LUCAS (European Land Use/Cover Area Frame Statistical Survey) is a project managed by the Eurostat, which main purpose is to provide harmonised information on agri-environment for Europe. LUCAS records land use (LU) and land cover (LC) information and stores digital field photographs in a two-level regular grid.
The purpose of this report is to assess the thematic accuracy of CLC2000 by means of an independent data source, the LUCAS data (not used in the compilation of the target database — CLC2000) and statistical methods homogeneously applied to as many of the participating countries as possible. The purpose of validating geographical data is to derive final accuracy/reliability figures by means of independent, high resolution and more accurate data, which is related to a similar data acquisition period.
LUCAS data is the only information that is available in many countries (18) for a Europeanwide validation of CLC2000. LUCAS fulfils the criteria of validation data as a result of its following characteristics:
- high geometric accuracy;
- data acquisition nearly coincident with CLC2000;
- independent, as not used during production of CLC2000.
The specific questions that the validation procedure is supposed to answer are:
- Was the 85 % thematic accuracy target of CLC2000 fulfilled?
- Which classes were determined with high accuracy and which ones with low?
- What were the reasons for misinterpretations?
- How subjective was photo-interpretation?
Two kinds of methods were used for validating CLC2000 data:
- Reinterpretation of IMAGE2000 data based on LUCAS codes and photographs;
- Automatic comparison of CLC2000 codes and LUCAS LU and LC codes.
The main result of the reinterpretation approach is that the total reliability of CLC2000 is 87.0 ± 0.8 %, which leads us to the conclusion that the 85 % accuracy requirement specified in the Technical Guidelines of CLC2000 was fulfilled.
The result of the automatic comparison was that the percentage of total agreement (PTA) found between CLC2000 and LUCAS LU and LC is 74.8 ± 0.6 %, meaning that CLC2000 approximates LUCAS thematic data with a 74.8 % average accuracy.
The analysis of results at class level shows that only 22 of the 44 classes of CLC2000 could be validated. The 22 classes that could not be validated belonged mostly to artificial surfaces, wetlands and water.