Search

Accurate streamflow (Qt) prediction can provide critical information for urban hydrological management strategies such as flood mitigation, long-term water resources management, land use planning and agricultural and irrigation operations. Since the mid-20th century, Artificial Intelligence (AI) models have been used in a wide range of engineering and scientific fields, and their application has increased in the last few years. In this study, the predictive capabilities of the reduced error pruning tree (REPT) model, used both as a standalone model and within five ensemble-approaches, were evaluated to predict streamflow in the Kurkursar basin in Iran. The ensemble-approaches combined the REPT model with the bootstrap aggregation (BA), random committee (RC), random subspace (RS), additive regression (AR) and disjoint aggregating (DA) (i.e. BA-REPT, RC-REPT, RS-REPT, AR-REPT and DA-REPT). The models were developed using 15 years of daily rainfall and streamflow data for the period 23 September 1997 to 22 September 2012. A set of eight different input scenarios was constructed using different combinations of the input variables to find the most effective scenario based on the linear correlation coefficient. A comprehensive suite of graphical (time-variation graph, scatter-plot, violin plot and Taylor diagram) and quantitative metrics (root mean square error (RMSE), mean absolute error (MAE), Nash-Sutcliff efficiency (NSE), Percent of BIAS (PBIAS) and the ratio of RMSE to the standard deviation of observation (RSR)) was applied to evaluate the prediction accuracy of the six models developed. The outcomes indicated that all models performed well but the AR-REPT outperformed all the other models by rendering lower errors and higher precision across a number of statistical measures. The use of the BA, RC, RS, AR and DA models enhanced the performance of the standalone REPT model by about 26.82%, 18.91%, 7.69%, 28.99% and 28.05% respectively.

Direct soil temperature (ST) measurement is time-consuming and costly; thus, the use of a simple and cost-effective machine learning (ML) tool is helpful. In this study, ML approaches, including KStar, instance-based K-nearest learner (IBK) and locally weighted learner (LWL) coupled with resampling algorithms of bagging (BA) and dagging (DA) were developed and tested for multi-step ahead (3, 6 and 9 days ahead) ST forecasting. In addition, a linear regression model (LR) was used as a benchmark to compare the results. A dataset with daily ST time-series (as models’ output) along with meteorological data (mean (TMean), minimum (TMin) and maximum (TMax) air temperature, evaporation (Eva), sunshine hours (SSH) and solar radiation (SR); as models’ input) were collected at Isfahan synoptic station (Iran), in a farmland, during 13 years (1992–2005) at 5 and 50 cm soil depths. Six different input combination scenarios were proposed to the models based on Pearson’s correlation coefficients between inputs and outputs. For the model building, we used 70% of the data and the remaining 30% was considered for model evaluation through different visual and quantitative metrics. Our findings showed that variable TMean is the most effective input variable for ST forecasting in most of the developed algorithms, while in some cases the combination of several variables including TMean, TMax and as well as the integration of TMean, TMax, TMin, Eva and SSH proved to be the best input combinations. Among the evaluated models, KStar showed more compatibility with the BA algorithm, while, in most cases and depending on soil depth, IBK and LWL obtained more accurate results when they were hybridized with DA. For soil depth of 5 cm, BA-KStar has superior performance (i.e. Nash-Sutcliffe Efficiency (NSE) = 0.90, 0.87 and 0.85 for 3, 6 and 9 months ahead forecasting, respectively) while for soil depth of 50 cm, DA-KStar outperforms other algorithms (i.e. NSE = 0.88, 0.89 and 0.89 for 3, 6 and 9 months ahead forecasting, respectively). Also, results confirmed that all hybrid models had higher prediction capability than the LR model.

Land subsidence (LS), which mainly results from poor watershed management, is a complex and nonlinear phenomenon. In the present study, LS at a country-wide assessment of Iran was mapped by using several geo-environmental conditioning factors (namely, altitude, slope degree and aspect, plan and profile curvature, distance from a river, road or fault, rainfall, geology and land use) into a machine learning algorithm-based artificial neural network (ANN), and a powerful group method of data handling (GMDH). The total dataset includes historical LS and non-LS locations, identified by the interferometric synthetic aperture radar (InSAR). The whole dataset was divided into two subsets at a ratio of 70:30 for training and validating the model, respectively. ANN- and GMDH-based LS maps were evaluated using receiver-operator characteristic (ROC) curves. The information gain ratio (IGR) was calculated to determine the relative importance of the conditioning factors. The results showed that all of the considered factors contributed significantly to the LS mapping in Iran, with geology having the strongest impact. According to the ROC curve analysis, both ANN and GMDH-based LS maps were accurate, but the map obtained by the GMDH approach had a higher accuracy than that of ANN. Southwestern, northeastern and some parts of the central region of Iran were shown to be susceptible to LS in the future. According to the GMDH susceptibility map, 10% of Iran exhibits high or very high susceptibility to LS in the future. The provinces of Hamedan and Khouzestan had the highest percentage of areas at risk of LS. According to the InSAR deformation map, 39%, 20%, 25%, 13% and 3% of the investigated areas are subject to a yearly LS of −1 to −2.5, −2.5 to −5, −5 to −7.5, −7.5 to −10 and −10 to −20 cm, respectively. The province of Razavi Khorasan in the northeast of Iran had the largest area (about 3500 km2) vulnerable to LS occurrence. Based on the LS susceptibility map, the provinces of Ardebil, Kurdistan, West and East Azerbaijan, Sistan and Baluchistan and Kermanshah, although not currently undergoing a high rate of LS, will be at high risk of severe LS in the future.

Labyrinth weirs are utilized to increase the weir crest length to transport a greater discharge during floods in contrast to conventional weirs. Nevertheless, due to the increased geometric complexity of labyrinth weirs, determination of accurate discharge coefficients and accordingly, head-discharge ratings are quite essential issues in practical application. Hence, as a first step the present study proposes the following eight standalone algorithms: decision table (DT), Kstar, least median square (LMS), M5 prime (M5P), M5 rule (M5R), pace regression (PR), random forest (RF), and sequential minimal optimization (SMO). Then, applying the stacking (ST) algorithm, these stand-alone models were hybridized to develop ST-LMS, ST-PR, ST-SMO, ST-Kstar, ST-DT, ST-M5R, ST-M5P, and ST-RF to predict the discharge coefficient (Cd) for sharp-crested labyrinth weirs. Modeling resulted in 123 experimental data sets including consideration of vertex angle (θ), channel width (B), head over the crest of the weir (h), crest heights (W), crest length of the weir (L), Cd, and flow discharge (Q). These effective variables were re-arranged in the form of several independent dimensionless parameters (θ, h/W, L/B, L/h, Froude number (Fr), B/W and L/W) to predict Cd as an output. Datasets were randomly divided into two groups; 70% of data used for model training while 30% used for model validation. The accuracy of the developed models was examined in terms of different statistical error measurement criteria of visually-based (line graph, scatter plot, box plot) and quantitative-based [root mean square error (RMSE), mean absolute error (MAE), the Nash-Sutcliffe efficiency (NSE), and percentage of bias (PBIAS)]. Results illustrate that h/W and B/W parameters have the highest and lowest effect on the Cd prediction, respectively. It was found that the most effective input combination included all input parameters except B/W. According to NSE, all developed algorithms provided accurate performances, while ST-Kstar has the highest prediction power (NSE=0.976, RMSE=0.011, MAE=0.008, PBIAS=0.027). Through incorporation of predicted Cd into discharge equation, promising results are obtained for accurate discharge computation.

The accurate forecasts and estimations of the amount of sediment transported by rivers are critical concerns in water resource management and soil and water conservation. The identification of appropriate and applicable models or improvements in existing approaches is needed to accurately estimate the suspended sediment concentration (SSC). In recent decades, the utilization of intelligent models has substantially improved SSC estimation. The identification of beneficial and proper input parameters can greatly improve the performance of these smart models. In this regard, we assessed the C-factor of the revised universal soil loss equation (RUSLE) as a new input along with hydrological variables for modeling SSC. Four data-driven models (feed-forward neural network (FFNN); support vector regression (SVR); adaptive neuro-fuzzy inference system (ANFIS); and radial basis function (RBF)) were applied in the Boostan Dam Watershed, Iran. The cross-correlation function (CCF) and partial autocorrelation function (PAFC) approaches were applied to determine the effective lag times of the flow rate and suspended sediment, respectively. Additionally, several input scenarios were constructed, and finally, the best input combination and model were identified through trial and error and standard statistics (coefficient of determination (R2); root mean square error (RMSE); mean absolute error (MAE); and Nash–Sutcliffe efficiency coefficient (NS)). Our findings revealed that using the C-factor can considerably improve model efficiency. The best input scenario in which the C-factor was combined with hydrological data improved the NS by 16.4%, 21.4%, 0.17.5%, and 23.2% for SVR, ANFIS, FFNN, and RBF models, respectively, compared with the models using only hydrological inputs. Additionally, a comparison among the different models showed that the SVR model had about 4.1%, 13.7%, and 23.3% (based on the NS metric) higher accuracy than ANFIS, FFNN, and RBF for SSC estimation, respectively. Thus, the SVR model using hydrological data along with the C-factor can be a cost-effective and promising tool in SSC prediction at the watershed scale.

Landslides are a geological hazard that can pose a serious threat to human health and the environment of highlands or mountain slopes. Landslide susceptibility mapping is an essential tool for predicting and mitigating landslides. This study aimed to investigate the application of deep learning algorithms based on convolutional neural networks (CNNs) with metaheuristic optimization algorithms, namely the grey wolf optimizer (GWO) and imperialist competitive algorithm (ICA), to landslide susceptibility mapping. The study area was Icheon City, South Korea, for which an accurate landslide inventory dataset was available. The landslide inventory map was prepared and randomly divided into datasets of 70% for training and 30% for validation. Additionally, 18 landslide-related factors, including geo-environmental and topo-hydrological factors, were considered as predictive variables. The models were compared using area under the curve (AUC) values in receiver operating characteristic (ROC) curve analysis. The validation results showed that optimized models based on CNN-GWO (AUC = 0.876, RMSE = 0.08) and CNN-ICA (AUC = 0.852, RMSE = 0.09) outperformed the standalone CNN model (AUC = 0.847, RMSE = 0.12). Nevertheless, the CNN model outperformed previous research that used a machine learning algorithm alone. Thus, the deep learning algorithm with optimization algorithms proposed in this study can generate more suitable models for landslide susceptibility mapping in the study area due to its improved accuracy.