A new Artificial Intelligence-based year-round temperature prediction model has been developed for use on the AEMN website. While previous temperature prediction models were developed specifically for use during the winter and early spring season, the new model simultaneously improves accuracy during this period while expanding coverage to the rest of the year. The model provides a series of hourly air temperature predictions up to 12 hours ahead for any of the more than 70 current AEMN weather stations throughout the state of Georgia. In addition, the model may be applied to any new station that will be added to the network in the future without the need for additional model development.

The model makes use of artificial neural networks (ANNs) to predict temperature. As the name suggests, ANNs are machine-learning tools inspired by the functioning of human brains. Just as biological brains involve networks of brain cells, ANNs are composed of layers of simple processing nodes, or neurons. Each artificial neuron can respond to a collection of input signals by firing a single ouput signal of its own. Outputs of one layer may serve as inputs to subsequent layers. An ANN learns by tuning the relationships between its inputs and various layers of neurons. As a result, an ANN acts a complex, adjustable mathematical function. A twelve-hour plot of predicted air temperature can be generated every 15 minutes for any of the 70 AEMN weather station sites. See, for instance, examples for Albany, Alpharetta, Duluth, Tifton and Valdosta. This information can then be used to aid in decision support when localized future air temperature is an important factor. Examples include frost protection for fruits and ornamentals or heat stress for athletes.

Please contact Dr. Ron McClendon or Dr. Gerrit Hoogenboom if you would like to receive further information about this activity.

For scientific information, please check the following web site: Improving Air Temperature Prediction with Artificial Neural Networks

This work was made possible through a Partnership with USDA’s Risk Management Agency.