wind pattern during winter (Sen et al., 2017).High PM level in Jorhat during winter/post monsoon is due to large number of forest fires in its near vicinity as reported by Sen et al. (2016) on the basis of open fire-count satellite data, in addition to festive biomass burning episode (meji burning) during this time (Deka and Hoque, 2014).
In southern India, at an urban coastal industrial sampling site Vishakhapatnam, PM 10 average mass concentration was observed to be 110±48 μg/m 3 (January 2014-December 2019) (Varaprasad et al., 2021).Important sources of PM pollution at this site are emissions from industries, vehicles and port activities.Tirumala which is high-altitude rural region also famous as a pilgrim site, reported moderate PM levels i.e., 60±25 μg/m 3 (January 2014-December 2019) (Varaprasad et al., 2021) whereby, emissions from transport sector is the major source of pollution.During August 2017-December 2019, Varaprasad et al. (2021) reported PM 10 levels in Amravati (residential site, capital of Andhra Pradesh) and Rajahmundry (industrial site) to be 79±45 μg/m 3 and 75±42 μg/ m 3 , respectively.In Amravati, the major sources of PM pollution are emissions from transport sector and construction activities, while in Rajahmundry, emissions from paper industries and powerplants located in its vicinity are the dominant sources.

1 Data Analysis
NASA's (National Aeronautics and Space Administration) MERRA-2 dataset (Modern-Era Retrospective Analysis for Research and Applications, Version 2) is a global atmospheric reanalysis tool developed by GMAO (Global Modelling and Assimilation Office) which is an advanced version of MERRA-1, Goddard Earth Observing System (GEOS) model.Atmospheric reanalysis integrating satellite observations is a helpful tool to provide the most complete picture of global climate change and spatial environment due to continual temporal and spatial resolution, high data quality and wide aerosol species (Randles et al., 2017).MERRA-1 was available from 1979 to February 2016, and after its discontinuation the subsequent dataset is present with MERRA-2 (runs a few weeks prior real-time).Basic aspects of both the reanalyses such as resolution and data products are the same as given in Gelaro et al. (2017).MERRA-2 dataset can be accessed online through NASA Goddard Earth Sciences Data Information Services Center (GES DISC).The optical depth observations assimilated at 550 nm in MERRA-2, are obtained by reflectance from MODIS (Moderate Resolution Imaging Spectroradiometer) on board Aqua and Terra satellites (Gelaro et al., 2017).The GOCART (Goddard Chemistry Aerosol Radiation and Transport) model coupled with GEOS provides information about the life cycle of predominant aerosol species, incorporating sea-salt (SS), dust (DS), organic carbon (OC), black carbon (BC) and sulfate (SO 4 ).At 550 nm, AOD is obtained by summation product of extinction coefficient of each species derived from cloud datasets, optical properties and mass concentrations of aerosols in the form of a column-species-integrated optical quantity.
Aerosol assimilation in MERRA-2 is done at a resolution of 0.5° latitude by 0.625° longitude and 73 vertical levels (from the surface to 80 km) all over the globe (Randles et al., 2017;Chin et al., 2002), thereby assessing the global distribution of PM 2.5 at high spatio-temporal resolutions.It is a very important tool to study particulate concentration and composition, especially in regions where ground monitoring is scarce or unreliable, or areas which are inaccessible.Having said that, there are also certain limitations in estimation methodologies used for simulation of MERRA-2 PM 2.5 mass concentrations, which must be carefully studied in order to better understand their merits and weaknesses.The purpose of this study is to assess MERRA-2's simulation of fine PM by comparison with ground-based measurements conducted at 27 stations across the Indian region between 2015 and 2019.The evaluation is based on seasonal and spatial variation, as the air quality of a region is predominantly influenced by life-style, meteorological and geographical factors.
To obtain ground-based PM 2.5 mass concentrations, a literature survey of available previous studies during 2015-2019 in India was done as shown in Table 1.Since PM 2.5 sampling durations varied in different studies, only those ground stations/cities are selected for which at least one month of continuous data is available, for better comparison with satellite data.It is also important to note that the MERRA-2 PM 2.5 mass concentration values at specific latitude and longitude coordinates were selected, corresponding to the coordinates of ground monitoring stations, in order to get spatially accurate data.Further, it maybe noted that there are many published studies which have used the data obtained from monitoring stations managed by the government, (for example pollution board monitoring stations) and then compared with satellite data (Navinya et al., 2020;Mahesh et al., 2019).However, to the best of our knowledge no study has used ground PM 2.5 mass concentration other than that of government managed stations for comparison with MERRA-2.

2 Evaluation method: Reconstruction of PM 2.5 Mass Concentration
Monthly MERRA-2 simulation data of five PM 2.5 species: SO 4 , OC, BC, DS 2.5 and SS 2.5 has been evaluated from 2015 to 2019 to estimate the total concentration of PM 2.5 .Using these individual aerosol species concentrations, the total mass concentration of PM 2.5 can be evaluated applying mass reconstruction method as shown in the equation below (Chow et al., 2015): where, DU 2.5 , BC, SS 2.5 , OC and SO 4 stand for dust (size≤2.5 μm), black carbon, sea salt (size≤2.5 μm), organic carbon and sulfate concentration, respectively.The factor of 1.375 is used to estimate absolute sulfate concentration as it is present in the form of neutralised (NH 4 ) 2 SO 4 in MERRA-2 data (Buchard et al., 2016).
Similarly, a factor of 1.6 is used to calculate particulate organic carbon to account for organic matter contributions of Indian region (Aggarwal and Kawamura, 2009).
It is important to note that eq. 1 lacks the concentration of nitrate particles predominantly anthropogenic in nature (Delmas et al., 1997), which may result to biases when compared with ground data (He et al., 2019;Mah-sh et al., 2019).
In this study, MERRA-2's simulated PM 2.5 concentration calculated by the equation above, is compared to the ground-based PM 2.5 concentration measured at 27 locations on seasonal as well as regional basis for a time period of January 2015-December 2019.In order to ensure the data reliability of the simulated and observed (ground-based) PM 2.5 data, a spatial consistency algorithm is used which is as follows: since trace concentrations are usually lognormally distributed, the bias between log-simulated concentration and log observed concentration (B log = log (C S ) -log (C O ); where C S : simulated concentration, C O : observed concentration) is calculated at all study locations.Thereby, reliability interval which justifies 95% of the normal distribution is calculated as: Blog±2σBlog; excluding all data pairs outside this interval.
Various statistical parameters were applied on monthly mean PM 2.5 concentrations for quantifying MERRA-2's accuracy.These are correlation coefficient (R), mean bias (B = C S -C O ), the standard deviation of bias (SD-B) mean fraction (F = C S / C O ), where C O and C S are observed and simulated mass concentration, respectively.Also, log transformed statistics were computed: B log , and R log .To evaluate air quality models, Chang and Hanna (2004) proposed Factor of Two (FAC2) index proposing that the proportion of data which satisfies 0.5≤C O /C S ≤2.0, should be equal to or more than 0.5 to be considered a good model performance.A detailed statistical comparison of ground and MERRA-2 PM 2.5 concentrations is given in section 4.1.

1 Seasonal and Spatial Comparison of PM 2.5 Mass Concentration between MERRA-2 and Ground-based Data
The aerosol distribution at surface and in the atmospheric column is heterogeneous and shows strong seasonal trend owing to meteorological conditions and seasonally varying emission source strengths over India (Sreekanth et al., 2007).Therefore, to evaluate the level of agreement between ground and MERRA-2 data, statistical metrics are applied on seasonal mean PM 2.5 concentrations shown in Table 2 (Navinya et al., 2020;Mahesh et al., 2019).We observed that the correlation coefficients are low during monsoon (0.22), indicating that PM 2.5 diurnal variability is not well captured by MERRA-2 in this season possibly due to restricted assimilation of MERRA-2 in presence of cloud cover (He et al., 2019;Buchard et al., 2016), while it is moderate during post monsoon (0.37), summer (0.38), and better during winter (0.59).However, Willmott (1982) discouraged the use of R to evaluate model performance since it is sensitive to extreme values and does not directly compare simulated with observed data.Therefore, to directly relate simulated and ground-based PM 2.5 concentrations, FAC2 index is evaluated given that this index is not disproportionately sensitive to extreme values.Based on FAC2 values, MERRA-2 and ground data performed well in all seasons (FAC2≥0.5)i.e., 1.60, 1.52, 1.40 and 1.84 for winter, summer, monsoon and post-monsoon, respectively, suggesting good model performance and comparability between the two datasets.The mean bias was less during monsoon (-10.75 μg/m 3 ) and summer (-24.33 μg/m 3 ) as compared to post-monsoon (-35 μg/m 3 ) and winter (-32.76 μg/m 3 ).Also, the average observed PM 2.5 concentration reported at ground locations was higher in post-monsoon (103 μg/m 3 ) and winter (86 μg/m 3 ) as compared to summer (73 μg/m 3 ) and monsoon (45 μg/ m 3 ).These results suggest MERRA-2's inability to simulate PM 2.5 mass concentration during high PM mass loading episodes.This is further supported by an underestimation of 34% and 38% during post-monsoon and winter respectively.Similar study by Navinya et al. (2020) reported the average observed PM 2.5 concentration over 20 Indian cities from 2015-2018 to be 80 μg/m 3 , while 35 μg/m 3 simulated by MERRA-2, indicating MERRA PM 2.5 underestimation by 34%.The explanation for disagreement between the two datasets is to be discussed in section 4.2.
Fig. 2 shows the spatial pattern of MERRA-2 simulated  PM 2.5 mass concentration throughout India for different seasons.The ground PM 2.5 mass concentration measured at 27 different sampling sites mentioned in this study are also depicted.A clear seasonal trend of PM 2.5 mass concentration is observed.Some key notable features are as follows: the air masses over IGP region are observed to have high dominance of PM 2.5 especially during the post-monsoon season (Fig. 2d) predominantly due to stubble burning over the western IGP which increases along the north eastern region during winter season (Fig. 2a), due to transport of aerosols from IGP (Sarkar et al., 2018).Also, throughout the study period, relatively higher PM mass concentration is observed over the IGP region compared to other parts of the country indicated by both MERRA-2 as well as ground data.Rapid industrialisation, urbanisation and elevated prominent PM sources such as biomass burning, waste burning, vehicular emissions are some of the reasons due to which IGP has been regarded as one of the most polluted regions in India ( Jain et al., 2021;Ojha et al., 2020).In contrast, less PM mass concentration is observed in IHR and south India due to pristine conditions and vegetation cover respectively (as explained in previous sections), which is indicated by both MERRA-2 and ground PM 2.5 concentrations.Therefore, we can say that PM 2.5 spatial trends observed from MERRA-2 data in different seasons reasonably conform to that of ground-based measurements.
However, it may be noted that MERRA-2 does not completely capture the size-resolved dust concentration in the IGP (Fig. 2b), since IGP is characterised by high dust loading during summer season (Goel et al., 2015).The underestimation of fine dust particle concentration has been reported in previous study by Kramer et al. (2018).Over central and eastern regions, coal-based power and steel plants in addition to small and medium scale industries as well as mining activities significantly contribute to PM mass loading (Guttikunda et al., 2019).It is important to highlight that the sampling site (located near the industrial site) at Chhattisgarh showed huge deviations (~3 times) between MERRA-2 and ground PM 2.5 concentrations throughout the study duration.This deviation could possibly be due to the inability of MERRA-2 to simulate nitrate concentrations from industries located in the vicinity of sampling location.Similar deviations between MERRA-2 and ground PM 2.5 (~4 times) was observed in Jorhat, where industrial emissions are the major factors for PM 2.5 emissions.
In terms of spatial distribution, when comparing the observed concentration (from ground data) and simulated concentration (from MERRA-2) an apparent deviation is evident in mass concentration of PM 2.5 .We observed high mean bias between ground and simulated data over IGP (-28 μg/m 3 ) as compared to IHR (-7 μg/ m 3 ) and southern region (-3 μg/m 3 ), and the corresponding underestimation by MERRA-2 to be 29%, 15% and 8%, respectively.Hence, it can be clearly observed that MERRA-2 underestimates mass concentration at regions showing very high pollution levels.the equiline, indicating both under and overestimations by MERRA-2.The scatter becomes wider for higher ground-based concentrations.The plot reveals that for significantly high (low) loading conditions, MERRA-2 underestimates (overestimates) the mass concentrations.Similar analysis is supported by log-transformed data.To better understand agreement between ground-based and MERRA-2 simulated PM 2.5 concentrations, we compared monthly averaged observed and simulated data from 2015 to 2019 (Fig. 4).Similar trends in PM concentrations by both ground and MERRA-2 data is observed in all months.Moreover, the bias values indicate better model performance in summer and monsoon months (i.e., March to April) as compared to winter and postmonsoon (October to February), possibly attributing to the absence of nitrate in MERRA-2 PM 2.5 reanalysis which is more significant in India given that PM 2.5 is mostly composed of anthropogenic particles (He et al., 2019;Mahesh et al., 2019).Another reason for high deviations in winter and post-monsoon could be due to MERRA-2's inability to unresolve local sources of pollution since its resolution is too coarse to evaluate the urban core of cities (Provencal et al., 2016).

2 Factors Resulting in PM
In addition, MERRA-2 does not completely capture the size resolved dust concentration resulting in underes-timation of fine dust particle concentration of size less than 2 μm (Kramer et al., 2018).Further, there are missing AOD values in MERRA-2 reanalysis dataset due to cloud cover, orbital gaps, etc. (Mahesh et al., 2019;Buchard et al., 2016).Therefore, these datasets require extensive regional scale validations for complete data interpretation, especially over the highly diverse Indian region.

VARIATIONS AND LIMITATIONS IN MEASUREMENT OF GROUND-BASED PM MASS CONCENTRATION
Since PM monitoring plays an important role in improving air quality and health, therefore the accurate and precise measurement of PM mass concentration is of utmost concern.To monitor PM concentration across India, Central Pollution Control Board (CPCB) has notified three methods namely, Gravimetric Method (GMM), Tapered Element Oscillating Microbalance (TEOM) and Beta Attenuation Monitor (BAM).Gravimetric Method is the primary method to quantify PM mass concentration which is based on the difference in mass of the filter used before and after sampling, while BAM and TEOM are the online/real-time measurement methods where the mass of PM collected on filter substrate is determined using beta attenuation, and change in frequency of oscillating element in TEOM, respectively.From the available ground-based data given in Table 1, we conclude that of all methods, Gravimetric Method is widely used in India since it is comparatively simpler, cost-effective and involves the primary measurement of PM mass.
In addition, some studies show inconsistency in mass concentration measurement even for the similar measurement technique used at same location for similar sampling duration.For example, Sen et al. (2014) reported PM 2.5 concentration in Varanasi to be 188±36.5μg/m 3 using IPM-FDS, Instrumex (flow rate of 1 m 3 h -1 , designed as per USEPA guidelines) during 20 th January-5 th February 2014, whereas another study by Kumar et al. (2015), reported PM 2.5 mass concentration as 138± 47.12 μg/m 3 in Varanasi during 1 st January-31 st March 2014 using the same instrument.Therefore, about 30% variation in PM mass concentration is observed despite similar conditions of measurement.A part of deviation in results could be attributed to the different local (site-specific) sources and measurement period (which was not exactly the same), but the inconsistency is primarily due to lack of standardisation of measurement method (Malik and Aggarwal, 2021).The calibration of instruments and the measurement methods deployed are very important, however in several studies most often this issue is not given adequate attention.The available literature shows that many studies lack the calibration details of instruments used, which is crucial for QA/QC purpose, i.e., Ezhilkumar et al., 2021;Budhavant et al., 2020;Islam et al., 2020a;Islam and Saikia, 2020b;Begam et al., 2017;Das et al., 2015 etc.Another issue of significant importance is variation in PM mass loading introduced due to different techniques.For comparison, PM 2.5 mass measured by MiniVol sampler during December 2013-January 2014 in Delhi was observed to be 277±100 μg/m 3 (Pant et al., 2015), while in another study PM mass measured using APM550 (high-volume sampler) in Delhi during January-February 2014 (Sen et al., 2014) was reported to be 178±56.5μg/ m 3 , i.e. about 43% variation in mass measurement.Here both the techniques used are gravimetric mass measurement, however calibration of impactor and cyclone are the major issues to be addressed.Thus, the performance evaluation and limitations of different measurement techniques should be well understood as per country's ambient conditions before deploying an instrument which would be beneficial to effectively monitor air quality.

CONCLUSION AND SUGGESTIONS FOR BETTER AND ACCURATE GROUND-BASED PM MEASUREMENT TECHNIQUES, PROTOCOLS AND POLICIES
For better air quality management (AQM), extensive technical and scientific efforts of government regulatory bodies along with research institutions/local universities is crucial to generate reliable air monitoring data.In order to ensure data reliability following protocols should be followed; (1) periodic calibration of flow rate (2) comparison/intermediate check of data obtained with reference instrument (3) calibration certificate provided by the manufacturer (Aggarwal et al., 2013).Some other important suggestions for improved air quality management are as follows: 1.For measurement of pollutants at air monitoring stations, the guidelines recommended by CPCB are followed.This review summarises ambient PM 2.5 and PM 10 mass concentration across different sampling locations in India during 2015-2019, and also compare PM 2.5 mass concentration with MERRA-2 reanalysis data.The objective was to map this data with temporal and spatial distribution to better understand PM pollution and thereby help policy makers to formulate better control policies.Some important conclusions drawn from the study are: 1. Due to pristine conditions of IHR, low PM mass concentration is observed when compared to other locations however due to urbanisation and long-range transport of pollutants from IGP, slight increase in the average mass concentration of PM is reported in many locations in last few years.Amongst the IHR sites, lowest PM 10 and PM 2.5 concentration were recorded in Kashmir in contrast to Dehradun which observed high PM mass concentration values showing uptrend/increasing concentrations in subsequent years, while Kullu recorded a slight drop in PM 10 concentration from 2014 to 2015. 2. Throughout the study period, ground sampling locations in IGP like Delhi, Agra, Varanasi, Kolkata reported the highest fine and coarse PM mass loading, due to rapid urbanisation, industrialisation and increased anthropogenic emissions from biomass/waste burning activities.The combined effect of local sources and stable atmospheric conditions further deteriorated the air quality during winters in these regions.Especially in Delhi, air quality is worsened due its landlocked geographical location.3. Most of the ground studies have been conducted in larger cities when compared to small cities and rural areas.This needs to be improved for better understanding of PM concentration levels and trends in order to comprehend the overall state of PM pollution in India.4. Variations in ground PM concentration are observed for the studies conducted in the same city.This may be due to different sampling locations and meteorological conditions, but to large extent variations are observed due to lack of standard operating methods.5.According to National Ambient Air Quality Standards, monitoring should be done for 24 h and at least 104 days in a year.However, the literature clearly indicates that most of the data are for limited days.6.There are associated uncertainties in data from ground monitoring networks due to lack of quality control procedures.In the absence of QA/QC, data quality remains questionable.7. MERRA-2 PM 2.5 exhibited seasonal variation with high mass loading in winter and low in monsoon.However, MERRA-2 mostly underestimated PM 2.5 mass concentration relative to corresponding ground measurement, which was even higher during pollution episodes indicating inability of MERRA-2's data assimilation when predicting higher mass concentration.
Effective air quality management has become one of the most important tools to be ensured in order to combat and control the negative impacts of atmospheric pollution on human health and environment.Hence, an accurate, continuous and precise air monitoring network is crucial for formulating effective abatement and control policies.The present study highlights better measurement techniques, protocols and policies to maintain quality and reliability in the air monitoring data.The development of RM would prove to be beneficial for monitoring bodies working in this domain.Regular quality and calibration checks are mandatory to ensure quality of data generated.All resources and efforts need to be prioritized in improving air quality network and encouraging scientific community towards advancement of better techniques and standards.

Fig. 2 .
Fig. 2. Distribution of average PM 2.5 and location of measurement sites along with corresponding Ground PM 2.5 and MERRA-2 PM 2.5 concentration during the time period 2015-2019 Fig. 3. Scatter plot for monthly mean (a) ground-based and MERRA-2 simulated PM 2.5 mass concentration and (b) log transformed ground-based and MERRA-2 simulated PM 2.5 mass concentration for the ensemble of locations in India.

Fig. 4 .
Fig. 4. Monthly average PM 2.5 mass concentration by ground-based and MERRA-2 simulation, mean bias and FAC2 values, for the ensemble of locations in India.

Table 1 .
Summary of ground-based ambient average mass concentration measurements of PM from several cities across India in the past five years (2015-2019).

Table 1 .
Summary of ground-based ambient average mass concentration measurements of PM from several cities across India in the past five years (2015-2019).

Table 1 .
Summary of ground-based ambient average mass concentration measurements of PM from several cities across India in the past five years (2015-2019).

Table 1 .
Summary of ground-based ambient average mass concentration measurements of PM from several cities across India in the past five years (2015-2019).

Table 2 .
Performance statistics of MERRA-2 PM 2.5 concentrations with regard to Ground PM 2.5 for the ensemble of locations in India in different seasons.AOC stands for "average observed concentration".
Available literature shows that greater number of studies are carried out in larger cities when compared to small cities and rural areas.This issue should be given more attention.The air monitoring network should be reliable, accurate and uniformly available so that data from different stations is comparable to better understand PM concentration levels and trends in India.3. Variations in PM concentration are observed for studies conducted in the same city.A part of this deviation could be due to different sampling locations and meteorological conditions, but to large extent variations are observed due to lack of standard operating procedures and limited availability of trained manpower.Therefore, all laboratories and stations involved in air quality monitoring should be accredited to ensure QA/QC procedures.Periodic workshops and capacity building training of all manpower, involved directly or indirectly in generating and reporting data, whether in fields and laboratories, should be arranged to familiarise them with appropriate sampling and analysis methods.