Get reliable science

A 39,000-ha semi- arid range l and test
site in Starr and Zapata Counties, Texas,
was used to compare rangeland inventory
and other 1anduse categories estimated
by digital pattern recognition methods
(maximum likelihood ratio classification)
with percentages estimated from a ground correlated print enlarged to 1:100,000
scale from a LANDSAT color composite
transparency. Five land-use categories
were identified (grassland, mixed brush
rangeland, saline rangeland, cropland,
and water ).We found a highly signifi-
cant correlation(r= 0.997**) between
the photo-and computer-estimated hectarage for the June LANDSAT-2 overpass.
The correlation was not significant for
the August overpass largely because a
large percentage of the most extensive
rangeland category(mixed brush range-
land) was misc1assified as cropland.
The misc1assification was probably
related to some of the spectra in the
rangeland area resembling that of crop
residue, and volunteer plants and weeds
on the idle cropland.Computer-estimated
hectarages for grassland, saline range-land, and water were similar to the
photo- estimated hectarages for both the
June and August overpasses, indicating
the feasibility of estimating these landuse categories for either date.

source:abstract