key: cord-0810339-byv8h2em authors: Meda, Bala Naga Manikanta; Mathew, Aneesh title: Temporal variation analysis, impact of COVID-19 on air pollutant concentrations, and forecasting of air pollutants over the cities of Bangalore and Delhi in India date: 2022-04-09 journal: Arab J Geosci DOI: 10.1007/s12517-022-09996-2 sha: 6634f2e4977b6dfc98ef1f8d13c7752a27e3e5fe doc_id: 810339 cord_uid: byv8h2em Indian cities are highly vulnerable to atmospheric pollution in recent years, due to exponential growth in urbanisation and industrialisation, and the increased pollution has been made to focus on the temporal variation analysis and forecasting of air pollutants over major Indian cities like Delhi and Bangalore. PM(2.5) concentrations are nearly 60.5% less than the annual average value during monsoon season while 76.3% more during the winter months. Ozone concentrations increase during the summer months (~ 46.3% more than the annual average) in Delhi, whereas in Bangalore, ozone concentrations are more (~ 75% more than the annual average) during the winter months. Variations of carbon monoxide and nitrogen oxides are significantly less comparatively. COVID-19 lockdown has a substantial positive impact on air pollution. Air pollutant concentrations are reduced during phase I and phase II of the lockdown. Pollutants, especially NOx and PM(2.5) concentrations, are drastically reduced compared to the previous years. NOx concentrations are reduced by ~ 20% in Bangalore, whereas ~ 50% in Delhi. PM(2.5) concentrations are reduced by ~ 41% in Delhi and ~ 55% in Bangalore. Forecasting of pollutants will be helpful in providing the valuable information for the optimal air pollution control strategies. It has been observed that linear model gives better results compared to ARIMA and Exponential Smoothening models. By forecasting, the concentration of NO(2) is 115.288 µg/m(3), the ozone is 30.636 µg/m(3), SO(2) is 11.798 µg/m(3), and CO is 2.758 mg/m(3) over Delhi in 2021. All the pollutants during forecasting showed a rising trend except sulphur dioxide. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s12517-022-09996-2. Air pollution is the presence in the ambient atmosphere of substances, generally resulting from human activities, in sufficient concentrations, for long enough periods, and under conditions that obstruct people's comfort or enjoyment of property (Indian Standard Institution IS-4167 1980) . It is the presence of one or more contaminants in the outdoor atmosphere, their characteristics, concentration, time duration affecting human health, vegetation, damages properties, and interference to the comfortable enjoyment of life (Wark et al. 1999 ). The concentration of air pollutants is primarily determined by the total amount of pollution released into the environment and the atmospheric conditions that influence the pollutants' fate. Vehicles, smokestacks, and other industrial emissions into the air and wind erosion of soil are significant sources of air pollution (Brusseau et al. 2019) . Anthropogenic and natural source emissions over long periods with enhanced concentrations alter pollutants' physical and chemical properties. For example, when oxides of nitrogen and volatile organic compounds (VOCs) in car exhaust have been emitted into warm, sunlight air experiences the immediate formation of O 3 . Hence, air pollution can be caused due to both physical and chemical actions. Air pollutants can fluctuate greatly, even when emissions are pretty consistent, in response to fluctuations in atmospheric conditions. When atmospheric conditions are very stable, even tiny emissions can lead to extreme levels of pollution (Ahammed et al. 2006) . Carbon monoxide, lead, groundlevel ozone, particulate matter, nitrogen dioxide, and sulphur dioxide, also identified as criteria pollutants, are major air pollutants with concern. Criteria pollutants are the only air pollutants with national air quality standards that establish allowable concentrations in ambient air (US EPA). Tropospheric ozone is one of the more severe pollutants, and the large abundances of ozone observed within urban regions are identified with emissions of ozone precursors, specifically, NO x , CO, VOCs, etc. (Beig et al. 2007 ). Higher level of ozone, especially in urban regions, is a major air quality issue and has huge impact on urban life (Li et al. 2019; Paoletti et al. 2014; Wang et al. 2009; Fuhrer et al. 1997) . Nitrogen oxides are essential in ozone formation chemistry and secondary aerosol formation, among other compounds. Natural (mostly lightning and microbiological processes in soil) and anthropogenic sources can cause the formation of ozone precursors. Vehicles and power plants are the primary sources of anthropogenic NO x concentrations (Rai et al. 2011 ). In the lower troposphere, the NO x sources are anthropogenic such as the combustion of fossil fuels. Biomass burning and microbial emissions from agricultural soils can be substantial contributors to rural regions (Van der A et al. 2008) . VOCs, one of the major precursors of ozone, contribute to formation of ozone at ground level at a major level through photochemical oxidation reactions. Quantitative assessment of VOCs impact on ozone will be helpful in emission reduction strategies of ozone in cities (Zhang et al. 2020) . The major sources of VOCs emissions are vehicles, petrochemical industries, paints, cooking gas usage, and fossil fuel burning (Lyu et al. 2016) . The quantities of non-methane hydrocarbons (NMHCs) which are emitted from vegetation mainly also contribute to the ozone formations (Trainer et al. 1987) . Carbon monoxide (CO) is another trace element in the troposphere whose significance has been well recognised. Its oxidation leads to O 3 formation or destruction, depending upon the NO concentration. The reaction of CO with hydroxyl radicals is the primary removal process from the atmosphere. Through this mechanism, CO acts as a significant precursor to photochemical ozone (Ahammed et al. 2006) . During hot and sunny summer episodes, ambient concentrations were found to be at their highest (Luna et al. 2014; Biancofiore et al. 2015) . Due to high relative humidity, there is a reduction in photochemical production efficiency and an increase in wet deposition, which reduces concentrations of O 3 (Lelieveld and Crutzen 1990; García et al. 2011) . Atmospheric movements of the air spread concentrations of pollutants like ozone and its precursors. Hence, wind speed and direction are also highly correlated with variations in ozone level concentrations (Revlett 1978; García et al. 2011) . Van der et al. (2008) derived trends of NOx globally over the period of 1996-2006 and found 7% declined emission annually. Peshin et al. (2017) have conducted spatio-temporal variation of atmospheric pollutants and anthropogenic effects on the ozone formation across the Delhi region during 2010-2014 and found that ozone is > 65% during October-February compared to other months. COVID-19 lockdown measures had resulted in a considerable change in air pollution worldwide. The complete shutdown of many industries and minimal usage of vehicle reduce pollution worldwide (Singh et al. 2020 ). There is a huge positive impact on the air pollution due to lockdown during this global pandemic (Dales et al. 2021; Liu et al. 2021; Barua and Nath 2021; Tian et al. 2021; Othman and Latif 2021) . Singh et al. (2020) have conducted a study over different regions of India and found significant reduction in particulate matter during lockdown. The highest decrease in particulate matter (~ 50-70%) was found for the northwest and Indo-Gangetic Plains. A significant reduction (~ 30-70%) in NO 2 was found except for a few sites in the central region. Similar reductions were observed for CO having a ~ 20-40% reduction. Kerimray et al. (2020) have investigated the effects of traffic-free urban settings on air quality in big cities during the COVID-19 lockdown in Almaty, Kazakhstan, and found PM 2.5 concentrations were lowered by 20% during the lockdown, variations ranging from 7 to 34% in various regions, compared to the average on the same days in 2018-2019. Many pollutant concentrations declined in during lockdown while few studies (Kerimray et al. 2020; Singh et al. 2020) reported ozone concentrations actually increased during lockdown due to suitable meteorological conditions. The present study primarily focusses on the temporal variations of air pollution, the impact of COVID-19 lockdown on air quality, and forecasting of pollutant concentrations for 2021. Atmospheric pollutants exhibit spatio-temporal variations. Spatial variations are generally due to variations in emissions in various locations. While temporal variations are mainly due to seasonal climatic changes. It is very important to study temporal variations to assess the behavioural pattern of pollutants for making strategies in reducing air pollution. Also, the study aims to analyse the effect of lockdown due to COVID-19 on the air pollutant concentration over the cities of Bangalore and Delhi. The formation of the various primary and secondary pollutants is very much complex in the atmosphere. Hence, it is required to develop forecasting model using advanced statistical models. Various air pollutants are also highly variable temporally as well as spatially. So, another objective of the present study is to perform forecasting modelling of various air pollutants using time-based regression models in terms of performance efficiency. The present work mainly focussed on two study areas: Bangalore and New Delhi. Bangalore is a south Indian city, while New Delhi is a north Indian city with extreme climates comparatively. Difference in climatic conditions and pollution level made me choose the cities for the study. Bangalore is a metropolitan city in Karnataka state. Being most significant information technology (IT) industry in India is in Bangalore; it is called Silicon city. The population of Bangalore is about 11 million. Bangalore has a tropical savanna climate as per Koppen climatic classification 'Aw' with distinct wet and dry seasons. Bangalore usually has a more moderate climate throughout the year because of its higher elevation. The average annual rainfall is about 974.5 mm (http:// www. banga lore. clima temps. com/). Bangalore is home to a wide range of heavy and light industries and high-tech and service industries, including IT and electronics, telecommunications, aerospace, and many other industries. Vehicular emissions account for 60-70% of Bangalore's pollution (Karnataka State Pollution Board). One of the most significant sources of air pollution is vehicular pollution. Inadequate urban governance also affects Bangalore, causing waste treatment and the use of diesel generators for electricity. Figure 1 shows the geographic location of Bangalore study area. New Delhi, India's capital city, is a cosmopolitan city with 30.29 million people and a land area of about 42.7 km 2 . Delhi's climate mixes monsoon-influenced humid subtropical (Koppen climate classification Cwa) and semi-arid (Koppen climate classification BSh), with significant differences in summer and winter temperatures and precipitation. The average annual rainfall is about 800 mm (http:// www. newdelhi. clima temps. com/). Figure 2 represents the geographic location of Delhi. IT, telecommunications, banking, hotels, media, and tourism are among the key industries. Delhi's manufacturing industries have also grown as many consumer goods companies have established manufacturing units and offices in the region. Apart from industries, animal agriculture contributes to Delhi's pollution problem, as smog and other harmful particles are produced by farmers burning their crops in other states. Animal agriculture accounts for approximately 80% of agricultural land. Animal agriculture is also a contributing factor to Delhi's air pollution. The data required for the study have been collected for both cities, i.e. Bangalore and New Delhi. The data have been collected from all India CAAQMS (Continuous Ambient Air Quality Monitoring Station) portal (https:// app. cpcbc cr. com/ ccr/#/ caaqm-dashb oard-all/ caaqm-landi ng). For Delhi, the data have been collected from four CAAQM stations. They are Mandir Marg, Anandvihar station, R K Puram, and Shadipur stations. For Bangalore, the data have been collected from BTM Layout, City Metro Station, and Silk Board stations. For temporal variations, the daily average concentrations of various pollutants, viz., carbon monoxide, oxides of nitrogen, ozone, and particulate matter, are collected for years 2017, 2018, and 2019 from all the above stations. For forecasting analysis, the annual average of various pollutants is collected from 2013 to 2019 and forecasted for 2021. Pre-processing of data has been carried out after the collection of data. Data pre-processing helps to deal with missing data and inconsistent data. Data pre-processing includes cleaning, normalisation, transformation, and deletion of noisy data. The removal of out of range data is an essential task in data pre-processing. For example, due to errors, the concentration value may be negative in the data. Such data must be removed. This is called out of rand range data removal. Conversion of pollutant concentration values into required units is also done in this step. Figure 3 shows the flowchart of methodology which illustrates the step by step procedures of methodology adopted for the present study. The collected data is pre-processed and then transformed into categories of interest. For temporal variation in the pollution, the daily average concentrations are clustered according to the months for further analysis. To analyse hourly variations, each hour data over the period is clustered separately, and various graphs have been plotted. Then, categorised data is used for finding temporal variation analysis and variations in pollutants due to COVID-19 lockdowns. The annual average data of pollutants are collected, and using various predictive analysis techniques, forecasting of pollutants has been conducted. Time series regression is a statistical method for predicting future responses using previous experiences and the transfer of dynamics from relevant predictors. From observational or experimental data, time series regression can help understand and predict the behaviour of dynamic systems (Ibrahim et al. 2009 ). Various models address time-based regression like the linear model, the Auto-Regressive Integrated Moving Average model, and Exponential smoothening. Here, y dependent variable x i i th independent variable or predictor variable θ intercept or bias variable β coefficient of predictor variable The Ordinary Least Squares technique is used to optimise the error in the linear model. The function to be optimised is where, ARIMA is an acronym that stands for Auto-Regressive Integrated Moving Average. This acronym is descriptive, capturing the key aspects of the model itself (Hyndman and Khandakar 2008). Briefly, they are: • AR: Autoregression. A model that uses the dependent relationship between an observation and some number of lagged observations. • I: Integrated. The use of differencing of raw observations (e.g. subtracting an observation from an observation at the previous time step) in order to make the time series stationary. • MA: Moving Average. A model that uses the dependency between an observation and a residual error from a moving average model applied to lagged observations. It is a type of model that explains a time series based on its past values, lags, and forecast errors. As a result, that equation can be used to forecast. A standard notation is used of ARIMA (p, d, q) where the parameters are substituted with integer values to quickly indicate the specific ARIMA model being used. The parameters of the ARIMA model are defined as follows: • p: The number of lag observations included in the model, also called the lag order. • d: The number of times that the raw observations are differenced, also called the degree of differencing. • q: The size of the moving average window, also called the order of moving average. Exponential Smoothening is one of the forecasting methods which is helpful in forecasting the data of no clear trend or seasonal pattern. In simple exponential smoothening, forecasts are calculated using weighted averages, where the weights decrease exponentially as observations come from further in the past the smallest weights which are associated with the oldest observations: where, y T+1 forecasted value for T + 1 observation from T observation α smoothening parameter and 0 ≤ α ≤ 1 The model performance has been evaluated by calculating various error standards, correlation coefficient (R), and determination (R 2 ). R 2 value determines the goodness of fit between actual model predicted values. Also, various error standards for a model like Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Square Error (MSE), and Root Mean Square Error (RMSE) will give a better view of how the actual and predicted values vary. Also, the correlation plot (actual vs. predicted) shows a clear relation between actual and predicted. The coefficient of determination (R 2 ) value gives the proportion of the variance in the dependent variable that is predictable from the independent variable. MAE gives the average difference between actual and predicted values. MAPE tells the percentage deviation of the prediction value from the actual value. (3) MSE is a combined measurement of the mean and variance of the error. RMSE is the square root of MSE where, a i actual observation of i th observation in data p i predicted observation of i th observation in data Various air pollutant concentration data have been collected from CAAQMS for Delhi and Bangalore cities for carrying out temporal and seasonal variation analysis from 2017 to 2019. The data have been pre-processed, and the average monthly pollutant concentrations are tabulated. The monthly and annual average pollutant concentrations of NO x , ozone, PM 2.5 , and CO have been tabulated for Delhi city. It has been observed that PM 2.5 concentrations are very high during winter months and least during monsoon months. The PM 2.5 has been following a trend with a higher value in January, and it keeps reducing up to September, i.e. completing summer and monsoon, and then hikes up to the end of the year, i.e. winter in Delhi. Page 7 of 20 736 The trend of ozone has behaved quite differently compared to other pollutants. The ozone has encountered higher concentrations in summer and least during the winter months of January and February. There is a sudden spike in the trend during April and May. The reason behind this trend variation is because of the ozone dependence on the meteorology of the area. In summer, due to higher solar radiations, higher ozone concentrations are formed. The monthly and annual average pollutant concentrations of NO x , ozone, PM 2.5 , and CO have been tabulated for Bangalore city. Various charts have been developed for analysing temporal variations in Bangalore. Various graphs have been developed and showing trend analysis in Bangalore as depicted in Fig. 5 . The annual average PM 2.5 concentrations have been found to be 43.45, 38.08, and 41.55 µg/m 3 for the years 2019, 2018, and 2017, respectively, in Bangalore. In 2019, the highest concentration was occurred in January and was 74.75 µg/ m 3 , while the least concentration was 20.9 µg/m 3 in August. In 2018, the highest and the least were 73.29 and 14.59 µg/ m 3 in December and August, respectively. During monsoon, the least concentrations have been recorded, and the highest was recorded in the winter months. The difference in concentrations between the summer months and winter months is pretty less. The trend has started with higher values in January and continued at a steady pace until April and has been falling until September (i.e. up to monsoon season) and again raised until December (due to winter). The trend of PM 2.5 in both Delhi and Bangalore cities is similar in comparison. The concentrations of PM 2.5 in Delhi have been found to be 3 to 4 times greater than those in Bangalore. The concentrations in Bangalore are mostly under permissible limits, while in Delhi, PM 2.5 is one of the primary pollutants that is devastating the environment. The The NO x pollution in Delhi is very much higher than that of Bangalore. The trend analysis in Delhi shows a seasonal variation, while Bangalore is not showing any seasonal variations. The pollutant concentrations in Delhi have been found to be 10-12 times higher than those in Bangalore. The The CO concentrations have been reduced from 2017 to 2019 in Delhi, while it was observed that there is an increase in CO concentrations in Bangalore, which is an adverse effect. Seasonal variations have also been noticed in Delhi for CO, while no such trend has been observed in Bangalore. The CO concentrations in Delhi have been observed to be 2 to 5 times more than those in Bangalore. Page 9 of 20 736 There is a strong trend followed by the ozone in Bangalore with higher concentrations during winter and least during monsoon. Higher concentrations of ozone have been recorded in winter in Bangalore, while in Delhi, the higher concentrations are found in summer. Meteorology may also influence the air pollutants' concentration variations in both cities. The ozone formation is influenced by meteorology. Even though pollutant concentrations are higher in Delhi, it has been observed that ozone concentrations are higher in Bangalore. Compared to Delhi city, the correlation of meteorological parameters with mean ozone concentrations in Bangalore is much higher. The meteorology in Bangalore city has a higher impact on ozone concentration at the ground level than in Delhi city. Wind speed and wind direction affect the mixing and spread of pollutant concentrations (Revlett 1978 ) and its correlation with ozone in Bangalore is higher than that in Delhi city. However, the amount of the variations of pollutant concentrations is also needed to find. By making yearly average as a benchmark for the pollutant concentration over the year, the percentage of pollutant concentration varied with the average value has been calculated. The charts have been developed showing percentage variations for different years and have been depicted in Figs. 6 and 7. All the pollutant concentrations have been found to be higher than the annual average concentrations during the winter months. The monthly variation of carbon monoxide over the years for both cities is less compared to other pollutants. The highest positive and negative variation has been found to be 61.73% and − 55.16% during months of January and July, respectively for the year 2019 in Bangalore. While in Delhi, it has been noticed to be 53.01% and − 43.25% during November and June, respectively, for the year 2017. Other pollutants like ozone have experienced more than 100% variations. The highest positive and negative percentage variation of ozone was 109.19% during Jan 2017 and − 55.168% during July 2019 in Bangalore, and it was 110% during April 2017 and − 33.31% during August 2019 for Delhi. Ozone is the only pollutant among the considered (9) %Variation = concentration in a month − Average value in that year Average value in that year * 100 pollutants whose trend is varied between the two cities because ozone is regulated by meteorology. Oxides of nitrogen did not show significant variations over the period. Even though the highest positive variation in NOx in Bangalore was 92.79%, the variation of more than 50% never happened over 3 years. The highest negative variation that happened in Bangalore was − 55.94%. Particulate matter (PM 2.5 ) is one of the pollutants which follow a strong trend for both cities. Monsoon is the season with the least PM 2.5 pollution, but during the winter, PM 2.5 pollution is at the worst level. The average monthly variation is higher than the annual average values during winter, with a highest of 92.47% in Bangalore and 103.24% in Delhi, while the highest negative is − 67.13% and − 69.89%, respectively for Bangalore and Delhi cities. Simply considering annual data is not advised and temporal variations must be taken into consideration in order to analyse the pollutant concentration variations. During the COVID-19 lockdown period in India, due to the shutdown of industries, minimal traffic and lessened anthropogenic activities accounting for pollution helped to improve the urban cities' air quality. The phase I lockdown was put in place from 25 th March 2020 with a complete shutdown of all services and factories for 21 days, i.e. up to 14 April 2020. At the end of phase 1, it was announced by the Government of India to extend the lockdown up to 3 rd May 2020, i.e. for 20 more days as phase II of lockdown. How the COVID-19 lockdown influenced the reduction in air pollution has been carried out by comparing pollution during the lockdown period with the previous years (2019 and 2018) which is tabulated, for Delhi and Bangalore cities, respectively. Along with the period of lockdown, the comparison in air pollution variations has also been made for 25 days before and after lockdown. The analysis has been made featuring the impact of COVID-19 lockdown on air pollution in Delhi and Bangalore cities. Various graphs have been developed showing the comparison of daily average concentrations of various air pollutants like NO 2 , ozone, PM 2.5 , and CO during lockdown period and the same period's pollutant concentrations during 2018 and 2019. Also 10 days before and after lockdown, the pollutant concentration variations are also depicted in Figs. 8 and 9 for Delhi and Bangalore, respectively. In Delhi, all the pollutants show a reduction during lockdown period. During COVID-19 period (i.e. 25 March 2020 to 5 May 2020), the concentrations of NO 2 have reduced drastically. The concentrations had not reached value of 30 µg/m 3 during this period, while during same days in 2018 and 2019, NO 2 concentrations were as high as 100 µg/m 3 . No days during lockdown have exceeded the concentrations of respective day in 2018 and 2019. The concentrations of PM 2.5 have been reduced drastically during lockdown in comparison with previous years. The daily average concentrations have not reached 100 µg/m 3 during lockdown and the highest was 72.18 µg/m 3 . In 2018 and 2019, during the same period of time, the concentrations on some certain days had reached the higher values up to 217.87 µg/m 3 which was quite higher than daily average standard value of 60 µg/m 3 as per NAAQS (National Ambient Air Quality Standards) standards. During the lockdown period, the concentrations of PM 2.5 were highly controlled and under NAAQS standards for most of the days, while during the same period in previous years, the concentrations were quite higher than daily average standard value. The concentrations of ozone were also reduced during the lockdown period. The concentration has reached as low as 50 µg/m 3 during lockdown, while in previous years during the same period, the daily average concentrations have reached as high as 94.68 µg/m 3 . During phase II of lockdown, the concentrations of ozone have been on higher side compared to previous years. During phase I, the daily average ozone concentrations exceeded those of previous year which was two, while during phase II, it was 15 days. Even though the concentrations of NO 2 have been reduced significantly, ozone concentrations did not reduce. This is because of the regulation of ozone not only by ozone precursors but also due to meteorology (Feng et al. 2019) . The concentrations of CO also have been reduced but not as drastically as NO 2 and PM 2.5 in Delhi. This has happened even though anthropogenic emissions reduced because natural source emissions were higher in Delhi. The daily average CO concentrations were reduced up to 0.7 mg/m 3 , and in previous years, it was about 2.5 µg/m 3 . The air pollution concentrations were reduced in Bangalore also due to lockdown. The air quality has been increased gradually during lockdown. The daily average PM 2.5 concentrations in Bangalore have been reduced drastically compared to previous years 2018 and 2019. The highest daily average concentration during lockdown was less than 47.02 µg/m 3 , while in 2019, it was 114.07 µg/m 3 and in 2018, it was 231.38 µg/m 3 . The concentrations of NO 2 have been reduced slightly during lockdown in Bangalore. The NO 2 concentrations were very high during 2018 and 2019; the concentrations have reduced compared to 2018. The reductions in daily Analysing the various pollutant concentrations during COVID-19 with previous years (2018 and 2019), concentrations of the same day show the significant reduction in pollution over both cities. Overall pollution has been reduced in 2020 when compared to 2019. The percentage reduction of air pollution during lockdown phases I and II is depicted in Table 1 and Table 2 . So, to know the extent of reduction during the lockdown, the comparison also included 25 days prior to and later the lockdown and is depicted in Figs. 10 and 11. Twenty-five days before lockdown, the concentrations during 2020 were likely a little less than those in 2019. However, during the lockdown, the concentration has reduced more than before the lockdown. After lockdown, some of the pollutant concentrations have increased more, even a partial lockdown was imposed. The daily average concentrations of CO in Bangalore and Delhi were reduced by 37.06% and 13.23%, respectively, during phase I and 21.95% and 20.12% during phase II. The daily average concentrations of CO were 0.69 mg/m 3 and 1.26 mg/ m 3 during 25 days before lockdown, and during lockdown in Bangalore, average concentrations of CO were 1.12 mg/ m 3 and 0.88 mg/m 3 . There is almost 50% reduction in daily average concentrations in comparison to 25 days before and during lockdown. PM 2.5 is one of the major pollutant concentrations which were also reduced drastically during the lockdown. There was 60.43% reduction in Bangalore, and it was 45.57% in Delhi during phase I. During phase II, the reduction was lesser than that during phase I and was 50.14% and 36.21% in Bangalore and Delhi, respectively. The daily annual average concentrations were reduced by 18.5% and 25.13% in Bangalore and Delhi cities during phase I. It was observed that the ozone concentrations rather than reduced were actually increased by 4.13% which is due to favourable meteorology in Delhi especially due to increase in solar radiation, and in Bangalore, it was reduced by 10.12% during phase II. The oxides of nitrogen were also reduced by 14.05% during phase I in Bangalore and there is drastic reduction of 44.25% in Delhi. During phase II of lockdown, the reduction in Bangalore was 15.96%, and in Delhi, it was 55.62%. Unlike other pollutants, NO 2 concentrations were reduced more during phase II than phase I. On an overall scrutiny, lockdown has shown an impeccable reduction in pollution. The results show that PM 2.5 , CO, and ozone reductions in Bangalore are more compared to Delhi, while NO 2 reductions in Delhi are more compared to Bangalore. The forecasting of the pollutants will help to protect the health by warning early and setting better pollution control policies. The forecasting of the air pollutants over Delhi city has been performed using time-based regression models. Various models like linear, ARIMA, and Exponential smoothening were run for forecasting of pollutants. Among all the models, the linear models show a good correlation in modelling compared to other models with very good R 2 values ranging from 0.8 to 0.9. Table 3 depicts the R 2 values of various models used in forecasting of various pollutants, viz., NO 2 , SO 2 , ozone, and CO. Comparison of forecasted concentrations with actual concentrations for year 2020 in Delhi city is shown in Table 3 . Based on R 2 values found for models used for forecasting and comparison of actual concentrations in 2020 with forecasted values of various models, it has been found that linear models are best fit for forecasting. The forecasted pollutant concentrations for 2021 are depicted in Table 4 . For forecasting, the data of annual averages of pollutants from years 2015 to 2019 is only used due to availability of limited data. Average data of year 2020 is used for validation. Being the data is limited, the forecasting is done for year 2021 only. All the models show good correlation between time and annual averages. There was an increase in pollution levels from 2015 to 2017 and has been found the trend of pollution levels started decreasing from 2018 for all the pollutants of interest which is a good sign for the cities. The reduction in pollution levels in Delhi is due to few pollution control programmes launched by Delhi government. Table 5 shows the forecasted pollutant levels in Delhi. Figure 12 shows the linear trend followed by the annual averages of pollutants and forecasted value for year 2021 with 95% and 85% confidence limits. Forecasted pollutant concentrations for the year 2021 for Delhi are shown in Fig. 12 . The forecasted pollutant concentrations of NO 2 , SO 2 , ozone, and CO are 115.266 µg/m 3 , 11.798 µg/m 3 , 30.636 µg/m 3 , and 2.758 mg/m 3 , respectively. This shows that pollution is not increasing rapidly for the past 5 years. This has shown the pollution control programmes taken place in the recent years are successful which is a positive sign. The monthly variations in the various pollutants over Delhi and Bangalore cities have been analysed, along with the impact of COVID-19 lockdown on air pollution. Forecasting analysis has also been carried out for various pollutants to find the air pollutants concentrations in the years 2020 and 2021. In Bangalore, from 2017 to 2019, annual average concentrations of CO were increased by 79.6%, PM 2.5 by 4.910%, and NOx by 105%. In contrast, ozone concentration was reduced by 1.7%. In Delhi, the annual average of many pollutants showed a negative trend. Annual average concentrations were reduced by 38.1%, 11.4%, and 4.9% from 2017 to 2019 for NOx, PM 2.5 , and CO, while ozone concentrations were increased by 9%. NOx concentrations showed a steady monthly average over the years with 1 or 2 months with slightly elevated values which may be due to biomass or agricultural waste burning. The monthly average values of ozone concentrations in Bangalore were higher than the annual average by 70--100%. In comparison, the ozone in Delhi was higher during the summer months ~ 46.3% more than the annual average. Carbon monoxide concentrations were almost steady in other months of the year. PM 2.5 concentrations showed many variations during the months of the year in both Bangalore and Delhi. PM 2.5 concentrations were nearly 60% less than the annual average value during the monsoon season, while it was nearly 75% more during the winter months in both cities. In Bangalore, various pollutants, viz., CO, NO 2 , PM 2.5 , and ozone concentrations, during lockdown were reduced by 29.9%, 55.4%, 16.2%, and 14.9%, while in Delhi, concentrations were reduced by 16.7%, 40.9%, 10.5%, and 49.9%, respectively, compared to the previous year. In contrast, meteorology is almost similar to the previous year. In Delhi, during phase 2, the ozone concentration increased by 4.1% while NO 2 concentrations decreased by 49.93%, which shows ozone formation does not solely depend on NOx concentrations. Time-based regression models have been used to predict various air pollutants and it has been observed that linear model gives better results compared to ARIMA and Exponential Smoothening models. Linear models show a good correlation in modelling compared to ARIMA and Exponential Smoothening models with R 2 values ranging from 0.8 to 0.9. By forecasting, the concentration of NO 2 is 115.288 µg/m 3 , the ozone is 30.636 µg/m 3 , SO 2 is 11.798 µg/m 3 , and CO is 2.758 mg/m 3 over Delhi in 2021. For further scope of research, forecast studies using various machine learning techniques can be conducted, and thus, the potential of new models can be exploited for predicting pollutant concentrations and thereby giving continuous ambient air quality predictions. The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s12517-022-09996-2. Data availability The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. Seasonal variation of the surface ozone and its precursor gases during 2001-2003, measured at Anantapur (14.62°N), a semi-arid site in India The impact of COVID-19 on air pollution: Evidence from global data 12 Forecasted pollutant concentrations for 2021 for Delhi Page 19 of Modelling the effects of meteorological variables on ozone concentration -A quantile regression approach Simultaneous measurements of ozone and its precursors on a diurnal scale at a semi urban site in India Analysis of surface ozone using a recurrent neural network Ambient air pollution exposure estimation for the global burden of disease 2013 Environmental Impacts on Human Health and Well-Being. Environmental and Pollution Science Visibility graphs of ground-level ozone time series: A multifractal analysis The association between air pollution and COVID-19 related mortality in Santiago, Chile: A daily time series analysis A comprehensive analysis of the spatiotemporal variation of urban air pollution in China during Unveiling tropospheric ozone by the traditional atmospheric model and machine learning, and their comparison: A case study in hangzhou Critical levels for Ozone effects on vegetation in Europe Artificial neural network model for ozone concentration estimation and Monte Carlo analysis Artificial Neural Network Models for Prediction of Ozone Concentrations in Guadalajara Prediction of ground-level ozone concentration in São Paulo, Brazil: Deterministic versus statistic models Automatic time series forecasting: The forecast package for R Forecasting and Time Series Analysis of Air Pollutants in Several Area of Malaysia Glossary of terms relating to air pollution Trends in ozone concentration distributions in the UK since 1990: Local, regional and global influences Assessing air quality changes in large cities during COVID-19 lockdowns: The impacts of traffic-free urban conditions in Almaty Influences of cloud photochemical processes on tropospheric ozone Anthropogenic drivers of 2013-2017 trends in summer surface ozone in China Ground-level ozone pollution and its health impacts in China Spatiotemporal impacts of COVID-19 on air pollution in Prediction of ozone concentration in tropospheric levels using artificial neural networks and support vector machine at Rio de Janeiro, Brazil Ambient volatile organic compounds and their effect on ozone production in Wuhan, central China Quantifying temporal trends in ground level ozone concentration in the UK Rethinking the Ozone Problem in Urban and Regional Air Pollution. National Academies Press Othman M, Latif MT (2021) Air pollution impacts from COVID-19 pandemic control strategies in Malaysia Ozone levels in European and USA cities are increasing more than at rural sites, while peak values are decreasing Artificial Neural Networks: Multilayer Perceptron for Ecological Modeling Comparison of neural network models with ARIMA and regression models for prediction of Houston's daily maximum ozone concentrations Gaseous air pollutants: a review on current and future trends of emissions and impact on agriculture Spatial modeling of daily concentrations of groundlevel ozone in Montreal, Canada: A comparison of geostatistical approaches Ozone Forecasting Using Empirical Modeling Spatio-temporal variation of air pollutants and the impact of anthropogenic effects on the photochemical buildup of ozone across Delhi-NCR Modeling and Prediction of Hourly Ambient Ozone (O 3 ) and Oxides of Nitrogen (NO x ) Concentrations Using Artificial Neural Network and Decision Tree Algorithms for an Urban Intersection in India Outdoor and Indoor Ozone Concentration Estimation Based on Artificial Neural Network and Single Zone Mass Balance Model Diurnal and temporal changes in air pollution during COVID-19 strict lockdown over different regions of India Impacts of primary emissions and secondary aerosol formation on air pollution in an urban area of China during the COVID-19 lockdown Models and observations of the impact of natural hydrocarbons on rural ozone Trends, seasonal variability and dominant NOx source derived from a ten year record of NO 2 measured from space Increasing surface ozone concentrations in the background atmosphere of Southern China Air Pollution: its origin and control A neural network model forecasting for prediction of daily maximum Ozone concentration in an industrialized urban area The impact of volatile organic compounds on ozone formation in the suburban area of Shanghai The authors declare no competing interests.