The objective of this study was to map the grass cover in Dutch heathland nature reserves using Landsat-TM satellite imagery. The grass cover data will be used to initialize the HEATHSOL-model. This model is used for biomass predictions of heather and grass species in natural heathlands in the Netherlands, as influenced by nitrogen deposition from the atmosphere.
Two images of the Landsat TM satellite sensor were used, both of 7 September 1993. To calibrate the images, field data were collected in the 'Dwingelose heide' and 'Strabrechtse heide' nature reserves by measuring the vegetation cover. The MICROFIND software, a module within the HEIMON heathland monitoring system operated by K&M, was used for the calibration and subsequent prediction. MICROFIND predicts cover values of heather, grass and bare soil for areas of 30x30 m which correspond to the Landsat TM pixel size. We used digitized 1:50,000 scale topographical map to isolate the heathland nature reserves in the images.
Remaining irrelevant cover types like forests, meres and roads were then removed. Very small reserves and heathlands in dune areas were excluded. A few reserves in the eastern part of the country were not covered by the images. After masking some 24,000 ha of heathland remained for the analysis. It was found that 45% of the study area was covered by grass, with a tendency of increasing grass cover from the north of the country to the south. Similar inventories of grass cover have been made earlier in 1986 and 1989.
This time the procedure was hampered by the presence of haze over southern scene. As a result predictions may be less accurate for part of the southern scene. The main accuracy of a grass cover prediction for an individual pixel was comparable to the accuracy found in the earlier studies: +- 1% when the models were applied to the reserves in which field data had been collected, up to +- 30% when they were extrapolated to other areas.