The most straightforward way to intensify Remote Sensing based approach is to maximise the coverage of TM figures to cover up to 100% of the tropical belt. This, however, would require plenty of resources for purchasing the data and staff for working on it using present visual interpretation approach.
There are other potential useful sources of remote sensing imagery which may be incorporated into the final analysis. In 1999 FRA commissioned a global forest cover map to be produced by the Eros Data Center (EDC) of the US Geologic Survey (USGS). This map, based on AVHRR coverage, provides a global snapshot of forest cover at a relatively coarse scale. There are also other AVHRR –based forest cover mapping projects ongoing (see for instance http://fellini.gvm.sai.jrc.it/trees/ or http://www.geog.umd.edu/tropical/main.html.
The primary advantage of increased use of satellite imagery is that it provides a tool for objectively filling in existing data gaps by use of an information source which is universally available. Linking TM and AVHRR data - using multi-stage or multi-phase sampling or ratio estimates - would be objective and simple to implement once the data and methods were available. The primary weakness of this approach is the lack of existing strong relationships between TM and AVHRR. Significant amounts of research have sought such correlations, with mixed results. Correlations may not be universally strong enough to act as a reliable source for filling in data gaps.
Given the lack of globally consistent forest inventory data, some kind of synthesis based on partial data plus expert opinion is not an unreasonable way to arrive at answers. Certainly expert opinion can be subject to bias and imprecision; yet it will generally be better than no information at all. Two of these methods are the Delphi Technique and Convergence of Evidence.
The Delphi Technique applies elementary statistics to a sample of expert opinions in an iterative fashion, with the intention of eliminating outliers, providing feedback, and converging to a level of consensus. The technique was originally developed for making forecasts of socio-economic trends and has since been generalized for use in a variety of estimation or problem solving situations (Sackman 1975).
The original Delphi Technique had several distinguishing characteristics. A pool of experts was chosen to make predictions on some topic. However, rather than assembling the group together, each group member was asked to submit their estimates anonymously, to avoid the chance that more prestigious or dominant group members would unduly influence the estimates made by other group members. The set of estimates was then averaged, and individual estimates which were outside some range of standard error were returned to the expert who had submitted it along with a request for justification or for a new estimate. The process would iterate several times until all there were no further changes in individual estimates, at which time a consensus estimated would be adopted.
Advantages of the Delphi Technique are that it will always provide an answer in the time available, and that it takes advantage of expertise and experience which may be relevant but non quantitative. Disadvantages are that the method may lack repeatability, and that the quality of the answers depends heavily on the quality of the experts. In our case, it may be difficult to find a qualified pool of experts for each of the many countries for which we wish to derive estimates.
A second qualitative technique called Convergence of Evidence is currently used by the Foreign Agriculture Service of the US to produce annual assessments of agriculture crops in major producing or consuming countries of the world. Whereas the Delphi Technique relies on a group of experts arriving at consensus based on their (somewhat different) experience, Convergence of Evidence instead relies on a single expert who has at their disposal a variety of independent or quasi-independent information sources regarding the parameter of interest (Roke 1999, www.fas.usda.gov/pecad/who_what.htm).
For example, for estimating future wheat harvest in country X, an analyst has at their disposal historical records of past crop yields, recent weather data affecting the current year’s crop, short term future weather forecasts, historical land ownership and farming data, agricultural statistics and forecasts produced by country X, and satellite data at different points over the current growing season showing potential areas and biomass accumulation of the target crop. The analyst synthesizes this information internally, using their experience and knowledge for assigning more weight to information that they feel is more reliable and arriving at a final answer. Ideally, over time, the expert has a chance to validate their estimate against a future measurement, allowing for learning over time.
This technique most closely resembles the methods used in the 1980 Assessment, where a small group of individuals immersed themselves in various available data sources. This trend continues to some extent in the current FRA 2000 where a small group of regional analysts are compiling, interpreting, and preparing to analyse regional and country level data. Clearly, the quality of the estimates for each country relies heavily upon the expertise and experience of the analyst.
One advantage of this method is flexibility in that estimates can be derived when the data sources vary in availability and quality from country to country. Like the Delphi technique, an estimate can be generated for every country and can be completed within a finite time window, recognizing that lack of time can limit the depth of the analysis. Disadvantages of this approach are dependence upon the skill of a single expert which can readily lead to bias, imprecision, and lack of repeatability.
