Rainfall–runoff modelling is complicated due to numerous complex interactions and feedback in the water cycle among precipitation and evapotranspiration processes, and also geophysical characteristics. Consequently, the lack of geophysical characteristics such as soil properties leads to difficulties in developing physical and analytical models when traditional statistical methods cannot simulate rainfall–runoff accurately. Machine learning techniques with data-driven methods, which can capture the nonlinear relationship between prediction and predictors, have been rapidly developed in the last decades and have many applications in the field of water resources. This study attempts to develop a novel 1D convolutional neural network (CNN), a deep learning technique, with a ReLU activation function for rainfall–runoff modelling. The modelling paradigm includes applying two convolutional filters in parallel to separate time series, which allows for the fast processing of data and the exploitation of the correlation structure between the multivariate time series. The developed modelling framework is evaluated with measured data at Chau Doc and Can Tho hydro-meteorological stations in the Vietnamese Mekong Delta. The proposed model results are compared with simulations of long short-term memory (LSTM) and traditional models. Both CNN and LSTM have better performance than the traditional models, and the statistical performance of the CNN model is slightly better than the LSTM results. We demonstrate that the convolutional network is suitable for regression-type problems and can effectively learn dependencies in and between the series without the need for a long historical time series, is a time-efficient and easy to implement alternative to recurrent-type networks and tends to outperform linear and recurrent models.

Landslides are dangerous events which threaten both human life and property. The study aims to analyze the landslide susceptibility (LS) in the Kysuca river basin, Slovakia. For this reason, previous landslide events were analyzed with 16 landslide conditioning factors. Landslide inventory was divided into training (70% of landslide locations) and validating dataset (30% of landslide locations). The heuristic approach of Fuzzy Decision Making Trial and Evaluation Laboratory (FDEMATEL)-Analytic Network Process (ANP) was applied first, followed by bivariate Frequency Ratio (FR), multivariate Logistic Regression (LR), Random Forest Classifier (RFC), Naïve Bayes Classifier (NBC) and Extreme Gradient Boosting (XGBoost), respectively. The results showed that 52.2%, 36.5%, 40.7%, 50.6%, 43.6% and 40.3% of the total basin area had very high to high LS corresponding to FDEMATEL-ANP, FR, LR, RFC, NBC and XGBoost model, respectively. The analysis revealed that RFC was the most accurate model (overall accuracy of 98.3% and AUC of 97.0%). Besides, the heuristic approach of FDEMATEL-ANP model (overall accuracy of 93.8% and AUC of 92.4%) had better prediction capability than bivariate FR (overall accuracy of 86.9% and AUC of 86.1%), multivariate LR (overall accuracy of 90.5% and AUC of 91.2%), machine learning NBC (overall accuracy of 76.3% and AUC of 90.9%) and even deep learning XGBoost (overall accuracy of 92.3% and AUC of 87.1%) models. The study revealed that the FDEMATEL-ANP outweighed the NBC and XGBoost machine learning models, which suggests that heuristic methods should be tested out before directly applying machine learning models.

Groundwater has a significant contribution to water storage and is considered to be one of the sources for agricultural irrigation; industrial; and domestic water use. The Gravity Recovery and Climate Experiment (GRACE) satellite provides a unique opportunity to evaluate terrestrial water storage (TWS) and groundwater storage (GWS) at a large spatial scale. However; the coarse resolution of GRACE limits its ability to investigate the water storage change at a small scale. It is; therefore; needed to improve the resolution of GRACE data at a spatial scale applicable for regional-level studies. In this study; a machine-learning-based downscaling random forest model (RFM) and artificial neural network (ANN) model were developed to downscale GRACE data (TWS and GWS) from 1° to a higher resolution (0.25°). The spatial maps of downscaled TWS and GWS were generated over the Indus basin irrigation system (IBIS). Variations in TWS of GRACE in combination with geospatial variables; including digital elevation model (DEM), slope; aspect; and hydrological variables; including soil moisture; evapotranspiration; rainfall; surface runoff; canopy water; and temperature; were used. The geospatial and hydrological variables could potentially contribute to; or correlate with; GRACE TWS. The RFM outperformed the ANN model and results show Pearson correlation coefficient (R) (0.97), root mean square error (RMSE) (11.83 mm), mean absolute error (MAE) (7.71 mm), and Nash–Sutcliffe efficiency (NSE) (0.94) while comparing with the training dataset from 2003 to 2016. These results indicate the suitability of RFM to downscale GRACE data at a regional scale. The downscaled GWS data were analyzed; and we observed that the region has lost GWS of about −9.54 ± 1.27 km³ at the rate of −0.68 ± 0.09 km³/year from 2003 to 2016. The validation results showed that R between downscaled GWS and observational wells GWS are 0.67 and 0.77 at seasonal and annual scales with a confidence level of 95%, respectively. It can; therefore; be concluded that the RFM has the potential to downscale GRACE data at a spatial scale suitable to predict GWS at regional scales.

The reliable prediction of dissolved oxygen concentration (DO) is significantly crucial for protecting the health of the aquatic ecosystem. The current research employed four different single AI-based models, namely long short-term memory neural network (LSTM), extreme learning machine (ELM), Hammerstein-Weiner (HW) and general regression neural network (GRNN) for modeling the DO concentration of Kinta River, Malaysia using available water quality (WQ) parameters. Afterwards, the first scenario used four different ensemble techniques (ET). Two linear, i.e. simple averaging ensemble (SAE) and weighted averaging ensemble (WAE) and two nonlinear namely; backpropagation neural network ensemble (BPNN-E) and HW ensemble (HW-E). The second scenario employed a hybrid random forest (RF) ensemble in order to enhance the prediction accuracy of the single models. The WQ parameters were subjected to a different pre-analysis test to ascertain their stability. The four-model combinations are generated using the nonlinear sensitivity input selection approach. The modeling performance was assessed using the statistical measures of Nash-Sutcliffe coefficient efficiency (NSE), Willmott’s index of agreement (WI), root mean square error (RMSE), mean absolute error (MAE) and mean square error (MSE) and correlation coefficient (CC). The results of the single AI-based models demonstrated that HW (M3) served as the best model for predicting DO concentration. For ensemble results, BPNN-E (WI=0.9764) was superior to the other three ET with average decreased of more than 2% with regards to MAE. Investigation on the hybrid RF ensemble demonstrated the reliable accuracy for all the hybrid models with better predictive skill shown by the HW-RF (CC=0.981) ensemble. The overall results verified the promising impact of HW-M3, ET and hybrid RF ensemble for the prediction of the DO concentration in the Kinta River, Malaysia.