Evaluating the spatiotemporal dynamics of driving factors for multiple drought types in different climate regions of China

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 640, P. 131710 - 131710

Published: July 23, 2024

Language: Английский

---

Interpretable machine learning on large samples for supporting runoff estimation in ungauged basins

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 639, P. 131598 - 131598

Published: July 5, 2024

Language: Английский

---

Enhancing dynamic flood risk assessment and zoning using a coupled hydrological-hydrodynamic model and spatiotemporal information weighting method

Journal of Environmental Management, Journal Year: 2024, Volume and Issue: 366, P. 121831 - 121831

Published: July 16, 2024

Language: Английский

---

Runoff Simulation in Data-Scarce Alpine Regions: Comparative Analysis Based on LSTM and Physically Based Models

Water, Journal Year: 2024, Volume and Issue: 16(15), P. 2161 - 2161

Published: July 31, 2024

Runoff simulation is essential for effective water resource management and plays a pivotal role in hydrological forecasting. Improving the quality of runoff forecasting continues to be highly relevant research area. The complexity terrain scarcity long-term observation data have significantly limited application Physically Based Models (PBMs) Qinghai–Tibet Plateau (QTP). Recently, Long Short-Term Memory (LSTM) network has been found learning dynamic characteristics watersheds outperforming some traditional PBMs simulation. However, extent which LSTM works data-scarce alpine regions remains unclear. This study aims evaluate applicability basins QTP, as well performance transfer-based (T-LSTM) regions. Lhasa River Basin (LRB) Nyang (NRB) were areas, model was compared that by relying solely on meteorological inputs. results show average values Nash–Sutcliffe efficiency (NSE), Kling–Gupta (KGE), Relative Bias (RBias) B-LSTM 0.80, 0.85, 4.21%, respectively, while corresponding G-LSTM 0.81, 0.84, 3.19%. In comparison PBM- Block-Wise use TOPMEDEL (BTOP), an enhancement 0.23, 0.36, −18.36%, respectively. both basins, outperforms BTOP model. Furthermore, transfer learning-based at multi-watershed scale demonstrates that, when input are somewhat representative, even if amount limited, T-LSTM can obtain more accurate than models specifically calibrated individual watersheds. result indicates effectively improve applied

Language: Английский

---

Improving Hydrological Simulation Accuracy through a Three-Step Bias Correction Method for Satellite Precipitation Products with Limited Gauge Data

Water, Journal Year: 2023, Volume and Issue: 15(20), P. 3615 - 3615

Published: Oct. 16, 2023

Satellite precipitation products (SPPs) have advanced remarkably in recent decades. However, the bias correction of SPPs still performs unsatisfactorily case a limited rain-gauge network. This study proposes new real-time approach that includes three steps to improve quality with gauges and facilitate hydrological simulation Min River Basin, China. paper employed 66 as available ground observation precipitation, Asian Precipitation—Highly Resolved Observational Data Integration Towards Evaluation Water Resources (APHRODITE) historical correct Global Mapping Precipitation NOW (GNOW) NRT (GNRT) 2020. A total 1020 auto-rainfall stations were used benchmark evaluate original corrected six criteria. The results show statistic dynamic method (SDBC) improved significantly cumulative distribution function matching (CDF) could reduce overcorrection error from SDBC. inverse variance weighting (IEVW) integrations GNOW GNRT did not noticeable improvement they use similar hardware software processes. better performance simulations. It is recommended employ different for integration. proposed significant estimation flood prediction data-sparse basins worldwide.

Language: Английский

---

Enhancing Runoff Simulation Using BTOP-LSTM Hybrid Model in the Shinano River Basin

Water, Journal Year: 2023, Volume and Issue: 15(21), P. 3758 - 3758

Published: Oct. 27, 2023

Runoff simulation is an ongoing challenge in the field of hydrology. Process-based (PB) hydrological models often gain unsatisfactory accuracy due to incomplete physical process representations. While deep learning (DL) demonstrate their capacity grasp intricate response processes, they still face constraints pertaining representative training data and comprehensive observations. In order provide unobservable variables from PB model DL model, this study constructed hybrid by feeding output (BTOP) into (LSTM) as additional input features. These underwent feature dimensionality reduction using selection method (Pearson Correlation Coefficient, PCC) extraction (Principal Component Analysis, PCA) before LSTM. The results showed that standalone LSTM performed well across basin, with NSE values all exceeding 0.70. enhanced performance increased 0.75 nearly 0.80 a sub-basin. Lastly, if BTOP directly fed without reduction, model’s significantly decreases noise interference. value decreased 0.09 compared demonstrated effectiveness PCC PCA removing redundant information within variables. findings new insights incorporating constructing models.

Language: Английский

---

Ungauged Basin Flood Prediction Using Long Short-Term Memory and Unstructured Social Media Data

Water, Journal Year: 2023, Volume and Issue: 15(21), P. 3818 - 3818

Published: Nov. 1, 2023

Floods are highly perilous and recurring natural disasters that cause extensive property damage threaten human life. However, the paucity of hydrological observational data hampers precision physical flood models, particularly in ungauged basins. Recent advances disaster monitoring have explored potential social media as a valuable source information. This study investigates spatiotemporal consistency during flooding events evaluates its viability substitute for catchments. To assess utility an input factor prediction conducted time-series spatial correlation analyses by employing scan statistics confusion matrices. Subsequently, long short-term memory model was used to forecast outflow volume Ui Stream basin South Korea. A comparative analysis various combinations revealed datasets incorporating rainfall, exhibited highest accuracy, with Nash–Sutcliffe efficiency 94%, coefficient 97%, minimal normalized root mean square error 0.92%. demonstrated viable alternative data-scarce basins, highlighting effectiveness enhancing accuracy.

Language: Английский

---

A decomposition approach to evaluating the local performance of global streamflow reanalysis

Hydrology and earth system sciences, Journal Year: 2024, Volume and Issue: 28(15), P. 3597 - 3611

Published: Aug. 8, 2024

Abstract. While global streamflow reanalysis has been evaluated at different spatial scales to facilitate practical applications, its local performance in the time–frequency domain is yet be investigated. This paper presents a novel decomposition approach evaluating by combining wavelet transform with machine learning. Specifically, time series of and observation are respectively decomposed then approximation components against those observed streamflow. Furthermore, accumulated effects derived showcase influences catchment attributes on scales. For generated Global Flood Awareness System, case study devised based observations from Catchment Attributes Meteorology for Large-sample Studies. The results highlight that tends more effective characterizing seasonal, annual multi-annual features than daily, weekly monthly features. Kling–Gupta efficiency (KGE) values original primarily influenced precipitation seasonality. High KGE tend catchments where there winter, which can due low evaporation reasonable simulations soil moisture baseflow processes. longitude, mean slope also influence components. On other hand, geology, soils vegetation appear play relatively small part Overall, this provides useful information applications reanalysis.

Language: Английский

---

Accuracy evaluation and comparison of GSMaP series for retrieving precipitation on the eastern edge of the Qinghai-Tibet Plateau

Journal of Hydrology Regional Studies, Journal Year: 2024, Volume and Issue: 56, P. 102017 - 102017

Published: Oct. 19, 2024

Language: Английский

---

Perspective Chapter: Big Data and Deep Learning in Hydrological Modeling

IntechOpen eBooks, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 4, 2024

This chapter delves into the integration of physical mechanisms with deep learning models to enhance interpretability and accuracy hydrological process modeling. In era big data rapid advancements in AI, synergy between traditional principles machine opens new opportunities for improved water resource management, flood prediction, drought monitoring. The presents a comprehensive framework that leverages vast datasets from sources such as remote sensing, reanalysis data, situ It explores potential models, particularly when combined insights, address challenges data-scarce regions, improving transparency predictions. By analyzing strengths limitations current approaches, study highlights value hybrid balancing interpretability. These not only predictive performance but also provide more transparent insights underlying processes. contributes sustainable disaster resilience, climate adaptation, pushing forward both scientific progress practical applications. offers valuable methodologies case studies underscore importance domain knowledge development explainable reliable reshaping future forecasting.

Language: Английский

---