445 
of 
to classify on the ground. In several notable instances 
photo interpretation proved more reliable than the initial 
ground reconnaissance because the photos enabled a pattern 
of deaths to be detected which was not readily observable 
on the ground. 
Drey 
Qia 
Interpretation - Small-scale Photography 
sr of 
riously 
*e 
>f the 
not 
! area 
The type boundaries interpreted on the 70 mm 
photographs were plotted onto the 1 : 40,000 photographs 
along the previously located flight lines. The 1: 40,000 
photographs were then interpreted by extending these 
boundaries using whatever distinguishing features that 
existed. Additional areas of.Type B remote from the 70 mm 
runs were also interpreted using comparison with the appear 
ance of "known" Type B. 
asso- 
1 for 
' prove 
te 
ded as 
ntify 
Although the general appearance of Type B was a 
lighter tone, this was not always the case, particularly 
along watercourses and flats which had a higher proportion 
of non-susceptible tea-tree (Leptospermum spp) in the under- 
storey. Dieback most frequently occurs along creeks and 
where two. adjacent 70 mm runs located dieback in a creek, 
these boundaries were linked along the creek even though 
there was little visual evidence of dieback on the 1s 40,000 
0 
photos. The reliability of this practice decreases with 
distance from the 70 mm runs. 
3 ns of 
deaths 
S of 
RESULTS 
A total of 16 km of field transect mapping along 
nerely 
iduals, 
the flight lines was carried out on a portion of the area 
covered by the 70 mm photos and an approximately equal 
distance in the area covered by the 1s 40,000 photos only. 
For the purpose of comparison with ground obser- 
?ey 
vations, Type G adjacent to Type B was grouped with Type B 
(tabulated as Type BC). This was to avoid large errors 
Lnter- 
These 
the 
iy be 
it 
h that 
)Ut 
irrah 
which would occur purely as a result of plotting errors. 
These are significant when it is considered that Type G may 
be a band of perhaps only 20 m in width. Type G in isola 
tion was recorded separately. The results are shown in 
Table 1 . 
Reliability of the interpretation can be expressed 
in several ways. Although the overall accuracy of the 
detection of ’’advanced dieback" was +k% for the 1s 3>000 
photos and -15# for the 1s 40,000 photos, this gives no 
;how 
aths 
s in 
en 
indication of the effect of compensating errors. 
A more useful expression is map reliability, 
defined as the probability of a point selected on the map, 
e 
ult 
falling into the correct type in the field. This also 
requires qualification with respect to errors of commission 
and omission. Errors of commission with respect to dieback