50
of remote sensing data, by an increasing amount of GIS information and by the necessity to
administer and operationally integrate different types of data and information. Land cover
mapping applications provide a good field for demonstrating the administration, integration
and automation of large and diverse data sets. Land cover mapping procedures usually require 6£
extensive user interaction, particularly in the assignment of meaningful land cover labels to Si 1
classified remotely sensed data and in the assessment of classification accuracies. The cost of 1-
user interaction rapidly becomes prohibitive as the amount of remotely sensed data increases. ar
Spatial and aspatial information held in a GIS may be used to reduce the amount of user Si
interaction, to increase the data processing throughput, and perhaps to produce more reliable V' c
classification results. These points are illustrated in this paper by examination of European m
land cover and forest cover mapping procedures. The land cover map was produced using an ei
unsupervised classification approach and the forest cover map was produced using a re
supervised classification approach. A
v<
There is a rich history of using NOAA-AVHRR sensor data for land cover classification at
regional, continental and global scales. This is due to the moderate spatial resolution of the ^
AVHRR sensor (1.1km pixels) and its daily world-wide coverage (Hoffmann and Belward, re
1996). Almost without exception these efforts have involved conversion of red and near ^
infrared satellite measurements into vegetation indices such as the normalised difference
vegetation index (NDVI). The NDVI serves as an indicator of surface biophysical activity, S(
being related to photosynthetic activity (Sellers, 1985) and above ground green biomass
(Tucker et al., 1981, 1985) but does not provide land cover type directly. However, statistical a
analysis of time series of NDVI values may be performed to discriminate between land cover R
types which have different phenologies (Townshend et al., 1991). The incorporation of
satellite derived surface temperature (Ts) with NDVI data has been shown to be useful for
monitoring surface energy absorption and exchange processes (e.g. Goward and Hope, 1989;
Nemani et al., 1993) and has been suggested as being potentially useful for characterising land
cover (Achard and Blasco, 1990; Running et al., 1994). This has been demonstrated recently
for America (Nemani and Running, 1996), Europe (Roy et al., 1996) and Africa (Lambin and
Ehrlich, 1995; Ehrlich and Lambin, 1996). ^
i-
c
r
F
1