Spatiotemporal Changes (1945–2020) in a Grazed Landscape of Northern Greece, in Relation to Socioeconomic Changes

Abstract

The spatiotemporal changes of the grazed Greek landscapes in the last 75 years resemble those evidenced in most parts of the northern Mediterranean region, where woody vegetation encroached on open areas changing landscape structure and diversity. These landscape transitions are deeply influenced by demographic and socioeconomic changes that exacerbate the abandonment of traditional management practices including livestock farming and wood harvesting. The aim of this paper was to examine the spatiotemporal changes regarding land use/land cover (LULC) types in a typical grazed landscape of the Lagadas area in northern Greece in the period 1945–2020 and try to associate them with socioeconomic changes. Special attention was given to grassland evolution. Cartographic material in various forms, such as historic (LULC) data sets in shapefile format (1945, 1960 and 1993), recent land use maps (Corine Land Cover of 2018), and satellite images (Google Earth images from 2017 to 2020) was analyzed with Geographic Information Systems software and landscape metrics. Socioeconomic inventory data and grazing animal numbers were also collected and analyzed from diachronic census reports of Greek authorities. Spatiotemporal changes in the Lagadas landscape showed that grasslands, open shrublands and silvopastoral areas decreased during the examined period in favor of dense shrublands and forests, causing a significant reduction in landscape diversity and heterogeneity. Main demographic and socioeconomic drivers were the decrease of the local population, population aging and a significant reduction of employment in the primary economic sector over time. These changes were coupled with reductions in the number of grazing animals (sheep, goats, and cattle), firewood harvesting and charcoal production and were identified as the main reasons for landscape change. Grasslands have become increasingly fragmented and isolated over the years. Future sustainable livestock husbandry in the area is seriously threatened by the ongoing reduction of grasslands and open shrublands.

1. Introduction

In the last 75 years, spatiotemporal changes in the grazed Greek landscapes have taken the form of woody plants (trees or shrubs) expansion, altering landscape structure and diversity [1,2,3,4,5,6]. These changes mainly concern the size and distribution of their Land Use/Land Cover (LULC) units and have as a result the large expansion of forests and dense shrublands into former open areas such as grasslands, open shrublands, silvopastoral, and abandoned agricultural areas [7,8,9,10,11,12]. Demographic and socioeconomic changes played a significant role in the transformations of these landscapes mainly by causing the abandonment of traditional management practices, most notably of extensive pastoral activities [8,13,14,15]. Grazing by livestock animals has been an important traditional management element of the Greek landscapes since the ancient times [16,17]. Significant changes have occurred in the Greek traditional livestock production systems in recent years, primarily due to the reduction of the number of local and transhumant flocks of grazing animals, mainly of sheep and goats [5,14,15], that have altered the structure of the traditional Greek pastoral landscapes [5,18] in the form of forest expansion.

Forest expansion is a common trait in most parts of the northern Mediterranean region [3,19] and is mainly caused by the abandonment of marginal agricultural lands [20,21], following the socioeconomic trends of land abandonment of rural areas [22,23]. Land abandonment in the Mediterranean region has also led to a dramatic reduction of traditional human activities, including extensive livestock farming and the collection of firewood and wood products [3,13,24,25,26,27,28], despite their potential to support development in less privileged regions [29]. More recently, climate changes, along with unbalanced land use activities such as coastal urbanization and undergrazing, are also considered to be a major threat to the integrity of Mediterranean ecosystems [30]. The results of these changes cause, particularly for the Mediterranean region, reduced biodiversity and an increased woody cover that have been frequently associated with an increased risk of wildfires [3,9,22,31]. Furthermore, these changes are reported to cause the invasion and dominance of woody species in grassland ecosystems to such an extent that grasslands are threatened with extinction [24,32,33,34]. In general, grasslands stand out among the world ecosystems as being at greatest risk because of extensive habitat loss and their low degree of protection [35]. In addition to the direct loss of their areas, grasslands are becoming more fragmented and isolated, causing a significant decline in grassland species [36].

Studies investigating spatiotemporal changes of landscape and vegetation types are usually conducted by means of photointerpretation analysis of aerial photograph time series sets [6,22,37] and/or digital processing of multispectral satellite images with a variety of techniques [38,39,40], or through object recognition techniques of remote sensing [41,42]. Geographic Information Systems (GIS) have also become an essential tool for evaluating landscape changes through time [2,40,43]. Recognizing LULC transition patterns from spatiotemporal landscape changes is an important element in the study of land use conflicts and influences, whereas several new tools are available for addressing them. Such tools include the dynamic annual rate [14], the annual rate of changes [43], the relative gain and loss [6], the persistence and net changes as quantity difference and swap as allocation difference [44], and the systematic or main transitions of change [22,39,45].

Spatiotemporal change in landscapes significantly alters landscape structure (composition and configuration) and can be easily evaluated with the use of landscape metrics [1,46,47,48]. Forest transitions and land abandonment are also evaluated by landscape metrics [6,49], as well as the impacts of agricultural intensification and abandonment on grassland patches and landscape structure [4,50,51].

The study of socioeconomic changes in combination with the results of spatiotemporal analysis of landscape development can provide valuable information in assessing the causes that affect landscape evolution. Thus, it comes as no surprise that in recent years, a considerable number of landscape studies have tried to examine and quantify the human impact on landscape evolution by co-examining the impacts exerted by socio-economic changes [52,53,54,55]. Nevertheless, there seems to be a limited body of literature on the spatiotemporal changes evidenced in grazed landscapes that suffer from land abandonment, for which the influences of socioeconomic changes are also considered. This lack of knowledge seems to be more pronounced in the case of the Eastern Mediterranean region. Hence, the present research primarily aims to investigate and to quantify the spatiotemporal changes in a typical grazed landscape of northern Greece, and to identify the effect of these changes on land use, focusing on grasslands. Furthermore, it aspires to elucidate the types of socioeconomic and demographic changes that serve as the main driving factors of landscape change, and identify the LULC types characterized by a high degradation risk. Finally, the effects of these spatiotemporal transitions on landscape structure and grassland patches are studied.

2. Materials and Methods

2.1. Study Area

The study area (hereafter Lagadas landscape) belongs to the Lagadas County of the Thessaloniki Prefecture (northern Greece) and is located 30 km NE from the second largest metropolitan area of the country, the city of Thessaloniki (population 1,110,000). The area consists of five village communities—municipality subdistricts (Kolchiko, Exalofos, Lofiskos, Ossa and Kryoneri) that cover an area of about 250 km², which ranges from less than 100 m a.s.l. in the south to more than 1100 m a.s.l. in the northeast and extends north of the Lake Koronia (Figure 1). Lake Koronia belongs to the EU’s Natura 2000 network sites [56], is extremely important worldwide for wildlife conservation, and is also protected by the Ramsar Convention [57]. The topography of the Lagadas landscape is hilly/semi-mountainous, with a rather dense drainage pattern. The climate is semi-arid to sub-humid Mediterranean with a mean annual precipitation of around 500 mm and a mean annual temperature of 11.5 °C. The main land uses are shrublands, agricultural areas, and forests. Shrublands are covered mainly by Quercus coccifera L. (kermes oak) and are grazed almost all year round by sheep and goats. Agricultural areas are cultivated mainly with annual crops and forests are dominated by deciduous oaks (e.g., Quercus pubescens Willd., Quercus frainetto Ten.) [7].

2.2. Socioeconomic Changes

The study period extended from 1945 to 2020. In order to examine the effect of socio-economic changes (population, employment data) and pastoral activities (number of sheep, goats, and cattle) on landscape evolution, statistical data were collected and analyzed from national census reports retrieved from the Hellenic Statistical Authority (available from 1951 to 2011) and historical records, such as the valuable transhumance records of the 1950s reported in Chatzimichali [58]. Additional relevant data concerning forest management, as in the cases of afforestation programs and charcoal and fuelwood production, was collected from previous studies [7,59].

2.3. Data Acquisition and Land Use/Land Cover Changes

The following cartographic material was collected and processed: Historic records of LULC data of the study area for the years 1945, 1960, and 1993 in shapefile format, georeferenced to the Hellenic Geodetic Reference System 1987 (HGRS87). This data was recovered from: (a) a previous European research project (GeoRange—Geomatics in the assessment and sustainable management of Mediterranean rangelands, contract EVK2-2000-21089 [7,59,60], (b) Digital maps of the Corine Land Cover 2018 (reprojected to HGRS87) and (c) Satellite images obtained from the Google Earth Pro program for the years 2017, 2019, and 2020 (georeferenced to HGRS87).

Initially, the historic data of 1945, 1960, and 1993 were reprocessed by means of the ArcGIS^® software (ver.10.8.1, ESRI Inc., Redlands, CA, USA), in a procedure aimed at reducing the original number of LULC types to those required by the selected classification scheme for further analysis. More specifically, the selected classification system consisted of eight categories of LULC types (

Table 1. Classification scheme of Land Use/Land Cover types in the Lagadas landscape.

Land Use/Land Cover Types	Description	Codes
Arable lands	Areas covered with temporary or permanent field crops	AL
Grasslands	Areas dominated by herbaceous plants, with woody shrubs or/with tree cover of less than 10%	GR
Open Shrublands	Areas dominated by sparse woody shrubs with a cover less than 40%	OS
Dense Shrublands	Areas dominated by dense woody shrubs with a cover higher than 40%	DS
Silvopastoral areas	Open grazed forest with a tree cover between 10 and 40%	SP
Forests	Forest areas with a tree cover higher than 40%	F
Barren areas	Bare lands with little or no vegetation	B
Urban areas	Areas with manmade features mainly villages	UR

) and was based on the LULC classification system used by the Greek Forest Service [7]. To identify the LULC types from 2020, visual (on screen) photointerpretation and manual digitization of LULC polygons were performed in shapefile format on recent Google Earth satellite images. Visual photointerpretation of various features on the satellite images was based on standard photographic keys (tone, texture, pattern, shade, shape, and size) and association of features to identify the different LULC types [6,7,22,37,39,59]. The most recent historic data of 1993 were used as a reference source and guided the photointerpretation of 2020. The selected minimum mapping unit was 1 ha, which is in compliance with all historic data sets. The 2020 LULC mapping procedure included the 2018 Corine Land Cover digital map as additional supporting materials in combination with the Google Earth^® (Alphabet Inc., Mountain View, CA, USA) software. More specifically, visual photointerpretation was supported by available 3D views, street view images of vegetation along the roads connecting the village communities, and summer and winter period vegetation images viewed in Google Earth^® software. Visual photointerpretation was also supported by extensive field sampling verifications by a group of scientists with research experience in the specific areas of more than 20 years. Finally, ArcGIS^® and Microsoft Excel^® (Redmond, WA, USA) software were also used for further processing of the above-mentioned cartographic material that resulted in the preparation of tables and digital maps exhibiting the temporal evolution of LULC types.

The annual rate of change of LULC types was calculated for: (a) the study period (1945–2020) and (b) three different intermediate time periods (1945–1960, 1960–1993 and 1993–2020), according to a procedure that was first introduced by Puyravad [61]. The annual rate of change (r) was based on the formula derived from the Compound Interest Law, which provides a more accurate estimation and a better biological assessment in terms of LULC change comparisons, but is also insensitive to the differing time periods between observation dates [39]. The annual rate of change was calculated following the formula in Equation (1) below:

$$r=\left(\frac{1}{t_2-t_1}\right)\times ln\left(\frac{A_2}{A_1}\right)$$

(1)

where A₁ and A₂ are the LULC class areas at time t₁ and t₂, respectively.

2.4. Spatiotemporal Transitions and Landscape Structure

The spatiotemporal transitions analysis of the Lagadas landscape, as well as the diachronic transitions of all LULC types from 1945 to 2020, was conducted by using a common Post-Classification Comparison (PCC) change detection method between the different dates of the study period [43]. The use of the PCC technique resulted in a LULC change transition matrix, which was computed using overlay functions in the ArcGIS© software for all time periods. Moreover, a relevant map of spatiotemporal transition of LULC types was also created. Transition matrices allowed the LULC change analysis to include additional components of land change, such as gains and losses, net changes, total changes, and swap. Gains (P_+j) and losses (P_j+) are the proportion of the landscape that experiences gross gain or loss of LULC type j within the examined time interval. Net change is the difference between gain and loss and is indicated as D_j (Equation (2)):

$$Dj=P_{+j}-P_{j+}$$

(2)

Swap (S_j) is the simultaneous gain and loss of a LULC type j. The diagonal elements of the LULC types transition matrices (P_jj) indicate the proportion of the landscape that shows persistence of category j [43]. S_j is calculated as two times the minimum of the gain and loss (Equation (3)):

$$S_j=2\times MIN\left(P_{j+}-P_{jj},P_{+j}-P_{jj}\right)$$

(3)

Total change for each LULC type j (C_j) is the sum of the net change and swap or the sum of gains and losses [39] (Equation (4)):

$$C_j=D_j+S_j$$

(4)

Current scientific views specify net change as quantity difference (or quantity disagreement) and swap as allocation difference (or allocation disagreement) [44].

Another crucial factor for the correct evaluation of LULC changes is the identification of the most systematic transitions or dominant signals of change [39,62]. Based on the transition matrix data, the most important type of transition can be evaluated by summing up the total area of change for each LULC type during the different time periods. This approach fails to consider the random process of LULC change caused by the dominant LULC types. Therefore, a proper way of evaluating LULC transitions is to interpret them based on their size [39]. This evaluation is based on a methodology introduced by Pontius [62], which calculates the expected gains (G_ij) and expected losses (L_ij) if random changes occur among the examined LULC types (Equations (5) and (6)).

$$G_{ij}=\left(P_{+j}-P_{jj}\right)\left(\frac{P_{i+}}{100-P_{+i}}\right)$$

(5)

$$L_{ij}=\left(P_{i+}-P_{ii}\right)\left(\frac{P_{i+}}{100-P_{+i}}\right)$$

(6)

The difference between the observed (P_ij) and the expected transition (G_ij or L_ij) under a random process of gain (P_ij − G_ij) or loss (P_ij − L_ij), is indicated as D_ij, and the ratios meaning (P_ij − G_ij)/G_ij or (P_ij − L_ij)/L_ij, are indicated as R_ij. D_ij and R_ij values reveal the tendency of a LULC type j to gain from type i (focus on gains) and the tendency of LULC type i to lose from type j (focus on losses). Values with a large positive or negative deviation from zero indicate systematic transitions or dominant signals of change [39]. R_ij ratios resemble the form of chi-square tests (observed value − expected value/expected value) [62].

Finally, the ArcGIS Patch Analyst^® (CNFER, Thunder Bay ON, Canada) software [63] was used to analyze landscape pattern (landscape metrics) [7,46,48]. Seven indices of spatial heterogeneity in landscape and class level were calculated: Number of Patches (NumP) and Mean Patch Size (MPS, ha) as an overall measure of landscape fragmentation, Edge Density (ED, m/ha) as a measure of the number of ecotones [47], Interspersion Juxtaposition Index (IJI, %) as a measure of patch dispersal, Mean Nearest Neighbor (MNN, m) as a measure of patch isolation, and Shannon’s Diversity and Evenness Index (SDI and SEI) as a measure of landscape diversity. The mathematical formulas of the chosen indices can be found in Patch Analyst and Arc Fragstats user manuals [46,63]. Among those, the most complex ones are presented below in Equations (7)–(11).

$$ED=\frac{E}{A}\left(\mathrm{10,000}\right)$$

(7)

$$IJI=\frac{-{{\displaystyle \sum }}_{i=1}^{m^{\prime }}{{\displaystyle \sum }}_{k=i+1}^{m^{\prime }}\left[\left(\frac{e_{ik}}{E}\right)\times ln\left(\frac{e_{ik}}{E}\right)\right]}{ln\left(0.5[m^{\prime }(m^{\prime }-1\right)])}\times \left(100\right)$$

(8)

$$MNN=\frac{{{\displaystyle \sum }}_{j=1}^{n^{\prime }}\left(hij\right)}{{n^{\prime }}_i}$$

(9)

$$SDI=-{\displaystyle \sum _{i=1}^m}\left(Pi\times lnPi\right)$$

(10)

$$SEI=\frac{-{{\displaystyle \sum }}_{i=1}^m\left(Pi\times lnPi\right)}{lnm}$$

(11)

where: E = total length (in m) of all edges in the landscape or sum of length of all patch edges of the same type (class); A = total landscape area (in m²); e_ik = total length (m) of the edge in a landscape between patch types i and k; m′ = number of patch types present in the landscape including the landscape border, if present; h_ij = distance (in m) from patch ij to nearest neighboring patch of the same type (class), based on edge-to-edge distance; n′_i = number of patches in the landscape of patch type (class) i that has nearest neighbors; P_i = proportion of the landscape occupied by patch type (class) i; m = number of patch types present in the landscape excluding the landscape border, if present.

3. Results and Discussion

3.1. Socioeconomic Changes

The total population of the five local communities decreased by 18% between 1945 and 2011 (Figure 2).

During the study period, local population reached its peak in 1961, because at that time local people were fully relocated back to their villages after the end of the Second World War (1940–1945) and the Greek Civil War (1945–1949). Taking this into account, the evolution of the local population, from its peak value (1961) until the year with the most recent census data (2011), showed a decrease of 27%. The decrease was more intense in the villages Lofiskos, Ossa, and Kryoneri, which are located in semi-mountainous and mountainous parts of the study area (Figure 1).

In terms of population age structure, it was observed that the percentage of the population with an age of up to 44 years old was reduced from 76 to 44%, whereas the percentage of the local population with an age over 45 years old increased from 24 to 56%, for the same period. This change appeared in all village communities but was less severe in the Kolchiko village, which is in the lowland area of the landscape (

Table 2. Evolution of age structure distribution of the local population in the Lagadas village communities between 1961 and 2011 (in %).

Villages	1961		1991		2001		2011
	(0–44)	(≥45)	(0–44)	(≥45)	(0–44)	(≥45)	(0–44)	(≥45)
Kolchiko	75.16	24.84	56.58	43.42	54.58	45.42	52.81	47.19
Exalofos	76.09	23.91	58.18	41.82	54.32	45.68	44.25	55.75
Lofiskos	79.00	21.00	55.28	44.72	44.24	55.76	40.94	59.06
Ossa	71.14	28.86	49.33	50.67	40.99	59.01	35.75	64.25
Kryoneri	79.00	21.00	60.27	39.73	52.89	47.11	48.70	51.30
Total	76.08	23.92	55.93	44.07	49.40	50.60	44.49	55.51

(Source: Hellenic Statistical Authority).

, Figure 1).

Regarding the employment census reports from the Hellenic Statistical Authority (

Table 3. Evolution of the employment in the primary sector of the active workforce in the Lagadas village communities between 1961 and 2011.

Year	Total Active Workforce	Employees in the Primary Sector	Employment in the Primary Sector (%)
1961	4031	3780	93.77
1991	2350	969	41.23
2001	1996	764	38.28
2011	1266	385	30.41

(Source: Hellenic Statistical Authority).

), it was observed that in 1961, 94% of the economically active population was employed in the primary sector of the economy (agriculture, livestock, forestry), while this percentage was reduced more than threefold (30%) in 2011. Similarly, during the study period, the total active workforce of the five villages decreased by 69%, indicating that the numbers of retired or officially unemployed in these villages is increasing.

All the above data are in line with the general trend described for the northern Mediterranean region that demographic change consists in the movement of the population from the villages to the urban centers and that this movement mainly concerns the younger age groups [15,52]. This change has the effect of reducing the available workforce and diversifying the ways and techniques of managing traditional land uses causing land abandonment [13,22,47,64]. The great decline of the number of people engaged in the primary sector, in combination with the shortage of young and productive people in the area, had a direct effect on agriculture (i.e., loss of traditional form of agriculture in terraces) and in forestry, as well as in the animal husbandry system in the area. Moreover, the transhumant/extensive livestock system was reduced, while household husbandry was increased only in a few numbers of animal species, as a result of the difficulty of old people to shepherd livestock far away from the villages, especially in high altitude areas [13]. On the other hand, the declining numbers of employees in livestock production until the late 1980s were renewed after the collapse of the former communist countries of eastern Europe in early 1990s, when cheap labor in the form of new immigrants started to work on the Greek animal farms [65]. However, even this labor force boost could not reverse the general trend of land abandonment, especially in the highlands [66]. Moreover, the recent reforms of the Common Agricultural Policy (CAP, starting with the MacSharry reform in 1992) of the European Union tried to support the multifunctional roles of extensive grazing systems, but failed to practically address the specific needs of extensive animal breeders and to halt rural abandonment [15,66,67].

The temporal evolution of farm animal numbers in the Lagadas landscape is presented in

Table 4. Evolution of the livestock numbers (heads) and livestock percentage changes (Δ) in the Lagadas landscape between 1961 and 2011.

Animals	1961	1991	2001	2011	Δ% 1961–2011	Δ% 2001–2011
Cattle	4111	1750	1691	2216	−46.10	30.00
Sheep	21,343	13,425	13,338	9883	−53.69	−25.74
Goats	17,410	14,951	16,575	13,009	−25.28	−23.85
Total	42,864	30,126	31,604	25,108	−41.42	−21.56

(Source: Hellenic Statistical Authority).

and

Table 5. Temporal evolution of the farm animals’ population (sheep, goats and cattle) in numbers (heads) and percentage changes (Δ) per village community in the Lagadas landscape between 1961 and 2011.

Villages	1961	1991	2001	2011	Δ% 1961–2011	Δ% 2001–2011
Kolchiko	11,090	7457	8148	7606	−31.42	−7.27
Exalofos	5550	2414	2361	1904	−65.69	−18.93
Lofiskos	10,165	5572	5240	3591	−64.67	−29.59
Ossa	8041	4842	4585	3502	−56.45	−22.37
Kryoneri	8018	9841	11,270	8505	6.07	−28.10
Total	42,864	30,126	31,604	25,108	−41.42	−21.56

(Source: Hellenic Statistical Authority).

. The total number of farm animals decreased between 1961 and 2011 by 41%. This reduction was more intensive for sheep and cattle (by more than 45%) and less for goats (almost 25%). This trend of change was more intense for the last ten years (2001 to 2011), where the reduction of farm animals was more than 20% (Table 4 and Table 5). On the other hand, during the study period farm sizes increased from an average of 65 animals/farm to 195 animals/farm (Hellenic Statistical Authority), as an attempt by the remaining farmers to increase their incomes under CAP suggestions [15].

These livestock changes reflect the effects of the socioeconomic changes mentioned earlier. The movement of inhabitants from the area to the big urban centers (mainly Thessaloniki) that started in the 1960s resulted in the reduction of the livestock activity. However, after 1981, the year that Greece joined the European Union, the CAP policy of subsidizing livestock animals led to a temporal increase in the number of farm animals. This increase became more apparent in goats between 1991 and 2001 (Table 4), probably due to the fact that shrublands are the dominant LULC type in the study area. Reductions in livestock numbers were very severe in most village communities (>50% between 1961 and 2011) with the exception of Kolchiko and Kryoneri (Table 5). Kolchiko was the only village community that had a slight increase of its population and a less severe aging trend (Figure 2 and Table 2), whereas Kryoneri was the only village community that kept a significant number of people employed in the primary sector in 2011 (164 employees—more than 45% of its total workforce, Table 3).

In order to have a better assessment of livestock pressure during the study period, it is important to mention the number of farm animals (sheep and goats) of the Sarakatsan nomads. The Sarakatsan nomads were, a Greek nomadic (transhumant) ethnic group who traditionally visited the broader area of Lagadas landscape. They did not have permanent residences; instead, they used to build new huts every year, or repair old ones [18]. The exact number of transhumant sheep and goats camping within the Lagadas landscape is not completely known, but an estimate of around 11,000 nomadic farm animals in the broader area at the end of the 1950s seems possible [58]. Unfortunately, the number of animals was not recorded by the State Authorities at that time, which is the reason for not including this data in Table 4 and Table 5. Since the 1990s, the presence of the Sarakatsan nomads in the area has been significantly restricted.

Finally, during the study period, the socioeconomic effects οn forestry activities are linked to the reduction of charcoal production (oak species were used for charcoal), and the decreased collection of kermes oak fuelwood. Fuelwood collection was gradually declining and eventually stopped as households turned to fossil fuels for their cooking and heating needs [7,59]. On the other hand, the Forest Service responsible for management of forests and forest lands initiated an afforestation program to increase the area covered by high forests. Until the early 1990s, more than 4000 ha of grasslands and open shrublands were afforested with pine species in several parts of the Lagadas County, including the study area [7,59]. However, forest operations belong among the most strenuous professions [68] and are characterized by increased difficulty [69], high frequency and severity of accidents, and professional health problems [70], most notably musculoskeletal disorders [69]. These attributes, combined with a poor professional profile and low income, have resulted in a sharp decline in the number of forest workers and a profound difficulty in keeping younger people in their villages and continuing to work in forest management [26,27]. Ironically, firewood production is very important for Greece, amounting to 72% of the total production of forest products, thus comprising the most important energy source, especially for remote and mountainous communities.

In general, the previously mentioned developments in the study area are also characteristic of the contemporary management of the Mediterranean vegetation. They have been frequently reported throughout the European part of the Mediterranean region [3,8,13,14,18,19,24,33,59,71] and possess key roles in landscape change.

3.2. Land Use/Land Cover Changes

The results of photo interpretation and LULC changes and annual rate of change over the last 75 years (1945–2020) are shown in

Table 6. Evolution of Land Use/Land Cover types in the Lagadas landscape between 1945 and 2020.

Land Use/Land Cover Type	Area (ha)				Change (ha)	Change (%)
	1945	1960	1993	2020	1945–2020	1945–2020
Arable lands	7411.16	7533.20	8308.26	7995.26	584.10	7.88
Grasslands	5869.14	4446.61	1280.71	717.26	−5151.88	−87.78
Open shrublands (10–40% cover)	3520.33	3820.25	2943.74	2159.38	−1360.95	−38.66
Dense shrublands (>40% cover)	3334.33	3818.21	5571.99	6099.15	2764.82	82.92
Silvopastoral areas (10–40% cover)	1600.56	1425.78	1734.50	1268.73	−331.83	−20.73
Forests (>40% cover)	2426.96	3105.03	4384.48	5943.85	3516.89	144.91
Barren areas	121.49	110.16	35.27	26.71	−94.78	−78.01
Urban areas	150.42	195.63	285.29	379.73	229.31	152.44
Total	24,434.39	24,454.87	24,544.24	24,590.07	155.65	0.64

and

Table 7. Annual rate of change of Land Use/Land Cover types in the Lagadas landscape between 1945 and 2020.

Land Use/Land Cover Type	Annual Rate of Change (% Per Year)
	1945–1960	1960–1993	1993–2020	1945–2020
Arable lands	0.11	0.30	−0.14	0.10
Grasslands	−1.85	−3.77	−2.15	−2.80
Open shrublands (10–40% cover)	0.55	−0.79	−1.15	−0.65
Dense shrublands (>40% cover)	0.90	1.15	0.33	0.81
Silvopastoral areas (10–40% cover)	−0.77	0.59	−1.16	−0.31
Forest (>40% cover)	1.64	1.05	1.13	1.19
Barren areas	−0.65	−3.45	−1.03	−2.02
Urban areas	1.75	1.14	1.06	1.23

. During this period, the woody vegetation in the Lagadas landscape has been considerably increased at the expense of open areas (mainly grasslands, open shrublands, and silvopastoral areas) (Table 6). In particular, the LULC types that increased were forests (>100%), dense shrublands (>80%), and urban areas (>100%), and to a lesser extent, arable lands. On the contrary, grasslands and open shrublands were drastically reduced by more than 80% and 35%, respectively. Barren areas were also reduced (>70%) but they cover only a limited part of the study area.

The annual change rate (Table 7) indicated a variety of trends for all LULC types examined during the study period. The most significant ones were the decline of grasslands and the increase of forest areas. The decrease of grasslands was more intense during the second (1960–1993) and third (1993–2020) study period as a direct effect of land abandonment processes which started in the late 1960s and the 1970s. On the other hand, the increase in forest areas was more uniform along all time periods with a steeper increase trend appearing during the first study period (1945–1960). Among the other LULC types, the annual increase of dense shrublands was rather limited during the third time period, during which open shrublands and silvopastoral areas appear to have the most significant decrease of their annual rates (Table 7) between 1945 and 2020.

The spatiotemporal evolution of LULC type changes is presented in Figure 3. The gradual conversion of grasslands and open shrublands to dense shrublands is observed in the central and southeastern parts of the landscape, along with a gradual conversion of grasslands and silvopastoral areas into forest in the northern-northeastern part of the landscape (Figure 3). Grasslands that maintained a wide dispersion up to the 1960s were considerably reduced, especially after 1993, to a narrow area near Kolchiko village to the south and to a few scattered parts to the north (Figure 1 and Figure 3). Arable lands appear spatiotemporally stable except for an area to the east side of the landscape (next to the village territories of Lofiskos and Kryoneri), where a significant part of grassland was transformed into arable lands after 1993 (Figure 1 and Figure 3).

This trend of forest-shrubland expansion/grassland reduction is consistent with studies from Greece [2,6,8,10,11,14] and other Mediterranean countries [25,33,72], indicating that policy makers and land managers should pay extra attention to these unique LULC interactions that rapidly change the traditional form of Mediterranean landscapes [67,73].

Forest expansion is not necessarily a problem; in some cases, authorities have tried to increase forest coverage to promote and distribute the benefits of ecosystem services to the local communities. However, the decisive element is the management of these resources [31]. In terms of forest utilization, a well-trained forestry workforce can ensure a high level of forest management and utilization [74], which still is a critical deficit of the Greek forest practice. Furthermore, the workforce size should be sufficient for the increased needs resulting from the enhanced size of the forest areas. Unfortunately, the forest professions are considered very difficult and poorly paid [26], resulting in a considerable decline in the number of forest cooperative members in Greece [27]. Therefore, the expansion of forests in the area will hardly meet the expectations of increased volumes, and possibly, of higher quality of forest products; rather, they will be left practically unmanaged in the future. Possible positive impacts of forest expansion include soil recovery, carbon sequestration increase, and nutrient cycling among others [75], which may even lead to ecosystem restoration if they are coupled with targeted initiatives [76]. On the contrary, negative impacts can be detrimental, most notably in the form of accumulation of large biomass quantities in the abandoned and more densely forested areas that increase the risk and severity of wildfires, especially in the context of climate change [31,77].

Moreover, the declining number of livestock farm workers in the area is expected to further promote forest expansion, especially over the traditional agrosilvopastoral areas of the landscapes. Eventually, if this trend prevails, it will be associated with reduced biodiversity [9,67] and alteration of the cultural values of the pastoral landscapes [73,78].

3.3. Spatiotemporal Transitions and Landscape Structure

Spatiotemporal transitions of Lagadas landscape are presented in

Table 8. Land Use/Land Cover change transition matrix between 1945 and 2020 (presented as % of the total area).

					2020
1945	AL	GR	OS	DS	SP	F	B	UR	Total 1945	Loss
AL	24.52	0.19	0.95	1.29	0.76	2.09	0.02	0.54	30.36	5.84
GR	4.85	2.11	4.91	4.87	3.15	3.93	0.00	0.20	24.02	21.91
OS	0.99	0.17	2.23	8.90	0.26	1.76	0.00	0.10	14.41	12.18
DS	0.81	0.05	0.55	9.50	0.04	2.68	0.00	0.01	13.64	4.14
SP	0.41	0.05	0.05	0.19	0.54	5.31	0.00	0.00	6.55	6.01
F	0.63	0.05	0.05	0.22	0.44	8.53	0.00	0.01	9.93	1.40
B	0.27	0.00	0.00	0.00	0.01	0.02	0.09	0.10	0.49	0.40
UR	0.01	0.00	0.00	0.00	0.00	0.00	0.00	0.59	0.60	0.01
Total 2020	32.49	2.62	8.74	24.97	5.20	24.32	0.11	1.55	100.00	51.89
Gain	7.97	0.51	6.51	15.47	4.66	15.79	0.02	0.96	51.89

Note: The values noted in shaded box (diagonally) indicate the unchanged LULC proportions between 1945 and 2020. AL: Arable lands, GR: Grasslands, OS: Open shrublands, DS: Dense shrublands, SP: Silvopastoral areas, F: Forest, B: Barren areas, UR: Urban areas.

and Figure 4.

Between 1945 and 2020, almost 52% of the study area changed use (Table 8). The most important changes were the conversion of open shrublands (OS) to dense ones (DS) and of silvopastoral areas (SP) into forest (F). Another significant change was grassland (GR) conversion into shrublands (open or dense), silvopastoral areas, and forests, and finally, the conversion of grasslands into arable lands (AL, Table 8). However, in only less than 3% of the total landscape the opposite trend has occurred, where dense forest and shrublands were transformed into open areas (e.g., grasslands, open shrublands), and arable lands were transformed into grasslands. The most important change in the Lagadas landscape during the study period was the encroachment of woody plants in open areas, possibly as a direct effect of the reduction in livestock numbers, firewood collection, and charcoal production. Woody plant expansion was also triggered by the afforestation practices of the Greek forest service in grasslands, open shrublands, and arable lands. In addition, a significant part of arable lands (about 5% of the total landscape, Table 8) was transformed into shrublands, silvopastoral areas, and urban areas, indicating the effect of the abandonment of traditional agricultural practices, a trend noticeable mainly during the period 1993–2020 (Table 6 and Table 7).

The map of spatiotemporal transition of LULC types (Figure 4) during the study period showed that shrubland transitions (from open to dense cover types) were mainly observed in the central and southern parts of the landscape, while silvopastoral transitions into forest were observed in the northeastern part of the landscape (Figure 4).

The most widely evident transition in the landscape (highlighted in the Figure 4 map with light and dark blue colors) was the transition of grasslands into shrublands, silvopastoral areas, and forests. In conclusion, several spatiotemporal transitions occurred in various LULC types, but these changes were neither linear from one type to another nor reciprocal in all of them. In addition, the magnitude of the changes was different among the various types.

During all the examined periods, the most significant changes of net values are observed in grasslands and forests, and to a lesser extent in shrublands (

Table 9. Temporal evolution of the Lagadas landscape in terms of percentual Land Use/Land Cover (LULC) gains, losses, net changes, and swaps.

Time Period	LULC Type	Gain	Loss	Total Change	Swap	Absolute Value of Net Change
	Arable lands	4.64	4.14	8.78	8.28	0.50
	Grasslands	3.58	9.40	12.98	7.16	5.82
	Open shrublands	5.51	4.28	9.79	8.56	1.23
	Dense shrublands	4.17	2.19	6.36	4.38	1.98
1945–1960	Silvopastoral areas	2.25	2.97	5.22	4.50	0.72
	Forests	4.50	1.73	6.23	3.46	2.77
	Barren areas	0.09	0.21	0.30	0.18	0.12
	Urban areas	0.20	0.02	0.22	0.04	0.18
	Landscape	24.94	24.94	24.94	18.28	6.66
	Arable lands	7.68	4.59	12.27	9.18	3.09
	Grasslands	1.32	14.49	15.81	2.64	13.17
	Open shrublands	7.07	10.66	17.73	14.14	3.59
	Dense shrublands	11.24	4.07	15.31	8.14	7.17
1960–1993	Silvopastoral areas	5.31	4.04	9.35	8.08	1.27
	Forests	7.62	2.39	10.01	4.78	5.23
	Barren areas	0.01	0.38	0.39	0.02	0.37
	Urban areas	0.39	0.02	0.41	0.04	0.37
	Landscape	40.64	40.64	40.64	23.51	17.13
	Arable lands	4.31	5.58	9.89	8.62	1.27
	Grasslands	0.85	3.27	4.12	1.70	2.42
	Open shrublands	2.65	5.90	8.55	5.30	3.25
	Dense shrublands	6.82	4.67	11.49	9.34	2.15
1993–2020	Silvopastoral areas	2.41	4.31	6.72	4.82	1.90
	Forests	8.89	2.54	11.43	5.08	6.35
	Barren areas	0.05	0.09	0.14	0.10	0.04
	Urban areas	0.48	0.10	0.58	0.20	0.38
	Landscape	26.46	26.46	26.46	17.58	8.88
	Arable lands	7.97	5.84	13.81	11.68	2.13
	Grasslands	0.51	21.91	22.42	1.02	21.40
	Open shrublands	6.51	12.18	18.69	13.02	5.67
	Dense shrublands	15.47	4.14	19.61	8.28	11.33
1945–2020	Silvopastoral areas	4.66	6.01	10.67	9.32	1.35
	Forests	15.79	1.40	17.19	2.80	14.39
	Barren areas	0.02	0.40	0.42	0.04	0.38
	Urban areas	0.96	0.01	0.97	0.02	0.95
	Landscape	51.89	51.89	51.89	23.09	28.80

). Net change values in grasslands and forests were higher in comparison to their swap values for the total study period, indicating that the quantity difference (net value) is by far more significant than their allocation difference (swap value). The latter suggested that the recorded total woody plant expansion against open areas in the landscape (Table 6) is focused on transition trends of these two LULC types.

During the intermediate time period this trend is more obvious in grasslands and for the second period 1960–1993 (Table 9). On the contrary, shrubland (open or dense) values of swap change appear to be higher than net values almost for all time periods, suggesting an extensive instability in their spatial location (continuous reallocation). Similarly, silvopastoral and arable lands show an even greater intertemporal reallocation (Table 7, swap values > net change values). At landscape level, the reported changes are attributable more to quantity difference (net value) than to allocation difference (swap value).

The net change (gains–losses) in ha for each LULC type between 1945 and 2020 is presented in Figure 5.

The net changes (gains–losses) for each LULC type between 1945 and 2020 showed that the largest losses were observed in grasslands and open shrublands, and the highest gains in forests and dense shrublands (Figure 5). Similar results were presented with regard to temporal changes in LULC areas (Table 6). Consequently, the two most significant LULC changes, in terms of total area change (Table 6) and net changes value (Table 8 and Table 9), were the temporal reduction of grassland areas and forest expansion. Grasslands appeared to lose the most significant part of their area during the second time period (1960 to 1993), a fact that was also observed for the same time period in the declining annual rates of change for grasslands in Table 7. On the other hand, forests had a more or less temporary balanced increase rate during all time periods. The declining trend of grassland losses during the third time period (1993 to 2020, Table 9 and Figure 5) suggests that the remaining areas of grassland appeared to have a higher level of persistence to further reduction. A similar persistence trend of grasslands in recent years was reported by Kiziridis et al. [6], who attributed it to a phenomenon they call “the mountain effect”. According to their explanation, the resistance to change of a remaining part of grassland in higher altitude areas is tougher due to the prevailing poor soil conditions.

Table 10. Gains and losses of the most systematic transitions of Land Use/Land Cover types for the Lagadas landscape between 1945 and 2020 (in %).

Transitions	Gains (%)		Losses (%)
	Dij	Rij	Dij	Rij
AL to DS	−4.15	−0.76	−0.87	−0.41
AL to F	−3.23	−0.61	−0.01	−0.01
GR to AL	2.10	0.76	−4.52	−0.48
GR to OS	3.08	1.69	2.39	0.95
GR to DS	0.57	0.13	−2.32	−0.32
GR to SP	1.95	1.63	1.65	1.10
GR to F	−0.28	−0.07	−3.09	−0.44
OS to DS	6.32	2.45	5.57	1.67
OS to F	−0.77	−0.30	−1.49	−0.46
DS to F	0.29	0.12	1.34	1.00
SP to F	4.16	3.62	3.77	2.44

Note: D_ij = The difference between observed and expected value, Rij = The difference between observed and expected value, relative to the expected value. This table includes the most important land use/covers transitions (>1%, Table 8). AL: Arable lands, GR: Grasslands, OS: Open shrublands, DS: Dense shrublands, SP: Silvopastoral areas, F: Forest.

includes the most important total area LULC transitions that exceed 1% of the total landscape area (Table 8). The differences of the observed minus the expected transitions under a random process of gain (D_ij) and their ratios (R_ij) indicate the systematic nature of many LULC types, whose values were distanced from zero. The largest positive values, indicating the most systematic transitions, appear in the grasslands to open shrublands (GR to OS), grasslands to silvopastoral (GR to SP), open to dense shrublands (OS to DS), and silvopastorals to forest (SP to F) transitions. The later implies that open shrublands and silvopastoral areas are systematically gaining area from grasslands, and that closed shrublands and forests are systematically gaining area from open shrublands and silvopastoral areas, respectively. On the other hand, the transition of grasslands to dense shrublands and forests (GR to DS, GR to F), and of shrublands (OS and DS) to forests, have values close to zero, which suggests random changes (weak transition signal).

The differences of D_ij and R_ij values under the random process of losses show similar systematic transitions of LULC types. The results presented in Table 10, indicate that grasslands systematically lose area in favor of open shrublands and silvopastoral areas and that open shrublands and silvopastoral areas also systematically lose area in favor of dense shrublands and forests. On the contrary, arable lands systematically avoid losing area to dense shrublands, as well as grasslands to arable lands, as indicated by their relative Dij and Rij negative values.

The aforementioned results of systematic transitions reveal the strong directions of woody (tree and shrub) expansion over grasslands and the strength of forest expansion. This trend has been reported by other scientists and designates the irreversibility of woody extension following land abandonment [22].

Temporal differences in landscape structure (in landscape level and class level for grasslands) over time are presented in

Table 11. Landscape metrics values for the Lagadas landscape between 1945 and 2020.

Years	NumP 1	MPS 2	ED 3	IJI 4	SDI 5	SEI 6
1945	653	37.42	164.49	75.89	1.72	0.83
1960	567	43.13	141.63	77.22	1.74	0.84
1993	392	62.61	114.48	75.04	1.67	0.80
2020	487	50.49	115.20	71.29	1.60	0.77

Note: ¹ Number of Patches, ² Mean Patch Size (ha), ³ Edge Density (m/ha), ⁴ Interspersion Juxtaposition Index (%), ⁵ Shannon’ s Diversity and ⁶ Evenness Index.

and

Table 12. Landscape metrics values (grassland class) from 1945 to 2020 for the Lagadas landscape.

Years	NumP 1	MPS 2	ED 3	MNN 4
1945	176	33.35	45.37	152.49
1960	160	27.79	30.62	149.05
1993	68	18.83	9.73	230.46
2020	38	18.88	4.87	506.27

Note: ¹ Number of Patches, ² Mean Patch Size (ha), ³ Edge Density (m/ha), ⁴ Mean Nearest Neighbor (m).

.

Τhe differences of NumP and MPS values show that the Lagadas landscape intertemporally becomes less fragmented and therefore, more homogeneous (Table 11). At the same time, ED values decrease, indicating a decrease of ecotone length per unit area, while the IJI shows a slight degradation of patch distribution in the landscape, meaning that similar patch types are less equally adjacent to each other. Finally, the SDI and SEI indices suggest a decrease in diversity over time. The above results are in line with the evolutionary trends of many other Greek landscapes, where the expansion of forest reduces landscape diversity and produces more homogeneous landscapes [4,6,7,10,11,18]. The later observation is common among many scientists studying the effect of land abandonment in the mountainous landscapes of the Mediterranean region, who point out that the advance of woody cover creates homogeneous landscapes with a lesser capacity to host biodiversity [49,79], causing a decrease in the heterogeneity, connectivity, and diversity of grassland ecosystems [25].

The temporal evolution of grasslands (NumP and MPS) shows that grassland patches are becoming more fragmented, reduced both in terms of frequency and size in the area (Table 12). In addition, grassland edges (ED), showed a great reduction of their values (Table 12) indicating the strong negative effect of ecotone reduction that temporal transformations of grassland cause. Moreover, the MNN distance was increasing, exhibiting that grassland patches are becoming increasingly isolated. These trends can be related to grasslands reduction (Table 6 and Table 8 and Figure 3 and Figure 4), suggesting that patches belonging to this LULC type are becoming more fragmented and dispersed over the years, and therefore, sparser and more isolated. Researchers have linked the fragmentation and isolation increase of grassland patches to the decline of grassland species [36]. Furthermore, edge density is reduced for grassland patches, indicating an even stronger negative effect in ecotone length, considering that grasslands produce typical ecotones and corridors associated with high diversity [80,81,82]. Open shrublands show similar results (temporal reduction from 1945 to 2020 in NumP, ED and MNN values). These changes show that grasslands and, to a lesser extent, open shrublands are at a risk of becoming extinct in the Greek grazed landscapes, stressing the need for direct measures that will preserve and restore these two important pastoral resources. Grassland and open shrubland preservation is considered necessary for sustaining extensive livestock husbandry [30,32,83].

3.4. Study Limitations

Although the findings of this study can be proven useful for future research and recognition of the land abandonment effects in grazed Mediterranean landscapes, it would be fair to refer to its shortcomings. The socioeconomic data presented and discussed could be further correlated to additional spatial data by applying multivariable analysis techniques. Such additional data could include several physiographic and accessibility factors related to landscape changes, comparable to the “Landscape driving factors” approach of the CLUE’s modelling framework [32,84]. Analogous approaches could offer a better understanding of the ways that human activities influence and shape several elements of the traditional cultural landscapes. This kind of investigation would require a considerable amount of extra effort in collecting and analyzing the necessary additional spatial data, which exceeded the scope of the present study.

4. Conclusions

Overall, the spatiotemporal transition and structural analysis of the Lagadas landscape suggested that the strongest trend of landscape evolution was the expansion of woody plants in open areas coupled with the reduction of grasslands, changes that were more obvious during the second time period (1960 to 1993). Moreover, the most systematic LULC transitions included those of grassland reduction to open shrubland and silvopastoral areas, of open to dense shrublands and, finally, of silvopastoral areas to forests. Temporal developments of landscape structure suggest that landscape diversity is declining, and instead, homogeneous landscapes are being created. Grassland patches are deteriorating by becoming more fragmented, distant, and isolated.

The above trends of change can be mainly attributed to land abandonment issues that are closely interwoven with socioeconomic conditions. Socioeconomic changes in the area had the form of local population decrease, population aging, and a significant temporal reduction in employment in the primary economic sector, which caused the abandonment of traditional management practices. Extensive pastoral activities (including transhumant or nomadic pastoralism) were strongly reduced and forest product collection (charcoal and fuel woods) was practically halted, despite the afforestation efforts aiming to increase the size of the forested areas. Agricultural production, especially in less favorable areas, was also affected (e.g., loss of traditional form of agriculture in terraces). Recent CAP reforms have also failed to resolve the declining trend of extensive rural activities. All these developments have inevitably influenced the contemporary management of the Mediterranean vegetation and have been identified as the main reasons for landscape change.

It should be noted that the steep reduction in livestock numbers, forest product harvesting, and of extensive rural activities in general, can possibly continue in the future, causing a further expansion of forests and dense shrublands at the expense of grasslands and open shrublands, a trend that would negatively affect landscape evolution and diversity. Thus, land managers must take into consideration the trends presented in this study for the Lagadas landscape. Forest expansion can be beneficial in terms of carbon sequestration and by promoting some aspects of ecosystems services mainly related to soil protection, but at the same time, it increases the biomass amount in poorly managed and underutilized forest ecosystems. Wood and shrub encroachment increases wildfire risk, a crucial problem of the Mediterranean forests, especially in the context of climatic change. Forest and dense shrubland expansion over grassland and open shrubland areas exerts a series of negative impacts by altering the traditional form of the Mediterranean grazed landscapes in terms of species diversity, cultural heritage, and the sustainable development of livestock husbandry. Considering that grasslands and open shrublands are the main natural resource types necessary for applying extensive pastoral practices, their conservation is critical to support ecological integrity and social benefits for the people historically linked to these landscapes.

Tailormade actions that further support and increase traditional human activities are deemed as critical for restoring Mediterranean grazed landscapes. Promoting knowledge and public awareness of the vital role of extensive pastoral practices and sustainable forest management at landscape conservation level could create new marketing opportunities for local dairy and forest products. Higher prices for these products can be achieved by featuring their positive environmental impact and promoting their quality aspects. Additionally, the poor professional profile of livestock and forest workers could be upgraded, possibly increasing the attractiveness of their professions, especially in the younger generations.