Jared Williams, John C. Stella, M. B. Singer, Adam M. Lambert, Steven L. Voelker, John E. Drake, Jonathan M. Friedman, L. Pelletier, Li Kui, Dar A. Roberts
Drought‐induced groundwater decline and warming associated with climate change are primary threats to dryland riparian woodlands. We used the extreme 2012–2019 drought in southern California as a natural experiment to assess how differences in water‐use strategies and groundwater dependence may influence the drought susceptibility of dryland riparian tree species with overlapping distributions. We analyzed tree‐ring stable carbon and oxygen isotopes collected from two cottonwood species (Populus trichocarpa and P. fremontii) along the semi‐arid Santa Clara River. We also modeled tree source water δ18O composition to compare with observed source water δ18O within the floodplain to infer patterns of groundwater reliance. Our results suggest that both species functioned as facultative phreatophytes that used shallow soil moisture when available but ultimately relied on groundwater to maintain physiological function during drought. We also observed apparent species differences in water‐use strategies and groundwater dependence related to their regional distributions. P. fremontii was constrained to more arid river segments and ostensibly used a greater proportion of groundwater to satisfy higher evaporative demand. P. fremontii maintained ∆13C at pre‐drought levels up until the peak of the drought, when trees experienced a precipitous decline in ∆13C. This response pattern suggests that trees prioritized maintaining photosynthetic processes over hydraulic safety, until a critical point. In contrast, P. trichocarpa showed a more gradual and sustained reduction in ∆13C, indicating that drought conditions induced stomatal closure and higher water use efficiency. This strategy may confer drought avoidance for P. trichocarpa while increasing its susceptibility to anticipated climate warming.
{"title":"Seasonal and Species‐Level Water‐Use Strategies and Groundwater Dependence in Dryland Riparian Woodlands During Extreme Drought","authors":"Jared Williams, John C. Stella, M. B. Singer, Adam M. Lambert, Steven L. Voelker, John E. Drake, Jonathan M. Friedman, L. Pelletier, Li Kui, Dar A. Roberts","doi":"10.1029/2023wr035928","DOIUrl":"https://doi.org/10.1029/2023wr035928","url":null,"abstract":"Drought‐induced groundwater decline and warming associated with climate change are primary threats to dryland riparian woodlands. We used the extreme 2012–2019 drought in southern California as a natural experiment to assess how differences in water‐use strategies and groundwater dependence may influence the drought susceptibility of dryland riparian tree species with overlapping distributions. We analyzed tree‐ring stable carbon and oxygen isotopes collected from two cottonwood species (Populus trichocarpa and P. fremontii) along the semi‐arid Santa Clara River. We also modeled tree source water δ18O composition to compare with observed source water δ18O within the floodplain to infer patterns of groundwater reliance. Our results suggest that both species functioned as facultative phreatophytes that used shallow soil moisture when available but ultimately relied on groundwater to maintain physiological function during drought. We also observed apparent species differences in water‐use strategies and groundwater dependence related to their regional distributions. P. fremontii was constrained to more arid river segments and ostensibly used a greater proportion of groundwater to satisfy higher evaporative demand. P. fremontii maintained ∆13C at pre‐drought levels up until the peak of the drought, when trees experienced a precipitous decline in ∆13C. This response pattern suggests that trees prioritized maintaining photosynthetic processes over hydraulic safety, until a critical point. In contrast, P. trichocarpa showed a more gradual and sustained reduction in ∆13C, indicating that drought conditions induced stomatal closure and higher water use efficiency. This strategy may confer drought avoidance for P. trichocarpa while increasing its susceptibility to anticipated climate warming.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"15 9","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140785133","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In fractured and stress‐sensitive reservoirs and aquifers, hydromechanical coupling is important, in connection with their heat and solute transport properties, and because the fluid production or extraction leads to land subsidence and potentially to induced seismicity. Classical dual‐porosity poroelasticity (DPP) models cannot upscale pressure diffusion and deformation in fractured porous media, which are characterized by anomalous behaviors that manifest in strong tailing in the temporal evolution of flow rate and subsidence. We study these behaviors using detailed numerical simulations of fluid production in naturally fractured formations characterized by multi‐Gaussian distributions of the matrix permeability. We find that the tailing behaviors depend on the permeability contrast between fracture and matrix, on the permeability distribution in the matrix, and on the correlation length. We use a non–equilibrium, multi‐porosity model to quantify the coupled behaviors of anomalous pressure diffusion, fluid flow and deformation. The model is parameterized by medium and fluid properties, which set the characteristic pressure diffusion time scales. It allows to identify the emerging scaling regimes and scaling behaviors of flow rate and subsidence. We propose a model implementation that captures the full anomalous evolution of flow rates and displacements observed in the detailed numerical simulations in terms of the permeability distribution and matrix length scales. The presented results shed new light on the controls of medium heterogeneity and geometry on pressure diffusion, fluid production and subsidence in highly heterogeneous fractured media.
{"title":"Anomalous Pressure Diffusion and Deformation in Two‐ and Three‐Dimensional Heterogeneous Fractured Media","authors":"Sandro Andrés, M. Dentz, Luis Cueto‐Felgueroso","doi":"10.1029/2023wr036529","DOIUrl":"https://doi.org/10.1029/2023wr036529","url":null,"abstract":"In fractured and stress‐sensitive reservoirs and aquifers, hydromechanical coupling is important, in connection with their heat and solute transport properties, and because the fluid production or extraction leads to land subsidence and potentially to induced seismicity. Classical dual‐porosity poroelasticity (DPP) models cannot upscale pressure diffusion and deformation in fractured porous media, which are characterized by anomalous behaviors that manifest in strong tailing in the temporal evolution of flow rate and subsidence. We study these behaviors using detailed numerical simulations of fluid production in naturally fractured formations characterized by multi‐Gaussian distributions of the matrix permeability. We find that the tailing behaviors depend on the permeability contrast between fracture and matrix, on the permeability distribution in the matrix, and on the correlation length. We use a non–equilibrium, multi‐porosity model to quantify the coupled behaviors of anomalous pressure diffusion, fluid flow and deformation. The model is parameterized by medium and fluid properties, which set the characteristic pressure diffusion time scales. It allows to identify the emerging scaling regimes and scaling behaviors of flow rate and subsidence. We propose a model implementation that captures the full anomalous evolution of flow rates and displacements observed in the detailed numerical simulations in terms of the permeability distribution and matrix length scales. The presented results shed new light on the controls of medium heterogeneity and geometry on pressure diffusion, fluid production and subsidence in highly heterogeneous fractured media.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"365 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140758581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Analog methods (AMs) have long been used for precipitation prediction and climate studies. However, they rely on manual selections of parameters, such as predictor variables and analogy criteria. Previous work showed the potential of genetic algorithms (GAs) to optimize most of the AM parameters. This research goes one step further and investigates the potential of GAs for automating the selection of the input variables and the analogy criteria (distance metric between two data fields) in AMs. Our study focuses on the prediction of daily precipitation in central Europe, specifically Switzerland, as a representative case. Comparative analysis against established methods demonstrates the superiority of GA‐optimized AMs in terms of predictive accuracy. The selected input variables exhibit strong associations with key meteorological processes that influence the generation of precipitation. Further, we identify a new analogy criterion inspired by the Teweles‐Wobus criterion, which consistently performs better than other Euclidean distances and could be used in classic AMs. In contrast to conventional stepwise selection approaches, GA‐optimized AMs display a preference for a flatter structure characterized by a single level of analogy and an increased number of variables. Overall, our study demonstrates the successful application of GAs in automating input variable selection for AMs, with potential implications for application in diverse locations and data exploration to predict alternative predictands. In a broader context, GAs could be used to perform input variable selection in other data‐driven methods, opening perspectives for a broad range of applications.
模拟法（AMs）长期以来一直用于降水预测和气候研究。然而，它们依赖于人工选择参数，如预测变量和类比标准。之前的研究表明，遗传算法（GA）具有优化大部分 AM 参数的潜力。本研究则更进一步，研究了遗传算法在自动监测中自动选择输入变量和类比标准（两个数据字段之间的距离度量）的潜力。我们的研究以欧洲中部（特别是瑞士）的日降水量预测为代表性案例。与已有方法的对比分析表明，GA 优化的 AM 在预测准确性方面更具优势。所选输入变量与影响降水生成的关键气象过程有密切联系。此外，我们受 Teweles-Wobus 准则的启发，确定了一种新的类比准则，其性能始终优于其他欧氏距离，可用于经典的 AMs。与传统的逐步选择方法相比，GA 优化的 AMs 更倾向于以单级类比和增加变量数量为特征的扁平结构。总之，我们的研究证明了在自动测试输入变量选择自动化中应用遗传算法是成功的，这对应用于不同地点和数据探索以预测替代预测因子具有潜在的意义。在更广泛的背景下，遗传效应可以用于在其他数据驱动方法中执行输入变量选择，为广泛的应用开辟前景。
{"title":"Automated Input Variable Selection for Analog Methods Using Genetic Algorithms","authors":"P. Horton, O. Martius, S. Grimm","doi":"10.1029/2023wr035715","DOIUrl":"https://doi.org/10.1029/2023wr035715","url":null,"abstract":"Analog methods (AMs) have long been used for precipitation prediction and climate studies. However, they rely on manual selections of parameters, such as predictor variables and analogy criteria. Previous work showed the potential of genetic algorithms (GAs) to optimize most of the AM parameters. This research goes one step further and investigates the potential of GAs for automating the selection of the input variables and the analogy criteria (distance metric between two data fields) in AMs. Our study focuses on the prediction of daily precipitation in central Europe, specifically Switzerland, as a representative case. Comparative analysis against established methods demonstrates the superiority of GA‐optimized AMs in terms of predictive accuracy. The selected input variables exhibit strong associations with key meteorological processes that influence the generation of precipitation. Further, we identify a new analogy criterion inspired by the Teweles‐Wobus criterion, which consistently performs better than other Euclidean distances and could be used in classic AMs. In contrast to conventional stepwise selection approaches, GA‐optimized AMs display a preference for a flatter structure characterized by a single level of analogy and an increased number of variables. Overall, our study demonstrates the successful application of GAs in automating input variable selection for AMs, with potential implications for application in diverse locations and data exploration to predict alternative predictands. In a broader context, GAs could be used to perform input variable selection in other data‐driven methods, opening perspectives for a broad range of applications.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"432 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140780611","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
River channels shape landscapes through gradual migration and abrupt avulsion. Measuring the motion of braided rivers, which have multiple channel threads, is particularly challenging, limiting predictions for landscape evolution and fluvial architecture. To address this challenge, we extended the capabilities of image‐based particle image velocimetry (PIV)—a technique for tracking channel threads in images of the surface—by adapting it to analyze topographic change. We applied this method in a laboratory experiment where a straight channel set in non‐cohesive sediment evolved into a braided channel under constant water and sediment fluxes. Topography‐based PIV successfully tracked the motion of channel threads if displacements between observations were less than the channel‐thread width, consistent with earlier results from image‐based PIV. We filtered spurious migration vectors with magnitudes less than the elevation grid spacing, or with high uncertainties in magnitude and/or direction. During braided channel initiation, migration rates varied with the channel planform development, showing an increase as incipient meanders developed, a decrease during the transitional braiding phase, and consistently low values during the established braiding phase. In this experimental setup, migration rates varied quasi‐periodically along stream at the half scale of initial meander bends. Lateral migration with respect to the mean flow direction was much more pronounced than streamwise migration, accounting for approximately 80% of all detected motion. Results demonstrate that topography‐based PIV has the potential to advance predictions for bank erosion and landscape evolution in natural braided rivers as well as bar preservation and stratigraphic architecture in geological records.
{"title":"Topography‐Based Particle Image Velocimetry of Braided Channel Initiation","authors":"Youwei Wang, Ajay B. Limaye, A. Chadwick","doi":"10.1029/2023wr035229","DOIUrl":"https://doi.org/10.1029/2023wr035229","url":null,"abstract":"River channels shape landscapes through gradual migration and abrupt avulsion. Measuring the motion of braided rivers, which have multiple channel threads, is particularly challenging, limiting predictions for landscape evolution and fluvial architecture. To address this challenge, we extended the capabilities of image‐based particle image velocimetry (PIV)—a technique for tracking channel threads in images of the surface—by adapting it to analyze topographic change. We applied this method in a laboratory experiment where a straight channel set in non‐cohesive sediment evolved into a braided channel under constant water and sediment fluxes. Topography‐based PIV successfully tracked the motion of channel threads if displacements between observations were less than the channel‐thread width, consistent with earlier results from image‐based PIV. We filtered spurious migration vectors with magnitudes less than the elevation grid spacing, or with high uncertainties in magnitude and/or direction. During braided channel initiation, migration rates varied with the channel planform development, showing an increase as incipient meanders developed, a decrease during the transitional braiding phase, and consistently low values during the established braiding phase. In this experimental setup, migration rates varied quasi‐periodically along stream at the half scale of initial meander bends. Lateral migration with respect to the mean flow direction was much more pronounced than streamwise migration, accounting for approximately 80% of all detected motion. Results demonstrate that topography‐based PIV has the potential to advance predictions for bank erosion and landscape evolution in natural braided rivers as well as bar preservation and stratigraphic architecture in geological records.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"312 2","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140776418","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This study developed and evaluated 30‐m daily evapotranspiration (ET) estimates using the Priestley‐Taylor Jet Propulsion Laboratory (PT‐JPL) model with ECOSTRESS, Moderate MODIS, harmonized Landsat Sentinel‐2 (HLS) imagery, ERA5‐Land reanalysis, and eddy covariance measurements. The new daily 30‐m ET showed significantly improved performance (overall, r = 0.8, RMSE = 1.736, KGE = 0.466) at 145 EC sites over contiguous United States compared to the current 70‐m ECOSTRESS ET (overall, r = 0.485, RMSE = 4.696, KGE = −0.841). A deep neural network postprocessing model trained with ET measurements from EC sites further improved the performance on test sites that were not used for model training (overall, r = 0.842, RMSE = 0.88, KGE = 0.792). The 30‐m ET estimation biases were significantly related to the biases in the upwelling longwave (RUL) and downwelling shortwave radiation (RDS) inputs, with ET estimates driven by MODIS radiation showing higher biases compared to those driven by ERA5‐Land radiation. The error diagnosis using random forest indicates that ET biases tend to be larger under higher ET estimates, and RUL and RDS were the primary contributors to the high bias at the higher ET ranges, with partial dependence plots revealing that the estimation biases tend to be higher under more humid environment, denser vegetation covers, and high net radiation conditions. In conclusion, higher spatial resolution satellite imagery of vegetation characteristics and higher temporal resolution radiation data, combined with continent‐wide EC measurements and deep learning, provided substantial added value for improving ET estimations at the field scale (30‐m).
{"title":"Improved 30‐m Evapotranspiration Estimates Over 145 Eddy Covariance Sites in the Contiguous United States: The Role of ECOSTRESS, Harmonized Landsat Sentinel‐2 Imagery, Climate Reanalysis, and Deep Neural Network Postprocessing","authors":"Taufiq Rashid, D. Tian","doi":"10.1029/2023wr036313","DOIUrl":"https://doi.org/10.1029/2023wr036313","url":null,"abstract":"This study developed and evaluated 30‐m daily evapotranspiration (ET) estimates using the Priestley‐Taylor Jet Propulsion Laboratory (PT‐JPL) model with ECOSTRESS, Moderate MODIS, harmonized Landsat Sentinel‐2 (HLS) imagery, ERA5‐Land reanalysis, and eddy covariance measurements. The new daily 30‐m ET showed significantly improved performance (overall, r = 0.8, RMSE = 1.736, KGE = 0.466) at 145 EC sites over contiguous United States compared to the current 70‐m ECOSTRESS ET (overall, r = 0.485, RMSE = 4.696, KGE = −0.841). A deep neural network postprocessing model trained with ET measurements from EC sites further improved the performance on test sites that were not used for model training (overall, r = 0.842, RMSE = 0.88, KGE = 0.792). The 30‐m ET estimation biases were significantly related to the biases in the upwelling longwave (RUL) and downwelling shortwave radiation (RDS) inputs, with ET estimates driven by MODIS radiation showing higher biases compared to those driven by ERA5‐Land radiation. The error diagnosis using random forest indicates that ET biases tend to be larger under higher ET estimates, and RUL and RDS were the primary contributors to the high bias at the higher ET ranges, with partial dependence plots revealing that the estimation biases tend to be higher under more humid environment, denser vegetation covers, and high net radiation conditions. In conclusion, higher spatial resolution satellite imagery of vegetation characteristics and higher temporal resolution radiation data, combined with continent‐wide EC measurements and deep learning, provided substantial added value for improving ET estimations at the field scale (30‐m).","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"63 2","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140794651","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The Mackenzie River Delta in Canada is a mediator of hydrological transport between the expansive Mackenzie River watershed and the Beaufort Sea. Within the delta, lakes frequently act as water and sediment traps, limiting or delaying the movement of material to the coastal ocean. The degree to which this filtering takes place depends on the ease with which sediment‐laden water is transported from distributary channels into deltaic lakes, referred to as functional lake‐to‐channel connectivity, which varies both spatially and temporally. Tracking of connectivity has previously been limited to either small regions of the delta or has focused on a snapshot of connectivity at a single instance in time. Here we describe an algorithm that uses Landsat imagery to track summertime functional lake‐to‐channel connectivity of 10,362 lakes between 1984 and 2022 on an image‐by‐image basis. We calculate a total average connected lake area of 1400.7 km2 during the 2 weeks after peak discharge, 763.6 km2 higher than previous estimates, suggesting a larger influence of connected lakes on water movement through the delta than previously estimated. We also identify water level thresholds that lead to the initiation of high sediment river water movement into 5,989 lakes (908 lakes with uncertainty ≤±0.5 m), and identify an additional 2899 lakes whose connectivity does not vary at all. As the Arctic hydrological cycle responds to climate change, this work lays a foundation for tracking the movement of water, and the matter it carries, from the Mackenzie River watershed to the Beaufort Sea.
{"title":"Remote Sensing of Multitemporal Functional Lake‐To‐Channel Connectivity and Implications for Water Movement Through the Mackenzie River Delta, Canada","authors":"W. Dolan, T. Pavelsky, A. Piliouras","doi":"10.1029/2023wr036614","DOIUrl":"https://doi.org/10.1029/2023wr036614","url":null,"abstract":"The Mackenzie River Delta in Canada is a mediator of hydrological transport between the expansive Mackenzie River watershed and the Beaufort Sea. Within the delta, lakes frequently act as water and sediment traps, limiting or delaying the movement of material to the coastal ocean. The degree to which this filtering takes place depends on the ease with which sediment‐laden water is transported from distributary channels into deltaic lakes, referred to as functional lake‐to‐channel connectivity, which varies both spatially and temporally. Tracking of connectivity has previously been limited to either small regions of the delta or has focused on a snapshot of connectivity at a single instance in time. Here we describe an algorithm that uses Landsat imagery to track summertime functional lake‐to‐channel connectivity of 10,362 lakes between 1984 and 2022 on an image‐by‐image basis. We calculate a total average connected lake area of 1400.7 km2 during the 2 weeks after peak discharge, 763.6 km2 higher than previous estimates, suggesting a larger influence of connected lakes on water movement through the delta than previously estimated. We also identify water level thresholds that lead to the initiation of high sediment river water movement into 5,989 lakes (908 lakes with uncertainty ≤±0.5 m), and identify an additional 2899 lakes whose connectivity does not vary at all. As the Arctic hydrological cycle responds to climate change, this work lays a foundation for tracking the movement of water, and the matter it carries, from the Mackenzie River watershed to the Beaufort Sea.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"749 ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140782380","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Model predictive control (MPC) is used to manage water systems, and its performance depends on the (internal or control‐oriented) model it is based on. Several models for the hydraulics of open water systems are presented in literature and used in applications, but their performance has not yet been investigated systematically, and no guideline exists on which model to select for a certain channel. The aim of this research is to present a guideline for model choice based on the geometry of the channel and the flow conditions. The guideline is developed by first categorizing the channels into four types, followed by performing time‐domain, frequency domain, and closed‐loop tests for all models and channel types. The evaluation of the tests shows that for short and wave‐dominated channels, the Muskingum, Integrator Delay, and Integrator Delay Zero models perform the best, while for longer channels the linear inertial model is the most suitable. Finally, a decision‐tree is presented how to choose the model. Lastly, a decision‐tree is introduced to aid in the selection of the most appropriate model.
模型预测控制（MPC）用于水系统管理，其性能取决于所依据的（内部或面向控制的）模型。文献中介绍了几种开放式水系的水力学模型，并已在应用中使用，但尚未对这些模型的性能进行系统研究，也不存在针对特定渠道选择哪种模型的指导原则。本研究的目的是根据水道的几何形状和水流条件，提出选择模型的指导原则。制定该指南时，首先将渠道分为四种类型，然后对所有模型和渠道类型进行时域、频域和闭环测试。测试评估结果表明，对于短水道和以波浪为主的水道，Muskingum、Integrator Delay 和 Integrator Delay Zero 模型的性能最佳，而对于长水道，线性惯性模型最为合适。最后，介绍了如何选择模型的决策树。最后，介绍了一种决策树，以帮助选择最合适的模型。
{"title":"How to Choose Suitable Physics‐Based Models Without Tuning and System Identification for Model‐Predictive Control of Open Water Channels?","authors":"K. Horváth, B. V. van Esch, I. Pothof","doi":"10.1029/2023wr035687","DOIUrl":"https://doi.org/10.1029/2023wr035687","url":null,"abstract":"Model predictive control (MPC) is used to manage water systems, and its performance depends on the (internal or control‐oriented) model it is based on. Several models for the hydraulics of open water systems are presented in literature and used in applications, but their performance has not yet been investigated systematically, and no guideline exists on which model to select for a certain channel. The aim of this research is to present a guideline for model choice based on the geometry of the channel and the flow conditions. The guideline is developed by first categorizing the channels into four types, followed by performing time‐domain, frequency domain, and closed‐loop tests for all models and channel types. The evaluation of the tests shows that for short and wave‐dominated channels, the Muskingum, Integrator Delay, and Integrator Delay Zero models perform the best, while for longer channels the linear inertial model is the most suitable. Finally, a decision‐tree is presented how to choose the model. Lastly, a decision‐tree is introduced to aid in the selection of the most appropriate model.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"18 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140775081","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Chao Wang, Shijie Jiang, Yi Zheng, Feng Han, Rohini Kumar, O. Rakovec, Siqi Li
While deep learning (DL) models exhibit superior simulation accuracy over traditional distributed hydrological models (DHMs), their main limitations lie in opacity and the absence of underlying physical mechanisms. The pursuit of synergies between DL and DHMs is an engaging research domain, yet a definitive roadmap remains elusive. In this study, a novel framework that seamlessly integrates a process‐based hydrological model encoded as a neural network (NN), an additional NN for mapping spatially distributed and physically meaningful parameters from watershed attributes, and NN‐based replacement models representing inadequately understood processes is developed. Multi‐source observations are used as training data, and the framework is fully differentiable, enabling fast parameter tuning by backpropagation. A hybrid DL model of the Amazon Basin (∼6 × 106 km2) was established based on the framework, and HydroPy, a global‐scale DHM, was encoded as its physical backbone. Trained simultaneously with streamflow observations and Gravity Recovery and Climate Experiment satellite data, the hybrid model yielded median Nash‐Sutcliffe efficiencies of 0.83 and 0.77 for dynamic and distributed simulations of streamflow and total water storage, respectively, 41% and 35% higher than those of the original HydroPy model. Replacing the original Penman‒Monteith formulation in HydroPy with a replacement NN produces more plausible potential evapotranspiration (PET) estimates, and unravels the spatial pattern of PET in this giant basin. The NN used for parameterization was interpreted to identify the factors controlling the spatial variability in key parameters. Overall, this study lays out a feasible technical roadmap for distributed hydrological modeling in the big data era.
与传统的分布式水文模型（DHMs）相比，深度学习（DL）模型显示出更高的模拟精度，但其主要局限性在于不透明性和缺乏基本物理机制。追求 DL 与 DHM 之间的协同效应是一个引人入胜的研究领域，但明确的路线图仍遥遥无期。在本研究中，开发了一个新颖的框架，该框架无缝集成了以神经网络（NN）编码的基于过程的水文模型、用于从流域属性映射空间分布和物理意义参数的附加 NN，以及代表不充分理解的过程的基于 NN 的替代模型。多源观测数据被用作训练数据，该框架是完全可微分的，可通过反向传播快速调整参数。基于该框架建立了亚马逊流域（6×106 平方公里）的混合 DL 模型，并将全球尺度 DHM HydroPy 作为其物理骨干进行编码。通过同时使用流场观测数据和重力恢复与气候实验卫星数据进行训练，混合模型在动态和分布式模拟流场和总蓄水量时的纳什-萨特克利夫效率中值分别为 0.83 和 0.77，比原始 HydroPy 模型分别高出 41% 和 35%。用替代 NN 取代 HydroPy 中的原始 Penman-Monteith 公式，可得出更合理的潜在蒸散量（PET）估算值，并揭示了这一巨大盆地中潜在蒸散量的空间模式。对用于参数化的 NN 进行了解释，以确定控制关键参数空间变化的因素。总之，这项研究为大数据时代的分布式水文建模提供了可行的技术路线图。
{"title":"Distributed Hydrological Modeling With Physics‐Encoded Deep Learning: A General Framework and Its Application in the Amazon","authors":"Chao Wang, Shijie Jiang, Yi Zheng, Feng Han, Rohini Kumar, O. Rakovec, Siqi Li","doi":"10.1029/2023wr036170","DOIUrl":"https://doi.org/10.1029/2023wr036170","url":null,"abstract":"While deep learning (DL) models exhibit superior simulation accuracy over traditional distributed hydrological models (DHMs), their main limitations lie in opacity and the absence of underlying physical mechanisms. The pursuit of synergies between DL and DHMs is an engaging research domain, yet a definitive roadmap remains elusive. In this study, a novel framework that seamlessly integrates a process‐based hydrological model encoded as a neural network (NN), an additional NN for mapping spatially distributed and physically meaningful parameters from watershed attributes, and NN‐based replacement models representing inadequately understood processes is developed. Multi‐source observations are used as training data, and the framework is fully differentiable, enabling fast parameter tuning by backpropagation. A hybrid DL model of the Amazon Basin (∼6 × 106 km2) was established based on the framework, and HydroPy, a global‐scale DHM, was encoded as its physical backbone. Trained simultaneously with streamflow observations and Gravity Recovery and Climate Experiment satellite data, the hybrid model yielded median Nash‐Sutcliffe efficiencies of 0.83 and 0.77 for dynamic and distributed simulations of streamflow and total water storage, respectively, 41% and 35% higher than those of the original HydroPy model. Replacing the original Penman‒Monteith formulation in HydroPy with a replacement NN produces more plausible potential evapotranspiration (PET) estimates, and unravels the spatial pattern of PET in this giant basin. The NN used for parameterization was interpreted to identify the factors controlling the spatial variability in key parameters. Overall, this study lays out a feasible technical roadmap for distributed hydrological modeling in the big data era.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"112 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140783292","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Anna Pölz, A. Blaschke, J. Komma, A. Farnleitner, J. Derx
Karst springs are essential drinking water resources, however, modeling them poses challenges due to complex subsurface flow processes. Deep learning models can capture complex relationships due to their ability to learn non‐linear patterns. This study evaluates the performance of the Transformer in forecasting spring discharges for up to 4 days. We compare it to the Long Short‐Term Memory (LSTM) Neural Network and a common baseline model on a well‐studied Austrian karst spring (LKAS2) with an extensive hourly database. We evaluated the models for two further karst springs with diverse discharge characteristics for comparing the performances based on four metrics. In the discharge‐based scenario, the Transformer performed significantly better than the LSTM for the spring with the longest response times (9% mean difference across metrics), while it performed poorer for the spring with the shortest response time (4% difference). Moreover, the Transformer better predicted the shape of the discharge during snowmelt. Both models performed well across all lead times and springs with 0.64–0.92 for the Nash–Sutcliffe efficiency and 10.8%–28.7% for the symmetric mean absolute percentage error for the LKAS2 spring. The temporal information, rainfall and electrical conductivity were the controlling input variables for the non‐discharge based scenario. The uncertainty analysis revealed that the prediction intervals are smallest in winter and autumn and highest during snowmelt. Our results thus suggest that the Transformer is a promising model to support the drinking water abstraction management, and can have advantages due to its attention mechanism particularly for longer response times.
{"title":"Transformer Versus LSTM: A Comparison of Deep Learning Models for Karst Spring Discharge Forecasting","authors":"Anna Pölz, A. Blaschke, J. Komma, A. Farnleitner, J. Derx","doi":"10.1029/2022wr032602","DOIUrl":"https://doi.org/10.1029/2022wr032602","url":null,"abstract":"Karst springs are essential drinking water resources, however, modeling them poses challenges due to complex subsurface flow processes. Deep learning models can capture complex relationships due to their ability to learn non‐linear patterns. This study evaluates the performance of the Transformer in forecasting spring discharges for up to 4 days. We compare it to the Long Short‐Term Memory (LSTM) Neural Network and a common baseline model on a well‐studied Austrian karst spring (LKAS2) with an extensive hourly database. We evaluated the models for two further karst springs with diverse discharge characteristics for comparing the performances based on four metrics. In the discharge‐based scenario, the Transformer performed significantly better than the LSTM for the spring with the longest response times (9% mean difference across metrics), while it performed poorer for the spring with the shortest response time (4% difference). Moreover, the Transformer better predicted the shape of the discharge during snowmelt. Both models performed well across all lead times and springs with 0.64–0.92 for the Nash–Sutcliffe efficiency and 10.8%–28.7% for the symmetric mean absolute percentage error for the LKAS2 spring. The temporal information, rainfall and electrical conductivity were the controlling input variables for the non‐discharge based scenario. The uncertainty analysis revealed that the prediction intervals are smallest in winter and autumn and highest during snowmelt. Our results thus suggest that the Transformer is a promising model to support the drinking water abstraction management, and can have advantages due to its attention mechanism particularly for longer response times.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"45 4","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140766934","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Freshwater chloride concentrations have been increasing in North American surface waters for decades, largely driven by increases in the use of road salt, which is commonly applied as a deicer. In Chicago, thousands of tons of road salt are applied to roadways each winter, and increases in surface water chloride concentrations have been noted across the region since the mid‐1960s. While much of the applied salt runs directly off to nearby waterways during snowmelt events, some percolates to groundwater, affecting public supply wells and increasing the amount of chloride released to streams as baseflow during the non‐salting season. In the present study we have developed a spatially distributed chloride mass balance across the Chicago Metropolitan Statistical Area (CMSA) for a 30‐year period (1990–2020) to better our understanding of long‐term chloride fluxes and storage. Our results show that inputs of road salt to the region increased by 33% between 1990 and 2020. During this same period, riverine chloride loads across the region increased by 60%. Despite these increases in riverine chloride export, we find that chloride is accumulating in CMSA groundwater at a rate of ∼480 ktons year−1. We show that shallow aquifers, <30 m, exhibit only seasonal chloride storage, without long‐term accumulation. In contrast, at depths below 30 m, we find chloride concentrations to be increasing over time, indicating that legacy chloride is accumulating at deeper depths in CMSA groundwater. The present results highlight the importance of legacy chloride to long‐term water quality dynamics in North American cities.
{"title":"Road Salt Legacies: Quantifying Fluxes of Chloride to Groundwater and Surface Water Across the Chicago Metropolitan Statistical Area","authors":"K. V. Van Meter, E. Ceisel","doi":"10.1029/2023wr035103","DOIUrl":"https://doi.org/10.1029/2023wr035103","url":null,"abstract":"Freshwater chloride concentrations have been increasing in North American surface waters for decades, largely driven by increases in the use of road salt, which is commonly applied as a deicer. In Chicago, thousands of tons of road salt are applied to roadways each winter, and increases in surface water chloride concentrations have been noted across the region since the mid‐1960s. While much of the applied salt runs directly off to nearby waterways during snowmelt events, some percolates to groundwater, affecting public supply wells and increasing the amount of chloride released to streams as baseflow during the non‐salting season. In the present study we have developed a spatially distributed chloride mass balance across the Chicago Metropolitan Statistical Area (CMSA) for a 30‐year period (1990–2020) to better our understanding of long‐term chloride fluxes and storage. Our results show that inputs of road salt to the region increased by 33% between 1990 and 2020. During this same period, riverine chloride loads across the region increased by 60%. Despite these increases in riverine chloride export, we find that chloride is accumulating in CMSA groundwater at a rate of ∼480 ktons year−1. We show that shallow aquifers, <30 m, exhibit only seasonal chloride storage, without long‐term accumulation. In contrast, at depths below 30 m, we find chloride concentrations to be increasing over time, indicating that legacy chloride is accumulating at deeper depths in CMSA groundwater. The present results highlight the importance of legacy chloride to long‐term water quality dynamics in North American cities.","PeriodicalId":507642,"journal":{"name":"Water Resources Research","volume":"247 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140468820","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：