Machine Learning Classification to Inform the Search Space Extension Variant for the Analog Ensemble Technique

Technical and scientific fields, including artificial intelligence, require historical repositories of data in order to execute a task, learn a pattern, enable a prediction, or accomplish a function. However, an adequate historical repository at a specific location of interest are not always available for experimentation or analysis. The analog ensemble technique is used to determine a probabilistic outcome based on a historical repository of numerical weather prediction forecasts and corresponding observations. The analog ensemble technique is a technique that also requires a search history (sometimes called a training dataset) and can suffer from inadequate historical data.

Currently, the analog ensemble only considers historical data for each point in the domain (single location or grid). This technique can be limited by the availability of historical data at a particular site. Therefore, a variant called the Search Space Extension (SSE) which uses additional historical repositories from nearby sites (points) or areas (grids) to artificially extend the search space. In particular cases, the SSE fosters a better selection of analogs used to generate a probabilistic forecast while in other cases there is no marked improvement. This research seeks to utilize machine learning to classify geographic regions that result in improved analog selection and regions that do not. Data generated using the search space extension variant to the analog ensemble technique are used to explore classification techniques that can provide an understanding of when to use and when not to use the search space extension variant.