Parallel Session 1B

Blast-induced ground vibration is a key environmental and operational concern in quarrying and mining industries because excessive peak particle velocity (PPV) can affect nearby structures, regulatory compliance and public confidence. This study proposes a physics-guided neural network (PGNN) for PPV estimation that combines the interpretability of scaled distance (SD) attenuation theory with neural residual learning. A revised dataset was curated from the Blast-Induced Vibration Data Evaluation Program (BIVDEP) example records by U.S. Department of Interior, representing monitored surface-mine blasts with distance, charge weight per delay, SD and measured PPV. After cleaning, 184 records were retained for model training and testing. The proposed approach was assessed against empirical, optimized empirical and data-driven benchmarks using standard error and goodness-of-fit metrics. The United States Bureau of Mines (USBM) formula baseline produced K = 428.768 and n = 1.190. On the test set, the fixed-parameter PGNN achieved best accuracy, with root mean squared error (RMSE) = 21.621 mm/s and R² = 0.8988, followed closely by the trainable-parameter PGNN. These results demonstrate that embedding blast-vibration physics in neural modelling can improve prediction reliability while preserving physically meaningful PPV attenuation trends. The framework offers an interpretable tool for blast design review, vibration compliance screening and safer ground-vibration management.

As populations expand in tropical coastal regions, the demand for freshwater has intensified, leading to groundwater depletion and saltwater intrusion, and Sri Lanka is no exception. Coastal reservoirs provide a sustainable alternative by capturing riverine freshwater before it discharges into the sea. This study develops a satellite‑derived screening framework to evaluate the suitability of six major Sri Lankan estuaries, Kelani, Kalu, Gin, Nilwala, Maha Oya, and Deduru Oya, for coastal reservoir development. Sentinel‑2 and Landsat imagery (2017-2025) were processed using a sequential funnel screening approach to assess estuaries across four criteria: geomorphological classification, identifying bar‑built systems; morphological stability, quantified through river mouth width trends; siltation risk, estimated via suspended sediment loads; and nearshore bathymetry, derived from the Log‑Ratio method to minimize construction cost of coastal reservoir. Results identified the Maha Oya estuary as the most viable candidate, demonstrating exceptional stability (coefficient of variation of estuarine width 5.87%), minimal turbidity (maximum NDTI 0.06), and a shallow nearshore shelf. The Nilwala River ranked second, while Kelani, Gin, Kalu, and Deduru Oya were excluded due to unfavourable geomorphology, deep bathymetry, extreme siltation, or instability. Validated against Singapore’s Marina Barrage, this framework establishes a cost‑effective, remote data‑driven methodology for coastal reservoir site selection in data‑scarce regions.

Mountain glaciers are highly sensitive to global warming-based climatic and environmental change. Comprehensive research applications are more important to conserve the cryosphere ecosystem and sustainable water resource management. This study aims to analyze the spatial and multi-temporal changes of snow cover, to analyze spatial and multi-temporal changes of glacier extent, and to assess the associations among snow cover, glacier extent, and land surface temperature in the Kuhumbuhimal Glacier, Nepal, in 1995, 2001, 2011, 2018, 2020, and 2024. To achieve these objectives, Landsat Level 2 satellite images were used to compute the Normalized Difference Snow Index (NDSI), Normalized Difference Glacier Index (NDGI), and Land Surface Temperature (LST) using ArcGIS Pro. Pearson’s correlation analysis was computed using 160 random sampling points to assess the above relationships using R. The results revealed that snow-covered areas were limited in the upper accumulation and eastern parts, while snow-free areas expanded in the lower accumulation of the glacier. Ice-mixed debris areas were distributed towards the glacier tongue, indicating intense glacier ablation. Correlation analysis showed a positive relationship between snow cover and glacier extent. Negative correlations were demonstrated by NDSI and NDGI with LST. The findings highlighted that LST directly influenced the spatial-multitemporal changes of this glacier.

Frequent water quality monitoring is essential for estuarine ecosystem management, yet traditional field sampling is often time consuming, expensive and spatially constrained. Most existing studies often rely on near 2-hour satellite-to-in-situ data matching windows and basic spectral bands and normalized difference indices (e.g., NDCI, NDTI), which limits the predictive accuracy in highly dynamic, optically complex waters. These gaps are addressed during this study by using a 30-minute data matching window and an extensive feature extraction followed by two-stage feature selection process based on Pearson correlation (|r| ≥ 0.3) and minimum Redundancy Maximum Relevance (mRMR) algorithm. Random Forest (RF), Support Vector Regression (SVR) and Artificial Neural Networks (ANN) were deployed to estimate Chlorophyll-a (Chl-a), Coloured Dissolved Organic Matter (CDOM), and Turbidity in Caloosahatchee River Estuary (CRE), Florida, using Sentinel-2 Level-2A imagery. SVR had the best accuracy for CDOM (R2 = 0.819), for Chl-a RF performed best (R2=0.63) and poor predictive accuracy was obtained for turbidity in all developed models. This study provides a scalable and cost-effective framework for continuous water quality monitoring in optically complex estuarine waters, reducing the need for extensive in-situ measurements.

Sustainable nitrogen management has become an important environmental concern because reactive nitrogen, including nitrogen oxides (NOx), contributes to air pollution, human health impacts, and climate-related atmospheric processes. Therefore, increased attention has been given to monitoring and reducing NOx emissions. In Sri Lanka, the limited availability of ground-based atmospheric pollutant monitoring stations restricts detailed assessment of NO2 plume behavior near individual emission sources. Satellite observations therefore provide a useful alternative for assessing tropospheric NO2 around major emission sources where ground observations are limited. This study addresses this gap by assessing the tropospheric NO2 plume associated with the Norochcholai Coal Power Plant using Sentinel-5P TROPOMI observations from 2023 to 2025, supported by ERA5 wind data. The results showed temporal variation, strong spatial clustering, and downwind changes in NO2 plume structure. For the selected overpasses on 6 April 2023 and 6 April 2024, the estimated NOx source rates were 131.21 kg h⁻¹ and 205.00 kg h⁻¹, respectively. The variation in these estimates indicates the influence of wind and plume dispersion conditions. The findings demonstrate that satellite-based plume assessment can support air-quality monitoring, emission-source identification, and environmental decision-making in regions with limited ground observations.

This study introduces a distributed rainfall–runoff modelling approach that uses a storage function-based subsurface representation to model the streamflow in the Kubukkan Oya catchment, Sri Lanka. The study addresses the issue of hydrological modelling in data scarce regions, where ground-based observations are insufficient for accurate model simulation. To address this limitation, satellite-based precipitation products were used and merged together by using an inverse error variance weighing method to minimize the bias and enhance reliability. The distributed model has been developed by combining the surface runoff, infiltration, evapotranspiration and groundwater flow processes at the grid level, and evapotranspiration was calculated by the Hargreaves–Samani methodology. The period 2018-2025 was used for the model calibration and validation with observed streamflow data available at Nakkala gauging station, and the data split was 80-20. The results show that the blended precipitation product enhances the performance of the model over the raw satellite precipitation inputs with a Nash–Sutcliffe Efficiency (NSE) of 0.463 during the calibration period and 0.570 during the validation period. The results show the potential of using a simplified distributed model, coupled with bias-corrected satellite data, for simulation of streamflow for ungauged basins. This study helps to enhances the water resources assessment and assists future climate resilience water management strategies in Sri Lanka.

This final segment invites reflections from presenters and attendees, synthesizing key insights from the session. The session chair will conclude the discussion by summarizing thematic threads, highlighting interdisciplinary contributions, and outlining potential collaborative directions.