Jr. Dymond et al., LARGE-AREA VEGETATION MAPPING IN THE GISBORNE DISTRICT, NEW-ZEALAND, FROM LANDSAT TM, International journal of remote sensing, 17(2), 1996, pp. 263-275
Debate over a forestry incentive scheme in the Gisborne district, New
Zealand, highlighted the need for up to date information on the vegeta
tion cover. Maps of vegetation at a scale of 1:100 000 were produced b
y automatically classifying Landsat Thematic Mapper (TM) imagery. The
classified imagery was compared with existing vegetation information (
20-years-old) from a GIS database to identify gross errors. Through fi
eld checking the discrepancies were identified as either real changes
or errors in classification. Correction of errors increased the overal
l classification accuracy from 84 to 90 per cent. The digital vegetati
on map was intersected with land use suitability data to provide a two
-way table that provided land managers with quantitative information s
uitable for making regional planning decisions. Although the 90 per ce
nt accuracy is high enough to permit the calculation of vegetation are
as and to achieve an adequate representation of regional vegetation pa
tterns, it is not high enough to permit the digital vegetation map to
be used as a vegetation database where point queries are important.