A methodology was developed and applied to estimating forest area and produ
cing forest maps. The method utilizes satellite data and ground reference d
ata. It takes into consideration the fact that a pixel rarely represents an
y single ground cover class. This is particularly true for low-spatial-reso
lution data. It also takes into consideration that the spectral classes ove
rlap. The image was first classified using an unsupervised clustering metho
d. A (multinormal) spectral density function was estimated for each class b
ased on the spectral vectors (reflectance values) of the cluster members. V
alues of the target variable - the proportion of forested area - were deter
mined for the spectral classes using sampling from CORINE (Coordination of
Information on the Environment) Land Cover database. Each pixel was assigne
d class membership probabilities, which were proportional to the value of t
he density function of the respective class evaluated at the spectral value
of the pixel. The estimate of forest area for the pixel was finally comput
ed by multiplying the class membership probabilities by the class forest ar
ea and summing over all the classes. The method was applied over a mosaic o
f 49 Advanced Very High Resolution Radiometer (AVHRR) images acquired from
the National Oceanic and Atmospheric Administration (NOAA)-14 satellite. Th
e estimated forest areas were compared with those extracted from the full-c
overage CORINE data and with official Forest statistics reported to the Eur
opean Commission's Statistical Office (EUROSTAT). The forest percentage (pr
oportion of forest area of the total land area) of 12 countries of the Euro
pean Union was underestimated by 1.8% compared to the CORINE data. It was u
nderestimated by 4.2% when compared with EUROSTAT's statistics and 6.0% whe
n compared to United Nations Economic Commission for Europe/Food and Agricu
ltural Organization (UN-ECE/FAO) statistics. The largest underestimation of
forest percentage within a country (compared to CORINE) was in France (5.9
%). The largest overestimation was found in Ireland. 15.6%. (C) 2001 Elsevi
er Science Inc. All rights reserved.