Geoinformation, monitoring and assessment Environment

Posted March 1996

The role of remote sensing in FAO's Global Forest Resources Assessment and Monitoring Programme


M. Bied-Charreton
Chief, Environment and Natural Resources Service (SDRN)
FAO Research, Extension and Training Division

P.G. Reichert
Remote Sensing Officer, SDRN

K. Janz,
Senior Forestry Officer
FAO Forest Resources Division


The need for assessment and systematic observations of forest cover and to establish and/or strengthen the necessary capacities has been strongly expressed in UNCED's Agenda 21, par. 11.29:

"...assessment and systematic observations are essential components of long-term planning ... and ... in many cases, even the basic information related to the area and type of lacking".

In keeping with its mandate, which includes the collection, analysis and dissemination of information on renewable natural resources, FAO has executed a series of forest resources assessment projects since 1946. Satellite remote sensing data were used for the first time for the tropical forest resources assessment project 1980 when vegetation maps were prepared using Landsat MSS data for Cameroon, Togo and Benin. The last forest resources assessment was undertaken for the reference year 1990 and was global in scope, i.e. it included the forest resources of non-tropical and industrialized countries as well as former USSR. Annual forest cover change rates have also been published by region, except for data from former USSR, where only data of Belarus and Ukraine were available.

Results published in 1995 show an annual rate of global deforestation (excluding former USSR, except Belarus and Ukraine) of about 10 million ha for forests and wooded land, equivalent to 0.2% per year. These results were obtained through compilation of existing reliable statistics on area and volume and adjustment to a common concept, classification and reference date.

Simultaneously another approach was used which assessed the forest cover in 1990 and changes since 1980 using sample high resolution satellite data. A report with the results of this assessment will be published in about April 1996 (FAO Forestry Paper 130: Survey of Tropical Forest Cover and Study of Change Processes based on Multi-date High Resolution Satellite Data).

This paper focuses on the remote sensing approach which was used for the tropical part of the "Forest Resources Assessment 1990" (FRA 1990). This, however, does not exclude, that a major part of the country statistics (first approach) has been obtained by analysing remote sensing data.

The remote sensing approach of the Forest Resources Assessment 1990 Project

Although desirable, wall-to-wall coverage of the tropical belt with high resolution satellite data was not possible for FRA 1990 due to cost and time constraints and also due to the high cloud cover risks which make the compilation of a time-wise consistent data set difficult. FRA 1990 has therefore chosen a stratified sampling design to select 117 samples, each equivalent to one Landsat frame in size. The stratification was based on a subdivision of the tropical belt into sub-regions, which were further sub-divided into forest cover strata. The samples were then selected proportional to the estimated deforestation. The samples cover the whole range of woody vegetation from the rainforests to the semi-arid shrub and tree savannas.

The 117 samples, each covering approximately 3.4 million ha, lead to a sampling intensity of about 10% (approximately 1,200 Landsat frames are required to cover the tropical belt). For each of these samples, two Landsat false colour composites at 1:250,000 were selected, one as close to 1990 as possible, the second one close to 1980, and visually analysed to assess the present state of the forests as well as the changes they have undergone since 1980.

This approach allows for the estimation of areas of forest cover and its rate of change with a high level of reliability together with the associated error at pan-tropical and regional scales. An inter-dependent approach was used for the simultaneous interpretation of the multi-temporal images to ensure highest possible accuracy of the analysis. This inter-dependent approach implies the visual interpretation of the two images of the same area acquired in the different years by the same interpreter, who is normally familiar with the local conditions. This way, possible problems that could result from different quality of the images due to different sensors (Landsat TM versus Landsat MSS), different atmospheric conditions and different seasonal aspects are being minimized; also, a common legend for forest classification is used for the whole tropical belt.

Interpretation results are analysed with a dot grid matrix of 100 x 100 dots, each dot representing an area of 400 ha. The land cover classes for each dot in each of the two satellite images were recorded and analysed using standard spreadsheet software, allowing for the quantification of forest classes at present as well as for change assessments between the two acquisition dates. The land cover classes are:

Results are then presented in tabular, matrix or map format (see table below). On the pan-tropical level these results confirm the results based on existing reliable statistics.

Pan-tropical area transition matrix for the period 1980-1990
Land cover classes, 1990 (million ha)
Land cover
Total 1980
millions ha
( % )
1275.98.979.279.172.5321.5734.791.783.951,367.96 (44.6)
0.86283.311.305.181.462.4010.180.050.21304.94 (9.9)
1.100.2648.611.080.792.352.270.050.0156.54 (1.8)
0.580.630.63159.330.451.4111.400.250.39175.06 (5.7)
Shrubs0. (5.6)
0.560.290.460.390.16119.797.300.190.17129.31 (4.2)
land cov.
0.710.700.261.351.942.03834.231.580.44843.26 (27.5)
Water0. (0.1)
Plantations0. (0.5)

The interpretation and validation of the satellite images has been carried out in cooperation with regional and national remote sensing institutions. The FRA 1990-Project is providing methodological guidance with the objective of strengthening capacity of these institutions, thus enabling them to continue monitoring of forest resources on their own in the future.

During the FRA 1990 project, a global GIS data base has been developed containing various information layers such as country boundaries, administrative boundaries within the countries, vegetation maps, forest cover statistics, eco-floristic zones, population density, population growth etc. These ancillary data are used for the development of the sampling design as well as for studies of the underlying causes of tropical deforestation and forest degradation. For instance, a high correlation has been established in many countries between population growth and deforestation.

Future plans

In follow-up proposals to FRA 1990, which should lead to a continuous forest cover assessment programme, it is planned to add another 120 samples each year so that the accuracy of the global assessment figures would increase and a complete coverage would be obtained in 10-year intervals. Inclusion of environmental parameters, yet to be precisely defined, has been discussed with UNEP for future forest resources assessments.

Future activities of FAO's Forest Department will focus on building and/or further strengthening of the countries capacity to monitor their own forest resources by providing additional advisory support, assistance in institution building, training, fellowships and equipment. It is foreseen to establish Lead Centres as focal points in the sub-regions, which have significant experience in the field of forest assessment and monitoring to encourage technical cooperation between developing countries (TCDC). In addition, twinning arrangements between specialized institutions in developing and developed countries are planned.

Development of a dedicated hardware and software configuration

The image analysis undertaken for FRA-1990 was based on standard colour composites at 1:250,000 scale. No geometric or radiometric correction nor specific enhancement was undertaken, although this would have facilitated the exact overlay of multi-temporal data, the adjustment of different atmospheric conditions or the better identification of forest classes. Geometric correction would also allow integration of interpretation results into a GIS data base and the combined analysis with other information layers, such as (for instance) administrative boundaries for compilation of statistics at sub-national level.

Present technology allows for integration of combined Image Processing and GIS software in user-friendly form on easy-to-maintain PC-platforms. However, a low-cost system specifically dedicated to forest applications at various scales from sub-national to regional, which would produce standardized outputs that could be aggregated to higher levels (i.e. from sub-national to national or national to regional) is not available.

Therefore, feasibility studies on the development of a dedicated hardware and software system for forest assessment and change monitoring, called RESPAS (Remote Sensing Processing and Archiving System), have jointly been prepared by FAO, the Dutch Aerospace Laboratory (NLR), the Agricultural University of Wageningen (AUW) and the International Institute for Aerospace Survey and Earth Sciences (ITC).

Based on these studies and on a user requirement survey, it is planned to develop a prototype Remote Sensing and Archiving System in close cooperation with selected developing countries, which represent different ecological environments, i.e. tropical rain forest, moist or semi-arid forest types. RESPAS aims at providing corrected and optimized remote sensing products to extract information required for the timely and efficient assessment, monitoring and sustainable management of forest resources.

The pilot RESPAS facility will be able to process and archive different types of satellite remote sensing data such as high and low resolution optical data, infrared data and microwave data. The architecture of RESPAS should be as flexible and as open as possible to be able to preprocess data from newly launched earth observation satellites and to account for additional user requirements which may arise in the future. RESPAS will be linked with (and at a later stage fully integrate) geographic information systems with information storage, retrieving and processing (simulation and modelling) capacities.

The system will primarily support forest planning and policy definition at national/sub-national level but it will also serve the needs of the Forest Resources Assessment Programme for provision of enhanced and rectified satellite data from selected samples to undertake regular and consistent estimation of the state of forest resources and the rate of its change at global, regional and national level.

RESPAS will also assist the participating countries in building up their own capacity to assess and monitor their forest resources (provision of hardware, software and training and logistic support), as required by UNCED Agenda 21.


The Forest Resources Assessment 1990 Project has demonstrated that a sampling design using high resolution satellite images (LANDSAT TM/MSS, SPOT, IRS-1, MOS and others) can provide a reliable estimate of global vegetation cover and its changes over time, on a continuous basis.

Moreover, it provides a unique possibility to describe in detail the direction of land cover change. Thus, it contributes to the understanding of degradation and deforestation processes in a way that has not previously been possible.

In accordance with Agenda 21 of UNCED, the capacity of the developing countries to monitor their own forest resources has to be developed and/or strengthened. This requires further efforts to assist the countries in need through technical assistance, institution building, twinning arrangements and training.

The enormous task to monitor the forest resources at global scale requires a cooperation of all institutions and organizations working in this field. A cooperation between FAO's Continuous Forest Resources Assessment Programme and the TREES project of the EU has already been agreed upon in principle.

Back to Top FAO Homepage