First page Back Continue Last page Overview Graphics

Anomaly Analysis

Notes:

Anomaly analysis is very useful for identifying anomalous data, capture problems and processing errors. It's less useful for identifying anomalous trends as such trends are indistinguishable from anomalous data.

Consider the above plot. This shows the raw monthly ISCCP satellite temperature data in red, monthly anomaly in dashed black and the 12 month running average as a thick red line. The largest feature in the anomaly plot (late 2001) is a data related issues. The Oct 2001 anomaly is caused by the confluence of a number of changes, including the transition from the NOAA-14 satellite to the NOAA-16 satellite.

You should notice how the 12 month running average and the anomaly are both increased after the data anomaly. A 5 year running average will smooth out the transition even more and make it seem to dissappear.