No abstract

Downscaling of general circulation model (GCM) outputs extracted from CMIP5 datasets to monthly precipitation for the Gediz Basin, Turkey, under Representative Concentration Pathways (RCPs) was performed by statistical downscaling models, multi-GCM ensemble and bias correction. The output databases from 12 GCMs were used for the projections. To determine explanatory predictor variables, the correlation analysis was applied between precipitation observed at 39 meteorological stations located over the Basin and potential predictors of ERA-Interim reanalysis data. After setting both artificial neural networks and least-squares support vector machine-based statistical downscaling models calibrated with determined predictor variables, downscaling models producing the most suitable results were chosen for each meteorological station. The selected downscaling model structure for each station was then operated with historical and future scenarios RCP4.5, RCP6.0 and RCP8.5. Afterwards, the monthly precipitation forecasts were obtained from a multi-GCM ensemble based on Bayesian model averaging and bias correction applications. The statistical significance of the foreseen changes for the future period 2015-2050 was investigated using Student's t test. The projected decrease trend in precipitation is significant for the RCP8.5 scenario, whereas it is less significant for the RCP4.5 and RCP6.0 scenarios.

ABSTRACT:In this study, statistical downscaling of general circulation model (GCM) simulations to monthly inflows of Kemer Dam in Turkey under A1B, A2, and B1 emission scenarios has been performed using machine learning methods, multi-model ensemble and bias correction approaches. Principal component analysis (PCA) has been used to reduce the dimension of potential predictors of National Centers for Environmental Prediction and National Center for Atmospheric Research (NCEP/NCAR) reanalysis data. Then, the reasonable GCMs were selected by investigating the rank correlations between the selected predictors in NCEP/NCAR reanalysis data and those in GCMs for 20C3M scenario between periods 1979 and 1999. Upon the training of feedforward neural network (FFNN), least squares support vector machine (LSSVM) and relevance vector machine (RVM) downscaling models, the general performance of the downscaled predictions using NCEP/NCAR reanalysis data for Kemer watershed showed that the trained RVM model produced adequate results. The effectiveness of RVM model was illustrated by its integration with 20C3M scenario between periods 1979 and 1999 and A1B, A2, and B1 future climate scenarios between periods 2010 and 2039. Afterwards, the flow forecasts were obtained by building a multi-model ensemble through the selected GCMs followed by a bias correction approach. Finally, the significance of the probable changes in trends was identified through statistical tests based on the corrected forecasts. Results showed that decreasing flows trends in winter, spring and fall seasons have been foreseen over the study area for the period between 2010 and 2039.

Over the past decade, artificial neural networks (ANN) have been widely used in the runoff modeling studies. In spite of a number of advantages, ANN models have some drawbacks, including the possibility of getting trapped in local minima, over training, subjectivity in the determining of model parameters, initialization of the weights in each simulation randomly, and the components of its complex structure. In the past decade, a new alternative kernel‐based technique called a support vector machine (SVM) has been found to be popular in modeling studies because of its advantages over ANN. Least squares version of support vector machines (LS‐SVM) provides a computational advantage over standard support vector machines by converting quadratic optimization problem into a system of linear equations. The LS‐SVM method is preferred in this study. The main purposes of this study are to examine the applicability and capability of LS‐SVM for the prediction of runoff values of Tahtali and Gordes watersheds, which are the major surface water resources for the city of Izmir in Turkey, and to compare its performance with ANN and other traditional techniques such as autoregressive moving average and multiple linear regression models. For these purposes, meteorological data (rainfall and temperature) and lagged data of runoff were used in modeling applications. Some favorite statistical performance evaluation measures were used to assess models. The results in study indicate that the LS‐SVM and ANN methods are successful tools to model the monthly runoff series of two study regions and can give better prediction performances than conventional statistical models. Although these two methods are powerful artificial intelligence techniques, LS‐SVM makes the running time considerably faster with the same or higher accuracy. In terms of accuracy, the LS‐SVM models, which involve different normalization types, resulted in increased accuracy to that of the ANN models. Copyright © 2012 John Wiley & Sons, Ltd.

No abstract

No abstract

No abstract

An experimental investigation was conducted to evaluate the suitability of hazelnut oil methyl ester (HOME) for engine performance and exhaust emissions responses of a turbocharged direct injection (TDI) diesel engine. HOME was tested at full load with various engine speeds by changing fuel injection timing (12, 15, and 18 deg CA) in a TDI diesel engine. Response surface methodology (RSM) and least-squares support vector machine (LSSVM) were used for modeling the relations between the engine performance and exhaust emission parameters, which are the measured responses and factors such as fuel injection timing (t) and engine speed (n) parameters as the controllable input variables. For this purpose, RSM and LSSVM models from experimental results were constructed for each response, namely, brake power, brake-specific fuel consumption (BSFC), brake thermal efficiency (BTE), exhaust gas temperature (EGT), oxides of nitrogen (NOx), carbon dioxide (CO2), carbon monoxide (CO), and smoke opacity (N), which are affected by the factors t and n. The results of RSM and LSSVM were compared with the observed experimental results. These results showed that RSM and LSSVM were effective modeling methods with high accuracy for these types of cases. Also, the prediction performance of LSSVM was slightly better than that of RSM.