PE&RS August 2015 - page 657

PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING August 2015 657
Using International Standards to Control the
Positional Quality of Spatial Data
F.J. Ariza-López, and J. Rodríguez-Avi
Abstract
A positional quality control method based on the application
of the International Standard ISO 2859 is proposed. This entails
a common framework for dealing with the control of all other
spatial data quality components (e.g., completeness, consis-
tency, etc.). We propose a relationship between the parameters
“acceptable quality level” and “limiting quality” of the inter-
national standard and positional quality by means of observed
error models. This proposal does not require any assumption
for positional errors (e.g., normality), which means that the ap-
plication is universal. It can be applied to any type of positional
and geometric controls (points, line-strings), to any dimension
(1D, 2D, 3D, etc.) and with parametric or non-parametric error
models (e.g., lidar). This paper introduces ISO 2859, presents
the statistical bases of the proposal and develops two examples
of application, the first dealing with a lot-by-lot control and the
second, isolated lot control.
Introduction
Positional quality is one of the most desirable characteristics of
spatial data and is determined by positional accuracy. Positional
quality is assessed by positional accuracy, which is a matter of
renewed interest because of the capabilities offered by Global
Navigation Satellite Systems (GNSS) and the need for greater spa-
tial interoperability to support the Spatial Data Infrastructures.
In a Spatial Data Set (SDS) the position of a real world
entity (feature) is described/recorded with position values of
geometric objects (e.g., points, line-strings, shapes, etc.) in an
appropriate coordinate system. Positional accuracy represents
the nearness of those values to the entity’s “true” position in
that system. Positional accuracy has traditionally been evalu-
ated using control points. Following this idea, there are very
many statistical Positional Accuracy Assessment Methodol-
ogies (PAAM), for example: National Map Accuracy Standard
(NMAS) ( USBB, 1947), Engineering Map Accuracy Standard
(EMAS) ( ASCE, 1983), National Standard for Spatial Data Accu-
racy (NSSDA) ( FGDC, 1998), STANAG 2215 (STANAG, 2002), ASPRS
Accuracy Standards for Large-Scale Maps (ASLSM) ( ASPRS,
1990) and the ASPRS Positional Accuracy Standards for Digital
Geospatial Data (ASPRS, 2015).
The majority of PAAMs take as an underlying hypothesis
the Gaussian distribution of positional errors, but several
studies indicate that this hypothesis is not true. For instance,
in the case of GNSS-error distribution, the Rayleigh distribu-
tion (Logsdon, 1995); and for the case of geocoding errors, a
log-normal distribution (Cayo and Talbot, 2003). For the case
of vertical errors in digital elevation models there are many
references (e.g., Bonin and Rousseaux 2005, Oksanen and Sar-
jakoski 2006) indicating that error distribution is not Normal,
and for this reason, it is proposed to express the results of
quality control checks by means of percentiles (Maune, 2007)
of the observed distribution. The latter has been introduced
by the ASPRS into the Guidelines for reporting vertical accura-
cy for lidar data (ASPRS, 2004).
Another source of criticism is related to the use of points
as control elements. Some researchers (e.g., Joao, 1998) have
criticized PAAMs for being limited to well-defined points, and
also for failing to address more complex control elements like
linear and aerial ones. In the last two decades the positional
assessment (planimetric) of line-strings elements has been an
issue of great interest for researchers, and there are some pro-
posed methods for their positional accuracy assessment. There
are methods based on distances (e.g., the Hausdorff Distance
Method by Abbas et al. (1995) and the Mean Distance Meth-
od by Skidmore and Turner (1992), and methods based on
buffers (e.g., the Single Buffer Overlay Method by Goodchild
and Hunter (1997), and the Double Buffer Overlay Method by
Tveite and Langaas (1999). In many of these cases the resulting
errors follow non-parametric distributions (distribution free).
PAAMs can be classified by means of the statistical ap-
proach: estimating methods (e.g., NSSDA) and control methods
(e.g., NMAS, EMAS, ASLSM). In the first group the estimation is
expressed by means of a confidence interval using an un-
derlying statistical model. If PAAMs are going to be used for
the acceptance or rejection of products (control approach),
explicit information about user (buyers or consumers) and
producer risks are needed. The introduction of such risks is
a new paradigm that has not been considered in any of the
traditional PAAMs. Some research has focused on users’ and
producers’ risks, e.g., for the cases of the EMAS (Ariza-López et
al., 2008) and the ASLSM (Ariza-López et al., 2010).
The use of industrial standards for quality control of
spatial data is possible, for instance through the use of ISO
2859 (Sampling procedures for inspection by attributes) for
counting errors (e.g., completeness, consistency, etc.) and ISO
3951 (Sampling procedures for inspection by variables) for
continuous errors (e.g., positional errors). Interest in applying
these standards to spatial data has been increasing since the
inclusion of some guides and examples of use previously in
ISO 19114 (Geographic Information - Quality Evaluation Proce-
dures) and now in ISO 19157 (Geographic Information - Data
Quality). For instance, ISO 2859 is applied widely throughout
the world to control attributes, e.g., for controlling the Land
Parcel Identification System in Europe (Milenov et al., 2010),
for geological data in China (Xie et al., 2008) and for diverse
spatial data quality controls in New Zealand by the National
Topographic Hydrographic Authority (NTHA, 2004). ISO 3951
has a much lower diffusion and application in the field of
F.J. Ariza-López is with the Universidad de Jaén, Departa-
mento de Ingeniería Cartográfica, Geodésica y Fotogrametría,
E-23071-Jaén, Spain ( fjariza@ujaen.es).
J. Rodríguez-Avi is with the Universidad de Jaén, Departa-
mento de Estadística e Investigación Operativa, E-23071-Jaén,
Spain.
Photogrammetric Engineering & Remote Sensing
Vol. 81, No. 8, August 2015, pp. 657–668.
0099-1112/15/657–668
© 2015 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.81.8.657