Description of the project taken from the final report I wrote at the end of the summer:

“During oceanic expeditions, pre-modern sailors meticulously recorded information on their longitude and latitude, the local wind conditions, and the state of the sea. The nature of the data changed as new measurement instruments and reference points were introduced. For a long time, sailors provided qualitative recordings of wind speed instead of quantitative (e.g.: “light breeze” instead of 5 meters/second). For that reason, this data requires additional processing before being usable for climate modeling. 

In particular, the unique phrases used in wind descriptions must be classified into one of thirteen base wind force levels to be accurately translated into a numerical value. Manually categorizing this information takes an incredibly long time.  Because of historical weather data’s importance for climate science, we investigated if machine learning could speed up this process while producing accurate results. 

Here we show that k-means nearest neighbors clustering, while efficient, generates outputs with reduced accuracy when compared to the data classified by humans. However, there is a noticeable improvement in the quality of the clustering when we introduce the Beaufort Wind Force Scale’s thirteen categories as starting centroids. 

These results show that machine learning could be a useful tool for wind term processing and that well-placed human input aids in the accuracy of outcomes. Therefore, we intend to employ cross-validation methods to help with the interpretability of the machine models utilized. Additionally, we plan to test how neural network clustering fares, as it has more room for iteration and human input.”