For each flagged anomaly, the automatic checking system assigns a severity level depending on how far statistics are from the expected values. When an anomaly persists over many data assimilation cycles the severity level is adjusted to highlight persistence.