7. Concluding remarks

7.1 Data quality

Observations and measurements are the basis for all data analysis and estimations in forest assessments. It is therefore imperative to guarantee a high level of data quality. Data quality means that
  • each single measurement is made with care and accurate, and that
  • all measurements of a particular attribute follow the same specifications (in terms of definition and measurement procedure).
Measurement errors are usually not considered when accounting for errors in forest inventories. For the statistical analysis it is commonly assumed that all measurements had been taken without error. That means that the error figures that statistical estimations provide must generally be taken as nominal errors, being a lower bound of the actual real error. However, studies of error budgets for a large area forest inventory (Gertner et al. 1992) suggest that the standard error originating from the sampling design is, in fact, the error component with the greatest relative relevance.

What can be done in forest assessments to achieve good data quality?

  • Assessment and measurement protocol: complete and clear documentation and description.
  • Staff: careful selection and training of field crews.
  • Supervision: of field work, of measurement devices (calibration), of data delivered.
  • Check and calibration of measurement devices.
  • Plausibility checks when data are entered into data base.
For more information about data quality visit Standards, metadata and data quality and for information about data quality in data collection through interviews, visit How does NFA verify data quality?

7.2 Non-response

Non-response is a sampling ¿feature¿ that is a default point of discussion in the social sciences where data collection bases mainly on interviews. There, for a subset of sampled ¿elements¿ observations cannot be taken because the persons do not respond.

Similar situations occur also in forest inventories, when, for example, sample plot locations are not accessible or when clouds and shadow do not allow to make observations in some sections of satellite images. It is important to treat those cases correctly as non-response, possibly applying imputation techniques. It is not a good practice to shift those plots, for example, to more easily accessible areas or to simply forget them.

7.3 Practical issues

One should keep in mind that field work to gather data is often physically extremely demanding. Data quality depends heavily also on the motivation of the field crews, where frequently lesser educated, lesser paid and maybe also lesser motivated members collect the data which are the basis for all subsequent analysis.
last updated:  Wednesday, March 2, 2005