ecological indicators 9 (2009) 357–363

available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/ecolind

Case study

Assessing the impact of ecological and economic factors on

land degradation vulnerability through multiway analysis§
Luca Salvati a,b,*, Marco Zitti c
a
Italian National Institute of Statistics (ISTAT), Viale Liegi 13, I-00198 Rome, Italy
b
c/o Department of National Accounting and Analysis of Social Processes, Faculty of Statistics, University of Rome ‘La Sapienza’,
Piazzale A. Moro 5, I-00185 Rome, Italy
c
Consultant of Central Office for Crop Ecology, National Research Council in Agriculture (CRA-UCEA), Via del Caravita 7a,
I-00186 Rome, Italy

article info abstract

Article history: Land degradation (LD) is a global problem which involves climate, soil, vegetation, economic,
Received 26 December 2007 and population conditions. In Mediterranean Europe climatic variability and human pressure
Received in revised form combine to produce soil sealing, erosion, salinisation, fire risk, and landscape fragmentation,
31 March 2008 all regarded as important factors to start LD. The aim of this paper is to introduce a time-series
Accepted 3 April 2008 evaluation of land vulnerability to degradation based on nine ecological and economic
variables. The analysis was carried out over 1970–2000 at the municipality level in Latium
(central Italy), a region which has shown increasing land vulnerability in the last years. A
Keywords: multiway data analysis (MDA) was applied in order to explore the relationship among
Land degradation indicators over the study period. Their importance in determining LD vulnerability was
Indicators estimated through a weighting system based on MDA results. A composite index of land
Multiway analysis vulnerability (LVI) was obtained as the weighted average of the nine variables transformed into
Composite index single indicators, according to their relationship with LD. Considerable increases in LVI were
Weighting system observed in dry coastal and lowland municipalities close to Rome, thus indicating that climate
Mediterranean basin aridity, population growth, and land use changes are important determinants of land vulner-
ability in Latium. LVI was positively correlated to the environmental sensitive area index
(ESAI) measured on the same spatial and time scales, thus suggesting that a sound evaluation
of land vulnerability is possible through LVI score.
# 2008 Elsevier Ltd. All rights reserved.

1. Introduction loss of the biological and economic productivity of irrigated

and non-irrigated agricultural land, pastures, rangeland, and
The term land degradation (LD) is often used to describe an woodlands (Brandt et al., 2003; Tanrivermis, 2003; Salvati et al.,
environmental phenomenon affecting dry lands, sometimes 2008). It results from various factors, including climatic
without a clear understanding of the involved processes (Le dryness, poor soil and vegetation quality, pressure due to
Houerou, 1993; Thornes and Brandt, 1995; Puigdefabregas and agriculture intensification, population growth, urban sprawl,
Mendizabal, 1998). LD usually means reduction or temporary and industrial concentration (Kosmas et al., 2000a; Garcia

§
This paper reflects the ideas and research activities of the authors. Findings, interpretations, and conclusions should not be attributed
to ISTAT or CRA.
* Corresponding author at: Piazza F. Morosini 12, I-00136 Rome, Italy. Fax: +39 06 69 53 12 14.
E-mail addresses: bayes00@yahoo.it (L. Salvati), mzitti@ucea.it (M. Zitti).
1470-160X/$ – see front matter # 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ecolind.2008.04.001
358 ecological indicators 9 (2009) 357–363

Latorre et al., 2001; Salvati and Zitti, 2005). Coastal and lowland approximately 17,065 km2 featuring a complex topography
areas in the Mediterranean basin are generally depicted as and various climatic zones according to elevation (Salvati
vulnerable to LD due to both anthropogenic factors (e.g. et al., 2007). In the last 30 years the study area has been
Loumou et al., 2000; Tanrivermis, 2003; Salvati et al., 2008) and subjected to a number of land use changes due to urban
the impact of climate change (e.g. Sharma, 1998; Incerti et al., growth, crop intensification, forest fires, and tourism con-
2007; Sivakumar, 2007). Although the environmental char- centration. Moreover, climate conditions became drier espe-
acteristics of vulnerable areas are similar to those of degraded cially along the coastal rim, and severe drought episodes
land, some factors (e.g. vegetation, agriculture, irrigation, and occurred more frequently over the whole region.
policy strategies) mitigate this process in the short term
(Thornes and Brandt, 1995; Montanarella, 2007). 2.1. Data and indicators
The complexity of LD represents a limitation for monitor-
ing, modelling, and projection approaches (Rubio and Bochet, An indirect estimation of LD in the Mediterranean basin was
1998; D’Angelo et al., 2000; Feoli et al., 2003). In the developed through the use of indicators describing the impact
Mediterranean basin the assessment of vulnerable land was of different factors on land vulnerability to degradation
conducted mainly through the use of proxy indicators (Puigdefabregas and Mendizabal, 1998; Rubio and Bochet,
depicting climate, soil, and vegetation at an adequate spatial 1998; Basso et al., 2000; Salvati et al., 2008). A number of
scale (Rubio and Bochet, 1998; Feoli et al., 2002; Salvati and indicators is commonly used which especially describes
Zitti, 2008a and references therein). These indicators were including those describing climate, soil characteristics and
usually aggregated into a composite index of land vulner- erosion risk, vegetation quality and plant productivity, fire
ability by standard procedures (e.g. environmental sensitive risk, land fragmentation and management (Kosmas et al.,
area index, ESAI, see Basso et al., 2000). However, time-series 2000a, 2000b; Salvati and Zitti, 2005). However, it should be
analyses focusing trends in both land vulnerability and noted that few indicators are available over a long time, thus
its main drivers are generally lacking in this area (e.g. representing a serious constraint for the objective of this
Montanarella, 2007). An integrated approach based on multi- study. We therefore identified a restricted number of variables
variate time-series analysis may better explore latent patterns which cover the whole national territory and are continuously
and trends of the main factors affecting LD (e.g. Feoli et al., available from the statistical sources: (i) over the last 30 years
2002; Salvati and Zitti, 2008a). Such approaches are mean- at least and (ii) at an adequate geographical scale (i.e.
ingful because they make clear that land vulnerability should municipalities). We believe that variables selected represent
not be treated as something static but as something that an acceptable compromise between accuracy and time/space
changes over time (e.g. Salvati and Zitti, 2008b). Moreover, the resolution (e.g. Yli-Viikari et al., 2007).
choice of relevant indicators, the method used to normalise According to the regional scale of this study three main
the indicators themselves, and the weighting technique have a research themes were identified (Salvati and Zitti, 2008b): (i)
considerable influence on the estimate of land vulnerability climate–soil, (ii) landscape, and (iii) human pressure. The
and need to be further studied (Basso et al., 2000; Salvati and climate–soil dimension was described by three factors (Diodato
Zitti, 2005). and Ceccarelli, 2004): the (i) bioclimatic, (ii) pedologic, and (iii)
The aim of this work is to built-up a regional vulnerability geomorphologic. These factors regard respectively with climate
evaluation model (VEM) able to assess land vulnerability over aridity, available water capacity of the soil, and soil erosion
time by way of a composite index integrating ecological and (Kosmas et al., 2000a; Venezian Scarascia et al., 2006; Incerti
economic indicators of LD vulnerability. The following steps et al., 2007). The landscape dimension includes three variables
were implemented in order to achieve this goal: (i) selecting linked to natural and agriculture land use (e.g. Kosmas et al.,
environmental, economic, and social indicators, and integrat- 2000b; Tanrivermis, 2003; Salvati et al., 2007): crop intensifica-
ing the information associated to those research dimensions; tion, woodland cover, and loss of agricultural surface. The
(ii) determining a weight for each indicator through a multi- impact of human pressure was finally described by three
variate time-series approach; (iii) estimating trends in land variables concerning population density and growth (Salvati
vulnerability from 1970 to 2000 by way of a composite index and Zitti, 2007), as well as concentration of industrial activities
(LVI) aggregating indicators on the basis of their weight. with a potential impact on soil (Salvati and Zitti, 2005; Salvati
VEM was built-up at local scale (i.e. municipalities) in order et al., 2008). Variables used in this work and related data sources
to provide politicians and other stakeholders with a simple were described in Salvati and Zitti (2005, 2007, 2008b) and Salvati
monitoring tool (e.g. Nader et al., 2008). We believe that an et al. (2007, 2008). All the variables were made available at 4
empirical framework like the one introduced here provides years: 1970, 1980, 1990, and 2000. According to Basso et al. (2000)
valuable results stimulating more sophisticated approaches to some variables can be considered static as they change slowly
the problem (e.g. Salvati et al., 2008). and by their nature are infrequently measured. This was the
case for AWC, which was regarded as constant in the following
analyses (Salvati and Zitti, 2008a,b).
2. Materials and methods Variables were then transformed into indicators ranging
from 0 to 1 as follows:
The study area includes the administrative region of Latium,
one of the twenty NUTS-2 Italian regions. In 2000, it includes x0i; j  x0min; j
five provinces (Viterbo, Rieti, Rome, Latina, and Frosinone) and Xt;i; j ¼ (1)
x0max; j  x0min; j
377 municipalities (Salvati and Zitti, 2007). It covers an area of
 ecological indicators 9 (2009) 357–363 359

Table 1 – Indicators entered VEM, research dimensions, and related data sources
Research dimension Indicator (acronym) Unit of measure (relation with LD) Source

Climate and soil Aridity index (IAR) mm/mm () Salvati et al. (2008)
Water capacity of the soil (AWC) mm () Salvati et al. (2008)
Estimated soil erosion rate (ERO) T ha1 y1 (+) Salvati et al. (2008)
Land use/cover Woodland cover (BOS) % () Census of agriculture
Variation in cultivated surface (SAV) % () Census of agriculture
Index of crop intensification (INT) % (+) Salvati et al. (2007)
Human pressure Population density (DEN) People km2 (+) Census of population
Demographic variation (VAR) % (+) Census of population
Industrial concentration (COI) Equivalents (+) Census of industry and services

" #
x0i; j  x0min; j A weight was computed for each indicator i and year of
Xt;i; j ¼ 1  (2) study y by multiplying the contribution of each indicator (Vi,y)
x0max; j  x0min; j
to the m most important (i.e. explaining >10% of total variance)
where x0i; j represents the observed value for the variable i factorial axes (Coppi and Bolasco, 1989) by their proportion of
measured in the municipality j, and x0min; j and x0max; j respec- explained variance (Ck,y). The sum of these products for all the
tively represent the minimum and maximum observed values m selected axes represents the absolute weight (wi,y) attributed
for the variable i measured in the municipalities of Latium. to each indicator (Salvati and Zitti, 2008c):
Eq. (1) was applied to variables showing a positive relationship
with LD, Eq. (2) to variables which showed a negative associa- X
m

tion to LD (Table 1). Indicator’s relation with LD was identified wi;y ¼ ðVi;y Ck;y Þ (3)
k¼1
according to recent literature (Rubio and Bochet, 1998; Kosmas
et al., 2000a; Salvati et al., 2007, 2008 and references therein). Relative weights (Wi,y) were obtained by dividing absolute
Each indicator ranges from 0 (the lowest contribution to land weights by the sum of the indicators’ weights estimated in the
vulnerability) to 1 (the highest contribution to land vulner- year y:
ability).
wi;y
Wi;y ¼ P9 (4)
2.2. Statistical analyses i¼1 wi;y

A multiway data analysis (MDA) was performed in order to The final index (so-called land vulnerability index, LVI) was
depict changes over time (1970–2000) of the indicators entered evaluated in each municipality as the weighted average of the
VEM (Salvati and Zitti, 2008c). Multiway data analysis (Lavit nine indicators (Salvati and Zitti, 2005) according to the
et al., 1994) was applied for the 4 years to the matrix composed equation:
by the nine indicators measured on all the municipalities of
Latium. The general goals of MDA are: (i) to analyse the X
9
0
relationship between different data sets, (ii) to combine them LVI j;y ¼ ðWi;y X0i; j;y Þ (3 )
i¼1
into a common structure called ‘compromise’ which is then
studied via principal component analysis (PCA) to reveal the LVI scores range between 0 (the lowest land vulnerability)
common structure between the variables, and finally, and (iii) and 1 (the highest land vulnerability). Municipalities were
to project each of the original data sets onto the compromise classified into quartile classes representing increasing levels
space to analyse communalities and discrepancies (Coppi and of land vulnerability (e.g. Salvati et al., 2008).
Bolasco, 1989). The structure of the set of indicators was
explored by computing loadings (i.e. the correlation among 2.4. Comparing LVI and a standard index of land
indicators and MDA axes; Salvati and Zitti, 2008c). sensitivity to degradation

2.3. Model building The ability of the proposed model to estimate different
levels of land vulnerability was evaluated by assessing the
Vulnerability to LD was interpreted by means of the nine relationship between LVI and a standard ESA index (ESAI)
indicators operating in association (Basso et al., 2000; Salvati calculated at the same period and geographical scale
and Zitti, 2005). A vulnerability evaluation model was (Basso et al., 2000). Two ESAI maps of Latium referring to
developed at the municipality level on the basis of a GIS 1990 and 2000 were built-up (Salvati and Zitti, 2008a)
procedure (Fig. 1). The spatial unit chosen here allows to according to the procedure stated in Basso et al. (2000).
perform a detailed analysis of changes in land vulnerability at The mean ESAI value was derived for each municipality by a
a scale which is easily interpretable also for not-technical ‘zonal statistic’ procedure carried out by way of GIS tools
users (e.g. policy-makers and other stakeholders acting at the (Salvati and Zitti, 2008b). The relationship between LVI and
local level) and consistent with the spatial resolution of the ESAI scores was assessed in both years by Spearman non-
indicators selected (e.g. Dumanski et al., 1998; Iosifides and parametric rank correlation test. Probability level was set up
Politidis, 2005). to a = 0.05.
360 ecological indicators 9 (2009) 357–363

Fig. 2 – Loading plot of the nine indicators analysed by MDA

(arrows indicate the shift in the position of each indicator
from 1970 to 2000; see Table 1 for abbreviations).

accounts for 14% of the total variance, may be interpreted as a

proxy of soil quality. ERO was strictly associated to positive
values of this axis while AWC was associated to the negative
values.
Arrows indicate over the study period the most important
changes in the position of the indicators as revealed by MDA.
DEN, COI, INT, and IAR showed a consistent shift along the
first axis going towards the positive values. This suggests that
both ecological and economic conditions affecting LD became
worse over the last years. Although associated to the first axis,
VAR and BOS showed different trends compared to the other
indicators. VAR moved along the first axis from positive to
negative values suggesting that the impact of this factor on
land vulnerability decreased in the last years. BOS showed two
distinct behaviours over the periods 1970–1990 and 1990–2000.
In the former period it moved towards the positive values,
while showing the opposite behaviour in the latter period.
Other important changes include SAV movements along the
second axis indicating a progressive reduction of agricultural
land consumption over time.

3.2. Weighting system

The relative weights attributed to each indicator were reported

in Table 2. On average, the weight attributed to climate–soil
theme (CLI) amounted to 0.48, those attributed to human

Table 2 – Relative weights attributed to each indicator

Fig. 1 – Flow chart of the procedure estimating LVI. entered VEM by year
Indicator 1970 1980 1990 2000 Average

IAR 0.13 0.12 0.10 0.12 0.12

3. Results AWC 0.19 0.19 0.19 0.19 0.19
ERO 0.18 0.17 0.17 0.18 0.17
CLI 0.50 0.48 0.47 0.49 0.48
3.1. Multiway analysis DEN 0.13 0.14 0.14 0.14 0.14
VAR 0.12 0.07 0.06 0.06 0.08
The results of MDA carried out on the entire dataset were COI 0.11 0.13 0.14 0.12 0.13
synthesised in Fig. 2, where the loadings of each indicator with POP 0.37 0.34 0.34 0.33 0.35
the two principal factorial axes were plotted. Factor 1, INT 0.08 0.09 0.11 0.12 0.10
explaining 23% of the total variance, represents a gradient BOS 0.04 0.07 0.08 0.06 0.06
SAV 0.01 0.01 0.00 0.00 0.01
depicting human pressure, with demographic and landscape
LAN 0.13 0.17 0.19 0.18 0.17
indicators positively associated to the axis. Factor 2, which
 ecological indicators 9 (2009) 357–363 361

pressure (POP) and landscape (LAN) themes were respectively whereas BOS showed a complex pattern increasing in 1970–
0.35 and 0.17. CLI weights were rather stable over time, 1990 and decreasing during the next 10 years.
whereas POP weights decreased from 0.37 to 0.33. LAN showed
the reverse pattern increasing from 0.13 to 0.18. Considering 3.3. Land vulnerability index
the single indicators, DEN, COI, and INT weights increased
from 1970 to 2000. VAR decreased over the same period, Fig. 3 reported maps of LVI estimated at the municipality level
in 2000 and as a difference between 1970 and 2000. Highest LVI
scores were found in coastal areas close to Rome, in farmlands
south of Rome featuring marked population growth and urban
sprawl, and in lowlands north of Rome featuring crop
intensification and climate dryness. From 1970 to 2000 the
major changes in LVI were observed in both urban munici-
palities close to Rome and coastal areas north of Rome. These
results highlight the importance of factors such as climate
aridity, population growth, and land use changes in determin-
ing land vulnerability at the local level.

3.4. Comparing LVI and ESAI

The relationship between LVI and ESAI scores estimated in

both 1990 and 2000 was significant (1990: rs = 0.53, p < 0.001;
2000: rs = 0.62, p < 0.001, both n = 377). These results should be
not regarded as allometric relationship but are able to
illustrate that, in general, LVI is producing vulnerability
estimates consistent to those obtained by way of ESA
procedure. It was already demonstrated (Basso et al., 2000)
that ESAI scores were significantly correlated to some
indicators of soil degradation.

4. Discussion

The establishment of a system of information is vital in order

to develop a dynamic assessment of natural resource
depletion (Yli-Viikari et al., 2007). Indicators are meaningful
when they improve knowledge of a multidimensional phe-
nomenon such as LD (Rubio and Bochet, 1998). This study
introduces a possible integration of some geo-physical and
socio-economic indicators of LD in order to obtain a sound
evaluation of land vulnerability. Economic, demographic, and
social indicators were rarely considered in LD assessment,
partly because of their very nature, and partly because the
information gleaned about them is of a qualitative type or is
available at a low-resolution scale (e.g. Basso et al., 2000).
Since the indicators selected for environmental evaluation
are different in their characters, this paper suggests a
procedure to enable comparisons to be made. Adopting both
(i) a balanced approach when selecting indicators (rather than
choosing indicators of a narrowly focused range of para-
meters) and (ii) an objective procedure in the weighting
process are adequate tools to achieve a more precise
evaluation of LD vulnerability (e.g. Niemeijer, 2002).
Although MDA attributed the highest weights to climate
and soil factors, it was suggested that also socio-economic
factors may have an important role in determining land
Fig. 3 – Estimating land vulnerability to LD in Latium: vulnerability (Blaikie and Brookfield, 1987). This study identi-
municipalities by elevation zones (a); distribution of LVI fied population density as a factor considerably affecting land
scores by quartiles in 2000 (b), and as a difference between vulnerability and has shown that its importance increased
1970 and 2000 (c). Arrows indicate increasing over the study period. In fact, a marked population growth has
vulnerability; stars indicate the municipality of Rome. been taking place during the last 30 years in many areas of
362 ecological indicators 9 (2009) 357–363

Latium (Salvati and Zitti, 2007) thus widening the demo- Based on these tools, an integrated evaluation of ecological
graphic divergence between Rome and the rest of the region and economic aspects of LD may be carried out thought a
(Salvati and Zitti, 2008b). Population growth has a direct multi-temporal approach, taking into account the structural
consequence in the soil sealing by human expansion into changes occurred in the societies and the economies of
productive lands (Brandt et al., 2003). Loss of agricultural and southern Europe during the last 50 years (Salvati and Zitti,
semi-natural land, degradation of high-quality soils, salinisa- 2008b). Considering a wider number of human variables and
tion of groundwater, increase of fire risk, and landscape weighting them according to their impact on LD (Montanar-
fragmentation are documented with a marked relationship ella, 2007) may therefore shade new light in the assessment of
with urbanisation (Garcia Latorre et al., 2001; Tanrivermis, vulnerable areas (Rubio and Bochet, 1998). This also indicates
2003; Salvati and Zitti, 2007; Salvati et al., 2007). possible directions taken by the degradation process, as
Among landscape indicators, the intensification of agri- human pressure and social changes generally act more rapidly
culture has shown the higher weights over the study period. It than variations in climate regimes and soil quality.
poses a question about the long-term sustainability of this
human activity in terms of LD (Tanrivermis, 2003). Crop
intensity, growth of mechanisation (with a consequent risk of references
soil compacting), and low-efficiency irrigation methods con-
tribute to the overexploitation of the water resources
especially in drier areas of Latium (Salvati et al., 2007). On Basso, F., Bove, E., Dumontet, S., Ferrara, A., Pisante, M.,
the other hand, the impact of deforestation (which has Quaranta, G., Taberner, M., 2000. Evaluating environmental
reached in lowlands the high rates in late eighties) on LD sensitivity at the basin scale through the use of geographic
vulnerability is now reducing (Salvati and Zitti, 2008c) due to information systems and remotely sensed data: an example
covering the Agri basin - Southern Italy. Catena 40, 19–35.
the general forestation trend involving part of hilly and
Blaikie, P., Brookfield, H.C., 1987. Land Degradation and Society.
mountain zones of Latium, as widely observed in other areas Methuen, London.
of southern Europe (e.g. Bouma et al., 1998). Finally, the Bouma, J., Varallyay, G., Batjes, N.H., 1998. Principal land use
abandonment of marginal lands represented a factor con- changes anticipated in Europe. Agric. Ecosyst. Environ. 67,
tributing to land vulnerability over the seventies and the 103–119.
eighties (D’Angelo et al., 2000; Kosmas et al., 2000b; Tanri- Brandt, J., Geeson, N., Imeson, A., 2003. A desertification
vermis, 2003). In Latium it was prevalently due to industria- indicator system for Mediterranean Europe. DESERTLINKS
Project, Bruxelles (www.kcl.ac.uk/desertlinks).
lisation processes, the increase in the cost of cultivation, and
Coppi, R., Bolasco, S., 1989. Multiway Data Analysis. North
the decrease of agriculture profits (Salvati et al., 2007). Actually Holland, Amsterdam.
land abandonment is reducing and appears as a possible cause D’Angelo, M., Enne, G., Madrau, S., Percich, L., Previtali, F.,
of LD only in inland, marginal areas, especially south of Rome Pulina, G., Zucca, C., 2000. Mitigating land degradation in
(Salvati and Zitti, 2008a). Here lands on sloping terrain are Mediterranean agro-silvo-pastoral systems: a GIS-based
vulnerable because of soil erosion accelerates when they are approach. Catena 40, 37–49.
Diodato, N., Ceccarelli, M., 2004. Multivariate indicator Kriging
abandoned (e.g. Kosmas et al., 2000a; Zalidis et al., 2002;
approach using a GIS to classify soil degradation for
Salvati and Zitti, 2005).
Mediterranean agricultural lands. Ecol. Indicat. 4, 177–187.
Dumanski, J., Pettapiece, W.W., McGregor, R.J., 1998. Relevance
of scale dependent approaches for integrating biophysical
5. Conclusion and socio-economic information and development of
agroecological indicators. Nutr. Cycl. Agroecos. 50, 13–22.
Land vulnerability to degradation, environmental quality Feoli, E., Gallizia Vuerich, L., Zerihun, W., 2002. Evaluation of
envoironmental degradation in northern Etiopia using GIS to
and management are all dynamic entities. Developing a
integrate vegetation, geomorphological, erosion, and socio-
decision support system like the one proposed in this study economic factors. Agric. Ecosyst. Environ. 91 pp. 313–225.
appears as a promising tool to define trends and predict Feoli, E., Giacomich, P., Mignozzi, K., Ozturk, M., Scimone, M.,
changes in land vulnerability and to promote efficient 2003. Monitoring desertification risk with an index
management of LD (Rubio and Bochet, 1998; Basso et al., integrating climatic and remotely-sensed data: an example
2000; D’Angelo et al., 2000). from the coastal area of Turkey. Manage. Environ. Qual. 14,
To identify ecological and socio-economic indicators as 10–21.
Garcia Latorre, J., Garcia-Latorre, J., Sanchez-Picon, A., 2001.
reliable inputs for decision making and implementation of
Dealing with aridity: socio-economic structures and
strategies to mitigate LD is necessary to establish links of environmental changes in an arid Mediterranean region.
causality within these processes (Dumanski et al., 1998; Zalidis Land Use Policy 18, 53–64.
et al., 2002). The main actions useful in this assessment Incerti, G., Feoli, E., Giovacchini, A., Salvati, L., Brunetti, A., 2007.
include: (i) the definition of statistical methods to explore the Analysis of bioclimatic time series and their neural
impact of different drivers on LD vulnerability (Salvati et al., network-based classification to characterize drought risk
patterns in south Italy. Int. J. Biometeorol. 51, 253–263.
2008); (ii) the set up of metadata collection and data ware-
Iosifides, T., Politidis, T., 2005. Socio-economic dynamics, local
houses including an ensemble of multivariate tools able to
development and desertification in western Lesvos, Greece.
analyse complex datasets (e.g. Stassopoulou et al., 1996); (iii) Local Environ. 10, 487–499.
the identification of adequate geographical scales to recognise Kosmas, C., Danalatos, N.G., Gerontidis, S., 2000a. The effect of
social, demographic, and economic indicators (see Dumanski land parameters on vegetation performance and degree of
et al., 1998 for a discussion). erosion under Mediterranean conditions. Catena 40, 3–17.
 ecological indicators 9 (2009) 357–363 363

Kosmas, C., Gerontidis, S., Marathianou, M., 2000b. The Salvati, L., Zitti, M., 2008c. Exploring rural land cover changes
effect of land use change on soil and vegetation over through multiway analysis: a case study in Latium, central
various lithological formations on Lesvos. Catena 40, Italy. Biota, in press.
51–68. Salvati, L., Zitti, M., 2005. Land degradation in the
Lavit, C., Escoufier, Y., Sabatier, R., Triassac, P., 1994. The ACT Mediterranean basin: linking bio-physical and economic
(STATIS) method. Comp. Stat. Data Anal. 18, 97–119. factors into an ecological perspective. Biota 5, 67–77.
Le Houerou, H.N., 1993. Land degradation in Mediterranean Salvati, L., Zitti, M., Ceccarelli, T., 2008. Integrating economic
Europe: can agroforestry be a part of the solution? A and environmental indicators in the assessment of
prospective review. Agroforest. Syst. 21, 43–61. desertification risk: a case study. Appl. Ecol. Environ. Res. 6,
Loumou, A., Giourga, C., Dimitrakopoulos, P., Koukoulas, S., 129–138.
2000. Tourism contribution to agro-ecosystems Salvati, L., Macculi, F., Toscano, S., Zitti, M., 2007. Comparing
conservation; the case of Lesbos island, Greece. Environ. indicators of intensive agriculture from different statistical
Manage. 26, 363–370. sources. Biota 8, 51–59.
Montanarella, L., 2007. Trends in land degradation in Europe. In: Sivakumar, M.V.K., 2007. Interactions between climate and
Sivakumar, M.V., N’diangui, N. (Eds.), Climate and Land desertification. Agr. Forest Meteorol. 142, 143–155.
Degradation. Springer, Berlin. Sharma, K.D., 1998. The hydrological indicators of
Nader, M.R., Salloum, B.A., Karam, N., 2008. Environment desertification. J. Arid Environ. 39, 121–132.
and sustainable development indicators in Lebanon: a Stassopoulou, A., Petrou, M., Kittler, J., 1996. Bayesian and
practical municipal level approach. Ecol. Indicat. 8, neural networks for geographic information processing.
771–777. Patt. Rec. Lett. 17, 1325–1330.
Niemeijer, D., 2002. Developing indicators for environmental Tanrivermis, H., 2003. Agricultural land use change and
policy: data-driven and theory-driven approaches sustainable use of land resources in the Mediterranean
examined by example. Environ. Sci. Policy 5 (2), 91–103. region of Turkey. J. Arid Environ. 54, 553–564.
Puigdefabregas, J., Mendizabal, T., 1998. Perspectives on Thornes, J.B., Brandt, J., 1995. Mediterranean Desertification and
desertification: western Mediterranean. J. Arid Environ. 39, Land Use. John Wiley & Sons, Chichester.
209–224. Venezian Scarascia, M.E., Di Battista, F., Salvati, L., 2006. Water
Rubio, J.L., Bochet, E., 1998. Desertification indicators as resources in Italy: availability and agricultural uses. Irrig.
diagnosis criteria for desertification risk assessment in Drain. 55, 115–127.
Europe. J. Arid Environ. 39, 113–120. Yli-Viikari, A., Hietala-Koivu, R., Huusela-Veistola, E., Hyvonen,
Salvati, L., Zitti, M., 2007. Long term demographic dynamics T., Perala, P., Turtola, E., 2007. Evaluating agri-
along an urban–rural gradient: implications for land environmental indicators (AEIs)—use and limitations of
degradation. Biota 8, 61–69. international indicators at national level. Ecol. Indicat. 7,
Salvati, L, Zitti, M., 2008a. Regional convergence of 150–163.
environmental variables: empirical evidences from land Zalidis, G., Stamatiadis, S., Takavakoglou, V., Eskridge, K.,
degradation. Ecol Econ., in press. Misopolinos, N., 2002. Impacts of agricultural practices on
Salvati L., Zitti M., 2008b. Territorial disparities, natural resource soil and water quality in the Mediterranean region and
distribution, and land degradation: a case study in southern proposed assessment methodology. Agric. Ecosyst. Environ.
Europe. Geojournal., in press. 88, 137–146.