ABSTRACT
Protected areas in the Ecuadorian highlands have been threatened recently by increasingly severe anthropogenic disturbances. Given both the biodiversity and spatial heterogeneity of Andean ecosystems, geographers should develop reliable methods for mapping these areas. While remote sensing techniques have rarely been applied to studies of tropical montane landscapes, these tools have the potential to provide useful maps of vegetation and zones of anthropogenic disturbance. This article documents the use of aerial photographs and Landsat digital imagery for mapping vegetation communities and assessing rates of deforestation in the southern Ecuadorian Andes. The utility and limitations of several mapping procedures will be discussed in relation to the conservation of natural resources.
INTRODUCTION
Having lost up to 350,000 ha of forest per year during the past fifteen years (Sánchez and Vivanco, 1991), Ecuador experiences one of the highest rates of deforestation in Latin America. Patterns and processes of deforestation have been described for both the western section of the country (Dodson and Gentry, 1991; Sierra, 1996) and for the Oriente (e.g., Fundación Natura, 1990; Southgate et al. 1991). Very few published studies, however, have attempted to document recent patterns of deforestation in Ecuador's montane regions (Keating, in press). Although both past and present human activities in the northern Andes have been studied by several researchers (e.g., Troll, 1973; Ellenberg, 1979; Laegaard, 1992), few attempts have been made to map deforestation and other human disturbances within most Andean regions (but see Echavarria, 1993; Keating, 1995; Sierra, 1996).
Because indigenous peoples have practiced various forms of land use in the Ecuadorian Andes for the past several thousand years (e.g., Engel, 1976), landscapes there have been altered repeatedly. During the last 500 years, however, significant changes in land-tenure systems and agricultural techniques have enhanced both the intensity and spatial extent of human disturbances (Ellenberg, 1979). Montane ecosystems around the world may soon be degraded irreversibly (Eckholm, 1975), and ecologists have become especially concerned about the consequences of this degradation in the Andes (e.g., Glaser and Celecia, 1981; Parsons, 1982). Destruction of Andean forests may reduce biodiverrsity (e.g., Young, 1994) and exacerbate soil erosion (Millones, 1982; Harden, 1991). Given the current need for more agricultural land, those upper montane forests and páramo communities (tropical alpine) not cut or burned in the recent past may be severely degraded in the near future (Hess, 1990). [end p. 77]
While several descriptions of environmental degradation in the Ecuadorian highlands have been made (e.g., Grubb, 1970; Morris, 1985), few studies have demonstrated how past and current human activities impart long-term environmental changes within these regions (but see Sierra, 1996). Both the severity and spatial distribution of anthropogenic disturbances vary significantly along ecological gradients, yet relatively little attention has been given to how human impacts may differentially modify the landscape over time. When several land-use techniques are practiced simultaneously in one region that includes complex ecological gradients, some biological communities may be more threatened than others. Therefore, conservationists should under-stand both the distributions of plant coomunities and their disturbance regimes (e. g., Botkin, 1990).
Nine protected areas have been established in Ecuador's montane zone during the past twenty-five years (Fundación Natura, 1992a). Although they contain numerous endemic species and complex ecosystems, few financial resources have been allocated to developing and enforcing environmental policies. Given that proper resource management is frequently hampered by ignorance ofthe biogeography and disturbance regimes within protected areas, the imperative of developing reliable mapping techniques for the tropical highlands is obvious. Ideally, projects designed to monitor existing biological resources would involve a combination of field studies and geographic information technologies. Studies conducted at several spatial scales simultaneously would enable researchers to determine the characteristics and areal extent of ecological communities and to monitor changes in land cover.1
While geographic information technologies have not been utilized in most of the tropical Andes, they have recently been used to map sections of Podocarpus National Park in southern Ecuador (Figure 1). Created in 1982, this Park receives substantial attention by conservationists not only because it contains tremendous biological diversity, but also because it includes the last zone of intact forest in southern Ecuador. Although most of its area is relatively undisturbed, some sections are currently being exploited by miners and colonists. Land use often involves intense burning of hill slopes and contamination of rivers (Fundación Natura, 1992b).
Until recently, relatively little has been known about plant communities in the Park, or about the impacts of human activities on its montane ecosystems. While few researchers have conducted detailed studies in the Park, some generalizations may now be made about both the disturbance ecology and effectiveness of mapping techniques available for this region. In this article, I discuss the use of several remote sensing procedures for mapping vegetation and environmental impacts in Podocarpus National Park. After patterns of environmental degradation and current modification of disturbance regimes are documented, suggestions will be made for optimizing the use of mapping techniques to assist present conservation efforts. Remaining technical and logistical difficulties will receive special emphasis. ]
USING GEOGRAPHIC INFORMATION TECHNOLOGIES TO MAP LAND COVER
While field research is crucial for understanding the impacts of human disturbances on tropical ecosystems, geographic information technologies are also necessary for understanding how fine-scale processes impart pattern to vegetation at the landscape scale. Not only do they enable researchers to map extensive areas of vegetation in rugged terrain, but they may provide a basis for documenting [end p. 78] rates, magnitudes and directions of deforestation and other disturbances. After determining how vegetation communities are differentially impacted by currrent disturbances, researchers may be able to identify which areas are in greatest danger of being exploited or degraded (see Zimmerer, 1994). Land cover mapping, therefore, may give resource managers a basis for making complex decisions concerning environmental policy and its enforcement.
Several categories of remotely sensed data have been applied to tropical ecology studies, but aerial photographs are still the most reliable and available sources of data in most developing countries (Sader et al. 1990). While cloud cover and complex vegetation structure impede the photo interpretation process, sequential aerial photographs are commonly utilized by resource managers to detect land-use changes over time (Adeniyi, 1980). Moreover, they are used to measure terrain characteristics and to help stratify forests with respect to accessibility (Sayn-Wittgenstein, 1980). Surprisingly, however, relatively few articles have discussed the use of photogrammetry and air photo interpretation for mapping land cover in neotropical areas (but see Echavarria, 1991; Keating, 1995); the most recent reviews of air photo mapping in the tropics were written before 1983 (Lanley, 1982).
Remote sensing studies that involve identifying tropical deforestation with the use of digital imagery are common, but most have been performed in areas below 1000m elevation (Nelson et al. 1987; Malingreau et al. 1989). Many researchers have found that the spectral resolution of AVHRR2 imagery makes it appropriate for assessing regional-scale changes in forest cover (e.g., Archard and Blasco, 1990), but its 1.1 km spatial resolution renders it ineffective for mapping land cover within relatively small areas. Most tropical remote sensing projects to date have utilized Landsat Multi-Spectral Scanner (MSS) data (Sader et al. 1990). In part because this imagery has relatively poor spectral resolution3 and a spatial resolution of 57 x 57 m, it appears to be inappropriate for mapping complex tropical vegetation or for accurately delineating zones of deforestation (Tucker et al. 1984).
When mapping land cover, geographers must be careful to select a mapping tool whose spatial resolution is commensurate with the scale of the phenomena being studied. Due to both inconsistent definitions of "deforestation" and the wide variety of remote sensors now used, there exists a wide range of estimates for the amount of deforestation in Latin America. Rates of clearing for the Amazon Basin, for example, have been estimated to be as low as 10,000 km2 per year to as high as 100,000 km2 (Malingreau and Tucker, 1988). In general, coarse-scale remote sensing studies may overestimate rates of deforestation by as much as 50 percent (Skole and Tucker, 1993).
Because Landsat Thematic Mapper (TM) imagery has a spatial resolution of 30 x 30m and a spectral resolution that is superior to the image types mentioned above, it may enable geographers to produce high quality maps of land cover. Due in part to the scarcity of cloud-free images for most tropical regions, TM data have been utilized less frequently in developing countries than in temperate areas (but see Mulders et al. 1992; Hill and Foody, 1994; Colby and Keating, submitted). TM imagery may be quite useful for mapping vegetation when ground-truthing is possible and when the selected cover categories are spectrally distinct. However, the utility of TM data for mapping areas of diverse tropical land cover in areas characterized by mountainous, rugged terrain has yet to be proven.
One goal of remote sensing specialists is "change detection," which may be performed if images are co-registered precisely and if the biophysical characteristics of the study area are well-understood (Jensen, 1996, 257-262). Land cover change is documented most effectively with a combination of remote sensing and Geographic Information Systems (Sader et al. 1990; Vanclay and Preston, 1990; Welch et al. 1990). GIS data storage characteristics permit the maintenance of spatial and temporal individuality of multiple data sets, which facilitates long-term monitoring of vegetation change (J akubauskas et al. 1990). Presently, however, there are no standardized techniques for monitoring tropical deforestation with "geographic tools" (Sader et al. 1990). While some ecologists (e.g., Veblen et al. 1994) have used GIS to map disturbances in public lands of developed countries, few researchers have done so for parks in the Third World (but see Sirait et al. 1994). Because relatively little is known about rates of tropical forest depletion (Myers, 1988), it is appropriate to integrate GIS into Ecuadorian park management strategies. [end p. 79]
LOJA PROVINCE AND WESTERN PODOCARPUS NATIONAL PARK
Research for this article was conducted in Loja Province, the southwestern-most province in the Ecuadorian highlands (Figure 1). The city of Loja, its capital, is situated immediately to the northwest of the Park within a broad valley at 2150m elevation. By 1990, the city contained approximately 87,000 people (Gomez, 1989), and its population continues to grow rapidly. Throughout the southern section of the Province, the population is predominately mestizo, and relatively few of the Saraguro Indians speak Quichua exclusively (see Knapp, 1991). Economic activities have traditionally included subsistence farming, dairy farming, grain production, cattle raising, and mining (Miller, 1959; Fauroux, 1986). Logging and mining have occurred increasingly since the 1940s.
The physiography of the Loja Canton, which includes western Podocarpus National Park, is dominated by a series of low mountain ranges (cordilleras) that are often more than 40 km in length (Figure 2). These mountains developed during the late Pliocene (Van der Hammen, 1974) and are not of volcanic origin. In contrast to the rich soils of central and northern Ecuador, the soils here often degrade quickly after vegetation removal (Christensen, 1989). Heavy rain occurs during most of the year, but a dry season may occur sometime between October and January (INMH, 1993). The upper montane zone often experiences cold temperatures and very strong easterly winds, and these inhospitable conditions may have protected the Park from more severe land degradation in the past.
Covering approximately 146,280 ha, Podocarpus National Park is one of the few protected areas between Cuenca, Ecuador and Juárez, Peru. Its northwestern section, in which the studies described below were conducted, is located at approximately 4°10' S by 79° W. The Park's northern border lies immediately to the northeast of Loja, and the southern border is nearly forty kilometers from the Peruvian frontier (Figure 2). Major highways surround the northern and western sides of the park, while smaller roads provide access from the south.
The Park's elevation varies between 1,000m and 3,695m elevation (Apolo, 1984), and includes several high ridges. Unlike the northern section of the country, where isolated peaks dominate the landscape, this region is characterized by a series of interconnected ridges, most of which have very steep slopes. The physiography is dominated by the Cordillera de Los Andes, a ridge that extends in a north-south direction. Numerous streams dissect the ridges that connect the high Cordillera with the highway to the west, so that the terrain is quite rugged (Figure 3).
Not only does the Park include unique landscapes, but it encompasses more than 100,000 ha of forest on the eastern cordillera of the Andes. The montane forests and paramos provide habit for more than 540 bird species (Andrade, 1996) and several endangered mammal species. Moreover, plant species diversity appears to be extremely high in all life zones studied to date (Bogh, 1992; Madsen and Ollgaard, 1993; Keating, in prep.). At the Cajanuma park station, for example, tree species diversity appears to be higher than that recorded at any other site in the world at this elevation (see Madsen and Ollgaard, 1993). Although botanists have collected specimens extensively in western Podocarpus National Park for more than fifty years (Espinosa, 1948), species lists for most life zones remain incomplete. [end p. 80]
Recent Anthropogenic Disturbances
Loja is the only Ecuadorian province that now exhibits net out-migration (Petri-Levy, 1993). Campesinos first left the province in large numbers after 1968, when an extended drought and several earthquakes occurred. The local land tenure system restricted access to remaining fertile lands, so that many people were forced to migrate either to large cities or adjacent provinces (Brownrigg, 1981). The exodus of people continues due to high unemployment, declining agricultural yields, and land degradation.
Because the Park contains the only remaining large tract of forest in southern Ecuador, many people have chosen to exploit resources or live within it. Therefore, the Park is now affected by a variety of severe anthropogenic disturbances. More than 100 families have established small farms within it (Espinosa et al. 1992), and other people enter the Park periodically to extract valuable timber. Although most farms are located within the eastern and southern section of the Park, the western edge of the Park has been degraded by the gradual expansion of farms. Selective logging, intense burning and livestock grazing are usually associated with colonization activities.
While most of the burning occurs at the periphery of pastures below 2,600m elevation, large fires occasionally consume sections of the paramo and timberline forest. Three of them have occurred since 1985, the largest of which affected 800 ha of upper montane plant communities (Keating, 1995). Paramo communities that experience intense burns during the veranillo4 exhibit very slow regeneration of species even after ten years. Because lighting strikes almost always occur during the wettest season, forest fires are almost certainly not part of the natural disturbance regime. According to Andre (1987), 100 percent of the fires in the Loja region are caused by humans, often by negligence. Landslides, which occur frequently during the wet season, appear to be the only major natural disturbance in the Park.
Conservation Efforts in Western Podocarpus National Park
While Podocarpus National Park contains many spectacular and unusual landscapes, its creation and protection have not been appreciated by the local people. Many Lojanos are unaware of the Park's existence (pers. obser.), and it is regarded by others as an impediment to regional economic development. In part because the boundary is poorly delimited, few of the local farmers refrain from cutting and burning the remaining forests. Current research efforts by the United Nations' Food and Agriculture Organization (FAO) and various other development groups have aimed to develop viable agricultural projects outside the Park and have sought ways of reducing deforestation by campesinos within the Park. While these ideas may be incorporated into the next comprehensive management plan, cutting and burning of the vegetation will most likely continue into the foreseeable future.
MAPPING WESTERN PODOCARPUS NATIONAL PARK AND ITS SURROUNDING AREAS
Elevation within the study area varies from 3,432m at the highest point of the Cordillera Oriental down to 2,120m near the western highway (Figure 2). This region is characterized by numerous knife-edged ridges, and slope angles vary from 0° to 57°, where the mean is 25° and the standard deviation is 16° (Keating, 1995). This area was chosen because it includes all common land cover types in the western section of the canton, as well as deforested areas of different ages. Both grass and shrub paramo communities dominate the upper sections of the Cordillera Oriental and the adjacent ridge tops. Because several areas of the ridge top have been burned at least once since 1983, the vegetation varies tremendously in structure and composition. Various stages of recovery after disturbances have been observed in the field.
Most of the study area is covered with mature forest, but numerous farms and pastures are located adjacent to and above the western highway. The upper montane forests (>2600 m) in this region have not experienced many anthropogenic disturbances during the past twenty-five years. Below 2,500m elevation, however, most areas have been disturbed chronically by humans during the past four decades (Figure 4). Most of the deforestation within 10 km of Loja occurred before 1960 (Dodson, pers. comm.), while deforestation to the south of the Cajanuma Park station has occurred more recently (Figure 2). Farms, pastures and clusters of houses occupy most of this zone, but large sections of it have been abandoned. These degraded lands are typically covered with scrub or secondary forest.
Interpretation of Aerial Photography
After the land-cover polygons were mapped on transparencies, they were carefully transferred to 1: 50,000 scale topographic maps. IDRISI, a raster-based GIS package (Eastman, 1993), was used to determine the areas and spatial distributions of [end p. 82] disturbed zones and other land cover types. To examine the relationships among terrain features and cover types, I created a digital elevation model (DEM) using both AutoCAD and ArcInfo software (Autodesk, 1992; ESRI, 1994). Initially, this process involved digitizing a portion of a 1: 50,000 scale map (contour interval is 40 m); all index contour lines between 2000 and 3400m were digitized.
While 16,480 hectares of forest existed in the study area in 1976, 15,939 hectares remained by 1989. Therefore, 541 ha, or 3.28 percent, of the forest had been cut and converted to agricultural land between 1976 and 1989. The average size of recently deforested areas was 26 ha, and most of these areas are located adjacent to or immediately above older zones of deforestation. More significantly, both the average elevation and the slope angle of deforested zones have increased markedly during the past several decades. The average elevation of deforested areas was 2,303m elevation in 1976 but increased to 2,502m by 1989. Whereas the average slope angle of older deforestation was 14.9°, the average slope angle for recent deforestation was 20.9°.
The photo analysis demonstrated, therefore, that the trends observed in other sections of the Andes are valid for this area as well. While rates of deforestation near Loja are not striking, the terrain features of the recently deforested areas give cause for concern. The more recent deforestation occurs on land that is rarely suitable for most agricultural activities. As subsequent fires occur near and above the peripheries of these pastures, an increasing area of forest may be degraded above the pastures due to the "pre-heating" effect that occurs on steeper slopes.
Ecologists typically seek air photos of scale 1:24,000 or larger for mapping projects (Lillesand and Kiefer, 1987). Because much smaller scale photos were used for this project, the analysis was only partially successful. The grass and shrub páramo communities were indistinguishable, and I [end p. 83] was unable to distinguish between the páramo and the timberline forest immediately below it. Therefore, the primary utility of air photo analysis was delineating deforested areas below 2,600m elevation.
Digital Image Processing
Image processing work for this project was performed with Erdas Imagine 8.0 software (Erdas, 1994). Before images were classified, I performed several "pre-processing" techniques, including black-body subtraction5 and topographic normalization.6 The latter process removes the uneven reflectance patterns (especially shadows) in the images and may enhance image classification accuracies (Colby and Keating, submitted). After several topographic normalization techniques were applied to both images, I was able to compare classified images that had received different treatments.
Using supervised classification techniques, I successfully mapped the following land cover categories with the TM images: grass páramo, shrub páramo, forest, and deforestation. Initially, I had attempted to subdivide "deforestation" into pastures, degraded forests, abandoned fields, and shrubby areas. Because these categories are not spectrally distinct enough to map accurately, I was only able to delineate mature forest from areas that had been cleared at some unknown time in the past. Due to both cloud cover and poor spectral resolution, the MSS images were only used to delineate forested areas from deforested zones.
Before accuracy assessments were performed on the classified images, a digital "ground truth" map was produced. Maps created from the air photo analysis, and numerous others based upon field observations, were digitized with AutoCAD (Autodesk, 1992). These data layers were then combined in IDRISI to create a land cover map that could be compared to the classified images. Using the Errmat module of IDRISI, I obtained overall accuracy percentages, error matrices and several other statistics.
Table 1 gives the overall image classification accuracies. The most accurate classified images were produced from normalized TM and MSS imagery; the overall best map (Figure 5) was created from the TM image classified with the minimum distance classifier, a non-parametric algorithm. Whereas the MSS images were classified with poor accuracy even when only two categories were selected, the classification accuracies for the TM images were significantly higher. Topographic normalization enhanced classification accuracies considerably and provided for a much better delineation of the páramo communities. This process provided striking visual changes in the imagery and enhanced classification accuracies more than the data indicates. Due to various technical difficulties, the ground truthing procedure underestimates the improvement of image classification accuracies (Keating, 1995). [end p. 84]
THE UTILITY OF VEGETATION MAPPING FOR RESOURCE CONSERVATION
Limitations of Available Sources of Remotely Sensed Data
Even higher accuracies would have been obtained if past human disturbances had not imparted so much fine-scale spatial heterogeneity to the landscape. Near both the upper and lower treelines in the study area, I encountered the "mixed-pixel" problem (e.g., Vanclay and Preston, 1990). One pixel, which covers 900 m2 on the ground, may include two or three different plant communities. Therefore, separating páramo communities at the higher elevations, or different types of degraded areas at the lower elevations, is very difficult. Many Andean landscapes are characterized by a complex mosaic of disturbed areas, which may exhibit various states of regeneration. Analysis of TM data alone will probably not enable geographers to produce accurate, large-scale land cover maps of these areas.
Topographic normalization does enhance land-cover classification accuracy, and the differences between maps produced from normalized and uncorrected images are readily apparent, especially in steeper terrain. However, two substantial problems may preclude this technique from becoming widespread. First, the algorithms are not part of standard software programs, nor are they easy to use (see Colby and Keating, submitted). Second, and more significantly, topographic normalization cannot be implemented without a high-quality, detailed DEM, which is utilized to calculate slope and aspect for each pixel in the image to be corrected. While creating a relatively crude DEM for the air photo map analysis required only several days of work, digitizing each contour line to create a precise DEM for the small study site required an inordinate amount of time. Given the amount of time required to correct for anisotropic reflectance, it is doubtful that most Latin American researchers will adopt this technique until both digital elevation models and more powerful computers for image analysis are available in developing countries.
Another significant difficulty impeding the application of remote sensing techniques in Latin America is the availability of images. While MSS images can be purchased for most areas in Latin America, they are unsuitable for mapping complex [end p. 85] tropical land cover and frequently contain significant cloud cover. High-quality TM images for most Andean regions are not archived by the EOSAT corporation (pers. obser.), or any other North American organization. While the Ecuadorian military receives TM imagery directly from satellites at the CLIRSEN-Cotopaxi data receiving station,7 obtaining these images from military sources may be problematical. In particular, because the Loja Province is located near the Peruvian border, obtaining any form of remotely sensed data for this region is difficult.
Although aerial photographs are cheaper and more readily obtainable than digital images, Ecuadorian photos are typically inadequate for monitoring fine-scale land cover changes due to several reasons. First, because photos are flown infrequently in most provinces, ecologists have few opportunities to monitor deforestation at regular intervals; air photos for the Loja province were taken only in 1963, 1976 and 1989. Second, the quality of air photos is often unsuitable for most mapping projects. Both the small scale and the inferior contrast ratios of these photographs render them useful only for delineating forest from "deforestation."
Finally, another problem concerns differences between the photography of 1976 and 1989. Because the flight lines of the two years were slightly offset, differences in the amount of relief displacement were rather pronounced. Polygons of the same size often appeared to be very different in 1989 than in the earlier photos. Therefore, the position of terrain features, i.e., "site" and "association," were emphasized as primary interpretation cues (sensu Lillesand and Kiefer, 1987).
Optimizing the Use of Geographic Information Technologies in Loja
To document the spatial extent of patchy ecosystems, large scale aerial photos could be interpreted, and digitized polygons would ultimately be put into a raster-based GIS. Given the difficulty of obtaining high-quality aerial photographs in most developing countries, researchers should have air photos flown when possible. Large-scale (1: 24,000) infrared photos are often superior to digital imagery for delineating cultural features and fine-scale vegetation patterns in areas the size of western Podocarpus National Park. Moreover, if fiducial marks8 are present on each photo, digital elevation models could created directly from the photos.
Maps created from either of these two image types could be updated periodically in a GIS with recent maps of cut and burned areas. Moreover, other data sources could be added to the preliminary digital maps to create detailed vegetation maps (e.g., Sirait et al. 1994). As more knowledge concerning the floristic composition of different altitudinal zones becomes available, park managers would be able to determine which areas face the highest risk of degradation in the future. To date, few studies in any region have examined changes in disturbance regimes along altitudinal gradients and how these disturbances interact differentially with vegetation structure (but see Veblen et al. 1992). If more detailed maps of plant communities were analyzed in relation to current disturbance regimes, more effective park policies could be formulated.
Whereas North American-style preservation (see Vale, 1987) has been unsuccessfully practiced in Podocarpus National Park during the past twelve years, some government officials now foresee the implementation of a "zonation scheme" in the near future. They envision a management plan that includes 1) a large "wilderness" area, in which no forms of farming or resource extraction would be practiced, 2) a zone of intensive use, in which pre-existing settlements would remain functional, and 3) a "buffer zone" that would separate the latter two zones. [end p. 86]
Clearly a GIS system would enable members of local NGOs and government officials to implement such a system more effectively. As more farms and mining sites are established in the park, habitats for numerous species of plants and animals will either be destroyed or fragmented. Forest fragmentation has become a severe problem in many Andean areas (see Young and Le6n, 1995) and may become one of the principal threats to the Park's ecosystems in the future. It would certainly be desirable to concentrate human use of the Park in several zones, so that fragile communities, particularly on the higher slopes, remain intact.
Remote Sensing and Conservation Efforts:Prospects for the Future
Accurate maps, which could be stored and managed with an inexpensive GIS such as IDRISI,9 would serve the local community by facilitating an understanding of where the park is, what it contains, and how it is being influenced by current human activities. Moreover, it would provide a substantive basis for dialogue between NGOs, INEFAN (the National Forestry Institute), and users of the Park.
ARCOIRIS, the principal conservation group in Loja, now receives substantial support from Fundación Natura and other large conservation groups. They have initiated and coordinated conserrvation projects that include foreign researchers, government officials and Ecuadorian conservationists. Not only have they conducted biological inventories in some sections of the Park, but they have surveyed communities in the Park to determine the needs and activities of local campesinos.
Some members of ARCOIRIS use GIS programs, and with further technical assistance, may soon be in a position to recommend policy changes throughout the region. Many of the geographic information processing techniques described in this article could conceivably be implemented on a park-wide basis. While economic and environmental conditions of the Loja canton may continue to deteriorate, mapping techniques could at least give conservationists a set of tools for developing more viable resource management strategies.
ACKNOWLEDGMENTS
NOTES
2. AVHRR stands for Advanced Very High Resolution Radiometer. This device is contained in satellites operated by NOAA, the National Oceanic and Atmospheric Administration.
3. A remote sensor's spectral resolution is determined by both the number and width of wavelength intervals in the electromagnetic spectrum that it detects. Landsat MSS records energy in only four relatively wide bands, whereas TM has six narrower bands. MSS imagery is less suitable than TM imagery for vegetation mapping in part because it detects fewer of the subtle differences between land cover types.
4. The verano, or summer, in Ecuador is a season during which precipitation is lower than normal. In Loja Province, this season is very short and called a veranillo. It usually occurs between mid-November and late December.
5. Black-body subtraction is an algorithm that reduces the effects of atmospheric scattering, or haze. To implement this procedure, one first constructs histograms that depict the frequency of brightness values for each band. The minimum value of the histogram, or some number close to it, is then subtracted from each pixel in the band so that the entire histogram is shifted to the left; the minimum value for each band is then zero. Jensen (1996, 107-122) covers this technique in greater detail.
6. Land cover classification with digital imagery is often unsuccessful in mountainous terrain. Surfaces that exhibit the same cover type but which are characterized by different slope and aspect angles may reflect different quantities of radiation (see Colby and Keating, submitted). These uneven reflectance patterns, or anisotropic reflectance, may impair the image classification process so that areas of identical land cover are classified as different cover types.
7. CLIRSEN (Centro de Levantamientos Integrados de Recursos Naturales por Sensores Remotos) is an Ecuadorian government agency responsible for collecting and distributing Landsat digital imagery. The CLIRSEN-Cotopaxi data receiving station is located approximately thirty kilometers south of Quito.
8. Fiducial marks define a frame of reference on aerial photographs that allow one to make precise spatial measurements on the photos. High-quality mapping cameras are calibrated so that exact distances between fiducial marks are known (Lillesand and Kiefer, 1987). These marks appear either in the corners or the margins of the photographs.
9 IDRISI is a raster-based GIS produced by the Clark Labs for Cartographic Technology and Geographic Analysis, Clark University. Because this package is very powerful yet inexpensive and easy to use, it has been adopted by government agencies and NGOs in many developing countries.
REFERENCES
Andrade, P. 1996. "Riqueza del Parque Nacional Podocarpus". Arcoiris II: 1-4.
Andre, G.G. 1987. Diagnóstico de Los Incendios Forestales. Uso y Manejo del Fuego Con Fines Agropecuarios en la Provincia de Loja. Thesis, Universidad Nacional de Loja, Ecuador.
Apolo, W. 1984. El Plan de Manejo: El Parque Nacional Podocarpus. (Quito: EI Ministerio de Agricultura y Ganadería)
Archard, F. and F. Blasco. 1990. "Analysis of vegetation seasonal evolution and mapping forest cover in West Africa with the use of NOAA AVHRR HRPT data," Photogrammetric Engineering and Remote Sensing 56: 1359-1365.
Autodesk. 1992. AutoCAD Release 12: reference manual. (San Rafael, California: Autodesk, Inc.)
Bogh, A. 1992. "Composition and distribution of the vascular epiphyte flora of an Ecuadorian montane rain forest," Selbyena 13: 25-34.
Botkin, D. B. 1990. Discordant Harmonies: a new ecology for the twenty-first century. (New York: Oxford University Press)
Brownrigg, L. A. 1981. "Economic and ecological strategies of Lojano migrants to El Oro. In: Whitten, N.E. (ed.), Cultural Transformations and Ethnicity in Modern Ecuador. (Urbana, Illinois: University of Illinois Press): 303-326.
Christensen, L. J. 1989. An Early Status of Regeneration in Gaps In a Montane Rain Forest in Southern Ecuador. Botanisk Institut, University of Aarhus, Denmark. [end p. 88]
Colby, J.D. and P.L. Keating. Submitted. "Land cover classification using Landsat TM imagery in the tropical highlands: the influence of anisotropic reflectance," Under consideration by the International Journal of Remote Sensing.
Dodson, E. H. and A.H. Gentry. 1991. "Biological extinction in western Ecuador," Annals of the Missouri Botanical Garden 78: 273-295.
Eastman, R. J. 1993. IDRISI Update Manual, version 4. J. (Worcester, Massachusetts: Clark University)
Echavarria, F. R. 1991. "Cuantificacion de la deforestación en el valle del Huallaga, Peru," Revista Geográfica 114: 37-53.
____. 1993. Remote Sensing of Montane Forest Degradation in Southern Ecuador. Unpublished Ph.D. dissertation, Department of Geography, University of South Carolina.
Eckholm, E. P. 1975. "The deterioration of mountain environments," Science 189: 764-770.
Ellenberg, H. 1979. "Man's influence on tropical mountain ecosystems in South America," Journal of Ecology 67: 401-416.
Engel, FA. 1976. An Ancient World Preserved. Relics and Records of Prehistory in the Andes. (New York: Crown Publishers)
Erdas. 1994. Erdas Field Guide. (Atlanta, Georgia: Erdas, Inc.)
Espinosa, D.A., E. Gallagher, P.A. Cueva, E.G. Gómez, and E.S. Bermeo. 1992. "Diagnóstico socio-económico del area de amortiguamiento occidental y sur oriental del Parque Nacional Podocarpus," Boletín Informativo Sobre Biología, Conservación y Vida Silvestre. (Loja, Ecuador: Fundación Podocarpus)
Espinosa, R. 1948. Estudios Botánicas en el Sur del Ecuador, Volumen J. (Loja, Ecuador: Universidad Nacional de Loja)
ESRI. 1994. ARC/INFO User's Guide. (Redlands, California: Environmental Science Research Institute)
Fauroux, E. 1986. "Cambio social y utilización diferencial del medio natural: el ejemplo de Loja," Revista Cultura 24: 6733689.
Fundacion Natura. 1990. Evidencias del Deterioro Ambiental en el Ecuador. (Quito, Ecuador: Fundación Natura)
__________. 1992a. Acciones de Desarollo y Areas Naturales Protegidas en el Ecuador. (Quito, Ecuador: Fundación Natura)
__________. 1992b. Acciones de Desarollo y Areas Naturales Protegidas en el Ecuador #5: Podocarpus National Park. (Quito, Ecuador: Fundación Natura)
Garda, M.C. and D.R. Alvarez. 1994. ''TM digital processing of a tropical forest region in southeastern Mexico," International Journal of Remote Sensing 15: 1611-1632.
Glaser, G. and J. Celecia. 1981. "Guidelines for integrated ecological research in the Andean region," Mountain Research and Development 1: 171-186.
Gómez, N. 1989. Elementos de Geografía del Ecuador. (Quito, Ecuador: Coleccion Imagenes de la Tierra)
Grubb, P.J. 1970. "The impact of man on the páramos of Cerro Antisana, Ecuador," Journal of Applied Ecology 2: 7-8.
Harden, e.P. 1991. "Andean soil erosion," National Geographic Research and Exploration 7: 216-231.
Hess, E.G. 1990. "'Moving up-moving down': agro-pastoral land-use patterns in the Ecuadorian páramos," Mountain Research and Development 10: 333-342.
Hill, R.A. and G.M. Foody. 1994. "Separability of tropical rainnforest types in Tambopata-Candamo Reserved Zone, Peru," International Journal of Remote Sensing 15: 2687-2693.
Hodgson, M.E. and B.M. Shelly. 1994. "Removing the topographic effect in remotely sensed imagery," ERDAS Monitor 6: 4-6.
INMH. 1993. Meteorological data sets for Loja Province. (lnstituto Nacional de Meteorología e Hidrografía, Quito, Ecuador).
Jakubauskas, M.E., K.P. Lulla, and P.W. Mausel. 1990. "Assessment of vegetation change in a fire-altered forest landscape," Photogrammetric Engineering and Remote Sensing 56:371-377.
Jensen, J.R. 1996. Introductory Digital Image Processing: a remote sensing perspective, second edition. (Upper Saddle River, New Jersey: Prentice Hall): 107-122, 257-262.
Keating, P.L. 1995. Disturbance Regimes and Regeneration Dynamics of Upper Montane Forests and Páramos in the Southern Ecuadorian Andes. Unpublished Ph.D. Dissertation, Department of Geography, University of Colorado at Boulder.
__________. In press. "Patrones y procesos de deforestación en el Ecuador," Revista Cultura 33 (March, 1997). (Quito, Ecuador: EI Banco Central)
__________. In prep. "Changes in the composition and structure of paramo communities along environmental gradients in Podocarpus National Park, Ecuador," For submission to the Journal of Biogeography.
Knapp, G. 1991. Geografía Quichua de la Sierra del Ecuador: núcleos, dominios, y esfera, third edition. (Quito, Ecuador: Abya Yala)
Kremen, C., A.M. Merlenlender, and D.D. Murphy. 1994. "Ecological monitoring: a vital need for integrated conservation and development programs in the tropics," Conservation Biology 8: 388-397.
Laegaard, S. 1992. "Influence of fire in the grass páramo vegetation of Ecuador," in Paramo: an Andean ecosystem under human influence, (ed.) Balslev, H. and J.L. Luteyn (New York: Academic Press): 151-170.
Lanley, J. 1982. Tropical Forest Resources, Food and Agriculture Organization of the United Nations, FAO Forestry Paper 30.
Lillesand, T.M. and R.W. Kiefer. 1987. Remote Sensing and Image Interpretation, second edition (New York: John Wiley and Sons).
Madsen, J.E. and B. Ollgaard. 1993. "Floristic composition, structure, and dynamics of an upper montane rain forest in southern Ecuador," Nordic Journal of Botany 14: 403-423. [end p. 89]
Malingreau, J.P., C.J. Tucker, and N. Laporte. 1989. "AVHRR for monitoring tropical deforestation," International Journal of Remote Sensing 10: 855-867.
Malingreau, J.P. and C.J. Tucker. 1988. "Large-scale deforestation in the southeastern Amazon Basin in Brazil," Ambio 17: 49-55.
Miller, E.Y. 1959. "Agricultural Ecuador," Geographical Review 49: 183-207.
Millones 0. J. 1982. "Patterns of landuse and associated environmental problems of the central Andes: an integrated summary," Mountain Research and Development 2: 49-61.
Morris, A. 1985. "Forestry and land-use conflicts in Cuenca, Ecuador," Mountain Research and Development 5: 183-196.
Mulders, M.A., S. De Bruin, and B.P. Schuilig.1992. "Structured approach to land cover mapping of the Atlantic zone of Costa Rica using single date TM data," International Journal of Remote Sensing 13: 3017-3033.
Myers, N. 1988. "Tropical deforestation and remote sensing," Forest Ecology and Management 23: 215-225.
Nelson, R., N. Horning, and T.A. Stone. 1987. "Determining the rate of forest conversion in Mato Grosso, Brazil, using Landsat MSS and AVHRR data," International Journal of Remote Sensing 8: 1767-1784.
Parsons, J. J. 1982. "The northern Andean environment," Mountain Research and Development 2: 253-262.
Pietri-Levy, A. 1993. Loja: una provincia del Ecuador. (Quito, Ecuador: EI Banco Central del Ecuador)
Sader, S.A., T.A. Stone and AT. Joyce. 1990. "Remote sensing of tropical forests: an overview of research and applications using non-photographic sensors," Photogrammetric Engineering and Remote Sensing 56: 1343-1351.
Sanchez, R. and O. Vivanco. 1991. Perspectivos Para el Monitoreo de la Deforestación en la Región Amazónica Ecuatoriana. (Quito, Ecuador: CLlRSEN)
Sayn- Wittgenstein, L. 1980. "New developments in tropical forest inventories," Proceedings of the 14th Symposium in Remote Sensing of the Environment, San Jose, Costa Rica: Pp. 118771194.
Sierra, R. 1996. La Deforestación en el Nor-ccidente del Ecuador: 1983-1993. (Quito, Ecuador: EcoCiencia)
Sirait, M., S. Prasodjo, N. Podger, A. Flavelle, and J. Fox. 1994. Mapping customary land use in East Kalimantan, Indonesia: a tool for forest management. Ambio 23: 411-417.
Skole, D. and E. Tucker. 1993. Tropical deforestation and habitat fragmentation in the Amazon: satellite data from 1978 and 1988. Science 260: 1905-1910.
Southgate, D. and H.L. Clark. 1993. Can conservation projects save biodiversity in Latin America? Ambio 22: 163-166.
Southgate, D., R. Sierra, and L. Brown. 1991. The causes of tropical deforestation in Ecuador: a statistical analysis. World Development 19: 1145-1151.
Troll, E. 1973. The upper timberlines in different climatic zones. Arctic and Alpine Research 5: A3-A 18.
Tucker, C.J., B.N. Holben, and T.E. Goff. 1984. "Intensive forest clearing in Rondonia, Brazil, as detected by satellite remote sensing," Remote Sensing of Environment 15: 255-261.
Vale, T.R. 1987. "Vegetation change and park purposes in the high elevations of Yosemite National Park, California," Annals of the Association of American Geographers 77: 1-18.
Vanclay, J.K. and R.A. Preston. 1990. "Utility of Landsat Thematic Mapper data for mapping site productivity in tropical moist forests," Photogrammetric Engineering and Remote Sensing 56: 1383-1388.
Van der Hammen, T. 1974. "The Pleistocene changes of vegetation and climate in tropical South America," Journal of Biogeography I: 3-26.
Veblen, T.T., K.S. Hadley, E.M. Nel, T. Kitzberger, M. Reid, and R. Villalba. 1994. "Disturbance regime and disturbance interactions in a Rocky Mountain subalpine forest," Journal of Ecology 82: 125-135.
Veblen, T.T., T. Kitzberger, and A. Lara. 1992. "Disturbance and forest dynamics along a transect from Andean rainforest to Patagonian shrubland," Journal of Vegetation Science 3: 2499258.
Welch, R., T.R. Jordan and J.e. Luvall. 1990. "Geocoding and stereo display of tropical forest multi sensor data sets," Photogrammetric Engineering and Remote Sensing 56: 138991392.
Young, K.R. 1994. "Roads and environmental degradation of tropical montane forests," Conservation Biology 8: 972-976.
Young, K.R. and B. León. 1995. "Distribution and conservation of Peru's montane forests: interactions between the biota and human society," in Tropical Montane Cloud Forests (ed.) Hamilton, L.S., 1.0. Juvik and EN. Scatena (New York: Springer-Verlag): 363-376.
Zimmerer, K.S. 1994. "Human geography and the 'new ecology': the prospect and promise of integration." Annals of the Association of American Geographers 84: 108-125.
RESUMEN
Due to mining, logging and inappropriate agricultural techniques, most of Loja Province has experienced constant environmental degradation since the 1700s (see Fauroux, 1986). Local anthropogenic disturbances, however, have become increasingly severe since 1940. Valuable tree species, such as >i
After a series of droughts in the late 1960s and early 1970s resulted in dramatically lower agricultural output (Pietri-Levy, 1993), several conservation groups strongly recommended creating a protected area. A comprehensive management plan was written (Apolo, 1984), but it satisfied the political objectives of large conservation groups more than it met the needs of the local people. Designed primarily to preserve both the biological diversity and water resources of the region, it excluded nearly all forms of land use within the Park. The Ministry of Agriculture did not make provisions [end p. 81] for pre-existing farms within the newly protected area, as the park boundary had not been adequately surveyed (see Apolo, 1984). Although only five park guards patrol this section of the Park, they are expected to ensure that absolutely no colonization, mining, or hunting occurs.
The research described below was conducted to delineate as many land cover types as possible in northwest Podocarpus National Park. Initially, black-and-white aerial photographs from both 1976 and 1989 were obtained for the western section of Podocarpus National Park and adjacent areas. The study area for the air photo project includes 27,518 ha of forest and páramo, extending from the lake district down to the highway. Patterns and magnitudes of deforestation along the western section of were documented over a thirteen year interval. Secondly, Landsat imagery collected in 1987 and 1989 was used to map various land cover types within a 8 km x 13.5 km (108 km2) portion of this area. The digital image processing was performed for a smaller area of the Park so that detailed ground-truthing and accuracy assessments could be performed.
Aerial Photographs (scale 1:60,000) were obtained from the Instituto Geográfico Militar in Quito. Topographic maps of scales 1:100,000 and 1:50,000 were also used in the interpretation process. With the use of a mirror stereoscope, I classified land cover as páramo, forest or "deforested" areas («2600m elevation); fine-scale delineation of upper montane plant communities was not possible due to the poor contrast ratio of the 1989 photographs. For this article, "deforested" refers to areas that have been cleared of mature forests and then disturbed repeatedly. To my knowledge, no deforested areas have undergone complete recovery after cutting and burning.
Because southern Ecuador is covered with clouds most of the year, few Landsat images are suitable for remote sensing projects in this area. With the assistance of the Ecuadorian military, I obtained the only cloud and haze-free Landsat TM image available for western Podocarpus National Park. Collected on 9 November 1989, this 15 x 15 km subscene includes six bands. Secondly, a Landsat MSS image from 2 November 1986 was obtained from the EOSAT corporation. In this image, clouds cover the top of the Cordillera Oriental as well as some of the forested areas below 2600m elevation.
Classified TM images included more land cover categories than did the other images examined, and may be used to map deforestation quite accurately. Moreover, the overall accuracy values obtained with the classified TM images compare favorably with those obtained by other researchers (e.g., García and Alvarez, 1994); very few researchers working in tropical areas report values above 85 percent. This study has demonstrated, therefore, that TM imagery has considerable utility in tropical highland areas if sophisticated digital image processing techniques are included in the mapping process.
Ultimately, park managers and conservationists in Loja need a spatial data base that would enable users to 1) map land cover, 2) measure the area and spatial distributions of cover types, 3) monitor rates and directions of disturbances, and 4) assess the risk of environmental degradation in different life zones. Meeting these needs would clearly involve several different geographic information technologies. Given the difficulties associated with separating tropical vegetation communities, TM images would be most suitable for mapping sections of Podocarpus National Park at the regional scale. Even when topographic normalization is not performed, classified TM images may be used to produce reasonably accurate maps of forest, deforestation and paramo. The relatively crude (but accurate) maps created with TM imagery could be used primarily for determining where more detailed mapping projects should be conducted.
While digital maps and spatial data bases would provide some information necessary for park management, they would certainly not be sufficient to ensure the long-term success of local conservation efforts. As Southgate and Clark (1992) have indicated, conservation projects in South America are often only partially successful due to the failure of local communities to appreciate natural ecosystems. Many Lojanos who recognize the existence of the Park either do not understand the conservation objectives or do not believe that they would benefit the local people. The notion of conservation itself may be alien to peoples who have experienced extreme poverty. It is hoped that both financial resources and the techniques discussed above will be applied to the integration of conservation efforts with long-term economic development of the region (e.g., Kremen et al. 1994).
This project was funded by the John D. and Catherine T. MacArthur Foundation, the Fulbright Commission of Ecuador, and the Ibero-Latin American Studies Center at the University of Colorado. CLIRSEN and the Instituto Geográfico Militar of Quito provided the necessary digital imagery, aerial photographs and maps for all remote sensing projects. I wish to thank Drs. M. Hodgson and C. Wessman for giving technical assistance and advice as I acquired expertise in remote sensing. I am also grateful to J. Robb for making the maps necessary for this publication. Ing. Santos Calderón and Pedro Cabrera of Loja were very generous in providing logistical support and lodging as I performed field work in Podocarpus National Park. Finally, I am very grateful to Dr. T. Veblen for his support and the use of his equipment while I conducted the research for this article. [end p. 87]
1. Whereas landuse refers to the human activities that occur in a given area, land cover refers to the type of physical feature present in that area (Lilies and and Kiefer, 1987). Lakes, forest, paramo and pastures are examples of land cover types.
Several topographic normalization techniques may be used to remove anisotropic reflectance before image classification is performed. If one assumes that a surface scatters light equally in all directions, one may employ a Lambertian model. In this study, a topographic normalization algorithm based upon this model (Hodgson and Shelly, 1994) was used to remove anisotropic reflectance from both images. Secondly, because this assumption is not always met, I also utilized a non-Lambertian model. This procedure is quite complex and entails the calculation of a Minnaert constant for each band, which describes a bidirectional reflection distribution function. To implement the non-Lambertian model, I used the algorithm mentioned above in conjunction with a regression technique that is performed with a spreadsheet (Keating, 1995). Both techniques require a very detailed digital elevation model and knowledge of the solar elevation and azimuth angles. For both TM and MSS images, I classified uncorrected, Lambertian-corrected and non-Lambertian-corrected images.
Relatively few organizations in Ecuador use GIS. To my knowledge, ARCOIRIS is the only organization in southern Ecuador that has acquired IDRISI. While members of ARCOIRIS understand the basics of GIS, they have not yet undertaken complex GIS projects. Located in Quito, Ecociencia and CDC (Centro de Datos para la Conservación) both have large geographic information processing laboratories that are equipped with state-of-the-art computers and advanced software packages.
Adeniyi, P.O. 1980. "Land-cover change analysis using sequential aerial photography and computer techniques," Photogrammetric Engineering and Remote Sensing 46: 1447-1464.
Recientemente, las áreas protegidas del altiplano ecuatoriano han sido amenazadas por severas perturbaciones antropogénicas. Dada la biodiversidad y heterogeneidad espacial del ecosistema andino, los geógrafos deben desarrollar métodos confiables para preparar mapas de estas áreas. Mientras que las técnicas de sensores remotos no han sido extensamente aplicadas a paisajes tropicales de montaña, estas técnicas tienen el potencial de proveer mapas de vegetación y zonas de alteración antropogénicas muy útiles. Este artículo documenta el uso de fotografías áereas e imágenes digitales Landsat para la preparación de mapas de comunidades de vegetación y para la evaluación de las tasas de deforestación en el sur de los Andes ecuatorianos. Discute también la utilidad y limitaciones de varios procedimientos para la creación de mapas en relación a la conservación de los recursos naturales. [end p. 90]