Enhancing Soil Moisture Forecasting Accuracy with REDF-LSTM: Integrating Residual En-Decoding and Feature Attention Mechanisms

Water, Journal Year: 2024, Volume and Issue: 16(10), P. 1376 - 1376

Published: May 11, 2024

This study introduces an innovative deep learning model, Residual-EnDecode-Feedforward Attention Mechanism-Long Short-Term Memory (REDF-LSTM), designed to overcome the high uncertainty challenges faced by traditional soil moisture prediction methods. The REDF-LSTM integrating a residual encoder–decoder LSTM layer, enhanced layers, and feedforward attention, not only captures features of time series data but also optimizes model’s ability identify key influencing factors, including land surface features, atmospheric conditions, other static environmental variables. Unlike existing methods, innovation this model lies in its first-time combination attention mechanisms field. It delves into complex patterns through structure accurately locates factors mechanism, significantly improving predictive performance. choice combine mechanism with is fully leverage their advantages processing sequences enhancing focus on important aiming for more accurate prediction. After comparison current advanced models such as EDLSTM, FAMLSTM, GANBiLSTM, our demonstrated best Compared models, it achieved average improvement 13.07% R2, 20.98% RMSE, 24.86% BIAS, 11.1% KGE performance indicators, proving superior capability potential application value precision agriculture ecosystem management.

Language: Английский

---

A state-of-the-art review of long short-term memory models with applications in hydrology and water resources

Applied Soft Computing, Journal Year: 2024, Volume and Issue: unknown, P. 112352 - 112352

Published: Oct. 1, 2024

Language: Английский

---

A deep adaptive bidirectional generative adversarial neural network (Bi-GAN) for groundwater contamination source estimation

Journal of Hydrology, Journal Year: 2025, Volume and Issue: unknown, P. 132753 - 132753

Published: Feb. 1, 2025

Language: Английский

---

Multimodal deep learning water level forecasting model for multiscale drought alert in Feiyun River basin

Expert Systems with Applications, Journal Year: 2023, Volume and Issue: 244, P. 122951 - 122951

Published: Dec. 15, 2023

Language: Английский

---

VMDI-LSTM-ED: A novel enhanced decomposition ensemble model incorporating data integration for accurate non-stationary daily streamflow forecasting

Journal of Hydrology, Journal Year: 2025, Volume and Issue: unknown, P. 132769 - 132769

Published: Jan. 1, 2025

Language: Английский

---

Multi-machine learning methods for rapid and synergistic inversion of groundwater contamination source, hydrogeologic parameter and boundary condition

Journal of Contaminant Hydrology, Journal Year: 2025, Volume and Issue: 273, P. 104599 - 104599

Published: May 6, 2025

Language: Английский

---

Aquifer flow parameter estimation using coupled meshless methods and metaheuristic algorithms

Environmental Modelling & Software, Journal Year: 2024, Volume and Issue: 177, P. 106050 - 106050

Published: April 18, 2024

Language: Английский

---

Groundwater Contamination Source Recognition Based on a Two-Stage Inversion Framework with a Deep Learning Surrogate

Water, Journal Year: 2024, Volume and Issue: 16(13), P. 1907 - 1907

Published: July 3, 2024

Groundwater contamination source recognition is an important prerequisite for subsequent remediation efforts. To overcome the limitations of single inversion methods, this study proposed a two-stage framework by integrating two primary approaches—simulation-optimization and simulation-data assimilation—thereby enhancing accuracy. In first stage, ensemble smoother with multiple data assimilation method (a type assimilation) conducted global broad search to provide better initial values ranges second stage. collective decision optimization algorithm simulation-optimization) was used refined deep search, further final Additionally, learning method, multilayer perceptron, utilized establish surrogate simulation model, reducing computational costs. These theories methods were applied validated in hypothetical scenario synchronous identification boundary conditions. The results demonstrated that significantly improved accuracy compared mean relative error absolute just 4.95% 0.1756, respectively. Moreover, perceptron model offered greater approximation than traditional shallow model. Specifically, coefficient determination, error, root square 0.9860, 9.72%, 0.1727, 0.47, respectively, highlighting its significant advantages. findings can more reliable technical support practical case applications improve efficiency.

Language: Английский

---

Small Data Insights for Groundwater Management

Environmental Science & Technology, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 12, 2025

InfoMetricsFiguresRef. Environmental Science & TechnologyASAPArticle This publication is free to access through this site. Learn More CiteCitationCitation and abstractCitation referencesMore citation options ShareShare onFacebookX (Twitter)WeChatLinkedInRedditEmailJump toExpandCollapse ViewpointFebruary 12, 2025Small Data Insights for Groundwater ManagementClick copy article linkArticle link copied!Zi ZhanZi ZhanSchool of Chemical Engineering, Shanghai University, 200444, ChinaMore by Zi ZhanView BiographyYaqiang Wei*Yaqiang WeiSchool ChinaState Protection Key Laboratory Source Apportionment Control Aquatic Pollution, China University Geosciences, Wuhan 430078, China*Email: [email protected]More Yaqiang Weihttps://orcid.org/0000-0001-6317-4735Tian-Chyi Jim YehTian-Chyi YehDepartment Hydrology Atmospheric Science, Arizona, Tucson, Arizona 85721, United StatesMore Tian-Chyi YehYiran ChenYiran ChenSchool Yiran ChenYuling Yuling ChenYu LiYu LiSchool Yu LiJiao ZhangJiao ZhangSchool Jiao ZhangYi Wen*Yi WenTechnical Centre Soil, Agriculture Rural Ecology Environment, Ministry Beijing 100012, Yi WenHui Li*Hui Hui LiOpen PDFEnvironmental TechnologyCite this: Environ. Sci. Technol. 2025, XXXX, XXX, XXX-XXXClick citationCitation copied!https://pubs.acs.org/doi/10.1021/acs.est.5c01025https://doi.org/10.1021/acs.est.5c01025Published February 2025 Publication History Received 21 January 2025Published online 12 2025article-commentary© American Society. available under these Terms Use. Request reuse permissionsThis licensed personal use The ACS Publications© SocietySubjectswhat are subjectsArticle subjects automatically applied from the Subject Taxonomy describe scientific concepts themes article.GroundwatersImpuritiesOptimizationPhysical chemical propertiesRemediationData scarcity poses significant challenges in environmental fields, including agricultural systems, (1) ecosystem management, (2) water resource assessment, (3) where incomplete or fragmented data sets often hinder accurate analysis decision-making. These particularly pronounced groundwater concealed nature subsurface environments, logistical difficulties, high monitoring costs severely limit collection. Furthermore, spatial heterogeneity geological formations amplify uncertainties modeling flow contaminant transport. Such complexities necessitate development a small approach that can extract meaningful insights limited sets, enabling timely, cost-effective decision-making management enhancing efficiency remediation efforts data-constrained conditions.Origins Issues Small DataClick section linkSection copied!The hidden characteristics systems create directly observing soil contamination, complex underground media. (4) Monitoring techniques, such as borehole sampling intermittent quality testing, provide localized but spatiotemporal discontinuities (5) (Figure 1). results substantial gaps increase migration studies heterogeneous environments with properties. (6) Additionally, confined delays detection, complicating accurately trace source information. (7) along difficulties continuous frequently result scenarios. gathering large volumes quickly rarely feasible practical applications, especially time constraints. (8) Despite their limitations, serve only information during critical moments. (9) Consequently, importance increases they offer timely actionable align immediate needs real-world engineering management. Therefore, effectively leveraging essential decision-makers respond efficiently, optimizing allocation impact.Figure 1Figure 1. generation, enhancement methods, within improving outcomes at contaminated sites. Constrained challenges, collected boreholes be leveraged in-depth site groundwater. Optimized methods maximizing utility not enhance evaluation subsequent also support efficient, high-precision, strategies.High Resolution ImageDownload MS PowerPoint SlideNecessity Optimizing UtilizationClick copied!Researchers have explored impact varying number pumping tests wells understand how additional points influence model accuracy field system characterization. For instance, increasing four nine 49 158 yields <10% improvement predictive accuracy, (10−12) highlighting diminishing returns volume. phenomenon suggests after balance point, accumulating more adds value, when weighed against associated time, equipment, computational complexity. expanding volume inadvertently introduce redundancy, saturating point marginal each new decreases. (13) redundancy resources required training escalates risk overfitting, ultimately reducing generalizability prediction reliability applications.Researchers Bayesian–MCMC (Markov chain Monte Carlo) pinpoint accumulation, ensuring maximized without incurring disproportionate demands. Identifying optimal threshold becomes guiding decisions, resource-constrained scenarios which paramount. approaches decision significantly reduce improve efficiency. While rely on large-scale collection require decision-making, valuable data, quicker responses efficient decisions. important help strike better between acquisition cost investment. By allocation, ensure forward-thinking reliable decisions Ultimately, application improves while maintaining providing robust sustainable development.Lessons Applications across FieldsClick copied!Recognizing research, existing show data-based techniques deliver models even extensive offering guidance strategies. (14) molecular science, highly constrained, machine learning like random forests vector machines been predict drug–target interactions drug toxicity promising results. (15−17) Similarly, researchers addressed utilizing augmentation transfer learning, combining convolutional neural networks image quantitative structure–activity relationship modeling, predictions quantum structures activity relationships chemistry. (18,19) has similarly benefited sample it played crucial role understanding habitat requirements population dynamics rare species, supporting biodiversity conservation efforts. (20,21) Meanwhile, shift toward personalized production batch manufacturing highlighted limitations traditional big models, prone overfitting To counter this, incorporated recurrent variational autoencoders conditional generative adversarial networks, alongside optimization algorithms proximal policy regression. improved product environments. (22,23) science hydrological innovative long short-term memory prototypical successfully employed address runoff data-scarce river basins, overcoming posed sparse data. (24,25) With respect hydrogeology, utilization minimization response times systems. Approaches focus extracting potential, addressing medicine, ecology, advancing data-limited fields climate urban planning, monitoring.Small Opportunities GroundwaterClick copied!Building achievements applications disciplines, underutilized opportunities exist fall short. demonstrated its value managing achieved sets. However, similar yet widely adopted hydrogeological studies, depends well-established, data-intensive MODFLOW MT3DMS surrogates trained (26) become unreliable situations, gap address.Due concealment aquifers, well constraints collection, present an effective test case methods. (27,28) factors set acquisition, thereby uncertainty transport modeling. surveys face (29,30) Common interpolation kriging inverse distance weighting, assume stationarity, limits ability complexity nonstationarity. (27) network availability, restrict potential improvements situations. (31)Given hydraulic tomography (HT) differs capturing connectivity permeable zones, parameter estimation conditions. (32−35) advantages, residual remain major challenge uncertainties, advanced genetic algorithms, particle swarm optimization, Bayesian increasingly used distribution Among these, inversion excels refining posterior distributions mitigating (30,31) exploring combination technologies characterize both aquifer contaminants, quantifying uncertainties. could establish framework plays phenomena common. Building advancements, proven other research. arises due points, low frequency, infrequency contamination events, impacts identification sources predictions. (36) issue, (GANs), knowledge generate (37,38) filling generalization 1).Challenges ProspectsClick copied!Achieving fraught several distinct First, parameters variability properties generalization, causing reduced broader applications. (39) Second, increases, benefit diminishes, making difficult identify further performance. Lastly, determining efficiency, practices compromising capabilities.Addressing requires strategies robustness, integrating physical regularization generalizability. (14,40) allow adapt diverse conditions mitigate risks. Advanced filtering optimize dynamically estimating uncertainty, (9,41) allowing identifying redundant. In combination, imputation, (15) ensemble (42) enhances stability. (43) addition, semisupervised unlabeled (44)While advantages still inherent limitations. Regularization designed may regions heterogeneity. extreme areas means cannot fully capture local differences, replicate complexity, potentially introducing artificial patterns fail reflect true aquifer. quantification dependent prior knowledge. (45) imprecise information, dependency bias distributions, estimates conductivity. When unreliable, yield misleading results, (46) Advances methodologies transforming precise control, fostering practices. innovations promote resilient ecosystems, informed, long-term policies.Author InformationClick copied!Corresponding AuthorsYaqiang Wei - School China; State https://orcid.org/0000-0001-6317-4735; Email: protected]Yi Wen Technical protected]Hui Li protected]AuthorsZi Zhan ChinaTian-Chyi Yeh Department StatesYiran Chen ChinaYuling ChinaYu ChinaJiao Zhang ChinaAuthor ContributionsZ.Z. Y.W. led conceptualization, writing, figure drafting. Y.C., J.Z. assisted writing figures. H.L. conceived ideas framework. T.-C.J.Y. Y.L. helped paper. All authors approved final form publication.NotesThe declare no competing financial interest.BiographyClick ZhanHigh SlideDr. currently Associate Professor University. He completed his postdoctoral research Tong obtained Ph.D. Chinese Academy Sciences 2017, joint program States. His primary migration, transformation, fate contaminants projects, project key fund Youth Fund National Natural Foundation, two subprojects Research Development Program Soil Pollution Causes Technologies. serves young editorial board member journals Eco-Environment Health Communications.AcknowledgmentsClick copied!This work was supported Foundation (42477004, 42330706, 42125706).ReferencesClick copied! references 46 publications. 1Pradeleix, L.; Roux, P.; Bouarfa, S.; Bellon-Maurel, V. Multilevel Assessment Regional Farming Activities Life Cycle Assessment: Tackling Scarcity Farm Diversity Inventories Based Agrarian System Diagnosis. Agricultural Systems 2022, 196, 103328, DOI: 10.1016/j.agsy.2021.103328 Google ScholarThere corresponding record reference.2Zuquim, G.; Stropp, J.; Moulatlet, G. M.; Van doninck, Quesada, C. A.; Figueiredo, F. O. Costa, R. C.; Ruokolainen, K.; Tuomisto, H. Making Most Scarce Data: Mapping Gradients Data-Poor Areas Using Species Occurrence Records. Methods Ecol. Evol. 2019, 10 (6), 788– 801, 10.1111/2041-210X.13178 reference.3Dutta, Das, M. Remote Sensing Scene Classification Labelled Samples─A Survey State-of-the-Arts. Comput. Geosci. 2023, 171, 105295, 10.1016/j.cageo.2022.105295 reference.4Wu, Y.; Xu, Liu, S. Generative Artificial Intelligence: A New Engine Advancing Engineering. 2024, 58, 17524, 10.1021/acs.est.4c07216 reference.5Berg, Illman, W. A. Capturing Aquifer Heterogeneity: Comparison Controlled Sandbox Experiments. Water Resour. Res. 2011, 47 (9), 1– 17, 10.1029/2011WR010429 reference.6Liu, X.; Craig, Zhu, Yeh, T. J. Validation Transient Hydraulic Tomography. 2007, 43 (5), 13, 10.1029/2006WR005144 reference.7Anshuman, Eldho, I. Parallel Workflow Framework Encoder-Decoder LSTMs Uncertainty Quantification Contaminant Identification Groundwater. Hydrol. 619, 129296, 10.1016/j.jhydrol.2023.129296 reference.8Liu, F.; Wang, Y. Hao, Wen, Characterization Basin-Scale Heterogeneity Tomography Responses Induced Exploitation Reduction. 2020, 588, 125137, 10.1016/j.jhydrol.2020.125137 reference.9Yang, R.; Jiang, Pang, T.; Yang, Z.; Han, Li, H.; Zheng, Crucial Time Emergency Reliable Numerical Identification. 265 (April), 122303, 10.1016/j.watres.2024.122303 reference.10Hao, Xiang, Ando, Hsu, K. Lee, Detecting Fracture Zone Connectivity. Ground 2008, (2), 183– 192, 10.1111/j.1745-6584.2007.00388.x Scholar10Hydraulic detecting fracture zone connectivityHao, Yonghong; Jianwei; Walter Kenichi; Kuo-Chin; Cheng-HawGround (2008), 183-192CODEN: GRWAAP; ISSN:0017-467X. (Blackwell Publishing, Inc.) zones geol. media great ground prevention remediation. paper, we recently developed tomog. technique anal. algorithm (sequential successive linear estimator) synthetic fractured aims explore characterizing connectivity. Results investigation showed using HT no. wells, (general pattern) mapped satisfactorily although estd. property smooth. As ports vivid values. We hope success generations technol. (i.e., hydraulic, tracer, pneumatic surveys) mapping fractures features >> SciFinder ®https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXjsVKms7k%253D&md5=d572f0c652ec3f3c86a6043313c0f06a11Wei, Chen, Estimation Properties Fractured Experiments Discrete Network Model. 2021, 66 (11), 1685– 1694, 10.1080/02626667.2021.1962887 reference.12Zha, Mao, D.; Lu, Usefulness Flux Measurements Tomographic Conductivity Distribution Medium. Adv. 2014, 71, 162– 176, 10.1016/j.advwatres.2014.06.008 reference.13Aliouache, Fischer, Massonnat, Jourde, An Inverse Approach Integrating Flowmeter Pumping Test Three-Dimensional Characterization. 603 (PB), 126939, 10.1016/j.jhydrol.2021.126939 reference.14Xu, Ji, Machine Learning Materials Science. npj Mater. 9 (1), 15, 10.1038/s41524-023-01000-z reference.15Dou, B.; Merkurjev, E.; Ke, Zhang, Wei, Challenges Molecular Chem. Rev. 123 (13), 8736– 8780, 10.1021/acs.chemrev.3c00189 Scholar15Machine ScienceDou, Bozheng; Zailiang; Ekaterina; Lu; Long; Jian; Yueying; Jie; Bengong; Guo-WeiChemical Reviews (Washington, DC, States) (2023), 8736-8780CODEN: CHREAY; ISSN:0009-2665. (American Society) review. presence various constraints, cost, ethics, privacy, security, tech. acquisition. past decade; received little attention, though severe (ML) deep (DL) studies. Overall, compounded issues, diversity, noise, imbalance, high-dimensionality. Fortunately, current era characterized breakthroughs ML, DL, intelligence (AI), enable data-driven discovery, many ML DL provided solns. problems. result, progress made decade. review, summarize analyze emerging mol. chem. biol. sciences. review basic regression, logistic regression (LR), k-nearest neighbors (KNN), (SVM), kernel (KL), forest (RF), gradient boosting trees (GBT), (ANN), (CNN), U-Net, graph (GNN), (GAN), (LSTM), autoencoder, transformer, active graph-based phys. model-based augmentation. briefly discuss latest advances Finally, conclude survey discussion trends science. ®https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=

Language: Английский

---

Breaking the mold of simulation-optimization: Direct forward machine learning methods for groundwater contaminant source identification

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 642, P. 131759 - 131759

Published: Aug. 10, 2024

Language: Английский

---