Action Plans to Reduce PM 2.5 Concentrations in Hotspots of Delhi-NCR Using a One-way Coupled Modeling Approach

The concentration of PM 2.5 in Delhi, one of the most polluted capital cities globally, frequently exceeds the Indian National Ambient Air Quality Standards, especially during the post-monsoon and winter months. This study evaluates the changes in PM 2.5 concentrations across Delhi using a one way coupled model (WRF-CMAQ-AERMOD) for various hotspot-specific intervention scenarios during post-monsoon and winter of 2018. PM 2.5 concentrations reduced up to 15% by scaling down total emissions across Delhi by 20%. An additional 9% reduction across entire Delhi and ~ 28% reduction at the top ten observation-based hotspots could be achieved if emissions of industry, unpaved road dust and construction in selected emission hotspots are made zero. Non-local contribution in hotspots of the city varied significantly. For example, the difference in the reductions of PM 2.5 concentrations from citywide versus hotspot-specific interventions is estimated to be 28% at DU North Campus and 11% at Anand Vihar. An average reduction of 12% was computed when construction and MSW burning emissions were down sized 100% in the locations identified based on compliance complaints received at the Central Pollution Control Board’s online complaint portal. A marginal reduction of 4% estimated for the previously implemented traffic rationing measure, the odd-even rule, indicates that such regulation of vehicles alone might be inefficacious. The results suggest that stakeholders must focus on — (a) source and hotspot-specific interventions alongside city wide interventions to significantly reduce ambient PM 2.5 concentrations, (b) local and non-local contributions from regions outside the hotspot grid needs to be carefully considered for estimating efficacy of an action plan.


INTRODUCTION
Air pollution in Delhi, the national capital of India, has come up as the most significant challenge and requires immediate attention.Rapid urbanization, industrialization, and increasing vehicle numbers have led to elevated air pollutants concentrations in Delhi and are expected to further deteriorate in the coming decades (Chowdhury et al., 2018;Conibear et al., 2018).In addition, the pollutants transported over from neighbouring states have made the situation further complex.As a result, Delhi was ranked as the most polluted capital city globally for the years 2018, 2019, and 2020(IQAir 2018, 2019, 2020)).With an estimated population of over 21 million (MoHFW, 2019) and a high population density, people in Delhi are exposed to high enough levels of pollutants, especially PM 2.5 , that may cause many acute as well as chronic effects (Lim et al., 2012;Forouzanfar et al., 2015;Feng et al., 2016;Forouzanfar et al., 2016;Sharma et al., 2020a).Sahu and Kota (2017) found out that for the years 2011-2014, PM 2.5 exceeded the National ambient air quality standards for 85% of the total days and was the dominant pollutant for 75-90% of the total days PM 2.5 in Delhi.PM 2.5 resulted in 227.47 years of life lost per 1000 population (Sahu et al., 2020) and was associated with an excess risk of 6% (Guo et al., 2019b) and ~1.06 million premature mortality (Guo et al., 2018) in Delhi.The population-weighted annual mean PM 2.5 loading in Delhi was found to be about 3.5 times higher than the global populationweighted value of 20 µg m -3 and ~22% higher than average Chinese city-level value during the years 2015-2018, and caused 10,200 (95% confidence interval: 6800-14,300) premature mortality each year (Chen et al., 2020).Pandey et al. (2021) estimated the population-weighted mean ambient PM 2.5 concentrations in the year 2019 to be 217.6 (95% UI: 117.9-297.3)µg m -3 with 16,595 (14,345) deaths and 2,353 (1,997-2,739) years of life lost attributable to the particulate matter air pollution.Also, many studies have found out that the concentrations of PM 2.5 in Delhi have been found to be the highest in the post-monsoon season and winters (Tiwari et al., 2012;Peshin et al., 2017;Kota et al., 2018;Hama et al., 2020;Mogno et al., 2021).
Many studies in the past have studied the contributions of various sources to the total PM 2.5 concentrations in Delhi.Sharma et al. (2016) concluded that secondary aerosols contributed the maximum of 21.3%, followed by soil dust (20.5%), vehicular emissions (19.7%), biomass burning (14.3%), fossil fuel combustion (13.7%), industrial emissions (6.2%) and seas salt (4.3%) to the PM 2.5 in central Delhi.Similarly, Jain et al. (2020) estimated that biomass burning contributed the maximum of up to 23% to total PM 2.5 , followed by vehicular emissions (16%), fossil fuel combustion (10%) and industrial emissions (10%) for central Delhi.Two other studies estimated that major emission sources for northern Delhi, were vehicular emissions (32-35%), biomass burning (26-30%), cooking emissions (15-17%), plastic burning (13.5%) and other sources (7-9%) including biogenic and industrial emissions (Gadi et al., 2019;Shivani et al., 2019).However, Guo et al. (2017) studied the source contributions to PM 2.5 in Delhi using a chemical transport model and estimated that industries (35%) were the significant contributors, followed by residential (23%) and energy (15%) sectors were the major contributors.Many other source apportionment studies that studied the samples collected at various sites over Delhi, have also highlighted how the source contributions change over different locations across Delhi (Sharma and Dikshit, 2016;ARAI and TERI, 2018;IITM and IITD, 2019).All these studies emphasize the need for targeted source-and hotspot-specific mitigation strategies, which can be very useful in improving PM 2.5 concentrations in Delhi.
In the past, the central and state governments have taken many steps to curb the high pollutant concentrations in Delhi, especially during post-monsoon and winter.In the year 2019, the Central Government launched the National Clean Air Programme (NCAP) to reduce the particulate matter concentrations in 132 non-attainment cities (https://cpcb.nic.in/uploads/Non-Attainment_Cities.pdf)across India by 20-30% by 2024 (MoEFCC, 2019).However, it does not include sector-specific emission reduction targets and control strategies to achieve the concentration targets.For Delhi and the National Capital Region, a Graded Response Action Plan (GRAP) (https://app.cpcbccr.com/ccr_docs/final_graded_table.pdf) was notified in January 2017 by the Ministry of Environment, Forest and Climate Change.The plan includes a set of interventions triggered in stages based on the deterioration of the air quality.Apart from these, the Government of India has introduced the Swachh Bharat Mission (SBM) to improve solid waste management, Unnat Chulha Abhiyan to promote the use of improved cookstoves and clean fuel, the Pradhan Mantri Ujjwala Yojana (PMUY) and the National Electric Mobility Mission Plan (NEMMP) to scale up the adoption of zero-emission vehicles to tackle the problem of air pollution (PIB, 2018;Gulati, 2020;PMUY, 2021;SBM, 2021).The Government of Delhi implemented traffic rationing measures in the form of the odd-even scheme (GNCTD, 2015), which is also one of the interventions proposed in GRAP in the year 2016.The average PM 2.5 concentrations within 25 km around the city before, during and after the vehicle-rationing intervention were greater than 160 µg m -3 , six times higher than the 24-hour WHO standards (Sinha and Kumar, 2019).Also, the effects of such interventions were short-lived and unobservable as Delhi witnessed a 34% increase in PM 2.5 during the two weeks after the interventions (Tiwari et al., 2018;Sinha and Kumar, 2019).
However, the recent COVID-19 lockdowns implemented in the country have shown that strict control over the sources can significantly reduce pollutant concentrations.During the lockdown period, the concentrations of PM 10 and PM 2.5 in Delhi reduced by about half compared to prelockdown concentrations (Dhaka et al., 2020;Mahato et al., 2020;Sharma et al., 2020b).These reductions prove that effective source control strategies can be beneficial to control the PM 2.5 concentrations in Delhi.The significant contributions of the local sources to the total PM 2.5 (Guo et al., 2019a) in Delhi also emphasize the need for interventions targeting the local emission sources.
Delhi has 2459 registered industries, including food and beverages, textiles, paper and paper products, primary metal and alloy, leather and leather products, chemicals, etc. (GNCTD, 2020).Three natural gas power plants are currently operational in Delhi (GNCTD, 2021).Delhi had approximately 11 million registered vehicles in 2019-20, out of which the public transport in the city runs mainly on compressed natural gas (CNG) while other vehicles run on diesel, gasoline and CNG (GNCTD, 2020(GNCTD, , 2021)).The areas where these sources are concentrated may experience very high concentrations of pollutants.Such locations, where emissions from specific sources may expose individuals and population groups to elevated risks of adverse health effects, are defined as hotspots (Bahadori and Smith, 2016).Identifying such hotspots can be an effective strategy to manage the air quality as curbs on the emissions from just these hotspots can improve the air quality around these hotspots.Hotspot identification also allows the regulatory authorities to effectively implement curbs on these selected locations to manage air quality in the hotspot and area in the vicinity.Thirteen such air pollution hotspots were identified based on the annual average concentration of PM 2.5 and PM 10 in Delhi by the Delhi Pollution Control Committee (DPCC) (https://www.dpcc.delhigovt.nic.in/uploads/sitedata/hotspot.pdf).
Thus, there is a pressing need for efficient, effective, and validated air quality management, including interventions that must be immediately implemented to counter the increasing air pollution and its associated effects.This study's primary goal is to analyze and simulate PM 2.5 concentrations using the U.S. EPA's AERMOD coupled with WRF (Weather Research Forecast Model) to estimate the area-specific potential decrease in concentrations under various intervention scenarios during the post-monsoon and winter seasons when the concentrations of PM 2.5 are the highest.

Study Area and Data Sources
New Delhi, India's capital, is a land-locked city in northern India at 28.61°N 77.23°E.Delhi is one of the most densely populated cities of India with a population density of 11,320 km 2 (GNCTD, 2020) with an estimated total population of 16.78 million in the year 2011 and is projected to be 20.57 million in the year 2021 (MoHFW, 2019;GNCTD, 2021).It faces extreme temperatures of as low as 4°C in winter to 45°C in summer.Delhi receives an annual average rainfall of 600-800 mm.
The hourly meteorological data, including wind speed, wind direction, temperature, and relative humidity and PM 2.5 concentration data for Delhi for the past three years (2016)(2017)(2018) for the CPCB, DPCC, and IMD monitoring stations over Delhi, was obtained from the CPCB online portal for air quality data dissemination (https://app.cpcbccr.com/ccr/#/caaqm-dashboard-all/caaqmlanding).The details of the data used for the simulations have been shown in Table 1.

WRF-AERMOD Modelling System
PM 2.5 concentrations in Delhi were modelled using the Air Quality Dispersion modelling system (AERMOD).Fig. 1 shows the methodology adopted to simulate PM 2.5 in this study.The small-scale model used in this study, AERMOD is a steady-state plume model developed by American Meteorological Society (AMS) and the U.S. Environmental Protection Agency (U.S. EPA) and adopts a Gaussian approach to dispersion modelling.As AERMOD is dispersion model with limitation in domain size and capability to handle longer range transport, a one-way coupled modeling network was used.This modeling network has a chemical transport model, Community Multiscale Air Quality Model (CMAQ), to estimate the concentrations from emissions not originating in the model domain and provide source specific boundary concentrations to the small-scale regulatory model (AERMOD).The details about the physical and chemical setting of the chemical transport model used are available in Kota et al. (2018) and Guo et al. (2018) and are not repeated here.
The basic input requirements of AERMOD can be broadly classified into meteorological data, emission data, and pollutant background data.AERMOD tackles the meteorological data and the terrain data by separate pre-processors called AERMET and AERMAP respectively.AERMET processes the raw meteorological data and generates the parameters required by AERMOD interface to carry out the calculation for the output concentration.The terrain pre-processor AERMAP characterizes terrain elevation and is used to produce source/receptor heights and elevation contours.The required meteorological input files in AERMOD-ready format were generated using the Weather Research Forecasting (WRF) model version 3.8.1, a Eulerian-based model, with initial and boundary conditions from FNL (Final) Operational Global Analysis data on 1.0 × 1.0 degree grids from NCAR for every six hours (http://dss.ucar.edu/datasets/ds083.2/).The wind speed, wind direction, temperature, and PM 2.5 concentration data for the years 2015-2018 at the 31 continuous ambient air quality monitoring stations in Delhi was obtained from CPCB's online data dissemination portal (https://app.cpcbccr.com/ccr/#/caaqm-dashboard-all/caaqmlanding).
The 400 m × 400 m gridded source-specific emission inventory for industry, residential, transport (heavy commercial vehicles, cars, three wheelers, buses, two wheelers and other transport), thermal power plants, paved and unpaved road dust, municipal solid waste burning, brick kilns and others (crop burning, crematoriums and miscellaneous) for a domain of 65 km × 70 km (4550 km 2 area) covering Delhi and its adjacent region by the SAFAR-Indian Ministry of Earth Sciences for the year 2018 (IITM-MoES) (Beig et al., 2018) was used to prepare emissions.Emissions from construction was taken from the TERI emission inventory (ARAI and TERI, 2018), which provides gridded emissions at a grid resolution of 4 km × 4 km.The emission inventory is available in the .csvfile format with the emissions in gridded format and the units tons grid -1 month -1 .The emissions were converted into g m -2 s -1 and were read by AERMOD from the .csvfile.
The simulations were carried out for post-monsoon (October, November) and winter (December, January, and February) of the year 2018.A 'no intervention' scenario was simulated to estimate the PM 2.5 concentrations with the original emission inventory (no interventions applied).Further, combinations of interventions were simulated, and the results were compared with the 'no intervention' scenario to calculate the percentage reduction achieved during each intervention scenario.

Model Performance
The model performance was evaluated after running the 'no intervention' simulation for Delhi for the post-monsoon and winter season.To check the model performance, mean fractional bias (MFB) and mean fractional error (MFE) were calculated after comparing the model's predicted concentrations to the observed concentrations at various CPCB air quality monitoring stations (Fig. 2).The model performed well at 17 stations, with the MFB lying in the U.S. EPA prescribed limits of -0.6 to 0.6 and the MFE less than 0.75.Moderate performance was observed for five stations, and the model performed poorly to predict the concentrations at 9 out of 31 stations.However, overall, the model was able to predict the concentrations for most of the stations in Delhi satisfactorily.

Reduction from Interventions
NCAP targets 20-30% reduction in average PM 2.5 concentrations in non-attainment cities in India.However, considering that the relation between emission and concentrations isn't linear, estimating percentage reduction in total emissions across sectors is required to meet this NCAP target.Thus, to estimate this in Delhi, a range of total percentage reductions were explored, but the results pertaining to only 20% is discussed, as an example.

Effects of intervention scenarios on PM 2.5 concentrations during postmonsoon and winters
The study evaluated the percentage reductions achievable from six intervention scenarios against the base case where no interventions were applied.Scenarios of interventions across the entire study domain and additional interventions on selected hotspots were simulated.The emission-based hotspots in Delhi were identified based on the total emissions from different sectors in the IITM-MoES and TERI emission inventory.An area of 5 km × 5 km was considered as the hotspot, and the hotspot specific interventions were applied to these 5 km × 5 km grids.The identified emission-based hotspots across Delhi have been listed in Table S1.Six intervention scenarios were simulated for the post-monsoon and winter seasons and are listed below: i. Scenario1 -Total emissions across the study domain were reduced by 20%.ii.Scenario2 -Total emissions across the study domain were reduced by 20%.Additionally, the emissions from crop burning from the National Capital Region outside Delhi in the study domain were made zero.iii.Scenario3 -Total emissions across the study domain were reduced by 20%.Additionally, the emissions from the construction sector in the study domain were made zero.iv.Scenario4 -Total emissions across the study domain were reduced by 20%.Additionally, the emissions from unpaved road dust in the selected hotspots were made zero.v. Scenario5 -Total emissions across the study domain were reduced by 20%.Additionally, the emissions from industries in the selected hotspots were made zero.vi.Scenario6 -Total emissions across the study domain were reduced by 20%.Additionally, the emissions from industries and unpaved road dust in the selected hotspots were made zero.vii.Scenario7 -Total emissions across the study domain were reduced by 20%.Additionally, the emissions from industries, unpaved road dust, construction and crop burning (NCR region) in the selected hotspots were made zero.Fig. 3 shows the percentage reduction in the concentrations estimated by implementing these six interventions during the post-monsoon and winter seasons, when the concentrations of PM 2.5 are the highest in Delhi.
It was estimated that an average reduction of ~15.75% could be achieved by reducing the overall emissions across the entire study domain by 20% for post-monsoon in Scenario1.For, Scenario2, an average reduction of up to ~17.63% in the predicted PM 2.5 concentrations was estimated during the two fortnight periods.Scenario3 yielded a reduction of ~17.74% and ~17.28% during the first and the second fortnight, respectively.The estimated reduction in the PM 2.5 concentrations for Scenario4 was found to be slightly higher in the first fortnight (~17.91%)than the second fortnight (~17.04%).For the Scenario5, a significant average reduction of ~19.36% was observed during both fortnights of the post-monsoon season.However, for the Scenario6, the highest reduction of ~21.52% was observed during the first fortnight, while ~20.65% reduction was observed for the second fortnight.Finally, for Scenario7, where all the interventions in previous scenarios were implemented together, ~25.28% and ~24.17% reductions in the concentrations were observed for the first and the second fortnights during the post-monsoon.The results indicate that just mitigating crop burning and construction sectors emissions during the postmonsoon could improve the reductions further by ~2-3%.Similar effect was also observed due to the reduction in emissions of unpaved road dust in the hotspots.Industries influenced the reductions significantly, as up to ~4% increase in the reductions was achieved just by reducing the industrial emissions in the hotspots grids to zero.The reductions are justified as the source contributions of sources also reveal the same pattern (see Fig. S1).All the interventions together increased the reductions further by ~10%.
During winter, for Scenario1, the percentage reduction in the PM 2.5 concentration was found to be ~15.75% for both the fortnights, which is equal to that observed during post-monsoon.However, for Scenario2 and Scenario3, the percentage reduction was found to be slightly (~0.50% and ~0.69%, respectively) higher than that observed in post-monsoons.For the Scenario4, the percentage reductions observed was ~0.82% lower in the first fortnight and almost equal in the second fortnight to that observed in post-monsoon.An average reduction of around ~19.17% was observed for the Scenario5 during winters.However, for the Scenario6, an average percentage reduction of ~20.51% was observed during the two fortnights, which was slightly lower than that observed in the post-monsoon.For Scenario7, the highest reduction of ~25.43% was observed during the second fortnight, while ~25.26% reduction was observed for the first fortnight.For the seven scenarios in winter, reductions similar to those observed in post-monsoon season were observed.However, the minor changes in the reductions in certain scenarios may be attributed to the change in the meteorology during winters when compared to post-monsoon.
The results reveal that reducing emissions from seven hotspots alone could reduce the concentrations by up to ~9.28%, highlighting that the hotspots can play a crucial role in air quality management in Delhi.The results highlight those emissions from unpaved road dust and industries should be aggressively targeted to achieve maximum reduction.

Citywide vs. hotspot specific interventions
Implementing citywide interventions can be very tough due to the number of resources required, including economic, human, and technical resources.Thus, we compare the reductions in PM 2.5 concentrations estimated for the citywide and hotspot-specific interventions.For this, the effect of citywide versus hotspot-specific interventions was studied at three stations, i.e., Anand Vihar, Mandir Marg, and DU North Campus, as these stations were found to be in the top five hotspots for all the months in Winters.While for the citywide interventions, the emissions of the particular sources were reduced across the whole study domain, the emissions in an area of 5 km × 5 km, with the station at the centre, were reduced for the hotspot-specific interventions.
The overall reduction in the concentrations was higher across all three stations in the case of citywide interventions, as shown in Fig. 4. Interestingly, the difference in the reductions in PM 2.5 concentrations from citywide versus hotspot-specific interventions was observed to be the highest in DU North Campus (27.65%).The difference was of much smaller proportions in the case of Mandir Marg (17.35%) and Anand Vihar (10.70%).This variation in the reduction in the PM 2.5 concentrations at the three stations can be attributed to the influence of the local sources in the hotspots (5 km × 5 km grid).The local sources in the 5 km × 5 km hotspot grid at Anand Vihar and Mandir Marg contribute significantly to PM 2.5 concentrations in these hotspots.Thus, just reducing the emissions in these hotspots leads to a significant reduction in the PM 2.5 concentrations.However, in DU North Campus, the contributions to the PM 2.5 concentrations are mainly from sources located outside the hotspot grid.Thus, the reduction observed for hotspotspecific interventions is meagre compared to that in citywide interventions.
The 27.65% difference in the reductions of PM 2.5 concentrations from citywide versus hotspotspecific interventions at DU North Campus, indicates significant contributions from non-local sources, whereas a lesser 10.70% at Anand Vihar, suggests a substantial contribution from the local sources in the 5 km × 5 km hotspot grid, thus leading to small reductions at Anand Vihar.Thus, local, and non-local contributions to a source need to be analyzed before selecting hotspots.
Simulations were also carried out to understand the impact of interventions for different categories of stations.For this, three stations were considered, Jhangirpuri, NSIT Dwarka, and Wazirpur.Jahangirpuri was categorized as residential, NSIT Dwarka as commercial, and Wazirpur as an industrial station.Simulations were carried out where the emissions from HCVs were reduced by 50%, MSW burning was reduced by 100%, unpaved road dust by 100%, and industry by 100% in the hotspots identified in Table S1 for 15-28 February 2018.Fig. 5 shows the percentage reduction for the intervention scenarios at the three selected stations.It was observed that for a 50% reduction in the emissions from HCVs, the maximum reduction (4.90%) was observed at the commercial station, NSIT Dwarka.Reduction of MSW emissions by 100% in the hotspots reduced the concentrations by only 2.6 and 2.2% at the residential and commercial stations.However, reducing unpaved road dust emissions by 100% in the hotspots reduced concentrations in the residential station by 10.80% and nearly half of that (4.8%) at the commercial station.
Interestingly, the industrial station recorded a negligible reduction in the concentrations of PM 2.5 for the first three hotspot-specific intervention scenarios.However, for the intervention scenario where the emissions from industries in the hotspots were made zero, a very significant reduction of ~56.2% was observed at the industrial station.As per the source contributions for the entire study period, as shown in Fig. S1, it can be observed that the percentage source contribution from industries is one of the highest at Wazirpur.Significant reductions of ~23% and ~30.60% were observed for the residential and commercial stations.Thus, significant reductions were observed at Wazirpur when the emissions from industries were reduced in the hotspot grid for all three stations, as industries contribute significantly at all three stations.The analysis thus stresses that categorizing hotspots based on dominant sources can further increase the reductions observed in the PM 2.5 concentrations.Similar analysis in future can be used to predict possible source specific hotspots in advance to optimize the reductions achievable in PM 2.5 concentrations.Thus, hotspot-specific interventions can be beneficial in reducing the concentrations at specific hotspots as they are easier to implement than citywide interventions.Policymakers should carefully consider local and non-local contributions from regions outside the hotspot grid when selecting hotspots.

Complaints based interventions
CPCB has an online portal (https://cpcb.nic.in/query-form1.php)for reporting air pollution complaints in Delhi NCR.Any citizen in Delhi can use this portal to report any incidents of noncompliance.Data of such complaints of non-compliance from various areas of Delhi was collected for December 2018.Hotspots associated with municipal solid waste burning, transport congestions and construction and demolition activities were identified using the data of district-wise complaints received and listed in Table S2.Simulations were carried out to simulate the scenarios where the citywide emissions were reduced by 20% and the emissions of MSW burning, construction, industry, HCVs, and unpaved road dust were made zero at the complaints based hotspots.The results showed that an average reduction of 27.91% was observed when emissions of MSW, construction, industry, HCVs, and unpaved road dust were made zero at the complaint-based hotspots along with a 20% citywide reduction of emissions from all sources.A maximum reduction of around ~45.33% was observed for Nehru Nagar station.An average reduction of 12.16% was estimated across the study domain by reducing the emissions of MSW burning and construction to zero from all the complaint-based hotspots.Thus, just interventions based on complaints-based hotspots can also yield significant reductions in the PM 2.5 concentrations in Delhi.Long term analysis of locations associated with frequent complaints can help policymakers identify prospective hotspot regions and devise necessary hotspot specific plans.

Odd-even Effect
As part of the effort to bring down the concentrations of PM 2.5 in Delhi, the Delhi Government implemented the vehicle rationing scheme or the odd-even scheme in the year 2016 (GNCTD, 2015).According to this scheme, vehicles with number plates ending with even numbers were allowed to run on even dates, while those with odd were allowed to run only on odd dates.However, vehicles of VIPs, emergency vehicles such as ambulance, fire brigade, enforcement vehicles, vehicles running on CNG, women-only vehicles, and vehicles carrying children in school uniforms were exempted from the odd-even rule.To evaluate the reduction in PM 2.5 by implementing the odd-even rule, we simulated the concentrations by reducing the emissions from cars by 50%, assuming that half of the registered cars in Delhi have number plates ending with even or odd numbers.The emissions from taxis running on gasoline were included in the simulated concentrations.Fig. 6 shows the percentage reduction in the PM 2.5 concentrations observed at various stations after implementing the odd-even scheme.Maximum reduction was observed at Nehru Nagar with a reduction of 6.55% in the predicted concentrations from the base case scenario (without any reductions in emissions).An average reduction of ~4.39% was observed for the entire study domain for the odd-even intervention scenario.Chowdhury et al. (2017) also observed only 4-6% reduction in PM 2.5 concentrations by the traffic restrictions during the odd-even policy in January (1-15) 2016.Similarly, Mohan et al. (2017) also concluded that there were no improvements in the levels of concentration of PM 2.5 .CPCB, in a report (CPCB, 2016), concluded that the oddeven scheme was not able to reduce the air pollution levels in Delhi substantially.
Thus, the odd-even rule alone is insufficient for reducing the PM 2.5 concentrations by a significant amount.The results also suggest that stakeholders must focus on source and hotspot-specific interventions concurrently with other feasible interventions like the odd-even rule to significantly reduce ambient PM 2.5 concentrations.

Strategies for Top Ten Hotspots in Post-monsoon and Winter
For post-monsoon and winter, the top ten hotspots were identified based on the average concentrations observed over the past three years.Simulations were carried out to study the changes in the concentrations of the PM 2.5 at these hotspots when two separate intervention scenarios were used.In the first intervention scenario (Case1), the emissions for all the sources were reduced by 20% across the entire study domain, and the emissions from industry and unpaved road dust from the seven identified hotspots (Table S1) were made zero.For Case2, the emissions from construction and crop burning (from NCR regions in the study domain) were made zero in the seven hotspots apart from the interventions in Case1.This analysis aimed to identify the intervention scenarios that could help reduce the concentrations by around 25% in these hotspots by applying interventions on a minimum number of sources.The percentage reductions observed at the hotspots have been presented in Fig. 7.
During 15-31 October, the top hotspot was observed to be Rohini.For most hotspots, a minimum of 20% reduction in the PM 2.5 could be achieved by Case1 interventions.However, Case2 could reduce the concentrations in Rohini, Jahangirpuri, Anand Vihar, Narela, and Nehru Nagar by more than 30%.Thus, significant reductions could be achieved by implementing Case2 interventions for October 15-31 to reduce the concentrations in the top 4 hotspots in Delhi by more than 30%.
During November 1-15, DTU, Anand Vihar, and Wazirpur were observed as the top hotspots.An average reduction of ~19.33, ~23.73, and ~17.82% was estimated for these three hotspots in the case of Case1.However, for Case2, the average reduction was found to be ~24.45,~25.69, and ~17.97%.For Lodhi Road, Ashok Vihar, and Mandir Marg, the reduction in Case2 was estimated to be greater than 30%.Thus, intervention Case2 could effectively reduce the PM 2.5 concentration by more than 25% for about 5 of the top ten hotspots.
During November 15-30, more than 15% reduction in the PM 2.5 concentrations was observed across all ten hotspots for the Case1.A reduction of ~35% was observed for Ashok Vihar and Nehru Nagar for Case1.For Case2, a reduction of more than 20% was observed for all hotspots.For Case2, Ashok Vihar and Nehru Nagar saw a reduction of more than 35%.The top hotspot, DTU, also saw a reduction of ~27.60% for the Case2 interventions.However, for Wazirpur and Punjabi Bagh, the reductions in Case2 barely increased from Case1.
During December 1-15, a more than 25% reduction was observed at five hotspots for Case1.Nehru Nagar, which was the top hotspot, saw a reduction of ~43.59% for Case1.For Case2, the concentrations of PM 2.5 reduced by more than 25% at seven hotspots.Like November, the percentage reduction at Wazirpur and Punjabi Bagh sites was found to be similar to that observed for the Case1.For December 15-31, a reduction of more than 20% was observed at five hotspot locations for Case1 interventions.Among the top hotspots, the maximum reduction for Case1 was observed at Nehru Nagar, where the concentrations reduced by 44.5% for Scenario1 and 57.37% for Case2.Ashok Vihar also saw significant reductions of up to 53.12% for Case2 interventions.Six hotspots saw a reduction of more than 30% for Case2 intervention.
Thus, it can be concluded that significant reductions of up to 30% can be achieved in the case of Case2 interventions for most of the stations.However, to achieve over 30% reductions in all the top ten hotspots, additional interventions such as reducing emissions from HCVs and MSW burning to zero in the hotspots may be required.Apart from this, policies such as the odd-even rule implemented along with the Case2 interventions can further decrease the concentrations.These results can further be used by policymakers to make appropriate graded responses or update the current GRAP.

CONCLUSIONS
The study attempts to identify the strategies that can be useful in regulating the PM 2.5 concentrations over Delhi.The impact of implementing different intervention scenarios for the post-monsoon and winter seasons when Delhi experiences the highest PM 2.5 concentrations were evaluated.
• A reduction of up to ~15% in the total PM 2.5 concentrations is estimated by reducing the total emissions by 20% in the entire study domain and an additional ~10% more reduction could be achieved if emissions of industry, unpaved road dust, construction and crop burning (NCR region) in the selected emission-based hotspots for the post-monsoon and winter seasons are reduced by 100%.Therefore, the contributions from non-local sources must be assessed when identifying the hotspots, as this can help in better estimation of the implementation of the mitigation strategies.
• It was concluded that up to 28% reduction in the PM 2.5 concentrations could be achieved at the top ten observational hotspot by reducing the emissions from all the sources by 20% across the entire study domain, along with a 100% reduction of the emissions from industries, unpaved road dust, crop burning, and construction in the seven identified emission-based hotspots.
• By reducing emissions from construction and MSW burning by 100% at the locations determined based on compliance complaints received at the Central Pollution Control Board's online complaint portal, an average reduction of 12% in the PM 2.5 concentrations was estimated.• The study also highlights that previously implemented interventions such as the odd-even rule may be inadequate by themselves in regulating the PM 2.5 concentrations in Delhi.Thus, for effective regulation of PM 2.5 concentrations in Delhi, there is an urgent need to develop the source and location-specific intervention strategies for Delhi.A decision support mechanism based on the methods used in this study can be used to devise hotspot specific action plans.The data generated in this study can be further used to create a suitable regulation mechanism, update existing mitigation plans and devise long-term strategies to address the high PM 2.5 concentrations in Delhi.

Fig. 1 .
Fig. 1.Flowchart showing the methodology used for this study.

Fig. 3 .
Fig. 3. Bar plots showing estimated percentage reduction in PM 2.5 concentrations for different intervention scenarios for postmonsoon and winter for the study domain.

Fig. 4 .
Fig. 4. Percentage reduction in PM 2.5 concentrations for citywide vs. hotspot specific intervention scenarios at three CPCB observation stations.(Intervention scenarios: HCV: Reducing the emissions from heavy commercial vehicles by 100%; Unpaved: Reducing the emission from unpaved road dust by 100%; HCV & Unpaved: Reducing the emissions from heavy commercial vehicles and unpaved road dust by 100%; MSW: Reducing the emissions from municipal solid waste burning by 100% and Cars -Reducing the emissions from cars by 50%.)

Fig. 5 .
Fig. 5. Difference in the reduction in PM 2.5 concentrations for four intervention scenarios at residential, commercial, and industrial CPCB observation stations (Intervention scenarios: HCV_50: Emissions from HCVs reduced by 50% in the hotspots; MSW_100: Emissions from MSW burning reduced by 100% in the hotspots; Unpaved_100: Emissions from Unpaved Road dust reduced by 100% in the hotspots and Industry_100: Emissions from industry reduced by 100% in the hotspots.)

Fig. 6 .
Fig. 6.Percentage reduction in PM 2.5 concentrations at 31 CPCB observation stations for Odd-Even scheme intervention scenario.

Fig. 7 .
Fig. 7. Percentage reduction in the PM 2.5 concentrations at the top 10 observational hotspots for two intervention scenarios (Case1 and Case2).

Table 1 .
Model setup, configuration, and data sources.
• It is observed that for certain hotspots, the percentage reductions in the concentrations of PM 2.5 by only implementing hotspot specific interventions are similar to those estimated for citywide interventions.The difference in the reductions from citywide versus hotspot-specific interventions was observed to be the highest in DU North Campus (28%) compared to Mandir Marg (17%) and Anand Vihar (11%), emphasizing that local sources at Anand Vihar and Mandir Marg contribute more to PM 2.5 concentrations, compared to DU North Campus.