The FRA 2000 continues the use of satellite image assessment for estimating rates of forest cover and change at the regional level. However, resources do not provide for extension of this approach to the country level, nor has FAO been given the mandate to conduct this level of work. Since estimates at the country level are still needed, such estimates will have to be based on other types of data. Sources of information available for completing FRA 2000 include the following:
The strategy for the Tropical 2000 Update approach is to revisit the topical FRA 1990 sampling units, and update them with a third date of TM imagery for the reference year 2000. Such a survey will provide up-to-date information on change processes, as well as new and unprecedented insights on deforestation trends over the periods 1980-1990 and 1990-2000.
The key strength of the TM change assessment approach is that it utilises advanced satellite technology in a way that provides consistent, reliable results. This approach also helps reduce program costs by allowing use of the photographic image archive already maintained by FAO. Although the method of manual interpretation does not take full advantage of current image processing technology, the method is simple to replicate, relatively inexpensive, and is readily familiar to foresters all over the world. The primary weakness of the TM change assessment as implemented by FRA is that, due to financial and time constraints, the sample intensity is designed for regional results only; few countries will have sufficient samples to support estimates at the country level. Thus, some subtle forms of forest degradation or deforestation may be missed in the stratification process, which may lead to false estimates of rates of forest loss.
FRA 2000 will feature a global forest cover map produced at 1-km spatial resolution, produced at the U.S. Geological Survey (USGS) EROS Data Centre (EDC) in cooperation with FAO. This map is a complete global coverage produced with a consistent methodology and a flexible database philosophy, using AVHRR Normalized Difference Vegetation Index (NDVI) as the primary input data. Two levels of land cover are available from the database: the full classification based on seasonal definitions and the aggregated 17-category International Geosphere and Biosphere Programme (IGBP) classification. The FAO forest cover mapping effort builds on the full USGS seasonal database, refining forest classes to reflect forest density classification.
The primary advantage of AVHRR imagery is that the imagery is acquired for all lands (wall to wall) on a daily basis. This presents a low cost alternative to TM imagery, and the frequent availability makes it more likely that cloud free imagery will be available for any given target area close to any given target date. The primary weakness of AVHRR imagery is the lower resolution (1 km pixel size vs. 30m for TM), which leads to much coarser interpretability and less precision than interpretation TM imagery. The wider angle and interference of atmospheric effects cause problems with interpretation near to the edge of images. Thus, while AVHRR may be very suited for change assessment in large (e.g. regional) areas, it may not be so useful for assessment at the national level, particularly for small or medium sized land areas
Many countries have in place systems for monitoring their forests on a periodic or occasional basis. Other countries have conducted studies of subsets of their forests, or complete or partial forest inventories sponsored by a variety of donor organisations. The result is a huge set of information regarding status and trends in forest statistics at the national and sub national level. FRA 2000 professional staff have been working for years to make and keep current a metadata base describing what data exist, by country, throughout the tropical world, as well as archiving the actual data in the FORIS (Forest Resources Information System) database. Regional specialists travel to the countries in their region, gleaning available information and assembling it for potential incorporation into the FRA 2000 and future studies. FRA will make these data available, along with appropriate metadata, via the World Wide Web.
The strength of using data reported by countries is that results tend to be of higher precision than results from studies conducted at the regional scale. National forest statistics would be harmonised with other reports and analyses produced by and used in the target country, increasing consistency between FAO and National reports. Presumably, forest statistics gathered at the National level would be overseen by experts with more familiarity with the target country and with more time to spend on a single country, which should lead to improved data quality. The primary weakness with the aggregation of country level is that there is tremendous inconsistency and incompleteness in country inventory data, both within countries over time as well as between countries. Assessments are conducted with different objectives and populations of interest, using different definitions, standards, sample designs, and field methods from those used by FAO or by neighbouring countries. The results can be difficult to interpret and harmonise for comparison across countries, as FRA must do. Additionally, many countries particularly in Africa simply lack the resources to conduct forest assessments, so that data are lacking.
The strength of the modelling approach used already in FRA1990 is that it provides a platform for generating estimates for all countries regardless of availability of country level data. Data gaps would simply be modelled by more generalised functions. Generation of models is a useful exercise in that it forces study of cause and effect relationships and can lead to insights regarding true mechanisms or drivers of forest cover change, which in turn could lead to better investment decisions by national governments seeking to manage their forests. The results of existing models are probably reasonable at the global level. However, the greatest weakness of the modelling approach continues to be inadequacy of existing models when applied to specific countries. The process of deforestation is such a complex process, involving physical, climatic, political, and socio-economic forces which are themselves very complex, that simple generalised models of forest change have so far not been developed. Current models are oversimplified and yield similar predictions of forest cover change rates for countries which are known to be very different. More complex models are yet to be developed and tested.