A novel hybrid of meta-optimization approach for flash flood-susceptibility assessment in a monsoon-dominated watershed, Eastern India

Environmental Earth Sciences, Journal Year: 2022, Volume and Issue: 81(5)

Published: Feb. 21, 2022

Language: Английский

---

Deep learning versus gradient boosting machine for pan evaporation prediction

Engineering Applications of Computational Fluid Mechanics, Journal Year: 2022, Volume and Issue: 16(1), P. 570 - 587

Published: Feb. 7, 2022

In the present study, two innovative techniques namely, Deep Learning (DL) and Gradient boosting Machine (GBM) models are developed based on a maximum air temperature 'univariate modeling scheme' for monthly pan evaporation (Epan) process. Monthly used to build predictive models. These evaluating prediction Kiashahr meteorological station located in north of Iran Ranichauri positioned Uttarakhand State India. Findings indicated that deep learning model was found best at testing datasets MAE (0.5691, mm/month), RMSE (0.7111, NSE (0.7496), IOA (0.9413). It can be concluded semi-arid climate both methods had good capability Epan. However, DL predicted Epan better than GBM. Moreover, highest accuracy also observed terms = 0.3693 mm/month, 0.4357 0.8344, & 0.9507 stage. Overall, results expose superior performance DL-based study stations utilized various other environmental modeling.

Language: Английский

---

Flash-flood hazard using deep learning based on H2O R package and fuzzy-multicriteria decision-making analysis

Journal of Hydrology, Journal Year: 2022, Volume and Issue: 609, P. 127747 - 127747

Published: March 24, 2022

Language: Английский

---

Modeling Various Drought Time Scales via a Merged Artificial Neural Network with a Firefly Algorithm

Hydrology, Journal Year: 2023, Volume and Issue: 10(3), P. 58 - 58

Published: Feb. 27, 2023

Drought monitoring and prediction have important roles in various aspects of hydrological studies. In the current research, standardized precipitation index (SPI) was monitored predicted Peru between 1990 2015. The study proposed a hybrid model, called ANN-FA, for SPI time scales (SPI3, SPI6, SPI18, SPI24). A state-of-the-art firefly algorithm (FA) has been documented as powerful tool to support modeling issues. ANN-FA uses an artificial neural network (ANN) which is coupled with FA Lima via other stations. Through intelligent utilization series from neighbors’ stations model inputs, suggested approach might be used forecast at meteorological station insufficient data. To conduct this, SPI3, SPI24 were modeled using stations’ datasets Peru. Various error criteria employed investigate performance model. Results showed that effective promising drought also multi-station strategy lack results can help predict mean absolute = 0.22, root square 0.29, Pearson correlation coefficient 0.94, agreement 0.97 testing phase best estimation (SPI3).

Language: Английский

---

Reinforcing long lead time drought forecasting with a novel hybrid deep learning model: a case study in Iran

Applied Water Science, Journal Year: 2025, Volume and Issue: 15(3)

Published: Feb. 18, 2025

Abstract Drought assessment is inherently complex, particularly under the influences of climate change, which complicates long-term forecasting. This study introduces a novel hybrid deep learning model, Deep Feedforward Natural Networks (DFFNN), enhanced by War Strategy Optimization (WSO), aimed at forecasting Standardized Precipitation Evapotranspiration Index (SPEI) for lead times one, three, six, nine, and twelve months. Key parameters DFFNN, including number neurons layers, rate, training function, weight initialization, were optimized using WSO algorithm. The model’s performance was validated against two established optimizers: Particle Swarm (PSO) Genetic Algorithm (GA). Evaluations conducted synoptic stations with distinct climatic conditions in Iran. Results demonstrated that WSO-DFFNN model achieved superior SPEI 12 (t + 1) correlation coefficient (r) 0.9961 Normalized Root Mean Square Error (NRMSE) 0.1028; 3) r = 0.8856 NRMSE 0.1833; 6) 0.8573 0.2203; 9) 0.7951 0.2479; 12) 0.7840 0.3279 Chabahar station. Additionally, outperformed 0.9118 0.1704; 0.8386 0.2048; 0.7602 0.2919; 0.6379 0.2843; 0.6044 0.3463 Anzali results obtained from this have potential to improve drought management strategies.

Language: Английский

---

Implementation of hybrid neuro-fuzzy and self-turning predictive model for the prediction of concrete carbonation depth: A soft computing technique

Results in Engineering, Journal Year: 2021, Volume and Issue: 10, P. 100228 - 100228

Published: May 26, 2021

Carbonation is one of the critical problems that affects durability reinforced concrete; it a reaction between CO2 gas and Ca (OH)2 when H2O available, which forms powdery CaCO3 alters microstructure concrete by reducing its pH level initiating corrosion reduces structure's service life. This study provides experimental information on carbonation depths samples from 10 separate existing structures, where five are located in inland area (Nicosia), while other coastal (Kyrenia) Turkish Republic North Cyprus. The found buildings have higher depth compared to buildings. building structures Cyprus exhibit rate than expected threshold within their life span. Constant values B were yielded, useful predicting depth. Using AI, potential Hybrid Neuro-fuzzy model, comprised an Adaptive Inference System (ANFIS), Extreme Learning Machine (ELM), Support Vector (SVM) Conventional Multilinear Regression (MLR) employed for estimation using data, including age, compressive strength, current density, constant. Four different performance indexes used verify modelling accuracy, namely Mean Absolute Error (MAE), Root Square (RMSE), Nash- Coefficient (NSE), Correlation (CC). results indicated AI models (ANFIS, ELM, SVM) performed better linear model with NSE-values 0.97 both testing training stages. also prediction skills ANFIS-M2 increased accuracy ELM-M2, SVM-M2, MLR-M2, ANFIS-M1 ELM-1, SVM-1 MLR-1 terms accuracy. final outcomes capability non-linear Cd.

Language: Английский

---

The potential of a novel support vector machine trained with modified mayfly optimization algorithm for streamflow prediction

Hydrological Sciences Journal, Journal Year: 2021, Volume and Issue: 67(2), P. 161 - 174

Published: Nov. 30, 2021

This paper focuses on the development of a robust accurate streamflow prediction model by balancing abilities exploitation and exploration to find best parameters machine learning model. To do so, simulated annealing (SA) algorithm is integrated with mayfly optimization (MOA) as SAMOA determine optimal hyper-parameters support vector regression (SVR) overcome weakness MOA method. The proposed method compared classical SVR hybrid SVR-MOA. examine accuracy selected methods, monthly hydroclimatic data from Jhelum River Basin used predict basis RMSE, MAE, NSE, R2 indices. Test results show that SVR-SAMOA outperformed SVR-MOA models. reduced errors models decreasing RMSE MSE 21.4% 14.7% 21.7% 15.1%, respectively, in test stage.

Language: Английский

---

A comparative analysis of data mining techniques for agricultural and hydrological drought prediction in the eastern Mediterranean

Computers and Electronics in Agriculture, Journal Year: 2022, Volume and Issue: 197, P. 106925 - 106925

Published: April 10, 2022

Drought is a natural hazard which affects ecosystems in the eastern Mediterranean. However, limited historical data for drought monitoring and forecasting are available Thus, implementing machine learning (ML) algorithms could allow prediction of future events. In this context, main goals research were to capture agricultural hydrological trends by using Standardized Precipitation Index (SPI) assess applicability four ML (bagging (BG), random subspace (RSS), tree (RT), forest (RF)) predicting events Mediterranean based on SPI-3 SPI-12. The results reveal that (SPI-12, −24) was more severe over study area, where most stations showed significant (p < 0.05) negative trend. accuracy varied relation implementation stage. training stage, RT outperformed other (Root mean square error (RMSE) = 0.3, Correlation Coefficient (r) 0.97); performance can be ranked as follows: > RF BG RSS both testing had highest correlation r (observed vs. predicted) (0.58–0.64) lowest RMSE (0.68–0.88). contrast, (0.3–0.41) (0.94–1.10) calculated algorithm. dynamic capturing, with correlation. validation satisfactory (RMSE 0.62–0.83, 0.58–0.79). output will help decision-makers mitigation plans new algorithms.

Language: Английский

---

Drought Forecasting: A Review and Assessment of the Hybrid Techniques and Data Pre-Processing

Hydrology, Journal Year: 2022, Volume and Issue: 9(7), P. 115 - 115

Published: June 26, 2022

Drought is a prolonged period of low precipitation that negatively impacts agriculture, animals, and people. Over the last decades, gradual changes in drought indices have been observed. Therefore, understanding forecasting essential to avoid its economic appropriate water resource planning management. This paper presents recent literature review, including brief description data pre-processing, data-driven modelling strategies (i.e., univariate or multivariate), machine learning algorithms advantages disadvantages), hybrid models, performance metrics. Combining various prediction methods create efficient models has become most popular use years. Accordingly, increasingly used for predicting drought. As such, these will be extensively reviewed, preprocessing-based parameter optimisation-based hybridisation components combination-based with models. In addition, using statistical criteria, such as RMSE, MAE, NSE, MPE, SI, BIC, AIC, AAD, evaluate

Language: Английский

---

Pre- and post-dam river water temperature alteration prediction using advanced machine learning models

Environmental Science and Pollution Research, Journal Year: 2022, Volume and Issue: 29(55), P. 83321 - 83346

Published: June 28, 2022

Language: Английский

---