Statistical characterization of SAPS velocities

B. S. R. Kunduri, J. B. H. Baker, J. M. Ruohoniemi, S. G. Shepherd, P. J. Erickon, A. J. Coster and E. G. Thomas

Introduction - SD SAPS Location model

Kunduri et al [2017], under review.

Kunduri et al [2017], under review.

In Kunduri et al., [2017] (under review), we
- developed SuperDARN based SAPS location model
- discussed SAPS speeds using gridded line-of-sight velocities assuming SAPS were perfectly westwards.
A detailed characterization of SAPS velocities is required to complete the SAPS model. In this study
- we analyze variations in SAPS flow direction with MLT.
- we analyze mean SAPS velocities at different Dst bins and MLTs.
- we derive kernel density estimates of SAPS velocities.

Clausen et al [2012] approach was to derive one L-shell fit for each radar pair. The approach was ideally suitable for case studies.
Not directly adaptable to statistical studies.
Three L-shell fit velocities (3 radar pairs) for data spanning more than 7 hours in MLT would be under utilization of resources.
The first step is make L-shell fitting applicable to statistical studies.

Instead of one velocity for each radar pair. We search the entire "map" to detect locations where L-shell fitting can be applied (irrespective of radars).
Regions where common volume measurements are available (highlighted above) are especially suited for applying L-shell fitting technique.

Apply a standard grid over SAPS observations.
Each cell in the grid is 1 hour MLT, 0.5 degrees MLAT (SAID features have narrow latitudinal range).
Cells with "good" scatter will:
- have measurements with l-o-s azimuth range > 35.
- Atleast have 3 unique azimuth measurements.
- When fitting a sinusoid, determined SAPS azimuth should be within -90 +/- 20 degrees.
- Fitting error should be less than 25%.
NOTE : L-o-s velocities below 150 m/s are discarded.

Solid lines indicate velocities where "good" L-shell fit could be derived,
Dashed lines indicate velocities whose directions were assumed to be same as the nearest "good" fit.
Results are comparable to Clausen et al, [2012].
The L-shell map approach is suitable for statistical studies. - 1) results are standardized for all events. 2) not impacted by data availability at radar pairs.

Mean and std. dev. of SAPS velocity directions at different Dst bins. -90 is perfectly westwards. The solid lines represent a linear fit to mean azimuths.
Flow direction becomes increasingly polewards towards dusk. Indicative of SAPS merging with auroral flows eventually? The pattern is also observed (not discussed) in Clausen et al [2012] event.

The Figure presents histograms, kernel density estimates (red) and fitted skewed gaussian curves (dashed blue) for SAPS velocities for different Dst bins at a selected location.
Skewed gaussian appears to be a better fit for the KDEs (after trying several others).
The likelihood of observing higher velocities increases with geomagnetic activity.
With decreasing Dst gaussian curve changes from right skewed to left skewed.

Figure shows KDEs of velocities for lowest dst bin at different MLTs.
Likelihood of observing higher velocities increases near dusk.
Each KDE can be modeled separately, but a universal model (function of Dst, MLAT, MLT) like SAPS location model is difficult.

Used L-shell fitting to estimate SAPS velocities
The mean SAPS velocities increased as we move towards dusk and at higher disturbance levels.
Flow direction was more polewards near dusk than near mid-night.
Kernel density estimates of velocities at a given location and dst-bin suggest a skewed gaussian distribution of velocities.
Developing a dst and geomagnetic location based common model is proving difficult. Tried different fits/distributions - skewed gaussian, rayleigh etc.