Translator Disclaimer
1 August 2003 A Landscape Perspective on Biodiversity Conservation
Author Affiliations +

The aim of this article is two fold: first, to depict key species and relate them to landscape units (LUs) to define habitats; second, to generate a sound network of protected areas to ensure the functional integrity of the ecosystem. In the present study, 6500 vertebrate and vascular plant species records were gathered into a database. A total of 137 sampling units were surveyed to verify key species and depict discrete LUs (scale 1:25,000). Using multivariate statistics (detrended correspondence analysis and canonical correspondence analysis), key species were selected and associated with LUs in a geographical information system. From the 1162 species recorded, 122 were identified as key species based on their endemicity and conservation status (12 amphibians, 42 reptiles, 37 birds, 11 mammals, and 20 vascular plants). Volcanic bodies and Holocene lava flows contain most key species but harbor less species overall, whereas mixed forest, meadows and crops, foot slopes, and accumulation plains harbor fewer key species but a greater number overall. These patterns were spatially displayed and discussed in light of their role in conservation and participatory management after over 15 years of research.


Large-scale human-induced environmental degradation is currently causing drastic reduction in the quality of life worldwide (Lambin et al 2001). Most conservation efforts, as well as current scientific knowledge, lag far behind in addressing the current rate of environmental deterioration (Myers 2000). Given this situation, there is a need for innovative approaches to retard current loss of biodiversity (Velázquez et al 2001). Traditionally, conservation has aimed at numerous levels of organization (Chung and Weaver 1994), although major emphasis has been given to single species (Mayr 2000). Sound species conservation relies on habitat maintenance (Velázquez et al 1996; Velázquez and Romero 1999). A habitat approach implies coherent integration of biotic and environmental attributes linked in geographical space (Kirby 1996). But this conceptual framework has not yet been fully implemented because most conservation efforts approach habitat management from a purely ecological viewpoint (Spellerberg 1996), ignoring social driving forces and relations between biotic communities and their physical environment. Hence, there is a need to search for innovative ways of achieving sound, long-term conservation because protected areas alone will not meet the demands of conservation (Hansen et al 1991; Vanclay et al 2001). This is especially relevant in tropical and subtropical regions, where most biodiversity occurs.

A major challenge in the integration of a holistic habitat management approach into conservation of biodiversity is the translation of point data into spatial units. Gap analysis (Scott et al 1993) has been widely used as a more robust approach to provide geographic reference to biological records. This approach aims to predict species distribution ranges by using environmental data such as land cover, assuming that species recorded at a given site may also be encountered at other sites with similar environmental conditions (eg, Butterfield et al 1994; Bojorquez-Tapia et al 1995). Species are considered the major attribute to depict habitats. In landscape ecology, geomorphological features are used to delineate natural entities and can then be used to depict biophysical habitats as discrete units (Velázquez and Bocco 2001). Species and geomorphological data sets together enhance the possibility of more robust delineation of habitats, rather than blocks delineated by Cartesian boundaries.

The present article depicts key species and links them to landscape units (LUs) to generate a sound network of protected areas to ensure the functional integrity of the ecosystem. This approach also provided a robust basis for further spatial modeling and conservation planning efforts using a participatory approach (Velázquez and Bocco 1994). The approach was tested using environmental data from Neotropical Mexican Plio-Pleistocene volcanic mountain ecosystems and biological data of vertebrates and vascular plant species. The study area is currently experiencing rapid habitat depletion, harbors an outstanding number of species, and has been regarded as crucial for balancing the functional integrity of the watershed of Mexico City.


Study area

The approach was tested in Sierra Chichinautzin and Sierra de las Cruces, Quaternary volcanic units in Central Mexico, 15 km south of Mexico City (Figure 1). These are located between 19°23′N–99°22′W and 19°03′N–98°55′W (extreme coordinates). The elevation ranges between 2500 and 3660 m, and the total surface covers about 900 km2. Leptosol and Regosol soil types prevail. Climate types vary from temperate mild to temperate cool (mean annual temperature, 8°C), and mean annual rainfall is 1000 mm. Temperate vegetation types prevail, dominated by fir, pine, oak, and alder forests. The area experienced drastic landscape transformation at the end of the 19th century, from pristine forest types to forage production (mainly wheat and oat). Only areas with bare soil, rock outcrops, and steep slopes remained forested. Forage demand declined at the end of the 20th century, so the remaining forested areas were expected to encroach. Currently, both traditional agriculture and pristine forests are threatened by urbanization. A thorough description of the study area is presented elsewhere by Velázquez and Romero (1999).

Definition of LUs

Geomorphological units were delineated on a set of 1:20,000 panchromatic black and white aerial photographs; contacts were digitized and geometrically corrected (according to Zonneveld [1995]). The area was stratified into major cover–geomorphological zones. A digital elevation model was created by digitizing and further interpolating data from topographic maps (1:50,000 scale) with a contour interval of 20 m. Slope gradient and slope aspect maps were derived using standard filtering techniques (Burrough 1988), and these, in turn, were used to check for homogeneity within geomorphological units.

Within every geomorphological unit, sampling units were randomly selected. In total, 137 sampling units (625 m2 each) were surveyed to describe all cover types hierarchically. The sampling units were allocated in such a way that all geomorphological units were surveyed proportionally. During fieldwork, geomorphological boundaries depicted from aerial photographs were verified and adjusted as necessary.

Plant communities were distinguished by means of classification analysis. Two nested levels of vegetation clusters, alliances and associations, were depicted and characterized by floristic composition and physiognomy, according to the method of Velázquez and Cleef (1993). Furthermore, 5 environmental variables were measured at each sampling unit: soil moisture, soil depth, elevation, slope steepness, and slope length (Velázquez 1994). Geomorphological units and vegetation clusters (at the alliance level) were used to typify and delineate LUs (Küchler and Zonneveld 1988). Final LUs were plotted on a scale of 1:25,000.

Vertebrate and vascular plant assemblages

To document the number of sympatric species within the study area, 6500 species records of vertebrates and vascular plants were gathered from different biological collections and from those cited in the literature. A species record means a species was found at a specific location and a given time. The collection sources were: Instituto de Biología (IBUNAM) and Museo de Zoología and Plantas Vasculares, Facultad de Ciencias (FCUNAM) of the Universidad Nacional Autónoma de México (UNAM); Escuela Nacional de Ciencias Biológicas (ENCBIPN); Colección de Mamíferos de la Universidad Autónoma Metropolitana (UAMI); and Centro de Investigaciones Biológicas (CIBUAEM) of the Universidad Autónoma del Estado de México.

A total of 846 vascular plant species and 316 vertebrate species were included in the final database obtained from records starting from the late 19th century up to 1995. To focus the conservation effort, all endemic and threatened species were considered as potential “key” species (CIPAMEX 1993; Instituto Nacional de Ecología 1994; Baillie and Groombridge 1996). All species records from potential key species were pooled into a database. A normalized frequency value of every potential key species per LU was calculated. This value was obtained by plotting every record on maps (scale 1:50,000) where landscape boundaries had been transferred previously. Because not all species were recorded in the same way and not all LUs sampled evenly, a random selection of a proportionally equal number of biological records was made to achieve sound comparisons.

A database on biological records was derived from preexisting collections and literature. In addition, data from fieldwork conducted during the 1980s by the authors were also included. We conducted 3 months of intensive fieldwork (October to December 1997) to survey all LUs evenly and to validate the current presence of potential key species. In this way, our database was updated, and it was comparable among LUs. During fieldwork, 3 different actions took place: the first was to sample vegetation communities using relevés (according to Velázquez and Cleef [1993], the method used since the beginning of the 20th century by most European vegetation scientists to describe, classify, and name vegetation communities); the second was to sight and trap vertebrates occurring at the relevé site; and the third was to validate land form and soil attributes depicting LUs. A detailed description of the results is provided by Velázquez and Romero (1999).

To detect species with similar distribution patterns along LUs, we performed ordination analysis (detrended correspondence analysis, Hill and Gauch 1980). Species with specific habitat requirements and restricted distribution ranges were finally selected as key species. All species grouped within a similar ordination space were considered to have similar environmental requirements (Velázquez and Heil 1996). Individual species records were further treated as species assemblages (SAs).

The 5 land attributes measured during fieldwork (soil moisture, soil depth, elevation, slope steepness, and slope length) were also analyzed using ordination techniques (canonical correspondence analysis [CCA], Ter Braak 1990). This procedure was conducted to detect (dis)similarities among LUs. LUs distributed along similar ordination space have similar environmental affinity. This procedure resulted in landscape clusters (LCs) that were further used as the basis for spatial analysis.

Spatial analysis

Plant communities, derived from clustering species records by ordination techniques, were related to geomorphological (geomorphology and soil) units by point-in-polygon operations (Aronoff 1989) using a geographic information system (Integrated Land and Watershed Information System 1999). The rules implemented in the simulations and used throughout the entire spatial modeling procedure were based on field-verified data (Velázquez and Bocco 1994) and validated habitat distribution patterns of key species (eg, Velázquez and Heil 1996). These rules included significant correlation among SAs and LCs. Those SAs and LCs with equal distribution patterns along ordination space were represented as different spatial units.

Statistical analysis

Floristic affinity among sampling units was calculated through reciprocal averaging (Hill 1979). The (dis)similarity among vegetation clusters (alliances and associations) was measured from dendrograms, where eigenvalues ≥0.250 were considered as significant among vegetation clusters. LCs were then correlated with vegetation clusters through CCA (Ter Braak 1990; Velázquez 1994). Validated criteria were thus obtained to identify differences among SAs and LCs. The score coordinates were corrected according to the percentage of variance explained by each axis. CANOCO version 3.1 (Ter Braak 1990) was used for classification and ordination purposes, and the Monte Carlo Permutation test was used to measure the significance of the eigenvalue along ordination axes. In all tests, 99 permutations were run.


The following 5 major geomorphological units were recognized in the study area: volcanic cone, foot slope, low hills, Holocene lava flow, and accumulation plain. The first occurs under cooler climatic conditions in higher areas, whereas the last is indicative of milder conditions found at lower elevations. Furthermore, 6 vegetation clusters were distinguished (Figure 2). These were the following: (1) Subalpine bunch grassland. (2) Fir forest. (3) Pine forest. (4) Mixed forest and mega-rosettes. (5) Mixed forest and meadows. (6) Crops. As a major output, 9 LUs (LU I–IX) were recognized from the combination of geomorphological and vegetation clusters. In other words, certain vegetation clusters occur in certain geomorphological units, so that the unique relation between geomorphological units and vegetation clusters described landscape patterns. The 9 LUs were integrated into the Integrated Land and Watershed Information System. To achieve geographic correction, polygons drawn on the aerial photographs were georeferenced by linking ground- and photocontrol points.

On the whole, 1162 sympatric terrestrial vertebrate and vascular plant species were recorded in the study area from 1890 to 1996. Of these, 122 were taken as potential key species (about 9% of the total number of species recorded during 116 years of records). The vast majority of amphibians (71%) and reptiles (75%) were reckoned as key species, whereas other taxa contained lower percentage values (mammals, 38%; birds, 19%; vascular plants, 2%). Only a small percentage (12%) of the total number of records was suitable for mapping purposes at a scale of 1:25,000 (Table 1). In other words, 780 records for 122 species were georeferenced within a range of ±1 second latitude and longitude (minimum spatial resolution, 3.5 km2). Additionally, 20 vascular plant species considered endemic were included as key species.

From the ordination analysis, the total cumulative variance explained by the axes reached 40% (axis I = 24%, λ = 0.799 and axis II = 16%, λ = 0.368). Clusters obtained by using the scores of the species along axes I and II suggest that species clusters were distributed along minor environmental differences but were dissimilar enough from one cluster to the other to form significantly different SAs. The clusters followed a species richness gradient, from the group comprising the most species (high biodiversity [HB]) to the group with a medium number of species (medium biodiversity [MB]) and ending with the group that included the fewest species (low biodiversity [LB]). The latter differed significantly from groups HB and MB along axis I. In contrast, significant differences were found among the groups HB and MB along axis II, whereas HB and LB showed a similar distribution along axis II (Figure 3).

LUs were also clustered into 3 groups (Figure 4). The total cumulative variance explained by the axes reached 64% (axis I = 37%, l = 0.798 and axis II = 27%, l = 0.494). The first LC (LC I) was formed by the mixed-forest mega-rosettes on low hills and Holocene lava flows (LUs VI and VII) and meadows and crops on accumulation plains (LUs VIII and IX). The second cluster (LC II) comprised fir and pine forests on foot slopes (LUs IV and V), and LC III included subalpine bunch grassland and fir forest found on volcanic cone and foot slopes.

Discussion and conclusions

Habitat mapping

Habitat definition on a species basis is not feasible worldwide because data (quality and quantity) on current species distribution patterns for most species are scanty (Grehan 1993; Bojorquez-Tapia et al 1994). Current spatial approaches, such as the so-called gap analysis (Scott et al 1993), assume that present species records have been sampled without bias, although this assumption has been shown to be inaccurate in most cases (Bojorquez-Tapia et al 1994). The landscape approach taken in the present study allowed linking of biological records to aerial units. These units can be regarded as polygons of a homogenous nature that share similar environmental conditions, which in turn is reflected in similar species composition. This approach is independent of the causes that have favored such patterns, either historical or ecological, because LUs integrate interactions between physical conditions and biological records at all temporal and spatial scales (Zonneveld 1995). For mapping purposes, accurate spatial reference (both geometric and topologic) needs to be considered. Topographic attributes (altitude, longitude, and latitude) have been favored to depict biodiversity patterns. Topologic (land form, soils, and vegetation) relationships between species records and geographic patterns have been largely neglected. These topologic attributes become relevant when representing biological data spatially. Species distribution patterns at all scales are strongly influenced by environmental features such as topologic and human actions. In our study, it was inferred that LUs represent homogenous areas sharing topological attributes, which in turn may better explain the biodiversity patterns found in the region. This tied relation between topological and biodiversity attributes has been reported in different situations regardless of climatic, edaphic, ecological, and cultural conditions (Butterfield et al 1994; Zonneveld 1995; Velázquez and Heil 1996; Hansen and Rotella 2002; Hoersch et al 2002).

Translating science into applied conservation action

The results have served to translate pure academic outputs into sound conservation actions (Velázquez and Romero 1999). This was also inherent in the landscape approach, given that the social component is regarded as the major driving force in delineating current landscape patterns (Velázquez et al 2001). This last experience comprised a thorough evaluation of the environmental services provided by the study region to people living in Mexico City (see Figure 1). Every LU was ranked according to its importance in providing environmental services, mainly water harvest, land-use change, and large concentrations of biodiversity. The project became known by the local government, which eventually supported it financially. The specific request from the government was to assemble all current biophysical information to pinpoint areas of focus for conservation efforts. Because all data were spatially explicit through the geographical information system, cross-matrices were built by map overlay functions. Three types of areas were distinguished: buffer I (area for low-impact traditional cropping activities); buffer II (area for sound agroforestry practices); and core areas typifying the different degrees of conservation need.

  • Buffer I is comprised of LUs with large anthropogenic transformation where rural land use still prevails, with LB and low value for water harvest (LB in Figure 3 and LC I in Figure 4).

  • Buffer II embodies LUs harboring MB values (Figure 3), with no land-use change, covered by agricultural fields devoted to forage production (LC I and LC II in Figure 4) but playing an important role in rainwater filtration.

  • Core areas include LUs crucial for watershed functional integrity, with no drastic land-use change in original forest vegetation, HB (in Figure 3), and steep slopes and lava flows (LC III in Figure 4).

The output made it possible to define a new network of priority areas for conservation to guarantee watershed integrity and biodiversity conservation (Figure 5). Local indigenous communities, which have helped us during these 15 years of research, also became acquainted with the ecological importance of their lands, for themselves as well as for urban areas. Through a large number of participatory workshops with over 20 communities, 3 have so far signed up to adopt the new network for conservation. New mechanisms for re-funding conservation efforts for rural people by urban settlers need to be designed. Finally, social, political, and ecological driving forces need to be holistically addressed to achieve sound, long-term conservation for most tropical and subtropical regions. In this sense, the landscape approach presented here attempts to provide a conceptual as well as a pragmatic alternative for land-use planning and conservation.



S Aronoff 1989. Geographic Information Systems. A Management Perspective. Ottawa: WDL Publications. Google Scholar


J Baillie and B Groombridge . editors. 1996. IUCN Red List of Globally Threatened Animals. Gland, Switzerland: IUCN. Google Scholar


LA Bojorquez-Tapia, I Azuara, E Ezcurra, and Flores-Villela O . 1995. Identifying conservation priorities in Mexico through geographic information systems and modeling. Ecological Applications 5:215–231. Google Scholar


LA Bojorquez-Tapia, P Balvanera, and AD Cuarón . 1994. Biological inventories and computer data bases: their role in environmental assessment. Environmental Management 18:775–785. Google Scholar


PA Burrough 1988. Principles of Geographical Information Systems and Land Resources Assessment. Oxford: Clarendon Press. Google Scholar


BR Butterfield, B Csuti, and JM Scott . 1994. Modeling vertebrate distribution for GAP analysis. In: Miller RI, editor. Mapping the Diversity of Nature. London: Chapman and Hall, pp. 53–69. Google Scholar


[CIPAMEX] Consejo Internacional para la Preservación de las Aves en México 1993. Propuesta de especies y subespecies de aves silvestres y acuáticas, raras, endémicas, amenazadas, en peligro de extinción y sujetas a protección especial para incluirse en la NOM-PA-CRN-001/93. Reunión Anual. Veracruz, Mexico: CIPAMEX. Google Scholar


KK Chung and RD Weaver . 1994. Biodiversity and Landscapes. Cambridge, UK: Cambridge University Press. Google Scholar


JR Grehan 1993. Conservation biogeography and the biodiversity crisis: a global problem in space/time. Biodiversity Letters 1:134–140. Google Scholar


AJ Hansen and JJ Rotella . 2002. Biophysical factors, land use, and species viability in and around nature reserves. Conservation Biology 16 4:1112–1122. Google Scholar


AJ Hansen, TA Spies, FJ Swanson, and JL Ohmann . 1991. Conserving biodiversity in managed forest. Bioscience 41:382–392. Google Scholar


M Hill 1979. TWINSPAN-A FORTRAN Program for Detrended Correspondence Analysis and Reciprocal Averaging. Ithaca NY: Cornell University. Google Scholar


M Hill and HG Gauch Jr . 1980. Detrended correspondence analysis: an improved ordination technique. Vegetatio 42:47–58. Google Scholar


B Hoersch, G Braun, and U Schmidt . 2002. Relation between landform and vegetation in alpine regions of Wallis, Switzerland. A multiscale remote sensing and GIS approach. Computers, Environment and Urban Systems 26:113–139. Google Scholar


Instituto Nacional de Ecología 1994. NOM-PA-CRN-001. Diario oficial de la federación. Mexico: Publicaciones de la INE-SEMARNAT. Google Scholar


Integrated Land and Water Information System (ILWIS) 1999. Version 2.2. Enschede, The Netherlands: International Institute for Geo-Information Science and Earth Observation (ITC) Publications. Google Scholar


KJ Kirby 1996. Conservation of habitats. In: Spellerberg IF, editor. Conservation Biology. London: Longman Press, pp. 141–153. Google Scholar


EW Küchler and IS Zonneveld . 1988. Vegetation Mapping. Dordrecht, The Netherlands: Kluwer Academic Publishers. Google Scholar


EF Lambin, BL Turner, and HJ Geist . 2001. The causes of land-use and land-cover change: moving beyond the myths. Global Environmental Change 11:261–269. Google Scholar


RM Mayr 2000. Living together. Science 288:5464. Google Scholar


N Myers 2000. Sustainable consumption. Science 287:2419. Google Scholar


JM Scott, F Davis, B Csuti, R Noos, B Butterfield, C Groves, H Anderson, S Caicco, FD Erchia, TC Edwards, J Ulliman, and RG Wright . 1993. Gap Analysis: A Geographic Approach to Protection of Biological Diversity. Wildlife Monographs 123. Supplement to The Journal of Wildlife Management 57.1. Available at Scholar


CJF Ter Braak 1990. Update Notes: CANOCO version 3.10. Wageningen, The Netherlands: Agricultural Mathematics Group. Google Scholar


JK Vanclay, AG Bruner, RE Gullison, RE Rice, and GAB da Fonseca . 2001. The effectiveness of parks. Science 293:1007–1008. Google Scholar


A Velázquez 1994. Multivariate analysis of the vegetation of the volcanoes Tláloc and Pelado, Mexico. Journal of Vegetation Science 5:263–270. Google Scholar


A Velázquez and G Bocco . 1994. Modeling conservation alternatives with ILWIS: a case study of the volcano rabbit. ITC Journal 3:197–204. Google Scholar


A Velázquez and G Bocco . 2001. Land unit approach for biodiversity mapping. In: Van der Zee D, Zonneveld I, editors. Landscape Ecology Applied in Land Evaluation, Development and Conservation: Some Worldwide Selected Examples. ITC Publications No. 81, IALE publication MM-1. Enschede, The Netherlands: ITC–IALE (International Training Centre for Aerial Survey–International Association for Landscape Ecology), pp. 273–286. Google Scholar


A Velázquez, G Bocco, and A Torres . 2001. Turning scientific approaches into practical conservation actions: the case of Comunidad Indígena de Nuevo San Juan Parangaricutiro, México. Environmental Management 5:216–231. Google Scholar


A Velázquez and AM Cleef . 1993. Plant communities of the volcanoes Tláloc and Pelado, Mexico. Phytocoenologia 22:145–197. Google Scholar


A Velázquez and GW Heil . 1996. Habitat analysis of the volcano rabbit (Romerolagus diazi) by different statistical methods. Journal of Applied Ecology 33:543–554. Google Scholar


A Velázquez and FJ Romero . editors. 1999. Biodiversidad de la región de montaña del sur de la Cuenca de México: Bases para el ordenamiento ecológico. Mexico City: UAM-X-SEMARNAP (Universidad Autónoma Metropolitana, Unidad Xochimilco)–Secretaría de Medio Ambiente, Recursos Naturales y Pesca. Google Scholar


A Velázquez, FJ Romero, and J Lopez-Paniagua . 1996. Ecología y conservación del conejo zacatuche y su hábitat. Mexico City: Fondo de Cultura Económica. Google Scholar


IS Zonneveld 1995. Landscape ecology. Amsterdam: SPB Academic Publishing. Google Scholar


Digital terrain model of the study area. Land mosaic derived from a satellite image (Landsat ET+7) including Sierra de las Cruces and Sierra Chichinautzin as the main mountainous landscapes. Purple color shows human settlements, blue shows crops, red and brown show forested areas. (Courtesy of Institute of Geography, UNAM)



Dendrogram obtained by classification analysis to distinguish major vegetation clusters. In total, 137 sampling units and 345 plant species were analyzed to define the 6 vegetation clusters. A thorough description of all vegetation types occurring in the area is given in Velázquez and Cleef (1993).



Ordination of (terrestrial vertebrate–vascular plant) key species obtained by DCA. The cumulative variance explained by the axes reached 40% (axis I = 24%, λ = 0.799 and axis II = 16%, λ = 0.368). The circle embodies 3 major SA defined on the basis of the mean value of scores (at 95% CI). The 3 clusters obtained were high biodiversity (HB), medium biodiversity (MB), and low biodiversity (LB).



Ordination of LUs (LCs I–X) based on their DCA attributes, resulting in 64% cumulative variance shown on 2 axes (axis I = 37%, l = 0.798 and axis II = 27%, l = 0.494). Three major significantly different LCs were distinguished.



Spatial representation of the current proposal obtained by the participatory approach. Three main clusters of regions were distinguished: buffer I, area for low-impact traditional cropping activities; buffer II, area for sound agroforestry practices; 4 core areas according to different conservation needs.



List of key species used in spatial modeling, with an indication of their conservation status (categories: T = threatened; E = endangered), according to CIPAMEX (1993), Instituto Nacional de Ecología (1994), and Baillie and Groombridge (1996).

Alejandro Velázquez, Gerardo Bocco, Francisco J. Romero, and Azucena Pérez Vega "A Landscape Perspective on Biodiversity Conservation," Mountain Research and Development 23(3), 240-246, (1 August 2003).[0240:ALPOBC]2.0.CO;2
Accepted: 1 April 2003; Published: 1 August 2003

Back to Top