Inderscience Publishers

Prediction of meteorological variables using artificial neural networks

In the present study, artificial neural network (ANN) models are developed to predict seven meteorological parameters such as maximum temperature, minimum temperature, relative humidity, wind speed, sunshine hours, dew point and evaporation individually. To predict each of the meteorological parameters, recurrent MISO-ANN network is used. Totally seven ANN models are developed and the development includes determination of networks architecture, optimisation of the connection weights, model testing and the choice of performance criteria. Out of 120 datasets of each parameter (ten years monthly data), 70% and 30% of the length is used for training and testing respectively, achieved through trial and error procedure. The results showed that recurrent MISO-ANN models has captured the patterns in the input data exceptionally well with a correlation coefficient varying from 0.93 to 0.98 during training and from 0.90 to 0.97 during testing. The reason for the high accuracy may be due to the well defined and very good pattern in the observed time series. However, parameters such as wind speed, sunshine hours and evaporation models showed slightly poor performance during testing due to irregular patterns.

Keywords: maximum temperature, evaporation, meteorological parameters, recurrent ANNs, artificial neural networks, time series modelling, wind speed, sunshine hours, evaporation models, relative humidity, dew point

Customer comments

No comments were found for Prediction of meteorological variables using artificial neural networks. Be the first to comment!