The study of rosette jets in crossflow is essential for understanding the transport and mixing of hot and saline discharges in crossflow ambient. Conventional approaches utilize physical experiments and Computational Fluid Dynamics (CFD) for study, which are both costly and time-consuming. This highlights the necessity for more time- and cost-efficient models, such as machine learning techniques, which offer a promising alternative for achieving accurate and efficient predictions. This study aims to evaluate the performance of machine learning algorithms using datasets generated by the CFD tool OpenFOAM, validated by experimental results. The datasets included key parameters influencing the jet trajectory, such as jet-to-ambient velocity ratios, Reynolds number, jet angle, and jet trajectory. Various machine learning algorithms, including support vector machines (SVM), Extreme Learning Machine (ELM), and Adaptive Neuro-Fuzzy Inference System (ANFIS), were trained and tested using these datasets. The models’ performance was evaluated based on their predictive accuracy and computational efficiency. The machine learning models were developed to predict the jet trajectory based on the jet-to-ambient velocity ratios, Reynolds number, dimensionless jet angle. The performance of these models was assessed using multiple statistical metrics, and the results were benchmarked against the previous study’s findings. The machine learning models demonstrated varying degrees of success in predicting the jet trajectory. The results were compared against the previous study, showing that some machine learning models effectively captured the complex dynamics of rosette jet trajectories, validating the models’ robust generalization capabilities. These models provide a rapid and feasible alternative to traditional CFD methods for accurately predicting the trajectories of rosette jets, which support the design and environmental assessment of coastal outfall systems.

Ensuring the safety and quality of drinking water is crucial for public health, particularly in regions where water contamination is a significant concern. This study investigates the application of machine learning techniques for water quality analysis in the Indian state of Kerala. A total of 328 water samples were collected and analyzed for various parameters including pH, dissolved oxygen, total coliform, fecal coliform, conductivity, nitrate, and biochemical oxygen demand. These parameters were used to compute the Water Quality Index (WQI), which was subsequently classified into four categories: clean, unclean, polluted, and highly polluted. Five machine learning classifiers were employed to classify the water quality data: Support Vector Machine (SVM), Decision Tree (DT), k-Nearest Neighbors (k-NN), Logistic Regression (LR), and XGBoost. The classifiers were trained and tested on the dataset to determine their accuracy in predicting water quality classes. Among these, XGBoost emerged as the most accurate classifier, achieving a classification accuracy of 91%. The study highlights the effectiveness of machine learning in environmental monitoring and demonstrates the potential of these techniques to aid in water quality management. The high accuracy of XGBoost suggests that it can be a valuable tool for predicting water quality and identifying areas at risk of pollution. By providing reliable classifications, machine learning models can support decision-makers in implementing timely and appropriate interventions to ensure the safety and cleanliness of drinking water.

The efficient management of irrigation networks is a pressing need, and it requires huge datasets to calibrate and simulate varying annual conditions. However, acquiring such datasets with the desired statistical properties might be challenging. This study presents a detailed methodology to generate synthetic years of flow rate data in irrigation networks. This approach is centred on the parameters of the best-adjusted distribution function, target volume, minimum and maximum constraints, and the number of values to generate, which is a crucial step towards more precise modelling and management of water resources.

The methodology was developed and implemented in MATLAB using the Statistics and Machine Learning Toolbox. First, the ideal distribution function was determined for each month of the dataset (e.g., Normal, Gamma, Lognormal). The input parameters werethe parameters of the best-adjusting distribution, the total volume for each month, the maximum and minimum flow values, and the total number of entries to generate. The function generates an initial set of random numbers following the specified distribution, then normalises and transforms the data. An iterative optimization process is carried out to adjust the values to match the desired monthly volume, ensuring the convergence criteria are met. Thus, the synthetic data represent the variations in the demands.

The methodology was tested with various distribution functions and target values to validate its performance. The generated synthetic years closely followed the input best-fitting distribution patterns and matched the target volume with minimal deviation. This study introduces a MATLAB-based methodology that effectively generates synthetic years of data tailored for irrigation networks. The approach facilitates realistic and reliable simulations of water usage patterns by ensuring adherence to the best-fitting distribution and the input constraints. This tool is valuable for planning, optimizing, and managing irrigation networks.

Check dams generally consist of a vertical barrier constructed on ditches, small streams, channels and gullies that have often been formed by the erosive activity of water. A check dam serves many purposes, such as reducing runoff velocity, reducing erosive activities, reducing the original channel gradient, improving bed sediment moisture in adjoining areas, sediment retention and allowing for percolation to recharge aquifers. A check dam interferes with flows in the upstream and downstream channels and dissipates the energy of flowing water. Therefore, identifying the quantitative variables that influence the volume of these structures is crucial for accurately estimating construction costs and their effectiveness. This study aimed to model check dam volumes across 100 sub-basins in eight provinces of Iran (Alborz, East Azerbaijan, Ilam, Bushehr, Qazvin, Fars, Mazandaran, and Hamadan). The database for modeling included 27 environmental features from each of the 100 sub-basins, and Gene Expression Programming (GEP) was used for the modeling process. The results indicated that the key features for estimating check dam volume among the 27 variables studied are precipitation, slope, drainage density, TWI index, shape factor, elevation difference, concentration time and NDVI index. The evaluation of the modeling, based on R², RRMSE, RAE and NSE values, revealed that the most accurate model for Qazvin province had values of 0.97, 0.18, 0.17 and 0.96, respectively. In contrast, the least accurate model for Mazandaran province had values of 0.80, 0.38, 0.35 and 0.80. Additionally, the results demonstrated that environmental characteristics could be used with high accuracy to estimate check dam volumes quickly. This allows for the relevant costs to be estimated before implementing check dams, facilitating the prioritization of areas effectively.

Introduction: This paper examines the transformative role of drone technology in water sciences, focusing on its applications in hydrology, water quality assessment, and aquatic ecosystem monitoring. With the growing need for high-resolution, spatially comprehensive data in water sciences, drones offer an unprecedented opportunity to enhance data collection and analysis. Methods: We review current methodologies, including drone-based remote sensing, water sampling, and thermal imaging techniques, to assess water bodies and their surrounding environments. Results: Findings highlight the efficiency of drones in capturing detailed, real-time data on water quality parameters, surface water movements, and vegetation health, contributing to more accurate hydrological models and environmental assessments. Conclusions: The incorporation of drone technology in water sciences significantly advances the field, offering more agile, precise, and cost-effective methods for water monitoring and management. This shift not only supports better-informed decision-making but also paves the way for innovative research avenues in understanding and protecting aquatic ecosystems. Future research should focus on developing standardized protocols for drone operations in water sciences and exploring the integration of drone data with traditional monitoring systems for comprehensive watershed management. Moreover, as drone technology continues to evolve, its integration into water sciences promises further enhancements in spatial and temporal data resolution. This progression is crucial for addressing the complex challenges of water management in the face of climate change and increasing human impact on natural water systems.

The objective of this study is to compare two modeling approaches, the Takagi-Sugeno Fuzzy System (TSFS) and Generalized Regression Neural Network (GRNN) models, with evapotranspiration calculated by FAO-56. The selected sites are situated in Constantine, Guelma, Mascara, Saida, Setif, Souk Ahras, Tiaret, and Tlemcen, known for their semiarid climates. The daily data registered from 2000 to 2022 include air temperature at 2m (°C), relative humidity at 2m (%), dew point at 2m (°C), precipitation (mm), surface pressure (hPa), ET0 FAO evapotranspiration (mm), vapor pressure deficit (kPa), wind speed at 10m (km/h), soil temperature from 0 to 7cm (°C), soil moisture from 0 to 7cm (m³/m³), sunshine duration (s), and terrestrial radiation (W/m²). The data were split into training (70%), validation (15%), and testing (15%) sets. To evaluate the two models, several indices were calculated, including Nash–Sutcliffe efficiency, coefficient of determination, root mean square error, mean absolute error, ratio sum ratio, and Willmott index.

The statistical results indicate that the GRNN model provides more accurate estimations of evapotranspiration compared to the TSFS model in semiarid regions. This is evidenced by a root mean square error (RMSE) of ≤ 0.285, a mean absolute error (MAE) of ≤ 0.212, a minimum coefficient of determination (R²) of 0.976, a Nash–Sutcliffe efficiency (NSE) of ≥ 0.976, a ratio sum ratio (RSR) of ≤ 0.156, and a Willmott index (WI) of > 0.882 for training, validation, and testing. In contrast, the TSFS model shows an RMSE of ≤ 0.513, an MAE of ≤ 0.405, an R² of > 0.923, an NSE of ≥ 0.965, an RSR of ≤ 0.277, and a WI of > 0.799.

The findings of this study confirm that the GRNN model is more suitable for accurately estimating evapotranspiration in semiarid regions, contributing to efficient water resource management in these areas . Future research should focus on expanding the dataset to include diverse climatic regions to enhance the models' applicability.

Drought is a major climatic hazard in the Mediterranean region, particularly in the Tensift river basin in Morocco. It has severe implications for water availability, agriculture, and local economies. However, traditional monitoring systems often fail to provide timely drought warnings. This study explores the integration of satellite and reanalysis data to enhance early warning systems, aiming to improve drought detection and monitoring across the Mediterranean. The effectiveness of the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI) based on satellite data from the Climate Hazards Group Infrared Precipitation (CHIRPS) and reanalysis data (ERA5-Land) was assessed using Receiver Operating Characteristic (ROC) curve analysis, comparing them to ground observations from 1981 to 2021. This study focused on optimizing thresholds and timescales for these indices to improve drought detection. The integration of CHIRPS and ERA5-Land significantly enhanced the detection of drought events compared to conventional methods. The ROC analysis identified optimal threshold levels for SPI and SPEI, which improved their detection performance. The Pearson Type III distribution was found to be the most suitable for SPI calculations, while the Log-logistic distribution was best for calculating SPEI. Integrating satellite and reanalysis data significantly advances drought characterization in the Tensift basin, facilitating more proactive drought management. This method proves crucial for mitigating impacts and supporting decision-makers in sustainable water resource management. Future research should aim to integrate these indices with socio-economic impacts to develop a comprehensive drought risk management strategy.

Given the recent increased attention towards "tipping points" in understanding the effects of climate change, it is vital to comprehend Arctic sea ice variability over the past twenty-one thousand years before the present (ka). Here, we have analyzed a quasi-continuous coupled general circulation model (CGCM) simulation to gain further insight into how the Arctic sea ice coverage (SIC) (ocean surface covered with more than 5% sea ice concentration) spatially responds to changes in orbital, greenhouse gas, meltwater, and continental ice sheet forcings over the past 21 ka. We found that the annual mean Arctic SIC predominantly varied in the Atlantic Ocean sector, particularly during the Heinrich 1 (H1; ∼19–17 ka) and Younger Dryas (YD; ∼12.9–11.3 ka) cold events and Bølling–Allerød (BA; 17–14.35 ka) warm periods. By altering the poleward oceanic heat transport into the Arctic, a weakened (strengthened) AMOC resulted in the expansion (reduction) of Arctic SIC. Meanwhile, we observed a two-way relationship between the AMOC and Arctic SIC with an out-of-phase (in-phase) relationship when AMOC leads (lags) Arctic SIC by 20 years (4 ka). Subsequently, we observed a sea ice-capping mechanism wherein an increase (decrease) in Atlantic SIC during H1/YD (BA) reduces (increases) net ocean surface heat flux and deep convection, thereby influencing the AMOC strength. Furthermore, it was found that the SPG and AMOC strengths have been in-phase throughout the past 21 ka, except during the abrupt termination and input of freshwater flux during the BA and Meltwater Pulse 1A (∼14.4–13.9 ka) events, respectively. Recent studies have observed increased Greenland ice sheet melting and precipitation in the Arctic, which would increase the influx of freshwater into the Arctic Ocean. Meanwhile, our paleoclimate study suggests that a sudden change in freshwater discharge into the subpolar North Atlantic may alter the polar ocean dynamics, which may help in understanding near-future Arctic conditions.

China ranks sixth in the world in terms of freshwater resources, yet its per capita freshwater resources are only one-fourth of the global average. Additionally, freshwater resources exhibit spatial unevenness. Given the rapid pace of urbanization, the safety of China's freshwater resources in terms of water quality is at risk due to contamination. Consequently, the utilization of water quality models for predicting changes in water quality has emerged as a prominent research focus. The data-driven water quality prediction model provides a scientific basis for water resource management departments to provide early warning of algal blooms and formulate effective control measures in advance. In this study, we predicted the dominant algae (blue-green algae) population in a water source based on the Long Short-Term Memory (LSTM) network model. Additionally, we explored the effects of the type of feature combination and the step size of the time window on the prediction performance of the LSTM by establishing different feature combinations and different time windows as input methods. The experimental results show that the performance of multi-feature prediction is consistently superior to single-feature prediction. Simultaneously, increasing the number of features in the input model tends to diminish the model's predictive performance. The time window impacts the performance of LSTM predictions. As the step size increases, the model prediction performance gradually enhances, eventually stabilizing while concurrently incurring significant time overhead. Lastly, this study offers insights into future research directions from three key dimensions: the input indicator, optimization algorithm, and model combination.

Timor-Leste is located in Southeast Asia and the South Pacific, about 500 km from the sea border with Australia and the land border with Indonesia. It has an area of approximately 14,954 km² and an estimated population of 1.293 million people. The climate is influenced by the Asian Monsoon regime, in addition to the strong influence of El Niño Southern Oscillation and intra-seasonal variability. Despite its small territory it presents an interesting climatic diversity and until now, studies on the efficiency of large-scale dynamic models on it have not been reported. Thus, the objective is to analyze the skill of two reanalyses in simulating precipitation and temperature over the territory of Timor-Leste. For that, we used the precipitation and temperature data collected in eight (8) cities in the period of 2013 and 2014 with monthly sampling. The models were the ERA-5 and ERA-interim, which are re-analyzes of the European Center for Medium-Range Weather Forecasts. The results show that ERA-5 reanalysis performed better compared to ERA-interim, both for precipitation and temperature. However, the result of the model depends on the geographic location of the station. In general, the models were worse in cities located in mountainous regions. We conclude that, in general, the ERA-5 reanalysis managed to reproduce in a more adequate way both the monthly temperature and the rainfall in Timor-Leste.