Pekka Härmä

Production of CORINE2000 land cover data using calibrated LANDSAT 7 ETM satellite image mosaics and digital maps in Finland

CORINE Land Cover is a Europeanwide land cover and land use classification. The CORINE2000 classification of Finland is based on automated interpretation of satellite images and data integration with existing digital map data. Satellite images are used in estimation of continuous variables describing vegetation type and coverage, as well as in updating map data. Continuous land cover variables were transformed into discrete CORINE classes by thresholding these variables according to class descriptions in CORINE nomenclature. The output of Finnish CLC2000 project consists of calibrated IMAGE2000 mosaic, national CORINE classification (25 m raster) which is generalized and vectorized to European CORINE classification. According to the first validation utilization of non-standard, automated data production approach together with exploitation of digital map data was proved to be successful in Finnish conditions.

Change detection using spatial data problems and challenges

This article describes the SYKEs approach to change detection using high and medium resolution satellite images and present experiences concerning change detection. If it is possible, information about the land use is acquired using digital map data. The main problems with satellite images have been the availability of cloud-free Landsat images and atmospheric correction of images. The lack of historical ground truth information makes the classification of old images and the validation of result difficult. The size of reliably detectable change in the case of forest clear cuts seems to be 5 pixels. The most important characteristics of map data are metadata and accuracy assessment.

Topographic correction of landsat ETM-images in finnish lapland

Different topographic correction methods for Landsat ETM-image were compared. Corrected images were compared using land cover classification and estimation of forest inventory variables. Also the effect of the amount of vegetation to the correction coefficients were studied. Generally the best correction methods were Ekstrand and C-correction when their coefficients were determined by stratifying data according to the amount of vegetation.

Using phenological information derived from MODIS-data to aid nutrient modeling

The use of MODIS data for helping nutrient modeling has been tested in Finnish Loimijoki river basin during year 2006. Phenological time-series was constructed by computing weekly maximum NDVI-mosaics for study period. The important dates of growing season like beginning of greening, peak of growing season and senescence were extracted from time-series. The nutrient modeling has been made using the dynamic INCA- N semi-distributed model which integrates hydrology and N processes. The use of phenological time-series improved predictions of model compared to the case when we used average growing season as input. This improvement happened in spring in the beginning of growing season, which is in the northern countries crucial time for nutrient leaching due to high runoff after snowmelt.

Accuracy of CORINE land cover classification in Northern Finland

The new CORINE 2000 land cover classification of Finland is based on automated interpretation of Landsat images and data integration with existing digital map data. Landsat images are used in estimation of continuous variables describing vegetation type and coverage, as well as updating existing map data. The estimation errors of land cover variables are reasonably small in Northern Finland. The estimation results are most accurate for polygons with size of 10-100 ha. The interpretation of whole mosaic composed of 9 images is feasible. Usually, the bias of estimates were smaller for whole mosaic and RMSE and correlation little better for individual images and submosaics