Methodology:

Following the data binning into an equal-area grid and 5-min medians as described in the abstract, the data gaps are infilled as follows. We initially infill the data gaps in the sparse binned data with zeros, and then we apply the method of data-interpolating Empirical Orthogonal Functions (EOFs). This allows global (i.e. the full extent of the binned data set) spatial and temporal basis vector patterns to be obtained. These basis vectors collectively describe the full variability of the dataset. The form (i.e. morphology/shape) of the basis vectors is controlled by the cross-correlations within the dataset. Since the ionospheric plasma velocity is strongly correlated in space and time, the spatial and temporal behaviour of the basis vectors with the largest eigenvalues (i.e. those which describe the majority of the variability in the dataset) are defined by the underlying physics of the ionospheric plasma. In contrast, since the missing data are relatively uncorrelated in space and time, the missing data contributes to lower-eigenvalue basis vectors. This provides us with a method to infill the missing values with the largest-eigenvalue basis vector, which is a better guess for the underlying plasma velocity field than the initial infill of zeros. Moreover, we have done this without any a priori specification of source geometry. The EOF-solution-and-infill process is repeated iteratively, until the amplitude of the infill converges with that of the data measurements, where both overlap. This infill only converges when it reinforces patterns present in the original data, thus providing a self-consistent description of the plasma velocity at the original temporal resolution of the SuperDARN data set. This gives complete spatial and temporal coverage without resorting to climatological averages, spatially smoothed models, or a priori relationships determined from solar wind drivers. Following this retrieval of the un-measured variability of the data, we fit a sinusoid model to translate the basis vectors from their line-of-sight (i.e. radar look direction) basis to a basis of cardinal north and east plasma velocity vector components. This method is described in full in a paper ('Data-Driven Basis Functions for SuperDARN Ionospheric Plasma Flow Characterisation and Prediction'), presently under review in JGR Space Physics (2021).

Data collection:

The data were gathered using the northern hemisphere radars of the SuperDARN global array, and the fitted Doppler velocities were processed from the original autocorrelation functions using version 4.0 of the radar software toolkit (RSTv4.0) and within that toolkit, fitting routine 'fitacfv2.5'.

Data quality:

The SuperDARN data were processed to remove ground scatter, and to eliminate measurements with too low power (lower than 3dB), or which had a poor-quality flag (identified in RSTv4.0). When binning the data, range gates below 11 and above 150 (where those values correspond to multiple of 45 km range distance from the radar array location) were not used, since these gave inaccurate locational estimates.