1.1 Main results by geographic region
The results presented in this section are based on a time-series analysis of reliable forest cover observations of 644 subnational units contained in FORIS. Some of the observations are single date and others multi-date spread over the period 1960 to 1990 as shown in Table 3.
Observations used for state and change assessment
|Frequency of observations||Number of observed subnational units|
The FORIS database is an important output of the assessment process and marks the starting point of tropical forest resources assessment on a continuing basis. Using the database, it is possible to make a study of historic trends and future prospects. As more multi-date observations become available and are added to the database, the precision of estimates, particularly of change rates, is expected to increase.
The current situation. As described in the chapter on Methodology, the estimations of forest cover and changes have been done at subnational level. Results are then aggregated at national level for reporting purposes (see Annex 1). The national estimates are further aggregated at subregional, regional and global tropical levels as presented in Table 4.
The pan-tropical forest cover was 1,756 million ha at end 1990 and 1,910 million ha at end 1980. Thus, average annual deforestation during the past decade amounts to 15.4 million ha (0.8 percent in compound annual rate of deforestation). The largest extent of forest cover was in Latin America and the Caribbean (918 million ha: 52 percent of the total tropical forest area), followed by Africa (528 million ha: 30 percent), and Asia and Pacific (311 million ha: 18 percent). The annual loss of forest cover by region was: Latin America and the Caribbean 7.4 million ha (0.8 percent), Asia and Pacific 3.9 million ha (1.2 percent) and Africa 4.1 million ha (0.7 percent).
Among the subregions, the following have relatively high rates of deforestation: West, East Sahelian and Southern Africa; Continental and Insular Asia; Central America and Mexico.
Estimates of forest cover area and rate of deforestation by geographical subregion
|Geographic subregion/region||Number of countries||Land area||Forest cover||Annual deforestation 1981–90|
|million ha||1980 million ha||1990 million ha||million ha||% per annum|
|West Sahelian Africa||6||528.0||43.7||40.8||0.3||0.7|
|East Sahelian Africa||9||489.7||71.4||65.5||0.6||0.9|
|Trop. Southern Africa||10||558.1||159.3||145.9||1.3||0.9|
|Asia & Pacific||17||892.1||349.6||310.6||3.9||1.2|
|Continental S.E. Asia||5||190.2||88.4||75.2||1.3||1.6|
|Insular S.E. Asia||5||244.4||154.7||135.4||1.9||1.3|
|Latin America & Caribbean||33||1,650.1||992.2||918.1||7.4||0.8|
|C. America & Mexico||7||239.6||79.2||68.1||1.1||1.5|
|Trop. South America||7||1,341.6||864.6||802.9||6.2||0.7|
Source: FORIS database
Comparison with FAO/UNEP 1980 Assessment. The results of the present assessment can be compared to those from the previous FAO/UNEP 1980 Forest Resources Assessment. For 76 countries assessed in common and covering approximately 98 percent of the tropical land area, the estimates are as presented in Table 5.
Comparison of FAO/UNEP 1980 assessment with FAO 1990 assessment
|Estimate||1980 FAO/UNEP assessment|
|1990 FAO assessment|
|Forest area by end 1980||1,935||1,910|
|Annual rate of deforestation|
|- during 1981–85||11.3||--|
|- during 1981–90||--||15.4|
Differences in the estimates of the 1980 forest cover and in annual deforestation rates for partly overlapping periods of the 1980s between both assessments at global (as well as regional and subregional) levels arise mainly from two factors: the increased amount of recent information from surveys related to 1980 and following years in the 1990 assessment; and the improved procedures for updating estimates at different years adopted by the latter. Therefore the differences in annual deforestation rates do not necessarily reflect an acceleration of the deforestation process during the 1980s.
1.2 The state of forest ecosystems
The report on the state of forest ecosystems, as presented in this section, is an important element of biodiversity assessment. Combined with data on species richness, the estimate of area losses by ecosystem can be used to make indicative estimates of loss of species richness.
|According to the international convention, biodiversity here means: “the variability among living organisms from all sources including, inter alia, terrestrial, marine and other aquatic ecosystems and the ecological complexes of which they are part; this includes diversity within species, between species and of ecosystems”.|
|The present assessment limits itself to a subset of ecosystems, namely tropical forests, and within that to mainly tree species, although the approach of the assessment is general and could be easily extended to include other plant and animal species.|
For the present reporting, the ecofloristic zones were grouped into four broad forest formations: one upland and three lowland. Figure 6 provides an overview of their distribution. The state of the tropical forests by formation and by country is given in Annex 1 and a summary in Table 6.
At end 1990 lowland formations were 1,544 million ha (88 percent of the total tropical forest area) and upland (hill and montane) formations 204 million ha (12 percent). Among the lowland formations, the tropical rainforests constituted the largest portion, namely 718 million ha (or 41 percent), the moist deciduous forests 587 million ha (33 percent) and the dry and very dry zone forests 238 million ha (14 percent). The rest of the forests, nearly 8 million ha, were scattered in the non-forest zone.
The annual losses of forest cover by ecological zone were: the tropical rainforests 4.6 million ha (0.6 percent), the moist deciduous forests 6.1 million ha (1.0 percent), the dry and very dry zone forests 2.2 million ha (0.9 percent) and the upland formations 2.5 million ha (1.1 percent).
Forest zones in the tropics
Estimates of forest cover area and rate of deforestation by main ecological zone
|Ecological zone||Land area||Population density 1990||Annual pop. growth 1981–901||Forest cover 1990||Annual deforestation 1981–901|
|million ha||inh / km2||% per annum||million ha||% land area||million ha||% per annum|
|Dry and very dry||1,241.0||70||2.3||238.3||19||2.2||0.9|
(hill and mountain forest)
(Alpine areas, deserts)
|TOTAL TROPICS 2||4,778.3||52||2.4||1,756.3||37||15.4||0.8|
1 Compound interest formulae were used for population growth and deforestation rate calculations.
2 Totals may not tally due to rounding.
Source: FORIS database
Comparing the forest area to land area ratio (see Figure 7) for each ecological zone, it appears that 76 percent of the tropical rainforest zone is still covered with forest. As expected, the percentage of forest cover declines with drier conditions and differences are quite obvious: 46 percent are forested in the moist deciduous zone, 30 percent in the dry deciduous zone and 19 percent for the dry and very dry zones together. One can assume that in the very dry zone the original forest area was considerably less than 100 percent of the total land area. This is in contrast to the rainforest and moist zones where the original forest area was close to 100 percent.
Assessment of biodiversity loss. The estimates of deforestation by ecofloristic zone, grouped by ecological region for reporting purposes, provide a basis for estimating biodiversity loss at ecosystem level. This information can be transformed into likely loss of species richness in the following manner.
The estimates of forest cover loss by main ecological region given in Table 6 are combined with information about human pressure by region. This is graphically presented in Figure 8.
Present forest cover by ecological zone
Data on distribution of higher plants (i.e. vascular plants) available for almost all tropical countries, are plotted against potential forest area of the respective country (see Figure 9). Different trends can be identified in the species area relationships according to the dominant ecological region in the countries. The gradients shown therein are representative of the forest formation sensitivity in terms of species loss due to deforestation. The same ecological conditions characterise also the forest formations for which the Project assessed the deforestation (see Figure 8). This fact is highlighted in Table 7 where annual rate of species loss was computed as due to deforestation.
The information provided permits a semi-quantitative evaluation of the risk of species loss due to forest area loss. It is important to highlight that the forest area loss, which is different in different forest formations and regions, has also different impact on the impoverishment of species richness.
The tropical rainforests of Asia seem to be most endangered in terms of likely impact of deforestation on species richness. The estimated loss is twice, though the area deforested is about half, that of Latin America and the Caribbean.
Forest proportion vs. population density by forest formation
Species/area relationships, plant species - 86 countries
Indicative estimates of species loss amongst higher plants due to deforestation
|Ecological zone||Forest cover area||Deforestation rate||Relative species richness||Indicative species loss|
|million ha||million ha||million ha||Species number||% Loss|
|Moist deciduous forest||274||251||22.5||13,000||2.5|
|Dry and very dry zone||162||151||10.8||1,900||1.0|
|Asia & Pacific|
|Moist deciduous forest||48||42||6.6||7,600||4.3|
|Dry and very dry zone||46||41||4.7||1,600||1.6|
|L. America & Caribbean|
|Moist deciduous forest||326||294||31.8||13,700||3.0|
|Dry and very dry zone||52||46||6.2||1,600||1.9|
2.1 Assessment by geographic region
The country level estimates are given in Annex 1 and a summary in Table 8. It is observed that the Caribbean, Central Africa and in Insular South East Asia subregions have still more than 200 tons per hectare, while the lowest values are found in tropical Southern Africa and in Sahelian Africa. It may also be noticed that the biomass per capita values are extremely low in South Asia due to very high population pressure. Low values are also found in West Africa, Sahelian Africa and Central America. It is remarkable that the average biomass per capita of Asia is only one fourth that of Africa and less than one tenth that of Latin America and the Caribbean.
Annual loss of biomass is estimated slightly over 2,500 million tons of which more than 50 percent is contributed by the Latin America region, nearly 30 percent by tropical Asia and about 20 percent by tropical Africa.
State of forest biomass and annual losses due to deforestation
|Region||Forest cover area||Forest biomass||Annual loss due to deforestation (1981–90)|
|1990||Average per ha||Total||Average per capita||Area||Biomass|
|million ha||tons / ha||106 tons||t/person||million ha||million tons||% of total|
|West Sahelian Africa||40.8||64.5||2,628||61||0.3||19||0|
|East Sahelian Africa||65.5||80.3||5,254||43||0.6||48||2|
|Tropical Southern Africa||145.9||60.5||8,824||93||1.3||81||3|
|Asia & Pacific||310.6||180.8||56,170||36||3.9||732||29|
|Continental South East||75.2||187.1||14,075||80||1.3||246||10|
|Insular South East||135.4||212.9||28,837||109||1.9||410||16|
|Latin America & Caribbean||918.1||185.0||169,844||426||7.4||1,303||52|
|Central America & Mexico Caribbean||68.1||95.2||6,483||55||1.1||106||4|
|Tropical South America||47.1||247.0||11,638||337||0.1||30||1|
Source: FORIS database
2.2 Assessment by ecological region
Estimates of potential and actual biomass density, as determined by modelling in GIS for most of tropical Asia are shown in Table 9. It is clear that the forests of insular Asia have both higher potential and actual biomass densities than those of continental Asia. The general trends in biomass density by ecoregion are consistent with expected patterns of biomass distribution: decreasing biomass density with decreasing moisture and increasing elevation. Forests in the continental dry zone and the insular seasonal zone appear to be the most degraded with actual biomass densities that are about 30 to 40 percent of their potential.
Potential (1980s forest area, without human impact) and actual (with human impact) biomass density (t/ha) for forests of tropical Asia by ecological zone.
|Ratio of actual to potential|
The study on the process of deforestation and forest degradation was conducted using the multi-date remote sensing based sample survey described in the chapter on Methodology.
3.1 Assessment by geographic region
The results are first presented for the Africa region based on 31 of 47 samples representing the entire geographical and ecological conditions, followed by a brief explanation of findings for the other regions.
For each sample a change matrix has been produced from the interdependent interpretation of two satellite images; the historical image acquired around year 1980 and the recent one acquired around year 1990.
The compilation of change matrices. The observed change matrices have been mathematically transformed into standardised matrices for the exact period 1980 to 1990 and further aggregated to produce the total given in Table 10. The class to class changes reported in this table highlight the complexity of the dynamics involved. As is expected from the process of such a large area almost all possible class-to-class changes are represented. However, most of the area changes are located above the diagonal implying a loss of forest area or cover, fragmentation, etc.
Another important inference can be derived from this table: thinking only in terms of deforestation is too simplistic. Deforestation is only one of the types of change in a complex process of area changes. It is important to consider all the facts of the change.
The change matrix for the Africa region based on the 31 samples
|Classes at year 1980||Classes at year 1990|
(areas in '000 ha)
|Closed forest||Open forest||Long fallow||Fragmented forest||Short fallow||Shrub||Other land cover||Water||Plantations||'000 ha||%|
|Other land cover||16.9||38.2||11.0||63.1||34.3||86.6||26,452||51.2||26,752.8||35.8|
Interpretation of the change matrix (Table 10) Comparing the row totals, showing the situation at year 1980, and the column totals, showing the situation at year 1990, illustrates the total area lost or gained by a class (see Figure 10).
Specific class-to-class changes can be read from each row. For example, the following changes from closed forest in 1980 can be read from the “closed forest” row. These are also illustrated in Figure 11.
During the period under consideration, 8.4 percent of the original closed forest cover (18,319,300 ha - 16,781,000 ha = 1,538,300 ha) was transferred to other classes.
247,500 ha (16.1 percent of the area changed) changed to other land cover;
524,300 ha (34.1 percent) was converted to short fallow agriculture which indicates the spontaneous pressure of rural population, rather than planned land use change;
291,800 ha (19.0 percent) changed to fragmented forest. Fragmented forest results from partial deforestation, usually by progressive clearing of small patches of forest, which creates a mosaic of forest and non-forest. In turn the fragmented forest mainly changed to other land cover (permanent agriculture) which implies that fragmentation is an intermediate stage toward permanent agriculture.
82,600 ha (5.4 percent of the area changed) changed to long fallow. This small portion of the total change indicates that this practice is relatively minor and of little relevance in this region;
382,100 ha (24.8 percent) changed to open forest, which indicates degradation by loss of density of the forest canopy. This is probably the effect of grazing and related fire practices, selective logging, fuelwood exploitation etc.
9,500 ha (0.6 percent) changed to shrub vegetation, which as in the case of conversion to long fallow agriculture, indicates that this practice is of little relevance in this region.
Land cover changes, Africa: 1981–90
Changes in closed forest cover, Africa: 1981–90
The changes described above refer only to the area originally in the land-cover class of closed forest. In order to achieve a more comprehensive overview, classes have been aggregated for analysis from nine land cover classes into five groups as follows:
|Land cover groups||Land cover classes|
|•||Continuous (or non-fragmented) forest||Closed forest|
|•||Fragmented forest||Fragmented forest|
|•||Other wooded land||Shrubs|
|•||Non-wooded area||Other land cover|
|•||Plantations||Plantations (forest and agric. plant.)|
Changes in continuous forest cover, Africa: 1981–90
The following points may be noted in the above grouping:
The changes have been grouped together on an increasing scale of biomass loss. The changes within the continuous forest group due to reduction of density (closed to open forest) or conversion of forest to long fallow are termed here “forest degradation”.
The changes from continuous forest cover to fragmented forest cover are termed “fragmentation” or “partial deforestation”.
Two types of “deforestation” are identified:
“Deforestation to other wooded land” where the forest is lost but a certain woody biomass remains;
“Deforestation to non-wooded area” where the forest is lost and no woody biomass remains.
The positive changes within the continuous forest are termed “amelioration”. This represents an increase in density within the forest (open forest to closed forest and long fallow to forest).
In 1980 the area of (continuous) forest cover was some 30 million ha in the sampled area of Africa. Figure 12 describes what has happened to the 2.6 million ha (8.7 percent) of continuous forest cover changed during the period 1980–1990.
With the exception of the negligible positive changes represented by Amelioration (1.8 percent, increase of density or decrease of perturbation in Continuous Forest) and by conversion to Plantation (0.05 percent), the bulk of the area has undergone negative changes as described below:
19.6% Degradation (decrease of canopy density or increase of perturbation)
25.4% Fragmentation (partial deforestation) On average this process represents a loss of two thirds of the original forest area, replaced by progressively increasing agricultural practices.
27.5% Deforestation: (forest to other wooded land cover group) This indicated two types of processes: loss of forest by change of its physiognomic characteristics (from forest to shrubs) and establishment of the traditional short fallow agriculture. The former process is rare while the latter is much more important in Africa. In both cases a certain amount of woody biomass remains.
25.7% Deforestation: (forest to non-wooded area) This represents total loss of woody biomass. It identifies the extreme level of degradation (denuded land), conversion to permanent agriculture or creation of water bodies (the latter being here negligible).
The sample survey data have been further analysed by ecological zone and corresponding change matrices produced.
3.2 Assessment by ecological region
In order to understand these processes in an ecological context, the results obtained from the remote sensing based sample survey were aggregated by ecological zone. Observations available so far justified main groups: (i) lowland wet and moist forest zone (with short or long dry period); (ii) Lowland dry forest zone (from dry to very dry); and (iii) Montane moist forest zone (moist montane and pre-montane with short or long dry period).
Results are first presented and discussed for the Africa region, followed by Asia and Latin America. Table 11 gives information about the sampled area in Africa, by ecological zone. The results indicate that the percentage of forest changes in the moist lowland zones (11 percent) is much greater than those of the dry lowland and moist montane zones (5.8 percent and 4.7 percent respectively). This can be explained by the higher suitability of the moist lowland zone to agricultural practices (and probably by the higher population density) in comparison with the other zones.
Table 11 and Figure 13 show, in tabular and graphical form, the types and extent of changes undergone during the period 1980–1990, by forest formation.
Changes in forest area 1 by ecological zones - Tropical Africa region
|Parameters||Moist lowland||Dry lowland||Moist montane||Total|
|'000 ha||%||'000 ha||%||'000 ha||%||'000 ha||%|
|Sampled area characteristics|
|Land area studied||30,175.0||27,693.3||13,751.0||71,619.3|
|Forest area 1980||17,503.6||8,697.0||3,805.8||30,006.4|
|% of land area||58.0||31.4||27.7||41.9|
|Forest area changed 1980–90||1,931.0||508.0||179.8||2,618.8|
|% of forest area 1980||11.0||5.8||4.7||8.7|
|Types of forest area changes (80–90)|
|Deforested to non-wooded||380.9||19.7||193.3||38.0||99.2||55.2||673.4||25.7|
|Deforested to other wooded||629.6||32.6||90.4||17.8||0.6||0.3||720.5||27.5|
|Total area changed||1,931.0||100.0||508.0||100.0||179.8||100.0||2,618.8||100.0|
1 Includes all forest classes except fragmented forest
Continuous forest cover change by main ecological zone, Africa: 1981–90
In all cases deforestation remains the major type of change. It is interesting to note, however, that while the larger share of the deforested area in the moist lowland zone goes to other wooded land, in the other two zones this type of deforestation is less important (dry lowlands) or even negligible (moist montane).
A preliminary pan-tropical result is given in Figure 14 which shows that the overall extent of changes are comparatively small in wet lowlands, increasing sharply in moist Lowlands, decreasing in dry lowlands and further decreasing in moist montane conditions. This pattern of change within the ecological zones seems to be relatively consistent across the tropics.
Changes in pan-tropic forest cover by main ecological zone: 1981–90
Figure 15 illustrates the ecological trend by geographical region: the higher rate of deforestation, fragmentation and degradation occurring in the moist lowland zones. A pattern similar to that observed in Africa is found in Asia and Latin America.
In both Brazil and tropical Asia change from Continuous Forest occurs at a rate much higher than in tropical Africa, probably owing to the higher population density (in Asia) and planned resettlements/resources exploitation programmes (in Brazil and Asia). It is interesting to see that in both Africa and Asia the area deforested is consistently accompanied by fragmentation and degradation processes, probably being preliminary stages to deforestation. The change process in Brazil is clearly different, where almost all the changed area is deforested, particularly to non-wooded land.
Forest cover changes by main ecological zone: 1981–90
3.3 Preliminary evaluation of the sampling design
The remote sensing survey was designed to be statistically sound and provide more precise estimates of forest cover and change at the regional and global levels than would be possible with the help of existing data, due to their inherent variations from country to country in data quality, inventory standards and reference dates. Further, the survey was expected to provide reliable and consistent multi-date information for modelling of changes and comprehensive information on the process of change.
The results obtained so far confirm that the level of expected precision at global and regional levels will most likely be reached as shown in Table 12.
Initial expected and present expected standard errors of forest cover estimates
|Region||Sampling units||Standard errors (%)|
|Total No.||Sample size||Initial expected||Present1 expected|
|Asia & Pacific||277||30||8.2||10.8|
|Latin America & Caribbean||480||40||4.7||5.4|
1 using information from the completed samples.
Further analysis shows that the stratification is of limited value in Africa, of some value in Asia and gives a substantial improvement in Latin America. It has been observed also that a stratification which is useful for forest cover estimation is not necessarily suitable for change estimation. In view of this different approach to reduce the sampling error, using appropriate ancillary variables in each situation has to be worked out. With proper selection of ancillary variables, the precision is expected to improve in the case of both forest cover and change estimates.
The information on the process of change provided by the sample survey is unique and most useful for policy-making. The change matrix is an invaluable tool for illustrating the future consequences for any time horizon if no action is taken to change the development of the process.
The results are given here for the West Africa and Amazon Basin subregions because they represent two interesting and contrasting fragmentation patterns:
The West African subregion shows an advanced phase of forest depletion: the closed forest is split into several islands in a “sea” of mixed agriculture/forest fallow;
In the Amazon basin deforestation is moving from the edges to the core of the main forest body. In this landscape, fragmentation is mainly represented by the complexity of the forest interface with other land cover types.
The two patterns represent different stages of demographic development: the former mature, heavily influenced by human pressure, the latter a relatively recent phase with low population pressure.
4.1 The case study in West Africa
This subregion is characterised by a very high ratio of forest fallow to closed forest even in “closed rain forest climax” ecofloristic zones and the remaining closed forest is separated into islands as shown in Figure 16.
A database with the geometrical properties of the forest islands was generated and statistics were produced at district and national level. The fragmentation was measured through the PAI.
As shown in the bar diagram (Figure 17), the PAI is well correlated with the deforestation figures estimated at national level by the Project. Note that in Liberia and Sierra Leone have been excluded due to the presence of clouds in the two countries at the time the satellite imagery was made, because this produced an apparently higher degree of fragmentation than the true values.
West Africa: NOAA derived forest map
Fragmentation vs. deforestation, West Africa
4.2 The case study in Brazil
The two Brazilian states, Acre and Rondonia, have different landscape features to West Africa. In Brazil the forest is the dominant cover type and deforestation proceeds along more or less linear transportation corridors such as the Amazon River and the main state roads. The effects of the deforestation are clearly visible from the maps (see Figures 18 and 19).
The two states have different deforestation rates and it is interesting to analyse how this process is accompanied by the presence of zones “at risk”, that are likely to be subject to further deforestation. These zones, quoted in the map legends as “edge forest”, were delineated on the forest map by arbitrarily setting a buffer zone of suitable width, 10 km in this case, from the edge of the mapped forest boundaries toward the core. This forest belt or zone covers the area already disturbed, or likely to be disturbed or degraded. Spatial statistics were produced for the two regions as shown in Tables 13 and 14.
Acre: NOAA derived forest map
Edge/core statistics in Acre State
|Map : Acre - NOAA vegetation map|
|Total of four classes||100.00|
Rondonia: NOAA derived forest map
Edge/core statistics in Rondonia State
|Map: Rondonia - NOAA vegetation map|
|Total of six classes||100.00|
Edge/core ratio in Acre and Rondonia, Brazil
|Forest Class||Forest region|
Table 16 shows the current figures on deforestation together with the edge/core ratio taken from the previous table.
Relation between edge/core ratio and deforestation
Source: FORIS database
The annual deforestation rate in Rondonia is estimated to be about 1.1 percent of the remaining forest, while in Acre it is 0.5 percent. The edge/core analysis reveals that more than 75 percent of Rondonia's forests fall in the “edge zone”, against the 25 percent in Acre. These statistics are clear indicators of the risk of further degradation of the forest resources.