Spring Symposium on UR and Community Engagement has ended
Back To Schedule
Tuesday, April 24 • 10:55am - 11:15am
Improving The Prediction Of Daily Maximum Temperatures Using A Neural Network

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

A weather forecaster uses model data, persistence, and historical data to predict values for weather variables on the following days. The application of a neural network to use the same data and train to identify a relationship between the predictors and the predicted value would result in a more refined source of data for weather forecasters to use. In this study a neural network is used to account for consistent bias in the models, overall trends in temperature, and current conditions to predict a maximum temperature for the next day. The model is trained on a dataset containing two models, historical daily maximum temperatures, and the previous day’s maximum temperature over two years from 2014 to 2016. The model is evaluated using data from 2016 to 2017, but will be able to take current data and predict the next day’s temperature with greater accuracy than model output. The neural network used is a multi-layer perceptron classifier that trains using backpropagation. The model will be assessed for accuracy and used to predict maximum daily temperatures for the Asheville area. This could result in improved reliability when predicting maximum temperatures. The model is easily scalable and could be used with various other weather variables to create more accurate data for forecasters to use and make short term weather forecasts that are more accurate than model output statistics.

Tuesday April 24, 2018 10:55am - 11:15am PDT
213 Rhoades Robinson Hall