Mid-Range Forecasting of Solar-Wind: A Case Study of Building Regression Model with Space Weather Forecast Testbed (SWFT)

Abstract:

The Space Weather Forecast Testbed (SWFT) is developed by a team of space weather scientists

and mathematicians at the University of Southern California (USC) and Jet Propulsion Laboratory

(JPL) to foster the creation of models for space weather forecast by exploration of existing historic

data using techniuqes of machine-learning. As an example to demonstrate the potential power

of SWFT, we present in this paper a multilinear regression based forecast model for solar wind.

Solar wind is one of the key drivers for numerous physics models for space weather including

thermosphere and ionosphere models. Many attempts have been made to produce forecast for

solar wind. SWFT provides an unified framework for forecast model formulation, training and

performance assessment. In particular, the preparation of training and validation data by SWFT

takes into account of realistic constraints on data latency and forecast lead time. In developping a

solar wind forecast model, SWFT allows fast exploration of many meta-parameters such as the list

of variables and their time history used in constructing a model. We shall present the performance

impact as results of meta-parameter selection, as well as, performance of these models to existing

solar wind forecast models.