(by Peter Holmgren)
The Global Forest Resources Assessment 2000 (FRA 2000) included a remote sensing survey of tropical forest area changes. The survey studied the period 1980-2000 using a time series of three points. Time and resources available made it necessary to sample the studied area. Considerable effort was made to ensure that the sample was representative for the tropics as a whole, yet affordable for the survey project (FAO 2001).
Predictably, the findings of FRA 2000 were given high attention in media, and widely commented by organisations active in the forestry field. This paper is a response to criticism raised by Tucker & Townsend (2000) who challenged the statistical design of the FRA 2000 remote sensing survey and claimed that a 10% area sample is not sufficient to capture the spatial variation of forest cover changes and that an 80% area sample would be required, using a much larger number of satellite images. Since this figure was later quoted by Matthews (2001) and Stokstad (2001) and thus given wide publicity, it is relevant to present the current analysis that addresses some underlying assumptions and findings in the Tucker & Townsend (2000) study.
The paper has been prepared by Ray Czaplewski who was involved in the design of the FRA 2000 remote sensing survey.
