Water Resources Research最新文献

{"title":"Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High-Resolution Hydrological Model: Application to the French Mediterranean Region","authors":"Ngo Nghi Truyen Huynh, Pierre-André Garambois, François Colleoni, Benjamin Renard, Hélène Roux, Julie Demargne, Maxime Jay-Allemand, Pierre Javelle","doi":"10.1029/2024wr037544","DOIUrl":"https://doi.org/10.1029/2024wr037544","url":null,"abstract":"Estimating spatially distributed hydrological parameters in ungauged catchments poses a challenging regionalization problem and requires imposing spatial constraints given the sparsity of discharge data. A possible approach is to search for a transfer function that quantitatively relates physical descriptors to conceptual model parameters. This paper introduces a Hybrid Data Assimilation and Parameter Regionalization (HDA-PR) approach incorporating learnable regionalization mappings, based on either multi-linear regression or artificial neural networks (ANNs), into a differentiable hydrological model. This approach demonstrates how two differentiable codes can be linked and their gradients chained, enabling the exploitation of heterogeneous data sets across extensive spatio-temporal computational domains within a high-dimensional regionalization context, using accurate adjoint-based gradients. The inverse problem is tackled with a multi-gauge calibration cost function accounting for information from multiple observation sites. HDA-PR was tested on high-resolution, hourly and kilometric regional modeling of 126 flash-flood-prone catchments in the French Mediterranean region. The results highlight a strong regionalization performance of HDA-PR especially in the most challenging upstream-to-downstream extrapolation scenario with ANN, achieving median Nash-Sutcliffe efficiency (NSE) scores from 0.6 to 0.71 for spatial, temporal, spatio-temporal validations, and improving NSE by up to 30% on average compared to the baseline model calibrated with lumped parameters. Multiple evaluation metrics based on flood-oriented hydrological signatures also indicate that the use of an ANN leads to better performances than a multi-linear regression in a validation context. ANN enables to learn a non-linear descriptors-to-parameters mapping which provides better model controllability than a linear mapping for complex calibration cases.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"41 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142563298","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Spectral Analysis of Hydrological Signals to Estimate Watershed Properties Considering Impacts of Unsaturated Zone","authors":"Yunqiu Zhou, Xiuyu Liang, Enze Ma, Kewei Chen, Keith Schilling, Tianyuan Zheng, Yuhu Zheng, You-Kuan Zhang, Chunmiao Zheng","doi":"10.1029/2023wr036680","DOIUrl":"https://doi.org/10.1029/2023wr036680","url":null,"abstract":"Understanding responses of stream discharge to precipitation in a watershed is important in gaining insights into watershed hydrology and estimating hydraulic parameters. Transfer functions in the spectral domain are commonly used to quantify the relationship between precipitation and discharge, and estimate watershed hydraulic parameters. However, previous models have not adequately accounted for the impact of the unsaturated zone. To address this, we have developed a novel analytical model that considers the effect of the unsaturated zone to obtain transfer functions within watersheds. These transfer functions are derived by the spectral method and verified through numerical simulations. The results indicate that the transfer functions are influenced significantly by the relative hydraulic conductivity exponent <i>α</i>_<i>k</i> in the moisture characteristic curve. A higher <i>α</i>_<i>k</i> results in a lower transfer function, indicating more robust filtering of hydrological signals. A thicker unsaturated zone results in lower transfer functions at higher frequencies. The traditional transfer functions, which neglect the retention capacity of the unsaturated zone, tend to overestimate hydrological responses at high frequencies. Our transfer functions agree well with integrated watershed-scale flow models and are also applied to observed data from four watersheds in Iowa, providing reasonable estimates for the hydraulic parameters. This study contributes to a deeper understanding of watershed behavior and offers an enhanced tool for estimating hydraulic parameters with practical applications.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"9 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142574493","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multivariate Evaluation of Flash Drought Across the United States","authors":"Jason A. Otkin, Yafang Zhong, Trent W. Ford, Martha C. Anderson, Christopher Hain, Andrew Hoell, Mark Svoboda, Hailan Wang","doi":"10.1029/2024wr037333","DOIUrl":"https://doi.org/10.1029/2024wr037333","url":null,"abstract":"This study uses the flash drought intensity index (FDII) to develop a multivariate flash drought climatology for the contiguous U.S. using data from 2001 to 2021. The FDII method uses the rate of intensification (FD-INT) and subsequent drought severity (DRO-SEV) to determine when a flash drought occurred and the strength of the event. Overall, the results showed that flash drought occurrence and severity varied with season and region and were sensitive to the drought indicator used to compute the FDII. Precipitation-based indicators identified more flash droughts across the western U.S. whereas soil moisture (SM) and evapotranspiration indicators identified more flash droughts across the central and eastern U.S. When assessed over the entire U.S., the most flash droughts were found when using an evaporative demand indicator. Though FD-INT was larger than DRO-SEV across the U.S. for most indicators, regional patterns were also evident in their relative importance. For example, a distinct east-west gradient was present in the SM and evapotranspiration FD-INT, with relatively large values in the central and eastern U.S. A combined data set synthesizing information from multiple indicators showed that the strongest flash droughts from a multivariate perspective were located in the central and southeastern U.S. A seasonal analysis revealed a distinct seasonal cycle in flash drought onset across the western and central U.S. Together, the results illustrate the need to use a multivariate framework to identify and characterize the occurrence and severity of flash droughts.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"2 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142563300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Experimental Visualization and Modeling of the Transport Behaviors of Monofilament Microplastic Fibers Through an Idealized Porous Media","authors":"Tyler T. Fouty, Nicholas B. Engdahl","doi":"10.1029/2024wr037901","DOIUrl":"https://doi.org/10.1029/2024wr037901","url":null,"abstract":"Microplastic fibers (MPF) are the largest fraction of microplastics in the environment by mass. The endpoints of these contaminants' movement is generally known at large-scale (i.e., their origins and where they end up), but the mechanics of how they get to those sinks remains poorly understood. The objective of this work was to improve understanding of the mechanisms driving MPF migration through terrestrial systems by directly imaging their motion through idealized representations of porous media. Monofilament line with 0.3 mm diameter was passed through a bench-scale, pseudo-2d flow cell to capture trajectories of MPFs of three different lengths and trajectories of passive micro-bead tracers were also captured. Video processing and automated image analysis converted the video of the experiments into a database of trajectories, allowing comparison of the experimental data to various numerical models. Simple advection-dispersion models were adequate for modeling the passive tracer but did not provide a good description of MPF transport. A physics-based, distributed model was able to generate realistic trajectories through the domain, but the speeds of the fibers in the initial simulation were too fast, despite working well for the passive tracer. Adding a delay (waiting time) process resulted in good description of the trajectories and travel times. The specifics of the delay process could not be deduced from these experiments, but its overall impact on transport provides mechanistic insights. These direct observation of the trajectories and speeds of MPFs moving through porous media show that MPFs likely have strong interactions with their surroundings.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"87 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142566207","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sea-Level Rise, Drinking Water Quality and the Economic Value of Coastal Tourism in North Carolina","authors":"J. C. Whitehead, W. P. Anderson, D. Guignet, C. E. Landry, O. A. Morgan","doi":"10.1029/2023wr036440","DOIUrl":"https://doi.org/10.1029/2023wr036440","url":null,"abstract":"We estimate the economic benefits of avoiding reductions in drinking water quality due to sea level rise accruing to North Carolina (NC) coastal tourists. Using stated preference methods and responses from recent coastal visitors, we find that tourists are 2%, 8%, and 11% less likely to take an overnight trip if drinking water tastes slightly, moderately, or very salty at their chosen destination. The majority of those who decline a trip would take a trip to another NC beach without water quality issues, others would take another type of recreational trip, with a minority opting to stay home. Willingness to pay for an overnight beach trip declines with the salty taste of drinking water. We find evidence of attribute non-attendance in the stated preference data, which impacts the regression model and estimates of the willingness to pay for trips. Combining economic and hydrological models, annual aggregate benefit losses due to low drinking water quality could be as high as $232 million by 2040.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"40 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142541714","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Using Commercial Satellite Imagery to Reconstruct 3 m and Daily Spring Snow Water Equivalent","authors":"Justin M. Pflug, Kehan Yang, Nicoleta Cristea, Emma T. Boudreau, Carrie M. Vuyovich, Sujay V. Kumar","doi":"10.1029/2024wr037983","DOIUrl":"https://doi.org/10.1029/2024wr037983","url":null,"abstract":"Snow water equivalent (SWE) distribution at fine spatial scales (≤10 m) is difficult to estimate due to modeling and observational constraints. However, the distribution of SWE throughout the spring snowmelt season is often correlated to the timing of snow disappearance. Here, we show that snow cover maps generated from PlanetScope's constellation of Dove Satellites can resolve the 3 m date of snow disappearance across seven alpine domains in California and Colorado. Across a 5-year period (2019–2023), the average uncertainty in the date of snow disappearance, or the period of time between the last date of observed snow cover and the first date of observed snow absence, was 3 days. Using a simple shortwave-based snowmelt model calibrated at nearby snow pillows, the PlanetScope date of snow disappearance could be used to reconstruct spring SWE. Relative to lidar SWE estimates, the SWE reconstruction had a spatial coefficient of correlation of 0.75, and SWE spatial variability that was biased by 9%, on average. SWE reconstruction biases were then improved to within 0.04 m, on average, by calibrating snowmelt rates to track the spring temporal evolution of fractional snow cover observed by PlanetScope, including fractional snow cover over the full modeling domain, and across domain subsections where snowmelt rates may differ. This study demonstrates the utility of fine-scale and high-frequency optical observations of snow cover, and the simple and annually repeatable connections between snow cover and spring snow water resources in regions with seasonal snowpack.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"126 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142562017","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Gas Transfer Across Air-Water Interfaces in Inland Waters: From Micro-Eddies to Super-Statistics","authors":"Gabriel Katul, Andrew Bragg, Ivan Mammarella, Heping Liu, Qi Li, Elie Bou-Zeid","doi":"10.1029/2023wr036615","DOIUrl":"https://doi.org/10.1029/2023wr036615","url":null,"abstract":"In inland water covering lakes, reservoirs, and ponds, the gas exchange of slightly soluble gases such as carbon dioxide, dimethyl sulfide, methane, or oxygen across a clean and nearly flat air-water interface is routinely described using a water-side mean gas transfer velocity &lt;span data-altimg=\"/cms/asset/d601d164-4c70-4fb9-8cc9-077f60e993c3/wrcr27452-math-0001.png\"&gt;&lt;/span&gt;&lt;mjx-container ctxtmenu_counter=\"533\" ctxtmenu_oldtabindex=\"1\" jax=\"CHTML\" role=\"application\" sre-explorer- style=\"font-size: 103%; position: relative;\" tabindex=\"0\"&gt;&lt;mjx-math aria-hidden=\"true\" location=\"graphic/wrcr27452-math-0001.png\"&gt;&lt;mjx-semantics&gt;&lt;mjx-mrow&gt;&lt;mjx-mrow&gt;&lt;mjx-mover data-semantic-children=\"2,3\" data-semantic- data-semantic-role=\"latinletter\" data-semantic-speech=\"k Subscript upper L Baseline overbar\" data-semantic-type=\"overscore\"&gt;&lt;mjx-over style=\"padding-bottom: 0.105em; margin-bottom: -0.544em;\"&gt;&lt;mjx-mo data-semantic- data-semantic-parent=\"4\" data-semantic-role=\"overaccent\" data-semantic-type=\"punctuation\"&gt;&lt;mjx-stretchy-h style=\"width: 1.086em;\"&gt;&lt;mjx-ext&gt;&lt;mjx-c&gt;&lt;/mjx-c&gt;&lt;/mjx-ext&gt;&lt;/mjx-stretchy-h&gt;&lt;/mjx-mo&gt;&lt;/mjx-over&gt;&lt;mjx-base&gt;&lt;mjx-msub data-semantic-children=\"0,1\" data-semantic- data-semantic-parent=\"4\" data-semantic-role=\"latinletter\" data-semantic-type=\"subscript\"&gt;&lt;mjx-mi data-semantic-annotation=\"clearspeak:simple\" data-semantic-font=\"italic\" data-semantic- data-semantic-parent=\"2\" data-semantic-role=\"latinletter\" data-semantic-type=\"identifier\"&gt;&lt;mjx-c&gt;&lt;/mjx-c&gt;&lt;/mjx-mi&gt;&lt;mjx-script style=\"vertical-align: -0.15em;\"&gt;&lt;mjx-mi data-semantic-annotation=\"clearspeak:simple\" data-semantic-font=\"italic\" data-semantic- data-semantic-parent=\"2\" data-semantic-role=\"latinletter\" data-semantic-type=\"identifier\" size=\"s\"&gt;&lt;mjx-c&gt;&lt;/mjx-c&gt;&lt;/mjx-mi&gt;&lt;/mjx-script&gt;&lt;/mjx-msub&gt;&lt;/mjx-base&gt;&lt;/mjx-mover&gt;&lt;/mjx-mrow&gt;&lt;/mjx-mrow&gt;&lt;/mjx-semantics&gt;&lt;/mjx-math&gt;&lt;mjx-assistive-mml display=\"inline\" unselectable=\"on\"&gt;&lt;math altimg=\"urn:x-wiley:00431397:media:wrcr27452:wrcr27452-math-0001\" display=\"inline\" location=\"graphic/wrcr27452-math-0001.png\" xmlns=\"http://www.w3.org/1998/Math/MathML\"&gt;&lt;semantics&gt;&lt;mrow&gt;&lt;mrow&gt;&lt;mover accent=\"true\" data-semantic-=\"\" data-semantic-children=\"2,3\" data-semantic-role=\"latinletter\" data-semantic-speech=\"k Subscript upper L Baseline overbar\" data-semantic-type=\"overscore\"&gt;&lt;msub data-semantic-=\"\" data-semantic-children=\"0,1\" data-semantic-parent=\"4\" data-semantic-role=\"latinletter\" data-semantic-type=\"subscript\"&gt;&lt;mi data-semantic-=\"\" data-semantic-annotation=\"clearspeak:simple\" data-semantic-font=\"italic\" data-semantic-parent=\"2\" data-semantic-role=\"latinletter\" data-semantic-type=\"identifier\"&gt;k&lt;/mi&gt;&lt;mi data-semantic-=\"\" data-semantic-annotation=\"clearspeak:simple\" data-semantic-font=\"italic\" data-semantic-parent=\"2\" data-semantic-role=\"latinletter\" data-semantic-type=\"identifier\"&gt;L&lt;/mi&gt;&lt;/msub&gt;&lt;mo data-semantic-=\"\" data-semantic-parent=\"4\" data-semantic-role=\"overaccent\" data-semantic-type=\"punctuation\"&gt;‾&lt;/mo&gt;&lt;/mover&gt;&lt;/mrow&gt;&lt;/mrow&gt;$overline{{k}_{L}}$&lt;/annotation&gt;&lt;/semant","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"45 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142541713","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing Streamflow Prediction in Ungauged Basins Using a Nonlinear Knowledge-Based Framework and Deep Learning","authors":"Parnian Ghaneei, Ehsan Foroumandi, Hamid Moradkhani","doi":"10.1029/2024wr037152","DOIUrl":"https://doi.org/10.1029/2024wr037152","url":null,"abstract":"In hydrology, a fundamental task involves enhancing the predictive power of a model in ungagged basins by transferring information on physical attributes and hydroclimate dynamics from gauged basins. Introducing an integrated nonlinear clustering framework, this study aims to develop a comprehensive framework that augments predictive performance in basins where direct measurements are sparse or absent. In this framework, uniform manifold approximation and projection (UMAP) is used as a nonlinear method to extract the essential features embedded in hydro-climatological attributes and physical properties. Then, the Growing Neural Gas (GNG) clustering model is used to find the basins that potentially share similar hydro-climatological behaviors. Besides UMAP-GNG, the integration of Principal Component Analysis (PCA) as a linear method to reduce dimensionality with common clustering methods are also assessed to serve as benchmarks. The results reveal that the combination of clustering algorithms with the PCA method may lead to loss of information while the nonlinear method (UMAP) can extract more informative features. The efficacy of the proposed framework is assessed across the Contiguous United States (CONUS) by training a single Base Model using long short-term memory (LSTM) for the centroids of all clusters and then, fine-tuning the model on the centroids of each cluster separately to create a regional model. The results indicate that using the information extracted by the UMAP-GNG method to guide a Base Model can significantly improve the accuracy in most of the clusters and enhance the median prediction accuracy within different clusters from 0.04 to 0.37 of KGE in ungauged basins.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"238 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142541715","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessing Nitrate Leaching During Drought and Extreme Precipitation: Exploring Deep Vadose-Zone Monitoring, Groundwater Observations, and Field Mass Balance","authors":"Iael Raij-Hoffman, Ofer Dahan, Helen E. Dahlke, Thomas Harter, Isaya Kisekka","doi":"10.1029/2024wr037973","DOIUrl":"https://doi.org/10.1029/2024wr037973","url":null,"abstract":"The increasing concern over agricultural practices' impact on groundwater quality necessitates comprehensive studies to evaluate and compare monitoring strategies for nitrate leaching. This work addresses this imperative by examining three methodologies: deep vadose-zone monitoring, shallow groundwater intensive monitoring, and field-level mass balance. The primary objective of the study was to assess nitrate leaching from an intensively cropped processing tomato rotation field using three different methods. Additionally, this study focuses on contrasting conditions between the growing season (characterized by drought in some years) and the winter/rainy season (characterized by extreme precipitation in some years). Results indicate varying degrees of nitrate leaching across methods, with all approaches detecting leaching events during the growing season and off-season precipitation. Despite uncertainties inherent in field-level mass balance estimates, they align reasonably with intensive in-situ monitoring results using the deep Vadose Monitoring System (VMS). Throughout two growing seasons and corresponding fall-winter rainy periods, the VMS effectively tracked seasonal nitrogen leaching below the root zone, correlating with observed groundwater nitrate concentrations increases following extreme precipitation events. Nitrate leaching increased during heavy rainfall in the winter following dry summer periods observed across the deep vadose zone using two VMS systems. This underscores the importance of continuous monitoring and assessment in understanding nitrate dynamics and groundwater contamination risks. In conclusion, this study contributes to knowledge and ongoing research by providing insights into effective monitoring strategies for nitrate leaching into groundwater from intensive cropping systems.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"6 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142541762","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Interpretable Transformer Neural Network Prediction of Diverse Environmental Time Series Using Weather Forecasts","authors":"Enrique Orozco López, David Kaplan, Anna Linhoss","doi":"10.1029/2023wr036337","DOIUrl":"https://doi.org/10.1029/2023wr036337","url":null,"abstract":"Transformer neural networks (TNNs) have caused a paradigm shift in deep learning domains like natural language processing, gathering immense interest due to their versatility in other fields such as time series forecasting (TSF). Most current TSF applications of TNNs use only historic observations to predict future events, ignoring information available in weather forecasts to inform better predictions, and with little attention given to the interpretability of the model's use of explanatory inputs. This work explores the potential for TNNs to perform TSF across multiple environmental variables (streamflow, stage, water temperature, and salinity) in two ecologically important regions: the Peace River watershed (Florida) and the northern Gulf of Mexico (Louisiana). The TNN was tested and its prediction uncertainty quantified for each response variable from one-to fourteen-day-ahead forecasts using past observations and spatially distributed weather forecasts. A sensitivity analysis (SA) was performed on the trained TNNs' attention weights to identify the relative influence of each input variable on each response variable across prediction windows. Overall model performance ranged from good to very good (0.78 &lt; NSE &lt; 0.99 for all variables and forecast horizons). Through the SA, we found that the TNN was able to learn the physical patterns behind the data, adapt the use of input variables to each forecast, and increasingly use weather forecast information as prediction windows increased. The TNN's excellent performance and flexibility, along with the intuitive interpretability highlighting the logic behind the models' forecasting decision-making process, provide evidence for the applicability of this architecture to other TSF variables and locations.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":"15 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142536906","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}