Time series of SOHO-SEM data after application of the AR outlier removal routine (blue [25 nm], green [30 nm]). We show the original data (black) without any processing done and the input data to the AR routine (red). Red dots mark points that are detected as outliers. This method picks obvious errors (e.g. around February 2002 there are some negative values), linearly interpolated data (mid-end of 1999, January 1999), and solar flares. However, it also detects some points that are not obviously (and necessarily) outliers. Their uncertainty is increased by the difference to the interpolated data. The inlet displays the whole time series, January 1996 to June 2014.