Do neural networks dream of falling snow?

Developing a deep learning radar-based precipitation retrieval

DeepPrecip network compute graph rendering — Image by Graphcore and Author

The application of machine learning (ML) in the Geosciences is not a new idea. Early examples of k-means clustering, Markov chains and decision trees have been actively used in several geographical contexts since the mid-1960’s (Preston et al., 1964; Krumbein et al., 1969; Newendorp, 1976).

However, advancements in cloud computing resources from companies like Graphcore, Amazon and …

editors pick networks neural networks precipitation remote-sensing snow