Land Use, Land-Use Change and Forestry

Other reports in this collection Determining Rates, Extent, and Locations of Forest Clearing and Regrowth

Estimating sources and sinks of carbon from forest clearing and regrowth using remote-sensing requires repeated measurements of forest clearing over large areas, fine spatial and temporal scale analyses to document regrowth, and in situ data on carbon pool changes that are associated with changes in land cover.

Coarse-resolution optical data (>1 km) are useful in describing broad distributions of different types of land cover, including different forest types (Belward et al., 1999). These coarse-resolution data generally are inadequate, however, for accurately quantifying changes in forest cover, such as clearing and regrowth. Justice and Townshend (1988) demonstrated a need for spatial resolution of optical data that is less than 250 m.

High-resolution data from the Landsat, SPOT, and similar series of Earth observation satellites have been employed to make regular regional measurements of forest clearing and regrowth (Skole and Tucker, 1993; INPE, 1999). Large amounts of these data exist in several national archives, dating back approximately 20 years. Thus, a continuous and consistent source of data is available from which a high-resolution, fine-scale (1:250,000 scale mapping) information system could be developed. Many countries routinely perform regular assessments of forest clearing and regrowth over large areas-in particular, Brazil, Thailand, and Indonesia (INPE, 1999). Accuracies that approach 15-25 percent of the area cleared have been demonstrated (Houghton et al., 2000).

Because cleared areas may revegetate rapidly, they must be observed frequently-as close to annually as possible. Frequent measurements make it easier to co-register areas of clearing and regrowth and reduce the probability of missing clearing on patches that begin to regrow quickly. Frequent observations will help in attributing changes to specific activities.

Although traditional optical remote-sensing techniques can distinguish between cultivated areas, pastures, and secondary growth, they are limited in terms of mapping various stages of fallow and secondary forests (Sader et al., 1989). On the other hand, synthetic aperture radars (SARs) operate independent of solar illumination, cloud cover, and smoke and can detect differences in forest structure and woody biomass associated with various stages of forest clearing and regrowth (Rignot et al., 1997; Saatchi et al., 1997). Use of both optical and SAR data can provide a better characterization of land cover. For example, a plot of newly cleared and partially burned tropical forest may contain a significant portion of dead woody debris (slash). Visible and near-infrared reflectance data will show that this area has been cleared but will provide little insight into the presence of the slash. The SAR data, however, would indicate a significant amount of biomass; as a result, this area could be confused with a secondary growth area with comparable radar cross-section. Using the optical and SAR data together would reveal that the site was deforested and not in secondary growth, yet still had a significant amount of residual woody biomass (Rignot et al., 1997).

The difficulty with these high-resolution satellites is not that they fail to identify cleared areas or areas where trees are returning but that they may fail to distinguish such clearing and regrowth from other changes, such as harvests, natural disturbances (fire, insects, storms), or other changes that are unrelated to human activity.

Very high-resolution data (1-m panchromatic and 3-m multi-spectral) that are now available from the commercial IKONOS II satellite may be useful for determining the actual activities on the ground that have led to forest clearing. Although such data can detect very small clearings, the scientific community as yet has very little experience with these data.

To obtain annual estimates of forest clearing and regrowth, a stratified sampling scheme might be employed to determine deforestation rates between the complete inventory/census years, spaced 3-5 years apart. The stratification might be based on the last complete inventory/census, assuming that deforestation is spatially persistent over intervals of 3-5 years. Research with Landsat- or SPOT-scale spatial resolution in some areas of the tropics suggests that a sample of 30 percent or less of the total forest area would be sufficient, but further research is necessary to determine sampling densities for other regions.

The costs of using remote-sensing data vary greatly on a per-hectare basis. AVHRR, other coarse-resolution optical sensors, and coarse-resolution radar sensors typically have very low costs per hectare for access to data. The costs of acquiring Landsat data vary according to the year in which the data were originally acquired, as well as whether the Landsat system was under private or public management. The most recent data for Landsat-7 are also the least expensive (~US$600 per Level 1 scene) because the system is now operated as a public resource. SPOT data are somewhat more expensive than Landsat and have finer spatial resolution but do not have global coverage. Very high spatial resolution data are only publicly available from IKONOS II, at somewhat higher costs per hectare, but practical considerations regarding data volume are likely to inhibit their use for broad-area surveys.

Once the spatial extent and rates of forest clearing and regrowth have been quantified, accurate calculations of changes in carbon stocks require techniques for measuring pre- and post-disturbance biomass, rates of secondary growth formation and turnover, and rates of biomass accumulation in secondary growth. There have been several attempts to use SAR data to determine aboveground biomass directly, through the known sensitivity of the radar backscatter to total aboveground material, its structure, and its dielectric properties. With current techniques and wavelengths, however, there is little ability to discriminate biomass levels greater than 50-100 t ha-1. The National Aeronautics and Space Administration's (NASA) planned Vegetation Canopy Lidar (VCL) mission-to be launched in 2001-is designed to provide data sets of the vertical distribution of vegetation canopies. These data may be a good proxy for aboveground biomass for many forested ecosystems, but full knowledge of the utility of the measurements will necessarily await the instrument's launch and subsequent experience in the scientific community.

Other reports in this collection