Digitizing Mt. Washington INPs Time Series Data

Abstract

As technology in computing has advanced, it has opened many new tools to aid researchers. Among these tools is neural networks. These will be utilized to digitize an old handwritten dataset taken from Mt. Washington Observatory. This dataset spans over 20 years and has the potential to aid our understanding of ice nucleating particles and their effects. Three overarching steps were identified when digitizing handwritten data: pre-processing, classification, and post-processing. The first two will be focused on here as they are the most important to the digitization process. It was also discovered how the errors in the pre-processing and classification affect the final model accuracy.