High-Level Project Summary
Project Icarus seeks to provide an easy way to correct discrepancies between solar flare data between the DSCOVER and WIND satellites. First, we derive Dynamic Time Warping mappings between the two satellites to synchronize the value of two datasets in time. These are then corrected using a feed forward network and LSTM machine learning, which outputs a single adjusted value. This value may then be further analyzed against time to predict the occurrence of the next large geomagnetic storm event.
Link to Final Project
Link to Project "Demo"
Detailed Project Description
The main difficulty in this project was understanding the data and what different values meant. As opposed to machine learning, the bulk of the work in this project is data processing. First, we must programmatically fetch data from the different sites, first the times and readings of DSCOVR and wind, which are used to perform dynamic time warping between the two. Then, we must collect and process the DSCOVR spectra data, and collect them into one value, which is our input. Then we collect velocity, density, and w-values from Wind, and use dynamic time warping to map the DSCOVR data to these values. Finally, we train a neural network, using the DSCOVR data as the inputs and the Wind velocity, density, and w-values as the output.
Space Agency Data
NASA
Hackathon Journey
Our first step was to find data from the DSCOVER and WIND sattellites. In this process we learned how to properly process .cdf files, which surprisingly took up a large amount of time due to the lack of integrable tools to properly unpack the data. Once our data was parsed, we had to account for Dynamic Time Warping. Learning and deriving a model for DTW was unlike anything we had done before, so we learned a significant amount here as well. This entire process (mainly data processing) took much longer than we expected, so we now know to do a better job of time management.
References
- González-Esparza, J. A., & Cuevas-Cardona, M. C. (2018). Observations of low-latitude red aurora in Mexico during the 1859 Carrington geomagnetic storm. Space Weather, 16, 593– 600. https://doi.org/10.1029/2017SW001789
- “The Great Aurora of 1859. - (to the Editor.) - the Daily News (Perth, WA : 1882 - 1955) - 8 Oct 1909.” Trove, https://trove.nla.gov.au/newspaper/article/77351480.
- Green, James L., et al. “Eyewitness Reports of the Great Auroral Storm of 1859.” Advances in Space Research, Pergamon, 6 Mar. 2006, https://www.sciencedirect.com/science/article/abs/pii/S0273117706000160.
- Solar Storm Risk to the North American Electric Grid - Lloyd's. https://assets.lloyds.com/assets/pdf-solar-storm-risk-to-the-north-american-electric-grid/1/pdf-Solar-Storm-Risk-to-the-North-American-Electric-Grid.pdf.
- Zell, Holly. “X1.9 Solar Flare Produces Wide Spread Aurora.” NASA, NASA, 9 Apr. 2015, https://www.nasa.gov/content/goddard/x19-solar-flare-produces-wide-spread-aurora.
- “Index of /Pub/Data/Wind/Swe/swe_h1.” NASA, NASA, https://cdaweb.gsfc.nasa.gov/pub/data/wind/swe/swe_h1/.
- “Index of /Pub/Data/Wind/MFI/mfi_h2.” NASA, NASA, https://cdaweb.gsfc.nasa.gov/pub/data/wind/mfi/mfi_h2/.
- “Index of /Pub/Data/DSCOVR/H0/Mag.” NASA, NASA, https://cdaweb.gsfc.nasa.gov/pub/data/dscovr/h0/mag/.
- Information, National Centers for Environmental. “Deep Space Climate Observatory (DSCOVR) Data: NCEI.” Deep Space Climate Observatory (DSCOVR) Data | NCEI, U.S. Department of Commerce, 13 Mar. 2014, https://www.ngdc.noaa.gov/dscovr/next/.
Tags
#DSCOVER #WIND #solar #ai