Jahangirnagar University Journal of Science

Spatiotemporal Analysis and Machine Learning-Based Profiling of Atmospheric Pollutants Using Sentinel-5P TROPOMI Satellite Data at Tier-Two Administrative Units in Bangladesh

Wed, 12 Aug 2026 00:00:00 +0600

Air pollution poses a critical environmental and public health threat in Bangladesh, yet a comprehensive, multi-pollutant analysis at a high administrative resolution has been lacking. This study bridges this gap by leveraging the high-resolution capabilities of Sentinel-5P TROPOMI satellite data and an unsupervised machine learning framework to conduct a district-level assessment of five key atmospheric pollutants—SO₂, NO₂, O₃, HCHO, and CH₄—from 2020 to 2024. Annual mean concentrations were extracted for all 64 districts using Google Earth Engine, followed by rigorous spatiotemporal trend analysis using the Mann-Kendall test. The core novelty of this research lies in the application of K-means clustering to classify districts based on their holistic, multi-annual pollutant profiles.

The analysis revealed distinct national trends: a significant increase in O₃ (54 districts) and CH₄ (61 districts), a volatile trend for SO₂, and a notable decrease in NO₂ in 21 urban districts. Critically, the K-means algorithm (k=5) segmented the country into five discrete pollution clusters, ranked from best to worst: 1) Cleanest (16 districts, e.g., coastal and hilly regions), 2) Moderate (7 districts, e.g., northeastern regions), 3) High O₃ & CH₄ (16 northern agricultural districts), 4) High Multi-Pollutant (23 southwestern and central industrial districts), and 5) Extreme Urban/Industrial (4 districts: Dhaka, Gazipur, Munshiganj, Narayanganj). This final cluster exhibited an extreme concentration of NO₂ and CH₄, over 50% higher than the national average.

The findings demonstrate that Bangladesh's air pollution landscape is not monolithic but a mosaic of distinct regional typologies, each driven by different dominant sources—from vehicular and industrial combustion in urban centers to agricultural emissions in the north. This study provides a novel, data-driven framework for "pollution zonation," moving beyond one-size-fits-all policies towards targeted, cluster-specific mitigation strategies. The methodology establishes a replicable model for evidence-based air quality management in data-scarce regions globally.

Comparative Study of ARAKAWA Scheme and WENO Scheme for Solving Advection Equation.

Prokriti Biswas — Wed, 12 Aug 2026 00:00:00 +0600

In this paper, we investigate the impact of various advection schemes in numerical ocean modelling. Numerical simulation in ocean models rely on the non- oscillatory advection scheme, which minimizes numerical dissipation and dispersion problems. We study the Arakawa scheme and WENO-5 scheme in the 2-D advection problem. Our key focus is to observe the oscillatory behavior of the numerical solutions, TVD property and numerical dissipation and dispersion these schemes. We observe that the Arakawa scheme shows an oscillatory solution with slight dissipation, yet preserves energy well. The WENO-5 scheme produces a non- oscillatory numerical solution and minimizes numerical dissipation in 2-D advection equations. The WENO-5 scheme performs better than other advection schemes and the WENO-5 reconstruction minimizes numerical dissipation.

Combination Models for Accurate Thunderstorm Forecasting in Bangladesh's Low-Risk Zone

Sharmin Akther, Mohammad Mahboob Hussain Khan, Rumana Rois — Wed, 12 Aug 2026 00:00:00 +0600

Thunderstorms can devastate buildings, crops and human life even in "low risk" areas like Sitakunda, Bangladesh. This conundrum underscores the need for timely and accurate forecasting to safeguard vulnerable, resource-poor populations. But traditional models often struggle to account for the nonlinear and seasonality of thunderstorms, and little research has been done on predicting the daily occurrence of thunderstorms in low-risk regions. This research seeks to enhance thunderstorm frequency forecasting by extensively evaluating and comparing statistical, machine learning, and deep learning approaches (ETS, ARIMA, STL-ETS, STL-ARIMA, NNAR, TBATS, Prophet, GARCH, SVM, and LSTM). An innovatively combined model is also developed. Applying daily thunderstorm counts from Sitakunda (1981-2023) and measuring performance using RMSE and MAE, the findings show the combination model achieves the lowest prediction errors and most reliable forecasts. Although Prophet and SVM show comparable results, GARCH has the lowest RMSE (1.3137). By contrast, LSTM and NNAR exhibit larger errors, suggesting that they struggle to model the unpredictable nature of thunderstorms. These results demonstrate the benefits of combination forecasting in improving thunderstorm predictions for low-risk areas. Incorporation of these models into early warning systems and local governance can enhance preparedness, minimize disruptions and injuries, and foster resilience to thunderstorm risks in the long term.