Seasonal Air Pollution Patterns in an Assam Monitoring Dataset for 2024
Subhalakhmi Changmai *
Department of Statistics, Jagannath Barooah University, Barpatra Ali, Jorhat 785001, Assam, India.
*Author to whom correspondence should be addressed.
Abstract
Aims: To describe seasonal variation in four ambient air pollutants and examine pollutant covariance and seasonal discrimination in a compiled 2024 Assam monitoring dataset.
Study Design: A retrospective exploratory analysis of monthly station-level environmental monitoring observations.
Place and Duration of Study: Thirty monitoring stations across 13 locations in Assam, January–December 2024.
Methodology: The analysis included 360 complete station-month observations of sulphur dioxide, nitrogen dioxide, PM₁₀ and PM₂.₅. Descriptive statistics characterised pollutant distributions. Principal axis factoring examined the correlation structure, and Friedman tests compared seasonal means within stations. Linear discriminant analysis used three pollutants with equal seasonal priors. Classification was assessed in-sample and using leave-one-record-out and leave-one-station-out validation.
Results: The Kaiser–Meyer–Olkin measure was 0.663 and Bartlett’s statistic was χ²(6) = 1645.66. A single factor accounted for 62.73% of total standardised variance, with particulate loadings of 0.967 and 0.954. Mean PM₁₀ and PM₂.₅ concentrations in winter were 114.74 and 55.41 µg/m³, respectively, compared with 47.39 and 22.67 µg/m³ during monsoon. Friedman statistics ranged from 34.60 to 84.28 (all P < .001). The first discriminant function accounted for 97.14% of between-group discrimination. Classification accuracy was 52.5% in-sample, 48.9% under leave-one-record-out validation and 50.8% under leave-one-station-out validation. Equality of seasonal covariance matrices was rejected.
Conclusion: Particulate concentrations displayed a consistent winter–monsoon contrast, but seasonal pollutant profiles overlapped substantially. Findings describe this compiled network-year and require source-record verification and multi-year validation before operational application.
Keywords: Assam, ambient air pollution, particulate matter, principal axis factoring, discriminant analysis, seasonal variation, grouped cross-validation