The study, an extension of previous data mining work that was carried out on measurements acquired at the same station, uses an additional 10 years of data, which is expected to enhance the reliability of the results and accuracy of the conclusions. In contrast to the previous studies that used data mining methods based on clustering and classification, the new approach looked at the mutual information between observed new-particle formation events and a variety of measured variables. The method was shown to be powerful and computationally light, highlighting its potential as a useful tool.

"The conclusions reached using our analysis were found to agree with previous findings; however, the analysis was achieved without supervision and did not require a deep understanding of the physics," study coauthor Adam Foster says. "Previous studies have required involved field, lab, and theory work, so being able to get to where we have through data analysis is a very positive step for future studies."

The work showed that new-particle formation was strongly correlated with water content and sulfuric acid concentration, as well as other factors such as temperature, relative humidity, condensation sink (how quickly molecules and small particles condense onto existing particles), and radiation.

"We hope that the method will provide a robust first option for analyzing atmospheric data sets," says Foster. "We aim to extend our approach to data sets from other SMEAR stations, as well as to use it to investigate different phenomena and their influences."

These findings are expected to be implemented widely in the field of atmospheric science. It is hoped that the effect of other variables such as volatile organic compounds and aerosol particles below 3 nm can be assessed using the reported approach.