Over the last two centuries, the concentration of carbon dioxide in the atmosphere has increased by more than 25%, from about 275 ppm in the eighteenth century to more than 350 ppm in 1989. Most of the increase is attributed to the combustion of fossil fuels, but up to a third is thought to have come from deforestation. Approximately 90% of the current annual net release of 1.8 x 10¹⁵ g of carbon is due to deforestation is from the tropics. Ten countries comprise two-thirds of the net release (Brazil, Columbia, Indonesia, Ivory Coast, Laos, Malaysia, Mexico, Peru, Thailand, and Zaire). At 0.6 x 10 ¹⁵ g C per year, the net flux from Brazil was the largest single biotic source of biogenic carbon in 1988.

The amount of carbon held in terrestrial ecosystems is changed as a result of 1) direct human effects of land use, such as deforestation, and 2) indirect human effects on ecosystems, such as increased concentrations of atmospheric CO₂ or climate change, which influence changes in ecosystem metabolism. At the present time, the net flux of carbon between the biota and atmosphere due to land use change (deforestation) can be calculated with more confidence than changes in carbon storage associated with large-scale changes in ecosystem metabolism. This is because of the large difference in biomass between forests and the agricultural systems which replace them, and because deforestation and reforestation can be readily documented and quantified, particularly if remote sensing data are utilized.

Changes in the stocks of carbon due to land use change cannot be measured directly for the Earth as a whole, or for an area the size of the Amazon Basin. Instead, the net flux of carbon must be modeled. Such an approach utilizes three types of data: 1) rates and geographic distribution of deforestation, 2) the fate of the deforested lands, and 3) changes in the stocks of carbon in biomass and soils as a result of disturbance and recovery over time.

Models developed with improved geographic and temporal data on deforestation rates, better parameterization of the dynamic nature of deforestation and reforestation, and improved data on above- and below-ground carbon response characteristics are needed.

Satellite remote sensing is the only means for resolving discrepancies or quantifying temporal and spatial variations in deforestation rates. There are no reasons why satellite-based techniques cannot be applied to a large area like the Amazon Basin, or a significant portion of the tropical forest belt, to resolve the aforementioned controversies and uncertainty, and thereby provide vastly improved forcing functions for global carbon models.

Project Objectives

The NASA Landsat Pathfinder Humid Tropical Forest (HTF) Project is a collaborative effort between the University of New Hampshire's Institute for the Study of Earth, Oceans and Space, University of Maryland's Geography Department, and NASA's Goddard Space Flight Center. The project has three major objectives:

1. Utilize Landsat data to map deforestation in closed tropical forests at sub-kilometer resolution for a base period (e.g., 1986);

2. Utilize Landsat data to quantify and map the rate of deforestation by mapping the change from other time periods (e.g., 1975, 1992); and

3. Create a Landsat data set and science products for distribution in a digital geographic information system (GIS) format, and de velop an information management system (IMS) to manage data orders, archiving and processing and distribution of products.

Through these objectives, the project aims at improving significantly the most important source of uncertainty in our understanding of the role of biota in the global carbon cycle. At the same time, the project will contribute significantly to improving the database for several international policy initiatives including the Framework Convention on Climate Change, the activities of the Intergovernmental Panel on Climate Change, national emission inventories, and many others which focus on the role of tropical forests. It is hoped that the results of this project will simultaneously fulfill the needs of the global change research community, the international policy community, and national-level forest resources and economic development programs. The project will focus on the three regions where most of the tropical deforestation in the world has occurred: (1) the Amazon Basin, (2) Central Africa, and (3) Southeast Asia. Mapping deforestation in these three regions will account for the majority of deforestation activities in closed tropical forests worldwide and will account for approximately 75-80% of the current net biotic flux of carbon.

As part of NASA's Pathfinder Program the project has two goals:

1. To utilize large amounts of existing satellite technologies and data in new ways to address important global change research questions in advance of, and leading toward, the launch of the Earth Observing System (EOS).

2. To develop a foundation of experience for managing large amounts of satellite data for global change research prior to the launch of the Earth Observing System, thereby testing and proving the technologies and approaches for information management which will be needed by the community at large with the launch of the Earth Observing System.

The approach to Pathfinder is straightforward. The first step is the identification and acquisition of a pan-tropical, wall-to-wall Landsat digital data set of over 2500 Multi Spectral Scanner (MSS) and Thematic Mapper (TM) scenes from the EROS Data Center (EDC) archive and the archives of the foreign ground receiving stations with coverage of the study areas. A three-date data set has been selected based on data availability. The three dates, or epochs, being used are early-1970s ( i. e., 1972-1974), mid-1980s (i. e., 1984 -1986), and early 1990s (i. e., 1989-1994). The exception to this plan is the use of 1978 data for the Amazon. This parsing of data analysis over three year epochs allows a much wider selection of low-cloud data from the archive. Landsat MSS data is utilized for the two earliest epochs. The last epoch utilizes Landsat TM data. Once acquired, the digital data is then analyzed to create a science product data set: a digital map database, in a geographic information system, of the rate and extent of deforestation.

Digital Data Processing

Research and development activities for this project and a prototype project for the International Space Year suggest that the use of digital image processing in conjunction with editing, georeferencing and spatial analysis in a Geographic Information System are effective means for quantifying deforestation. Findings also indicate that the use of high resolution Landsat data may in fact yield much better precision than AVHRR-based analyses. The use of digital pre-processing with visual post-processing greatly reduces analysis time over that of hand digitization of a photographic product, and greatly reduces the confusion of classes associated with purely digital processing techniques.

After EDC pre-processes the imagery into a uniform, project-specific format, the digital data are first classified using traditional techniques based on unsupervised clustering and knowledge-based assignment of clusters to land cover classes. The output classes of interest are forest, deforested areas, secondary growth, non-forest vegetation, cloud, cloud shadow, and water. The secondary growth class represents areas that have been deforested and then abandoned and are regrowing (accumulating carbon). The digital classification is then converted to polygons and then plotted on clear vellum at 1:250,000 scale and compared with the 1:250,000 scale color composite prints provided by EDC. Digitizers check the label on each and every polygon for accuracy, make any necessary changes and add any missing polygons. This process is iterative, with quality assurance checking until the coverage is accepted for archiving. Quality control is carried out by trained supervisors with GIS training, forestry backgrounds and field experience. The accepted final coverages are stitched together scene by scene to build regional coverages in an edge-matching process. The edge-matching is performed to insure thematic and positional consistency within the overlap areas of the individual Landsat scenes. The final product is a seamless database.

Validation and Accuracy Assessment

To obtain an estimate of the accuracy of the final analysis, the project has developed a field-based accuracy assessment program. Objectives are to quantify the thematic and positional variance. This is done at three levels of analysis:

1. Experts in each region are consulted and insights are gained from their extensive knowledge of local conditions. To facilitate a close working relationship with experts in the countries and regions under study, a visiting scientists program provides support to colleagues from tropical countries to spend time in residence at the University of New Hampshire and University of Maryland. Visiting scientists spend anywhere from 2 to 12 months in residence.

2. Preliminary and cursory field excursions to various areas are conducted to get a good sense of on-the-ground conditions and to establish initial classification rules and procedures.

3. Systematic field validation exercises are conducted, where points on the field are selected and measurements are made using a Global Positioning System. The results of these field exercises are used to develop a statistical accuracy assessment using standard methods of presentation in contingency tables. In these field exercises, two aspects of accuracy are tested. The first is thematic, assessing errors of omission and commission in classification of the images. The second, using the GPS and obvious features, assesses the geometric and positional accuracy of the image registration.

The project has established approximately 22 field test sites throughout the tropics. At each test site, as much ancillary data as possible is collected, as well as other sources of remote sensing data, including Spot 20 m multispectral data and JERS-1 ERS-1 SAR data. Field measurements are conducted at each test site. The test sites are also used for inter-laboratory comparisons, which helps with assessments of the variability between analyses conducted in the two different laboratories, and assessments of the variation in interpretation and classification within the laboratory. The latter is accomplished through repeated classification trials and comparisons with classifications performed using other data sources, such as Spot. Because the project also uses historical MSS data for the first two epochs, historical aerial photos are used to obtain reference data for the accuracy assessment. Some preliminary results from an accuracy assessment of the Amazon analysis revealed an overall accuracy of 89%.

Some of these field test sites are also used for examining the land cover change dynamics between the nearly decadal assessments for each epoch. Annual Landsat and Spot observations are analyzed to look at annual rates of clearing and abandonment in an attempt to understand the land use practices driving the deforestation. These sites are spread out across the tropics in an attempt to characterize many of the more wide spread land use/land cover change dynamics associated with tropical deforestation.

Results from the Humid Tropical Forest (HTF) Project to date indicate that in the Amazon in 1986 there was 245,415 km² of deforestation, and of the deforested area, 72,305 km² were in secondary growth. This suggests an annual rate of deforestation from 1978 to 1986 of 18,000 km² The secondary growth numbers highlight the fact that almost 30% of the disturbed forest area is in some form of secondary growth. Figures 18.1 and 18.2 present the extent of the 1986 deforestation and secondary growth, respectively. The full resolution data has been gridded to 16 km cells and shaded according to the percentage of each 16 km cell that has been deforested or is in secondary growth.

All of the Landsat data used by the HTF project is available from the Land Processes Distributed Active Archive Center (DAAC) at the EROS Data Center. Information on how to obtain data from the DAAC can be found at the EDC homepage (http://sun1.cr.usgs.gov /landdaac/landdaac.html). The derived products will initially be available from the Landsat Pathfinder HTF project (http://pathfinder-www.sr.unh.edu/pathfinder) and eventually from the DAAC at EDC. Preliminary data on the extent of forest in several countries in Southeast Asia is shown in Table 18.1.