ASPRS
1990 specifies that horizontal
GCP
s and checkpoints
should be surveyed to an accuracy three times greater than the
maximum allowable
RMSE
at the chosen map scale, and verti-
cal
GCP
s and checkpoints should be surveyed to an accuracy
of one twentieth of the contour interval chosen. So for Class
1 accuracy at a scale of 1:500 and with a contour interval of
0.5 m, it would be necessary to survey points to an accuracy
of better than 0.04 m in both X and Y and 0.025 m vertically.
While there are no absolute standards for the number and
location of horizontal and vertical checkpoints, a minimum
of 20 well-distributed points are recommended for accuracy
assessment purposes.
ASPRS
1990 states that for the purpose
of map compilation and checking it is permissible to use local
horizontal and vertical datums, as long as both the control
survey and the survey being tested use the same datums
.
While
ASPRS
1990 provides a good starting point as far
as defining project accuracy is concerned, it was primarily
developed for printed maps. As technology has evolved there
has increasingly been a need for standards that pertain to digi-
tal data. In recognition of the changing geomatics landscape,
the
ASPRS
released a set of guidelines for vertical accuracy
reporting of lidar data in 2004 (
ASPRS
, 2004), hereafter re-
ferred to as
ASPRS
2004. Because of the difficulty of accurately
identifying horizontal positions on lidar data, accuracy guide-
lines were only provided for elevation data in locally flat
areas. These guidelines represented an incremental update of
ASPRS
1990. Under
ASPRS
2004, project accuracy was defined
at the 95 percent confidence level, which can be calculated
by multiplying the
RMSE
z
by a factor of 1.96, assuming errors
are normally distributed. In addition,
ASPRS
2004 recognized
that different levels of accuracy may be obtained from differ-
ent cover types within the same job, such as when the project
area includes a closed forest canopy as well as open ground.
Under such circumstances errors cannot necessarily be as-
sumed to be normally distributed (
ASPRS
, 2004). Unlike
ASPRS
1990 and
ASPRS
2015,
ASPRS
2004 was published as a set of
guidelines, and as such does not provide statistically testable
accuracy thresholds for project classification.
ASPRS 2015
Under
ASPRS
2015, horizontal and vertical accuracy require-
ments have been tightened up from those of
ASPRS
1990,
reflecting recent improvements in data collection and triangu-
lation procedures, especially at larger scales. The December
2013 draft of
ASPRS
2015 followed the tradition of defining
both horizontal and vertical accuracy hierarchically as being
Class 1, Class 2, Class 3, etc., with horizontal accuracy classes
being defined relative to the pixel size of the output orthoim-
age mosaic, while the vertical accuracy was determined rela-
tive to defined vertical
RMSE
thresholds
.
The final version of
ASPRS
2015 contains several important
changes from the December 2013 draft (see
ASPRS
, 2013). Nu-
merical ranking of horizontal and vertical accuracy classes is
a legacy from
ASPRS
1990, which was included in the Decem-
ber 2013 draft, but has now been dropped. Instead, horizon-
tal accuracy is now expressed in terms of classes defined
by check point
RMSE
x
and
RMSE
y
, with both assumed to be
normally distributed. Corresponding estimates of horizontal
accuracy at the 95 percent confidence level can be computed
using
NSSDA
methodologies, with
RMSE
x
or
RMSE
y
being multi-
plied by a factor of 2.448.
In non-vegetated areas, where checkpoint
RMSE
z
is assumed
to be normally distributed, the non-vegetated vertical accura-
cy (
NVA
) is computed by multiplying the checkpoint
RMSE
z
by
a factor of 1.96. For vegetated areas, no assumption of normal
distribution is made, and the vegetated vertical accuracy (
VVA
)
at the 95 percent confidence limit is computed as the 95
th
per-
centile of the absolute value of vertical errors in all vegetated
land cover categories combined, including tall weeds and
crops, brush lands, and fully forested areas. (
ASPRS
, 2015).
The
VVA
standard is 3.0 times the allowable
RMSE
z
value of the
corresponding vertical accuracy class. Common horizontal
accuracy classes defined by
ASPRS
2015 are listed in Table 1,
with vertical accuracy classes listed in Table 2
.
Where accuracy testing is to be carried out,
ASPRS
2015 stip-
ulates that a minimum of 20 checkpoints be used for testing of
either planimetric or vertical accuracy. While the distribution
T
able
1. C
ommon
H
orizontal
A
ccuracy
C
lasses
for
D
igital
O
rthophotos
and
A
ssociated
H
orizontal
S
tandards
for
GCP
s
(
after
ASPRS 2015)
Horizontal Accuracy Class
RMSE
x
and RMSE
y
(cm)
RMSE
r
(cm)
Allowable x or y
error for GCPs (cm)
Typical GSD of
Source Imagery (cm)
Appropriate scale for ASPRS
1990 Class 1 mapping
Appropriate scale for ASPRS
1990 Class 2 mapping
0.63
0.9
0.16
0.31 to 0.63
1:25
1:12.5
1.25
1.8
0.31
0.63 to 1.25
1:50
1:25
2.5
3.5
0.62
1.25 to 2.5
1:100
1:50
5.0
7.1
1.25
2.5 to 5.0
1:200
1:100
7.5
10.6
1.88
3.8 to 7.5
1:300
1:150
10.0
14.1
2.50
5.0 to 10.0
1:400
1:200
12.5
17.7
3,12
6.3 to12.5
1:500
1:250
15.0
21.2
3.75
7.5 to 15.0
1:600
1:300
17.5
24.7
4.38
8.8 to 17.5
1:700
1:350
20.0
28.3
5.00
10.0 to 20.0
1:800
1:400
22.5
31.8
5.62
11.3 to 22.5
1:900
1:450
25.0
35.4
6.25
12.5 to 25.0
1:1000
1:500
T
able
2. C
ommon
V
ertical
A
ccuracy
C
lasses
for
DEM P
oints
in
N
on
-V
egetated
and
V
egetated
T
errain
(
after
ASPRS 2015)
Vertical
Accuracy Class
RMSE
z
Non-Vegetated (cm)
Non-vegetated vertical
accuracy (NVA) (cm)
Vegetated vertical (VVA)
at 95
th
Percentile (cm)
Equivalent Class 1
contour interval per
ASPRS 1990 (cm)
Equivalent Class 2
contour interval per
ASPRS 1990 (cm)
1-cm
1.0
2.0
3.0
3.0
1.5
2.5-cm
2.5
4.9
7.5
7.5
3.8
5-cm
5.0
9.8
15.0
15.0
7.5
10-cm
10.0
19.6
30.0
30.0
15.0
15-cm
15.0
29.4
45.0
45.0
22.5
20-cm
20.0
39.2
60.0
60.0
30.0
33.3-cm
33.3
65.3
99.9
99.9
50.0
66.7-cm
66.7
130.7
200.1
200.1
100.1
100-cm
100.0
196.0
300.0
300.0
150.0
333.3-cm
333.3
653.3
999.9
999.9
500.0
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
October 2015
789
