TIN - Triangulated Irregular Network. A vector-based surface topography representation, which uses point data and creates triangles that contain slope and elevation data. An alternative to rasterized DEM datasets.

Locations of each study, referenced by number.

Annotated Bibliography

1. Keese, James, Mastin, Thomas, and David Yun. 2007. Identifying and Assessing Tropical Montane Forests on the Eastern Flank of the Ecuadorian Andes. Journal of Latin American Geography 6(1): 63-84.

This study examined the potential for a conservation "corridor" between two national parks (Podocarpus and Sangay) in the Andes mountains of Ecuador. The mountain forests of the Andes exhibit high rates of biodiversity (total number of species) and endemism (exclusively native species), making them a "global hotspot" for conservation. Conversion of forest for other land-uses, including small-holder agriculture and grazing of livestock, threatens these forests by fragmenting them and creating more forest "edge," which is harmful to the native species of flora and fauna which are not adapted to these disturbances.

The authors of this study used Landsat satellite data (30 meter pixel size) to classify land-use in their study area, identifying four broad land-cover types: montane forest, paramo (alpine grassland), agriculture/pasture, and other (all other land cover types). Using field-gathered GPS data, they were able to "ground-truth" their classification types in a GIS. After the classification, the raster data structure was converted to vector to create polygons representative of coterminous land-cover types, the most important being forested area in this study. Spatial analysis of this data included buffering to determine the correlation of deforestation and roads. Results showed higher deforestation near the roads, with 23 percent of land within one kilometer of the roads deforested, and only 14 percent using a five kilometer buffer. An analysis of the correlation of slope of the land and deforestation was also performed, using a DEM (Digital Elevation Model) to derive slope percentages. Not surprisingly, the steepest areas had a higher percentage of forest cover, as they offer less access than flatter lands. Agriculture plots were not as strongly correlated with slope, being found in steep as well as flat areas. As a result of this analysis, the authors recommended focusing areas of conservation in the remote. road-free areas, and more topographically extreme parts of their study area.

2. Rosero-Bixby, Luis, and Alberto Palloni. 1998. Population and Deforestation in Costa Rica. Population & Environment 20(2): 149-185.

Population growth has been implicated as a strong determinant of deforestation, simply because the needs of a larger group of people often involve greater natural resource demands. This study looks more closely at population and deforestation in Costa Rica, which experienced rapid population growth after World War II and subsequent loss in forest cover (estimated at 50% of the total). The authors used a GIS to combine three layers of data for this study: land use data to derive forest cover, physical and cultural features, and census data. The census data was geo-coded (referenced to points on the earth's surface) to match census tracts. Forest and population layers were sampled to continuous (raster) datasets. This allowed for assignment of variables such as population potential, percentage below poverty line, and fertility to each pixel. Results showed a strong negative correlation between the presence of humans and presence of forests, as would be expected, but many interesting findings emerge from the exploration of other demographic variables. The increasing presence of poverty was tied to higher deforestation rates except for areas which had a population completely below the poverty line. Some unexpected results were also noted. For example, with each added child per woman, deforestation was reduced seven percent. Data from this study suggests that there are many dynamic factors at work in settlement and deforestation, and studies incorporating population and deforestation will never be able to show a direct correlation. Despite this, they can show important local aspects of deforestation that can be further examined from a social science perspective.

3. Arima, Eugenio Y., Robert T. Walker, Stephen G. Perz, and Marcellus Caldas. 2005. Loggers and Forest Fragmentation: Behavioral Models of Road Building in the Amazon Basin. Annals of the Association of American Geographers 95(3): 525-541.

The modeling potential of GIS was again on display in this study, but instead of modeling the forest, the authors modeled one of the drivers of deforestation: roads. This study took place in the largest tropical rain forest of the world (and definitely the most famous for deforestation occurring there), the Brazilian Amazon. Deforestation and forest fragmentation are driven largely by agricultural colonization along roads which penetrate into the forest, which spawn feeder roads. The resulting matrix of roads resembles a "fishbone" pattern, a term often applied to the deforestation in the Amazon. These smaller roads are often build perpendicular to the state-built highway. Though colonists fuel the initial road-building, logging companies continue the road construction when resources are exhausted. This study attempted to predict where a road would be built using a lowest-cost analysis in GIS, factoring in slope and other potential costs, like river crossings. They predicted both a "destination determined" and a "destination indeterminate" road, which they compared to the actual roads that were built. The undetermined road route model was not very successful, but the determined route model was. This leads to an interesting conclusion. The logging companies that build the roads seemed to follow the colonists' preferred road path rather than the most cost-effective path for them. Indeed, many local factors work to determine a route, and political maneuvers- by colonists, logging companies and governments- determine how the Amazon becomes fragmented.

4. Apan Armando, and James Peterson. 1998. Probing Tropical Deforestation: The Use of GIS and Statistical Analysis of Georeferenced Data. Applied Geography 18(2):137-152.

Armando and Peterson examined deforestation in Mindoro, Philippines, looking at a wide variety of possible controlling factors and analyzing their statistical correlation with deforested area. Several examples of these factors include slope, elevation. distance from roads and rivers, soil type, ownership of land, and population density. This data came from a variety of sources and needed to be digitized and geo-referenced by the authors, after which each factor became a layer in the GIS. A DEM was created using topographic maps, digitizing the data into a TIN structure, and then creating a 30 meter pixel raster data from the TIN. Landsat data was used to determine land cover, and a simple classification of either "forest" or "non-forest" was used. The statistical test of the data acquired surprisingly showed a very weak magnitude of relationship for all of the variables examined. This may have been due, the authors note, to possible errors in the original dataset, which may have been outdated. A GIS analysis include many attribute table data queries and analysis led to some interesting conclusions; for example, most (60 percent) of the total deforested area was in areas with slope greater than 60 percent. This led the authors to conclusions that forest fires may be one of the main causes of deforestation. Since fires are more random than anthropogenic disturbances, little correlation would be expected between most environmental variables and deforestation, and this was evident in their results.

5. Menon, Shaily, R. Gil Pontius Jr, Joseph Rose, M. L. Khan, and Kamaljit S. Bawa. 2001. Identifying Conservation-Priority Areas in the Tropics: A Land-Use Change Modeling Approach. Conservation Biology 15(2): 501-512.

Because of the high levels of biodiversity in tropical forests, studies often aim to determine conservation potential of these areas. This study examined Arunachal Pradesh, a state in northeast India. There are five distinct types of forest in this state, generally following elevation gradients. Elevation is highly varied in this state, ranging from 500 meters to 7000 in the Himalaya. A relatively low population density resulted in 70 percent forest cover in 1988. Data on forest type an extent was digitized to vector format from the 1988 map, as well as roads, cities, rivers, and protected areas. Land cover types were then changed to raster format to allow for future modeling by a computer program, GMODO2. This program estimates where future deforestation will occur based on controlling features such as roads, rivers, cities, elevation. The authors of this study used a per-capita deforestation estimate combined with population estimates to determine the future rate of deforestation, so the program knew how many pixels to change from forest to non-forest. Results were somewhat discouraging, predicting a 50% loss of current forest cover by 2021. Areas that were projected to be deforested and not protected were thought to be of highest priority for conservation, especially in the low elevation wet-evergreen forests. Areas of projected deforestation within currently protected areas also need attention, due to the lack of enforcement of this protected status. These "paper parks" (areas protected only on paper) are a common theme in the developing world and present challenges to conservation efforts.

6. Muller, Daniel, and Manfred Zeller. 2002. Land use dynamics in the central highlands of Vietnam: a spatial model combining village survey data with satellite imagery interpretation. Agricultural Economics 27(3): 333-354.

The authors of this study developed a complex statistical model which they used to determine the impact of a variety of geophysical and socioeconomic factors. The study are was in the province of Dak Lak in Vietnam, one of the less densely populated areas of the country. Forested area was determined at three separate years of analysis, corresponding to years of policy changes. GIS was used to vectorize several cultural features, import GPS points with household survey data, as well as analyze the images after their classification into five different land-cover types. To determine community boundaries, a GIS process creating Thiessen polygons was performed. These polygons are created from point data (in this case community center-points) and are constructed so that all areas within each polygon are closest to the points within that polygon. This allowed for delineation of the study area into "community" sections. Results of the land-cover calculations showed 10% deforestation from 1975 to 1992, but 8% regrowth from 1992 to 2000, indicating changes in land-use. These were largely due to policies of market integration of the rural area, making many new income-generation possibilities. Intensification of agricultural land-use- using labor-intensive practices combined with technologies like irrigation and pesticides for higher yields- instead of land-intensive regimes was also implicated as a driver of this change. It is important to study these rare areas of reforestation in the tropics, to understand how developing countries can colonize new areas without completely removing the forests, and the processes that help foster this type of situation.

7. Zhang, Quanfa, Christopher Justice, Mingxi Jiang, Jake Brunner, and David Wilkie. 2006. A Gis-based assessment on the vulnerability and future extent of the tropical forests of the Congo basin. Environmental Monitoring and Assessment. 114(1): 107-121.

This study used GIS to model potential forest cover in the Congo Basin, where most of the African tropical rain forest is found. With thirty million people living in the area and population estimates predicting two to three percent of growth per year, increases in deforested area as well as deforestation rates are expected. This study modeled deforestation out to the year 2050. They used pre-defined deforestation rates for each country (ranging from 0.3% to 0.6%) in the region and linearly increased them to a one percent yearly rate by 2050 to account for the increased demand for land, primarily for shifting agriculture. Their model was created a grid with ten kilometer pixel size. To develop their model, they assigned rankings for four variables- population density, road density, logging history, and existing forest fragmentation. A summation of these variables for each pixel allowed the authors to "rank" each pixel according to the probability of deforestation. With each pixel ranked (with the highest numbers being the most susceptible to deforestation), a pre-determined number of the highest pixels were changed from forest to non-forest for each year, depending on the pre-determined rate for that year. The authors predicted a fragmentation of Africa's rain forest into three large contiguous areas, but more importantly, noticed a 20% loss in the percentage of "interior" forest (forest ten kilometers away from other types of land uses). Forests further away from human disturbances tend to be more biodiverse, and some species need these large ranges of intact forests for a suitable habitat. This study revealed the use of GIS for predictive mapping, which could be applied in selecting areas for conservation in this important eco-region.

8. Baldyga, Tracy J., Scott N. Miller, Kenneth L. Driese, and Charles Maina Gichaba. 2008. Assessing land cover change in Kenya's Mau Forest region using remotely sensed data. African Journal of Ecology. 46(1): 46-54.

This study examines land use changes were the topic of in the River Njoro watershed of Kenya. This area in the central eastern part of the country possesses a varied topography and elevations ranging from 1700 to 3000 m.a.s.l. This study incorporated temporal change, meaning that several years of land cover data were analyzed to examine the rate and areas of land cover change. The authors used Landsat data from three separate dates (over a span of 17 years) and used classifications to assign land cover values to each pixel in the study area for each year. Spectral reflectance captured in Landsat image pixels allows for separation into classes, because similar land cover types will have similar spectral reflectance. The authors chose to separate the land cover types into nine classes, and they encountered problems in distinguishing between similar land cover types. This region contained both native and plantation forests, and the differences between these two land covers was difficult to assess. GIS spatial analyses were developed for the study which change out-of-place pixels. For example, a pixel assigned to an "urban" land cover within a national park was automatically reassigned to a more applicable class, such as vegetation. To measure accuracy, GPS points from field work in the region were compared with the classifications, a sort of "ground-truthing" for the satellite-derived data. Results from the study were examined at several spatial scales. For example, the uplands showed loss of forests at around a one percent rate per year, while at a regional or watershed scale, rates were much slower. Landsat data was found sufficient in both accuracy and scale in determining both local and regional changes, but more research on distinguishing vegetation types was suggested by the authors for a more complete analysis.

9. Kremen, Claire, Vincent Razafimahatratra, R. Philip Guillery, Jocelyn Rakotomalala, Andrew Weiss, and Jean-Solo Ratsisompatrarivo. 1999. Designing the Masoala National Park in Madagascar Based on Biological and Socioeconomic Data. Conservation Biology. 13(5): 1055-1068.

The practical applications of GIS were evident in this paper which describe the creation of a national park in Madagascar's Masoala Peninsula. Deforestation is a major threat to the unrivaled biodiversity of the island country, which has a growing population as well as a exceptionally economically-depressed populace. The authors set out to create a national park which could fit a model of "conservation with development," which would serve both to protect natural ecosystems while also providing economic opportunities for the locals which did not harm their environment. GIS was used to combine layers of socioeconomic and biological GPS data as well as SPOT satellite images classified to show forest extent. An estimate of future deforestation using metrics of elevation and slope was derived from the satellite images and elevation data. The results of this show an area of increased forest clearance close to current village centers, and this led the authors to create a "buffer zone" which could be used by the locals for timber and agriculture. A "core forest" area was selected as the main part of the park, with several unique forest fragments added as a result of species surveys, as well as areas which were not under threat of future deforestation. GIS's wide range of use is evident in this study, helping to create a sustainable and place-relevant park which can be beneficial to both humans and nature.

10. Mapedza, E., J. Wright, and R. Fawcett. 2001. An investigation of land cover change in Mafungautsi Forest, Zimbabwe, using GIS and participatory mapping. Applied Geography 23(1): 1-21.  

This study used a novel approach to integrate GIS and participatory research to investigate land-use changes, primarily deforestation. This study took place in the Zambezi River valley, near a forest reserve called Mafungautsi. Several communities are established near the edge of this reserve, and land-use practices are similar to many tropical areas of the world, with shifting cultivation, livestock grazing, and small-scale forestry. The authors interview villagers from two separate communities, asking them questions about their perceptions of the reserve, where use of the forest is not allowed, and a communal area which also contains forested land and can be used. Using aerial photos, villagers were asked to map out their perceptions of land-use changes over a 20-year time period, and their maps were integrated into a GIS through vectorization. In conjunction with this, an analysis of historical aerial photos was analyzed by the authors in the GIS to discern areal extent of forests and land-cover types. Perceptions did match reality, as villagers though that the forested area within the reserve had increased, due to stricter enforcement of the boundaries. They were also correct on the more obvious loss of forest in the communal areas, with the most rapid loss in recently deregulated former portions of the reserve. Increased populations are expected to further reduce forest extent in the communal area. With further understanding of the situation and attitudes on the ground combined with GIS analysis, as this study employed, a more thorough plan could be put into use to better protect the forest reserve.

Created by David Bucklin, last modified 13 March 2009.