Early Prediction Methods
Prediction Based on January SST Maps
The early prediction, published in February, is based on a prediction method I proposed in 2020, which analyzes January Sea Surface Temperature (SST) maps using a Convolutional Neural Network. Each hurricane season is classified into two classes: high and medium-low intensity. The intensity is measured by the season's Accumulated Cyclone Energy (ACE) index. This predictor has an average accuracy of 83.33%.
The paper describing my method that has been published in the IEEE Oceans 2022 Conference.
*Note: the prediction model was retrained in March, 2023, based on updated data visualization maps on NOAA's website.
Sea Surface Temperature Map Sample for January, 2019 (NOAA, 2019).
Updated Prediction
Prediction Based on July OLR Maps
The updated prediction, published in August, is based on a prediction method that I proposed in 2022, which analyzes July Outgoing Longwave Radiation (OLR) maps using a Convolutional Neural Network. OLR measures the amount of energy that the Earth’s surface, oceans, and atmosphere release into space (National Center for Atmospheric Research Staff, 2014). Each hurricane season is classified into three classes: high, medium, and low intensity. The intensity is measured by the season's Accumulated Cyclone Energy (ACE) index. This predictor has an average accuracy of 77.33% (for 3 classes).
Outgoing Longwave Radiation Map Sample for July, 2019 (NOAA, 2019).