This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain. Conservation of Madrean Archipelago and Regional Forest Development Proj ects in Mexico Luis A. Boj6rquez-Tapia, L.A. Pena, C. Alvarez, Ivan Azuara, M. Alquicira, and A. Ramirez1 Abstract.-The Madrean Archipelago is a priority for conservation of biodiversity in Northwestern Mexico. Also, the Chihuahua and Durango Forestry Development Project was proposed to manage the forest of those two Mexican states to modify current deforestation rates. An environmental baseline study was performed to provide baseline data for the protection of the biodiversity in the region and to comply with Mexican environmental law and the World Bank requirements. Biological data is biased to accessible sites and limited to a few taxonomic groups (several families of vascular plants and terrestrial vertebrates). Consequently, to optimize the use of available data, a GIS modeling approach was used to identify species-rich areas. Modeling was performed by means of mUltivariate statistics (correspondence analysis and generalized linear models). The GIS was used to environmentally characterize the collection sites and to display the spatial patterns of species-rich areas. The results delineated a set of priority areas for conservation. Within these areas, special forestry management considerations were recommended for more detailed environmental impact assessments required by law. The modeling approach was useful to derive legally defensible inferences to reduce environmental conflicts between conservation and forest development, despite the limitations of biological inventories. process, (2) data are insufficient to derive useful conclusions, (3) and results are not suitable for planning because of their small scale and lack of accuracy. The Madrean Archipelago is a critical region for biodiversity conservation. In Mexico, the Madrean Archipelago extends along the Sierra Madre Occidental in the states of Chihuahua and Durango. Nonetheless, the Forestry Development Project of Chihuahua and Durango has been perceived as a major threat to biological diversity. The purpose of this paper is to present the results of an investigation on biodiversity distribution patterns for Chihuahua and Durango. This research was carried out as part of the EA of the Chihuahua and Durango Regional Forestry Development Project The objective of the forestry development project was to introduce modern management practices and to expand village and INTRODUCTION Preservation of biological diversity has been acknowledged as an imperative task for sustainable development. Thus, governments all over the world and multinational developments banks are setting up policies aimed at integrating conservation priorities and regional development. Environmental assessments (EAs) are planning studies to evaluate the effect of development projects. Typically, EAs are short duration studies, performed with scarce funds and limited data (Bojorquez-Tapia et aI. 1994). Moreover, EAs of large regional development projects have been criticized in the following grounds: (1) EAs have not been properly integrated into the planning 1Centro de Ecologfa, UNAM, Apartado Postal 70-275, Mexico, D.F. 04510, Mexico. 206 commercial forest-based industries in the Sierra Madre Occidental. It was expected that with the new forestry practices current deforestation rates in the Sierra Madre Occidental would be curtailed. A major problem related to the absence of adequate, current, and usable data on the locations of the most biologically important areas. Therefore, a predictive approach was used in this study to map the biologically important areas. The approach was a modification of the gap analysis technique (Scott et al. 1987) to adjust for the constraints and biases of biological data in Mexico (Bojorquez-Tapia et al. 1994). In essence, the method consisted of a modeling exercise, carried out by means of computer databases, geographical information systems (GIS), and multivariate statistics. Through GIS modeling, the data constraints of EAs were overcome. Furthermore, the resulting maps provide a useful baseline for regional landuse planning and protection of biological diversity. Both the Mexican forestry authorities and the World Bank maintained that the project was feasib Ie form the technical, social, economic and environmental viewpoints. On the other hand, non-governmental organizations in both Mexico and the United States of America asserted that the impacts of this project on the environment, especially on biodiversity, could be extremely adverse. They claimed that protective measures were inadequate to provide a reasonable response to potential effects, because of the insufficient coordination, planning and monitoring capabilities of Mexican authorities. Although the World Bank loan was eventually canceled, the Project is being carried out by Mexican forestry authorities. The viewpoint of Mexican authorities is that the environmental risks are outweighed by the long and short-term benefits likely to be realized by the implementation of the Forestry Development Project. STUDY AREA The study area is located within the states of Chihuahua and Durango, along the Sierra Madre Occidental (fig. 1). This region is characterized by a variety of environmental conditions. The Sierra is rough and altitude ranges from 250 to 3,150 m a.s.l. The drainage pattern of the Sierra is complex; the principal rivers -Yaqui, Mayo, Fuerte p and Sinaloa- drain to the Pacific Ocean, though one -the Conchos- flows north to the Rio Bravo or Rio Grande. According to the Koeppen's classification, the Sierra Madre Occidental presents Bs and Cw climate types in ridges, and Aw along canyons (INEGI 1982). Mean annual precipitation ranges from SOO to 1,200 mm. Though rainfall is bimodal (summer and winter); 60-80% of the rainfall occurs from June to September. The spring drought period is more intense and longer than the fall one. Mean annual temperature ranges from 2426°C, in the lowest locations, to 6-8°C at the highest altitudes; maximum and minimum annual temperatures are 40°C and -15°C, respectively. The study area presents the following vegetation types (INEGI 1992, Rzedowski 1978): • Madrean coniferous forest.- It is located in the highest and coldest zones in the Sierra Madre, especially in northern aspects at 2,200 m a.s.l., although it can be found at higher elevations in southern aspects. The dominant species are Pseudosuga menziesii, P. ayacahuite" P. arizonica, P. strobiformis, Abies concolor, A. THE FORESTRY DEVELOPMENT PROJECT The Chihuahua and Durango Forestry Development Project originated in 1989. This year, Mexico solicited a $45.5 million loan to the World Bank to complement the financing of a $91.1 million five-year project. The project was expected to affect about 15% of the 9.3 million ha of forest land in the two states (which represents about 35% of remaining forest land in Mexico). Major problems of regional forestry development were lack of product classification systems and modern technology, combined with a deficient enforcement of forestry policies. These conditions resulted in extensive logging, mainly through high-grading, in the region. Hence, the Project rationale was that current environmental deforestation trends would be restrained by the introduction of better forestry practices into local communities and small private producers. Funding was directed to financing road work, new sawmills and equipment, although about 10% of the capital was aimed at environmental protection, technical training, supervision, and administration. To achieve the goal of introducing better forestry practices, about 90,000 credits would be given to people from local communities (200/0 of total population in the forests in the two states). p 207 durangensis, P. tremuJoides, A. acummata. Relict and isolated populations of Picea chihuahuana, an endangered species, are located in northern aspects in restricted zones, such as Cerro Mohinora, and Cascada de Baseseachic. • Pine forest.- The pine forest distributes at elevation from 1,500 to 3,000 m a.s.l. Dominant species include P. reflexa, P. arizonica, P. Jumbotzi, P. ayacahuite, and P. ponderosa. Microhabitats are important for the dominance of a particular species; for example, P. reflexa is dominant in the most xeric aspects, while P. ayacahuite prevails on more mesic slopes in canyons. • Pine-oak forest.- The similar ecological requirements of pines and oaks, their entangled successional relationships, and the diversity of microhabitats produce a mosaic of pines and oaks. Oak forest components often can be found above the lower limit of the pine forest, I I COAHUILA. I,,' I ! 2" ,25- SINALOA. ZACATECAS. o· ~III NAD"[ a OCCIDr/ll1'1L 10 so 100 NAYARIT. 1)5' ,04" Figure 1.-Study area. 208 especially in xeric aspects. On disturbed sites, oak s are more abundant than pines. Common species are Quercus emoryi, Q.oblongifolia, Q. grisea, Q. santaclarensis, Q. durifoJia, Q. arizonica, Q. albocincta, Q. coccolobifolia, Q. crassifolia, Q. hypoleucoides, Q. pennivenia, Q. sideroxyla, P. cembroides, P. emoryi, P. 00carpa, P engelmannii, P.leiophylJa, Cuppresus arizonica, and Juniperus deppeana. • Oak woodland.- Deciduous oaks dominate between 1,000 and 2,000 m a.s.l., in slopes and plateaus. The most abundant oak species are Q. chihuahuensis, Q. tuberculata, Q. sipuraca, Q. santaclarensis, and Q. fulva. Dominant understory species are bunchy grasses, such as Bouteloa, Eragrostis, Muhlenbergia, and Schizachyrium. • Grassland.- Grasslands locate in plateaus and valleys of moderately deep soils, between 200 and 2,000 m a.s.l. Important species are Bouteloa spp., Muhlenbergia spp., Bacharis spp., Schizachrium spp., and Hilaria spp. In disturbed areas by fire, overgrazing, overcutting, of abandoned agricultural lands, Aristida. spp. are common. • Desert scmb.- This vegetation type is widespread in Northwestern Mexico. In the study area, it extends along the Pacific Coastal Plain, the lowlands of Sonora, and the highlands in Chihuahua. In the highlands, the desert scrub can be found as high as 3,000 m a.s.l. It is composed by shrubs with small leaves or folioles. Typical species are Larrea tridentata, Yucca carnerosana, Fouqueria splendens, Opuntia spp., Celtis spp., Prosopis, and Acacia. • Tropical deciduous forest.- This vegetation type is typical of sub.humid hot climates. It extends along the lowest elevations until 29°N, due to the protection against northern winds of the Sierra Madre. Most of the individuals (75°/0-1000/0) loose their leaves after the summer rains and for long periods (6-8 months). The dominant trees lack of spines and the tallest are 15 m. The dominant species are: Lysiloma microphylJa, L. watsoni, Ceiba acuminata, Bombax palmeri, Cochlospermum vitifolium, Lamiocereus spp., Caesalpinia atomaria, Tababoua palmeri, Conzattia sericera, Bursera spp., Guazuma uimifolia, and Ipomea arborescens. Representative understory species are: Hintonia latiOora, Schopfia parvUolia, Sebastiana pringlei, Agonandra racem05a, Wimmeria mexicana, WiUardia mexicana, and Erythrina flabelliformis. models. Programs ORDEN version 1.4 by E. Ezcurra, and GLIM version 3.77 (London Royal Statistical Society 1985) were used for all statistical analyses. CASEY was used to detect the relations between ecological factors and species distributions. The relationships between axes and environmental variables were evaluated by visual examination to identify species assemblages. A frequency matrix was prepared for each species assemblage. Thus, a matrix cell showed the number of records that corresponded to that specific combination of the orthogonal environmental variables that explained the species distribution patterns. Non existent combinations of values between the orthogonal environmental variables were eliminated. A log-linear model was fitted to each frequency matrix (Atkin et al. 1990). Models were assessed according to their coefficient of determination (r2 > 0.25) and significance (p < 0.001. for linear terms; p < 0.05 for quadratic terms). The value combinations that predicted the highest frequencies were mapped in the GIS. We draw a final map by overlaying all the predicted distributions and the vegetation and land-use layer. ' MATERIALS AND METHODS Our methods combined GIS layers, a biological computer database, and multivariate statistics. The equipment included five IBM compatible microcomputers, and two digitizing tables (Kurta IS / ONE and Numonics 2200). The GIS consisted of the programs AU2 (ICFA 1987), Roots (CorsonRikert 1990), and CI/SIG (CI 1992). Our database was compiled species information obtained from comprehensive literature and scientific collection revisions. Species data were gathered from the literature, and biological collections. The data consisted of species names and collection sites. Each collection site was georeferenced in 1:250,000 topographic maps before it was transferred to the GIS. The following 1:1,000,000 maps (INEGI 1982) were digitized into the GIS: soils, geology, vegetation and land-use, topography, mean annual precipitation, and mean annual temperature. With respect to the vegetation and land-use map, this map was reclassified to reduce the number of categories; thus, the categories in the new layer were the following: Mixed coniferous forest, pine forest, pine-oak forest, oak woodland, grassland, desert scrub I tropical deciduous forest, and agriculture and ranching. A one km2 cell size was used for each raster layer. An additional layer was the binary map of collection sites. The digitized maps were overlaid on the layer of collection sites to environmentally characterize these sites. Results from the overlays were transferred to presence-absence matrices of species and environmental variables. From these matrices, corresponding contingency tables were prepared for the subsequent modeling. Modeling was carried out by means of ordinations, through correspondence analysis of species and environmental variables or CASEY (Montana and Greigh-Smith 1990) and generalized linear RESULTS The biological database included many records, although site description for about one half of the records was insufficient to determine their latitude and longitude (Table 1). With respect to the number of families, species, and records included in the database, birds encompassed most of the data, followed by mammals, reptiles, and amphibians. The majority of the georeferenced records corresponded to mammals, while amphibians represented a small percentage of the data (Table 1). Table 1.-Number of species and records compiled in the data base from Chihuahua and Durango. Taxa Families (#) Species (#) (%) Records (#) (%) Georeferenced records (#) (%) (%) 6 7 11 2 i 14 3 79 5 Reptiles 10 12 44 HI 510 15 231 15 Birds 49 59 327 71 1,516 45 428 27 Mammals 18 22 79 17 1,244 37 838 53 Totai 83 100 461 100 3,384 100 1,576 100 Amphibians 209 use was not included in the GLIM to simplify the models. However, it was used as an additional layer in the GIS to increase the accuracy of the predicted areas. The general linear model consisted of five terms: The following matrices were obtained: presence absence (species x environmental variables), amphibians (114 x 6), reptiles (510 x 6), birds (1,516 x 6), and mammals (871 x 6); contingency tables (species x categories of environmental variables), amphibians (10 x 26), reptiles (38 x 30), birds (296 x 26), and mammals (116 x 29). Twenty species assemblages with similar ecological requirements were obtained from the CASEV (Table 2). The variables that best explained the highest variance were elevation, mean annual temperature, mean annual precipitation, and vegetation and land-use (Table 2). Since elevation and mean annual temperature were highly correlated (r2 = -0.8), the former was discarded from further analyses. Vegetation type and land- y=ea +bt+cp+dt2 +ep2+fpt Where y is the predicted species frequency, t is mean annual temperature, and p is mean annual precipitation. The importance of each term varied between models (Table 3). Significant fits of GLIM were possible for 13 from the original 20 species assemblages (Table 4). These models were transfered to the GIS to display their spatial distribution, and were combined with the observed spatial distributions of the Table 2. -Species assemblages generated by correspondance analysis for terrestrial vertebrates of Chihuahua and Durango (Me=mlxed coniferous forest; P=plne forest; PO=plne.oak forest; O-oak woodland; TO-tropical deciduous forest; OS=desert scrub; A=agrlculture and cattle ranching). Species Assemblage Temperature Range cae) PreclpltaUon Range (cm) Elevation range (m a.s.I.) VegetaUon type Area (km2) Area MC,PO 3,128 1.0 (%) Amphibians 8-20 400-1,200 1.()()()..3,000 22-24 700-1,500 0- 600 TO 1,174 0.5 8-20 200- 500 0- 600 P,A 392 0.1 2 18-20 500-1,500 1,000-1,500 PO,O,MC 1,270 0.4 3 20-22 500- 700 1,000-1 ,500 NA 887 0.3 4 20-22 500- 600 500-1,000 TO 1,777 0.5 16-18 500- 600 2,000-3,000 MX,P 2 8-16 600-1,500 2,000-3,000 TO 3 16-18 600-1,500 2,000-3,000 O,PO 4 18-22 600-1,500 600-2,000 O,PO,MC 5 22-28 500-1,500 600-2,000 NA 2 BIIdUII !I!!:U 392 Mammals 2 8-16 700- 800 2,000-3,000 O,PO,MC 5,574 1.8 16-20 700- 800 1,000-2,000 O,PO 2,360 0.8 o ,TO 772 0.3 2,B56 0.9 3 20-22 600- BOO 1,000-2,000 4 22-2B 600- 700 0-1,000 5 16-20 600- 700 1.000-2,000 0 66 0.0 6 8-16 600- 700 2.000-3,000 NA 5,137 1.7 7 20-22 500- 600 1.000-2,000 A 4,621 1.5 8 18-20 500- 600 1.000-2,000 DS,P 672 0.2 9 8-18 500-600 2,000-3,000 P 7,593 2.5 210 TO servation by increasing the effectiveness of reserve areas (Franklin 1993). A landscape approach also facilitates the resolution of the current conflict between logging and forest preservation in the Sierra Madre Occidental. Much of the conflict is over the location of critical sites for conservation (for example, oldgrowth stands and species-rich hot spots) and the presence of priority species (Burns et aI. 1994). Clearly, a know ledge on species distributions and on ecosystem processes is obligatory for a successful implementation of a landscape strategy. Gap analysis is a technique designed to identify priority areas for conservation by comparing the location of existing nature preserves with the location of species-rich areas (Scott et aI. 1987, 1988). The gap analysis approach combines the emerging concepts in conservation biology with up-to-date advances in data management and analysis. However, utilization of gap analysis is severely limited because of the quality and quantity of the available information in Mexico. As in other regions in the country (Boj6rquez-Tapia et aJ. 1994), biological data of Chihuahua and Durango are biased to a few families of vascular plants and terrestrial vertebrates, and to accessible sites (eg. along major roads or close to cities). Similarly, most of the thematic maps are small scale and outdated, especially the themes on species distribution and on vegetation and land-use. Consequently; the design of a landscape level approach for the Sierra Madre depends on portraying species richness and species distribution patterns. No adequate theory exists for predicting areas of species richness (Austin 1991). Thus, approaches to minimizing the effects of forestry development on sensitive wildlife are generally derived from experience (Irwing and Wigley 1993). Nonetheless, inferences for regional landuse planning cannot be restricted to those originated from experience. On the contrary, effective regional land-use planning has to satisfy two conditions: (1) results have to be obtained from accurate methods, and (2) inferences have to be legally defensible (Steiner 1983) The method used for depicting species distribution patterns in Chihuahua and Durango was based upon readily available information. The baseline database included the readily available biological data and many collection sites (Table 1). Twelve high biodiversity areas were identified seven species assemblages that did not significantly fit a linear model (fig. 2). Predicted and observed distributions of species assemblages covered approximately 10% of the area of influence (fig. 2). Based on the spatial patterns, we delineated 12 areas critical for conservation (fig. 2). Some of these areas layed outside the project region, but all were placed within the area of influence of the project. For the forestry project, we considered areas 5, 6, 8, 9, 10, 11, and 12 as critical. Total predicted areas for each of the taxonomic Classes were the following: mammals, 2,965,100 ha; birds, 413,600 ha; reptiles, 432,600 ha; and amphibians, 486,200 ha. DISCUSSION Successful biodiversity conservation compels the use of a large-scale landscape strategy to merge nature reserves and sensible management of the semi-natural landscape matrix. While nature reserves are crucial for conservation, the semi-natural matrix renders a critical role in conTable 3.-Generallzed lineal models (GUM) for terrestrial vertebrates In Chihuahua and Durango. The order of the variables In the model Indicates their Importance (1 = mean annual temperature; 2 mean annual precipitation; and 3 Interaction between 1 and 2). = Group Model = Degrees of freedom f2 Amghlblans 4 0.45 1+ 2 3 0.52 2 1+ 2 4 0.43 3 2 + 1 4 0.36 1+2+3 4 0.29 1+3+2 4 0.25 1 + 2 +3 5 0.62 2 1+ 2 4 0.35 3 2 + 1 4 0.36 5 1+2 4 0.34 6 1 +2 4 0.57 8 2 + 1 +3 4 0.36 9 1+2+3 7 0.72 1+2 RegtUes Birds 2 Mammals 0 211 Figure 2.- Predicted distribution areas of terrestrial vertebrates and critical regions for conservation (numbered squares) in Chihuahua and Durango (white = study area; Iblack=predlcted distributions of terrestrial vertebrates; gray=no data). 212 through ordination analysis and GIS modeling (fig. 2). Albeit the models are yet to be validated, we assume that the predicted species-rich areas are accurate, based upon previous experiences with the application of empirical models elsewhere (Austin et ai. 1984, Boj6rquezTapia et ai. in press, Ezcurra et al 1987, Margules and Stein 1989, Miller et al 1989, .scott et ai. 1993). The species-rich areas included sites that have not been explored in past biological surveys. Consequently, while the use of available data was optimized, these results represent an improvement in the ability to identify ecologically sensitive areas. From the legal viewpoint, our study can be considered as an unbiased mapping exercise to outline priority areas for biodiversity conservation. The twelve high biodiversity areas correspond to regions where nature reserves should be established. Also, by overlying a map of proposed forestry operations on the species assemblages distribution map, areas where conflicts between forestry and conservation are more likely to occur can be delineated. According to environmental legislation, forestry practices that blend conservation with sustained forest development should be devised for the conflicting areas. As a first step, a detail survey should be done before the implementation of any forestry practices; detailed surveys should include a biological inventory, a ground truth vegetation map, and locate landscape corridors connecting areas of high species richness. Subsequent research and adaptive-feedback monitoring should refine the strategies and reduce risks of extinction and impacts caused by forestry development. Conservation International. 1992. CI/SIG, sistema de informaci6ngeografica, version 2.0: Manual del usuario. Conservation International, Washington, D.C., 151 pp. Corson-Rikert,J . 1990. Roots user's manual. Harvard University Graduate School of Design, Laboratory for Computer Graphics and Spatial Analysis. Cambridge, Massachusetts, 152 pp. Ezcurra, E., M. Equihua, and J. L6pez-Portillo. 1987. The desert vegetation for El Pinacate, Sonora, Mexico. Vegeta tio71 :49-60. Franklin, J.F. 1993. Preserving biodiversity: Species, ecosystems, or landscapes? Ecological Applications 3:202-205. ICFA. 1987. A U2, paquete general de cartograffa: Manual del usuario. Publicacion 88-AU-2-01, ICFA, Mexico, CONCLUSIONS The Forestry Development Project of Chihuahua and Durango has been perceived as a major threat to biological diversity. Given the limitations in data and know ledge, a predictive approach was carried out to delineate potential species-rich areas. The results included 12 areas where special forestry management considerations should be set up" Such considerations must be the base of detailed EAs of specific forestry practices. This research illustrates how a predictive approach -through multivariate models and GIS- can be used to delineate conservation priorities in regional land-use planning. Furthermore, the usefulness of the modeling approach is demonstrated by the optimization of available biological information to derive defensible inferences from rigorous methods. D.F. Instituto Nacional de Estadistica, Geograffa e Informatica. 1982. Atlas Nacional del Medio Fisico. Secretarfa de Programacion y Presupuesto. Mexico, D.F. 224 pp. Irwing, L.L. and T.B. Wigley. 1993. Toward an experimental basis for pretecting forest wildlife. Ecological Applications3:213-217. Margules, C.R. and J.L. Stein. 1989. Patterns in the distribution of species and the selection of nature reserves: An example from Eucalyptus forest in south-eastern New South Wales. Biological Conserva tion 50:219-238. Miller, R.I., S.N . Stuart, and K.M. Howell. 1989. A method ~ ology for analyzing rare species distribution patterns utilizing GIS technology: The rare birds of Tanzania. LITERATURE CITED Atkin, M., D. Anderson, B. Francis, and J. Hinde. 1990. Statistical modeling in GLIM. Clarendon Press, Oxford.374 pp. Austin, M.P. 1991. Vegetation theory in relation to cost-effective surveys. In: Pages 17-22, C.R. Margules and M.P. Austin, Nature conserva tion: Cost effective biological surveys and data analysis. CSIRO, Australia 207pp. Austin, M.P., R.B. Cunningham, and P.M. Flemming. 1984 New approaches to direct gradient analysis using environmental scalars and statistical curve-fitting procedures. Vegetatio55:11-27. Bojorquez-Tapia, L.A., I. Azuara, E. Ezcurra, and O. Flores-Villela. In press. Identifying conservation priorities in Mexico through geographical information systems and modeling. Ecological Applications. Bojorquez-Tapia, L.A., P. Balvanera, and A.D. Cuaron. 1994. Biological inventories and computer data bases: Their role in environmental assessments. Environ- mentalManagement18:775-785. Burns, B. M. Dress, D. Hadley, and W. Laird. 1994. Proyecto de Recursos Tarahumara. Unpublished report. Biodiversity Support Program. Landscape Ecology2:173-189. Montana, C. and P. Greigh-Smith. 1990. Correspondence analysis of species by environmental variable matrices .Journalof Vegeta tion Science 1:453-460. Royal Statistical Society. 1985. GLIM3.77. London. Rzedowski, J. 1978. Vegetaci6n de Mexico. Editorial Limusa, Mexico, D.F.432 pp. Scott,J.M., B.Csuti, K.Smith,J.E.Estes, and S.Caicco.1988, Beyond endangered species: An integrated conserva~ tion strategy for the preservating of biological diversity. Endangered Species Update5:43-48. Scott, J .M, B. Csuti, and J .E. Estes. 1987. Species richness, a '-, geographic approach to protecting future biological diversity. BioScience37:782-788. Scott, J.M, F. Davis, B. Csutti, R. Noss, B. Butterfield, C. Groves, H. Anderson, S. Caicco, F. D'Erchia, I.e. Edwards,J. Ulliman, and R.G. Wright. 1993.Gap analysis: A geographic approach to protection of biological diversity. Wildlife Monograph 123:1-141. Steiner, F. 1983. Resource suitability: Methods for analySt!S. Environmental Management5:401-420. 213