Search

Data uncertainty affects the accuracy of pollution source detection (PSD), particularly in the background of low‐cost water quality sensing and low‐accuracy data challenge. This study aims to develop a novel PSD method to use low‐accuracy sensor data, namely, the method of coupled forward data Assimilation and inverse Optimization in PSD (A&O‐PSD). This approach primarily employs filtering strategies to handle observation errors and extract hidden trend information during forward water quality data assimilation, and then optimal estimation of pollution source information through inverse optimization with enhanced trend information matching, avoiding the non‐Gaussian distribution challenge of pollution source information. Both real‐world pollution events and semi‐synthetic cases were used to evaluate the methodology and compare its performance with the traditional optimization approach (T‐PSD). The results indicated that T‐PSD is significantly affected by observational and parameter noise, engendering noticeable biases in PSD under the low‐accuracy sensor conditions. In contrast, the A&O‐PSD could accomplish the estimation task of PSD in real‐world pollution events, with improved robustness against noise interference. Furthermore, A&O‐PSD achieved an accuracy improvement of over 10% compared to T‐PSD in estimating pollution source locations within the typical noise distribution range of most low‐accuracy sensors currently available, making it possible to use low‐accuracy data that would otherwise be unusable in T‐PSD. Overall, the A&O‐PSD method, combined with low‐cost low‐accuracy water quality sensing, offers an effective solution for watershed environmental management. Plain Language Summary: Human activities significantly contribute to water pollution, such as illegal dumping or sudden pollution incidents. These events depend on the decision‐making behavior of polluting enterprises and socio‐economic factors, leading to concealed and uncorrelated pollution discharges with unpredictable outlet distributions. The challenges in PSD also lie in identifying discharges (location, time, and release mass) from limited and inaccurate water quality monitoring. Since the field sampling and laboratory chemical analysis for water quality observations are associated with high labor, financial, and time costs; while in in situ monitoring, the sensor‐based measurement is prone by significant errors, especially in complex, long‐term natural environments. To address the challenges of unknown distributions of pollution source information and low‐accuracy data, we propose the method of coupled forward data Assimilation and inverse Optimization (A&O‐PSD), which can extract credible trend information from inaccurate sensor data and optimally estimate pollution source information. A&O‐PSD has proven effective in both real‐world and semi‐synthetic pollution events, demonstrating robustness against data and parameter uncertainty. It offers a flexible framework for utilizing low‐cost low‐accuracy water quality sensing, with potential implications for sensor hardware development to improve accuracy. Key Points: A pollution source detection (PSD) paradigm of coupled forward assimilation and inverse optimization is proposedThe accuracy of PSD in low accuracy sensor data is improved by trend‐match strategyA&O‐PSD with better robustness tolerates inaccurate sensor data and model parameters [ABSTRACT FROM AUTHOR]

Water conservation is a crucial indicator that measures the available water resources needed for maintaining regional ecological services and socioeconomic development. The Yiluo River Basin plays an essential role in water conservation in the Yellow River Basin, which is one of the most important river basins with vulnerable ecological conditions and a large population in China. However, previous studies have a limited understanding of the distribution of water conservation in the Yiluo River Basin. To address this knowledge gap, we developed a SWAT model to evaluate water conservation in the Yiluo River Basin with high spatial and temporal details on a monthly scale. From a monthly perspective, water conservation accumulation primarily took place in July (54.6 mm), August (23.5 mm), and September (33.2 mm), which are in the flood season. From 1966 to 2018, we found a significant 47% reduction in basin-wide water conservation, and the reduction was primarily influenced by meteorological conditions and underlying surface dynamics. The results of the temporal correlation analysis identified precipitation as the most significant factor influencing water conservation, while the spatial correlation analysis revealed that potential evapotranspiration, vegetation, and elevation had the highest spatial correlation with water conservation. By combining SWAT outputs on the HRU (hydrological response unit) scale with the spatial distribution of HRUs, the study achieved the visualization of the spatial distribution of water conservation, identifying Luonan County, Luanchuan County, and Luoning County as the key regions that experienced water conservation decline over the past decades. These findings advance our understanding of the distributions of water conservation and their key driving factors in the study area and provide valuable policy implications to support ecological protection and water resource management in the Yellow River Basin. [ABSTRACT FROM AUTHOR]

Groundwater contamination induced by anthropogenic activities has long been a global issue. Characterizing and modeling contaminant transport processes is crucial to groundwater protection and management. However, challenges still exist in process complexity, data constraint, and computational cost. In the era of big data, the growth of machine learning has led to new opportunities in studying contaminant transport in groundwater systems. In this work, we introduce a new attention‐based graph neural network (aGNN) for modeling contaminant transport with limited monitoring data and quantifying causal connections between contaminant sources (drivers) and their spreading (outcomes). In five synthetic case studies that involve varying monitoring networks in heterogeneous aquifers, aGNN is shown to outperform LSTM‐based (long‐short term memory) and CNN‐ based (convolutional neural network) methods in multistep predictions (i.e., transductive learning). It also demonstrates a high level of applicability in inferring observations for unmonitored sites (i.e., inductive learning). Furthermore, an explanatory analysis based on aGNN quantifies the influence of each contaminant source, which has been validated by a physics‐based model with consistent outcomes with an R2 value exceeding 92%. The major advantage of aGNN is that it not only has a high level of predictive power in multiple scenario evaluations but also substantially reduces computational cost. Overall, this study shows that aGNN is efficient and robust for highly nonlinear spatiotemporal learning in subsurface contaminant transport, and provides a promising tool for groundwater management involving contaminant source attribution. Plain Language Summary: Groundwater contamination caused by human activities is a longstanding global challenge. Accurately characterizing and modeling the movement of contaminants is crucial for the protection and management of groundwater resources. However, the complexity of the processes, limitations in data availability, and high computational demands pose significant challenges. In the age of big data, machine learning offers new avenues for exploring contaminant transport in groundwater. In this study, we introduce a novel machine learning model called an attention‐based graph neural network (aGNN) designed to model contaminant transport with sparse monitoring data and to analyze the causal relationships between contaminant sources and observed concentrations at specific locations. We conducted five synthetic case studies across diverse aquifer systems with varying monitoring setups, where aGNN demonstrated superior performance over models based on other approaches. It also proved highly capable of making inferences about pollution levels at unmonitored sites. Moreover, an explanatory analysis using aGNN effectively quantified the impact of each contaminant source, with results validated by a physics‐based model. Overall, this study establishes aGNN as an efficient and robust method for complex spatiotemporal learning in subsurface contaminant transport, making it a valuable tool for groundwater management and contaminant source identification. Key Points: A novel graph‐based deep learning method is proposed for modeling contaminant transport constrained by monitoring dataThe proposed model quantifies the contribution of each potential contaminant source to the observed concentration at an arbitrary locationThe deep learning method substantially reduces the computational cost compared with a physics‐based contaminant transport model [ABSTRACT FROM AUTHOR]

In many flood-prone areas, it is essential for emergency responders to use advanced computer models to assess flood risk and develop informed flood evacuation plans. However, previous studies have had a limited understanding of how evacuation performance is affected by the arrangement of evacuation shelters (with respect to their number and geographical distribution) and human behaviors (with respect to the heterogeneity of household evacuation preparation times and route-searching strategies). In this study, we develop an integrated socio-hydrological modeling framework that couples (1) a hydrodynamic model for flood simulation, (2) an agent-based model for evacuation management policies and human behaviors, and (3) a transportation model for simulating household evacuation processes in a road network. We apply the model to the Xiong'an New Area and examine household evacuation outcomes for various shelter location plans and human behavior scenarios. The results show that household evacuation processes are significantly affected by the number and geographical distribution of evacuation shelters. Surprisingly, we find that establishing more shelters may not improve evacuation results if the shelters are not strategically located. We also find that low heterogeneity in evacuation preparation times can result in heavy traffic congestion and long evacuation clearance times. If each household selects their own shortest route without considering the effects of other evacuees' route choices, traffic congestion will likely occur, thereby reducing system-level evacuation performance. These results demonstrate the unique functionality of our model with respect to supporting flood risk assessment and advancing our understanding of how multiple management and behavioral factors jointly affect evacuation performance. [ABSTRACT FROM AUTHOR]

Flood forecasts and warnings are intended to reduce flood-related property damages and loss of human life. Considerable research has improved flood forecasting accuracy ( e.g., more accurate prediction of the occurrence of flood events) and lead time. However, the delivery of improved forecast information alone is not necessarily sufficient to reduce flood damage and loss of life, as people have varying responses and reactions to flood warnings. This study develops an agent-based modeling framework that evaluates the impacts of heterogeneity in human behaviors ( i.e., variation in behaviors in response to flood warnings), as well as residential density, on the benefits of flood warnings. The framework is coupled with a traffic model to simulate evacuation processes within a road network under various flood warning scenarios. The results show the marginal benefit associated with providing better flood warnings is significantly constrained if people behave in a more risk-tolerant manner, especially in high-density residential areas. The results also show significant impacts of human behavioral heterogeneity on the benefits of flood warnings, and thus stress the importance of considering human behavioral heterogeneity in simulating flood warning-response systems. Further study is suggested to more accurately model human responses and behavioral heterogeneity, as well as to include more attributes of residential areas to estimate and improve the benefits of flood warnings. [ABSTRACT FROM AUTHOR]