Evaluation of CatBoost Method for Predicting Weekly Pan Evaporation in Subtropical and Sub-Humid Regions

Pure and Applied Geophysics, Год журнала: 2024, Номер 181(2), С. 719 - 747

Опубликована: Фев. 1, 2024

Язык: Английский

---

Hybrid river stage forecasting based on machine learning with empirical mode decomposition

Applied Water Science, Год журнала: 2024, Номер 14(3)

Опубликована: Фев. 15, 2024

Abstract The river stage is certainly an important indicator of how the water level fluctuates overtime. Continuous control can help build early warning floods along rivers and streams. Hence, forecasting stages up to several days in advance very constitutes a challenging task. Over past few decades, use machine learning paradigm investigate complex hydrological systems has gained significant importance, one promising areas investigations. Traditional situ measurements, which are sometime restricted by existing handicaps especially terms regular access any points alongside streams rivers, be overpassed modeling approaches. For more accurate stages, we suggest new framework based on learning. A hybrid approach was developed combining techniques, namely random forest regression (RFR), bootstrap aggregating (Bagging), adaptive boosting (AdaBoost), artificial neural network (ANN), with empirical mode decomposition (EMD) provide robust model. singles models were first applied using only data without preprocessing, following step, decomposed into intrinsic functions (IMF), then used as input variables. According obtained results, proposed showed improved results compared standard RFR EMD for which, error performances metrics drastically reduced, correlation index increased remarkably great changes models’ have taken place. RFR_EMD, Bagging_EMD, AdaBoost_EMD less than ANN_EMD model, had higher R≈0.974, NSE≈0.949, RMSE≈0.330 MAE≈0.175 values. While RFR_EMD Bagging_EMD relatively equal exhibited same accuracies AdaBoost_EMD, superiority obvious. model shows potential signal learning, serve basis insights forecasting.

Язык: Английский

---

Daily suspended sediment yield estimation using soft-computing algorithms for hilly watersheds in a data-scarce situation: a case study of Bino watershed, Uttarakhand

Theoretical and Applied Climatology, Год журнала: 2024, Номер 155(5), С. 4023 - 4047

Опубликована: Фев. 10, 2024

Язык: Английский

---

Stacked hybridization to enhance the performance of artificial neural networks (ANN) for prediction of water quality index in the Bagh river basin, India

Heliyon, Год журнала: 2024, Номер 10(10), С. e31085 - e31085

Опубликована: Май 1, 2024

Water quality assessment is paramount for environmental monitoring and resource management, particularly in regions experiencing rapid urbanization industrialization. This study introduces Artificial Neural Networks (ANN) its hybrid machine learning models, namely ANN-RF (Random Forest), ANN-SVM (Support Vector Machine), ANN-RSS Subspace), ANN-M5P (M5 Pruned), ANN-AR (Additive Regression) water the rapidly urbanizing industrializing Bagh River Basin, India. The Relief algorithm was employed to select most influential input parameters, including Nitrate (NO

Язык: Английский

---

Estimation of crop evapotranspiration using statistical and machine learning techniques with limited meteorological data: a case study in Udham Singh Nagar, India

Theoretical and Applied Climatology, Год журнала: 2024, Номер 155(6), С. 5279 - 5296

Опубликована: Апрель 3, 2024

Язык: Английский

---

A comparative survey between cascade correlation neural network (CCNN) and feedforward neural network (FFNN) machine learning models for forecasting suspended sediment concentration

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Май 9, 2024

Abstract Suspended sediment concentration prediction is critical for the design of reservoirs, dams, rivers ecosystems, various operations aquatic resource structure, environmental safety, and water management. In this study, two different machine models, namely cascade correlation neural network (CCNN) feedforward (FFNN) were applied to predict daily-suspended (SSC) at Simga Jondhara stations in Sheonath basin, India. Daily-suspended discharge data from 2010 2015 collected used develop model suspended concentration. The developed models evaluated using statistical indices like Nash Sutcliffe efficiency coefficient (N ES ), root mean square error (RMSE), Willmott’s index agreement (WI), Legates–McCabe’s (LM), supplemented by a scatter plot, density plots, histograms Taylor diagram graphical representation. was compared with CCNN FFNN. Nine input combinations explored lag-times (Q t-n ) (S as variables, current desired output, FFNN models. CCNN4 4 lagged inputs t-1 , S t-2 t-3 t-4 outperformed other lowest RMSE = 95.02 mg/l highest N 0.0.662, WI 0.890 LM 0.668 Station while same secure best 53.71 0.785, 0.936 0.788 Station. result shows better than predicting both Overall, showed forecasting potential stations, demonstrating their applicability hydrological complex relationships.

Язык: Английский

---

Multi-ahead electrical conductivity forecasting of surface water based on machine learning algorithms

Applied Water Science, Год журнала: 2023, Номер 13(10)

Опубликована: Сен. 8, 2023

Abstract The present research work focused on predicting the electrical conductivity (EC) of surface water in Upper Ganga basin using four machine learning algorithms: multilayer perceptron (MLP), co-adaptive neuro-fuzzy inference system (CANFIS), random forest (RF), and decision tree (DT). study also utilized gamma test for selecting appropriate input output combinations. results revealed that total hardness (TH), magnesium (Mg), chloride (Cl) parameters were suitable variables EC prediction. performance models was evaluated statistical indices such as Percent Bias (PBIAS), correlation coefficient (R), Willmott’s index agreement (WI), Index Agreement (PI), root mean square error (RMSE) Legate-McCabe (LMI). Comparing these indices, it observed RF model outperformed other algorithms. During training period, algorithm has a small positive bias (PBIAS = 0.11) achieves high with values ( R 0.956). Additionally, shows low RMSE value (360.42), relatively good efficiency (CE 0.932), PI (0.083), WI (0.908) LMI (0.083). However, during testing algorithm’s negative − 0.46) 0.929). decreases significantly (26.57), indicating better accuracy, remains 0.915), (0.033), (0.965) (− 0.028). Similarly, periods Prayagraj. PBIAS 0.50, bias. It an 368.3, 0.909, CE 0.872, 0.015, 0.921, 0.083. demonstrates slight 0.06. reduces to 24.1, improved accuracy. maintains 0.903) 0.878). (PI) increases 0.035, suggesting fit. is 0.960, accuracy compared value, while slightly 0.038. Based comparative algorithms, concluded performed than DT, CANFIS, MLP. recommended current month’s multi-ahead forecasting (EC t+1 , t+2 t+3 ) future studies basin. findings indicated DT had superior MLP CANFIS models. These can be applied monthly at both Varanasi Prayagraj stations

Язык: Английский

---

Modeling of soil moisture movement and wetting behavior under point-source trickle irrigation

Scientific Reports, Год журнала: 2023, Номер 13(1)

Опубликована: Сен. 11, 2023

Abstract The design and selection of ideal emitter discharge rates can be aided by accurate information regarding the wetted soil pattern under surface drip irrigation. current field investigation was conducted in an apple orchard SKUAST- Kashmir, Jammu a Union Territory India, during 2017–2019. objective experiment to examine movement moisture over time assess extent wetting both horizontal vertical directions point source irrigation with 2, 4, 8 L h −1 . At 30, 60, 120 min since beginning irrigation, pit dug across length area on order measure pattern. For measuring width depth, three replicas samples were collected according treatment average value considered. As result, 54 different experiments conducted, resulting digging pits [3 × 3 application times replications 2 (after 24 after application)]. This study utilized Drip-Irriwater model evaluate validate accuracy predictions fronts dynamics orientations. Results showed that modeled values very close actual values, mean absolute error 0.018, bias 0.0005, percentage 7.3, root square 0.023, Pearson coefficient 0.951, correlation 0.918, Nash–Sutcliffe efficiency 0.887. just measured at 14.65, 16.65, 20.62 cm; 16.20, 20.25, 23.90 20.00, 24.50, 28.81 cm , min, respectively, while depth observed 13.10, 20.44 15.10, 21.50, 26.00 19.40, 25.00, 31.00 respectively. flow rate from increased, amount dissemination grew (both immediately irrigation). contents 0.4300, 0.3808, 0.2298, 0.1604, 0.1600 −3 0.3841, 0.2385, 0.1607, 4 0.3852, 0.2417, 0.1608, 5, 10, 15, 20, 25 30 time. Similar distinct increments found findings suggest this simple model, which only requires soil, simulation parameters, is valuable practical tool for design. It provides patterns distribution single emitter, important effectively planning designing system. Investigating redistribution profile helps promote efficient

Язык: Английский

---

Seasonal rainfall pattern using coupled neural network-wavelet technique of southern Uttarakhand, India

Theoretical and Applied Climatology, Год журнала: 2024, Номер 155(6), С. 5185 - 5201

Опубликована: Март 28, 2024

Язык: Английский

---

A data-driven approach to river discharge forecasting in the Himalayan region: Insights from Aglar and Paligaad rivers

Results in Engineering, Год журнала: 2024, Номер 22, С. 102044 - 102044

Опубликована: Март 26, 2024

This study aims to better understand the time series forecasting of Aglar and Paligaad rivers' discharge (which has a significant impact on Himalayan river) using advanced methods such as Holt-Winters (HW) additive method, Simple exponential smoothing (SES), Non-seasonal ARIMA models. used antecedent information forecast next event. Comprehensive statistical examinations were conducted analyzed. The highly stochastic nature these river trends adds complexity efforts requires sophisticated modeling techniques that are capable capturing interpreting variability accurately. models proposed in current provide reliable for 15 months 31 recorded data. analysis shows both HW non-seasonal model results indicate decay end 2016 early 2017. best performance long-term forecasting, indicating sharp increase spring small during fall months. However, short-term non-ARIMA should show more promising results. methodologies substantially improve accuracy all consecutive perennial rivers. While presents discharge, generalizing findings other systems or different geographical regions may be problematic due varying hydrological characteristics environmental conditions, which need further study.

Язык: Английский

---