Data-Mining Similar Scenarios for Uncertainty Quantification of Solar Wind Predictions at L1
Description
Accurate Uncertainty Quantification (UQ) for space weather forecasts is an ever-important supplementary variable to enable accurate risk response. Modeling uncertainty is itself a “model of a model”, and one of the best datasets to describe a model’s performance is a past database of it’s predictions and after-the-fact observations. In this work, we develop a method based on k-NN and kernel regression to quantify uncertainty in the WSA solar wind model and it’s predictions of the solar wind speed at L1. By constructing state vectors that describe the current forecasting context— recent observations, recent predictions, and future predictions, we build a catalog of “similar scenarios” from past data. With a set of similar scenarios at each timestep, we can base our uncertainty on the performance in those cases. This approach—suitable for low-dimensional datasets such as time series—is extremely fast and interpretable. We find that the resulting uncertainty estimates naturally capture structured patterns in forecast error, such as shifts between solar minimum and maximum, and periodic features on the scale of half a solar rotation.
Detailed example
Uncertainty Estimation (sigmas)
AI / analytics pattern
Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relationships.
Automation level / stage
a) Pre-deployment – The use case is in a development or acquisition status.
Expected benefit
Being integrated into WSA software for operational use in Moon 2 Mars and NOAA Space Weather Prediction Center
Controls / human review
ATO: Not reported; PIA: Not published