George Miliaresis - Terrain Modelling for Specific ... · Terrain Modelling for Specific Geomorphologic Processing ... Geomorphometry; Terrain analysis; ... Today, the mountain topography

Proceedings of International Symposium on Terrain Analysis and Digital Terrain Modelling, 2006, Nanjing

1

Terrain Modelling for Specific Geomorphologic Processing

George Miliaresis

Department of Geology, University of Patras, Rion 26504, Greece [email protected]

phone: +30-2610-996296; fax: +30-2610-991900

Abstract: The aim of this paper is to review the global digital elevation data availability for both earth (GTOPO30, GLOBE and SRTM DEMs) and Mars (MOLA DEM) and the quantitative techniques and methods used in modern digital specific geomorphology from the author subjective point of view. The modern specific geomorphology requires a terrain partition framework that is implemented with digital image processing techniques (region growing segmentation, edge enhancement, histogram density slicing etc.) allowing the partition of the landscape to elementary objects (e.g. mountain and fluvial landforms). Thus, abstractions of landforms (objects) are defined on the basis of geomorphological principles and understanding. Each object is parametrically represented on the basis of its spatial 3-dimensional arrangement and mapped according to a terrain classification scheme (e.g. density slicing, unsupervised classification, fuzzy sets and fuzzy pattern recognition). The object partition framework could be also formed by artificial manmade partitions (e.g. municipalities and local authorities) while parametric representation could be based on either spectral indices derived from satellite multi-spectral imagery or landcover classification schemes. Keywords: Geomorphometry; Terrain analysis; Terrain classification; Geomorphologic mapping; Interpretation

1. Introduction Specific geomorphology involves subdividing a landscape into landforms based on a terrain segmentation methodology and measurement of their size, shape and relation to each other (Evans, 1981). During the initial steps specific geomorphology was mainly concentrated to drainage basin analysis since basins could be defined at a rather continuous way in the majority of the geomorphologic landscapes evident in mid-latitudes.

Fig. 1: Psysiographic map (left) and the corresponding mountain objects delineated from DEMs (middle image) and objects identification and labelling (right).

Various landforms were also studied. For example:

• Orometry, the 19th-Century measurement of mountains was an attempt to interpret landscape evolution and physical process that reflect the interplay of mountain building and erosion in regions of active deformation. Today, the mountain topography (Miliaresis, 2001) is of great significance not only in regional studies but also in terrain analysis, in navigation of airplanes and in the InSAR processing chain. Nowadays, tectonic geomorphology studies elevation and slope distributions, drainage spacing as a function of mountain-belt width, valley height/width ratios and other


2

geomorphometric attributes, that contribute to understanding of the evolution of mountain topography (Miliaresis, 2006b).

• During the early part of the 20th century, the study of regional-scale geomorphology was termed "physiography" (Fenneman, 1931). Physiographic analysis was based on the partition of terrain to physiographic units by taking into account the form and spatial distribution of their component features through fieldwork and visual interpretation of topographic maps and aerial photographs. Today, physiography is being stimulated by the need to explain enigmatic landscapes, newly explored on the surfaces of other planets through remotely sensed data. For example the study of impact craters and volcanoes during the early exploration of Moon and Mars (Pike, 2001) and the mapping and interpretation of extra-terrestrial chamsmata in Mars (Miliaresis and Kokkas, 2004).

• While physiographic analysis is concerned with regional scale geomorphology, terrain analysis is concerned with local (medium scale) geomorphology and it involves the systematic study of image patterns relating to the origin, morphologic history and composition of distinct terrain units, called landforms (Way, 1978). Landforms are natural terrain units, which when developed from the same soil and bedrock or deposited by a similar process, under similar conditions of climate, weathering, and erosion exhibit a distinct and predictable range of visual and physical characteristics on aerial images, called pattern elements (Lillesand and Kiefer, 1987). Typical pattern elements examined include topographic form, drainage texture and pattern, gully characteristics, soil tone variation and texture, land use, vegetation, and special features (Short and Blair, 1986). Towards, this end, landform quantification and modelling is of great importance in an attempt to assist interpretation of the exogenic and endogenic forces acting and forming the earths relief (Miliaresis and Argialas, 2000; Miliaresis, 2001).

The analysis steps (Fig. 1) in specific geomorphology include:

• Defining of a measurable form from a data source (topographic map, imagery), • Sampling a population of landforms, • Choosing descriptive parameters, • Measuring of enumerating parameters, • Analyzing the data, • Making sense of the results.

The aim of this paper is to review the digital data availability and the quantitative techniques and methods used in modern digital specific geomorphology from the author subjective point of view!

2. Methodology First, the data sources of hypsometric and environmental information are introduced. Then terrain partition methodologies are discussed. Finally objects parametric representation and classification schemes are analysed and particular case studies are presented.

Fig. 1: Globe DEM of Asia Minor (shaded relief map and the borders of extracted mountain features).


3

Fig. 3: MOLA DEM of Vales Marineris.

2.1 Data While during the initial steps specific geomorphology was based on the interpretation and measurement applied on topographic maps and imagery, nowadays, a broad-scale quantification of topography and digital elevation models (DEMs) represents the earth’s relief at regional to moderate scale (Pike, 1999). For example:

• The Global Digital Elevation Model ( http://edc.usgs.gov/products/elevation/gtopo30/gtopo30.html )

(GTOPO30) and the Global Land One-kilometer Base Elevation DEM (http://www.ngdc.noaa.gov/mgg/topo/globe.html) are available at a global scale (Fig. 2) providing a digital representation of the earth’s relief at a 30 arc-seconds sampling interval (Hastings and Dunbar, 1998).

• A moderate-resolution DEM ( http://pds-geosciences.wustl.edu/missions/mgs/megdr.html ) is available for Mars (Fig. 3), acquired by the Mars Orbiter Laser Altimeter (MOLA) a 10-Hz pulsed infrared-ranging instrument, operated in orbit around the planet from 1997 to 2001 aboard the Mars Global Surveyor (Miliaresis and Kokkas, 2004).

• The Shuttle Radar Topography Mission (SRTM) successfully collected Interferometric Synthetic Aperture Radar (IFSAR) data over 80 percent of the landmass of the Earth between 60 degrees North and 56 degrees South latitudes in February 2000 (Miliaresis and Paraschou, 2005). The Consortium for Spatial Information (CSI) of the Consultative Group for International Agricultural Research (CGIAR) is offering post-processed 3-arc second SRTM DEM data ( http://srtm.csi.cgiar.org ) for the globe (Void-filled seamless SRTM data, 2005).

Fig. 2: SRTM DEM of Southern Greece.

Additionally, a comprehensive land cover database is available, for the 25 EC Member States and other European countries, named Corine ( http://dataservice.eea.eu.int/ ). It provides quantitative data on land cover (Fig. 5), consistent and comparable across Europe at an original scale of 1: 100000 using 44 landcover classes (CLCs) of a 3-level nomenclature (CLC 2005).


4

Fig. 5: Corine landcover map of Southern Greece (Peloponnesus).

2.2 The Evolution of Digital Specific Geomorphology At the same time various digital image processing and G.I.S. techniques are being developed in order to automate the segmentation of digital imagery and DEMs (Soller, 1998). Quantitative techniques have been developed and applied in order to automate the interpretation of terrain features from DEMs (Miliaresis and Argialas, 1999a) and various geomorphometric parameters were developed in an attempt to characterize the landscape (Miliaresis and Argialas, 2002). The modern specific geomorphology requires a terrain partition framework first, allowing the partition of the landscape to elementary objects. Whereas the segmentation defines abstractions of landforms based on geomorphological principles and understanding (Miliaresis and Argialas, 2000), the terrain classification separates and maps landforms that have similar and contrasting ranges of characteristics (Miliaresis and Kokkas, 2004). Thus, each object should be parametrically represented on the basis of its spatial 3-dimensional arrangement and mapped according to a terrain classification scheme.

Fig. 6: Segmentation of bajadas from 15 minute US Geological Survey DEMs.

2.3 Geomorphometric Segmentation Discontinous and continuous terrain partition schemes will be presented. Geomorphometric segmentation was applied in order to isolated specific landforms from the geomorphologic background, for example the mountains and the fluvial landforms that provide a discontinuous terrain partition framework. Geomorphometric segmentation extracted mountains (Fig. 1 and Fig.2) developed on different base levels (Miliaresis 2001a) and fluvial landforms (Fig. 6) developed along mountain fronts (Miliaresis 2001b) from moderate resolution DEMs. The technique is summarized as a region-growing segmentation algorithm that uses seeds (for example the ridge cells) and a growing criterion (usually based on slope and elevation). The procedure for mountain features extraction is described hereunder.


5

• The data used for the extraction of mountain features was the GTOPO30 DEM with spacing 30 arc-

seconds since it provides a digital global representation of the earth’s relief at a regional scale and it is appropriate for regional scale (1:1,000,000) comparative studies. In a mountain, two parts are often distinctive: a) the gently sloping summit and b) the steep mountainsides. The process for the identification of mountains is based on the assumption that the summit or ridge pixels form the initial set of mountain pixels which needs to be expanded downslope taken into account the gradient values present in their neighborhood. The employed algorithms first identify the summits and then label the pixels around the summits as mountain pixels as long as their gradient was greater that the certain threshold.

• The summits were extracted and labeled by implementing runoff simulation. In this approach, a single water unit is imported in every cell of the DEM and travels according to the upslope aspect pointing direction. The water units imported in each cell are counted and finally, the derived values represent the runoff per cell. The cells with runoff values greater than a certain threshold should belong to the ridge network. Human expertise is required in order to judge if the resulting ridge network resembles the usual ridge network observed on maps in the current physiographic context. In this case study, it was found out that the threshold should be equal to 9.

• Then the slope was computed. The slope value depends on (a) the computation method and (b) the accuracy specification and grid size of the DEM. Due to the accuracy specification of the GTOPO30 a larger kernel of size 9*9 was selected for gradient computation. So the slope differ to a degree from the values an interpreter observes in the field and additional expertise should be developed by landform specialists in order to deal with this kind of artificially derived image. Statistical analysis of training areas indicated that the gradient of the mountainsides should be greater than 6 degrees. Note that if the gradient threshold was chosen greater than 6 degrees then the resulting mountains would have been smaller in size while if it was chosen lesser than 6 degrees then the resulting mountains would have been larger in size. This threshold is by no means applicable to other physiographic regions since their mountainsides could be less or more steep than the mountainsides observed in the Great Basin. Additionally, if a different relief representation and/or a different algorithm for gradient computation were going to be used then a different gradient threshold would have been derived even for the Great Basin.

• Then, an iterative region growing segmentation algorithm was applied to label the mountain pixels. The boundary of the mountain features was delineated and a unique integer identifier was assigned to each mountain.

• Visually comparing the extracted mountain features to the mountain ranges compiled by Fenneman (1933), it is observed that there is a fairly good correspondence between them, e.g., for each of the Fenneman ranges there is at least one range in the map of extracted mountain ranges. It is however observed that some of the mountain ranges of Fenneman appear rather broken in the map of extracted mountains. This discrepancy could be explained either by the level of generalization induced by human and machine or by the intrinsic nature of the mountain ranges in Great Basin. Fenneman could have used human expertise and fieldwork and thus, he might have connected isolated mountains and adjacent mountain ranges applying a generalization process.

• The mountain feature extraction techniques are subjective to a degree. Human expertise is needed in order to deal with the discrete representation of the terrain at various scales and select the most suitable algorithms that could deal with the elevation and positioning errors of the available datasets. Usually one has to use a specific dataset that is available and thus a particular expertise should be developed in order to deal with the derived images and models. The selection of thresholds for gradient or for runoff is performed through a trial and error procedure and through comparison of the derived images to our mental images and models for this particular physiographic context. In the next section quantitative attributes for the mountains will be defined.

On the other hand a continuous terrain partition scheme might be defined. The interpretation of aspect image indicates that the landscape might be divided to a continuous terrain partition framework on the basis of elementary regions composed by adjacent pixels with the same aspect pointing direction. These regions are named aspect regions (Fig. 7) and they are developed between the valley and ridge network (Miliaresis et al. 2005) and were identified by a connected component labeling algorithm applied to the aspect image (Miliaresis and Kokkas, 2004).


6

Fig. 7: Aspect regions (left) and slope image in SW Greece (Miliaresis et al., 2005).

In another research effort terrain objects coincide to the 3-d spatial objects defined by the local authorities’ border in Greece, an artificial terrain partition framework that in many cases contradicts to the geomorphologic and environmental organization of the terrain. The aim was to characterize the local authority terrain partition framework and its sensitivity to the influence of unfavorable and dangerous natural events or phenomena (Miliaresis, 2006a) on the basis of the DEM and the Corine landcover map (Fig. 8).

Fig. 8: The artificial terrain partition framework (local authorities) superimposed over the DEM (left) and the CORINE landcover classes (right image).

2.4 Parametric Representation The parametric representation (Miliaresis and Argialas, 2002; Miliaresis and Illiopoulou, 2004; Miliaresis and Kokkas, 2004) of either geomorphologic (mountains, aspect regions etc.) or artificial objects (local authorities) that formed polygons was achieved by the a set of parameters the most common being by the logarithm of their size (lnA), mean elevation (H), roughness (sd), local relief (LR), mean gradient (G), and the hypsometric integral (HI), etc.. More specifically: • LnA is indication of the area extent. • H is correlated to the annual rainfall height and vegetation type. • LR equals to the elevation range and it is correlated to the terrain diversity within the object • SD is the standard deviation of elevation and correlated to the height variability. • G expresses mean slope per object and it is correlated to the intensity of the erosion/deposition processes. • HI is used as an indicator of the relative amount of land (from the base of the mountain to its top) that was

removed by the erosion process. HI reflects the stage of landscape development (an indicator of the cycle of erosion).


7

Fig. 9: Density slicing of the slope to 7 slices of the aspect regions in Vales Marineris.

Nevertheless landcover (Fig. 8) and biophysical attributes (temperature, humidity, etc.) might be assigned to each geomorphologic object (Miliaresis, 2006a).

2.5 Terrain classification and mapping Various methods were applied for terrain classification and mapping that included descriptive statistics, frequency distributions and the linear regression between attributes in an attempt to characterize landscape (Miliaresis, 2006b). More specifically:

• In Vales Marineris (Miliaresis and Kokkas, 2004) the domain of the attributes was sliced to intervals (Fig. 9) on the basis of geomorphometric criteria and the interpretation of the resulted maps aided the interpretation of morphotectonic structure and evolution of both terrestrial and extraterrestrial landscapes.

Fig. 10: The 5 clusters derived by K-means cluster analysis in Zagros Ranges.


8

o Cluster analysis a multivariate procedure was also used for regional classification of the parametric representation of geomorphologic objects. It is based on some measurement of distance among objects, (for example Euclidean distance), which is calculated in a c-dimensional space, where c represents the number of attributes used in the clustering process (Miliaresis and Illiopoulou, 2005). The centroid method was employed that requires a priori definition of the number of clusters. For example, the centroid clustering method revealed clearly the SE-NW stair-step topography observed in Zagros Ranges (Fig. 10) while the steepest and more massive mountains were also observed along this direction. The zones derived by the mapping of clusters were associated to the existing morphotectonic zones of the study area while geomorphometric processing proved capable of segmenting morphotectonic zones to sub-regions with different geomorphometry.

• Fuzzy sets have been developed as a calculus for the representation of natural language in various

domains and are being used in the following for representation of the imprecision of the qualitative geomorphologic attributes (linguistic variables) used in our knowledge base. A variable is called linguistic if it can take words in a natural language as its values (Ross, 1995). The words are represented by fuzzy sets defined in the domain of the linguistic variable. More specifically, a linguistic variable is characterized by (Zadeh, 1975):

1. the name of the variable (e.g., Local Relief), 2. the set of linguistic labels that the variable takes (e.g., low, moderate, high), 3. the actual physical domain in which the linguistic variable takes its quantitative values (e.g., {300,

1200}), and 4. a semantic rule that relates each linguistic label of a variable with a fuzzy set in the actual physical

domain. Thus in order to quantify the natural geomorphologic language all four elements should be determined. The names of the linguistic variables and their labels were determined directly by geomorphologic descriptions. The quantitative values of the actual physical measurements were computed through the geomorphometric parameterization of each extracted mountain feature to a set of attribute values. A fuzzy partition of the physical domain was next implemented and a sub-domain for each linguistic label was derived. This was achieved based on geomorphological knowledge and trial and error experimentation. The semantic rules that relate each linguistic label with a fuzzy set in the actual physical domain were expressed through membership function (Zadeh, 1975). For a continuous variable (x), the membership function (MBF) describes the compatibility between the linguistic label (DMB=MBF(x)). DMB is called the degree of membership and its values are in the interval 0 to 1. The membership function of a linguistic label is a) subjective, b) context-dependent, and c) influenced by new numerical data and knowledge.

Fig. 11: Fuzzy set representation of the linguistic variable Diameter (size) of the Mountains in the Great Basin physiographic context.

There is no general method to determine an MBF. Its specification is a matter of definition, rather than of objective analysis (Wang, 1996). Many different shapes of MBFs have been proposed in the literature and


9

the most practical implementations use the so-called "Standard MBFs" (Altrock, 1995) that are normalized (maximum is always 1 and minimum 0). The definition of a standard MBF includes the following steps (Altrock, 1995): a) Define the value of the domain that best fits to the meaning of the label and assign DMB equal to 1 and b) Define the rightmost and the leftmost values (DMB=0) of each linguistic label assuming that adjacent labels have usually 60% overlap. For example, for the fuzzy sets that correspond to each label of the linguistic variable Size of the Mountain Ranges in the Great basin physiographic context [Miliaresis and Argialas 1999b; 1999c) are given in Figure 11. The fuzzy sets allowed the fuzzy partitioning of the domain of geomorphic variables in Great Basin and the quantitative representation of the geomorphic language in that geomorphic context. Their definition was based on both (a) well-accepted geomorphic knowledge and (b) the geomorphometric data acquired for the study area (physical domain). Thus, the geomorphic language describing the mountain ranges was quantified for the Great Basin context (Miliaresis and Argialas 1999b). Fuzzy pattern recognition might be applied and the overall assessment of the resemblance objects can be assessed on the basis of the maximum approaching degree (Miliaresis and Argialas, 1999c). So a user of a computer system could be assisted during the interpretation process by recalling the knowledge base of Great Basin and by the numerical values observed. The computer system projects the values he gave to the domain of the Great Basin digital words and would give responses like this mountain feature is small in size and elongated in a Great Basin physiographic context (Miliaresis and Argialas, 1999b). Human factors are crucial for the fuzzy partition of the domain and for the selection of the MBF types. The last selection influences the interpretation process and is performed by a trial and error procedure on the basis of human expertise and the derived quantitative data (domain). Although there is degree of subjectivity, a novice interpreter could be assisted and make judgements on the basis of the relative (context-dependent) knowledge base of “digital words”. In the future when perhaps a more complete and tested knowledge base could be made available for various physiographic regions it will lead to the creation of an absolute definition (non-context dependent) of the geomorphic words and terms.

3. Conclusion Global digital elevation models of earth and other planets have fostered specific geomorphology and terrain modelling at broad spatial scales. The broad-scale quantification of topography and the DEM-based analyses transformed specific geomorphology into one of the most active and exciting fields in the Earth sciences. Segmentation techniques allowed the partition of the terrain to continuous (aspect regions) and discontinuous (mountains features, fluvial landforms, etc.) schemes. Additionally, non geomorphologic (artificial) terrain portioning schemes might be implemented as in the case of local authorities. The parametric representation of the derived objects can be achieved by means of both geomorphometric, landcover and biophysical attributes derived by DEMs, landcover maps and satellite imagery. Finally object classification schemes (density slicing, K-means cluster analysis, fuzzy set representation) allow the terrain classification and interpretation. Modern digital specific geomorphology provides a quantitative way to compare developed and developing landscapes in areas of both differing and similar geologic structure. Additionally, experience with the earth’s landscape assists the exploration and interpretation of various landscapes in inaccessible areas on other planets from digital elevation data.

References [1] Altrock C., 1995. Fuzzy logic and neurofuzzy applications explained. Upper Saddle River, NJ: Prentice Hall. [2] CLC 2005. Corine land cover 100 m – ver. 5. European Environment Agency {©EEA, Copenhagen},

http://dataservice.eea.eu.int/ [3] Evans I., 1981. General geomorphometry. In: Geomorphological Techniques, by A. Goudie (Editor),

British Geomorphological Research Group, George Allen & Unwin, London, pp. 31-37 [4] Fenneman N., 1931. Physiography of western United States. New York: McGraw-Hill. [5] Hastings D., & Dunbar P., 1998. Development & assessment of the Global Land One-km Base Elevation

digital elevation model (GLOBE). ISPRS Archives, 32(4): 218-221 [6] Lillesand T., & Kiefer W., 1987. Remote Sensing and Image Interpretation (second edition), John Wiley

and Sons, New York, 721 p. [7] Miliaresis G., 2001a. Geomorphometric Mapping of Zagros Ranges at Regional Scale. Computers &

Geosciences, 27, 775-786.


10

[8] Miliaresis G., 2001b. Extraction of Bajadas from Digital Elevation Models & Satellite Imagery. Computers & Geosciences, 27, 1157-1167.

[9] Miliaresis G., 2006a. Geometric and Landcover Signatures of Local Authorities in Peloponnesus. WSEAS Transactions on Environment and Development, 4, (2), 239-244 p.

[10] Miliaresis G., 2006b. Geomorphometric mapping of Asia Minor from Globe digital elevation model. Geografiska Annaler Series-A, 88 A (3),209-221

[11] Miliaresis G., Argialas D., 1999a. Segmentation of Physiographic Features from the Global Digital Elevation Model / GTOPO30. Computers & Geosciences, 25, 715-728.

[12] Miliaresis G, & Argialas, D 1999b. Formalization of the photo-interpretation process by a fuzzy set representation of mountain objects in the geomorphic context of the Great Basin section. Proceedings of the 25th Conference of the Remote Sensing Society. Cardiff, United Kingdom, 745-750.

[13] Miliaresis G., & Argialas D, 1999c. Fuzzy pattern recognition of compressional mountain ranges in Iran. Proceedings of the 5th International Conference of the Association for Mathematical Geology. Trondheim, Norway, 227-233.

[14] Miliaresis G., & Argialas D., 2000. Extraction & Delineation of Alluvial Fans from DΕΜs & Landsat TM Images. Photogrammetric Engineering & Remote Sensing, 66, 1093-1101.

[15] Miliaresis G., & Argialas D., 2002. Quantitative Representation of Mountain Objects Extracted from the GTOPO30 DEM. Int. Journal of Remote Sensing, 23, 949-964.

[16] Miliaresis G., Illiopoulou P., 2004. Clustering of Zagros Ranges from the Globe DEM representation. Int. Journal of Applied Earth Observation & GeoInformation , 5(1), 17-28.

[17] Miliaresis G., & Kokkas N., 2004. Segmentation and terrain modeling of extra-terrestrial chasmata. Journal of Spatial Sciences, 49(2), 89-99.

[18] Miliaresis G., Sabatakakis N., & Koukis G., 2005. Terrain pattern recognition and spatial decision making for regional slope stability studies. Natural Resources Research, 14(2), 91-100.

[19] Miliaresis G., & Paraschou C., 2005. Vertical accuracy of the SRTM DTED Level 1 of Crete. Int. Journal of Applied Earth Observation & GeoInformation, vol. 7(1), 49-59.

[20] Pike R., 2001. Scenes into numbers: facing the subjective in landform quantifiacation. In «Interpreting Remote Sensing Imagery: Human factors» by R. R. Hoffman and A. B. Markman (Eds), 304 pages, ISBN:1566704138 . Lewis Publishers-CRC PRESS , New York, pp. 83-114

[21] Pike R., 1999. A bibliography of geomorphometry the quantitative representation of topography – supplement 3- OF 99-140. Menlo Park, U.S. Geological Survey.

[22] Ross T., 1995. Fuzzy logic with engineering applications. New York, McGraw Hill. [23] Short N., & Blair R., 1986. Geomorphology From Space, a Global Overview of Regional Landforms,

(NASA SP-486), U.S. Government Printing Office, Washington DC, 709 p. [24] Soller D., 1998. Digital mapping techniques. Open-file report 98-487. Illinois, U.S. Geological Survey. [25] Void-filled seamless SRTM data, 2005. International Centre for Tropical Agriculture (CIAT), available

from the CGIAR-CSI SRTM 90m Database, URL: http://srtm.csi.cgiar.org [26] Wang P., 1996. The interpretation of fuzziness. IEEE Transactions on Systems, Man, and Cybernetics,

26(2), 321--326. [27] Way D., 1978. Terrain analysis. New York, McGraw-Hill. [28] Zadeh L., 1975. The concept of a linguistic variable and its application to approximate reasoning I, II, III.

Information Sciences, No.8, 199-251 (I); 8, 301-357 (II); 9, 43-80 (III).

Acknowledgements The author would like to thank the Organising Committee, the Prof Liu Xuejun (Symposium Secretary), and the Prof Brian Lees (Program Committee Chairman) of the International Symposium on Terrain Analysis and Digital Terrain Modelling for the keynote speaker invitation and to express his gratitude for the financial support offered.

Documents

George Miliaresis - Terrain Modelling for Specific ... · Terrain Modelling for Specific Geomorphologic Processing ... Geomorphometry; Terrain analysis; ... Today, the mountain topography