MORPHOLOGICAL TERRAIN CLASSIFICATION AND 
ANALYSIS USING GEOSTATISTICAL TECHNIQUES 
JM. Azañôn®*, J. Delgado ”*, A. Gômez° 
* Dpto. de Geodinámica, Facultad de Ciencias, Universidad de Granada. 
Avda. Fuentenueva s/n — 18071 Granada, Spain — jazanon@ugr.es 
? Dpto. de Ingeniería Cartográfica, Geodésica y Fotogrametría, Escuela Politécnica Superior, Universidad de Jaén, 
c/ Virgen de la Cabeza, 2 — 23071 Jaén, Spain — jdelgado@ujaen.es 
* Dpto. de Ingeniería Cartográfica, Geodésica y Fotogrametría — Expresión Gráfica, 
Escuela Técnica Superior Ingenieros Agrónomos, Universidad Politécnica de Madrid, 
Ciudad Universitaria s/n, 28040 Madrid (Spain) — agomez(@stereocarto.com 
KEY WORDS: Spatial Information Systems, DEM/DTM, Geomorphology, Classification, Statistics 
ABSTRACT: 
Nowadays there is a wide range of systems that provide an efficient and fast data capture for the DEM generation and the spread of 
the computer systems has allowed the use of DEM to a high number of technicians. Unfortunately, the DEM analysis tools do not 
have similar development. Actually, the analyses are mainly based in the application of local scale operators that provide additional 
information. In this paper we present a methodology based on the geostatistical techniques applied to the analysis and classification 
of the different terrain morphologies obtained from the DEM. It is based on the use of several parameters obtained from the spatial 
autocorrelation functions together with other parameters obtained directly from the DEM following a moving-windows schema. 
Through this process it is possible to obtain a multivariate digital model that include information relative to the terrain heights, like, 
for example, variability, continuity level and anisotropies. A non-supervised classification process is applied to the multivariate 
digital model. This classification allows selecting the more interesting parameters in order to make the final classification. The main 
advantage of this method is based on its capabilities of the DEM analysis following a multiscalar schema. The analysis can include 
very interesting aspects like the anisotropies and other information derived from the structural analysis. This methodology is 
illustrated with an application example to the Guadix-Baza basin (S of Spain). 
1. INTRODUCTION point). Using the geostatistical procedures a multiscalar work 
approach can be made. 
The application of geostatistical methodology to DEM analysis 
has been exposed in several papers from different points of 
view. Thus, for example, the semivariogram has been used in 2. METHODOLOGY 
several works in order to characterize the terrain morphology 
(Mulla, 1988) and the kriging techniques have been utilized for The main objective of the proposed methodology is the terrain 
the elevation estimation (for example in Moffat et al., 19806). classification using the information derived from the statistical 
and geostatistical treatment of the digital elevation model. The 
The basic tool for the most of the geostatistical analysis is the applied classification is a non-supervised classification based in 
variogram function. The variogram is a function that establishes the ISODATA procedure (Tou and Gonzalez, 1974). 
the variance per points when the points are considered in pairs 
separated a distance h (see Goovaerts, 1997 and Webster and The proposed methodology can be divided into the following 
Oliver, 2001 for additional information). The variogram steps (figure 1): 
provides very important information. Thus, the variogram — determination of the parameters of the moving- 
provides information about the influence range of the value windows schema that will be used. These parameters 
registered in a point into the surroundings points, the variability must be defined according to the zone's dimensions 
at different scales, the anisotropies presence, continuity level, and data number available. The schema presents an 
periodicities, among others. This information can be used in overlapping windows system. 
order to characterize the phenomenon that we are studying. In — basic statistics and directional variograms calculation 
this paper, we intend to apply this information on the terrain of the data that are included in the different windows 
characterization and automatic classification. that are considered. 
— parameters selection that will be used in the 
Usually, the terrain classification procedures are based only in classification process. The parameters can be obtained 
the heights of the area contained in a digital elevation model or from the statistic analysis (mean, variance, minimum, 
in the used of derived models such as slopes, orientation, maximum, quartiles, range, …) and the geostatistical 
gradients, ... . The main advantage of these procedures is that analysis (variogram values at different directions and 
they work only on local level (neighborhood of a determined distances, ratio between the variogram values - 
anisotropy ratios-, ... . 
  
* Corresponding author. 
62 
Inte 
  
3.1 
In 
met] 
sele 
prov 
Nev 
The 
clas: 
grid 
map 
The 
the 
unit: 
foot 
deve 
pres 
imp 
Basi 
relie 
The 
8,75