LUIS M. MARTINEZ
SAMUEL RIVERA
MONICA JONES
GERMAN SABILLON
It is estimated that for the decade of 1980-1990 the global deforestation rate in tropical countries reached 15.4 million hectares per year. According to the United Nations Food and Agriculture Organization (FAO) only 14 out of 104 tropical countries can provide information about the location of their forested areas and just a few of them are able to give information about the current location and advance of the deforestation. Knowing where the deforestation is taking place, might allow to make a more realistic and accurate assessment of the problem and a more appropriate design of alleviating strategies (Syngh, 1994).
Deforestation is highly related to growth-population rate, so Central America being one of the fastest growing regions in the world, has a deforestation rate even higher than South America, where the large Amazon jungle is located (1.5% and 0.6% respectively), and close to the deforestation rate of South Asia, which is 1.6% (Aldobous, 1993).
Following image shows a picture of deforestation in forest broadleaf in Honduras.
Honduras is located in the middle of Central America, between 13 degrees and 16 degrees latitude North, and 83 degrees, and 89.5 degrees longitude West.
Honduras has an area of 112,100 square kilometers. Various studies estimate that 25% of this area has agricultural potential (less than 5% has been developed for high value crops) and 66% of mountainous-forest lands. Its geography is dominated by rainforest mountains and pine hills. The highlands and the mountainous region extends from East to West. Honduras topography is considered the most rugged landscape in Central America because 75% of lands are located on slopes greater than 30%. There are few valleys with most of the land suited for agriculture, dedicated to high value crops like banana, grapefruit, pineapple, mango, which are grown by international companies for exportation.
The effect of a fast growing population has overcome the little or none efforts made by the government and the dominant classes to provide the basic educational and health means for a self sustained and ordinate growth, resulting in a process of deforestation that has disrupted the ecological equilibrium in the country. (Golling,1994)
This environmental equilibrium disruption is reflected by problems in the areas of agriculture, water management, marine resources, wildlife, and health, Depletion of forest resources (80,000 hectares are lost each year) is attributed to uncontrollable fires and wasteful logging practices. The major problems in agriculture are: (1) deforestation due to shifting cultivation, (2) declining land production resulting from population pressures, and (3) heavy pesticide use. (Campanella et al 1982).
The uneven distribution of agricultural land is a constraint to sustainable development (USAID, 1989). Adding the growing number of uneducated peasant farmers that are forced to practice migratory farming using the traditional slash-and-burn cultivation systems moving to increasingly marginal and sloping lands (Silliman and Hazelwood, 1981).
An obvious example of the status of deforestation in the country is the case of El Cajon, (the box), a $750 million hydroelectric generating power dam that was built in the early 1980's being the nations largest reservoir and the source for 70 percent of the power needs. From 1985 until 1990, beside supplying that energy the reservoir created $8 million a year for electricity exports. Early in 1994, the government announced that water levels in the reservoir were down 60 meters, this reduced the facility four giant turbines practically nonfunctional, with the water level in the reservoir well below the level of the power plant s main intake channels, and just 10 feet up from the less-efficient backup channels, resulting in a dramatic decrease in power generation. More than half of the 10,000-hectare reservoir system had become a giant mud flat. The country was estimated to spend more that $60 million on imported energy, and that would not be enough to sustain the economy. According to the Interamerican Development Bank the real effects of the crisis are difficult to measure, their best estimate for the lost in industrial production was $20 million a month; most private and government offices became chattering places, with computers and machine faxes inoperative, cement mixers silent, paralyzing the construction industry (Golling, 1994).
The watershed of the reservoir is 8,192 square km, 26 sawmills had been operating in the region, and 6 of then illegally near the reservoir. Though lumber extraction is a problem, the main concern is the "Campesinos" peasant farmers which cut as much as 7.5 times more wood for fire (6 million cubic meters versus 800,000 m3 extracted by sawmills). Migratory farmers burn tropical forests, plant corn and beans in the ash-enriched soil, then when the soil is washed away because it is no longer protected, they move deeper in the forest to repeat the cycle, when the land is exhausted. (Gollin, 1994).
Picture showing effect of forest fires.
In Honduras, a national assessment of the forest cover is urgently needed in order to determine where forest resources are located and how they have been depleted. Forest resources represent the country s largest natural resource, however, one hectare of forest is lost every five minutes, no specific studies have been conducted (using an accurate geographic procedure) to support this estimation (Honduras Corporation for Forestry Development, 1992).
In other tropical countries, several studies have been successfully conducted to detect forest cover changes using remote sensing and GIS techniques. Nicaragua, Phillippines, Thailand, Peru, Bolivia, India, and Brazil are some of the countries that have already initiated monitoring projects to detect forest cover changes (Zhu and Evans, 1992; Stone et al, 1994;Schereeirer et al 1994; Hall et al, 1991; Rasch, 1994).
This study intends to accomplish the following objectives:
1. To obtain a map of deforested areas of Honduras in a 30-year period (1965-1995) by using remote sensing and GIS techniques.
2. To estimate the deforested area between 1965 and 1995 for both pine and broadleaf forests. To calculate some statistical values (mean and standard deviation).
3. To find out the relationship between the changes and to estimate substitutions of land use(and their dynamics that occurs between 1965 and 1992.
The procedure was divided into the following steps:
1. Digitizing the Map
With the ARC/INFO (Version 7.0) software, we started this step by digitizing polygon by polygon of forest lands and other land uses of Honduras. This map was produced by FAO in 1965 by using conventional (input scale: 1-500,000) photogrammetric techniques. Normally 4 tics are digitized before digitizing, however in this study we digitized 32 tics in order to have a more accurate georeference at the moment of overlapping. We cleaned the map's dangles and overshoots. After the map was cleaned we created a polygon attribute table where the different land uses were labeled and then verified. After this process we built the map topology and we transformed tic coordinates into real-world coordinates.
2.AVHRR Image
This part of the process required an image to produce a new forest map. We wanted the most current data, so we ordered a 8 mm 1995 Advance Very High Resolution Radiometer(AVHRR) tape (Entity ID: AH 14031995185321) from the EROS (Earth Resources ObservationSystems) Data Center to use as our image This image has a resolution of 1.0 square km, whichis consider a coarse resolution, however for the objectives of this study it seems acceptable. The preprocessing and processing of the image started with the downloading of this image into the Earth Resources Data Analysis System (ERDAS), specifically the software Imagine Version 8.2.
The AVHRR is a broad-band, four or five channels scanner, sensing in the visible, near-infrared, and thermal infrared portions of the electromagnetic spectrum. The AVHRR provides for global on board collection of data from all spectral channels. Each pass of the satellite provides a 2399 km wide swath. The satellite orbits the Earth 14 times each day from 833 km above its surface. The EROS Data Center provides standard AVHRR digital Level 1b and georegistered products on 9-track tapes 6250 bpi, 3480 cartridge, mm cassette, and via network access. The sensor is carried by the National Oceanic and Atmospheric Administration NOAA) satellite, in this case the image was taken by the NOAA # 14 on 03-18-1995. Usually AVHRR data is used for studying and monitoring vegetation conditions in ecosystems including forests, tundra, grasslands. Other applications include agricultural assessments, land cover mapping, producing image maps of large areas such as countries or continents and tracking regional and continental snow covers (INTERNET AVHRR information).
COMPOSITE IMAGE
Because theAVHRR image (1995) had around 30% cloud coverage, we decided for our second approach to use a composite image. A composite image is made up from raw data taken from ten consecutive days of satellite images, were the best pixels over that time period are chosen by using some computer programs. We felt that the composite image was our best chance of getting a cloud free coverage, because the image is derived from the best raw data over a ten day period of time. This image was obtained from a recently started project. The United Nations Food and Agriculture Organization s Forest Resources Assessment 1990 Project requires 1-km AVHRR data for all the forested lands on the planet, especially the tropical countries, for their global forest mandate. For this reason the production of global 10-day composites have been undertaken by several world agencies in order to meet the needs of the international science community. The prototype 10-day composite used in this study was produced for the period of April 21-30, 1992.
This project is considering to make composite of 30 days in the future (Eindenshink and Faundeen, 1995).
With this new image we created a two band Normalized Difference Vegetation Index(NDVI), so that we could find the range for the different vegetation values in the image. The values that represent vegetation in a image range between 100 and 200, 100 is low vegetation density(least green) and 200 represents high vegetation density(the greenest areas). To find the range value of the broadleaf and coniferous forest, a statistical analysis was performed to find the mean, mode, and standard deviation of each forest type and the other land uses.
Next image shows the NDVI image from Honduras
The process that was used to gather the statistical information was achieved by randomly picking ten polygons from each forest types. Each polygon was made into a subset, saved as a file, and then statistically analyzed. After the statistics from the ten polygons were found, the mean, mode, and standard deviation was calculated for that individual forest type. With this statistical information the range of the vegetation value could be calculated, so that the whole vegetation ranges from the image was calculated. However, we found that there were overlaps among the found ranges, so vegetation patterns were extremely difficult to differentiate. Therefore, we decided to use the five-bands image instead.
After that, we made a subset of the original image in order to obtain only the area that covers the country. Having the image ready to be classified, We ran an unsupervised training (computer automated method) to identify the vegetation pattern of the five bands of the AVHRR image. In this process we used thirty-classes classifications to identify the different vegetation and landuses of the country. The thirty classes were manually grouped into three classes by using the digitized 1965's forest map and personal field experience.
These three classes were identified after grouping the original (computer generated) 30 classes:
(C) Coniferous Forest: including the seven species of pines that are found in Honduras.
(B) Broadleaf Forest: which includes: tropical rain forests, high elevation cloud forests, dry forest (found in the southern part of the country) and mangrove (coastal) forests.
(O) Other land uses: includes agricultural lands, grasslands, urban areas, etc., (all the other land uses).
These categories were chosen because they match exactly with the 1965 forest map categories which makes it much easier the comparison. Raster system was used in this part of the process. The resulting map was georeferenced by using a 1-1,000,000-scale topographic map of Honduras (procedure proposed by Peters and Reed, 1992).
Maps Overlapping
Polygon overlapping is a spatial operation that overlays one polygon coverage on another to create a new polygon coverage (ESRI, 1992). Before overlapping the 1965 forest map and the 1992 classified-image, the 1992 image had to be reformatted in vector system. Using Imagine scommands the image was exported to ARC where it was transformed to vector format. After this process both maps were overlapped in ARC/INFO using the intersection analysis. Arc's intersection in ARC/INFO is a spatial join that merges feature attributes based on the overlaying of points, lines, and polygons but keep only those portions of the input coverage features falling within the overlay coverage features (ESRI,1992).
To determine the amount of change for each type, we created a report to describe the results of the analysis. We ran the command statistical report on the intersected coverage. We also made two new coverages showing how the pine forest was substituted by other land uses and how the broadleaf forest was affected by the same phenomena by making xv's files they were exported to mosaic and ARC/VIEW.
RESULTS
From the Honduras '65 map we obtained the statistics from all the polygons. The next display is the final map and the table with the statistical information.
The following table shows statistical information about 1965 vegetation map:
Next image is the vegetation map from 1992 obtained from AVHRR image:
Statistical values from this map is following displayed:
The next image shows the changes in pine forest area from 1965-1992:
Areas in color red, show the difference in Pine area from the map 1965 and 1995. This area represent the deforestation in Pine forest.
Follow image is similar than the above, but now refered to deforestation in broadleaf. Same red area represent deforestation in Broadleaf forest.
The following diagram shows how each type of vegetation changed during this period.
This graph shows the deforestated areas from both vegetation types during this period of analysis.
CONCLUSIONS
About the methodology:
1. The use of NDVI to determine vegetation patterns in the tropical areas demonstrated low accuracy, because the different types of vegetation analyzed in this study were difficult to identify while using this index.
2. For the purpose of this study, the AVHRR data was able to provide the required information in detail, and was more accurate than the 1965 map.
3. The unsupervised classification used in this analysis provided accurate information about the Honduran forest. However, we found some polygons that did not correspond with the actual vegetation type, so we manually classified some of the polygons from personal field experience. The area this problem occured was in the Honduran Pacific Coasts, where the mangrove forest were identified as pine forest.
Data obtained from the Maps:
1. We found that 9,295 square km of pine forest, and 13,941 square km of broadleaf forest were deforested during this 28 year period (1965 - 1992).
2. The deforestation rate was 332 square km/year (33,200 has/year) for pine forest, and 498 square km/year (49,800 has/year) for broadleaf forest. This data shows that the broadleaf forest is being more rapidly deforested than the pine forest.
3. Forest pine areas were substituted by other land uses by 47% (17,509 square km or 1,750,900 has), and 40% (16,475 square km or 1,647,500 has) of broadleaf forest were also converted to other land uses. Thus, reducing both pine and broadleaf forest to almost half of their prior area.
4. The large areas of forest areas in 1965 were drastically fragmented in the 28 year period. In 1965 there were 69 large polygons of pine forest areas. However, in 1992 there are 2,686 polygons fragmented into smaller pine areas, and for the broadleaf forest the area changed from 73 polygons in 1965 to 2537 polygons in 1992.
REFERENCES
Aldobous, P., 1993. Tropical Deforestation, Not Just a Problem in Amazonia. Science, Pp.1390-1393.
Campanella, P. et al. 1982. Honduras Country Environmental Profile: A Field Study. Honduras. JRB Associates, Inc. - USAID.
Chuvieco, E., and M. P. Martin, 1994. Global Fire Mapping and Fire Danger Estimation Using AVHRR Images. Photogrammetric Engineering and Remote Sensing, Vol. 60, No. 5, Pp.563-570.
Congalton, R. G., K. Green, J. Teply, 1993. Mapping Old Growth Forest on National Forest and Park Lands in The Pacific Northwest from Remotely Sensed Data. Photogrammetric Engineering & Remote Sensing, Vol. 59, No. 4, Pp. 529-535.
ESRI, 1992. Understanding GIS, The Arc/Info Method. Environmental Research Institute,Inc. Redlands, CA.
Foody, G. M., 1994. Ordinal Level Classification of Sub-Pixel Tropical Forest Cover.Photogrammetric Engineering & Remote Sensing. Vol. 60, No. 1, Pp. 61-65.
Golling, J.D. 1994. Trees Down, Lights Out in Honduras. Habitat. The Christian Science Monitor. Nov. 15, 1994.
Hall, F. G., D. B. Botkin, D. E. Strebel, K. D. Woods, and S. J.Goetz, 1991. Large-Scale Patterns of Forest Succession As Determined by Remote Sensing. Ecology, Vol.72, No.2, Pp.628-640.
Honduran Corporation For Forestry Development, 1992. Annual Statistics. COHDEFOR Tegucigalpa M.D.C., Honduras, Central America.
Iverson, L. R., R. L. Graham, and E. A. Cook, 1989. Applications of Satellite Remote Sensing to Forested Ecosystems. Landscape Ecology, Vol. 3, No. 2, Pp. 131-143.
Maclean, A. L., D. D. Reed, G. D. Mroz, G. W. Lyon, and T. Edison, 1992. Using GIS to Estimate Forest resource Changes, A Case Study in Northern Michigan. Journal of Forestry, No. 12, Pp. 22-25.
Moody, A., and C.E. Woodcock, 1994. Scale-Dependent Errors in the Estimation of Land Cover Proportion: Implication for Global Land-Cover Data Sets. Photogrammetric Engineering & Remote Sensing, Vol. 60, No. 5, Pp. 585-594.
Myers, N., 1988. Tropical Deforestation and Remote Sensing. Forest Ecology and Management, Vol., 23, Pp. 215-225.
Peters, A. J., and B.C. Reed, 1992. A Technique for Processing NOAA-AVHRR Data into a Geographic Referenced Image Map. Photogrammetric Engineering & Remote Sensing, Vol. 58,No. 5, Pp. 569-571.
Ripple, W. J., 1994. Determing Coniferous Forest Cover and Forest Fragmentation with NOAA-9 Advance Very High Resolution Radiometer Data. Photogrammetric Engineering & Remote Sensing. Vol. 60, No. 5, Pp. 533-540.
Schereier, H., S. Brown, M. Schmidt, P. Shah, B. Shrestha, G. Nakarmi, K. Subba, S. Wymann, 1994. Gaining Forest But Losing Ground: A GIS Evaluation in a Himalayan Watershed. Environmental Management, Vol. 18, No. 1, Pp. 139-150.
Silliman, J.R. and P.T. Hazelwood. 1981. Draft Environmental Profile of Honduras. University of Arizona. Office of Arid Lands Studies-USAID. Honduras.
Stone, T. A., P. Scheesinger, R. A. Houghton, and G. M. Woodwell, 1994. A Map of the Vegetation of South America Based on Satellite Imagery. Photogrammetric Engineering & Remote Sensing, Col. 60, No. 5, Pp. 541-551.
Syngh, K. D., 1994. Tropical Forest Resources; An Analysis of the 1990 Assessment. Journal of Forestry, No. 2, Pp. 27-31.
Zhu, Z., and D. L. Evans, 1992. Mapping Midsouth Forest Distributions, AVHRR Satellite Data and GIS help meet RPA Mandate. Journal of Forestry. No. 12, Pp. 27-30.