Saccharide concentration prediction from proxy-sea surface microlayer samples analyzed via ATR-FTIR spectroscopy and quantitative machine learning

The physical and chemical properties of the sea surface microlayer (SSML) are dynamic and complex. With an enrichment of organics from dissolved organic carbon (DOC) and many mechanisms for their release into the atmosphere, high-throughput analysis of SSML samples is necessary. Collection of more detailed information about the SSML would enable greater understanding of the release of ice nucleating and cloud condensation particles and provide critical feedback for climate models. The work presented herein details an investigation to determine the most accurate and precise machine learning (ML) model for analyzing SSML samples. Support vector regression (SVR) models predict the true saccharide concentration best and we evaluate unknown SSML samples using the model to predict the amount of carbohydrate present. Model predictions were 60-90 mM saccharide concentrations from SSML samples. Our work presents an application combining fast spectroscopic techniques with ML to analyze SSML chemistry more efficiently, without sacrificing accuracy and precision.