عنوان مقاله [English]
Background: The average air temperature prediction great importance in the field of water resources management, agriculture, water and a lot of things everyday. Air temperature is also one of the components of the hydrological and ecological models is input, as well as land evaluation models. On the other hand, because the weather has a significant impact on social life, and individual centers worldwide scientific research on climate issues raised as a priority, nearly as fundamental.
Quantitative prediction of air temperature is one of the most important elements in managing and programming of surface water resources, especially take suitable decisions in occurrence of drought event.
Objective: In this study, recorded data sets in Sari Station (located in Mazandaran province), were used to investigate the precision of different Air temperature prediction models. The wavelet neural network model and artificial neural network models selected for modeling of daily Air temperature, and the results were compared to examine the accuracy of studied models.
Methods: Daily Air temperature were selected and observed of this basin in the Sari station that were applied for calibration and validation of models. For this purpose, at first 80 % of daily Air temperature data (2002-2010) were selected to calibrate selected models, and 20 % of data (2010-2012) were used to validate models. For modeling, meteorological data from sari synoptic station (Mazandaran Province) were used. Humidity parameters, maximum temperature, minimum temperature, wind speed and evaporation in daily time scale during a ten year time period (2002-2012) are inputs of the network, and mean temperature as network output is selected.
Wavenet called wavelet-based neural network which combined with wavelet theory and neural networks have been created.It also have supportive of the benefits and features of neural networks and charm and flexibility and strong mathematical foundations and analysis of multi-scale wavelet . a combination of wavelet theory with neural network concepts to the creation of wavelet neural network and feedforward neural shock can be a good alternative for estimating approximate nonlinear functions .Feedforward neural network with sigmoid activation function is in the hidden layer While at the nerve shocked wavelet ,wavelet functions as activation function of hidden layer feedforward networks are considered, In both these networks and scale wavelet transformation parameters are optimized with their weight.
Artificial neural networks inspired by the brain's information processing systems, design and emerged intoTo help the learning process and with the use of processors called neurons trying to understand the inherent relationships between data mapping between input space and optimal space. Hidden layer or layers, the information received from the input layer and output layer are the processing and disposal.Based on the artificial neural network structure, its major features high processing speed, the ability to learn the pattern,The ability to extend the model after learning, flexibility against unwanted errorsNo disruption to error on the part of the connection due to weight distribution network. The first practical application of synthetic networks with the introduction of Multilayer Perceptron network wasConsultants. for training this network back propagation algorithm is used.The basis of this algorithm is based on error correction learning ruleThat consists of two main routes.By adjusting the parameters in the MLP model error signal and input signal occurs.Determine the number of layers and neurons is the most important issues in simulation with artificial neural network.
Criteria of correlation coefficient, root mean square error and Nash Sutcliff coefficient were used to evaluate and compare the performance of models. . Coefficient of correlation, root mean square error and Nash Sutcliffe coefficient was used to evaluate and compare the performance of the models.
Results: Results showed that all two models (in a structure), consisting gives better than results any other structure. And also, based on results according to the evaluation criterion, the models was used to wavelet neural network model, most accurate (R=0.999), and the lowest root mean square error (RMSE=0.001) and the highest standards Nash Sutcliffe (NS=0.998) the validation phase is capable.
Conclusions: Finally, wavelet neural network model outperformed the artificial neural network. So, wavelet neural network model can be effective in forecasting the daily Air temperature and in turn facilitate the development and implementation of Prevent drought will be useful. , and the use of wavelet neural network model can estimate the drought effectively, Which in return facilitates the development and implementation of management strategies to avoid drought.