Data Archiving and Networked Services (DANS) · 2000 · 41 citations · 3 references
Open access
Natural SciencesEarth SciencesGeoinformaticsCrop SensingLand Use ManagementRemote SensingLand Cover Change DetectionLand Cover ClassificationLand Cover MappingLand UseRemote Sensing Change DetectionClassification AccuracyClassification Accuracy AnalysisClassification Accuracy AssessmentLand Cover AnalysisLand Cover Classification SchemesLand Cover DataLand Use DataSatellite Image Classification
Because of the lack of suitable land cover/use data for environmental studies it is was decided in 1987 to produce a national land cover/use database of The Netherlands (‘LGN data base’) using satellite images (Landsat TM and SPOT) and ancillary data. Today, three versions of the database are available (LGN1, 2 and 3), based on satellite images from respectively 1986, 1992/’94 and 1995/’97. During the up-dating of the LGN database the classification method improved considerably, resulting in a sharp increase of classification accuracy and number of land cover/use classes. The classification procedure of the LGN3 database is based upon an integrated approach using multi-temporal satellite imagery, digital and analogue ancillary data, reference data and expert knowledge. A combination of automatic and interactive classification was used. The LGN3 database consists of 39 classes. For most non-agricultural LGN3 classes the minimum required classification result of 90% and 70% for respectively level I and level II classes is amply reached. In general, the classification results of most agricultural crops in the LGN3 database are high (> 70%), unless the total acreage of a crop in an ‘agricultural stratum’ is relatively low or the concerning crops show a large spectral variability.
3