Long-term streamflow forecasting in data-scarce regions: Insightful investigation for leveraging satellite-derived data, Informer architecture, and concurrent fine-tuning transfer learning

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 631, P. 130772 - 130772

Published: Feb. 2, 2024

Language: Английский

---

Estimating reference crop evapotranspiration using improved convolutional bidirectional long short-term memory network by multi-head attention mechanism in the four climatic zones of China

Agricultural Water Management, Journal Year: 2024, Volume and Issue: 292, P. 108665 - 108665

Published: Jan. 9, 2024

Accurate reference crop evapotranspiration (ET0) estimation is essential for agricultural water management, productivity, and irrigation systems. As the standard ET0 method, Penman-Monteith equation has been widely recommended worldwide. However, its application still restricted to comprehensive meteorological data deficiency, making exploration of alternative simpler models acceptable highly meaningful. Concerning aforementioned requirement, this study developed novel deep learning model (MA-CNN-BiLSTM), which incorporates Multi-Head Attention mechanism (MA), Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory network (BiLSTM) as intricate relationship processor, feature extractor, regression component, estimate based on radiation-based (Rn-based), humidity-based (RH-based), temperature-based (T-based) input combinations at 600 stations during 1961–2020 throughout China under internal external cross-validation strategies. Besides, through a comparative evaluation among MA-CNN-BiLSTM, CNN-BiLSTM, BiLSTM, LSTM, Multivariate Adaptive Regression Splines (MARS), empirical models, result indicated that MA-CNN-BiLSTM achieved superior precision, with values Determination Coefficient (R2), Nash–Sutcliffe efficiency coefficient (NSE), Relative Root Mean Square Error (RRMSE), (RMSE), Absolute (MAE) ranging 0.877–0.972, 0.844–0.962, 0.129–0.292, 0.294–0.644 mm d−1, 0.244–0.566 d−1 strategy 0.797–0.927, 0.786–0.920, 0.162–0.335, 0.409–0.969 0.294–0.699 strategy. Specifically, Rn-based excelled in temperate continental zone (TCZ) mountain plateau (MPZ), while RH-based yielded best precision others. Furthermore, was by 2.74–106.04% R2, 1.11–120.49% NSE, 1.41–40.27% RRMSE, 1.68–45.53% RMSE, 1.21–38.87% MAE, respectively. In summary, main contribution present proposal LSTM-type (MA-CNN-BiLSTM) cope various data-missing scenarios China, can provide effective support decision-making regional agriculture management.

Language: Английский

---

Convergent and Transdisciplinary Integration: On the Future of Integrated Modeling of Human‐Water Systems

Water Resources Research, Journal Year: 2025, Volume and Issue: 61(2)

Published: Feb. 1, 2025

Abstract The notion of convergent and transdisciplinary integration, which is about braiding together different knowledge systems, becoming the mantra numerous initiatives aimed at tackling pressing water challenges. Yet, transition from rhetoric to actual implementation impeded by incongruence in semantics, methodologies, discourse among disciplinary scientists societal actors. Here, we embrace “integrated modeling”—both quantitatively qualitatively—as a vital exploratory instrument advance such providing means navigate complexity manage uncertainty associated with understanding, diagnosing, predicting, governing human‐water systems. From this standpoint, confront barriers offering seven focused reviews syntheses existing missing links across frontiers distinguishing surface groundwater hydrology, engineering, social sciences, economics, Indigenous place‐based knowledge, studies other interconnected natural systems as atmosphere, cryosphere, ecosphere. While there are, arguably, no bounds pursuit inclusivity representing spectrum human processes around resources, advocate that integrated modeling can provide approach delineating scope through lens three fundamental questions: (a) What “purpose”? (b) constitutes sound “boundary judgment”? (c) are “critical uncertainties” their compounding effects? More broadly, call for investigating what warranted “systems complexity,” opposed unjustified “computational complexity” when complex human‐natural careful attention interdependencies feedbacks, scaling issues, nonlinear dynamics thresholds, hysteresis, time lags, legacy effects.

Language: Английский

---

Virtual Hydrological Laboratories: Developing the Next Generation of Conceptual Models to Support Decision Making Under Change

Water Resources Research, Journal Year: 2024, Volume and Issue: 60(4)

Published: April 1, 2024

Abstract As hydrological systems are pushed outside the envelope of historical experience, ability current models to serve as a basis for credible prediction and decision making is increasingly challenged. Conceptual most common type surface water model used support due reasonable performance in absence change, ease use computational speed that facilitate scenario, sensitivity uncertainty analysis. Hence, conceptual effect represent “shopfront” science seen by practitioners. However, these have notable limitations their resolve internal catchment processes subsequently capture change. New thinking needed confront challenges faced generation dealing with changing environment. We argue next should combine parsimony our best available scientific understanding. propose strategy develop such using multiple lines evidence. This includes appropriately selected physically resolved “Virtual Hydrological Laboratories” test refine simpler models' predict future changes. approach moves beyond sole focus on “predictive skill” measured metrics performance, facilitating development fidelity (i.e., “get right answers reasons”). quest more than curiosity; it expected policy makers who need know what plan for.

Language: Английский

---

A spatio-temporal analysis of environmental fate and transport processes of pesticides and their transformation products in agricultural landscapes dominated by subsurface drainage with SWAT+

The Science of The Total Environment, Journal Year: 2024, Volume and Issue: 945, P. 173629 - 173629

Published: May 29, 2024

Pesticides are detected in surface water and groundwater, endangering the environment. In lowland regions with subsurface drainage systems, drained depressions become hotspots for transport of pesticides their transformation products (TPs). This study focuses on detailed modelling degradation different physico-chemical properties. The objective is to analyse complex hydrological processes, understand temporal spatial dynamics pesticides. ecohydrological model SWAT+ simulates processes as well agricultural management pesticide can therefore be used develop loss reduction strategies. three (pendimethalin, diflufenican, flufenacet), two TPs, flufenacet-oxalic acid (FOA) flufenacet sulfonic (FESA). area a 100-hectare farmland northern German lowlands Schleswig-Holstein that characterised by an extensive network 6.3 km managed according common conventional practice. modelled streamflow very good agreement between observed simulated data during calibration validation. Regarding pesticides, performance highly mobile substances better than non-mobile While moderately via tile drains played important role both wet dry conditions, no was sorptive pendimethalin. conclusion, reliably represent small-scale drainage-dominated catchments, runoff-induced peak loads. However, it has weaknesses accounting substances, which lead underestimation subsequent delivery after precipitation events thus underestimates total load.

Language: Английский

---

A Comprehensive Framework for Evaluation of Skeletonization Impacts on Urban Drainage Network Model Simulations

Water Resources Research, Journal Year: 2025, Volume and Issue: 61(2)

Published: Jan. 30, 2025

Abstract Urban drainage network models (UDNMs) have been widely used to facilitate flood management. Typically, a UDNM is developed using data from Geographic Information Systems (GIS), and hence it consists of many short pipes connection nodes or manholes. To improve modeling efficiency, GIS‐based model generally skeletonized by removing elements. However, there has surprisingly lack knowledge on what extent such skeletonization can affect the model's simulation accuracy, resulting in uncertainty risk estimation. This paper makes first attempt quantitatively evaluate multidimensional impacts different levels hydraulic properties UDNMs. goal achieved new evaluation framework comprising eight existing metrics that make use hydrographs, main pipe hydraulics distribution properties. A real‐life illustrate under various rainfall conditions levels. The also compare performance two compensation methods mitigating caused skeletonization. Results obtained show that: (a) significantly magnitude peak flow at outfall, with maximum overestimation up 20%, (b) be affected increasing 35%, (c) may alter which largely ignored past studies. These findings provide guidance for skeletonization, where their associated should aware engineering practice.

Language: Английский

---

Innovative use of corncob ash in concrete: a machine learning perspective on compressive strength prediction

Innovative Infrastructure Solutions, Journal Year: 2025, Volume and Issue: 10(3)

Published: Feb. 4, 2025

Language: Английский

---

Are Deep Learning Models in Hydrology Entity Aware?

Geophysical Research Letters, Journal Year: 2025, Volume and Issue: 52(6)

Published: March 22, 2025

Abstract Hydrology is experiencing a shift from process‐based toward deep learning (DL) models. Entity‐aware (EA) DL models with static features (predominantly physiographic proxies) merged to dynamic forcing show significant performance improvements. However, recent studies challenge the notion that combining forcings attributes make such entity aware, suggesting are not effectively leveraged for generalization. We examine awareness using state‐of‐the‐art Long‐Short Term Memory (LSTM) networks and CAMELS‐US data set. compare EA provided ablated variants inputs. Findings suggest superior of primarily driven by information meteorological data, limited contributions features, particularly when tested out‐of‐sample. These results previously held assumptions regarding how proxies contribute generalization ability in Models, highlighting need new approaches robust

Language: Английский

---

Exploring Soil Moisture Dynamics and Variability Across Scales and Geological Settings Using Gaussian Mixture-Long Short-Term Memory Networks

Published: Jan. 1, 2025

Language: Английский

---

Transdisciplinary coordination is essential for advancing agricultural modeling with machine learning

One Earth, Journal Year: 2025, Volume and Issue: 8(4), P. 101233 - 101233

Published: April 1, 2025

Language: Английский

---