The atmosphere of Jupiter's volcanic moon Io consists of mainly sulfur dioxide (SO2), and this main constituent has been studied with a variety of observing techniques across many wavelengths over the years. Here we study absorption by SO2 at the hydrogen Ly-alpha line (1216 angstrom) in a large set of images taken by the Space Telescope Imaging Spectrograph (STIS) onboard the Hubble Space Telescope (HST) between 1997 and 2018. An advanced statistical analysis using a Monte-Carlo trial method is applied to derive the SO2 column density from the Ly-alpha intensity, which includes the uncertainties of the used variables such as solar and background flux. Our analysis produces a probability distribution function of the SO2 column density and highlights some short-comings of the observing technique. Most importantly, the HST/STIS images of the surface-reflected Ly-alpha flux are only sensitive to SO2 column densities between similar to 10(15) cm(-2) and similar to 5x10(16) cm(-2) . Due to strong non-linearity in the relationship between the SO2 abundance and the Ly-alpha flux at the low and high values of detected flux, SO2 abundance directly retrieved from the STIS images will generally fall within these boundaries. This explains the relatively low equatorial column density of about 10(16) cm(-2) reported by previous studies using the Ly-alpha images (e.g., Feldman et al., 2000; Feaga et al., 2009) compared to other studies (e.g., Spencer et al., 2005; Tsang et al., 2013; Jessup and Spencer, 2015; Lellouch et al., 2015), where the obtained column density is often 10(17) cm(-2). By assuming a log-normal probability distribution function for the column density, a new estimate of the SO2 column density is then fitted, indirectly accounting for abundances beyond the detectability limits. This method suggests slightly higher equatorial SO2 abundances and much larger upper-limits, revealing that the Ly-alpha observations are in fact consistent with the higher abundances found in other studies. We then investigate the SO2 abundances at three volcanic sites (Loki, Marduk, Thor), where plumes were observed before and where the sensitivity in our images is comparably high. The observations did not reveal transient changes due to local outgassing at any of the three sites. Finally, the heliocentric distance of Io changed from 4.95 AU to 5.45 AU between the observation dates, potentially allowing us to investigate the influence of solar intensity changes on the SO2 column density via surface frost sublimation. However, the derived error bars are significantly larger than the derived variability, preventing any firm conclusion on seasonal changes and local volcanic outgassing.