QUANTIFYING THE DISTRIBUTION OF FOREST FUNCTIONAL TYPES AND FOREST LEAF AREA INDEX IN THE ALPS
- Land cover change is an important element of global environmental change processes. Most ecosystem processes strongly depend on land cover and its attributes. Mapping land cover, especially in mountain terrain is a difficult and challenging task. Remote sensing is an attractive source of thematic maps, such as those depicting land cover. Thematic mapping from remote sensing data is typically based on image classification. The image classification procedure synthesizes satellite data with field data and other ancillary data derived from a Geographic Information System (GIS - ArcInfo) coverage. The present study combines GIS and remote sensing data to produce a land cover map for the National Park Berchtesgaden and to build an extrapolation for other test areas in the Alps (Stubai and Ötz Valleys). Although a vast GIS data set had been assembled for the National Park, remote sensing was not previously used as a tool for land cover mapping and forest ecosystem analysis. For supervised classification, the maximum likelihood algorithm was used to sort and group data into discrete classes, which can be uniquely identified. Comparison and accuracy assessment with „ground truth“ data was carried out. An overall accuracy of 86% and 87% of the classification results in the National Park Berchtesgaden and in Stubai Valley, respectively, was achieved. Another important parameter determining gas exchange (water loss and carbon gain) of alpine forests is Leaf Area Index (LAI). Remote sensing provides a means to estimate LAI over large areas. To map LAI in mountain regions, Landsat TM NDVI index and SR index were examined together with forest inventory data of the Berchtesgaden National Park. “Ground truth” point grid maps for LAI were obtained through the use of allometric relationships (relating tree size and leaf area) as derived from tree harvests and together with the forest inventory database. On the basis of the forest mask derived from land classification and the Landsat vegetation indices, homogeneous forest polygons were identified. They were used for polygon by polygon correlation between LAI and vegetation indices. Mean forest polygon values were used to determine the relationships. With the derived equations, LAI was mapped at 30m resolution (Landsat data). Using the digital elevation model, the distribution of the vegetation types and LAI along elevation gradients was investigated. The results in National Park Berchtesgaden were further used in an extrapolation to classify land cover in Stubai and Ötz Valleys. Except to detect the distribution of land cover classes, supervised classification was used as a part of the algorithm for predicting forest leaf area index at the investigated sites. The digital LAI map of Stubai Valley was compared with LAI map derived from allometric relationships in Neustift (part of Stubai Valley). A correlation between NDVI and LAI in Neustift was derived. The validation results derived for coniferous forest in Neustift (Stubai Valley) show good correspondence to the results derived in Berchtesgaden. For both investigated sites, leaf area index can successfully be described with simple and reasonable correlation with NDVI.