Accurate streamflow prediction is essential for flood warning, reservoir operation, irrigation scheduling, hydropower planning, and sustainable water management, yet remains challenging due to the complexity of hydrological processes. Although data-driven models often outperform conventional physics-based hydrological modelling approaches, their real-world deployment is limited by cost, infrastructure demands, and the interdisciplinary expertise required. To bridge this gap, this study developed QPred, a regional, lightweight, cost-effective, web-delivered application for daily streamflow forecasting. The study executed an end-to-end workflow, from field data acquisition to accessible web-based deployment for on-demand forecasting. High-resolution rainfall data were recorded with tipping-bucket gauges and loggers, while river water depth in the Aglar and Paligaad watersheds was converted to discharge using site-specific rating curves, resulting in a daily dataset of precipitation, river water level and discharge. Four DL architectures were trained, including vanilla Long Short-Term Memory (LSTM), stacked LSTM, bidirectional LSTM, and Gated Recurrent Unit (GRU), and evaluated using Nash-Sutcliffe Efficiency (NSE), Coefficient of Determination (R²), Root-Mean-Square-Error-Standard-Deviation Ratio (RSR), and Percentage Bias (PBIAS) metrics. Performance was watershed-specific, as the vanilla LSTM demonstrated the best generalisation for the Aglar watershed (R² = 0.88, NSE = 0.82, RMSE = 0.12 during validation), while the GRU achieved the highest validation accuracy in Paligaad (R² = 0.88, NSE = 0.88, RMSE = 0.49). All models achieved satisfactory to excellent performance during calibration (R² > 0.91, NSE > 0.91 for both watersheds), demonstrating strong capability to capture streamflow dynamics. The highest performing models were selected and embedded into the QPred application. QPred was developed as a lightweight web pipeline, utilising Google Colab as the primary execution environment, Flask as the backend inference framework, Google Drive for artefact storage, and Ngrok for secure HTTPS tunnelling. A user-friendly front end utilises range sliders (bounded by observed minima and maxima) to gather inputs and provides discharge data along with metadata, thereby enhancing transparency. This work demonstrates that accurate, context-aware deep learning models can be delivered through low-cost, web-based platforms, providing a reproducible and scalable pipeline for hydrological applications in other watersheds and for practitioners.

Kala Oya is one of the drier river basins in Sri Lanka that is affected by droughts for certain time periods. Water shortages are visible in crop yields and public water supply due to climate change. Consequently, the Soil and Water Assessment Tool (SWAT) was used to develop hydrological models, and Water Evaluation and Planning (WEAP) software was used to analyze the water allocation for the basin. The software's evaluation and assessing capability of allocation of water, transmission, and diversion links for demands are some reasons to use WEAP as a separate water allocation model. The future land use and climate aspect (2040) has also been included in these models to enable the generation of "scenarios" that can be used to test the demand deficits for irrigation and public water supply. Three climatic conditions such as optimistic, pessimistic, and average for 2040 were considered for modeling. Our major findings include: 1. The pessimistic climate change scenario exhibits the highest rise in drought metrics while the optimistic represents the lowest. Under current land use conditions, annual Long-Term Average (LTA) public water supply deficits are 1.0 million cubic meters (MCM) (3.2%), and for future land use, in a pessimistic climate change scenario, annual LTA deficits are 4.7 MCM (4.4%). 2. For medium/major irrigated agriculture, annual LTA deficits for current conditions are 42.8 MCM (9.0%), and for future land use, pessimistic climate change scenarios are 56.1 MCM (12.7%). For minor irrigation, annual LTA deficits for current conditions are 20.8 MCM (17.9%) and future pessimistic climate change scenarios are 24.2 MCM (20.2%). 3. This study concludes that the public water supply demand deficits are considerably greater in the middle and lower catchments of Kala Oya basin for future land use (with basin developments) model simulations. This may create water scarcity and social stress for people who require immediate mitigation measures. 4. Overall, it was revealed that the agriculture-oriented drought losses (major/medium irrigation) are significant (around 66–67% of total demand deficits) in the Kala Oya basin, and they may create adverse impacts on the country's economy due to crop yield losses.

Floods have been identified as one of the world's most common and widely distributed natural disasters over the last few decades. Floods' negative impacts could be significantly reduced if accurately predicted or forecasted in advance. Apart from large-scale spatiotemporal data and greater attention to data from the Internet of Things, the worldwide volume of digital data is increasing. Artificial intelligence plays a vital role in analyzing and developing the corresponding flood mitigation plan, flood prediction, or forecast. Machine learning (ML)-based models have recently received much attention due to their self-learning capabilities from data without incorporating any complex physical processes. This study provides a comprehensive review of ML approaches used in flood prediction, forecasting, and classification tasks, serving as a guide for future challenges. The importance and challenges of applying these techniques to flood prediction are discussed. Finally, recommendations and future directions of ML models in flood analysis are presented.

In this study, a machine learning-based predictive model was developed for the Musa petti Wind Farm in Sri Lanka to address the need for localized forecasting solutions. Using data on wind speed, air temperature, nacelle position, and actual power, lagged features were generated to capture temporal dependencies. Among 24 evaluated models, the ensemble bagging approach achieved the best performance, with R² values of 0.89 at 0 min and 0.75 at 60 min. Shapley Additive exPlanations (SHAP) analysis revealed that while wind speed is the primary driver for short-term predictions, air temperature and nacelle position become more influential at longer forecasting horizons. These findings underscore the reliability of short-term predictions and the potential benefits of integrating hybrid AI and probabilistic models for extended forecasts. Our work contributes a robust and explainable framework to support Sri Lanka’s renewable energy transition, and future research will focus on real-time deployment and uncertainty quantification.