Expert system for evaluating the Environmental Sensitivity Index (ESI) of a local area: methodology
Author: Agostino Ferrara, University of Basilicata - Italy <ferrara@unibas.it>

This methodology is an updated version of that which was developed and tested in the MEDALUS III project by the O.Us. coordinated by: Agostino Ferrara, University of Basilicata - Italy  and  Constantinos Kostas, University of Athen - Greece

Main references

• Ferrara A., Bellotti A., Faretta S., Mancino G., Taberner M., 1999. Identification and assessment of Environmentally Sensitive Areas by Remote Sensing. MEDALUS III 2.6.2. - OU Final Report. King's College, London. Volume 2: 397-429
  
• Kosmas C., Ferrara A., Briasouli H., Imeson A. 1999. Methodology for mapping Environmentally Sensitive Areas (ESAs) to Desertification. In 'The Medalus project Mediterranean desertification and land use. Manual on key indicators of desertification and mapping environmentally sensitive areas to desertification. Edited by: C. Kosmas, M.Kirkby, N.Geeson. European Union 18882. pp:31-47 ISBN 92-828-6349-2
  

These categories are not necessarily independent: remotely sensed radiometric data and vegetation or phytoclimatic maps, are all influenced by different factors from many different origins. Data can be obtained from available documents or obtained by dedicated surveys where costs depend on the ease with which they are obtained. They can be nominal (e.g. crop data, crop type, and forms of tillage), presence or absence, ordinal, discrete (e.g. a pedological system, soil water, or organic matter content), or continuous (e.g. information from a Digital Elevation Model) to name but a few. The complexity of the information is related to the sophistication of the questions that have to be answered; yet the combination of different data and complex questions means that the data have to be analysed in an integrated way to extract succinct, and well founded, answers (De Jong, 1994; Ferrara et al., 1995; Yassoglu et al., 1995).

In this context, a comprehensive system was developed to evaluate and investigate the causes and responses which contribute to the Environmental Sensitivity to desertification of each fundamental land unit describing an Environmental Sensitive Area (ESA). The coverage of the scheme ranges from the local to basin-wide scales. The system presented here is an expert application of this system, presented as a web-based tool to enable the Environmental Sensitive Index to be determined for individual areas. In the system data from many different sources, such as qualitative and quantitative satellite radiometric measurements, available geographical data, and ad hoc ground surveys, can be integrated.

5 top

g Methodological hypothesis

The Environmental Degradation or Sensitivity of an area to desertification is a complex concept to rationalise since, depending on the context, it can be caused by many different factors operating in isolation or in association (Rubio, 1995; Thornes, 1995; UNEP, 1977; UNEP, 1992; Basso, 1996). An Environmental Sensitive Area (ESA) can be considered, in general, as a specific and delimited entity in which environmental and socio-economical factors are not balanced or are not sustainable for that particular environment. The Environmental Sensitivity Index (ESI) to degradation or desertification of an area can also be seen as the result of the interactions among elementary factors (information layers) that are differently linked to direct and indirect degradation or desertification phenomena (Basso et al., 1998). Severe, irreversible environmental degradation phenomena, for example, could result from a combination of poor management quality together with various combinations of critical environmental factors (soil, climate, and vegetation). In order to make informed decisions it is necessary to be able to characterise and identify the significant factors which produce critical situations. As these factors are rarely independent, it is also necessary to be able to establish their interrelationships so that their relative contributions can be determined. On the other hand decision makers require functional summaries highlighting major issues; the straightforward identification of sensitive areas irrespective of source; and the ability to determine the effect of remedial actions without recourse to the intricacies of the scientific background.

In terms of the data required for estimating the Environmental Sensitivity to desertification, three different primary types of information are essential: physical-structural; vegetal; socio-economic. The groups are not necessarily independent: remote sensed radiometric data, a potential vegetation map, or a phytoclimatic one, are all influenced by diverse factors arising from each context. The data can be obtained from available documents, or obtained in the field. They can be nominal (e.g. crop data, types of crops, and forms of tillage systems), presence or absence, ordinal, discrete (e.g. pedological system, soil water, or organic matter content), or continuous (e.g. information provided by digital elevation models or land surveys) to name but a few. The complexity of the information is related to the sophistication of the questions that have to be answered, yet the combination of complex data and complex questions means that the data has to be analysed in an integrated way to extract succinct, and well founded, answers.

Regarding data analysis, data processing techniques usually simplify the data when deriving results. Sampling and homogenisation cause a reduction in data accuracy as highly detailed information layers, and layers with different information, can only be joined to each other at a simpler level. Classification procedures used to interpret the different information layers also lose information. In fact, classifications simplify the data by summarising the multiplicity of sampled, or calculated, attribute values with a limited number of intervals, whereby the initial detailed information are lost in the coarser classification groupings. Interpretations and simplifications of the data are, however, required whether arising from the need to provide classifications, to group the data in a homogeneous way, to organise the data into common reference systems, or when comparing different types of environments.

In terms of the hardware and software requirements, the use of the new generation of low cost Geographic Information Systems facilitates: access to the information; data maintenance; and interpretation through various procedures of cross analysis. It also becomes possible to acquire and analyse the information quickly to identify the transformations which, in turn, enable the necessary implementations to be defined, selected, programmed and initiated.

5 top

g Layers used in the present system

Three essential considerations were taken into account when selecting the information layers:

- their correlation to degradation phenomena or environmental critical state;

- coverage;

- ease and economy of updating.

Establishing a system which requires information that is difficult to obtain, or is expensive to update, even if it is scientifically important, would be severely restrictive and be impractical in large, complex, environments or with continuous monitoring systems. The system developed permits the easy addition or removal of information layers. Layers can be added when there is a requirement to study specific aspects or areas in greater detail, layers can be removed when a first approximation of an ESI estimate is required and all the desired information is still not available over the area of interest.

The optimised set of layers currently used in evaluating ESI to desertification and their sources are summarised in Table 1 while in the following paragraphs a more detailed description is presented. Comprehensive descriptions on how the environmental layers are linked to the degradation or desertification phenomena is given: in dedicated pages of the web site; in Basso et al., 1997; Basso et al., 1998; FAO, 1976; Briggs et al. 1992; Kosmas et al. 1994; Kosmas et al. 1998; Kosmas et al. 1998; Kosmas 1998; Kosmas et al. 1999; Poesen et al 1996; and in the descriptions of the procedures for the selection of the layers.

Table 1 - Set of information layers used in evaluating ESI to desertification and their related sources.

Quality	Layer	Source	Used layers
Soil	Parent material, Rock Fragments, Soil Depth, Slope Angle, Drainage, Soil Texture	Published data at various scales and field samplings	All
Climate	Rainfall, Aspect, Aridity index (Bagnouls e Gaussen)	Published data at various scales, field samplings and DEM	All
Vegetation	Fire risk, Erosion protection, Drought resistance, Plant cover	Landsat TM, published data at various scales and field samplings	All
Management	Policy enforcements, Land use intensity	Statistical data, mainly at municipality level	All (*)

5 top

g Detailed scheme of the optimised layers-score system

Soil layers

Soil is a crucial factor in evaluating the Environmental Sensitivity of an ecosystem, especially in the arid, semi-arid and dry sub-humid zones. Soil properties related to desertification and degradation phenomena affect two principal parameters:
   (i) water storage and retention capacity;
   (ii) erosion resistance.

The complete evaluation of the quality of a soil with respect to these properties can be performed by using soil properties given in regular soil survey reports. Soil layers and relative scores are reported in Table 2.

Layer	Classes	Scores
Parent material	Shale, schist, basic, ultra basic, conglomerates,   unconsolidated, clays; , marl (with natural veg.); Limestone, marble, granite, rhyolite, ignibrite, gneiss, siltstone, sandstone, dolomyte;  Marl, Pyroclastics.	1 1.7 2
Soil texture	L, SCL, SL, LS, CL SC, SiL, SiCL Si, C, SiC S	1 1.2 1.6 2
Rock fragments cover, %	> 60 20 - 60 < 20	1 1.3 2
Soil depth, cm	deep (>75 cm) moderate (75-30 cm) shallow (15-30 cm) very shallow (<15 cm)	1 2 3 4
Drainage	well drained imperfectly poor drained	1 1.2 2
Slope, %	< 6 6 - 18 18 - 35 > 35	1 1.2 1.5 2

 

Climate layers

Climate quality is assessed on the basis of how it influences water availability to the plants. Consideration has been given to the amount of rainfall, air temperature and aridity. Climate layers and relative scores are reported in Table 3. In particular the selected layers are: Annual precipitation (a crucial parameter in plant growth); Bagnouls-Gaussen aridity index (a synthesis of precipitation, evapotranspiration and run-off information); Slope aspect (affects microclimatic conditions and erosion).

Climate quality classes are:

Classes	Ranges
High quality	< 1.15
Medium quality	>= 1.15 <= 1.81
Low quality	> 1.81

Layer	Classes	Scores
Plant cover, %	> 40 40 - 10 < 10	1 1.8 2
Fire risk	Bare soils; Bedrocks; Almonds; Orchards; Vines; Olives; Irrigated annual crops (maize, tobacco, sunflower, ... ); Horticulture. Bare soils; Bedrocks; Almonds; Orchards; Vines; Olives; Irrigated annual crops (maize, tobacco, sunflower, ... ); Horticolture.	1
	Perennial grasslands; Pastures; Cereals; Annual grasslands; Deciduous forests (oak, mixed); Mixed Mediterranean macchia-Evergreen forests (with Q. ilex); Very low vegetated; Shrublands.	1.3
	Mediterranean macchia.	1.6
	Pines and other conifer forests.	2
Erosion protection	Evergreen forest (except conifers); Mixed Mediterranean macchia-Evergreen forests (with Q. ilex); Bedrocks.	1
	Mediterranean Macchia; Conifer forests; Perennial grasslands; Pastures; Olives; Shrublands.	1.3
	Deciduous forests (oak, mixed).	1.6
	Almonds; Orchards; ...	1.8
	Vines; Horticulture; Annual crops (cereals, maize, rice, oats, barley, annual grasslands, ...., etc.);   Very low vegetated; Bare soils .	2
Drought resistance	Evergreen forest (except conifers);  Mediterranean macchia;  Mixed Mediterranean macchia - Evergreen forests (with Q. ilex); Bedrocks; Bare soils.	1
	Conifer forests;  Deciduous forests;  Olives.	1.2
	Almonds; Orchards; Vines.	1.4
	Perennial grasslands; Pastures; Shrublands.	1.7
	Annual crops (annual grassland, cereals, maize, tobacco, sunflower, ...); Horticulture; Very low vegetated .	2

 

Vegetation quality classes are:

Land use intensity

Agricultural land (cropland): The intensity of land use of a cropland is classified into three classes based on the frequency of irrigation, degree of mechanization of cultivation, application of fertilizers and agrochemicals, types of plant varieties used, etc.

Table 5a - Land use intensity for agricultural land (cropland)

Class	Description	Score
1 2 3	Low land use intensity (LLUI) Medium land use intensity (MLUI) High land use intensity (HLUI)	1 1.5 2

Agricultural land (pasture land): The intensity of land use a pasture land is defining by estimating the sustainable stocking rate (SSR) and the actual stocking rate (ASR) (described previously) for the various pieces of land under grazing. Then, the intensity of land use is assessed by using the ratio of ASR/SSR and classified into three following classes.

Table 5b - Land use intensity for agricultural land (pasture land)

Class	Description	Stocking rate	Score
1 2 3	Low Moderate High	ASR<SSR ASR =SSR to 1.5 *SSR ASR>1.5*SSR	1 1.5 2

Natural areas (forest, macchia, shrubland, bare land): In natural areas such as forests, shrubland, bare lands, macchia etc., the intensity of land use is defined by assessing the actual (A) and sustainable yield (S). Then, the intensity of land use is classified into three classes based on the ratio A/S.

Table 5c - Land use intensity for natural areas (forest, shrubland, bare land)

Class	Description	Management characteristics	Score
1 2 3	Low Moderate High	A/S=0 A/S<1 A/S= 1 or greater	1 1.2 2

Mining areas: The intensity of land use for areas with mining activities is defining by evaluating the measurements undertaken for soil erosion control such as terracing, vegetation cover, etc. Then, the intensity of land use is classified into three classes based on the evaluated degree of land protection from erosion.

Table 5d - Land use intensity for mining areas

Class	Description	Erosion control measurements	Score
1 2 3	Low Moderate High	Adequate Moderate Low	1 1.5 2

Recreation areas: In areas undergoing active recreational use such as skiing, rallies etc., the intensity of land use is evaluated by defining the actual and the permitted number of visitors per year (A/P). Then the land use intensity is classified into three classes based on the ratio A/P.

Table 5e - Land use intensity for recreation areas

Class	Description	A/P visitors ratio	Score
1 2 3	Low Moderate High	<1 1 to 2.5 > 2.5	1 1.5 2

Policy enforcement

The policies related to environmental protection are classified according to the degree to which they are enforced for each case of land use. The information on the existing policies are collected and then the degree of implementation/enforcement is evaluated. Three classes related to the policy on environmental protection are defined (Table 8).

Table 6 - Policy enforcement

Class	Description	Degree of enforcement	Score
1 2 3	Low Moderate High	complete (>75% of the area under protection) partial (25-75% of the area under protection) incomplete (<25% of the area under protection)	1 1.5 2

 

Management quality classes are:

Classes	Ranges
High quality	< 1.25
Medium quality	>= 1.25 <= 1.51
Low quality	> 1.51

5 top

g Computing algorithm

The quantification of different Environmental Sensitivity (ES) levels at the basin scale can be carried out by evaluating the overall influence that single information layers have on the phenomena under study (Ferrara et al. 1999). The goal was to develop a system which would function irrespective of the number and type of information layers at its most primitive level. This is achieved by adopting a two stage approach as illustrated in figure 1. In the first stage, in this case, the four single quality layers are first determined from the basic data layers and in the second phase the final sensitivity of an area is evaluated from the quality layers.

Figure 1 - Scheme of the ESI estimate.

Each elementary unit in each Quality Layer is estimated as the geometric mean of its own sub-layers:

Quality_x _ij = (layer_1 _ij * layer_2 _ij * layer_3 _ij * ...... * layer_n _ij) ^(1/n)                      [1]

where:
i,j = rows and columns of a single elementary land unit of each layer;
n = number of layers used
 
The first level, that of the basic data layers, isolates the rest of the system from the details of the data. The quality layer, level 2, acts as a buffer between the level 1 data layers and the derived ESA layer, level 3. With the four qualities obtained from the above, the ES is estimated by:

ES _ij = (Quality_1 _ij * Quality _2 _ij * Quality _3 _ij * Quality _4 _ij) ^(1/4)                           [2]

where:
i,j = rows and columns of a single elementary land unit of each quality;
Quality_nij = computed values
 
In present Expert System, computed values, as they result from the application of the algorithm and the scores above reported, are also re-ranged from 0 to 100 for simpler interpretation. Algorithms and used values are reported in following table:

Table 7 - Values and algorithms used in ranging data from 0 to 100

Sensitivity scores give an efficient and simply to use estimate of different levels of sensitivity present in a defined area. Computed values are continuous within the selected ranges, usually from 1 to ~ 2 or from 0 to 100 as in the ES Index, but for interpretation and rapresentation puropses it is suggested to group them into classes. Class grouping is an open process in which threshold values are selected depending on the phoenomena that is needed to put in evidence. In following table is reported an example of class grouping of the ESI values, that can be considered generally applicable in mediterranean environments (Ferrara et al. 1999, Kosmas et al. 1999).

Level of sensitivity Type of areas Sensitivity score Sensitivity Index (% of critical factors) Short description   Very low Not affected (N)   >= 1 < 1.170   >= 0 < 15.14 Areas in which critical factors are very low or not present, with a good balance between environmental and socio-economical factors.   Low Potential (P) >= 1.170 <= 1.225 >= 15.14 <= 20.04 Areas threatened by desertification under significant climate change, if a particular combination of land use is implemented or where offsite impacts will produce severe problems. This would also include abandoned land which is not properly managed.   Medium Fragile (F1) Fragile (F2) Fragile (F3) > 1.225 <= 1.265 > 1.265 <= 1.325 > 1.325 <= 1.375 > 20.04 <= 23.60 > 23.60 <= 28.94 > 28.94 <= 33.39 Areas in which any change in the delicate balance between natural and human activity is likely to bring about desertification.   High Critical (C1) Critical (C2) Critical (C3) > 1.375 <= 1.415 > 1.415 <= 1.530     > 1.530 > 33.39 <= 36.95 > 36.95 <= 47.19       > 47.19 Areas already highly degraded through past misuse, presenting a threat to the environment of the surrounding areas or with evident desertification processes.

5 top

g Example of the use of sensitivity scores

One of the most important aspects of the system is that the ES classes are not directly linked to an absolute value of sensitivity but are related indirectly, and relatively, through scores that define different levels of sensitivity, for different parameters, for a particular area. As a result, sensitivity calculated at the top layer imposes a common framework on the components of an area. The elements, which are grouped into broad categories, can be thus investigated and characterised in a different phase by other analyses. For example, the effects of different kinds of intervention on the sensitivity can be estimated by simulating the different intervention options (e.g. recovering the functionality of degraded deciduous forest versus the conversion of conifer stands into more efficient deciduous forests) or it is possible to identify critical factors by applying a simple cluster analysis to areas with high sensitivity values at the municipality level.

The following figures show how the ESI can be used to evaluate and characterise different levels of sensitivity, at the local area scale, in the Agri basin, Italy. Critical factors were identified by establishing the frequency of the layer classes (or scores) over all spatial samples in a municipality, for each layer taken separately. The process was repeated for each municipality. The resulting matrix was subjected to cluster analysis with complete linkage and Euclidean distance methods. Figure 2 illustrates the resulting dendrogram obtained using the municipality data, which gives five Sensitivity typologies inside the basin that corresponds to five defined territorial zones while figure 3 shows graphs of the percentage of different Environmental Sensitivity grades of the four qualities and municipality groups obtained from the cluster analysis.

Figure 2 — Cluster analysis of the Municipalities located in the Agri basin.

Figure 3 — The profiles of Environmental Sensitivity of the municipality groups in the Agri basin. Profiles are expressed in percentages of ES levels in function to the single qualities. (X-axis from 10 to 80; Y-axis from 0 to 75%).

As can be seen, municipality groups 1, 2 and 3 have similar climate (all three are located in the Upper Val d’Agri). Group 1 has critical socio-economic factors and the worst, overall, soil factors. Group 2 has better vegetation qualities but very critical socio-economic factors. Groups 4 and 5 differ, even though they are similar from a geographical perspective: group 5 is characterised by better socio-economical factors and has the worst climatic ones found in the basin, group 4, on the other hand, has worse vegetation conditions.

These examples emphasise the flexibility of the system: it can draw on diverse data yet is efficient; it can be used analyse the environmental system intensively; it enables the current state of, and changes to, the environment to be accurately and quickly assessed; and, by identifying critical factors, it helps to define and identify the most beneficial strategies that could be introduced to reduce the sensitivity of a given area. The use of cross analysis techniques, applied to pre-existing information with other ad hoc data collected as necessary, can also be used to investigate specific degradation or environmental sensitivity phenomena easily and effectively. Furthermore, this approach not only allows the identification of different degrees of environmental sensitivity but, at the same time, allows us to investigate the factors that cause the evolution as they happen.

5 top
 
g Conclusions

Systems which allow us to identify and understand the factors that combine and accelerate land degradation need to be developed in order to adequately manage the land and its resources. The system, outlined here, can be used to isolate current degradation phenomena. To do this, cross-analysis techniques can be applied to the data held in information layers. The information in these layers comes from a variety of sources - some based on pre-existing themes, some based on combinations of these themes, and some created ex-novo from other analyses. It must be emphasised that the main reason for this Environmental Sensitivity Evaluation Model is to define a reference framework to be used in analysing various situations within the Mediterranean Environment under the following operational constraints:

• the system must be reasonably simple to establish, robust in operation, and widely applicable;
• the selection of the information layers is made, not only on the basis of their actual information content (i.e. their relationship with the phenomena under study), but also as a function of our ability to obtain and update the data with ease and economy;
• the system must be adaptable and accommodate the development and refinement of the existing information content and the addition of new information.

5 top

g References

• Basso F., Bove E., Ferrara A., Pisante M., Quaranta, G. 1996. Land degradation and desertification processes in Agri Basin: prevention and management methodologies through use of remote sensing, low environment impact techniques and socio-economic issues. Proceedings International Conference on Mediterranean Desertification. Crete, Greece 29 October - 01 November 1996.
• Basso F., Bellotti A., De Natale F., Ferrara A., Pisante M., 1997. Analisi del rischio di degradazione del suolo in aree agricole della Basilicata: una proposta metodologica. Agronomia XXXI. 3: 864-871.
• Basso F., Bellotti A., Bove E., Faretta S., Ferrara A., Mancino G., Pisante M., Quaranta, G., Taberner M., 1998. Degradation processes in the Agri Basin: evaluating environmental sensitivity to desertification at basin scale. Proceedings International Seminar on 'Indicator for Assessing Desertification in the Mediterranean'. Porto Torres, Italy 18 - 20 September. Edited by G. Enne, M. D'Angelo, C. Zanolla. Supported by ANPA via Brancati 48 - 00144 Roma. pp 131-145
• Basso F., Bellotti A., Faretta S., Ferrara A., Mancino G., Pisante M., Quaranta G., Taberner M., 1999. Application of the proposed methodology for defining ESAs: The Agri Basin In 'The Medalus project Mediterranean desertification and land use. Manual on key indicators of desertification and mapping environmentally sensitive areas to desertification. Edited by: C. Kosmas, M.Kirkby, N.Geeson. European Union 18882. pp:74-79 ISBN 92-828-6349-2
• Basso F., Bove E., Dumontet S., Ferrara A., Pisante M., Quaranta, G., Taberner M., 2000. Evaluating Environmental Sensitivity at the basin scale through the use of Geographic Information Systems and Remote Sensed data: an example covering the Agri basin (southern Italy). Catena 40 : 19-35.
• Briggs D., Giordano A., Cornaert M., Peter D., Maef J., 1992. CORINE soil erosion risk and important land resources in the southern regions of the European Community. EUR 13233. Luxembourg. 97 pp.
• Faretta S., Ferrara A., Mancino G., Taberner M., 2004. Performance Evaluation Procedure of Key Indicators Based Systems. In press.
• Ferrara A., Pisante M., Harrison A.R., Taberner M., 1995. The use of spatial relationship analysis to study the Agri-basin with remotely sensed images. MEDALUS II Final Report. King's College, London. 67-83.
• Ferrara A., Bellotti A., Faretta S., Mancino G., Taberner M., 1999. Identification and assessment of environmentally sensitive areas by Remote Sensing. MEDALUS III 2.6.2. OU Final Report. King's College, London. Volume 2: 397-429
  
• Kosmas C., Moustakas N., Danalatos N.G., Yassoglu N., 1994. The effect of rock fragments on wheat biomass production under highly variable moisture conditions in Mediterranean environments. Catena, 23: 191-198.
• Kosmas C., Ferrara A., Briasouli H., Imeson A. 1999. Methodology for mapping Environmentally Sensitive Areas (ESAs) to Desertification. In 'The Medalus project Mediterranean desertification and land use. Manual on key indicators of desertification and mapping environmentally sensitive areas to desertification. Edited by: C. Kosmas, M.Kirkby, N.Geeson. European Union 18882. pp:31-47 ISBN 92-828-6349-2
• Kosmas C., Ferrara A., Bellotti A, Detsis V., Faretta S., Gerontidis St., Mancino G., Marathainou M. and Pisante M. 1998. A Comparative Analysis of the Physical Environment of two Mediterranean Areas Threatened by Desertification. Istituto Mediterranico, Universitade Nova De Lisboa. Mediterraneo n 12/13: 127-145
• Kosmas C., Poesen J., Briasouli H., 1999. Key indicators of desertification at the ESAa scale. In 'Manual on Key Indicators of desertification and Mapping Environmentally Sensitive Areas to Desertification'. MEDALUS III Project. King's College, London.
• Poesen J., Bunte K., 1996. The effects of rock fragments on desertification processes in Mediterranean environments. In ‘Mediterranean desertification and land use’ edited by J. Brandt & J. Thornes. John Wiley & Sons Ed. London. 247-267.
• Rubio J.L., 1995. Desertification: evolution of a concept. In EUR 15415 "Desertification in a European context: Physical and socio-economic aspects", edited by R. Fantechi, D. Peter, P. Balabanis, J.L. Rubio. Brussels, Luxembourg: Office for Official Publications of the European Communities: pp. 5-13.
• Thornes J.B., 1995. Mediterranean desertification and the vegetation cover. In EUR 15415 - "Desertification in a European context: Physical and socio-economic aspects", edited by R.Fantechi, D.Peter, P.Balabanis, J.L. Rubio. Brussels, Luxembourg: Office for Official Publications of the European Communities. 169-194.
• Yassoglu N., Kollias B., Tatsis E., Filis e., 1995. The use of geographical information systems in the assessment of natural hazards. In EUR 15415 ‘Desertification in a European context: Physical and socio-economic aspects’. Edited by R. Fantechi, D. Peter, P. Balabanis, J.L. Rubio. Brussels, Luxembourg: Office for Official Publications of the European Communities. 617-625.

5 top