Averaged spectrum for the values in Fig. 5, blue plots, where the FWHM of the lines are greater than 3.5 km s⁻¹. The fitting routine did not account for two clumps along the line-of-sight of a pixel for this range in velocities, and so tried to fit the two components as a single Gaussian with a broad FWHM (in green, corresponding to a FWHM of 4.4 km s⁻¹). When we allowed for a second Gaussian term to be added to the model, the two clumps are fitted by Gaussians with narrow FWHM (1.62 km s⁻¹). The data points from the cube are plotted in blue, the initial guess for the fit in dashed black, and the fit result is in red. The x-axis is in units of the cube velocity channels, where channel 200 corresponds to 0 km s⁻¹, and channelsare 0.5 km s⁻¹ apart.