The sediment transport, involving the movement of the bedload and suspended sediment in the basins, is a critical environmental concern that worsens water scarcity and leads to degradation of land and its ecosystems. Machine learning (ML) algorithms have emerged as powerful tools for predicting sediment yield. However, their use by decision-makers can be attributed to concerns regarding their consistency with the involved physical processes. In light of this issue, this study aims to develop a physics-informed ML approach for predicting sediment yield. To achieve this objective, Gaussian, Center, Regular, and Direct Copulas were employed to generate virtual combinations of physical of the sub-basins and hydrological datasets. These datasets were then utilized to train deep neural network (DNN), conventional neural network (CNN), Extra Tree, and XGBoost (XGB) models. The performance of these models was compared with the modified universal soil loss equation (MUSLE), which serves as a process-based model. The results demonstrated that the ML models outperformed the MUSLE model, exhibiting improvements in Nash–Sutcliffe efficiency (NSE) of approximately 10%, 18%, 32%, and 41% for the DNN, CNN, Extra Tree, and XGB models, respectively. Furthermore, through Sobol sensitivity and Shapley additive explanation–based interpretability analyses, it was revealed that the Extra Tree model displayed greater consistency with the physical processes underlying sediment transport as modeled by MUSLE. The proposed framework provides new insights into enhancing the accuracy and applicability of ML models in forecasting sediment yield while maintaining consistency with natural processes. Consequently, it can prove valuable in simulating process-related strategies aimed at mitigating sediment transport at watershed scales, such as the implementation of best management practices.

From hourly ozone observations obtained from three regions⸻Houston, Dallas, and West Texas⸻we investigated the contributions of meteorology to changes in surface daily maximum 8-h average (MDA8) ozone from 2000 to 2019. We applied a deep convolutional neural network and Shapely additive explanation (SHAP) to examine the complex underlying nonlinearity between variations of surface ozone and meteorological factors. Results of the models showed that between 2000 and 2019, specific humidity (38% and 27%) and temperature (28% and 37%) contributed the most to ozone formation over the Houston and Dallas metropolitan areas, respectively. On the other hand, the results show that solar radiation (50%) strongly impacted ozone variation over West Texas during this time. Using a combination of the Kolmogorov-Zurbenko (KZ) filter and multiple linear regression, we also evaluated the influence of meteorology on ozone and quantified the contributions of meteorological parameters to trends in surface ozone formation. Our findings showed that in Houston and Dallas, meteorology influenced ozone variations to a large extent. The impacts of meteorology on West Texas, however, showed meteorological factors had fewer influences on ozone variabilities from 2000 to 2019. This study showed that SHAP analysis and the KZ approach can investigate the contributions of the meteorological factors on ozone concentrations and help policymakers enact effective ozone mitigation policies.

The adverse health effects associated with the inhalation and ingestion of naturally occurring radon gas produced during the uranium decay chain mean that there is a need to identify high-risk areas. This study detected radon-prone areas using a geographic information system (GIS)-based probabilistic and machine learning methods, including the frequency ratio (FR) model and a convolutional neural network (CNN). Ten influencing factors, namely elevation, slope, the topographic wetness index (TWI), valley depth, fault density, lithology, and the average soil copper (Cu), calcium oxide (Cao), ferric oxide (Fe₂O₃), and lead (Pb) concentrations, were analyzed. In total, 27 rock samples with high activity concentration index values were divided randomly into training and validation datasets (70:30 ratio) to train the models. Areas were categorized as very high, high, moderate, low, and very low radon areas. According to the models, approximately 40% of the study area was classified as very high or high risk. Finally, the radon potential maps were validated using the area under the receiver operating characteristic curve (AUC) analysis. This showed that the CNN algorithm was superior to the FR method; for the former, AUC values of 0.844 and 0.840 were obtained using the training and validation datasets, respectively. However, both algorithms had high predictive power. Slope, lithology, and TWI were the best predictors of radon-affected areas. These results provide new information regarding the spatial distribution of radon, and could inform the development of new residential areas. Radon screening is important to reduce public exposure to high levels of naturally occurring radiation.

This research study uses Artificial Neural Networks (ANNs) to predict occupational accidents in Sivakasi firework industries. Atmospheric temperature, pressure and humidity are the causes of explosion during chemical mixing, drying, and pellet making. The Proposed ANN model predicts the accidents and the session of accidents (FN/AN) based on atmospheric conditions. This prediction takes values from historical accident data due to the atmospheric conditions of Sivakasi (2009–2021). In the development of ANN model, the Feed-Forward Back Propagation (FFBP) with the Levenberg-Marquardt function has been employed with hidden layers of 5 and 10 to train the network. The performance accuracy of both the hidden layers is evaluated and compared with other models like Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbor (K-NN). The accuracy of the proposed model for accident classification is 82.7% and 67.8% for hidden layers 5 and 10, respectively. Also, the model predicts the session of accident with the accuracy of 72% and 54%, specificity of 77.7% and 60.1%, sensitivity of 69% and 52.92% for hidden layers 5 and 10, respectively. It is found that hidden layer 5 gives higher accuracy than hidden layer 10. The proposed ANN model gives the highest accuracy when compared to other models. This study is helpful in the firework industry management, and workers improve safety precautions and avoid explosions due to atmospheric conditions.

Studies have reported significant reductions in air pollutant levels due to the COVID-19 outbreak worldwide global lockdowns. Nevertheless, all of the reports are limited compared to data from the same period over the past few years, providing mainly an overview of past events, with no future predictions. Lockdown level can be directly related to the number of new COVID-19 cases, air pollution, and economic restriction. As lockdown status varies considerably across the globe, there is a window for mega-cities to determine the optimum lockdown flexibility. To that end, firstly, we employed four different Artificial Neural Networks (ANN) to examine the compatibility to the original levels of CO, O₃, NO₂, NO, PM₂.₅, and PM₁₀, for São Paulo City, the current Pandemic epicenter in South America. After checking compatibility, we simulated four hypothetical scenarios: 10%, 30%, 70%, and 90% lockdown to predict air pollution levels. To our knowledge, ANN have not been applied to air pollution prediction by lockdown level. Using a limited database, the Multilayer Perceptron neural network has proven to be robust (with Mean Absolute Percentage Error ∼ 30%), with acceptable predictive power to estimate air pollution changes. We illustrate that air pollutant levels can effectively be controlled and predicted when flexible lockdown measures are implemented. The models will be a useful tool for governments to manage the delicate balance among lockdown, number of COVID-19 cases, and air pollution.

As a driving factor of global changes, microplastics have gradually attracted widespread attention. Although MPs are extensively studied in aquatic systems, their presence and fate in terrestrial systems and soil are not fully understood. In China, construction-land must be mulched by dust-proof nets to prevent and control fine particulate pollution, which may cause MPs pollution and increase ecological risks. In order to understand the pollution characteristics and sources of MP in the soil covered by dust nets, we conducted a case study in Beijing. Our results revealed that the abundance of MPs in soil mulched by dust-proof nets ranged from 272 to 13,752 items/kg. Large-sized particles (>1000 μm) made up a significant proportion (49.83%) of MPs in the study area. The dominant MP polymer types were polyethylene (50.12%) and polypropylene (41.25%). The accumulation of MPs in construction-site soil mulched by dust-proof nets (average, 4910.2 items/kg) was significantly higher (P < 0.05) than that in unmulched soil (average, 840.8 items/kg), which indicates a dust-proof nets as an essential source of microplastics in the soil of construction land. We applied a remote-sensing data analysis technique based on remote imagery acquired from a high-resolution remote-sensing satellite combined with deep-learning convolutional neural networks to automatically detect and segment dust-proof nets. Based on high-resolution remote sensing images and using a U-net convolutional neural network, we extract the coverage area of Beijing’s dust-proof nets (18.6 km²). Combined the abundance of MPs and the dust-proof nets’ coverage area, we roughly estimate that 7.616 × 10⁹ to 3.581 × 10¹¹ MPs accumulated in the soil mulched by the dust-proof nets in Beijing. Such a large amount of MPs may cause a series of environmental problems. This study will highlight the understanding of soil MPs pollution and its potential environmental impacts for scientists and policymakers. It provides suggestions for decision-makers to formulate effective legislation and policies, so as to protect human health and protect the soil and the wider environment.

The assessment of trace metal pollution in aquatic environments remains a challenge. Chemical methods are insufficient and bioindicators seem to be the most promising alternative. Finding an adequate species is important to ensure accurate data. The combined use of several bioindicators may help to overcome the limitations of species’ spatial distribution and specific reactions. The aims of the present study were to compare the contents and bioaccumulation capability of 11 trace elements in Ceratophyllum demersum and different organs of Hydrocharis morsus-ranae and to validate H. morsus-ranae as a bioindicator of pollution in aquatic reservoirs using C. demersum, an established bioindicator, as a reference species. The application of several statistical techniques allowed us to identify similarities in accumulation patterns and concentration gradients between the two species. The results showed that concentrations of Cd, Co, Cr, Cu, Mn, Ni, Pb, Rb, Zn, V in C. demersum and roots of H. morsus-ranae were similar and mostly higher than in the leaves and stems of H. morsus-ranae. The contents of Cd, Co, Cr, Li, Mn, Ni, Rb, V, Zn were positively correlated. The inner transport of metals in H. morsus-ranae was limited (TF < 1). Both species are accumulators (BF > 10³) of Ni and Zn, and H. morsus-ranae also of Cu and Pb. Frog-bit roots were chosen to be most promising in bioindication. Major axis regression analysis showed that the uptake of Cd, Cr, Co, Li and Pb was similar in the two species. Neural networks demonstrated substantial uniformity in responses of C. demersum and roots of H. morsus-ranae to the type of anthropogenic activity and land use and similar spatial distributions of Cd, Cr, Co, Li and Pb. When Nemerow Pollution Index was applied, both species showed congruent gradients of contamination. Thus, H. morsus-ranae was validated as a reliable bioindicator of trace metal pollution in freshwater.

Pollutant emissions in ship exhaust have been continually increasing. SO₂ is one of the main gaseous pollutants in ship exhaust, resulting from the use of marine heavy fuel oil with high sulfur content. Therefore, it is necessary to detect the fuel sulfur content (FSC) to regulate ship exhaust emissions. Optical remote sensing methods, such as differential optical absorption spectroscopy (DOAS), light detection and ranging (LIDAR), and ultraviolet (UV) camera techniques, are regarded as simple and effective remote monitoring methods. One common technique is to estimate the SO₂ concentration in a ship plume using its local optical characteristics and use this to calculate FSC. One drawback of this technique is that there are always errors in the estimations of the SO₂ concentration despite the continuous improvement of such estimations. Another drawback is that calculating FSC from SO₂ often requires additional measurement methods. Here, a sulfur content prediction model based on a deep convolutional neural network using a UV camera is introduced. First, a ship benchmark test is performed. In the test, a large number of ultraviolet characteristic images of the ship exhaust plume are taken with a UV camera and the corresponding FSC data are collected. Next, a visual geometry group (VGG)-16 convolutional neural network model based on transfer learning is built. The model extracts all the features of the exhaust plume image as input data to the deep neural network and outputs the predicted FSC as a classification label. The results show that the model can predict the FSC value with high accuracy corresponding to the exhaust plume image. This study proves that it is theoretically feasible to apply a convolutional neural network to learn features of ultraviolet ship exhaust plume images for FSC predictions, which can provide guidance for the remote regulation of ship exhaust emissions.

Real-time river chloride prediction has received a lot of attention for its importance in chloride control and management. In this study, an artificial neural network model (i.e., multi-layer perceptron, MLP) and a statistical inference model (i.e., stepwise-cluster analysis, SCA) are developed for predicting chloride concentration in stream water. Then, an ensemble learning model based on MLP and SCA is proposed to further improve the modeling accuracy. A case study of hourly river chloride prediction in the Grand River, Canada is presented to demonstrate the model applicability. The results show that the proposed ensemble learning model, MLP-SCA, provides the best overall performance compared with its two ensemble members in terms of RMSE, MAPE, NSE, and R² with values of 11.58 mg/L, 27.55%, 0.90, and 0.90, respectively. Moreover, MLP-SCA is more competent for predicting extremely high chloride concentration. The prediction of observed concentrations above 150 mg/L has RMSE and MAPE values of 9.88 mg/L and 4.40%, respectively. The outstanding performance of the proposed MLP-SCA, particularly in extreme value prediction, indicates that it can provide reliable chloride prediction using commonly available data (i.e., conductivity, water temperature, river flow rate, and rainfall). The high-frequency prediction of chloride concentration in the Grand River can supplement the existing water quality monitoring programs, and further support the real-time control and management of chloride in the watershed. MLP-SCA is the first ensemble learning model for river chloride prediction and can be extended to other river systems for water quality prediction.