Seasonal and Temporal Patterns of Urban Air Pollutants and Their Differential Influence on the Air Quality Index: A Comparative Empirical Study

Seasonal and Temporal Patterns of Urban Air Pollutants and Their Differential Influence on the Air Quality Index: A Comparative Empirical Study

Moses Nnanna Ibe

Department of Data Analytics and Technology University of Greater Manchester Manchester, United Kingdom

Chinoso Job

Department of Data Analytics and Technology University of Greater Manchester Manchester, United Kingdom

Onwe, Festus Chijioke

Information Technology Department, University of Port Harcourt, Rivers State, Nigeria.

Abstract – Urban air pollution constitutes one of the foremost environmental threats to public health globally, with the World Health Organisation estimating approximately 7 million annual deaths attributable to outdoor and household air pollution. This study presents a comparative empirical analysis of the seasonal and temporal dynamics of six key air pollutantssulphur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), ozone (O3), and particulate matter (PM2.5 and PM10)and their differential influence on the Air Quality Index (AQI), using a four-year dataset (20212024) processed via Apache PySpark on Google Colaboratory. Through exploratory data analysis, seasonal decomposition, Kruskal-Wallis statistical testing, and Spearman correlation analysis, the study addresses four empirical research questions concerning holiday-period emission reductions, seasonal AQI variation, pollutant-specific seasonal patterns, and pollutant- AQI correlation strength. Results reveal: (i) no statistically significant difference in SO2, NO2, or CO concentrations between holiday and non-holiday periods, suggesting structural rather than behavioural emission sources; (ii) significant winter-season AQI elevation (mean AQI 275), substantially exceeding summer values ( 100); (iii) PM2.5 and PM10 as the dominant AQI drivers (Spearman = 0.91 and 0.93 respectively); and (iv) contrarian ozone behaviour, peaking in summer via photochemical production. Findings carry significant implications for targeted seasonal pollution intervention policy and business operational planning in environments subject to AQI variability.

Index Terms – urban air quality, AQI, seasonal variation, PM2.5, PM10, PySpark, big data analytics, air pollution, NO2, ozone, public health

1. INTRODUCTION

   Urban air quality has emerged as a critical determinant of public health outcomes, economic productivity, and environmental sustainability in contemporary cities. The compound health burden of ambient air pollutionestimated by the Global Burden of Disease study at over 8 million deaths in 2021, making it the second leading environmental risk factor

   globally [1]demands rigorous, data-driven characterisation of pollutant dynamics to support evidence-based intervention.

   In the United Kingdom, air pollution accounts for an estimated 30,00035,000 premature deaths annually, with populations in urban corridors near high-traffic roads and industrial sites facing disproportionate exposure [2]. While regulatory progress has been madesulphur dioxide levels in major UK cities have declined substantially following coal phase-out legislation [3]NO2 and PM2.5 concentrations continue to exceed World Health Organisation (WHO) guideline limits in numerous urban areas.

   Air quality management requires granular understanding of the temporal, seasonal, and source-specific determinants of pollutant concentrations and their composite expression in the Air Quality Index (AQI). The AQI translates complex, multivariate pollutant data into a standardised public health risk indicator across a hazard spectrum from Good (050) to Hazardous (>300), based on U.S. Environmental Protection Agency (EPA) established thresholds. However, the differential contributions of individual pollutants to AQI valuesand how these contributions vary across seasons, holiday periods, and atmospheric conditionsremain incompletely characterised in the literature.

   This study addresses the following empirical research questions:

   RQ1. Are levels of SO2, NO2, and CO significantly lower during holiday periods, suggesting reduced industrial or transport activity?

   RQ2. Does urban AQI vary significantly across the four seasons (winter, spring, summer, autumn)?

   RQ3. How do key pollutant concentrations (PM2.5, PM10, NO2, O3, SO2, CO) vary seasonally, and what are the policy implications?

   RQ4. Which pollutant is most strongly associated with higher AQI, and thus poor urban air quality?

2. BACKGROUND AND RELATED WORK

   1. Air Pollution and Public Health

      The health consequences of urban air pollution are welldocumented. Long-term exposure to PM2.5 is causally associated with cardiopulmonary mortality, lung cancer, and adverse pregnancy outcomes [4]. NO2 exposure is linked to respiratory inflammation and increased asthma incidence, while CO reduces blood oxygen-carrying capacity. Ground- level ozonea secondary pollutant formed through photochemical reactions between NOx and volatile organic compounds in sunlightexacerbates respiratory conditions and impairs lung function [5].

   2. Seasonal Dynamics in Urban Pollution

      Seasonal variation in pollutant concentrations arises from the interplay of emission source variability, atmospheric chemistry, and meteorological conditions. Winter-elevated PM and NO2 concentrations reflect increased residential and industrial heating demand combined with atmospheric temperature inversions that trap pollutants near the surface [6]. Summer ozone elevation reflects photochemical production driven by intensified solar radiationa mechanism contrarian to most primary pollutant patterns.

   3. AQI as a Composite Indicator

   The AQI aggregates individual pollutant concentration data through pollutant-specific breakpoint functions, taking the maximum sub-index value across pollutants as the reported AQI. This design means that understanding AQI dynamics requires understanding the dominant pollutant contributor, which may shift across seasons. Previous studies have predominantly analysed single pollutants in isolation; comparative multipollutant approaches linking seasonal dynamics to AQI are underrepresented in the literature.

3. METHODOLOGY

   1. Research Design

      This study employs a quantitative, observational, correlational design to characterise seasonal and temporal patterns in urban air pollutant concentrations and their associations with AQI. The observational design leverages existing monitoring data from regulatory agencies, avoiding the sampling biases of self-reported data [7].

   2. Dataset

      The dataset comprises daily air quality observations from publicly available monitoring stations maintained by regional Environmental Protection Agencies and the Global Air Quality Index Repository, covering the period 1 January 2021 to 31 December 2024 (1,461 daily records). Variables include AQI,

      AQI category, and concentrations of PM2.5, PM10, NO2, O3, SO2, and CO in µg/m3. A binary holiday indicator marks UK public holiday dates.

   3. Data Preprocessing

      Data preprocessing was implemented in Apache PySpark on Google Colaboratory (Python 3.10). The processing pipeline comprised:

      1. Missing value detection and imputation: Missing values were identified using PySparks describe() and null- count functions. Median imputation was applied to continuous pollutant vaiables to avoid mean bias from outlier influence.

      2. Outlier detection and validation: Extreme values were flagged using the Interquartile Range (IQR) method (values < Q1 1.5×IQR or > Q3 + 1.5×IQR) and crossreferenced against EPA established AQI breakpoints to distinguish genuine extreme pollution events from data errors.

      3. Seasonal classification: Observations were categorised into four seasons based on date: Winter (December February), Spring (MarchMay), Summer (June August), and Autumn (SeptemberNovember).

      4. Holiday classification: Dates were labelled as Holiday or Non-Holiday using an authoritative UK public holiday calendar.

   4. Variable Definitions

      Dependent variable: AQIa composite indicator of air pollution severity, computed from the maximum sub-index value across all pollutants at each observation, following U.S. EPA breakpoint functions.

      Independent variables:

      • Pollutant concentrations (PM2.5, PM10, NO2, O3, SO2, CO) in µg/m3;

      • Season (categorical: Winter, Spring, Summer, Autumn);

      • Holiday marker (binary: Holiday / Non-Holiday).

   5. Analytical Techniques

   Descriptive analysis: Mean, median, and standard deviation of pollutant concentrations by season and holiday status.

   Inferential testing: Kruskal-Wallis H-test (non-parametric one-way ANOVA) for seasonal differences in AQI and pollutant concentrations; Mann-Whitney U-test for holiday/nonholiday comparisons. Non-parametric tests were selected given that Shapiro-Wilk normality tests rejected normality for all pollutant distributions (p < 0.001).

   Correlation analysis: Spearman rank correlation coefficients () between individual pollutant concentrations and AQI,

   providing a non-parametric measure of monotonic association strength.

   Visualisation: Grouped bar charts, seasonal line charts, and Spearman correlation heatmaps, implemented using PySpark DataFrames exported to Matplotlib/Seaborn.

4. RESULTS

   A. RQ1: Holiday vs. Non-Holiday Pollutant Concentrations

   Mann-Whitney U-tests revealed no statistically significant difference in SO2 (U = 48312, p = 0.43), NO2 (p = 0.31), or CO (p = 0.67) concentrations between holiday and nonholiday periods (Fig. 1). This finding suggests that these pollutants originate predominantly from stable structural emission sources (industrial processes, continuous traffic patterns, energy generation infrastructure) rather than from activity patterns that shift during public holidays.

   10

   with Bonferroni correction showed that all pairwise seasonal comparisons were significant (p < 0.001) except Spring vs. Autumn (p =0.18). Winter AQI (mean 275) falls in the Very Unhealthy category on the U.S. EPA AQI scale, while summer values ( 100) approach the Moderate threshold.

   C. RQ3: Seasonal Patterns in Key Pollutants

   Fig. 3 reveals distinct seasonal profiles:

   • PM2.5 and PM10: Peak in winter, lowest in summer, reflecting seasonal combustion (heating) demand and atmospheric dispersion conditions.

   • NO2: Winter-dominant, driven by energy-intensive heating and reduced atmospheric mixing.

   • O3: Contrarian summer peak, driven by photochemical reactions between NOx and hydrocarbons under intense solar radiation [5].

   • SO2 and CO: Relatively stable across seasons (seasonal variation < 15%), consistent with stable industrial point sources unaffected by seasonal demand variation.

   Fig. 1. Mean SO2, NO2, and CO concentrations during holiday versus nonholiday periods. Differences are not statistically significant (Mann-Whitney p > 0.05 for all three pollutants).

   Fig. 3. Seasonal mean concentrations of key urban air

   pollutants. PM2.5, PM10, and NO2 peak in winter; O3 peaks in summer. SO2 and CO remain relatively constant.

   D. RQ4: Pollutant-AQI Correlation Analysis

   Fig. 2. Mean AQI across four seasons. Winter exhibits the highest mean AQI (275, classified as Very Unhealthy), while summer records the lowest (100, classified as Moderate). Kruskal-Wallis H =312.4, p < 0.001.

   B. RQ2: Seasonal Variation in AQI

   Kruskal-Wallis testing confirmed highly significant seasonal differences in AQI (H =312.4, p < 0.001). Post-hoc Dunns test

   Fig. 4. Spearman rank correlation coefficients between individual air pollutants and AQI. PM10 and PM2.5 show the strongest positive associations ( 0.93 and 0.91). O3 shows a weak negative correlation.

   Spearman correlation analysis (Fig. 4) identifies particulate matter as the dominant AQI driver:

   • PM10: = 0.93 (p < 0.001) very strong positive association;

   • PM2.5: = 0.91 (p < 0.001) very strong positive association;

   • CO: = 0.71 (p < 0.001) moderate positive association;

   • NO2: =0.32 (p < 0.001) weak positive association; O3: =0.15 (p < 0.01) weak negative association (summer ozone elevation coincides with lower overall AQI);

   • SO2: =0.08 (p =0.09) negligible, non-significant association.

5. DISCUSSION

   1. RQ1: Structural Emission Sources

      The absence of significant holiday-period emission reductions for SO2, NO2, and CO challenges policy assumptions that short-term activity reductions (e.g., lockdowns, public holidays) meaningfully reduce urban air pollution. This finding aligns with observations from the COVID-19 lockdown period, where NO2 reductions were temporally limited and rebounded rapidly upon activity resumption [11]. The policy implication is that emission reduction strategies must target structural sourcesfuel composition standards, industrial process retrofits, continuous traffic managementrather than relying on voluntary or episodic activity reductions.

   2. RQ2 and RQ3: Winter Pollution Challenge

      The pronounced winter AQI elevation, driven predominantly by PM2.5, PM10, and NO2, identifies winter as the critical intervention period for urban air quality management. This pattern is consistent across the four-year observation period, confirming that winter pollution represents a systematic structural challenge rather than an episodic anomaly. Policy responses should include: winter-specific emission control regulations; real-time PM monitoring and public communication systems; targeted restrictions on high-emission heating fuels during temperature inversion events; and incentive programmes for district heating and heat pump adoption as alternatives to combustion heating.

   3. RQ4: Particulate Matter as the Primary AQI Driver

   The very strong correlation between PM2.5, PM10, and AQI confirms particulate matter as the dominant determinant of urban air quality standards exceedances. This finding has direct implications for regulatory prioritisation: interventions targeting PM reduction (low-emission zones, vehicle emission standards, construction dust management, wood-burning stove restrictions) offer the greatest potential for AQI improvement per unit of intervention effort. The moderate CO-AQI

   correlation suggests that combustion-related pollution source contribute substantially to AQI through multiple pathways.

   The negative O3-AQI correlation is counter-intuitive but explicable: summer ozone episodes occur during meteorological conditions (high pressure, low wind) that reduce PM and NO2 concentrations, thus paradoxically coinciding with moderate AQI values despite ozone exceedances. This pattern suggests that the standard AQI calculation may inadequately capture ozone health risks in summer, warranting regulatory review of AQI communication methodology for seasonal ozone exposure.

6. LIMITATIONS AND FUTURE WORK

   This study has three principal limitations. First, the dataset represents a single urban monitoring station; spatial heterogeneity in urban pollution patterns may limit generalisability to other urban contexts. Second, the analysis is observationalcausal claims about emission source contributions require complementary receptor modelling and source apportionment studies. Third, meteorological variables (wind speed, temperature, precipitation, boundary layer height) were not included in the analytical model; future work should incorporate atmospheric conditions as covariates to isolate anthropogenic emission effects from meteorological variation. Future research directions include: spatial analysis of pollutant distribution hotspots within urban areas; predictive modelling of AQI exceedance events 2448 hours in advance using ML algorithms; and longitudinal analysis of AQI trends

   under progressive emission reduction policies.

7. CONCLUSION

This study has provided a rigorous empirical characterisation of seasonal and temporal patterns in urban air pollutant concentrations and their differential influence on the AQI using a four-year dataset processed in Apache PySpark. Key findings are: (i) holiday-period activity reductions do not significantly reduce combustion-related pollutant concentrations, confirming structural emission dominance; (ii) winter AQI is substantially elevated relative to other seasons, driven by increased PM and NO2 from combustion heating; (iii) PM2.5 and PM10 are the dominant AQI drivers, with Spearman exceeding 0.90; and (iv) ozone exhibits contrarian summer behaviour driven by photochemical production.

These findings provide an empirical evidence base for targeted seasonal pollution intervention policy, with implications for energy policy, urban planning, and public health communication.

REFERENCES

1. A. Jakubowska, The burden of air pollution: A perspective on global health inequalities, Pollutants, vol. 3, no. 3, pp. 419436, 2023.

2. H. L. Macintyre et al., Impacts of emissions policies on future UK mortality burdens associated with air pollution, Environ. Int., vol. 174, p. 107862, 2023.

3. P. Walk and N. Stognief, From coal phase-out to net zero: Driving factors of UK climate policy, Environ. Sci. Policy, vol. 138, pp. 7684, 2022.

4. Q. Di et al., Association of short-term exposure to air pollution with mortality in older adults, JAMA, vol. 318, no. 24, pp. 24462456, 2017.

5. P. Vongelis et al., Air pollution and effects of tropospheric ozone (O3) on public health, Int. J. Environ. Res. Public Health, vol. 22, no. 5, pp. 709713, 2025.

6. F. Petracchini et al., Influence of transport from urban sources and domestic biomass combustion on the air quality of a mountain area, Environ. Sci., vol. 24, no. 5,

   pp. 47414754, 2017.

7. R. Shrestha et al., Environmental health related socio- spatial inequalities: Identifying hotspots of environmental burdens and social vulnerability, Int. J. Environ. Res. Public Health, vol. 13, no. 7, pp. 691695, 2016.

8. Global Change Data Lab, Number of deaths by risk factor, Our World in Data, 2021. [Online]. Available: https://ourworldindata.org/grapher/ number-of-deaths- by-risk-factor

9. S. Babatola, Global burden of diseases attributable to air pollution, J. Public Health Africa, vol. 9, no. 3, pp. 813 817, 2018.

10. G. Singh, S. Rai, and N. Jadon, Major ambient air pollutants and toxicity exposure on human health and their respiratory system: A review, J. Environ. Manag. Tourism, vol. 7, no. 55, pp. 17741788, 2021.

11. J. Domingo and J. Rovira, Effects of air pollutants on the transmission and severity of respiratory viral infections, Environ. Res., vol. 187, pp. 109650109652, 2020.

12. A. Raza, Air quality under a changing climate: Trends and implications for respiratory diseases, J. Environ. Sci. Health, vol. 1, no. 1, pp. 112, 2025.

13. E. De Angelis et al., Low emission road transport scenarios: An integrated assessment of energy demand, air quality, GHG emissions, and costs, IEEE Trans. Autom. Sci. Eng., vol. 19, no. 1, pp. 3747, 2021.