‘ment of the
acy that, in
he traditional
d to perform
using new
with enough
based on a
alysis, makes
ous way, the
eld, without
dure can be
'W surveyed
sation of the
ances of the
al systematic
rocrustes are
ponents and
> presence of
spects of the
ty to detect
tical tests of
stes Analysis
Optical 3-D
. 345-353.
onal scaling:
ork, pp. 337-
ions. DSWO
3-61.
omposizione
di analisi
Udine Univ.
d Procrustes
independent
i ote Sensing,
ipe analysis.
7.
ical analysis
hodological,
s analysis.
ninations of
78, pp. 279
ne matrix to
igid motion.
rotation for
7-276.
A PROPOSAL FOR NEW ALGORITHMS TO EVALUATE DATA QUALITY IN
MULTIDISCIPLINARY GIS
T. BELLONE(*), F. RINAUDO(*), A. SPANO’ (*%)
(*) DIGET - Politecnico di Torino, Turin, Italy
(**) DINSE — Politecnico di Torino — II Facolta di Architettura, Turin, Italy
tamara.bellone@polito.it - fulvio.rinaudo@polito.it - antonia.spano@polito.it
KEYWORDS: algorithms, data quality, quality control, Accuracy, GIS, Cultural Heritage
ABSTRACT
After carrying out a GIS aimed to the reconstruction of a medioeval setting in South-West Piedmont Region (a land of Alpine valleys
of North-West Italy where a large and important bishopric was settled), a selected working group, among the whole interdisciplinary
group working on a general project, started to attend to the data quality evaluation and control.
Historical nature of such GIS, whose first purpose is to define the human settlement in a certain territory, and qualitative
characteristics of most data managed in GIS have advised to use non-traditional statistical tools to evaluate data.
Actually, first of all, norms prescribed by proper Authorities for Standards due to archiving, representing and transmitting of
collected data have been taken into account, but the most important attention is given to focuse some methods or techniques to
manage the quality control when data sets have high degrees of uncertainty. So the main purpose was not the attainment of data
quality control in global datasets array, but studying and testing some quality control criteria suitable to evaluate several datasets
directly connected with historical and archaeological researches.
Tests from non-parametric inference sphere have been applied, and we are testing other algorithms to search proper tools to solve
problem of vagueness and lack of clarity of data; a proposal of applications of Bayesian theorem and Markov chains is also
presented.
1. INTRODUCTION
The early intention to work in the sight of spatial data
Standards, carrying out the interdisciplinary GIS named
“Marchesato di Saluzzo”, brought the data quality control. Even
more than in regional GIS, this special kind of GIS project,
where spatial and historical data are connected and often users
are not accustomed to treat spatial information, requires
particular care in data quality managing: different origin and
quality data are mixed and users have to be informed about the
usage threshold.
Most efforts of general GIS application concerned the
collection of features and their attributes generating from
documentary sources examination and archaeological data
gatherings; they have been organized in a datasets frame.
These datasets, referring the medioeval scenery that one wants
to reconstruct, are overlying on modern topographic and
thematic maps.
Maps used in GIS application are mostly selected from national
or regional mapping series, provided by public or private
institutions, so it has already been submitted to quality control
processes and just obtained quality assurance. So the quality
control procedures we have studied are oriented to evaluate new
accomplished datasets and their relationships with territorial
features.
Data quality for GIS is quite important: actually, this
technology lets often the users mistake innovation as real
increase of data quality; however, problems dealed with are
different from the ones of old maps, which had a uniform
quality for all objects: on the contrary, for GIS, data with
different origin and quality may often be mixed. Quality is an
extensive term, since managing quality is itself a complex
matter. In the last decade a lot of studies and reflections about
quality management have developed together with spatial data
standard statement.
This paper is the following of an earlier study started with a
review of quality parameters as they are defined by standards.
(Bellone, Spano', 2002). Our work has started with a basic
issue: GIS projects use maps that are reality models and a way
to check closeness of models to reality is to compare with
reference datasets.
Quality indices for a GIS are, as everybody knows: accuracy,
completeness, lineage, consistency. A way to deal with data
quality control in such multidisciplinary GIS, was to fit widely
accepted tools of quality evaluation to the particular contexts
met. Against a high grade of completeness for non spatial data,
we had to cope with the heavy difficulty to identify the spatial
location of stored structures. A feature of proposed GIS is that,
for many structures under study, location on maps may be
ambiguous, due to the weakness of their traces. So, at first, we
decided to place them according to acceptable rules, related to
the precision of available maps.
Thematical accuracy takes into account attributes given by
database, both quantitative or qualitative. In the first case, it is
obvious to use some type of statistical estimator, as r.m.s. In the
second case, we applied the error of classification matrix,
usually used to measure thematic accuracy of geometric features
(Giordano, Veregin, 1984) to evaluate qualitative attributes
generating from historical sources. Adding to the need of
evaluating thematic accuracy of historical and archaeological
attributes, one of the major difficulties noticed during data
collection was to refer many sets of data to terrain, since the
quality parameters we are interested the most is spatial and
thematic accuracy of data. It regards both preserved and
unpreserved structures (or ruins) because it is often difficult to
share attributes deriving from multidisciplinary investigations to
map entities without a terrain recognition. This item is one of
the most significant matters that modern archaeological GIS