Previous Page Table of Contents Next Page

Chapter III


The technique of assessment is presented in the form of a flow chart in Figure 2. It consists of the following three steps: (i) establishment of a computerised database; (ii) development of deforestation model (or an adjustment function); and (iii) computation of results for the standard reference years.

Figure 2
Forest cover state and change assessment (using existing reliable data)

Objectives: Produce standardised data relating to:

Figure 2

*    FORIS is the acronym for Forest Resources Information System. It is a computerised database to store/retrieve national/sub-national forest resource information

**   GIS stands for Geographic Information System used to store spatial/cartographic and related statistical data

***  Multi-date remote sensing data was used in the model for the Africa region as existing data was not considered adequate.

1.1 Establishment of a database

The data compiled by the Project consists of two categories:

  1. Tabular data including forest resources, population and socio-economic data at the subnational level (province, state).

    The Guidelines for assessment based on existing data were prepared in the three official FAO languages to compile the country statistics and bring them in a framework of common concepts and classifications. Using the guidelines, the country reports were reviewed and reliable information extracted, edited and stored as part of the Project database with the acronym FORIS which stands for FOrest Resources Information System.

Forests are defined as ecosystems with a minimum of 10 percent crown cover of trees and/or bamboos, generally associated with wild flora, fauna and natural soil conditions, and not subject to agricultural practices.
Deforestation refers to change of land use with depletion of tree crown cover to less than 10 percent. Changes within the forest class (from closed to open forest) which negatively affect the stand or site and, in particular, lower the production capacity, are termed forest degradation. Degradation is not reflected in the estimates.

FORIS is a database as well as a database management system. It runs in a PC environment using dBASE IVTM. The data entering facility is user-friendly. Only minimum computer knowledge is required in order to enter, correct and print the data. FORIS is a useful database and database management system for planners and researchers.

  1. Map data including vegetation types, ecofloristic zones and country and subnational boundaries

    Realising that deforestation is a location-specific process driven by, among other things, population pressure and environmental conditions (particularly the population carrying capacity of the area), demographic and ecological parameters were included in the database and integrated with the statistical data in the form of a GIS. From the outset it was realised that a GIS was essential to handle statistical and spatial data from a variety of sources, in a variety of formats and in different projections. Therefore, GIS development was given special attention by the Project.

    In the course of time, more spatial data were added to the Project GIS including LANDSAT 4.5 World Reference System Grid, potential global and forest biomass, mean annual precipitation and biotemperature, topography, protected areas, and vegetation map from NOAA/AVHRR satellite imagery (for some regions).

1.2 Deforestation model

The forest cover data contained in FORIS refer to different periods and need to be brought for reporting purposes to the standard years, namely 1980 and 1990. This was done with the help of a deforestation model (or a forest area adjustment function) which correlates forest cover change in time with other variables including population density and population growth for the corresponding period, initial forest cover area and the ecological zone under consideration. For developing the model only multi-date data at the subnational units are used and the function expressed in the form of a differential equation.

Deforestation Model
The model, expressed in the form of a differential equation, has the form:
-Y is the percentage of non-forested area in a subnational unit computed as: Y = 100 * (Total area - Forest cover area)/(Total area)
- P = 1n(1+Population density) with population density expressed in persons per square kilometer (1n is the natural logarithm)
 dY/dP is the derivative of Y with respect to P
- b1, b2 and b3 are the model parameters.

The model interprets the human/forest interaction as a biological growth process. Very much like some biological growth processes, deforestation is observed to increase relatively slowly at initial stages, much faster at intermediate stages, and to slow down at final stages (see Figure 3).

Validation of the model was performed using a set of 16 observations not used to build the model, for which two or more reliable estimates of forest cover in time were available. The mean difference between the observed and the predicted changes was ± 10.6 percent. The residual distribution, despite some large deviations in the case of some subnational units, was quite symmetrical with no evidence of bias.

Figure 3
Illustration of model curves for different ecological zones

Figure 3

1.3 Assessment by geographic region

The FORIS database, in conjunction with the model, serves: (i) to adjust forest cover area data of the subnational units to the standard reference year 1990; and (ii) to produce estimates of the forest cover area change over the period 1981 to 1990. For these purposes the most recent forest inventory data of a national/subnational unit are used as baseline, and forest cover area in 1980 and 1990 (standardised results) computed according to one of the following options:

  1. Reliable multi-date inventories available; in this case the existing multi-date information is used to calibrate the general model with the ‘local’ b1 parameter and the resulting model is then used to compute the standardised results. This is the optimum case.

  2. Reliable single-date inventory available; in this case the standardised results are computed using the general model.

  3. No reliable inventory available; in this case estimates of baseline forest cover area are extracted from calibrated vegetation maps existing in the projects' GIS and are then used as input to the modelling procedure described for option (ii).

The estimated values at subnational level are aggregated at national, regional and global levels. Keeping in view the law of propagation of errors the global estimates are expected to be more precise than the subregional; and the subregional estimates more precise than the national and subnational estimates.

For each option, procedures have been developed for integrating the FORIS and GIS data to compute the model parameters for each unit (subnational in most cases) and to provide forest cover in 1980 and 1990 as standardised output.

The model has proved to be a valid and flexible tool to estimate deforestation at a global level and to relate forest cover changes with demographic and ecologic variables. Given the scarcity of existing reliable multi-date observations, the model serves a very useful purpose in global forest resources assessment.

In the present system of assessment the up-dating of results is an almost continuous process. Both FORIS database and deforestation model parameters can be updated as and when additional country data become available. The addition of further records and fields in FORIS will improve the reliability of model parameters as well as precision of estimates for the country which in turn will improve the precision of the estimates at regional and global levels.

1.4 Assessment by ecological region

The environmental implications of deforestation and forest degradation are determined jointly by the type and magnitude of the human intervention as well as the ecological context of the land and forest area under change. Keeping this in view the Project developed a new system of global assessment whereby reports on the current state of forests and on-going changes are made by ecological region. As will be presented later, this type of assessment forms a basis of study on biodiversity loss at the ecosystem and species levels.

The ecological regions are defined with the help of ecofloristic zone (EFZ) map. The criteria applied in classifying ecofloristic zones are to begin with ecological: climatic, physiographic, and edaphic. Bioclimatic limits are determined based on parameters of mean annual rainfall, rainfall regime, length of the dry season, relative humidity and temperature. The physiographic contours, the soils and the leaf-retaining characteristics of the forest canopy (i.e. phenology) are used to further subdivide the bioclimatic zones. Within the ecological types thus defined, subtypes (syn. ecofloristic zones) are distinguished based on the dominant or characteristic woody species of the flora, with attention given to their successional positions. Thus, for each ecofloristic zone it is possible to define the vegetation types that it contains.

Vegetation types constitute another layer of important data for the state of forest ecosystem reporting. The vegetation classes are distinguished mainly by physiognomic criteria which can be seen in field and remote sensing documents, such as density, continuity of plant cover, height, etc, dense forests, secondary forests, woodlands, thickets, savannas, etc. are identified. The denomination of the formations is derived from the classification of Yangambi (1956) and of UNESCO (1973). These formations are further subdivided on the basis of density, ranging from the woody types that are the most closed, to the types that are the most open. This reflects the different stages of the regressive series or, in very rare cases, the evolutionary tendency of the vegetation within a zone.

Ecofloristic zones and vegetation types were mapped with consistent methodology across all regions, but have distinct legends. Separate maps exist for Africa, Continental South and Southeast Asia and South America.

The ecofloristic zone and vegetation maps are integrated with FORIS database to produce estimates of forest cover and change by ecological region.

A review of literature was made with a view to organising the existing knowledge on estimating biodiversity and its loss. The literature review suggested, as is well known, that the species distribution has a characteristic distribution pattern which is comparable within forests occurring in similar ecological regions. It was also observed that, within a given forest type, the number of species increases with the increase of forest area.

The main steps in estimating biodiversity loss were the following:

  1. Estimation of species richness in the inventoried areas, where reliable species identification has been done. This formed the main basis of species - area relation by main ecological region.

  2. Estimation of loss of forest area by ecological region and its likely implication for the loss of species richness.

It need not be mentioned that the present state of knowledge does not permit making reliable statements on the extent and rate of loss of biodiversity.


Forest biomass is the source of many products, including timber, fuel and fodder. Assessing the total aboveground biomass of forests (per unit area and for whole forested regions) is a useful step towards quantifying the amount of resource available for all these traditional purposes.

Here biomass is defined as the total amount of above-ground organic matter present in trees (leaves, twigs, branches, main bole and bark), expressed either as oven-dry tons per hectare (referred to as biomass density) or as oven-dry tons per country, region, etc (referred to as total biomass). For most forests, this will include only the biomass in trees with diameters greater than 10 cm, however for forests of smaller stature, such as those in the dry tropical zones or degraded forests, the minimum diameter could be smaller than 5 cm. Furthermore, other forest components such as palms and bamboos are included only where they comprise an important component of the forest and local use. Not included are (1) undergrowth (except where important locally, see above) which is generally less than 5 percent of the aboveground biomass density, (2) forest floor fine litter, or (3) lying and standing dead wood.

Biomass of forests has become very relevant to studies related to global change. The biomass of forests provides estimates of the carbon pools in vegetation (about 50 percent of biomass is carbon) and consequently the potential amount of carbon dioxide that can be added to the atmosphere when the forest is cleared and/or burned. Biomass is also a useful variable for comparing structural and functional attributes of forest ecosystems across a wide range of environmental conditions.

Not all biomass for domestic use originates from forests; significant quantities are obtained from non-forest lands such as woodlots, windbreaks and other line formations, home gardens, etc. It is recognised that this source should be assessed in the future, but it was beyond the present scope of the FRA 1990 Project.

2.1 Assessment by geographic region

To estimate the biomass density of forests use is made of existing volume over bark estimates (VOB) in the FORIS database which is converted to biomass density with the help of biomass estimation function and further “expanding” this value to take into account the biomass of the other above ground components.

Function for estimating biomass from volume information
BD (t/ha) = VOB * average WD * BEF
where: BD = biomass density
VOB = volume over bark of all trees to 10 cm minimum diameter (m3/ha)
WD = average wood density (t/m3); values obtained from FRA 1990 guidelines
BEF = biomass expansion factor

Research results based on actual inventory data across the tropics show that for tropical broadleaved forests biomass expansion factors are significantly related to stemwood biomass (SB) according to the following model:

BEF= exp{3.213 - 0.506*In(SB)}for SB < 190 t/ha
 = 1.74for SB ≥ 190 t/ha

No model for calculating biomass expansion factors for coniferous forests is available at present because of general lack of data for this type of analysis. Therefore the estimates presented in the project guidelines were used.

This approach is of unknown reliability because much of the VOB data were estimated generally through extrapolation from existing local and international forest inventories. However, this method has the advantage of being pan-tropic and can therefore be used with area estimates for 1990 to produce an assessment of the total biomass of tropical forests.

2.2 Assessment by ecological region

The approach of estimating biomass density and biomass degradation, so far implemented for tropical Asia only, is based on a modelling method with GIS technology using various existing databases such as reliable inventory, population density and climatic data to produce digital maps of climatic indices and existing digital maps of vegetation, ecofloristic zones, soils and topography. The work is a cooperative effort between the Project and the staff of the University of Illinois, USA. The advantage of this method is that it produces biomass estimates directly without having to resort to an empirical extrapolation process. The disadvantage is that very few inventories give complete stand tables including small diameter classes for all species, thus not all countries in the tropics are covered by these estimates.

In the present approach, the following data layers (GIS-digital maps) are used as inputs to the modelling:

Potential biomass density (PB) is assumed to be a function of bio-climatic factors according to the following simple model:

PB (t/ha)= f[WCI, precipitation, soil (texture, depth, slope), topography]

Each of these factors is spatially represented by a numerical scale whose values were ranked according to how the particular factor affected biomass. The digital maps are overlaid according to the above model, and the results validated and calibrated using existing forest inventories for mature forests and other literature sources and the EFZ map.

GIS modelling provides results on biomass density (t/ha) by ecofloristic zone. These data could be combined with the area estimates to obtain total biomass by ecological and geographic region. Using the Project's vegetation map as a indication of forest cover for 1980, a spatial representation of forest biomass density was prepared by region.


The estimates of forest cover area and deforestation rate at country level based on Phase 1 indicate the magnitude of deforestation.

To undertake control measures or simply to diagnose the problem, a better understanding of the change process is essential. For this purpose, the following types of questions need to be answered:

How do the tropical forest resources change?
How much is degraded? Fragmented?
What happens to deforested land?
Under which ecological and socio-economic conditions?
What are the causes of deforestation?

The only satisfactory way to provide reliable information on the process of change is to establish a forest resources monitoring system, using a globally compatible and consistent methodology. This provides reliable and location specific change information. In consideration of costs, precision and timeliness of results, a remote sensing based sampling approach was designed and used to cover the entire tropical zone in a time frame of about one year (see Figure 4). The specific objectives were:

  1. to achieve the highest level of consistency and precision in the assessment of forest cover state and change at global and regional levels;

  2. to develop and disseminate a simple and robust monitoring technique for producing estimates of forest cover state and change at global and regional levels with application also at national level;

  3. to provide spatial and statistical data for estimating class to class changes of land cover and forest cover categories between the two dates of interpretation at the sample locations and for producing change matrices at regional and global levels.

A distinctive feature of the methodology lies in the fact that it provides not only forest cover change data, but also change maps and matrices for each sample location. This enables estimation of class to class changes of land cover and forest categories between the two dates of interpretation at sample, regional and global levels: essential information for understanding the complex processes taking place, such as deforestation, fragmentation, degradation, afforestation, etc. An example of raster maps derived from image interpretation change maps and resulting transition matrix is given in Figure 5.

The Sample Survey Design
The survey is based on a sampling design covering all tropical countries. The World Reference System 2 (WRS2) for the LANDSAT satellites is used as the sampling frame. LANDSAT scenes covering approximately 3.4 million ha serve as the sampling unit.
A two stage stratified random sample is carried out. In the first stage the survey area is divided into sub-regions on a geographic basis. In the second stage the sub-regions are divided into forest cover strata. The allocation of samples to the sub-regions it made proportional to estimated deforestation. Within a sub-region sampling units are selected with equal probability.

Figure 4
Pan-tropical continuous forest resources survey design

Figure 4


Figure 4Figure 4
Date: 13/06/81 P/R 185/68Date: 20/07/79 P/R 184/68

Figure 5
Example of spatial and statistical outputs - sample no. 1510, located along Zaire/Zambia border line

a: Maps derived from interpretation of satellite images of sample no. 1510
Figure 5Figure 5
04/07/1989         13/06/81         20/07/79
b: Change maps produced by comparison of common areas of recent and historical images. The 100 possible class combinations (10×10) have been reduced to 14 change classes for display purposes.Figure 5
changes (1981–89)changes (1979–89)
c: Sample no. 1510
- Standardized transition matrix computed for the reference period 1980–1990
The two matrices reporting the cover class changes for periods 1981–1989 and 1980–1989 have been mathematically adjusted to the standard period 1980–1990 on the basis of transition probabilities and subsequently combined to form the comprehensive change matrix presented here.

Cover Classes at year 1990
Cover Classes at year 1980Closed ForestOpen ForestLong FallowFragm. ForestShort FallowShrubOther Land CoverWaterManmade Woody Veg.Total at year 1980
'000 ha%
Closed forest1672.3143.3  273.4 17.5  2106.571.2
Open forest0.984.9  4.2 0.0  90.03.0
Long fallow           
Fragm. forest           
Short fallow    574.9 1.3  576.119.5
Other land cover      160.6  160.65.4
Water       21.2 21.20.7
Manmade W.veg
Total at year 1990'000 Ha1673.2228.2  852.4 179.421.22.42956.8 
%56.67.7  28.8 100.0

The main characteristics of the survey design are as follows.

Table 2 gives the total number of sampling units by region, the number of sampling units selected for the study, and the expected sampling error at 95 percent probability level for forest area figures, which have been calculated using appropriate formulae for stratified random sampling.

Table 2
Allocation of sampling units by region and expected standard error

ContinentSampling unitsProportion of forest/land areaEstimated forest area standard error (%)
Total No.Samp.size
Asia & Pacific277300.458.2
Latin America & Caribbean480400.684.7

At each sample location, satellite images of the best quality and appropriate season, separated by an approximate ten year interval, are selected for observation. The image close to 1990 enables state assessment; whereas the area in common between the “1990” and “1980” images enables the change assessment.

The salient features from a remote sensing point of view are the following:

  1. Interdependent interpretation procedure; this interpretation approach secures the highest level of spatial consistency between historical and recent image classification.

  2. Standard classification of various forest classes (closed, open, with shifting cultivation, fragmented) on a pan-tropical basis.

  3. Change matrix; each sample produces a transition matrix from which the processes of change can be analysed.

  4. Image archive; all images used represent permanent reference as part of a continuous time series; in future these images will be used to estimate the speed of change (3 or more time series)

Inter-dependent Interpretation Procedure
  1. Image characteristics
    • LANDSAT MSS (20 percent of recent images will be TM) and IRS (Indian Remote Sensing Satellite) satellite data
    • Two dates, one close to 1990, one close to 1980, at least five years apart, same season (preferably beginning of dry season)
    • Standard enhancement for vegetation assessment
    • Common cloud-free overlay of one million ha minimum

  2. Interpretation procedure
    • Interdependent interpretation, namely visual interpretation of both images by the same interpreter (coming from the region and familiar with the location) at the same time
    • Standard and globally compatible land use classification

  3. Image interpretation recording
    • User-friendly data recording from dot grid using customized LOTUS 1–2–3® spreadsheet software
    • Generation of change matrices showing land use changes

  4. Field verification of selected samples in cooperation with national forest services and remote sensing agencies

The interpretation was implemented at selected regional and national forestry and/or remote sensing institutions which have a good knowledge of the sample locations and traditionally involved in forest resources assessment activities. With the two-fold objective of strengthening national capacities for forest monitoring and improving the quality of image interpretation, the Project has so far organised three regional workshops and eight training sessions with national institutions, benefitting 27 countries and 81 participants.

The results and quality of the interpretation done by local institutions are centrally reviewed and evaluated. A database is established and analyses carried out as will be presented later. The results of the remote sensing survey have found the following applications.

Considering the extremely poor information available on the tropical forests in general, and on the on-going processes of change of the tropical forest resources in particular, this project component could be considered an important achievement.


A continuous forest cover is termed fragmented when in the course of time it is broken down into disjunct parts by conversion of parts of the cover to non-forest uses, e.g. by building of roads or introduction of agricultural crops, etc.

Different indices have been developed in the past years to quantify the landscape variability including specifically forest fragmentation. In the framework of the Project, two of the indices were used: the Perimeter Area Index (PAI) and the edge/core ratio (ECR).

Fragmentation indicators
The Perimeter Area Index (PAI) for forests located in a geographic region is computed by the following formula:
Where N is the number of forest islands, Pa their total measured perimeter and A the total area.
The edge/core ratio (ECR) in a forest zone is derived from the following formula:
Where “edge forest” is the forest area on the edge zone, 10 km wide here, from the edge of the mapped forest to its interior. This is usually the area most subject to anthropic impact.
“Core forest”, is the inner forest areas, in the present case assumed to be 10 km beyond the edge of forest. This is usually the undisturbed forest area.

The main objective of the study on fragmentation was to investigate the relationship between fragmentation and variables such as deforestation and forest degradation.

Vegetation maps derived from NOAA-AVHRR satellite data (1 km resolution) were used as the source document due to their resolution and recent acquisition date. Two areas were chosen for the study, namely the West Africa and the Amazon Basin subregions, because they represent two interesting and contrasting patterns of forest fragmentation.

Previous Page Top of Page Next Page