Advanced Machine Learning Techniques to Improve Hydrological Prediction: A Comparative Analysis of Streamflow Prediction Models

Water, Journal Year: 2023, Volume and Issue: 15(14), P. 2572 - 2572

Published: July 13, 2023

The management of water resources depends heavily on hydrological prediction, and advances in machine learning (ML) present prospects for improving predictive modelling capabilities. This study investigates the use a variety widely used algorithms, such as CatBoost, ElasticNet, k-Nearest Neighbors (KNN), Lasso, Light Gradient Boosting Machine Regressor (LGBM), Linear Regression (LR), Multilayer Perceptron (MLP), Random Forest (RF), Ridge, Stochastic Descent (SGD), Extreme Model (XGBoost), to predict river inflow Garudeshwar watershed, key element planning flood control supply. substantial engineering feature study, which incorporates temporal lag contextual data based Indian seasons, leads it distinctiveness. concludes that CatBoost method demonstrated remarkable performance across various metrics, including Mean Absolute Error (MAE), Root Square (RMSE), R-squared (R2) values, both training testing datasets. was accomplished by an in-depth investigation model comparison. In contrast XGBoost LGBM higher percentage points with prediction errors exceeding 35% moderate numbers above 10,000. established itself reliable time-series modelling, easily managing categorical continuous variables, thereby greatly enhancing accuracy. results this highlight value promise algorithms hydrology offer valuable insights academics industry professionals.

Language: Английский

---

Daily River flow Simulation Using Ensemble Disjoint Aggregating M5-Prime Model

Heliyon, Journal Year: 2024, Volume and Issue: 10(20), P. e37965 - e37965

Published: Sept. 30, 2024

Accurate prediction of daily river flow (Q t ) remains a challenging yet essential task in hydrological modeling, particularly crucial for flood mitigation and water resource management. This study introduces an advanced M5 Prime (M5P) predictive model designed to estimate Q as well one- two-day-ahead forecasts (i.e. t+1 t+2 ). The performance M5P ensembles incorporating Bootstrap Aggregation (BA), Disjoint Aggregating (DA), Additive Regression (AR), Vote (V), Iterative classifier optimizer (ICO), Random Subspace (RS), Rotation Forest (ROF) were comprehensively evaluated. proposed models applied case data Tuolumne County, US, using dataset comprising measured precipitation (P ), evaporation (E t), . A wide range input scenarios explored predicting , t+1, t+2. Results indicate that P significantly influence accuracy. Notably, relying solely on the most correlated variable (e.g., t-1) does not guarantee robust However, extending forecast horizon mitigates low-correlation variables Performance metrics DA-M5P achieves superior results, with Nash-Sutcliff Efficiency 0.916 root mean square error 23 m3/s, followed by ROF-M5P, BA-M5P, AR-M5P, RS-M5P, V-M5P, ICO-M5P, standalone model. ensemble modeling framework enhanced capability stand-alone algorithm 1.2 %-22.6 %, underscoring its efficacy potential advancing forecasting.

Language: Английский

---

GTWS-MLrec: global terrestrial water storage reconstruction by machine learning from 1940 to present

Earth system science data, Journal Year: 2023, Volume and Issue: 15(12), P. 5597 - 5615

Published: Dec. 8, 2023

Abstract. Terrestrial water storage (TWS) includes all forms of stored on and below the land surface, is a key determinant global energy budgets. However, TWS data from measurements by Gravity Recovery Climate Experiment (GRACE) satellite mission are only available 2002, limiting regional understanding long-term trends variabilities in terrestrial cycle under climate change. This study presents (i.e., 1940–2022) relatively high-resolution 0.25∘) monthly time series anomalies over surface. The reconstruction achieved using set machine learning models with large number predictors, including climatic hydrological variables, use/land cover data, vegetation indicators (e.g., leaf area index). outcome, machine-learning-reconstructed estimates GTWS-MLrec), fits well GRACE/GRACE-FO measurements, showing high correlation coefficients low biases GRACE era. We also evaluate GTWS-MLrec other independent products such as land–ocean mass budget, atmospheric budget 341 river basins, streamflow at 10 168 gauges. results show that our proposed performs overall as, or more reliable than, previous datasets. Moreover, reconstructions successfully reproduce consequences variability strong El Niño events. dataset consists three based (a) mascons Jet Propulsion Laboratory California Institute Technology, Center for Space Research University Texas Austin, Goddard Flight NASA; (b) detrended de-seasonalized reconstructions; (c) six average areas, both without Greenland Antarctica. Along its extensive attributes, GTWS_MLrec can support wide range geoscience applications better constraining evaluating models, climate-carbon coupling, resources management. Zenodo through https://doi.org/10.5281/zenodo.10040927 (Yin, 2023).

Language: Английский

---

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

---

Assessment of the impact of climate change on streamflow of Ganjiang River catchment via LSTM-based models

Journal of Hydrology Regional Studies, Journal Year: 2024, Volume and Issue: 52, P. 101716 - 101716

Published: Feb. 24, 2024

Language: Английский

---

Novel time-lag informed deep learning framework for enhanced streamflow prediction and flood early warning in large-scale catchments

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 631, P. 130841 - 130841

Published: Feb. 5, 2024

Constrained by the sparsity of observational streamflow data, large-scale catchments face pressing challenges in prediction and flood management amid climate change. Deep learning excels simulation performance while flow lag information data-driven approaches is barely highlighted. In this study, we introduce a time-lag informed deep framework for catchments. Central to utilization between upstream downstream subbasins, enabling precise forecasting at outlet driven data. Taking monsoon-influenced Dulong-Irrawaddy River Basin (DIRB) as study area, determined peak (PFL) days relative annual scale (RAFS) defined subbasins. By incorporating with historical data different time intervals, developed optimal model DIRB. This was then applied evaluate processes 2008 2009, using selected indicators. The results indicate that led significant improvements, notably LSTM_PFL_RAFS Hkamti sub-basin which achieved Kling-Gupta Efficiency (KGE) 0.891 (Nash-Sutcliffe efficiency coefficient, NSE, 0.904), surpassing LSTM's 0.683 (NSE, 0.785). Further integration specific interval, model, H(15)_PFL utilizes reached an impressive KGE 0.948 0.940). outperformed standard LSTM accurately simulating key characteristics, including flows, initiation times, durations 2009 events. Notably, provides valuable 15-day lead forecasting, extending window emergency response preparations. Future research incorporates additional essential catchment features into holds great potential unraveling complex mechanisms hydrological responses human activities

Language: Английский

---

Improved prediction of monthly streamflow in a mountainous region by Metaheuristic-Enhanced deep learning and machine learning models using hydroclimatic data

Theoretical and Applied Climatology, Journal Year: 2023, Volume and Issue: 155(1), P. 205 - 228

Published: Sept. 6, 2023

Language: Английский

---

Coupling SWAT and LSTM for Improving Daily Streamflow Simulation in a Humid and Semi-humid River Basin

Water Resources Management, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 12, 2024

Language: Английский

---

Deep learning algorithms and their fuzzy extensions for streamflow prediction in climate change framework

Journal of Water and Climate Change, Journal Year: 2024, Volume and Issue: 15(2), P. 832 - 848

Published: Feb. 1, 2024

Abstract The present study analyzes the capability of convolutional neural network (CNN), long short-term memory (LSTM), CNN-LSTM, fuzzy CNN, LSTM, and CNN-LSTM to mimic streamflow for Lower Godavari Basin, India. Kling–Gupta efficiency (KGE) was used evaluate these algorithms. Fuzzy-based deep learning algorithms have shown significant improvement over classical ones, among which is best. Thus, it further considered projections in a climate change context four-time horizons using four shared socioeconomic pathways (SSPs). Average 2041–2060, 2061–2080, 2081–2090 are compared that 2021–2040 changed by +3.59, +7.90, +12.36% SSP126; +3.62, +8.28, +12.96% SSP245; +0.65, −0.01, −0.02% SSP370; +0.02, +0.71, +0.06% SSP585. In addition, two non-parametric tests, namely, Mann–Kendall Pettitt were conducted ascertain trend point projected streamflow. Results indicate provides more precise prediction than others. identified variations across different SSPs facilitate valuable insights policymakers relevant stakeholders. It also paves way adaptive decision-making.

Language: Английский

---

Futuristic Streamflow Prediction Based on CMIP6 Scenarios Using Machine Learning Models

Water Resources Management, Journal Year: 2023, Volume and Issue: 37(15), P. 6089 - 6106

Published: Oct. 26, 2023

Language: Английский

---