The project's objectives in statistical precision expected during the design phase were met or even exceeded, as shown by survey results.
The survey could be implemented in toto (117 out of 117 sampling units), in spite of some difficulties due to non-availability of few images. In addition, all theoretical cornerstones were made operational.
The survey maintained a decentralized approach through the active participation of a large number of institutions and individuals, from both developing and industrialized countries, in all stages of the survey, i.e., design, data interpretation and analysis. This cooperative effort ensured an increase of both territorial and theoretical knowledge, essential for sound results and effective methods, and permitted widespread dissemination of monitoring principles and methodology.
The remote sensing methodology used, with its land cover classification system and interdependent procedure of satellite image interpretation, focused particularly on the consistency of multi-temporal observation, produced reliable change estimates in the form of change matrices.
The rich information contained in the change matrices led to new pioneering methods and areas of investigation such as change processes and flux analyses, that provide more in-depth perception of forest changes and reveal the “behavior” of a much varied and dynamic landscape.
The whole survey required more time than expected and, in particular, image selection and acquisition have proved to be extremely time consuming.
As standard products, which represent the processing level within the project's budget, the images were often of poor quality and consistently manifested relative geometric distortion. These constraints have been taken into careful consideration while developing the interpretation and compilation procedures which ensured high consistency and spatial co-registration between interpretation manuscripts of historical and recent images. There is no doubt, however, that good quality images, co-registered geometrically and radiometrically, would facilitate the interdependent interpretation procedure and thus enhance the reliability of the results obtained.
Owing to funding constraints and access restrictions, field verification was carried out only in 15 percent of the sample. In the future, field verification should be effected with greater intensity and should include accuracy assessment techniques for the estimation of measurement errors.
The pan-tropical sample survey demonstrates that forest and land use change information can be produced on a global basis in a cost-effective, timely and statistically sound manner. The detailed land cover classification and interdependent image interpretation approach provide consistent information on the change process and, thus, valuable insight into the nature of man-land interaction.
If continued over time, such surveys would lend factual support to global environmental research and policy making through detailed description of the processes of change, and the quantification of essential parameters, on a reliable basis.
(from FP124, p41:Conclusions)
A comparison of the characteristics and results of the FORIS database and the RS survey shows that there is strong complementarity between the two approaches as well as close correlation between their results; these factors call for the integration and continuation of both approaches. The integration of the two data sets will result in: (i) more reliable prediction models and, consequently, more detailed, homogeneous and up-to-date country level data; and (ii) more efficient stratification systems to optimize future sample rounds and expansion of RS results.
The de-centralized approach followed during all phases of the survey, from its design to its full implementation, has fostered the creation of a network of institutions and individuals from developing and developed countries. This network of experienced interpreters and advisers, coherently focused on operational forest resource monitoring, has not only contributed to the success of the completed survey but it also constitutes a very important human and institutional resource/reference for future monitoring efforts at both global and national level. The far reaching impact of this cumulative experience should not be considered a small achievement.
Recommendations for future action
In view of the information needs of the international community and, in particular, of the studies on global change, it is recommended that the building up of a consistent and reliable time series of observations of forest and land use be continued. Such a time series should include return visits to part of the sampling units included in the first round and an increment in the sample, both in space (new rounds) and in time (three dates in selected sub-samples). Keeping this in view, it is recommended that further statistical designs and analytical systems for a series of transition matrices be developed.
It is recommended that the efficiency of the survey design be improved, by:
Focusing on improving change estimates by stratifying on parameters that would reduce the variance of forest area changes, such as demography, economic indicators, infrastructure, etc.
it is recommended that in future survey rounds important improvements be made in ensuring satellite image quality and that more resources be invested in field verification and accuracy assessment. Improved spatial and spectral resolution of new and future satellites, associated with multi-temporal processing techniques of co-registration, will guarantee better image quality and, consequently, more accurate interdependent interpretation results. Field verification and accuracy assessment should be carried out more regularly with special attention to areas of complex and poorly discernible land cover1
In order to permit the widest possible use and deeper analysis of results obtained, it is recommended that the RS survey be complemented with other data collected at field level:
Socio-economic parameters related to the cause-effect mechanisms driving the deforestation process and other land cover changes.
In order to respond to the issues mentioned above it is recommended that a multi-phase survey design, aimed at providing reliable data on the state and change of area, biomass and biodiversity for forests and other land cover/uses, be developed and implemented. A three-phase survey design is recommended:
|phase I:||wall-to-wall information (NOAA AVHRR classifications, FORIS spatial and statistical databases);|
|phase II:||remote sensing sample, based on multi-date Landsat images;|
|phase III:||field inventory sub-sample data.|
Such a comprehensive approach is the only one that would allow the expansion of field information to wall-to-wall pan-tropical and global levels, with known level of reliability and in a timely, cost-effective manner. There is sufficient experience for the first two phases, that can be linked with minor effort, while the development of the field phase, providing ground truth and essential data on biomass, biodiversity and socio-economy, is at present the most challenging element in view of its operational complexity. For this third phase, a series of pilot projects in cooperation with national institutions is recommended, during which flexible and consistent field survey designs will be developed and tested.
In view of the positive response to regional workshops and in order to enhance country capacity in producing reliable national change estimates, the further dissemination of sound monitoring procedures, such as the interdependent remote sensing analysis, and the promotion of its implementation to obtain location-specific information to support policy decisions, are recommended. It is also further recommended that at country level the classification system adopted in response to local needs be maintained compatible to the standard global classification system, in order to ensure full contribution of country-level data to the global data base.
1 Provisions for these improvements are already included in the FAO Project Proposal G.5221 “Programme for Continuous Assessment of Tropical Forest Resources using High Resolution Satellite Data”.