86 
RESULTS 
Full Study Area Evaluation 
The entire study area, minus areas covered by cloud and cloud-shadow, 
was used in the initial IDIMS validation study. Evaluation of before 
and after acreages indicated that removal of clouds and shadows did not 
significantly alter acreage proportions per cover type. An overall 
comparison made between the full resolution Landsat and the MAGI land 
use statistics determined using three different computation strategies: 
P, P', and (P* + S'), produced the results shown in Table I. 
The Landsat forest category showed excellent correspondence with all 
three sets of MAGI acreages, achieving an error rate under 5 percent 
even with the simplest approach, P. A similar result was produced for 
MDR, but the P option produced an error rate just in excess of 5 
percent. 
The adjusted primary land use, P', actually further skewed results pro 
ducing a worse correspondence of the Landsat and MAGI data for most 
cover types. Exceptions were crop/pasture and water, which were 
slightly improved. 
More realistic acreage estimates from the MAGI data base were derived 
by combining the secondary land use data layer with the primary land 
use (P T + S'). Comparison of the P and (P' + S') MAGI figures for each 
cover type revealed that the percent of P reassigned in the adjustment 
to (P* + S') in each case were: +6.3 percent for forest, -9.9 percent 
for crop/pasture -59.9 percent for water, +1.6 percent for transi 
tional, -11.8 percent for CII, +2.6 percent for MDR, and +12.9 percent 
for LDR. This adjustment to P produced excellent correspondence be 
tween the Landsat and MAGI acreages, with error rates under 5 percent, 
for three additional categories: crop/pasture, water, and MDR. 
In all, excellent correspondence of the MAGI and Landsat acreage esti 
mates was achieved with either the P option (forest) or (P' + S') 
option (C/P, water, and MDR). None of the land use computations from 
the MAGI data provided significantly different results for the transi 
tional category, which was considerably underestimated. Note that the 
transitional category is only found on the MAGI primary land use layer, 
and therefore cannot be adjusted to account for percentages less than 
60 percent. The (P' + S') option significantly improved the underesti 
mation of CII and considerable "overestimation" of LDR by Landsat, 
although error rates were still unacceptably high. 
Restricted Study Area Evaluation 
In order to determine if the high error rates for the transitional, LDR 
and CII categories were due to the MAGI non-specific land use desig 
nations for the several large institutional sites within the study 
area, a second comparison (Table II) was made on the remaining 85 per 
cent of the study area after excluding these inappropriate CII blocks, 
as discussed in the Approach. The statistics given here were computed 
using both the unadjusted primary (P) and adjusted primary and secon 
dary (P' + S') data layers. P' was not computed since it was not 
found to be useful in the whole area evaluation. 
Error rates for correspondence between Landsat and MAGI acreage esti 
mates for most cover types expressed as P were improved over the whole 
area P estimates. The only exceptions were MDR, for which the error