Zainab Tahir

Air pollution caused by fine particulate matter (PM_2.5) poses a serious public health threat in many South Asian megacities where monitoring networks remain limited. Lahore, Pakistan—frequently ranked among the world’s most polluted cities—still lacks reliable short-term PM_2.5 forecasting systems. This study develops a performance-weighted ensemble machine learning framework that integrates satellite observations, meteorological reanalysis data, and ground monitoring measurements to improve daily PM_2.5 prediction. Eleven predictor variables were processed using a unified Google Earth Engine pipeline, including MODIS aerosol optical depth, Sentinel-5P trace gases (CO, NO₂, SO₂), and ERA5 meteorological parameters. Four tree-based machine learning algorithms—Random Forest, XGBoost, LightGBM, and CatBoost—were trained using daily observations from 2019 to 2023. Model evaluation using an independent 2024 dataset showed strong predictive capability, with Random Forest achieving R² = 0.77 (RMSE = 24.75 µg m⁻³), XGBoost R² = 0.76 (RMSE = 26.32 µg m⁻³), CatBoost R² = 0.73 (RMSE = 30.39 µg m⁻³), and LightGBM R² = 0.70 (RMSE = 32.75 µg m⁻³). To further enhance performance, the best models were combined into a weighted ensemble (RF 0.5, XGBoost 0.3, and CatBoost 0.2), which produced the highest validation accuracy (R² = 0.77; RMSE = 23.37 µg m⁻³). Statistical testing using paired t-tests and Diebold–Mariano tests confirmed that the ensemble significantly reduced forecast errors compared with individual models. Feature importance analysis revealed that surface pressure, temperature, CO, and NO₂ were the most influential predictors of PM_2.5 variability. The proposed framework demonstrates that combining satellite data, reanalysis meteorology, and ground observations through ensemble learning can provide accurate and scalable air quality forecasting for data-limited urban environments.

Accurate mapping of inland and coastal water bodies is crucial for monitoring environmental changes, managing hydrological resources, and assessing the impacts of climatic variability. This study presents a deep learning-based semantic segmentation framework that leverages multi-band Sentinel-2 imagery for delineating glaciers and coastal lakes. The dataset comprises 400×400-pixel image patches, constructed using false-colour composites of Sentinel-2 Bands 8 (NIR), 4 (Red), and 3 (Green), which enhance spectral separation between water and non-water surfaces. These bands were strategically selected to improve water body contrast and boundary definition through multi-sensor data fusion, enabling more precise lake border extraction. Each image patch is paired with hand-labeled binary lake masks to serve as ground truth. We developed and trained a Simple U-Net in PyTorch and a shallow Convolutional Neural Network (CNN) in TensorFlow to evaluate model performance and architectural efficiency using the same dataset. Both models were assessed using standard performance metrics, including Precision, Recall, F1 Score, and Intersection over Union (IoU). Results show high segmentation accuracy across both platforms (F1 > 0.92, IoU > 0.86). The TensorFlow-based model exhibited faster training and inference, while the PyTorch U-Net provided more consistent and accurate border delineation. This work demonstrates the synergistic power of multi-band spectral fusion and deep learning for environmental feature extraction in remote sensing. The proposed models and methods are scalable and adaptable for broader applications in coastal monitoring, inland water mapping, and climate-related hydrological assessments, offering a valuable contribution to automated Earth observation workflows under changing climatic conditions.

Urban resilience is essential for cities to endure and adjust to environmental and socioeconomic upheavals. The static indicators and rule-based spatial frameworks that are the mainstays of traditional resilience assessment models frequently fall short of capturing the dynamic character of coastal and urban resilience. This paper suggests a deep learning-based categorization framework for identifying resilience levels in urban and coastal settings by combining Long Short-Term Memory (LSTM) networks with multi-sensor remote sensing data. The Copernicus Marine Data Service's spatiotemporal ocean physics data, namely the eastward (uo) and northward (vo) seawater velocity, are used in the model to increase the precision of resilience evaluations. The methodology includes a multi-step deep learning pipeline, incorporating data preprocessing, feature extraction, class balancing with SMOTE, and LSTM-based classification. The proposed LSTM model is optimized to enhance performance with dropout regularisation (0.3), an Adam optimizer (learning rate = 0.0003), and class weighting strategies. The model is evaluated using accuracy, F1-score, confusion matrices, and loss curves, ensuring reliable classification across different resilience categories. Results indicate that the framework achieves high classification accuracy (91.5%), demonstrating superior performance compared to traditional machine learning approaches. Regarding multi-sensor fusion and deep learning, this study provides a scalable, adaptive, and data-driven solution for resilience classification, supporting climate adaptation strategies, disaster risk management, and sustainable urban development. The proposed methodology offers a robust tool for policymakers and urban planners, enabling more effective resilience monitoring and decision-making in rapidly evolving urban and coastal environments.

Rapid urbanization in Lahore, Pakistan, has led to significant ecological and thermal challenges, particularly the intensification of Urban Heat Island (UHI) effects and increased thermal stress as measured by the Urban Thermal Field Variance Index (UTFVI). This study employs a multi-temporal evaluation of Landsat satellite imagery and GIS-based analysis to investigate the Spatio-temporal trends in land-use and land-cover (LULC) changes from 1994 to 2024. We detected substantial changes in urban growth, vegetation cover, and barren areas using supervised classification (Random Forest) methods and remote sensing indices such as NDVI (Normalized Difference Vegetation Index), NDMI (Normalized Difference Moisture Index), NDBI (Normalized Difference Built-up Index), and NDWI (Normalized Difference Water Index). The classification models demonstrated high accuracy, with combined values ranging from 96.7% to 98.4%. Our results indicate a significant increase in metropolitan regions, from 608.66 km² (32.43%) in 1994 to 968.50 km² (51.6%) in 2024. This urban expansion correlates with rising land surface temperatures (LST), exacerbating UHI effects. The UTFVI analysis reveals a progressive rise in heat stress from 1994 to 2024. In 1994, most areas ranged from “Good” (0.020), indicating severe urban heat intensification, particularly in central and southern regions providing evidence of a significant increase in urban heat, especially in the center and southern areas of the city. These findings underscore the critical need for sustainable land management strategies to mitigate the adverse effects of Lahore's rapid urbanization on the environment. This study comprehensively explains unplanned urban growth's ecological and climatic consequences. The importance of balanced development in protecting natural ecosystems and enhancing urban resilience is highlighted, offering valuable insights for urban planners and policymakers.

Rapid urbanization and industrialization in Lahore and Faisalabad have intensified air pollution issues, influencing nitrogen dioxide (NO ₂ ) concentrations, land surface temperature (LST), and vegetation. The study aims to comprehensively assess changes in NO ₂ , LST, and vegetation induced by industrialization, focusing on seasonal variations from 2019 to 2022. The study evaluates NO ₂ concentrations vegetation health using indices normalized difference vegetation index, enhanced vegetation index, atmospherically resistant vegetation index, and LST variations. The analysis reveals a notable increase in NO ₂ during both summer and winter, with approximately 0.021 (×10 ³ mol/m ² ) and 0.03 (×10 ³ mol/m ² ) rises observed in Lahore. In comparison, Faisalabad experienced more modest increases of around 0.0034 (×10 ³ mol/m ² ) and 0.007 (×10 ³ mol/m ² ) in the respective seasons. Simultaneously, vegetation indices decline in both cities, indicating substantial vegetation health deterioration. Moreover, a notable upward trend in LST occurred, with Lahore experiencing an increase of approximately 1.59 °C in summer and 0.92 °C in winter. Faisalabad also showed rises of around 1.64 and 0.54 °C in the corresponding seasons. Pearson correlation analysis highlights a robust negative correlation between NO ₂ and vegetation indices, underlining the impact of declining vegetation on air quality. A positive correlation between NO ₂ and LST indicates the interconnected nature of rising temperatures and air pollution. The findings emphasize the need for environmental regulations in Lahore and Faisalabad. Addressing rising NO ₂ levels and temperatures is critical for policymakers and urban planners. These insights contribute to the sustainable development goal 11, fostering strategies for sustainable cities and communities to combat pressing environmental challenges in these urban areas.

Soil erosion is a destructive consequence of land degradation caused by deforestation, improper farming practices, overgrazing, and urbanization. This irreversible effect negatively impacts the limited renewable soil resource, causing soil truncation, reduced fertility, and unstable slopes. To address the anticipation of erosion modulus resulting from long-term land use and land cover (LULC) changes, a study was conducted in the Swat District of Khyber Pakhtunkhwa (Kpk), Pakistan. The study aimed to predict and evaluate soil erosion concerning these changes using remote sensing (RS), geographic information systems (GIS), and the Revised Universal Soil Loss Equation (RUSLE) model. We also evaluated the impact of the Billion Tree Tsunami Project (BTTP) on soil erosion in the region. Model inputs, such as rainfall erosivity factor, topography factor, land cover and management factor, and erodibility factor, were used to calculate soil erosion. The results revealed that significant soil loss occurred under 2001, 2011, and 2021 LULC conditions, accounting for 67.26%, 61.78%, and 65.32%, falling within the category of low erosion potential. The vulnerable topographical features of the area indicated higher erosion modulus. The maximum soil loss rates observed in 2001, 2011, and 2021 were 80 t/ha⁻¹/year⁻¹, 120 t/ha⁻¹/year⁻¹, and 96 t/ha⁻¹/year⁻¹, respectively. However, the observed reduction in soil loss in 2021 as compared to 2001 and 2011 suggests a positive influence of the BTTP on soil conservation efforts. This study underscores the potential of afforestation initiatives like the BTTP in mitigating soil erosion and highlights the significance of environmental conservation programs in regions with vulnerable topography.