Tomographic Time Series

At UCSD to retain more slowly changing portions of the heliospheric signal, we first remove a two-dimensional modeled zodiacal-light map [Buffington et al., 2006] in heliographic coordinates. We then remove a sidereal brightness map from the residue in sidereal coordinates. We further remove a median filtered minimum map currently from 600 consecutive SMEI orbits in polar heliocentric coordinates. A selection of lines of sight from the sidereal sky map is chosen to be both separated from one another by approximately 3° and located to avoid bright stars in order that difficulties in their removal not influence our photometric result. These ~4000 sky locations are averaged to one-square-degree sections of the sky, and culled down to ~1000 to emphasize the population of lines of sight in the solar direction. This remaining collection of line-of-sight time series is then ready to proceed to further removal of unwanted signals.

First, each line-of-sight time series is cleaned of unwanted residues (left-over high-energy particle contamination, auroral light or veiling glare from the moon) by eliminating signals too large to be Thomson-scattering brightness changes. Next, an average Gaussian mean brightness with an e-1 weighting 200 orbits (~14 days) from the midpoint time is subtracted and the residual is presented as a two-dimensional orbit- versus-time-in-orbit map (see Figure). This map is normalized so that each point has approximately the same weight by normalizing brightness to solar elongation (ε) by multiplying each point by ε2.5/902.5. This is an empirical approximation to a correction of the brightness fall-off with elongation. The map is then edge-filtered using a several-step iterative process that eliminates the largest discrepant excursions from the mean value in time and at adjacent orbital locations and then recalculates the mean and again eliminates outliers using a more stringent criterion. At each step a new-long term average Gaussian mean is calculated as well as a rapidly time-varying filter that follows the rapid excursions from the mean. Excursions greater than approximately two standard deviations from this rapidly-varying filtered data set are removed from the time series. The filter parameters on each step have been set such that typically 5% of the remaining time series data points are eliminated on each iteration with fewer and fewer data points present in the time series that remains. This current UCSD-designed analysis provides a final stable base and variable time series over time scales of many weeks which are adequately free of most unwanted noise for the 3D reconstructions presented in the tomographic analysis. At the end of this process the base is subtracted from the points that remain to provide the time series used in the tomographic analysis. For more information about this procedure see Jackson et al., 2008.