4. Quality control 2017


4.1. On the national level
4.2. At delivery to PECBMS
4.3. For supranational indices and trends
4.4. For multispecies indicators

The overall aim of PECBMS is to use national data and to produce supranational indices and trends and multispecies indicators as precise and accurate as possible. Therefore, quality control is an extremely important part of the data flow process. Data quality control has been implemented at each level of data collation and analysis, and although many data quality control measures have been described in each computation step, we here present an overview.

Data quality of national schemes as well as PECBMS supranational outputs has been independently checked and approved by publications in peer-reviewed scientific journals.

Production of European population trends and indices and indicators: an overview (click on for the larger picture)

4.1. On the national level

Data quality control at the national level is implemented by each monitoring scheme. PECBMS facilitates contacts between the schemes to share experience and provides suggestions on good practice, as for instance by publishing the Best Practice Guide (Voříšek et al., 2008). Coordinators of national schemes ensure that their fieldworkers are sufficiently trained and skilled. Some schemes even test the knowledge and identification skills of fieldworkers, as for instance in Norway (http://www.birdid.no). Many schemes provide fieldworkers with identification tools (e.g. records of bird songs and calls), with maps and field recording sheets incl. detailed instructions.

The data collected by fieldworkers are examined once received, suspicious records are usually checked immediately and if not justified, they are abolished. Automatic tools (e.g. automatic warning signals issued when suspicious data are entered into a database) are of increasing importance especially in on-line data collation, e.g. to prevent typing errors. National schemes also check the data quality at the level of data analysis, for instance by setting filters for minimum amount of record/sites counted etc.

Most importantly, countries have implemented standardized procedures for field work and produce national indices and trends in the same way.

4.2. At delivery to PECBMS

When delivered to the PECBMS coordination unit, we check all national indices – species by species – for their interannual consistency (that is: we compare them with previous indices) and we examine all suspicious and inexplicable inconsistencies in detail. A special tool to automatically check for interannual consistency of each species/country combination has been developed and we use it on a routine basis.

Also national data on species trends are checked for inconsistencies and computation errors, using quantitative criteria. Suspicious results are again examined in detail and excluded if they cannot be justified as reflecting real population changes; otherwise, they are validated. If necessary, we consult coordinators of national monitoring schemes.

National indices, once checked for suspicious data, are scrutinized further. Supranational species indices are computed only for those species for which data are available from countries hosting at least 50% of its PECBMS European population.

For more details see chapter 1, Box Field methods.

After supranational indices and trends have been produced, a control for their consistency is carried out just as was done for national data. In case we find some inconsistencies, we examine them in detail to find out whether this occurs because the data set is enlarged or improved, or by a computation error.

4.4. For multispecies indicators

Each supranational species index is examined to find out whether the species concerned can be included in the multispecies indicators. Species with either a very large or a very low population index and less precise trend values are examined in detail and potentially discarded (for details see chapter 3).

We compare the indicators computed with their previous versions, just like we do for species trends and indices. Again, when we find any inconsistency, we try to unravel whether this is caused by improvements in the data (e.g. improved national data sets, longer time series, new countries contributing their data) or by a computation error.