Applied Geography 62 (2015) 347e356
Contents lists available at ScienceDirect
Applied Geography journal homepage: www.elsevier.com/locate/apgeog
Common features and different trajectories of land cover changes in six Western Mediterranean urban regions ry e, E. Marraccini a, g, *, 1, M. Debolini b, 1, M. Moulery c, P. Abrantes d, A. Bouchier f, J.-P. Che E. Sanz Sanz c, T. Sabbatini g, C. Napoleone c a
UP 2012-10-103 PICAR-T, Institut Polytechnique LaSalle Beauvais, France UMR 1114 INRA-UAPV EMMAH, Avignon, France Inra, UR Ecodeveloppement, Avignon, France d ficos, IGOT, Universidade de Lisboa, Lisbon, Portugal Centro de Estudos Geogra e AgroParisTech, UMR TETIS, Montpellier, France f INRA, UMR Innovation, Montpellier, France g Institute of Life Sciences, Scuola Superiore Sant'Anna, Pisa, Italy b c
a r t i c l e i n f o
a b s t r a c t
Article history: Available online 7 June 2015
Land use and land cover change (LULCC) dynamics have been particularly strong in the Mediterranean region, due to its historical development and to agro-pedoclimatic conditions favorable to human settlement. This area has undergone in the 1950s and the 1980s intense urbanization processes that has followed different trajectories. Urban expansion commonly occurs at the expense of agricultural land, leading to the fragmentation of natural areas and conflicts over access to land resources. These dynamics mainly concern the fringe between urban and agricultural land, e.g. the peri-urban areas usually included within functional urban regions. Here, to identify common features of LULCC in Western Mediterranean urban regions, we investigated two main features: direct changes due to urbanization and indirect changes affecting non-artificial land uses. We compared LULCC dynamics in 6 case studies from the north and south of the Western Mediterranean region: the urban regions of Montpellier and Avignon (France), Pisa (Italy), Madrid (Spain), Meknes (Morocco), and Constantine (Algeria), using a 30-year multitemporal spatial analysis (1980e2010). Two series of Landsat TM images were acquired for each case study and land cover data were analyzed both for dynamics and for land patterns, using landscape and class metrics. We found no significant north-south differences in LULCC dynamics between the investigated Western Mediterranean urban regions. Differences are more pronounced between smallemedium cities and large metropolitan areas in type of urban diffusion, which is more sprawled in smallemedium cities and more compact in large metropolitan areas. Rather, differences occur in LULCC not directly affected by urbanization, since in Northern Mediterranean urban regions afforestation and abandonment of agricultural areas are prevalent and closer to the urban areas, whereas transformation of natural areas into agricultural ones occurs mainly in Southern Mediterranean urban regions at a similar distance from urban areas than it happens for afforested or abandoned areas. In attempting for the first time to assess LULCC in these Mediterranean urban regions, we provide a preliminary comprehensive analysis that can contribute to the active LULCC research in the Mediterranean basin and that can be easily applied to other Mediterranean urban regions. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Land cover changes Peri-urban areas Land patterns Semi-supervized classification Regional studies
Introduction
* Corresponding author. UP 2012-10-103 PICAR-T, Institut Polytechnique LaSalle Beauvais, 19, rue Pierre Waguet, 60026 Beauvais, France. E-mail address:
[email protected] (E. Marraccini). 1 These authors equally contributed to the manuscript. http://dx.doi.org/10.1016/j.apgeog.2015.05.004 0143-6228/© 2015 Elsevier Ltd. All rights reserved.
Land use and land cover changes (hereafter LULCC) are among the main indicators of environmental and socio-economic global changes (Lambin & Meyfroidt, 2011). Moreover, current LULCC can have an impact on the supply of ecosystem services (Foley et al., 2005; Schroter et al., 2005; Tayyebi, Pijanowski, & Pekin, 2015). Existing and possible future LULCCs are usually analyzed through
348
E. Marraccini et al. / Applied Geography 62 (2015) 347e356
modeling approaches developed under the framework of the land change science (Turner, Lambin, & Reenberg, 2007) to understand the relations between observed dynamics and their contingent drivers (e.g. Tayyebi et al., 2015; Verburg, Veldkamp, & Bouma, 1999; Verburg et al., 2002). These approaches can improve our knowledge on the factors driving changes and subsequently on the possible policy measures to manage these changes (Lesschen, Kok, Verburg, & Cammeraat, 2007). LULCC dynamics in the Mediterranean region have been particularly strong, due to its historical development and to its agro-pedoclimatic conditions, favorable to human settlement (Bouma, Varallyay, & Batjes, 1998; Leontidou, 1993). Recently, this area has undergone an intense urbanization process. While the trajectories of the process differ in the different countries, they have all led to a common result: an urban population exceeding the rural population since the ‘60s with, in 2011, two thirds of Mediterranean inhabitants already living in urban areas (Houpin, 2011)’. Population growth (1950e2000) has been concentrated and fastest on the southern side, with an annual growth rate of þ3.6% instead of a þ1.2% in the Northern-West Mediterranean (Zlotnik, 2003). Extensive urban growth, in terms of sprawl, is experienced on the northern side due to housing policies that use more space per capita (Catalan, Saur, & Serra, 2008; Kasanko et al. 2006; Salvati, Sateriano, & Bajocco, 2013), whereas in the southern, the growth is more compact or addressed to the creation of new settlements and slums (UN-Habitat, 2012). Cities expand mainly at the expense of agricultural land, leading to the fragmentation of natural areas and conflicts over access to land resources (Brueckner, 2000; Irwin & Bockstael, 2007; Jongman, 2002; Zavala & Burkey, 1997). Existing studies on LULCC in the Mediterranean environment have highlighted differences in observed dynamics. On the one hand, abandonment of marginal and low-intensity agricultural areas leads to re-naturalization and to increased forest cover (Gellrich & Zimmermann, 2007; Millington, Perry, & Romero-Calcerrada, 2007; Romero-Calcerrada & Perry, 2004; San Roman Sanz et al., 2013). On the other hand, coastal and plain areas undergo opposing processes, with agricultural intensification in the more favorable zones (Nainggolan et al., 2012) and a simultaneous increase in urbanized areas (Aguilera-Benavente, Valenzuela, & Botequilha~o, 2011; Vimal, Geniaux, Pluvinet, Napoleone, & Lepart, Leita 2012). In many cases, the urbanization process is non-continuous, resulting in the fragmentation of neighboring agricultural and natural areas (Basnou et al., 2013; Salvati, Munafo, Morelli, & Sabbi, 2012). These dynamics are concentrated at the interface between urban and agricultural lands, usually identified as peri-urban areas (Morin, 1991; Mougeot, 2000; Serra, Vera, Tulla, & Salvati, 2014). In these areas, agricultural systems are complex because of their geographical and productive characteristics (Moustier & Fall, 2004; van Veenhuizen, 2006), as well as the multiplicity and diversity of stakeholders acting there (Allen, 2003; Dossa, Abdulkadir, Amadou, Sangare, & Schlecht, 2011). At the same time, these systems perform relevant functions for the cities, providing agroenvironmental and feeding services (De Bon, Parrot, & Moustier, 2010; Zasada, 2011), in addition to leisure and open space facilities (Torreggiani, Dall'Ara, & Tassinari, 2012; Vidal & Fleury, 2009). For this reason, case studies were developed in recent years around the Mediterranean basin to identify the main landscape dynamics at work in the peri-urban areas. Studies analyzed LULCC in cities with particular socio-economic and urban characteristics, mainly large metropolitan areas like Athens (Chorianopoulos, Pagonis, Koukoulas, & Drymoniti, 2010), Rome (Salvati, 2013; Salvati et al., 2012), Istambul (Çakir et al., 2008), Madrid (Navarro, 2012) and Barcelona (Basnou et al., 2013; Catalan et al., 2008), where urban sprawl and the consequent fragmentation of natural and
agricultural areas were assessed as the most pronounced. Studies on other smaller urban regions are less common (e.g. AguileraBenavente et al., 2011 in the Grenada urban region or Hepcan, € Turan, & Ozkan, 2011 in Turkey) even though medium-sized and small cities are considered an important pattern in Mediterranean area development (Benoit & Comeau, 2005). Kasanko et al., (2006) carried out a comparative study in 15 cities around Europe, analyzing the relationship between urban development and population increase and seeking to classify the different case studies according to their land pattern change trajectories. In particular, they revealed that in all the cities analyzed, urban development increased well beyond the population's housing needs. They also grouped together the Southern European cities considered (Palermo, Milan, Bilbao and Porto) which showed similar urban characteristics, i.e. compact and densely populated areas facing rapid, non-compact urban expansion. Few studies have been carried out in Southern Mediterranean countries, where urban pressure appears similar to or greater than in northern countries (Weber & Puissant, 2003). Studies have tended to focus on Sub-Saharan metropolitan areas, exploring development and poverty reduction issues (Raynaut, 2001; World Bank, 2005). Moreover, most of the existing studies consider different geopolitical areas of influence on the northern and southern sides of the Mediterranean, rather than taking the regions as a whole. In this paper, we compare LULCC in 6 European and African case studies in the Western Mediterranean: the metropolitan and urban regions of Montpellier and Avignon (France), the urban region of Pisa (Italy), the metropolitan region of Madrid (Spain), the urban region of Meknes (Morocco) and the metropolitan region of Constantine (Algeria). Performing multitemporal spatial analysis (the last three decades), we searched for features common to the Western Mediterranean urban regions, to answer three questions specifically related to the six case studies. First, are there common features directly linked to urban growth (i.e. changes from other land covers to an artificial/urban land cover)? Second, are there common features indirectly linked to urban growth (e.g. land cover changes not directly involving artificial/urban land cover, without any assumption about the drivers of such changes, e.g. urbanization)? Finally, are there differences in the features of the northern and southern urban regions of the Western Mediterranean? We start here from the hypothesis that both the peri-urban landscape configuration of medium or large cities around the Mediterranean region and their spatial dynamics are heterogeneous. However, we can assume that some common patterns can be identified, for instance the location of new urbanization areas: both on the northern and the southern sides of the Mediterranean, people tend to want to remain accessible to services or jobs (usually concentrated in the city-center and its surroundings), the largest slopes are generally avoided, and a game competition assigns land use to the most profitable activity. However, other spatial patterns can show different dynamics, such as the position of agricultural and natural areas around the cities. The six case studies have been chosen in the context of an international research program aimed to analyze the integration of agriculture within Mediterranean urban systems. These case studies have been selected to be illustrative of different geographic and socio-economic contexts, e.g. large and medium-sized cities, coastal and inland areas, Northern and Southern side of the Mediterranean basin. In the following, we describe the methods uniformly applied to all six case studies to produce land use maps and assess landscape dynamics. Then, we present our findings on both urban and nonurban dynamics (either agricultural abandonment and renaturalization or de-forestation and new agricultural land creation).
E. Marraccini et al. / Applied Geography 62 (2015) 347e356
Materials and methods Six urban regions in the Western Mediterranean We analyzed the land cover dynamics of six urban regions in the Western Mediterranean, four in the European part and two in the North African part (Fig. 1). Following Forman (2008), an urban
349
region has been here defined as a region recognized by local planning policies as urban, i.e. an urban inter-municipality or a metropolitan region. In the case of metropolitan areas (Fig. 2), only a part of the urban center and surrounding region was included in the case study in order to ensure a gradient from the urban center to the more rural areas. In every case, the urban regions analyzed are composed by groups of municipalities. The cities studied were
Fig. 1. Location, urban areas and topographical characteristics of the six case studies.
350
E. Marraccini et al. / Applied Geography 62 (2015) 347e356
Fig. 2. Classification of the case studies.
Avignon and Montpellier (France), Constantine (Algeria), Madrid (Spain), Meknes (Morocco) and Pisa (Italy). Table 1 presents the main characteristics of the corresponding urban regions. Despite the diversity of their location, size and number of inhabitants, the case studies present common features typical of the Mediterranean region, such as heterogeneous morphological conditions, diversity of crops and heterogeneity of farming and agricultural systems (Grasso & Feola, 2012; Ortiz-Miranda, MoraguesFaus, & Arnalte-Alegre, 2013). In terms of number of inhabitants and population growth over the last decade, the six cities differ in size. Pisa and Avignon have low population growth and a low ratio between city and urban region inhabitants, while Madrid is a large metropolitan area with high population growth and a predominance of city inhabitants (Fig. 2). These are typical of the differences between small/medium cities and metropolitan areas: population growth in the former tends to occupy surrounding areas, whereas metropolitan areas population concentration in the urban center coexists with scattered patterns in the enlarged periphery (Abrantes, Pimentel, & Tenedorio, 2010) Database LULCC for the six case studies was first characterized through remote sensing. To ensure homogeneity of data sources and
therefore comparable data, the following data sources, time interval and thematic resolution for the satellite images were selected. We decided to use Landsat data, since they have been recorded worldwide for a long time (since the seventies), are free of charge and are suitable for regional analysis (Wulder et al., 2008). A simple, economical and comparable data source was needed for the six case studies, in line with the objectives of the study. However, we were well aware of the limited resolution of Landsat images and the consequences for land use maps generated there from (AguileraBenavente, Botequilha-Leit~ ao, & Diaz-Varela, 2014). For the European countries, information from Corine Land Cover (EEA, 2006) was also available as a complementary data source. The Landsat images were selected independently for each case study, on the following basis: avoiding poor-quality or low-resolution images, opting for spring/summer images showing a wide range of vegetation, using images encompassing the whole region. In addition, already pre-processed images and other ancillary data were used to test reliability, e.g. Google Earth photos, other Landsat images at a different period of the same year as spring and autumn ones, expert knowledge. The same time interval was used for all six case studies (roughly 1980e2010), in order to highlights changes that have occurred since the latest peak of urban sprawl in Mediterranean basin (Catalan et al., 2008). Table 2 illustrates the main characteristics of the Landsat images selected for each case study.
Table 1 Main characteristics of the case studies. Case study
Location
Urban region Population dynamics Main city Ratio main city/urban Main agricultural land use surface inhabitants region inhabitants
Avignon (France) Constantine (Algeria) Madrid (Spain) Meknes (Morocco) Montpellier (France) Pisa (Italy)
Alluvial inland plain
386 km2
þ13% (1990e2010)
89,683
45%
Inland plain and hilly areas
485 km2
þ8% (1998e2008)
438,164
56%
Hilly areas and edge of plateau (600 m) Inland plain and edge of plateau Coastal plain and hilly area
340 km2
þ125% (1987e2011) 3265,038
95%
2
590 km
þ30% (1990e2010)
469,169
78%
282 km2
þ43% (1982e2010)
257,351
72%
Coastal alluvial plain and hilly 500 km2 (below 900 m) areas
þ10% (1980e2010)
86,236
45%
Arable crops (cereal, vegetables), permanent crops (vineyards and fruit groves) Arable crops (winter wheat) Arable crops (winter wheat, maize, barley, forage, vegetables), Permanent crops (vineyards, olive and fruit groves), arable crops (winter wheat and vegetables) Permanent crops (vineyards, olive groves), arable crops (winter wheat, vegetables) Arable crops (winter wheat, maize), Permanent crops (olive groves)
E. Marraccini et al. / Applied Geography 62 (2015) 347e356 Table 2 Main characteristics of the acquired remote sensing images. Case study
Landsat images (T0, T1)
Temporal interval (years)
Avignon (France) Constantine (Algeria) Madrid (Spain) Meknes (Morrocco) Montpellier (France) Pisa (Italy)
June 1987, June 2011 June 1987, June 2011 August 1984, August 2009 June 1987, June 2011 July 1989, July 2011 June 1985, August 2011
24 24 25 24 22 26
General methodology The methodology involved four steps. First, images were classified using a semi-supervized classification by the software MultiSpect® (Landgrebe & Biehl, 2011). Then, this classification was validated and corrected through field observation or secondary data, such as existing land use maps or aerial photos. Second, the main LULCCs were quantified. Third, spatial analysis was performed to determine the location of the main changes. Finally, an analysis of the land cover class metrics enabled us to assess the main land cover configurations.
Semi-supervized classification and land use change analysis All images were first classified using the software MultiSpect®, which performs a pixel-based classification. Because of the low resolution of the Landsat images (30 m), we decided that our objectives required a low thematic resolution analysis. Five land cover classes were assessed: agriculture (which included permanent and annual crops), natural vegetation and woods, urban areas, bare soil including natural (e.g. sand, rock areas) and artificial bare soils (e.g. quarries) and water (e.g. sea, rivers, and lakes). These classes have been chosen because they represents the main type of land cover in the study areas and they coherent with the main land cover standard classification already existing (e.g. Corine Land Cover, USGS). All the land cover classes were evaluated on the basis of the internal kappa indices, obtaining an indication of the overall accuracy of the maps through the confusion matrix. In particular, the overall accuracy of the maps was calculated through at least 30 control points in the study areas. For maps with a kappa lower than 0.7, we corrected the classification obtained. More control points were added
351
on the basis of secondary data sources (Google Earth, aerial photos and existing maps) and a post-processed and corrected map was produced. Once the two land cover maps for each case study were obtained, we evaluated the land use changes through a transition matrix analysis. According to the results of the transition matrix, we calculated for each case study the relative growth rate over the time interval considered as (UrbT1-UrbT0)/UrbT0 where UrbT1 are the total hectares of urban areas at the time 1 and UrbT0 the total hectares at time 0 (see Table 2). Moreover, to allow comparisons among the case studies, we calculated the annual rate of urban 1=Tn UrbT1 €ck et al. (2014) as growth following Taubenbo 1 UrbT0 where Tn is the time interval between the two land cover databases. Spatial patterns analysis To shed light on the different processes leading to the observed land cover changes, under our assumptions of the competition between spaces and the influence of distance from urban areas, we carried out an analysis of the distance patterns in the main land cover dynamics. We identified two main land cover changes: 1) from agricultural to natural areas (afforestation, abandonment) and 2) from natural to agricultural areas (new agricultural areas), and analyzed the distributions of these changing areas’ distance from urban land cover. We were thereby testing the hypothesis that proximity to urban areas is a factor influencing agricultural abandonment, as indicated by land cover change from agricultural to natural vegetation. Conversely, new agricultural areas are sometimes created at great distances from the urban center, and these dynamics may be indicated by land cover change from natural or woody vegetation to agriculture. In particular, these dynamics were assessed calculating the distance between each pixel passing from agricultural to natural land use and the nearest urban area (Fig. 4), or the distance between each pixel passing from natural to agricultural land use and the nearest urban area (Fig. 5). We calculated some basic statistics (number of pixel changed, average distance values and standard deviation) both for each case study and for the grouped Northern and Southern Mediterranean urban regions. Then, we calculate the frequencies of the distance distributions. The statistical significance of the average values among the two
Fig. 3. Extent of urban areas in 2011 and their origin in the other land cover classes.
352
E. Marraccini et al. / Applied Geography 62 (2015) 347e356
Fig. 4. Frequency distribution of distances from the nearest urban area for the transition from agricultural land cover to natural/woody land cover (abandonment, afforestation).
dynamics was tested using a KruskaleWallis test performed with the R-stats package (kruskal.test function). This analysis carried out in the six case studies was also expected to reveal potential differences in land use change patterns between the north and the south of the Mediterranean.
aggregation index (AI) considers the landscape aggregation for each land use class. These indicators were calculated for the six case studies and the results were compared.
Results Land cover metrics Starting from the remote sensing classification, we performed a landscape pattern analysis on Fragstat (McGarigal & Marks, 1994) based on the most commonly used landscape metrics as per the existing literature (e.g. Bailey et al., 2007; Hepcan, 2012; Lausch & Herzog, 2002). Seven indicators were selected: (1) patch density (PD) represents the number of patches in a fixed area and gives an indication of the landscape fragmentation; (2) edge density (ED) is a measure of the contour complexity for the different patches, corresponding to the ratio between the perimeter of each land use class and the total landscape surface; (3) largest patch index (LPI) is the ratio between the largest patch for each class and the total landscape surface, and gives an indication of the class dominance on the landscape; (4) Shannon diversity index (SHDI) represents the diversity of the landscape measured by the number of contiguous classes; (5) related circumscribing circle (CIRCLE) is the ratio between the patch surface and the surface of the smallest circle circumscribing the patch, useful because it gives an indication of the patch elongation; (6) interspersion and juxtaposition index (IJI) considers the proximity relations between the different land classes and is evaluated through the percentage of adjacent classes; (7)
The overall accuracy of the land cover maps obtained from the supervized classification was good (0.85 on average for the six case studies). In general, active vegetation performed very well as a classifier (kappa >0.9), as did water (kappa >0.8). Conversely, bare soil and urban areas were less efficient classifiers (the minimum kappa obtained was 0.24 in the Pisa urban region). This is probably due to the high similarity between the spectral signatures of the two classes, which did not allow them to be fully distinguished from each other. The use of ancillary data allowed distinguishing bare soil from urbanized areas. Moreover, in some cases, bare soils corresponded to bare agricultural soils between one crop and another. Also in these cases, the use of ancillary data helped us to distinguish bare soil to agricultural land covers. Land cover changes between 1980 and 2010 fell into two main classes of spatial dynamics: on the one hand, urban growth either around the main city center or as sprawl throughout the peri-urban area; on the other hand, “non-urban dynamics”, namely change from natural to agricultural areas and, conversely, from agricultural to natural areas. In the following sections we describe in detail our findings for these two types of spatial dynamics.
E. Marraccini et al. / Applied Geography 62 (2015) 347e356
353
Fig. 5. Frequency distribution of distances from the nearest urban area for the transition from natural/woody land cover to agricultural land cover.
Urban growth Results for the six case studies show no clear differences between urban growth in the northern and in the southern parts of the Mediterranean areas (Table 3). Southern cities do not seem to grow more or faster than northern cities, and metropolitan areas do not seem to grow faster than the other urban regions. Lower annual growth rate were from Avignon and Madrid (less than þ2%) followed by Pisa, Meknes and Montpellier (between þ3 and þ4%). The highest rate of urban growth was observed for the city of Constantine, with an annual growth rate of þ7.2%. Fig. 3 shows the origin of the urban areas detected in 2011. In some cases, more than 50% of these urban areas were already urban areas in the eighties (Avignon and Madrid). In these cases, urban growth takes place essentially on agricultural areas (or bare soils). Where less than 50% of the urban areas were already urban areas in the eighties (Constantine, Meknes, Montpellier, Pisa), the rate of soil sealing on natural or agricultural areas was different, probably due to factors like previous land uses, local policies (e.g. natural park in Pisa), land tenure (Debolini, Vallette, François, & Chery, 2015; for Meknes). Constantine's particular urbanization history makes it very different from the other study areas. Quite a small city until the 80s, Constantine has experienced very strong growth over the last 30 years. Moreover, in the peri-urban area of Constatine, new conurbations have been built in the southern part of the urban region, occupying former farmland.
Further insights into the different dynamics of urban growth were derived from the landscape metrics analysis (Table 4). These results fitted well with the type of urban region in terms of ratio of urban area to total area of the urban region and urban growth. Large metropolitan areas already occupying a large part of the urban region (e.g. Madrid and Montpellier) present the highest LPI values with the fastest growth. Furthermore, their AI is high and stable, whereas their IJI is slightly decreasing. These results show that urban areas are quite compact and that their urban growth is generally a densification process. However, other urban regions present different features, depending on their type of urban spatial organization. For Pisa and Meknes, which have a highly concentrated urban center, LPI values are low and stable over time, or show weak growth. The same holds for PD, even though starting values are different. In contrast, Constantine and Avignon, having more dispersed urban areas, present similar PD and LPI with comparable growth. Processes of urban growth vary, taking the form of sprawl for Pisa (AI having a different trajectory from the other cases) and densification and/or sprawl for the other cases. No significant changes were detected in ED, SHDI and CIRCLE CV (data not shown). Non-urban dynamics In addition to urbanization processes, we observed other nonurban dynamics, namely changes from agricultural to natural
354
E. Marraccini et al. / Applied Geography 62 (2015) 347e356
Table 3 Percentage of urban areas at the two analyzed dates (cf. Table 2), relative urban growth and annual growth rate (cf. formulas given in Section General methodology).
Meknes, where the areas changing from agricultural to natural land uses are generally located farther from the urban areas.
Case study
% Urban areas T0
% Urban areas T1
Relative urban growth
Annual growth rate
Discussion
Avignon (France) Constantine (Algeria) Madrid (Spain) Meknes (Morrocco) Montpellier (France) Pisa (Italy)
17% 3% 35% 5% 13% 4%
24% 16% 56% 10% 30% 9%
þ39% þ497% þ58% þ108% þ127% þ116%
þ1.4% þ7.2% þ1.8% þ3.1% þ4.0% þ3.0%
Urban planning disciplines tend to consider current local features as a result of social or historical characteristics, and relevant to explain the form of current urban sprawl (UN Habitat, 2009). This has been confirmed by previous studies, such as Zhang (2001), who identified social-economic factors and housing stock as the main drivers of urban development, or Wu (2006), who studied spatial distribution of environmental amenities. Yet the relationship between these characteristics and LULCC may differ, depending on environmental and social conditions (Parr, 2006). From a global perspective, however, segmentations are more complex and cities can exhibit common growth patterns despite different histories or social characteristics (Seto & Fragkias, 2005). Our contribution here is the finding that LULCC dynamics on a sample of six Mediterranean urban regions do not differ significantly from the northern to the southern side of the sea. Some differences occur in LULCC not directly affected by urbanization, since in Northern Mediterranean urban regions afforestation and abandonment of agricultural areas are prevalent and closer to the urban areas, whereas transformation of natural areas into agricultural ones occurs mainly in South Mediterranean urban regions at a similar distance from urban areas than afforested or abandoned areas. We explained this difference with a higher search of urban food security by urban dwellers in Southern regions. A future analysis of the associated agricultural land uses both in Northern and Southern urban regions may support this interpretation. Similar studies on Mediterranean urban areas showed comparable results in terms of urban growth, both on Northern and Southern Mediterranean basin. Weber and Puissant (2003) quantified the land cover changes in the Tunis urban region from 1986 to 1996, obtaining an increase of the builtup areas of 62% in the 10 years. Kasanko et al. (2006) compared 15 European case studies and four of them were considered as a homogenous group of compact southern cities, namely Palermo, Milan, Bilbao and Porto. In terms of urban growth these four cases have a similar trend compared to our six case studies, but they are characterized by a compact urban development, which in our cases is similar for the metropolitan urban areas such as Madrid and Montpellier, but not for the mediumesmall cities. At the same time, we do not identify a differing behavior in urban growth in smallemedium cities and large metropolitan areas already identified in Table 1, whereas differences appeared in type of urban diffusion, which was more sprawling in smallemedium cities and more compact in large metropolitan areas (Table 4). These results are consistent with those of Seto and Fragkias (2005) for four Chinese cities where urban growth induces multiple spatial configurations during the early stages of economic growth, which become convergent thereafter. One explanation may be the bias inherent to both the high degree of individual housing areas in smallemedium cities and the difficulty of clearly identifying this type of land use (Irwin & Bockstael, 2007). Another explanation, according to Salvati et al. (2013), could be the different growth time scales which can lead either to sprawl or to compact growth dynamics. Finally, urban growth and its patterns could also be influenced by topographical factors (e.g. proximity to the sea or a river, hilly systems) or by planning (e.g. a natural parks). However, the influence of these factors on urban growth has not been investigated in this work. These spatial configurations affect agricultural and natural areas differently. We show in Fig. 3 that in metropolitan areas, natural/ forest land cover is more affected than in the other urban regions
Table 4 Results of the urban class metrics for the case studies between T0 and T1. PD means Patch Density, PLI, Largest Patch Index, IJI Interspersion and Juxtaposition Index, AI Aggregation Index. PD
Avignon Constantine Madrid Meknes Montpellier Pisa
LPI
IJI
AI
T0
T1
T0
T1
T0
T1
T0
T1
3.5 2 3 0.8 7.7 7.4
5.2 3.9 2.6 1.5 9.7 7.7
5.7 0.2 22 1.6 6.4 0.4
10.1 3.8 36.9 3.5 17.4 0.6
47.4 15.9 75.9 67.4 63.3 10.7
57 46.5 68.6 72.2 60.8 10.7
84.5 59 86.2 60.8 80 67.9
84.9 77.8 89.7 71.4 84 60.6
areas and, conversely, from natural to agricultural areas (Table 5). Firstly, it is possible to notice that in Southern Mediterranean urban regions on the whole pixels concerned by non-urban dynamics, 61% were pixels converted to agriculture and 39% afforested/ abandoned. On the contrary, in Northern Mediterranean urban regions, the changes concerned even more pixels afforested/ abandoned (73%) than those converted to agriculture (27%). No statistically significant differences in average distances (KruskaleWallis test) were found in the distances to urban areas of afforested/abandoned pixels in the studied Southern and Northern urban regions, whereas significant differences (p-value < 0.0001) were found for in the case of newly agricultural converted areas. At the same time, statistically significant differences were found in Southern and Northern Mediterranean urban regions between pixels abandoned/afforested or converted to agriculture (both pvalues < 0.0001). While these changes occur in all the case studies, they follow different patterns (Figs. 4 and 5). Pisa stands out, with the areas changing from agricultural to natural land cover located closer to the urban areas than the average and the those areas changing from natural to agricultural land cover farer to the urban areas. The opposite pattern can be observed for Constantine and
Table 5 Descriptive statistics on the distances from urban areas of the pixels under nonurban changes (Nat to Agr: converted to agriculture, Agr to Nat: abandoned/afforested). Ns indicates a p-value > 0.05. Case study
Pixel changed
Mean (standard deviation)
Nat to Agr
Nat to Agr Agr to Nat Kruskale Wallis test
Agr to Nat
Avignon (France) 5651 12,776 Constantine (Algeria) 37,457 53,906 Madrid (Spain) 2249 17,824 Meknes (Morrocco) 79,864 21,039 Montpellier (France) 10,405 32,978 Pisa (Italy) 7350 7399 Southern Mediterranean 117,321 74,945 urban regions Northern Mediterranean 25,655 70,977 urban regions
198 210 294 357 133 786 310
(173) (235) (278) (328) (118) (653) (310)
349 (470)
183 318 261 529 130 232 377
p-value
(210) (312) (468) (676) (155) (219) (456)
* *** ** ns ns *** ***
183 (284)
***
E. Marraccini et al. / Applied Geography 62 (2015) 347e356
(13e5%), whereas in the other urban regions, agricultural areas are more affected than in metropolitan areas (30e50%). Thus, the more compact growth in large metropolitan areas makes it easier to maintain the previous agricultural systems and natural areas until building, while the sprawl occurring in smallemedium cities generates a particular disturbance of agricultural systems through the fragmentation of open spaces (Salvati, 2013; Schneider & Woodcock, 2008). In addition, the local dynamics of agricultural or natural areas may be different and may react differently to urban growth: in regions where agricultural systems are intensive and agricultural areas highly specialized, the fractioned spatial structure generated by smallemedium city sprawl facilitates agricultural abandonment (York et al., 2011). This is true of the orchards surrounding Avignon and the cereal-growing areas surrounding Constantine. Conversely, in regions where agriculture is diversified, with small-scale systems like family farming, smallemedium city sprawl allows some agricultural areas to be maintained between urban patches, as with Pisa. Conclusion Our comparison of LULCC over three decades in 6 Mediterranean case studies highlights both common features and differences in the analyzed Mediterranean urban regions. The common features relate to the spatial patterns of large metropolitan areas, while the differences are linked to city size (smallemedium vs large). Here, the mediumelarge cities experienced a strong increase in urban area (more than 10%), mainly centered around the urban core. Cities with less than 100,000 inhabitants showed a moderate increase in urban area (less than 10%) but with a strong effect on open space fragmentation. These results do not reveal any significant differences in LULCC dynamics between northern and southern Mediterranean urban regions for the six case studies, or between countries. Nevertheless, we found a higher proximity to urban areas of afforestation or abandonment dynamics in Northern urban regions rather than the transition to new agricultural areas, whereas these transitions are comparable in terms of proximity and size in Southern urban regions. A more comprehensive examination of LULCC dynamics at Mediterranean level should enable the consequences of current trends on open spaces to be properly assessed, with a view both to natural conservation and to food production for cities. Moreover, next development of these kinds of studies should be focused on modeling possible future LULCC in the Mediterranean area, in order to have insights on possible planning policies and also on the impact on ecosystem services which can be associated to the future identified dynamics. The results of this study suggest the need of a deeper knowledge on the land use and land cover dynamics on the whole Mediterranean area, in order to identify common pattern or possible dissimilarity and analyze the factors driving these trajectories. Acknowledgment The authors acknowledge ANR funding via the DAUME project n ANR-2010-STRA-007-01. We thank Marjorie Sweetko for English language revision. References Abrantes, P., Pimentel, D., & Tenedorio, J. A. (2010). Metropolitan dynamics typology of the Portuguese urban system. The Open Urban Studies Journal, 3, 68e77. Aguilera-Benavente, F., Botequilha-Leit~ ao, A., & Diaz-Varela, E. (2014). Detecting multi-scale urban growth patterns and processes in the Algarve region (Southern Portugal). Applied Geography, 53, 234e245.
355
~o, A. (2011). Landscape Aguilera-Benavente, F., Valenzuela, L. M., & Botequilha-Leita metrics in the analysis of urban land use patterns: a case study in a Spanish metropolitan area. Landscape and Urban Planning, 99, 226e238. Allen, A. (2003). Environmental planning and management of the peri-urban interface: perspectives on an emerging field. Environment and Urbanization, 15(1), 135e148. Bailey, D., Herzog, F., Augenstein, I., Aviron, S., Billeter, R., Szerencsits, E., et al. (2007). Thematic resolution matters: indicators of landscape pattern for European agro-ecosystems. Ecological Indicators, 7(3), 692e709. Basnou, C., Alvarez, E., Bagaria, G., Guardiola, M., Isern, R., Vicente, P., et al. (2013). Spatial patterns of land use changes across a Mediterranean metropolitan landscape: implications for biodiversity management. Environmental Management, 52, 971e980. Benoit, G., & Comeau, A. (2005). A sustainable future for the Mediterranean. The blue plan's e Environment and development outlook (pp. 197e302). Taylor and Francis. Bouma, J., Varallyay, G., & Batjes, N. H. (1998). Principal land use changes anticipated in Europe. Agriculture, Ecosystem and Environment, 67, 103e119. Brueckner, J. K. (2000). Urban sprawl: diagnosis and remedies. International Regional Science Review, 23(2), 160e171. €se, S., Sivrikaya, F., & Keleş, S. (2008). Evaluating Çakir, G., Ün, C., Baskent, E. Z., Ko urbanization, fragmentation and land use/land cover change pattern in Istanbul city, Turkey from 1971 to 2002. Land Degradation & Development, 19(6), 663e675. Catalan, B., Saur, D., & Serra, P. (2008). Urban sprawl in the Mediterranean? Patterns of growth and change in the Barcelona metropolitan region 1993e2000. Landscape and Urban Planning, 85, 174e184. Chorianopoulos, I., Pagonis, T., Koukoulas, S., & Drymoniti, S. (2010). Planning, competitiveness and sprawl in the Mediterranean city: the case of Athens. Cities, 27, 249e259. De Bon, H., Parrot, L., & Moustier, P. (2010). Sustainable urban agriculture in developping countries. A review. Agronomy for Sustainable Development, 30, 21e32. Debolini, M., Vallette, E., François, M., & Chery, J.-P. (2015). Mapping land use competition in the rural-urban fringe and futures perspectives for lands (Morocco). Land Use Policy, 47, scape policies: a case study of Mekne 373e381. Dossa, L. H., Abdulkadir, A., Amadou, H., Sangare, S., & Schlecht, E. (2011). Exploring the diversity of urban and peri-urban agricultural systems in Sudano-Sahelian West Africa: an attempt towards a regional typology. Landscape and Urban Planning, 102(3), 197e206. EEA e European Environmental Agency. (2006). Corine land cover 2006 raster data. From http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2006raster. Foley, J. A., Defries, R., Asner, G. P., Barford, C., Bonan, G., Carpenter, S. R., et al. (2005). Global consequences of land use. Science, 309, 570e574. Forman, R. T. T. (2008). The urban region: natural systems in our place, our nurishment, our home range, our future. Landscape Ecology, 23, 251e253. Gellrich, M., & Zimmermann, N. E. (2007). Investigating the regional-scale pattern of agricultural land abandonment in the Swiss mountains: a spatial statistical modelling approach. Landscape and Urban Planning, 79, 65e76. Grasso, M., & Feola, G. (2012). Mediterranean agriculture under climate change: adaptive capacity, adaptation, and ethics. Regional Environmental Change, 12(3), 607e618. Hepcan, S¸. (2012). Analyzing the pattern and connectivity of urban green spaces: a case study of Izmir, Turkey. Urban Ecosystems, 16(2), 279e293. € _ A., & Ozkan, Hepcan, Ç. C., Turan, I. M. B. (2011). Monitoring land use change in the Çes¸me coastal zone, Turkey using aerial photographs and satellite imaging. Land Degradation & Development, 22(3), 326e333. Houpin, S. (2011). Urban mobility and sustainable development in the Mediterranean: regional diagnostic outlook. Blue plan papers. UNEP. From http://www.planbleu. org/publications/Cahier9_mobilite_urbaine_uk.pdf. Irwin, E. G., & Bockstael, N. E. (2007). The evolution of urban sprawl: evidence of spatial heterogeneity and increasing land fragmentation. Proceedings of the National Academy of Sciences, 104(52), 20672e20677. Jongman, R. H. G. (2002). Homogenisation and fragmentation of the European landscape: ecological consequences and solutions. Landscape and Urban Planning, 58(2e4), 211e221. Kasanko, M., Barredo, J. I., Lavalle, C., McCormick, N., Demicheli, L., Sagris, V., et al. (2006). Are European cities becoming dispersed? A comparative analysis of 15 European urban areas. Landscape and Urban Planning, 77, 111e130. Lambin, E. F., & Meyfroidt, P. (2011). Global land use change, economic globalization, and the looming land scarcity. Proceedings of the National Academy of Sciences of the United States of America, 108(9), 3465e3472. Landgrebe, D., & Biehl, L. (2011). An introduction and reference for MultiSpect e Version 9., 2011 https://engineering.purdue.edu/~biehl/MultiSpec/MultiSpec_ Intro_9_11.pdf, 12 November 2014. Lausch, A., & Herzog, F. (2002). Applicability of landscape metrics for the monitoring of landscape change: issues of scale, resolution and interpretability. Ecological Indicators, 2, 3e15. Leontidou, L. (1993). Postmodernism and the city: Mediterranean versions. Urban Studies, 30(6), 949e965. Lesschen, J. P., Kok, K., Verburg, P. H., & Cammeraat, L. H. (2007). Identification of vulnerable areas for gully erosion under different scenarios of land abandonment in Southeast Spain. Catena, 71, 110e121.
356
E. Marraccini et al. / Applied Geography 62 (2015) 347e356
McGarigal, K., & Marks, B. J. (1994). Fragstat. Spatial pattern analysis program for quantifying landscape structure. From http://www.umass.edu/landeco/pubs/ mcgarigal.marks.1995.pdf. Millington, J. D. A., Perry, G. L. W., & Romero-Calcerrada, R. (2007). Regression techniques for examining land use/cover change: a case study of a Mediterranean landscape. Ecosystems, 10, 562e578. Morin, F. (1991). The evolution of counturbanisation around Saintes: from rural to periurban [Evolution de l’espace periurbain de Saintes du rural au periurbain]. Norois, 152, 397e408. Mougeot, L. J. A. (2000). Urban agriculture: definition, presence, potentials and risks. Agenda. In N. Bakker, et al. (Eds.), Growing cities, growing foods: Urban agriculture on the policy (pp. 1e42). Feldafing, DSE. Moustier, P., & Fall, S. A. (2004). Les dynamiques de l'agriculture urbaine: carrisation et e valuation. In O. B. Smith, P. Moustier, A. J. L. Mougeot, & A. Fall acte (Eds.), CIRAD/CRDI, Montpellier, France (pp. 23e43). Nainggolan, D., de Vente, J., Boix-Fayos, C., Termansen, M., Hubacek, K., & Reed, M. S. (2012). Afforestation, agricultural abandonment and intensification: competing trajectories in semi-arid Mediterranean agro-ecosystems. Agriculture, Ecosystems and Environment, 159, 90e104. Navarro, C. (2012). Metropolitanization of Spanish irban areas: trends, challenges and the special case of Madrid. European Political Science, 11(3), 420e429. Ortiz-Miranda, D., Moragues-Faus, A., & Arnalte-Alegre, E. (2013). Introduction: reframing agriculture in Mediterranean Europe. In D. Ortiz-Miranda, A. Moragues-Faus, & E. Arnalte-Alegre (Eds.), Agriculture in Mediterranean Europe. Between old and new paradigms (pp. 1e8). Parr, J. B. (2006). City hierarchies and the distribution of city size: a reconsideration of Beckmann's contribution. Journal of Regional Science, 9(2), 239e253. Raynaut, C. (2001). Societies and nature in the Sahel: ecological diversity and social dynamics. Global Environmental Change, 11, 9e18. Romero-Calcerrada, R., & Perry, G. L. W. (2004). The role of land abandonment in landscape dynamics in the SPA ‘Encinares del rıo Alberche y Cofio, Central Spain, 1984e1999. Landscape and Urban Planning, 66, 217e232. Salvati, L. (2013). Monitoring high-quality soil consumption driven by urban pressure in a growing city (Rome, Italy). Cities, 31, 349e356. Salvati, L., Munafo, M., Morelli, V. G., & Sabbi, A. (2012). Low-density settlements and land use changes in a Mediterranean urban region. Landscape and Urban Planning, 105, 43e52. Salvati, L., Sateriano, A., & Bajocco, S. (2013). To grow or to sprawl? Land cover relationships in a Mediterranean city region and implications for land use management. Cities, 30, 113e131. San Roman Sanz, A., Fernandez, C., Mouillot, F., Ferrat, L., Istria, D., & Pasqualini, V. (2013). Long-term forest dynamics and land-use abandonment in the Mediterranean mountains, Corsica, France. Ecology and Society, 18(2), 38. http://dx. doi.org/10.5751/ES-05556-180238. Schneider, A., & Woodcock, C. E. (2008). Compact, dispersed, fragmented, extensive? a comparison of urban growth in twenty-five global cities using remotely sensed data, pattern metrics and census information. Urban Studies, 45(3), 659e692. Schroter, D., Cramer, W., Leemans, R., Prentice, I. C., Araujo, M. B., Arnell, N. W., et al. (2005). Ecosystem service supply and vulnerability to global change in Europe. Science, 310, 1333e1337. Serra, P., Vera, A., Tulla, A. F., & Salvati, L. (2014). Beyond urbanerural dichotomy: exploring socioeconomic and land-use processes of change in Spain (1991e2011). Applied Geography, 55, 71e81. Seto, K. C., & Fragkias, M. (2005). Quantifying spatiotemporal patterns of urban land-use change in four cities of China with time series landscape metrics. ́ Landscape Ecology, 20, 871e888. € ck, H., Wiesner, M., Felbier, A., Marconcini, M., Esch, T., & Dech, S. (2014). Taubenbo New dimensions of urban landscapes: the spatio-temporal evolution from a
polynuclei area to a mega-region based on remote sensing data. Applied Geography, 47, 137e153. Tayyebi, A., Pijanowski, B. C., & Pekin, B. K. (2015). Land use legacies of the Ohio River Basin: using a spatially explicit land use change model to assess past and future impacts on aquatic resources. Applied Geography, 57, 100e111. Torreggiani, D., Dall'Ara, E., & Tassinari, P. (2012). The urban nature of agriculture: bidirectional trends between city and countryside. Cities, 29, 412e416. Turner, B. L., II, Lambin, E. F., & Reenberg, A. (2007). The emergence of land change science for global environmental change and sustainability. PNAS, 104, 20666e20671. UN Habitat. (2009). Planning sustainable cities: Global report on human settlements (Abridged ed.). London: Earthscan, 98 pp. UN-Habitat (United Nations Human Settlements Programme). (2012). The state of Magheb cities. In The state of Arab cities, chap 3: “The state” (pp. 88e127). Retrived online the April 20th, 2015 at: http://mirror.unhabitat.org/pmss/ listItemDetails.aspx?publicationID¼3320&AspxAutoDetectCookieSupport¼1. van Veenhuizen, R. (2006). Cities farming for the future. In R. van Veenhuizen (Ed.), Cities farming for the future: Urban agriculture for green and productive cities (pp. 1e17). Ottawa, Canada: RUAF Foundation, IIRR, IDRC. Verburg, P. H., Veldkamp, T., & Bouma, J. (1999). Land use change under conditions of high population pressure: the case of Java. Global Environmental ChangeHuman and Policy Dimensions, 9, 303e312. http://dx.doi.org/10.1016/s09593780(99)00175-2. Verburg, P. H., Soepboer, W., Veldkamp, A., Limpiada, R., Espaldon, V., & Mastura, S. S. A. (2002). Modeling the spatial dynamics of regional land use: the CLUE-S model. Environmental Management, 30, 391e405. http://dx.doi.org/ 10.1007/s00267-002-2630-x. Vidal, R., & Fleury, A. (2009). La place de l’agriculture dans la m etropole verte. Projets de paysage http://www.urbanagricultureeurope.la.rwth-aachen.de/files/vidal_ fleury_2009_metropole_verte.pdf. Vimal, R., Geniaux, G., Pluvinet, P., Napoleone, C., & Lepart, J. (2012). Detecting threatened biodiversity by urbanization at regional and local scales using an urban sprawl simulation approach: application on the French Mediterranean region. Landscape and Urban Planning, 104, 343e355. Weber, C., & Puissant, A. (2003). Urbanization pressure and modeling of urban growth: example of the Tunis metropolitan area. Remote Sensing of Environment, 86, 341e352. World Bank. (2005). The urban transition in Sub-Saharan Africa: Implications for economic growth and poverty reduction. Africa Region, Working Paper Series 97. Wu, J. (2006). Environmental amenities, urban sprawl, and community characteristics. Journal of Environmental Economics and Management, 52, 527e547. Wulder, M. A., White, J. C., Goward, S. N., Masek, J. G., Irons, J. R., Herold, M., et al. (2008). Landsat continuity: issues and opportunities for land cover monitoring. Remote Sensing of Environment, 112(3), 955e969. York, A. M., Shrestha, M., Boone, C. G., Zhang, S. A., Harrington, J. A., Prebyl, T. J., et al. (2011). Land fragmentation under rapid urbanization: a cross-site analysis of southwestern cities. Urban Ecosystems, 14(3), 429e455. Zasada, I. (2011). Multifunctional peri-urban agriculture e a review of societal demands and the provision of goods and services by farming. Land Use Policy, 28, 639e648. Zavala, M. A., & Burkey, T. V. (1997). Application of ecological models to landscape planning: the case of the Mediterranean basin. Landscape and Urban Planning, 38(3e4), 213e227. Zhang, T. (2001). Community features and urban sprawl: the case of the Chicago metropolitan region. Land Use Policy, 18, 221e232. Zlotnik, H. (2003). The population of the Mediterranean region during 1950e2000. In H. Brauch, P. Liotta, A. Marquina, P. Rogers, & M.-S. Selim (Eds.), Security and environment in the Mediterranean (pp. 593e614). Berlin Heidelberg: Springer.