Decoding Nonlinear Signals In Large Observational Datasets

Large observational datasets of precipitation are analyzed to decode complex patterns and improve understanding of climate processes. A robust, multidimensional dataset was curated over 10 years from 10 global sites, capturing detailed microphysical properties like particle size, shape, and fall speed using a specialized video instrument. This high-quality data, combined with meteorological variables, was made publicly available along with a Python API for exploration. The analysis begins by applying Principal Component Analysis (PCA) to a filtered subset of snowfall events to explore linear embeddings and uncover latent features.