|
|
|
Here is a
collection of hourly mean timeseries
plots showing a range of data anomalies. These data were downloaded directly
from public web sites. Recent data are obviously provisional, even if not
labelled, and subject to further quality control. Some data were provided as final data and
were already used in reports. The worst examples were from the final reports
produced by large reputable environmental consultancies so there is no provisional excuse.
Eventually all data should be processed and cleaned up ready for reporting.
If the plots look good and convincing, then nobody will know about the
internal compromises you have taken to produce your published reports! Identifying faults is a team effort between the data managers,
ratifiers, site owners, calibrators, independent auditors and engineers.
Input and insight is welcome from any source. Ultimately, leaving
questionable data within final datasets, published reports and websites is
the responsibility of the data ratifiers. All data are anonymous. Plots can be removed on request if you recognise
your data. Open the Rogues
Gallery and see the images for yourself. |
|
Should poor quality data remain on public web sites forever? What does this say about the care and attention of the
organisation? Can any data be trusted? |
|
Not all data are as bad as this PM10 example published by a very
well known university. The customer could not see the flat lining and spikes
because data was so difficult to plot. The customer assumed / trusted that
everything was perfect. |
|
Some common processing inaccuracies and bad practice are shown in
the section on NO2. |
|
All is not lost. You just need an expert to diagnose
the problem and provide the final polishing. These data just need a little
tender loving care. The images in the Before
and After section show had bad data can be transformed into a
reportable dataset. |
|
Contact me Geoff.Broughton@aqdm.co.uk |
