Forestry
Departmental Series No. 3
June, 1969
Identification of
on Near
Take
Forest Condition
Vertical Aerial
;n with a K-20
Classes
Photographs
Camera
A gricultural
AUBURN
E. V. SMITH, Director
Experiment Station
UNIVERSITY
Auburn. Alabama
/
Identification of Forest Condition Classes
on Near Vertical Aerial Photographs
Taken with a K-20 Camera
R. C. PARKER, Graduate Student of Forestry
E. W. JOHNSON, Professor of Forestry
THE CLASSIFICATION OF FOREST COVER into
condition classes is a basic operation in forest land man-
agement. Because of their high cost and relative inaccu-
racy, insofar as planimetry is concerned, conventional
stand mapping procedures involving on-site observations
are rapidly being replaced by methods employing aerial
photographs. In most cases, the land managers work with
existing aerial photographs such as those that are avail-
able through the U.S. Department of Agriculture. Unfor-
tunately, the USDA photographs have a relatively small
scale (1:20,000) and are made with panchromatic film,
which often is not well adapted to the work of vegetation
classification. Furthermore, the photographs are often
several years old and usually do not show current forest
conditions. Ideally, it would be desirable to have new
photographs made with the most efficient film-filter-season-
scale combination for each job, but the cost of such new
photography on a contract basis is often prohibitive.
A possible solution to this problem would be for the
land management organization to obtain its own photo-
graphs using rented aircraft and its own camera equip-
ment. Conventional camera equipment, however, is ex-
pensive, heavy, and complex. It is difficult to mount in
an aircraft without modifications that would be undesir-
able to the aircraft owner. This has brought about in-
creasing interest in both small camera aerial photography
and in modification of existing equipment to permit the
taking of suitable photographs from light aircraft, which
would be available at convenient times for use over small
areas. Small camera photography could be used inde-
pendently or to supplement and up-date large format
photographs that are out of date. In light of such in-
terest, this study was designed to explore one path pos-
sibly leading to a rapid and economical method of ob-
taining aerial photographs useful to the land manager.
OBJECTIVES OF THE STUDY
The primary objective of the study was to determine
whether or not the K-20 aerial reconnaissance camera,
FmsT PRINTING 2M, JUNE 1969
which produces 4 x 5-inch negatives, can be used to ob-
tain near vertical photographs that would be useful for
stand classification. A second objective was to determine
which of several film-filter-scale combinations would be
best adapted to the work of stand classification when the
photographs were taken with the K-20 camera.
MATERIALS USED
Camera. A K-20, Day Reconnaissance, aerial camera was
used to take the photographs. This is a hand-held, man-
ually operated camera. It has a 6%-inch focal length lens
with apertures ranging from f/4.5 to f/22 and a between-
the-lens shutter with speeds of 1/125, 1/250, and 1/500
second. The camera is capable of handling rolls of 51
inch film up to 20 feet in length. Virtually every type of
aerial film, both black-and-white and color, is available
for use in this camera. There is an electrically driven
equivalent, the K-25, which should provide results simi-
lar to those obtained with the K-20.
Camera mount. A special mount for the K-20 camera
was designed and built by the senior author for use on
high wing light monoplanes of the Cessna 150 type (8).
The mount, see cover photo, permitted positioning the
camera vertically on the side of the aircraft in such a
way that it could be operated manually. Because of
manual operation, an intervalometer could not be used
and the intervals between exposures had to be timed
with the sweep second hand of a watch.
Films, filters, and scales. Two black and white films, Ko-
dak Super XX Aerographic Panchromatic (No. 5425) and
Kodak Infrared Aerographic (No. 5424), were chosen for
the study. A Wratten No. 12 (Minus blue) filter was
used with both films. Photographs were taken at two
scales, 1:2,500 and 1:5,000, referenced to mean ground
elevation. Exposure settings were calculated using the
Kodak Aerial Exposure Computer. Using an Armed
Forces type B-5 developing tank, all film used in the pro-
ject was processed according to manufacturer's instruc-
tions (2). Because no suitable contact printer was avail-
able, contact prints were made on Kodak Polycontrast Pa-
per with the appropriate Polycontrast filters. The paper
was placed in direct contact with a negative and exposed
with light from an enlarger. The prints were finished so
as to produce a glossy, unferrotyped surface.
Interpretation equipment. All the interpretation work
was done under pocket stereoscopes of the Abrams CF-8
type. Abrams Height Finders were used to determine
tree heights and Moessner Crown Density Scales (5) were
used as aids in estimating stand density.
THE STUDY AREA
The study area is located approximately 4 miles north
of Loachapoka, in western Lee County, Alabama. The
area lies at a mean altitude of about 700 feet and is in
the transition zone between the Piedmont and Upper
Coastal Plain physiographic provinces (3). The ground
surface is gently rolling with a total relief of about 150
feet. The forest cover is mixed pine and hardwoods. The
pines are mostly loblolly (Pinus taeda L.) and shortleaf
(P. echinata Mill.). The hardwoods are those typical of
the region, such as southern red oak (Quercus falcata
Micxh.), water oak (Q. nigra L.), white oak (Q. alba L.),
hickories (Carya spp.), sweetgum (Liquidambar styraci-
flua L.), yellowpoplar (Liriodendron tulipifera L.), and
white ash (Fraxinus americana L.).
The study area was divided into two parts, one for
determination of relationships between ground truth and
photo-images, the other for the testing phase of the study.
The first of these areas was designated the "learning area"
and the second the "testing area."
THE FOREST COVER CLASSIFICATION SYSTEM
An unlimited number of forest cover classification sys-
tems could be devised. However, the most important
information to be provided is species composition, size of
individual trees, and density of the stands. The system
devised for this study recognizes all three of these ele-
ments. Table 1 summarizes the system and shows the
code designations used. Table 2 provides further infor-
mation on the tree size classification system.
PROCEDURE
Photography. All the photographs were made during the
summer of 1964. They were taken only on days classed
as photographic days, that is, days when the air was free
of haze and clouds. Air turbulence, which often was
severe on these days, made it difficult to fly straight and
level. Consequently, it is likely that most, if not all, of
the photographs contained considerable amounts of tilt.
Since the photographs were not to be used for metric
purposes and since no difficulty was encountered in view-
ing the photographs stereoscopically, the authors felt that
the tilt was not a problem.
A line between two clearly identifiable points was
drawn across the "learning area." This line was 2 2 miles
long and was deliberately located so as to cross as many
conditions as feasible. It served as the centerline of the
TABLE 1. CLASSIFICATION SCHEME 
USED FOR THE
INTERPRETATION OF CONDITION CLASSES
ON AERIAL PIHOTOGRAPHS
Forest cover type
Code Name
P Pine
1
BH Branchbottom
hardwoods
2
PBH Pine-branchbottom
hardwoods
3
UH Upland hardwoods
4
PUH Pine-upland
hardwoods
5
OFR Old field restocking
6
Size classes Density;
crown 
closure
Code Name Code Pct.
1 Sawtimber
2 Poles
3 Pulpwood
4 Saplings
5 Reproduction
A 81-100
B 61-80
C 41-60
D 6-40
E 0-5
1 S.A.F. (1962) Forest Cover Types 75 
(shortleaf pine), 80
(loblolly pine-shortleaf pine), and 81 (loblolly 
pine).
2 Types 87 (sweetgum-yellowpoplar) and 104 
(sweetbay-
swamp tupelo-red maple).
4 Types 40 (post oak-black 
oak), 52 (white oak-red 
oak-
hickory), and 72 (southern scrub oak).
5 Type 76 (shortleaf pine-oak)
6 Usually loblolly pine and/or sweetgum, but other 
species
were present in small numbers.
TABLE 2. CRITERIA FOR THE IDENTIFICATION OF
VEGETATIVE SIZE CLASS
Size class Range in D.B.H. Range in 
Crown
tree height sizes
In. Ft.
1-Sawtimber
2-Poles
3--Pulpwood
4-Saplings
14
5-Reproduction L
U-Mixture of all sizE
and larger 60-100 Large and
broadly
rounded
10-13 40-60 Intermediate
in size and
roundness
5-9 20-40 Small and
narrowly
rounded
1-4 10-20 Very small
and pointed
ess than 1 1-10 No definite
crown
e classes above
strips of photographs taken with each of the film-scale
combinations. A similar line was established in the "test
area" and, when photographs were taken, both lines
were flown to assure that photographic conditions for the
two areas would be comparable.
Exposure intervals were such as to obtain 60 per cent
endlap when the long dimension of the photograph was
oriented parallel to the flight line.
Ground truth. The boundaries of the stands in the actual
"learning" and "test" areas were located and delineated
on contact prints of standard USDA photographs, but no
attempt was made to assign the stands to condition classes
using these photographs. All these assignments were
made in the field by the senior author following on-site
visits. In the learning area these visits were made prior
to any photo-interpretation using the test photographs.
In the test area, however, the stands were visited and
classified only after the senior author had completed his
photo-interpretation of the area. Only in this way could
the probability of bias be eliminated from his part of the
work.
[31
Photo-interpretation - learning phase. Information gained
from intensive study of the photographs and from the on-
site visits was used as the basis for developing written
descriptions of the condition classes as they appeared on
the different film-scale combinations. Stereograms showing
each of the condition class-film-scale combinations also
were prepared. These descriptions and prepared stereo-
grams were used by the interpreters in their training and
testing operations.
Three interpreters participated in the study. Two were
the authors of this paper and the third was a senior stu-
dent in forest management who was highly interested in
the use of aerial photographs. All three interpreters had
perfect scores when given the Moessner Floating Circles
Test (6), indicating that all were able to see stereoscop-
ically without difficulty.
Prior to the testing phase of the work, each interpreter
worked with the photographs, the written descriptions,
and the stereograms of the learning area in order to
familiarize himself with the procedure. When an in-
terpreter felt competent to proceed, he began work on
the test.
Photo-interpretation - testing phase. Twenty sampling
points were randomly located on every second photograph
of the 1:2,500 infrared set. This resulted in 500 points
on 25 photographs. With a multiscope, these points were
transferred to all the other sets of photographs of the test
area. Each point was used as the center of a circular
sample plot. The plots on the 1:2,500 photographs had a
radius of 0.2 inch, while those on the 1:5,000 photo-
graphs had a radius of 0.1 inch.
Each of the 500 field plots was visited by the senior
author, who evaluated and recorded the ground truth
concerning the condition class.' Nineteen forest condition
classes were found to be present. Four sample plots were
randomly selected from each of these nineteen classes for
use in the photo-interpretation test.
Each of the interpreters attempted to classify the sample
plots on all sets of photographs. In order to minimize
the "memory carry-over" of information from one set of
photographs to another, each interpreter interpreted only
one set each day. The interpreters were not told how
many plots were in each condition class nor were they in-
formed of their scores until all the tests were completed.
Scoring. Scoring was carried out at two levels. At the first
level overall accuracy of each interpretation was graded,
while at the second each of the three elements of the
classification system was graded individually.
The overall grading system was based on the idea that
tree size and stand density values were meaningless if the
forest cover type was not correctly identified. Conse-
quently, the score for an interpretation that was com-
pletely correct was set at two. When the cover type was
incorrectly identified, the score was zero regardless of
how well the size and density components were judged.
If the cover type was correctly identified but the size
component or the density component was in error by
only one class the score was 1 2. If both the size and
density components were in error the score was zero.
1 Since the senior author was to be used as a test photo-
interpreter he interpreted all 500 plots on each of the sets
of photographs prior to going into the field and obtaining the
ground truth.
Thus, for each interpretation a score of either 2, 11, or
0 could be assigned.
The individual component scoring system was set up
so that the effect of film, scale, and interpreters on each
of the classification components could be evaluated. Con-
sequently, each component was scored without regard to
the accuracy of the evaluation of the other two compo-
nents. A correct species identification was given a score
of one. An incorrect species identification was given a
score of zero. A correct tree size or stand density evalua-
tion was given a score of one. If in either of these cases
the correct value was missed by only one class the inter-
pretation was given a score of 1/. An estimate with an
error greater than one class received a score of zero.
Method of analysis. The primary experimental design of
the study was a factorial with mixed factors and with
equal subclasses. Films, scales, and condition classes were
considered fixed while the interpreters were assumed to
be randomly selected from a population of similar in-
terpreters.
DISCUSSION
The 19 condition classes shown in Table 3 were en-
countered in the test area. It can be seen that most were
in the larger size classes and heavier density classes. This
skewed distribution of the sample resulted in the test
being essentially incomplete. A broader, more repre-
sentative, sample would be necessary before a true picture
of the effectiveness of the procedure could be obtained.
TABLE 3. CONDITION CLASSES ENCOUNTERED IN THE TEST AREA
Species Group Density Size
1 2 3 4 5 U
Pine (P) A X X
B X X
C X
Branch hwds (BH) A X X X X
B X X X
Upland hwds (UH) A X
B X
PBH A X
PUH A X
B X X X
Table 4 lists the mean accuracy percentage of interpre-
tation by condition classes and film-scale combinations
across all three interpreters. These mean per cent ac-
curacy values refer to the arithmetic mean of the per cent
of the total possible score achieved by the interpreters.
In general, the results are poor. In only one case, for
P3A on panchromatic photographs at a scale of 1:5,000,
did the mean accuracy reach 90 per cent. On the other
hand, in several cases the mean accuracy was zero. These
results are sufficiently poor that one might feel justified
in dismissing the procedure out of hand as a practical tool.
However, a further examination of the factors involved
in the experiment should be made before making a final
judgment of the procedure.
Table 5 shows the overall analysis of variance for the
study. The results of this analysis indicate that the inter-
preters differed in their ability to identify the condition
classes and that some condition classes were relatively
[4]
more easy to identify than others. It does not appear to
show either that the film type or the photographic scale
TABLE 4. ACCURACY OF INTERPRETATION OF CONDITION
CLASSES BY FILM-SCALE COMBINATIONS
Condition
class
P2A ---
P2B____
P2 C - - - - - - - -
P3A------
P3B -- - - - - - -
BH1A---
BH1B___
BH2A--- -
BH2B ----
BH3A--
B H 4A --------- --
U H 2B --------- --
PB H2A----------
PUH2B --------
PUH4B ---------
M eans-----------
Pan mean -----
Infrared mean,
1:2,500 mean--.
1:5,000 mean -.
Panchromatic film Infrared film
1:2,500
Pct.
50
44
54
54
56
6
6
0
33
0
0
48
23
35
13
42
29
* 33
13
28.4
1:5,000 1:2,500 1: 5,000
-27.3
_40.8
_35.3
-32.8
Pet. Pet.
63 69
56 75
40 54
90 73
42 31
21 63
6 21
15 37
0 48
6 25
6 58
56 35
8 25
19 25
21 21
6 42
21 67
15 25
8 10
26.3 42.3
Pet.
58
73
48
81
60
13
4
25
35
15
23
48
31
15
21
52
44
71
29
39.3
TABLE 5. ANALYSIS OF VARIANCE OF THE PHOTO-
INTERPRETATION DATA
Source
Films (F)------
Scales ()---
Condition classes ((
Interpreters (I)----
F X IC ----------
S X IC -,--------
F X S XIX -
E rror-------------
T otal ------------
--- - --- --- - --- - -- -
d.f. S.S. M.S. F
1 18.961 18.961 9.37N.S.
1 0.145 0.145 0.07N.S.
_18 112.910 6.273 3.42***
2 15.474 
7.737 24.10**4
1 0.121 0.121 0.36N.S.
].8 18.690 1.038 1.25N .S .
2 3.899 1.949 6.07***
18 14.173 0.787 1.47N.S.
2 4.152 2.076 6.47*
36 66.026 1.834 5.71***
18 18.655 1.036 2.43*
2 0.673 0.336 1.05N.S.
36 29.872 0.830 2.59***'
_36 19.327 0.537 1.6704
36 15.411 0.427 1.33N.S.
-684 219.311 0.321
-911 557.790
*0 Significant at the 0.05 level of probability.
** Significant at the 0.01 level of probability.
kSignificant at the 0.005 level of probability.
N.S. Non-significant at the 0.05 level of probability.
had any effect on the rate of correct interpretations. How-
ever, the situation with respect to-films and scales is not
this simple because the interactions between films and
interpreters (F X I) and between scales and interpreters
(S X I) are significant, indicating that the interpreters
reacted differently to both films and scales. In addition,
the film type means shown in Table 4 were sufficiently
different that the lack of significance in the F-test was un-
expected. Furthermore, the C X I, F X C X I, and
S X C X I terms in the analysis of variance are signifi-
cant, indicating that the interpreters reacted differently to
the different condition classes. In general, some condition
classes were easy to identify while others were difficult.
Superimposed on this overall pattern were interpreter
patterns, which differed from one another but not suffi-
ciently to mask completely the overall pattern. These
results warrant a study of the individual interpreters to
see if some reasons for the results can be determined.
Table 6 displays the independent analysis of variance
for each interpreter. In all three cases there was a sig-
nificant difference between films. This appears more rea-
sonable than the result of the overall analysis when one
considers the magnitude of the difference between film
type means shown in Table 4, which indicate that gen-
erally better results were obtained using infrared film. The
probable reason for the ]ack of significance in the overall
analysis of variance is that the F X I term instead of the
error term was used to test the films. Apparently since the
F X I term was large (it was highly significant), the film
component of variance alone was too small to be detected.
When the interpreters were separated and error terms
used for testing, the film effects were easily detected.
TABLE 6. ANALYSES OF VARIANCE OF THE PHOTO-
INTERPRETATION DATA KEEPING THE
INTERPRETERS SEPARATE
Source d. 5.5. M.S. F
Interpreter A
Film (F)- 1 1.895 1.895 6.45
Scale (S)______ 1 2.056 2.056 
6.99
Cond. class (C) 18 56.602 3.145 10.70***
F X .________ 1 0.398 0.398 1.35N.S.
F X C_________ 18 13.105 0.728 2.48
S X C- 18 9.257 0.514 1.75
F X S X C--------- 18 14.727 0.818 2.78***
Error 228 67.000 0.294
Total--- 303 165.040
Interpreter B
Film (F)- 1 16.582 16.582 34.55*
Scale (S)______ 1 1.451 1.451 3.02N.S.
Cond. class (C) 18 36.166 2.009 4.19000
F X 5 ------------- 1 0.211 0.211 0.44N.S.
F X C 18 19.262 1.070 2.33**0
S X C 18 16.456 0.914 1.90*
F X S X C 18 10.071 0.559 1.16N.S.
Error 228 109.500 0.480
Total---------------- 303 209.700
Interpreter C
Film (F)------------ 1 4.382 4.382 23.31000
Scale (5) ___________ 1 0.790 0.790 4.20*
Cond. class (C) ______18 86.163 4.787 25.46***'
F X 5 _____________ 1 0.185 0.185 0.98N.S.
F X C_____________ 18 16.196 0.900 4.790***
S X C-------------- 18 7.788 0.433 
2.30***
F X S X C________ 18 9.268 0.515 
2.74***
Error ---------------- 228 42.813 0.188
Total --------------- 303 167.580
in the overall analysis of variance. A possible explanation
for these results is that Interpreter B was so familiar with
the photographs and the ground conditions after setting
up the descriptive material and stereograms that the scale
effect was negligible in his case. Interpreter A had a
great deal of experience with small scale photography
(1:15,840 and 1:20,000) and found difficulty in shifting
to the large scales. Interpreter C had no previous experi-
ence to confuse him and did better with the photographs
with larger images.
Table 7 shows the per cent accuracy of interpretation
of condition classes by film-scale-interpreter combinations.
The evidence in this table is very strong that some condi-
tion classes were relatively easy to identify while others
were extremely difficult. Furthermore, a definite inter-
relationship between condition classes and interpreters is
evident, as was indicated in the overall analysis of vari-
ance by the highly significant C X I term. Since each
condition class designation is composed of three elements
and an error in any one of these elements would lower the
interpretation score, the percentage scores shown in Table
7 cannot be evaluated in detail. To make such an evalua-
tion possible, each element of the condition class descrip-
tion was studied independently of the other two elements.
The results of these studies are shown in Tables 8, 9, and
10, which are concerned, respectively, with cover type,
size class, and density class.
with the photographic characteristics and with ground con-
ditions. In his case, the major question was why his
scores were not higher.
Interpreter A had widespread experience in photo-in-
terpretation but not the intimate knowledge of the ground
conditions or of the photography used in this study pos-
sessed by Interpreter B. As was mentioned in the section
on procedure, each interpreter was introduced to instruc-
tional material prior to the test and worked with this ma-
terial until he judged himself competent to go on with the
test. It is likely that Interpreter A, because of his previous
experience, felt himself ready to proceed before he was
completely prepared. Probably the greatest problem he
had was adjusting to the small ground area covered per
photograph. His previous experience was largely with
9 x 9-inch photographs taken at scales of 1:15,840 or
1:20,000. With the former of these scale-format combina-
tions each photograph covers approximately 5 square
miles, while in the second each photograph covers about 8
square miles. In contrast, one of the 4 X 5-inch photo-
graphs at a scale of 1:2,500 covers only 0.3 square mile
and, at a scale of 1:5,000, 1.2 square miles. When the
photographs are viewed stereoscopically only about 0.2
or 0.8 square mile is seen at one time and the interpreter
finds difficulty in tracing out drainage systems, particularly
in areas of low relief. Contributing to this problem is re-
duced stereo-exaggeration in the small format photographs.
TABLE 7. ACCURACY OF INTERPRETATION OF CONDITION CLASSES BY FILM-SCALE-INTERPRETER COMBINATION
Condition class
Scale Inter BH BH BH BH BH BH BH UH UH PBH PUH PUH PUH PUH
preter P2A P2B P2C P3A P3B 
1A 1B 2A 2B 3A SB 
4A 2B 3A 2A 2B 3A 
3B 4B
Per cent
Pan film
1:2,500 A 25 37 37 19 13 0 0 0 0 0 
0 69 19 0 0 0 56 87 13
B 63 37 63 63 75 19 19 0 100 0 0 75 0 
44 37 13 31 13 25
C 6356 63 81 81 0 0 0 0 0 0 0 50 63 0 
0 0 0 0
1:5,000 A 69 63 37 87 50 50 19 44 0 0 19 18 
0 0 0 0 37 25 0
B 37 50 31 87 0 18 0 0 0 19 0 87 0 
31 63 19 25 19 25
C 8156 50 94 75 0 0 0 0 0 0 0 25 25 0 0 
0 0 0
Infrared film
1:2,500 A 50 81 37 50 19 44 19 13 0 13 
63 31 19 0 25 63 63 13 13
B 87 56 56 87 19 69 13 75 75 19 63 
50 56 75 25 63 100 63 19
C 69 87 69 81 56 75 31 25 69 44 50 25 
0 0 13 0 37 0 0
1:5,000 A 75 69 37 75 87 0 0 0 0 0 19 44 37 0 0 75 
63 81 25
B 37 75 44 87 19 37 13 75 81 25 25 100 25 
44 63 69 69 56 63
C 63 75 63 81 75 0 0 0 25 19 25 0 31 0 0 
13 0 75 0
All of the interpreters did relatively well with the pure
pine stands, with the accuracy rate ranging from 45 to
100 per cent and averaging 79 per cent. This accuracy
rate is what one would expect from theoretical considera-
tions and past experience. The difference between the
scale means is small and is largely the result of a poor per-
formance of Interpreter A when using the panchromatic,
1:2,500 photographs.
The presence of enough hardwoods in a stand to affect
the cover type designation caused an enormous amount of
trouble. Furthermore, different interpreters did not re-
spond to these difficulties in the same way. Interpreter B
did much better than either of the others, while Interpreter
A did better than Interpreter C. This ranking was not
unexpected. Interpreter B was responsible for the devel-
opment of the condition class descriptions and the stereo-
grams. Consequently, he was intimately familiar both
The base-height ratio in the case of 9 X 9-inch photo-
graphs, taken with a 6-inch lens at 1:15,840 or 1:20,000,
is approximately 0.58, while with an 8.25-inch lens it is
approximately 0.43. With the 4 X 5-inch photographs
used in this study the base-height ratio was only approxi-
mately 0.31. This tended to flatten noticeably the stereo-
image and made the detection of drainages difficult. With-
out the broad view needed for judging the character of the
topography it is not surprising that there was considerable
confusion between upland and branchbottom hardwoods.
This, of course, would be expected to extend into the
mixed hardwood classes. Although examination of the
data in Table 9 cannot confirm this hypothesis, the original
data imply this type of confusion. It is probable that the
situation could be greatly improved if the small format,
large scale photographs were used in conjunction with
accurate topographic maps or with small scale, large for-
[6]
TABLE 8. ACCURACY OF INTERPRETATION OF FOREST COVER
BY FILM-SCALE-INTERPRETER COMBINATIONS
Scale 
Interpreter 
Cover types
P BH UH PBH 
PUH
Per cent
Pan film
1:2,500 A.
B
C
1:5,000 A
B
C
Infrared film
1:2,500 A
C
1:5,000 A
B
C
Pan mean .e...
Infrared mean.
1:2,500 mean--
1:5,000 mean--
Overall mean
45
75
90
75
55
100
70
85
95
90
70
95
73
84
.76
.81
79
18
39
0
46
18
0
39
75
75
14
64
18
20
48
41
27
34
13
25
75
0
25
50
25
75
0
37
50
25
31
35
35
31
33
0
50
0
0
75
0
50
25
25
0
75
0
21
29
25
25
25
50
31
0
19
25
0
56
69
13
88
75
31
21
55
36
40
38
TABLE 10. ACCURACY OF INTERPRETATION OF DENSITY
CLASSES BY FILM-SCALE-INTERPRETER COMBINATIONS
Scale Interpreter
Pan film
1:2,500 A-
B
C
1:5,000 A
B
C__
Infrared film
1:2,500 A
B
C
1:5,000 A
B
C
Pan mean
Infrared mean
1:2,500 mean
1:5,000 mean
Overall mean
Density Classes
B
Per cent
60
81
35
61
81
40
47
87
50
60
89
54
60
65
60
64
62
69
72
71
76
69
69
69
56
71
78
65
81
71
70
68
71
71
TABLE 9. ACCURACY OF INTERPRETATION 
OF SIZE CLASSES
BY FILM-SCALE-INTERPRETER COMBINATIONS
Size Classes
Scale Interpreter 
1 2 
3 4
Per Cent
Pan film
1:2,500 A
B
C
1:5,000 A
B
C
Infrared film
1:2,500 A_
B
1:5,000 A
C
Pan mean
Infrared mean
1:2,500 mean
1:5,000 mean_
Overall mean
44
50
37
50
50
56
50
50
87
75
50
6
48
53
53
48
50
52
73
67
55
77
59
58
78
61
52
70
58
64
63
65
62
63
66
79
77
68
87
73
60
79
64
59
75
89
75
71
71
75
73
81
75
69
88
94
69
94
50
69
81
87
57
79
73
73
79
76
mat photographs of any age. In any case, the photo-
interpreters must have a feel for the terrain. If maps or
small scale photographs are not available, this feeling for
the terrain will have to be developed by studying a num-
ber of adjacent photographs and extending the drainage
pattern from one to the next.
Interpreter C was faced with all of the problems men-
tioned above. In addition, since he had no past experi-
ence, he was dependent entirely on his training program.
Hindsight leads one to the conclusion that this training
regime was perhaps inadequate or that the reference ma-
terials might have been improved. No attempt was made
to develop an identification key which would have sys-
tematized the search for clues as to the identity of the
cover type. If such a key, stressing the importance of the
drainage pattern, had been used it is likely that all in-
terpreters would have been more accurate in identifying
cover type.
Generally, as can be seen in Table 9, size classes were
easier to identify than were the cover types. As one prob-
ably should expect on the basis of theory and previous ex-
perience, neither film type nor photographic scale ap-
peared to influence the rate of correct interpretations. Al-
though the rate of correct identifications of size classes was
considerably better than the rate for cover types, it re-
mained low. The apparent reason for this lay in the cri-
teria used for classification. These were tied to timber
utilization categories which may or may not be closely
associated with tree size, particularly as the trees are seen
on the photographs. Tree height is a valuable criterion
for classifying stands at all stages of development. How-
ever, it must be remembered that the difference in total
height between pulpwood and sawtimber stands may be
small. In order to distinguish between the two classes, a
second criterion such as the average crown diameter of
the dominant-codominant portion of the stands is needed.
In this study, no provisions were made for the fact that
heights of utilization classes overlap or that definite di-
mensions of crown diameters could be measured. It is
probable that size class identification would have been im-
proved if these had been taken into account.
Another complication of size class identification was that
many of the stands had uneven canopies. This was far
more evident on the photographs than on the ground. This
led the interpreters to class these stands as uneven-aged
when they had been classified on the ground as even-aged.
Comparison of rates between size classes in Table 9 illus-
trated this factor. Stands of small trees, which tend to be
quite uniform, were correctly classified more often than
were stands of large trees, which tend to be less uniform.
An identification key based on tree height and crown di-
ameter dimensions and illustrated with prepared stereo-
grams might have helped increase the rate of correct in-
terpretations.
[71
C
38
38
75
50
25
25
37
25
50
50
13
25
42
33
44
38
38
----- -- ---
As in the case of the size classes, the rate of correct
identifications of density classes, Table 10, was approxi-
mately two out of three. This was not unexpected since
stand density evaluation from aerial photographs is known
to be erratic, particularly when it is done in the manner
chosen for this study. Complicating the problem was the
fact that the so-called true crown closure was visually
estimated. This estimation from the ground may have
been poorer than the one from photographs. Two pro-
visions could have been made to improve the procedure.
The photographic crown closure should have been meas-
ured or estimated objectively by some such procedure as
the one developed by Losee (4), or one using a dot-grid.
With the large-scale photographs used in this study, either
of these procedures would have been feasible. Another
means of increasing accuracy could have been using the
moosehorn (9) or other device to make an objective eval-
uation of crown closure on the ground.
In short, too much reliance was placed on subjective
evaluations of crown closure, both on the ground and on
the photographs in this study.
CONCLUSIONS AND RECOMMENDATIONS
The question posed by the main objective of the study,
"Can photographs taken with a K-20 camera be used for
stand classification purposes?", cannot be answered un-
equivocably from the evidence obtained. However, a test-
ing procedure that can be used as the basis for further
studies which should provide the information needed to
answer the question has been developed.
There was good evidence that infrared film would
yield better results than would panchromatic, a result to
be expected under conditions of mixed hardwoods and soft-
woods. On the other hand, there was little evidence to
indicate that one photographic scale was any better than
the other. However, the smaller scale might prove better
in the long run because the larger ground area covered
per photograph would permit an easier understanding of
the drainage system of the area. An added benefit of the
smaller scale would be to lower photographic cost.
Certain recommendations can be made regarding future
studies of this type.
(1) The descriptive material used to guide the inter-
preters should be organized into a key. This key should
contain as many stereograms of condition classes as pos-
sible.
(2) Whenever possible, the critical items in the key
should be evaluated by objective (actual measurement)
rather than subjective (estimation) procedures. Grey
scales, such as that in the Munsell color system (7), should
be used to express grey tones. Tree heights should be
measured with a floating dot instrument. Crown diameters
should be measured with a micrometer wedge or tube
magnifier. Stand densities should be estimated by Losee's
procedure or with a dot grid.
(3) The key should be given preliminary tests and any
ambiguities should be resolved prior to the final test.
(4) The interpreters who test the procedure shoull
have a thorough indoctrination over a fairly lengthy period
prior to the test. During this period they should be closely
supervised by the project leader and should not be per-
mitted to proceed to the testing phase until the project
leader is satisfied that they have attained competence.
(5) The developer of the procedure should not be one
of the test interpreters.
LITERATURE CITED
(1) ANONYMOUS. 1962. Forest cover types of North Amer-
ica. Society of American Foresters, Washington, D.C.
67 pp.
(2) EASTMAN KODAK Co. 1964. Kodak data for aerial
photography. Eastman Kodak Co., Rochester, N.Y.
4 pp.
(3) HODGKINS, E. J. et al. 1965. Southeastern forest habi-
tat regions based on physiography. Auburn Univ.
(Ala.) Agr. Exp. Sta., Forestry Dept. Series No. 2.
10 pp.
(4) LOSEE, S. T. B. 1953. Timber estimates from large
scale photographs. Photogrammetric Engineering. 19:
752-762.
(5) MOESSNER, K. E. 1949. A crown density scale for
photo interpreters. Jour. Forestry. 47:569.
(6) MOESSNER, K. E. 1954. A simple test for stereoscopic
perception. U.S. Forest Service, Central States For.
Exp. Sta. Tech. Paper 144. 14 pp.
(7) MUNSELL COLOR Co. Munsell Book of Color. Mun-
sell Color Co., Baltimore, Md. 2 vols.
(8) PARKER, R. C. AND E. W. JOHNSON. 1969. Small
camera aerial photography - the K-20 system.
Accepted for publication, Jour. Forestry.
(9) RoBINsoN, M. W. 1947. An instrument to measure
forest crown cover. Forestry Chronicle. 23:222-225.