Comparison of individual, combined and series artificial neural network models for air temperature and dew point temperature prediction

Agricultural producers suffer economic losses in crops and livestock due to frost, freeze and heat stress. Freezing temperatures are responsible for reduced crop yields due to damage to leaves and fruit, especially blueberries and peaches. Heat stress could severely impact livestock, and similar temperature conditions could cause heat stroke in humans. Accurate prediction of air temperature and dew point temperature can help managers minimize the losses to crops and livestock. The research presented in this thesis compares artificial neural network models predicting air and dew point temperatures for twelve prediction horizons. The models are compared, using mean absolute error and number of prediction anomaly, with current web-based models available on the University of Georgias Automated Environmental Monitoring Network website, www.georgiaweather.net.