Use of Imaging Spectrometer Data to Determine Seasonal Patterns and Changes in Wooded Parkland Landscape in SW Finland

摘要

We have tested the feasibility of the AISA imaging spectrometer data in the monitoring of a wooded parkland landscape in SW Finland, based on two flight campaigns: one in May (deciduous trees are leafless) and the other in August (mid of the growing season). The data were acquired in 40 and 53 spectral bands, respectively, using 2 × 2 m ground resolution along 4.3 km transects representing typical ground cover of this region. Several test areas were defined for each main ground cover type, and these pixel populations were subjected to a number of spatial, spectral and inter-seasonal analyses. The results show that correlations among the different spectral bands were different in the data sets. Correlations in May were high in general, while in August only the visible bands were highly correlated, with the exception of the spruce forest stands in which all the other bands but blue bands were highly correlated. In all vegetation classes, distinctive small-scale spatial variation was observed, and these patterns were different in May and August data. The use of principal components among correlating spectral bands allowed production of color composites that were highly informative and could be adjusted to the special needs of the field scientists. The differences that were observed in data acquired in different seasons within the same study sites, both in their average spectral properties and in the spatial distribution of pixel values, call for carefulness in any application of AISA data in land cover applications. We conclude that the added value of the AISA data as compared to other remote sensing tools, was in their suitability to distinguish minor variations in the land cover, and their likeliness to suit well for seasonal and inter-annual monitoring of vegetation change.