prediction of maximum temperatures through artificial neural network case study: ardabli township

â  the prediction of maximum temperatures as one of the most important climatic parameters due to climate change, global warming and the recent drought will provide definitely more opportunity for planning and the provision of necessary arrangements for the planners. maximum temperatures are much important in management of natural and water resources, agriculture, development of pests and diseases, flood and snow melting, evaporation and transpiration, drought, etc. today, with developing of intelligent and expanded models in experimental science, including climatology, the necessity for alternative compared to old models will important. one of these methods is artificial neural network derived from artificial intelligence components which has important applications in the field of atmospheric sciences through prediction and calculation of climatic parameters. in this study, using the variables including average relative humidity, average wind speed, total sunshine, average minimum temperature and monthly average maximum temperature as input multi layer perceptron (mlp) network, have been predicted monthly average maximum temperature in ardabil synoptic station. examined parameters include data period 1985 to 2005.out of 21 years statistical period about 85 percent of the available data, meaning 18 years (216 months) were used for training the network and 3 years (36 months) remaining in the test stage were applied. for this purpose, facilities and functions available in matlab software were made and for every month a network was designed with under 5 percent error. after studying network performance indicators, including the correlation coefficient, root mean square error, mean squares error, mean absolute error, mean percentage it was observed that the maximum temperature predicted with acceptable accuracy has been made in such away that the rate of correlation coefficient was 0.99 and the maximum difference with the real data was 0.83 ⺠c.