Monitoring land use and land cover change in mountain regions: An example in the Jalca grasslands of the Peruvian Andes

Landscape and Urban Planning 112 (2013) 40–49

Contents lists available at SciVerse ScienceDirect

Landscape and Urban Planning journal homepage: www.elsevier.com/locate/landurbplan

Research paper

Monitoring land use and land cover change in mountain regions: An example in the Jalca grasslands of the Peruvian Andes Carolina Tovar a,∗,1 , Arie C. Seijmonsbergen b , Joost F. Duivenvoorden b a b

Conservation Data Centre, La Molina National Agrarian University, Lima, Peru Institute for Biodiversity and Ecosystem Dynamics, Universiteit van Amsterdam, Amsterdam, The Netherlands

h i g h l i g h t s     

The Jalca grasslands decreased at 1.5%/yr, faster than in other Andean regions. Agriculture increased at 1.2%/yr and new agriculture mainly at 3600–3800 m a.s.l. Mining and tree plantation expanded most quickly despite the low absolute values. Mining increased at 9%/yr while tree plantation at 12.3%/yr. Agriculture may be preferably established near areas of high Jalca fragmentation.

a r t i c l e

i n f o

Article history: Received 16 November 2011 Received in revised form 29 September 2012 Accepted 8 December 2012 Available online 21 January 2013 Keywords: Elevation gradient Landscape analysis Land use change Object-based classification Tropical Andes

a b s t r a c t Mountains are rich in biodiversity and provide ecosystem services for their inhabitants. These regions are currently threatened by land use and land cover changes (LUCC), therefore an efficient monitoring is required to capture such changes. The aim of this study is to test a landscape change analysis in a mountain region to guide landscape management by including: (1) LUCC trends, (2) LUCC trends across the elevation gradient and (3) changes in spatial configuration. This framework was applied to the Peruvian Jalca grasslands (>3000 m a.s.l.), located in the Tropical Andes for the period 1987–2007. We used objectbased classification of Landsat TM and patch metrics for each land cover class. Our results show an overall loss of Jalca (−1.5%/yr) and montane forest and shrubland (−2.8%/yr) with higher rates than other Andean regions. Furthermore, fragmentation is observed for the Jalca while montane forest and shrubland class is not fragmenting but the largest patches are vanishing, potentially affecting the connectivity between natural areas. Agriculture has replaced the Jalca, especially in the upper zones of the Andes showing an upward expansion of crops. However tree plantation and mining had increased more dramatically than agriculture (>9%/yr). Upper and less fragmented Jalca areas may be suitable for conservation purposes while agriculture may better expand in already degraded natural areas. Records of changes across the elevation gradient and in spatial patterns result in useful information for decision makers and may improve ecosystem management not only in the Tropical Andes but also in other mountain regions. © 2012 Elsevier B.V. All rights reserved.

1. Introduction Mountain ecosystems are characterized by their topographic variety and climatic gradients. As a consequence, it is not a surprise that they are rich sources of biodiversity (Brooks et al., 2006) and host high plant endemism (Kruckeberg & Rabinowitz, 1985). At the

∗ Corresponding author at: Jesus College, Turl St, Oxford OX1 3DW, United Kingdom. Tel.: +44 01865 281319. E-mail addresses: [email protected], [email protected] (C. Tovar), [email protected] (A.C. Seijmonsbergen), [email protected] (J.F. Duivenvoorden). 1 Present address: Long-Term Ecology Laboratory, Department of Zoology, University of Oxford, Oxford, United Kingdom. 0169-2046/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.landurbplan.2012.12.003

same time they also provide many ecosystem services to human populations such as water supply and energy availability (Körner & Spehn, 2002). Despite their importance, mountains are continuously subjected to rapidly changing environmental factors of which land use change is the most important one (Körner & Spehn, 2002). The loss of traditional land use practices and transformation into modern agricultural areas are threatening mountain biodiversity and their ecosystem services (Spehn, Liberman, & Körner, 2006). Therefore it is of utmost importance to understand temporal land use and land cover changes in order to guide management strategies for both human activities and potential conservation (Seijmonsbergen, Sevink, Cammeraat, & Recharte, 2010). Remote sensing technology provides instruments for monitoring land use and land cover change (LUCC) and the development of a

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49

coherent categorization of land cover units has been a main focus. Traditional classifiers, such as the K-nearest neighbour (KNN) or maximum likelihood (ML), may perform well on Landsat-TM data sets (Smits, Dellepiane, & Schowengerdt, 1999). Most of these algorithms are pixel-based classifications, where each pixel is labelled in an image as representing particular ground cover types or classes. In the last 10 years the use of object-based image analysis has increased (Blaschke, 2010). This approach seeks to detect clusters of similar pixels that together form an ‘object’ instead of single pixels as an analysis unit (Benz, Hofmann, Willhauck, Lingenfelder, & Heynen, 2004). This method proved to be useful especially in heterogeneous regions (Burnett & Blaschke, 2003; Lucas, Rowlands, Brown, Keyworth, & Bunting, 2007) such as mountain areas. A further consideration in mountain ecosystems, related to the topographic heterogeneity, is the change in vegetation across the elevation range. Among other factors, mean annual temperature is the main factor determining the range of different plants (Guisan, Theurillat, & Kienast, 1998). In the last decades many authors have suggested that there is an upward displacement of some ˜ mountain biomes due to climate change (Penuelas & Boada, 2003; Sanz-Elorza, Dana, & Gonzales, 2003). Crops may follow the same pattern; therefore the inclusion of land use and land cover change analysis along an elevation gradient may improve our understanding of LUCC in the mountains. LUCC analyses based on changes in land cover classes alone might not be sufficient information in monitoring mountain regions. The heterogeneity and related functions of these fragile regions emphasize the need to integrate stratified LUCC information along with temporal landscape metric changes. For instance, ecosystems limited to mountain summits can be considered as naturally fragmented (Riebesell, 1982) and this fragmentation may increase over time by human activities such as agriculture (Andrén, 1994; Ewers & Didham, 2006). Many studies have incorporated LUCC analysis and landscape metrics which seems promising for analyzing LUCC in mountains (i.e. Kintz, Young, & Crews-Meyer, 2006; Zomeni, Tzanopoulos, & Pantis, 2008). In this context, the objective of this study is to test a more comprehensive landscape change analysis in a mountain region by addressing four research questions: (1) What are the main land use and land cover changes? (2) Is there any upward shift in the main land cover classes? (3) How is the spatial pattern of land cover classes changing? And finally (4) How to apply the results in a management context? This framework is applied to the Tropical Andes, which is recognized as a biodiversity hotspot (Myers, Mittermeier, Mittermeier, Fonseca, & Kent, 2000). This region provides a wide range of ecosystem services (Buytaert, Cuesta-Camacho, & Tobón, 2011) to about 100 million people living in both elevated areas and in the surrounding lowlands. Despite the coexistence of human and natural areas dates back to pre-Inca periods (Ellenberg, 1979) it is evident that human activities have seriously encroached natural areas in the last century (Spehn et al., 2006). For example in the Northern Andes, changes in agricultural practices have triggered extensive degradation of vegetation (Sarmiento, Llambi, Escalona, & Marquez, 2003) and intensive grazing and burning caused typical native species to disappear (Premauer & Vargas, 2004). Transformation of landscape functions may also affect its hydrological regulation, severely reducing the water retention capacity of e.g. tropical mountain grassland ecosystems (Buytaert et al., 2006; Podwojewski, Poulenard, Zambrana, & Hofstede, 2002). The proposed framework is tested in the Peruvian Jalca, a tropical alpine grassland ecosystem located between the Northern and Central Andes, above 3000 m for the period 1987–2007. The analysis is based on an object-based classification of Landsat TM images since there is not only evidence of good performance in heterogeneous regions but also in allowing applications in vegetation patchiness or landscape complexity and habitat

41

fragmentation studies (Blaschke, 2010; Newman, McLaren, & Wilson, 2011).

2. Materials and methods 2.1. Study area The study area is located approximately between the 6◦ 30 S and S in the south of the Cajamarca region, Peru. It covers an elevation range between 3000 and 4200 m and encompasses 6307 km2 (Fig. 1). This region is known as Jalca, a tropical alpine grassland ecosystem where natural vegetation typically consists of bunch grasses (Weberbauer, 1945). Climate conditions are drier than the paramos of the Northern Andes but more humid and lower than the punas of the Central Andes (Sánchez-Vega et al., 2005). In fact, many authors consider the Jalca as a transition area between the paramos and the punas as some vegetation from both regions can be found here (Luteyn, 1999; Sánchez-Vega, 1996). However, the Jalca also has more than 40 endemic species of plants (Hensold, 1999) such as Laccopetalum giganteum, Ascidiogyne sanchezvegae, Calceolaria caespitosa, Chuquiraga oblonguifolia and Belloa plicatifolia (SánchezVega, 1996). The Jalca is limited naturally in the lower border by montane forest and shrubland, especially in the western part of the study area. The montane forest (a broad-leaf evergreen forest) is part of the last continuous montane cloud forest of the Peruvian western Andes and it presents high endemism of plants (i.e. Asteraceae, Solanaceae, Loasaceae) and birds (Weigend, Rodriguez, & Arana, 2005). In the study area, total annual precipitation ranges from 650 mm in the west to 1370 mm in the east and mean annual temperature ranges from 5.7 ◦ C in the upper areas to 16.3 ◦ C in the lower areas (Hijmans, Cameron, Parra, Jones, & Jarvis, 2005). A main effect of human occupation in this area is that agriculture has largely replaced the natural Jalca vegetation (Sánchez-Vega et al., 2005), however this has not yet been quantified. Land tenure has been suggested as one of the main drivers of land use change for the tropics (van Gils & Loza Armand Ugon, 2006). In our study area, changes in land tenure such as those promoted by the agrarian reforms of 1963 and 1970 have lead to more occupancy of the Jalca (Sanchez, 2003). As a consequence of these reforms and before forced expropriation, the former owners of large farms sold smaller pieces to peasants, especially those areas with less productive value, which include the Jalca. New owners used these areas, mainly for intensive grazing of beef cattle (Sanchez, 2003). Later on, the presence of two important dairy factories in Cajamarca triggered intense overgrazing of Jalca areas as well (Sanchez, 2003) and promoted the sowing of pastures. Nowadays croplands typically consist of crop fields of potatoes, wheat, barley, and peas among others. Parcels are either irrigated or rainfed and fallow agriculture is a common practice. According to the last agricultural census of 1994, sowed pastures represent 5.4% of the cropland area of the Cajamarca’s highlands (INEI, 2012) but it is very likely that this number has increased considerably in the last years. Nearly 60% of the farmers from the Cajamarca region own parcels of less than 2 ha with low scale production (Zegarra & Calvelo, 2006) and for 1994, only 51% of Cajamarca’s farmers had legal land titles (INEI, 2012). For the remaining area, farmers are currently gaining titles or they belong to a traditional form of tenure (communal lands) or cooperatives. For 2001, 79 of the 107 recognized peasant communities had land titles (Owen, Morelli, & Hernández, 2007). More recently, two other human activities have affected the Jalca integrity in the study area. Mining has developed more intensively since the beginning of 1990 (Sanchez, 2003). Some mountain tops above 3600 m are being exploited as opencast mining for gold extraction. A first impact is an extensive land use and land cover change, and a second one is the use of cyanides for leaching. The 7◦ 30

42

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49

Fig. 1. Land use and land cover change between 1987 and 2007.

latter is a potential risk of water and soil contamination. Another recent problem relates to water availability, which may lead to conflicts between mining company and local farmers (Sanchez, 2003). Finally, exotic pine and eucalyptus tree plantations are becoming more widespread in the Jalca and in general in the Andean region, however, no clear quantification of this process exists. Tree plantations in Cajamarca started more massively in 1974 in private and communal lands mainly promoted by the Government and mostly with the financial support of international agencies (Llerena, Hermoza, & Llerena, 2007). Wood extraction is the main purpose of this afforestation process. Finally, there was only one protected area in our study area at the time of the analyzed period, the Coto de caza Sunchubamba

established in 1977. It was originally created to manage fauna for sport hunting and to protect forest but it also includes Jalca areas (Fig. 1). 2.2. Land cover classification Two orthorectified Landsat 5TM images (path09 row65) were downloaded from the Global Land Cover Facility site (http://www. landcover.org/index.shtml) for the analysis. The images were taken on 23/06/1987 and 01/08/2007 during the dry season, which minimizes possible differences in the reflectance of the vegetation (Martinuzzi, Gould, & Ramos González, 2007) in a postclassification comparison. The seven bands were stacked and then

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49 Table 1 Description of land cover classes. Land cover class

Description

Lake

Water bodies; cross-checked with topographical map (1:25,000) Natural grasslands, mostly bunch grasses. The Jalca occurs in areas above 3000 m Forest fragments (tree cover >70%, height of canopy between 20 and 35 m) and shrubs (canopy height between 4 and 5 m) occur between 2600 and 3200 m (Cuesta, Peralvo, & Valarezo, 2009; Weigend et al., 2005). The small fragments of native Polylepis forest could not be detected Crop fields (i.e. potatoes, wheat, corn, barley, peas), sown pastures (i.e. alfalfa) and fallow agriculture Areas with no vegetation cover. Appears very bright on the Landsat image and has irregular artificial boundaries with the adjacent land cover units Exotic species, typically Pinus sp. and Eucalyptus sp. Parcels of these trees should have reached a certain age level in order to be spectrally separated Areas with no vegetation cover others than mining such as roads

Jalca Montane forest and shrubland

Agriculture Mining

Tree plantation

Bare surface

reprojected to UTM zone 17S, datum WGS84. First we applied a multiresolution segmentation algorithm (Baatz & Schäpe, 2000) using Definiens Developer 7 software (Definiens A.G., 2007). In this process individual pixels are clustered into image objects based on three criteria: reflectance values (colour), shape and compactness (Benz et al., 2004). Each of these criteria is assigned with different weights where the sum of the three is 1. In this way the segmentation process considers both sensor measurements and the context within the scene (Benz et al., 2004). In the Andes the irregular topography results in irregular shapes for the various units of land cover categories, therefore weight values for shape (0.3) and compactness (0.2) were low, giving more importance to reflectance (0.5). An iterative trial and error procedure of these parameter values showed that these settings resulted in image objects with the best visual match with the land cover units in both images. After the segmentation, an object-based supervised classification was applied to each image using the nearest neighbour classifier. The output classification was then analyzed once again through visual inspection, to correct misclassified objects. The following land cover classes were identified: lake, Jalca, montane forest and shrubland, agriculture, mining, tree plantation (exotic species) and bare surface (Table 1). Clouds and shadows could be classified and masked out quite easily from both images to allow proper comparison. Following Congalton (1991), an accuracy assessment was performed by comparing the classification results against a reference dataset. This reference dataset consists of 290 points randomly placed over the 1987 and the 2007 images. For the 2007 image we crosschecked the reference points with more detailed images from Google Earth, when available and with the authors’ previous knowledge of the area. Information from a mining concession map (titles given before 1987) was crosschecked with mining points for 1987. For this same year, tree plantation control points were placed in areas were information of established reforestation areas before 1987 was available (Ansión, 1986; Llerena et al., 2007). For the 2007 image, we used 40 additional GPS points taken mostly randomly during a fieldwork campaign between May and June of 2009 to match the dry season of the 2007 image. A few GPS points were taken specifically in tree plantations and montane forest and shrubland. Of these 40 points, 18 were located in the Jalca, 1 in montane forest and shrubland, 15 in agriculture and 6 in tree plantation. The User’s, Producer’s and the Overall accuracy and KHAT statistics were calculated.

43

2.3. Patch metrics and class metrics Patches for all land cover classes were extracted from the classified land cover maps of 1987 and 2007 using the FRAGSTATS software (McGarigal & Marks, 1995). The extraction of patches was based on the 8 pixels rule where both orthogonal and diagonal adjacencies defined continuous patches (McGarigal & Marks, 1995). Most metrics used to measure patch characteristics are redundant but some authors have grouped them in relatively independent groups (Cain, Riiters, & Orvis, 1997; Hargis, Bissonette, & David, 1998). We selected patch size (AREA), perimeter complexity (FRAC: approaches 1 for very simple perimeters and 2 for highly convoluted perimeters) and distance between patches (ENN: Euclidean nearest neighbour) from those groups. Number of patches (NP), the largest patch index (LPI: largest patch area divided by total study area) and the average of the patch metrics were calculated per cover class as well. 2.4. LUCC and spatial pattern analysis Land use and land cover changes were analyzed through the percentage of change for the period 1987–2007 with regard to the area of 1987 for each land cover class following Eq. (1) x=

A2 − A1 × 100 A1

(1)

where x is the percentage of change (% change), A2 and A1 are the area of land cover class at time t1 and t2 (1987 and 2007 respectively). In order to make comparisons with other studies we calculated the annual rate of change for each land cover class using the formula derived from the Compound Interest Law (Puyravaud, 2003) shown in Eq. (2): r=

1 × ln t2 − t1

A  2

A1

(2)

where r is the annual rate of change (%/yr), A2 and A1 are the area of land cover class at time t1 (1987) and t2 (2007). A conversion matrix was calculated to analyze changes between classes for the period 1987–2007, and the overall and percentage change for each class. Changes along an elevation gradient were analyzed for the Jalca and agriculture. The study area was split in 200 m elevation intervals, ranging from 3000 to 4300 m. For each interval we calculated the percentage of agriculture and Jalca coverage for 1987 and 2007. In addition, the percentages covered by new agricultural areas for 2007 and regeneration areas of Jalca for 2007 were also calculated per elevation interval. Finally, changes in spatial patterns for the 2 natural land cover classes (Jalca, montane forest and shrubland) and the 3 human land cover classes (agriculture, mining and tree plantation) were evaluated using the patch metrics calculated for each class for both years. We looked for statistically significant changes in average values of each patch metric between years. Statistical significance was assessed with Mann–Whitney U test to a significance level of 0.05. P-values were corrected for multiple comparisons using the Bonferroni–Holm correction (Wright, 1992). 3. Results 3.1. Classification The classification results are presented in Table 2. The overall accuracy of the 1987 classification is 82.1% while the one of the 2007 classification is 80.3%. Tree plantation has the highest percentage of coincidence with the reference data (user’s accuracy) for both years, while mining has also a very high percentage for 1987 (100%) and montane forest and shrubland for 2007 (83.3%).

44

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49

Table 2 Accuracy assessment of the classification for Landsat 5TM of 1987 (a) and 2007 (b). Reference data

Classification Cloud and shadow

(a) Cloud and shadow Jalca Montane forest and shrubland Agriculture Mining Tree plantation Total User’s accuracy Producer’s accuracy Overall accuracy KHAT (b) Cloud and shadow Jalca Montane forest and shrubland Agriculture Mining tree plantation Total User’s accuracy (%) Producer’s accuracy (%) Overall accuracy (%) KHAT

Total Jalca

Montane forest and shrubland

Agriculture

1 20 1

13 9 106

22

2 130

Mining

Tree plantation

6 98 10 16

4 6 100 100 82.1 0.72

124 79 87.5

90.9 51.3

83 2 17

1 10 1

22 12 149

12

6 189

81.5 86.2

4 100 100

4 4 100 66.7

3

1

2

1

6 4 75 100 80.3 0.67

102 81.4 76.1

83.3 41.7

However, for the producer’s accuracy, tree plantation and montane forest and shrubland have the lowest values. The producer’s accuracy shows that mining has the highest values (100% for 1987 and 2007). Our classification performs better in identifying agriculture (producer’s accuracy values above 85%) than the Jalca, which has a higher producer’s accuracy value for 1987 (88%) than for 2007 (76%). Misclassification of Jalca and agriculture is related to spectral confusion between these classes. The main misclassification occurs between the class montane forest and shrubland and the classes Jalca and agriculture. 3.2. LUCC and changes along the elevation gradient There has been a considerable loss of natural areas, especially the Jalca areas between 1987 and 2007 (>750 km2 , −25.5% of change with regard to 1987, −1.5%/yr) (Table 3 and Fig. 1). The main Jalca loss is due to the expansion of agriculture and then tree plantation and mining areas. However, in relative terms, the latter two show the largest increase of the human land cover classes (Table 3). Montane forest and shrubland class shows the major loss in relative terms as well, and has been principally replaced by agriculture (Table 3). Agriculture increases in more than 710 km2 , representing a total change of 26% at an annual rate of 1.2%/yr and covering more than half our study area for 2007 (Table 3). The majority of the class agriculture is located between 3000 and 3200 m, however, the major increase is registered for the 3600–3800 m interval (Table 4). In this interval, new agricultural land represents 3.9% of the study area for 2007 while only 1.6% of the study area represents this change in the 3000–3200 m interval (Table 4). The largest Jalca areas in 1987 (11.5% of the study area) are also concentrated in the 3600–3800 elevation range. The total change for one class is the result of dynamics in LUCC. For example, more than 860 km2 have experienced a change from Jalca to agriculture class from 1987 to 2007 but this is partly compensated by nearly 250 km2 that change from the agriculture class

78.8 88.7

8 75 100

6 112 39 123 4 6 290

14 15

3 109 24 168 6 20 330

93.3 70

to the Jalca class (Table 3). This regeneration process is most evident in the 3400–3800 m range (2% of the study area, Table 4), but it is considerably lower than the Jalca loss due to transformation into agriculture for the same elevation interval (6.9% of the study area, Table 4). Areas of Jalca loss due to agriculture expansion are located principally along the eastern slopes in the headwaters of the tributaries ˜ of the Maranón River, while the area less affected by agriculture expansion is the south (Fig. 1). Mining expansion is mostly concentrated in the north of Cajamarca city where gold extraction is the main activity of Yanacocha Mining Company. Some other scatter mining areas are located further north around Hualgayoc town. To the west of Yanacocha mine, tree plantation is concentrated (Fig. 1). This area belongs to the Cooperativa Agraria de Trabajadores Atahualpa Jerusalén (Granja Porcon). Both mining and tree plantation areas are located in the centre of the study area where the Jalca narrows. Areas where montane forest and shrubland are lost are mostly located in the western slopes. The presence of montane forest and shrubland is registered in the literature for the northwest of our study area; however the presence of some clouds did not allow their detection in this region.

3.3. Changes in spatial patterns The number of Jalca patches increases from 744 in 1987 to 1196 in 2007. Montane forest and shrubland patches decrease from 1117 in 1987 to 1094 in 2007. The mean patch size of Jalca and montane forest and shrubland are significantly smaller in 2007 than in 1987 as the mean distance between patches of Jalca (Table 5). In contrast, the mean distance between montane forest and shrubland patches increases and the patch shape is less complex for 2007 than for 1987 (Table 5). The largest patch of Jalca for 1987 represents 22.8% of the study area (LPI) but it reduces drastically to 4.3% in 2007. Similarly, the LPI of montane forest and montane shrubland decreases from 0.32% to 0.10%.

−1.5 −2.8 −25.5 −42.7

26.0 499.6 1066.3 711.7 14.9 0.0

46.9 5.5

43.2 0.2 0.2 0.0 0.1 3.9

100.0 100.0 6307.0 100.0

2726.2 12.4 10.5 2.4 5.5 244.9

2960.7 344.5

1.2 9.0 12.3 10.6 0.7 0.0

Annual rate of change (%/yr) % Change

Change 1987–2007 % Study area 1987 Area 1987 (km2 )

45

In the case of human land cover classes, the mean patch size significantly increases for all classes between 1987 and 2007 (Table 5). The number of patches increases as well but more dramatically for tree plantation (from 62 to 588) than for mining (from 37 to 102) and agriculture (from 3498 to 3525). Only for agriculture mean FRAC values are significantly lower in 2007 than in 1987, indicating that patch border shapes are less complex in 2007 than in 1987. In terms of connectivity (mean distance between patches), only agriculture shows significant changes. Patches of agriculture are closer together in 2007 than in 1987. Finally, and contrary to what happens for natural areas, LPI values increase for human land cover classes. The largest patch of agriculture represents 6.5% of the study area for 1987 but for 2007 this increases to 28%. In the case of mining LPI increases from 0.05% to 0.9% and for tree plantation it increases from 0.03% to 0.6%.

4. Discussion

244.9 3.9 6.3 0.1 19.5 0.3 122.3 1.9 74.2 1.2 3436.3 54.5 197.3 3.1 2206.1 35.0 Area 2007 (km2 ) % Study area 2007

0.0 0.0 0.0 0.0 0.0 244.9 0.4 0.0 0.0 0.0 3.5 0.0 4.5 0.2 0.0 0.5 0.0 0.0 20.6 0.0 6.4 0.0 0.0 0.0 9.7 10.2 0.3 0.3 0.1 0.0 24.4 0.0 0.0 0.0 0.0 0.0 248.5 0.8 1.2 0.9 1.4 0.0

2418.1 1.2 2.6 0.7 0.4 0.0

0.0 0.0 2.3 0.1 14.3 0.1 89.1 6.2 52.4 1.2 866.0 147.4 32.0 140.9 1904.7 48.6

1987 Jalca Montane forest and shrubland Agriculture Mining Tree plantation Bare surface Lake Cloud and shadow

Jalca

2007

Montane forest and shrubland

Agriculture

Mining

Tree plantation

Bare surface

Lake

Cloud and shadow

4.1. LUCC and changes in spatial patterns in the Jalca grasslands

Land use/land cover class

Table 3 Conversion matrix of land cover classes between 1987 (rows) and 2007 (columns) (km2 ) where diagonal show areas of no change. The last two columns refer to the total land use and land cover change for 1987 and 2007 for each class.

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49

The Jalca has suffered from the impact of an extensive land use and land cover change for the period 1987–2007, mainly by encroaching agriculture. Its area reduction at a rate of 1.5%/yr is higher than that of the alpine grassland of the Abiseo National Park and its buffer zone (0.33%/yr, period 1987–2001), located approximately 100 km SE of our study area (calculation based on Kintz et al., 2006). Other alpine grasslands with a protected area in the southern Bolivian Andes have shown an increase in area at a rate of 0.1%/yr for the period 1985–2003 (calculation based on Brandt & Townsend, 2006). The only protected area in our study area represents approximately 5% of the total extent which may explain the difference in rates with other areas where protected areas have larger representation. However, in the paramo of Paute (Ecuador) where there is no protected area and an increasing trend in population density the rate of loss is 0.11%/yr, still lower than that of our study area (calculation based on Vanacker, Govers, Barros, Poesen, & Deckers, 2003 period 1989–1995). Different drivers may be responsible for the high rate and they will be further explained when agriculture is discussed. The increase in areas of human activity has not only caused Jalca habitat loss but also a change in its spatial patterns. The Jalca has suffered from fragmentation between 1987 and 2007 (more patches, less patch mean size and less total area). This same process is also observed for the alpine grasslands of Abiseo National Park (Kintz et al., 2006), in despite of the difference in habitat loss rate with our study area. The distance between new patches that results from the fragmentation of larger patches is still small, which contributes to the decrease in the mean distance between patches (Table 5). This fragmentation may affect water storage and supply since large areas of tropical alpine grasslands, such as the Jalca, provide these services for their surroundings (Buytaert et al., 2006). The other natural class, montane forest and shrubland has even a higher rate of loss than the Jalca (−2.8%/yr) totally opposite to the one observed for the Abiseo National Park where forest shows a recovery (Kintz et al., 2006). However the loss of montane forest and shrubland in our study area is lower than that one observed for the forest of Southern Bolivian Altiplano (−12.2%/yr) (calculation based on Brandt & Townsend, 2006). The montane forest and shrubland class do not show a fragmentation process between 1987 and 2007, but the lower number of patches for 2007 than for 1987 may indicate either disappearance or merging of patches. Given that the total area of this class reduces, it seems acceptable that some patches disappear. The significant decrease in patch size may indicate that patches disappearing are the largest ones. The

46

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49

Table 4 Percentage of study area occupied by total Jalca and agriculture area for 1987 and 2007, new agricultural areas for 2007 and Jalca regeneration for 2007 according to elevation interval. Elevation interval (m)

Total agriculture (%)

Total Jalca (%)

New agriculture 2007 (% of study area)

1987

2007

1987

2007

3000–3200 3200–3400 3400–3600 3600–3800 3800–4000 4000–4300

13.4 11.8 10.1 6.3 1.5 0.1

15.4 14.0 12.0 9.2 3.7 0.1

4.8 7.0 10.6 11.5 11.0 2.0

3.8 5.1 7.7 8.1 8.4 1.8

1.6 2.4 3.0 3.9 2.7 0.1

1.3 0.8 0.2 0.0 0.0 0.0

0.7 0.8 1.0 1.0 0.4 0.1

% Study area

43.2

54.4

46.9

35.0

13.7

2.3

3.9

disappearance and/or the reduction of large montane forest and shrubland patches are of major concern, given that they are part of the last continuous areas of the Peruvian western montane forest located in the North of Peru (Weigend et al., 2005). The forest relicts in our study area represent part of the upper limit of this former forest extent. In addition, the observed increase of distance between remaining patches of forest and shrubland may impact the future population survival. Individuals of these populations may be genetically adapted to broader climatic conditions and represent the source for future migration under climate change (Valencia, Urrego, Silman, & Bush, 2010). However, larger distances between forest patches may make dispersal more difficult and may impact migration. Therefore, urgent measures are necessary to protect this important forest relict. Natural habitat loss is mainly caused by agricultural expansion at a rate of 1.2%/yr which involves not only crops but also sown pastures and fallow agriculture. In the Cajamarca region one of the reasons of agricultural expansion is related to an increasing market for milk products which demands sowing of pastures for cattle. Currently, two milk factories are present in the Region (Nestle Company and GLORIA S.A.) (Escurra, 2001) and milk production increased from 88,052 tonnes in 1996 to 208,580 tonnes for 2003 (INEI, 2004). In addition, cultivated crop surface has increased for whole Cajamarca region with a higher rate (4.7%/yr) than in any other Andean region and this is an even higher value than that one of the total country (2.8%/yr) for the period 1993–2007 (calculation based on MINAG, 2008). These figures support our results and confirm the importance of agricultural expansion. In comparison to other studies in the upper Andes, our results contrast with

Was Jalca in 1987

Jalca regeneration (% of study area)

Was Montane forest and shrubland in 1987

those showing a decrease in agricultural areas either in zones with protected areas (Brandt & Townsend, 2006; Kintz et al., 2006) or those with no protection (Vanacker et al., 2003) for about a similar period. However for previous periods such as that of 1962–1989 in the Paute catchment of Ecuador, agriculture increased at a rate of 0.21%/yr (Vanacker et al., 2003). Interestingly, new agricultural areas in our study area are located principally in the upper areas, similarly to the northern Ecuadorian Andes for the period 1960–1990 (López Sandoval, 2004). In that study, in two of the three study sites, the upper boundary of agricultural areas moved towards the 3600–3800 m elevation interval and one shifted to the 3800–4000 interval. In areas of central and southern Ecuador an upslope agriculture displacement reached 3900 m (Hess, 1990). The results of López Sandoval, Hess and ours suggest the upslope agriculture displacement is a regional pattern for the last 30 or 40 years in the Northern Andes. The main reasons for this expansion might be associated with socio-economic factors such as the increase of population density and changes in land tenure and migration (López Sandoval, 2004; van Gils & Loza Armand Ugon, 2006). However, temperature has increased in the Andean Region between 1950 and 1998 (Vuille et al., 2003) and this may have stimulated crop cultivation also, at higher elevations. The landscape shows a highly dynamic LUCC. In addition to the large conversion from Jalca to agricultural areas, our results also show that nearly 250 km2 of agricultural fields were converted to Jalca areas in the period 1987–2007. These findings are not enough to estimate the fallow length, therefore an analysis with a higher temporal resolution is recommended. This is important for farm

Table 5 Mean of patch metrics for each land cover class. Differences between 1987 and 2007 were evaluated through the Mann–Whitney U test using Bonferroni–Holm correction. Average patch metric

land cover class

Mean patch area

Jalca Montane forest and shrubland Agriculture Mining Tree plantation

Mean FRAC values

Mean Euclidean distance between patches

* P < 0.05 (unadjusted P-value). ns, P ≥ 0.05 (unadjusted P-value).

Jalca Montane forest and shrubland Agriculture Mining Tree plantation Jalca Montane forest and shrubland Agriculture Mining Tree plantation

1987 398 31 78 33 17 1.100 1.085 1.071 1.090 1.076 261 238 166 1683 1893

2007

Mann–Whitney test

184 18 97 73 21

*

1.098 1.078 1.069 1.067 1.067 202 259 131 583 501

* * * *

ns * *

ns ns * * *

ns ns

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49

management because a fallow length of at least 12 years is necessary to recover vegetation physiognomy in alpine ecosystems (Sarmiento et al., 2003). Although natural species richness will not fully recover within this 12 years period (Sarmiento et al., 2003), the regeneration of vegetation physiognomy can help to maintain the water storage function of the Jalca. Agricultural activities have the highest absolute expansion in area, but in relative terms the cover of tree plantation and mining increases most strongly. More than 89 km2 of Jalca were converted to new tree plantation mainly due to pine plantation in the land of an Agrarian Cooperative known as Granja Porcon located in the centre of our study area. Pine species are the most popular for afforestation, especially for their resistance to cold at high elevation (Martinez Meier, Mondino, & Gallo, 2005) and for its economic value since wood production is a great source of income for local population in alpine grassland areas (Hofstede, 2003). Afforestation with pine and eucalyptus is not only common in Cajamarca but also in the paramos of Venezuela, Colombia and Ecuador (Hofstede, 2003). In an Ecuadorian catchment the afforestation process was observed at a much lower rate than in our study area (0.53%/yr) for the period 1989–1995, however the major afforestation occurred from 1962 to 1989 at a rate of 7.3%/yr (Vanacker et al., 2003), still, a lower rate than that one observed for our study area (12.3%/yr). Between 1978 and 1985 afforestation was promoted by the Peruvian Government mostly funded by the Interamerican Development Bank and Agrarian Bank that supported credits for reforestation (Llerena et al., 2007). These credits were adopted mainly by individual peasants but the large increase in tree plantations in the last 20 years is due to the expansion of Granja Porcon. The effects of this plantation on soils and hydrology have not yet been studied for Cajamarca. However, a study in the paramo of Ecuador demonstrated a reduction of nearly 50% of water yield in afforested areas in comparison to areas with natural grasslands ˜ (Buytaert, Iniguez, & De Bièvre, 2007). The effects of tree plantation on biodiversity were also studied in Colombia, where characteristic species of sub-paramo disappeared in areas with more pine cover (van Wesenbeeck, van Mourik, Duivenvoorden, & Cleef, 2003). The area of mining activities increases by 500% (9%/yr) between 1987 and 2007. Both Jalca and agriculture were converted to mining with more emphasis on the Jalca areas. However, our results only show the opencast extraction. The off site effects of the mining are unknown. According to MISA (2002, cited by Bury, 2005) Yanacocha Mining Company acquired more than 11,000 ha between 1992 and 2000 from private landholders and local communal organizations. Gold production (content of fine gold) was 41,350 kg for 1998 and 88,583 kg for 2003 for this Company (INEI, 2004). The increase in productivity is obviously related to the opencast mining area expansion, especially north to Cajamarca city. The presence of Yanacocha has brought some positive aspects such as economic development (i.e. infrastructure, credit programmes) or human development (i.e. more education and technical training) (Bury, 2005). Nevertheless other aspects such as disturbance of the natural environment (i.e. a reduced water resource availability, lower quality of water and soil, disturbance of potential geoconservation areas) and social aspects (i.e. conflicts with local communities, decrease number of local organizations and leadership) are negatively impacted (Bury, 2005; Seijmonsbergen et al., 2010). For tree plantation, mining and agriculture the higher number of patches combined with an increase in total area indicate that the number of patches increased from 1987 to 2007. These changes are not a result of fragmentation but occur through the creation of new patches since the average mean patch size increases. More regular areas (lower FRAC values) suggest that marginal agriculture might change towards a more extensive agriculture. In addition,

47

agricultural patches are closer between them than before, indicating an expansion near already established agricultural areas. 4.2. Assessing the method and implications for management Classification using an object-based approach performs well for the Jalca class but shows relatively low accuracy values for montane forest and shrubland. The producer’s accuracy is rather low but the user’s accuracy is high, therefore the area of montane forest and shrubland is underestimated. Since they only represent less than 4% of the study area the priority was focused on finding the best performance for the most important classes (Jalca and agriculture) as we defined this as our end-purpose. The class with the second lowest value of producer’s accuracy is tree plantation, suggesting also underestimation of the real area of this class as well. Given that peasants use tree plantation as thin hedges or plant trees as small patches, our classification was not able to identify some of them and this may explain the confusion between tree plantation and agriculture. On the contrary, large plantations were identified without problems. The Jalca is more accurately identified for 1987 than for 2007 which suggest that the reported area for 2007 may be underestimated and misclassified mostly as agriculture. However, agriculture is also misclassified as Jalca, although, to a lesser degree. Therefore possible bias in the total observed changes should be minimal. The final classification shows not only good results for the Jalca and agriculture but also for mining areas. There were minor misclassifications that incorrectly suggested transitions from mining to Jalca or from mining to agriculture within the conversion matrix; however the misclassified area represents roughly 1 km2 in each case, which is small relatively to our study area. Therefore the object-based approach in combination with an accuracy assessment and a post classification by visual inspection seems to work well in this heterogeneous mountainous environment. Our results may contribute to a well-planned ecosystem management to prevent more alterations of the natural environment in this region of the Tropical Andes or to minimize the impacts of inevitable human expansion. Recently the Cajamarca Regional Government finished the Ecological and Economical land use zoning of Cajamarca (Zonificación Ecológica y Económica) with the objective of delimiting the space according to the potentialities and limitations of each territory (Gobierno Regional de Cajamarca, 2011). Despite some of the remaining Jalca areas are identified as “potential for livestock grasslands, tree plantation and mining”, most of the remaining Jalca is identified as potential for different categories of protection, conservation and/or restoration. Our results may support the zoning by offering a dynamic perspective, based on the observed land use and land cover changes, in which, the inclusion of landscape metrics and elevation range allows for an adequate spatial management strategy especially under climate change (Young, 2009). In addition, they can be also be used as input to forecast future agriculture expansion. This method of spatial analysis may rapidly identify expansion of agricultural areas into the upper natural Jalca zone. It is likely that this expansion will continue but, in a management context, agriculture may be better established in areas where fragmentation of Jalca is already high, instead of close to large patches. In addition, it is important to maintain traditional practices of fallow agriculture which allows rapid recovery of the ecosystem since interaction between succession and fragmentation might influence species extinction dynamics (Collins, Holt, & Foster, 2009) and might negatively affect diversity and species composition in the long term. From a conservation perspective, upper areas of Jalca, especially large patches less close to agriculture may be suitable for conservation of vegetation cover and therefore adequate to preserve water supply functions. However, some areas of highly fragmented

48

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49

Jalca may be important to maintain connectivity and expansion of human activities should be avoided. Narrow connections are typical of the Andes and important for dispersal of alpine species but they are also fragile and prone to fragmentation (Young, 2009). Changes in connectivity between habitat patches may lead either to speciation or extinction (Young, 2009). Isolation can lead to speciation over evolutionary times, developing high endemism such as that observed for plants in the Jalca of Contumaza (Sagástegui, 1988 cited in Sánchez-Vega & Dillon, 2006), clearly isolated from the main Jalca area in the southwest (Fig. 1). However rapid changes in habitat loss may lead to extinction. Our LUCC map for 2007 shows loss of connectivity between large areas mainly by agriculture, but in addition to this, in the centre of our study area the Jalca is threatened due to expansion of tree plantation and mining as well. The total loss Jalca vegetation cover due to mining activities creates a more hostile matrix surrounding the Jalca area which may affect species composition (Tovar, Duivenvoorden, SánchezVega, & Seijmonsbergen, 2012). Therefore those fragmented areas may potentially represent sources of biodiversity to future restoration action. Montane forest and shrubland class deserves special attention since these are relicts of once a bigger and continuous area of this class. More research is needed to adequately select the “most important patches” of montane forest and shrubland for preservation. 5. Conclusions Monitoring landscape changes such as those presented in this study should be stimulated as a useful management tool to identify threatened areas and to design alternative development strategies. In our study area, a notable degradation at landscape level has occurred between 1987 and 2007. There was an overall loss of Jalca areas, montane forest and shrubland. These losses occurred at a higher rate than other alpine grassland or forest types within Peru, Ecuador and Bolivia. Particularly in the upper areas of the Andes, agriculture has replaced the Jalca showing upward shifting of agricultural activities. However, the most important land use and land cover changes in relative terms are linked to tree plantations and mining. Furthermore, the spatial pattern of the Jalca occurrence has been affected; it is suffering from an increasing fragmentation towards smaller patches. An improved ecosystem management may require delimiting areas of human activities expansion such as those where natural areas are already highly fragmented. Exceptionally, highly fragmented Jalca should be conserved if they represent an important connectivity path within the Jalca. This is for instance the case in the centre of our study area. However, generally it is preferable to conserve large patches in the upper areas, not only because of their biodiversity value, but also because they can be invaluable sources of water supply. Especially in mountain ecosystems, the combined information on land cover loss, shifts in land cover across the elevation gradient, and changes in spatial patterns may result in useful information for decision makers. It allows for suggesting suitable areas not only for conservation, but also for human activities expansion, where they will less affect natural areas. We believe that the proposed framework can be suitable to analyze other mountain regions of the world where both landscape planners and landscape conservationists need to have the most complete information to make better decisions. Acknowledgments We would like to thank Alicia Quispe from the Regional Goverment of Cajamarca (director of “Proyecto Zonificación Ecológica

Económica de la región Cajamarca”), Enrique Crisólogo and David Rosario for their help with GIS data. Also, this work was possible thanks to the financial support of Proyecto Páramo Andino implemented by CONDESAN and the Institute for Biodiversity and Ecosystem Dynamics (IBED) of the Universiteit van Amsterdam. References Andrén, H. (1994). Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: A review. Oikos, 71(3), 355–366. Ansión, J. (1986). (The tree and the forest in the Andean Society) El árbol y el bosque en la Sociedad Andina. Proyecto FAO/Holanda/INFOR. [in Spanish] Baatz, M., & Schäpe, A. (2000). Multiresolution segmentation – An optimization approach for high quality multi-scale image segmentation. In J. Strobl, T. Blaschke, & G. Griesebner (Eds.), Angewandte Geographische Informationsverarbeitung XII. Presented at the AGIT-Symposium (pp. 12–23). Salzburg: Wichmann Verlag. Benz, U., Hofmann, P., Willhauck, G., Lingenfelder, I., & Heynen, M. (2004). Multiresolution object-oriented fuzzy analysis of remote sensing data for GIS-ready information. ISPRS Journal of Photogrammetry & Remote Sensing, 58, 239–258. Blaschke, T. (2010). Object based image analysis for remote sensing. ISPRS Journal of Photogrammetry & Remote Sensing, 65, 2–16. Brandt, J. S., & Townsend, P. A. (2006). Land use–land cover conversion regeneration and degradation in the high elevation Bolivian Andes. Landscape Ecology, 21, 607–623. Brooks, T. M., Mittermeier, R., da Fonseca, G. A. B., Gerlach, J., Hoffmann, M., Lamoreux, J. F., et al. (2006). Global biodiversity conservation priorities. Science, 313, 58–61. Burnett, C., & Blaschke, T. (2003). A multi-scale segmentation/object relationship modelling methodology for landscape analysis. Ecological Modelling, 168, 233–249. Bury, J. (2005). Mining mountains: Neoliberalism, land tenure, livelihoods and the new Peruvian mining industry in Cajamarca. Environment and Planning A, 37, 221–239. Buytaert, W., Célleri, R., De Bièvre, B., Cisneros, F., Wyseure, G., Deckers, J., et al. (2006). Human impact on the hydrology of the Andean páramos. Earth-Science Reviews, 79, 53–72. Buytaert, W., Cuesta-Camacho, F., & Tobón, C. (2011). Potential impacts of climate change on the environmental services of humid tropical alpine regions. Global Ecology and Biogeography, 20(1), 19–33. ˜ Buytaert, W., Iniguez, V., & De Bièvre, B. (2007). The effects of afforestation and cultivation on water yield in the Andean páramo. Forest Ecology and Management, 25, 22–30. Cain, D. H., Riiters, K., & Orvis, K. (1997). A multi-scale analysis of landscape statistics. Landscape Ecology, 12, 199–212. Collins, C. D., Holt, R. D., & Foster, B. L. (2009). Patch size effects on plant species decline in an experimentally fragmented landscape. Ecology, 90(9), 2577–2588. Congalton, R. G. (1991). A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment, 37, 35–46. Cuesta, F., Peralvo, M., & Valarezo, N. (2009). (The montane forest of the Tropical Andes) Los bosques montanos de los Andes tropicales. La Paz/Lima/Quito: Ecobona. [in Spanish]. Definiens A.G. (2007). Definiens developer. München, Germany: Definiens A.G. Retrieved from http://www.definiens.com Ellenberg, H. (1979). Man’s influence on tropical mountain ecosystems in South America. Journal of Ecology, 67, 401–416. Escurra, E. M. (2001). Situación de la ganadería lechera en Cajamarca [Status of diary cattle in Cajamarca]. Revista de Investigaciones Veterinarias del Perú, 12(2), 21–26 [in Spanish]. Ewers, R. M., & Didham, R. K. (2006). Confounding factors in the detection of species responses to habitat fragmentation. Biological Reviews, 81, 117–142. Gobierno Regional de Cajamarca. (2011). (Ecological and Economical Land use zoning basis for land management of the department of Cajamarca) Zonificación ecológica y económica base para el ordenamiento territorial del departamento de Cajamarca. Retrieved from http://zeeot.regioncajamarca.gob.pe/ [in Spanish] Guisan, A., Theurillat, J.-P., & Kienast, F. (1998). Predicting the potential distribution of plant species in an alpine environment. Journal of Vegetation Science, 9(1), 65–74. Hargis, C. D., Bissonette, J. A., & David, J. L. (1998). The behavior of landscape metrics commonly used in the study of habitat fragmentation. Landscape Ecology, 13, 167–186. Hensold, N. (1999). Las Angiospermas endémicas del Dpto de Cajamarca, Perú [The endemic Angiosperms of the department of Cajamarca]. Arnaldoa, 6(2), 141–183 [in Spanish]. Hess, C. (1990). ‘Moving up–moving down’: Agro-pastoral land-use patterns in the Ecuadorian Paramos. Mountain Research and Development, 10(4), 333–342. Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. (2005). Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25, 1965–1978. Hofstede, R. (2003). Los páramos en el mundo: Su diversidad y sus habitantes. In R. Hofstede, P. Segarra, & P. Mena (Eds.), Los páramos del mundo. Proyecto Atlas Mundial de los Páramos (pp. 15–38). Quito: Global Peatlands Initiative/NCIUCN/EcoCiencia [in Spanish].

C. Tovar et al. / Landscape and Urban Planning 112 (2013) 40–49 INEI. (2004). (Peru Statistical Compendium 2004) Perú Compendio estadístico 2004. Lima: INEI. [in Spanish]. INEI. (April 1, 2012). III Censo Nacional Agropecuario 1994 (CENAGRO) (3rd National Agricultural Census 1994, CENAGRO). Retrieved from http://www.inei.gob.pe/ [in Spanish]. Kintz, D., Young, K. R., & Crews-Meyer, K. (2006). Implications of land use/land cover change in the buffer zone of a national park in the Tropical Andes. Environmental Management, 38(2), 238–252. Körner, C., & Spehn, E. (Eds.). (2002). Mountain biodiversity: A global assessment. USA: The Parthenon Publishing Group. Kruckeberg, A. R., & Rabinowitz, D. (1985). Biological aspects of endemism in higher plants. Annual Review of Ecology, Evolution, and Systematics, 16, 447–479. Llerena, C., Hermoza, R. M., & Llerena, L. M. (2007). Plantaciones forestales, agua y gestión de cuencas [Forestry plantations, water and watershed management]. Debate Agrario: Análisis y Alternativas, 42, 79–110 [in Spanish] López Sandoval, M. F. (2004). Agricultural and Settlement frontiers in the Tropical Andes: The paramo belt of northern Ecuador, 1960–1990. PhD Thesis. Institut für Geographie and der Universität Regensburg, Germany, Regensburg, Germany. Lucas, R., Rowlands, A., Brown, A., Keyworth, S., & Bunting, P. (2007). Rule-based classification of multi-temporal satellite imagery for habitat and agricultural land cover mapping. ISPRS Journal of Photogrammetry & Remote Sensing, 62, 165–185. Luteyn, J. L. (1999). Paramos: A checklist of plant diversity, geographical distribution, and botanical literature. Memoirs of the New York Botanical Garden The Bronx, USA: New York Botanical Garden Press. ˜ por heladas Martinez Meier, A., Mondino, V., & Gallo, L. (2005). Evaluación de danos tardías en ensayos de procedencias de pino oregón introducidos en el norte de la Región Andino Patagónica Argentina [Evaluation of damage by late frosts in provenances trials of Douglas-fir introduced in Argentinian North Patagonia]. Bosque, 26(3), 113–120 [in Spanish]. Martinuzzi, S., Gould, W. A., & Ramos González, O. M. (2007). Creating cloud-free Landsat ETM+ Data sets in tropical landscapes: Cloud and cloud-shadow removal (p. 12). General Technical Report No. IITF-GTR-32. Portland, Oregon, USA: United States Department of Agriculture. McGarigal, K., & Marks, B. J. (1995). FRAGSTATS: Spatial pattern analysis program for quantifying landscape structure (No. GTR-351). General technical report PNW (USA) (p. 122). Pacific Northwest Research Station, Portland, OR, USA. MINAG, D. G. de I. A. (2008). Boletín Estadística Agraria mensual – Julio 2008 (Monthly Agrarian Statistics Bulletin – July 2008) (p. 147). Lima: Ministerio de Agriculltura. Retrieved from http://www.minag.gob.pe/portal/download/pdf/ herramientas/boletines/boletineselectronicos/estadisticaagrariamensual/2008/ BemsaJulio2008.pdf [in Spanish]. Myers, N., Mittermeier, R., Mittermeier, C., Fonseca, G., & Kent, J. (2000). Biodiversity hotspots for conservation priorities. Nature, 403, 853–858. Newman, M., McLaren, K. P., & Wilson, B. S. (2011). Comparing the effects of classification techniques on landscape-level assessments: Pixel-based versus object-based classification. International Journal of Remote Sensing, 32(14), 4055–4073. Owen, R., Morelli, J., & Hernández, R. (2007). Evaluación del potencial de consorcios para la reforestación en la Sierra peruana (Potential Consortium assessment for the Peruvian Andes reforestation) (No. 07-014 CP-PER). Programas y proyectos Perú (p. 60). FAO/WORLD BANK [in Spanish]. ˜ Penuelas, J., & Boada, M. (2003). A global change-induced biome shift in the Montseny mountains (NE Spain). Global Change Biology, 9(2), 131–140. Podwojewski, P., Poulenard, J., Zambrana, T., & Hofstede, R. (2002). Overgrazing effects on vegetation cover and properties of volcanic ash soil in the paramo of Llangahua and La Esperanza (Tungurahua Ecuador). Soil Use and Management, 18, 45–55. Premauer, J., & Vargas, O. (2004). Patrones de diversidad en vegetación pastoreada y quemada en un páramo húmedo (Parque Natural Chingaza, Colombia) (Diversity patterns in grazed and burned vegetation in a humid bamboo paramo (Chingaza Natural Park, Colombia). Ecotropicos, 17(1–2), 52–66 [in Spanish] Puyravaud, J. P. (2003). Standardizing the calculation of the annual rate of deforestation. Forest Ecology and Management, 77, 593–596.

49

Riebesell, J. F. (1982). Arctic-alpine plants on mountaintops: Agreement with island biogeography theory. The American Naturalist, 119(5), 657–674. Sanchez, P. E. (2003). Perú (Peru). In R. Hofstede, P. Segarra, & P. Mena (Eds.), Los páramos del mundo. Proyecto Atlas Mundial de los Páramos (pp. 159–204). Quito, Ecuador: Global Peatlands Initiative/NC-IUCN/EcoCiencia [in Spanish]. Sánchez-Vega, I. (1996). Aspectos florísticos de la jalca y alternativas de manejo sustentable [Floristic aspects of the Jalca and alterntives for sustainable management]. Arnaldoa, 4(2), 25–49 [in Spanish]. Sánchez-Vega, I., Cabanillas, M., Miranda, A., Poma, W., Díaz, J., Terrones, F., et al. (2005). (The Jalca, the cold ecosystem of northwest Peru – Biological and ecological fundamentals) La Jalca, el ecosistema frío del noroeste peruano – Fundamentos biológicos y Ecológicos (1st ed.). Lima, Peru: Yanacocha – Geográfica EIRL. [in Spanish]. Sánchez-Vega, I., & Dillon, M. O. (2006). Jalcas (Jalcas). In R. M. Moraes, B. Øllgaard, L. P. Kvist, F. Borchsenius, & H. Balslev (Eds.), Botánica Económica de los Andes Centrales (pp. 77–90). La Paz, Bolivia: Universidad Mayor de San Andrés [in Spanish]. Sanz-Elorza, M., Dana, E. D., & Gonzales, A. (2003). Changes in the high mountain vegetation of the Central Iberian Peninsula as a probable sign of global warming. Annals of Botany, 92, 273–280. Sarmiento, L., Llambi, L. D., Escalona, A., & Marquez, N. (2003). Vegetation patterns regeneration rates and divergence in an old-field succession of the high Tropical Andes. Plant Ecology, 166, 63–74. Seijmonsbergen, A. C., Sevink, J., Cammeraat, L. H., & Recharte, J. (2010). A potential geoconservation map of the Las Lagunas area, northern Peru. Environmental Conservation, 37(2), 107–115. Smits, P. C., Dellepiane, S. G., & Schowengerdt, R. A. (1999). Quality assessment of image classification algorithms for land-cover mapping: A review and a proposal for a cost-based approach. International Journal of Remote Sensing, 20(8), 1461–1486. Spehn, E., Liberman, M., & Körner, C. (2006). Land use change and mountain biodiversity. USA: CRC Press Taylor & Francis Group. Tovar, C., Duivenvoorden, J. F., Sánchez-Vega, I., & Seijmonsbergen, A. C. (2012). Recent changes in patch characteristics and plant communities in the Jalca grasslands of the Peruvian Andes. Biotropica, 44(3), 321–330. Valencia, B. G., Urrego, D. H., Silman, M. R., & Bush, M. B. (2010). From ice age to modern: A record of landscape change in an Andean cloud forest. Journal of Biogeography, 37, 1637–1647. van Gils, H. A. M. J., & Loza Armand Ugon, A. V. (2006). What drives conversion of tropical forest in Carrasco province, Bolivia? Ambio, 35(2), 81–85. van Wesenbeeck, B. K., van Mourik, T., Duivenvoorden, J. F., & Cleef, A. (2003). Strong effects of a plantation with Pinus patula on Andean subpáramo vegetation: A case study from Colombia. Biological Conservation, 14, 207–218. Vanacker, V., Govers, G., Barros, S., Poesen, J., & Deckers, J. (2003). The effect of shortterm socio-economic and demographic change on landuse dynamics and its corresponding geomorphic response with relation to water erosion in a tropical mountainous catchment Ecuador. Landscape Ecology, 18, 1–15. Vuille, M., Bradley, R., Werner, M., & Keimig, F. M. (2003). 20th century climate change in the Tropical Andes: Observations and model results. Climatic Change, 59, 75–99. Weberbauer, A. (1945). (The Peruvian Andean vegetation) El mundo vegetal de los Andes peruanos. Lima, Peru: UNALM, Ministerio de Agricultura. [in Spanish]. Weigend, M., Rodriguez, E. F., & Arana, C. (2005). The relict forests of Northwest Peru and Southwest Ecuador. Revista Peruana de Biología, 12(2), 185–194. Wright, P. (1992). Adjusted p-values for simultaneous inference. Biometrics, 48(4), 1005–1013. Young, K. R. (2009). Andean land use and biodiversity: Humanized landscapes in a time of change. Annals of the Missouri Botanical Garden, 96(3), 492–507. Zegarra, E., & Calvelo, D. (2006). (Cajamarca: Guidelines for an agricultural regional policy) Cajamarca: Lineamientos para una política regional de agricultura. Lima, Peru: Contribuciones para una visión del desarrollo de Cajamarca. [in Spanish]. Zomeni, M., Tzanopoulos, J., & Pantis, J. D. (2008). Historical analysis of landscape change using remote sensing techniques: An explanatory tool for agricultural transformation in Greek rural areas. Landscape and Urban Planning, 86, 38–46.