Part I: Global issues

Chapter 1. Forest area and area change
Chapter 2. Wood volume and woody biomass
Chapter 3. Forest plantations
Chapter 4. Trees outside the forest
Chapter 5. Biological diversity
Chapter 6. Forest management
Chapter 7. Forests in protected areas
Chapter 8. Fires
Chapter 9. Wood supply
Chapter 10. Non-wood forest products

Forest area changes 1990-2000 (million hectares)

Chapter 1. Forest area and area change

ABSTRACT

Forest area and area change was a major theme in FRA 2000. Estimates were based on a comprehensive analysis of the latest forest inventory data available for each country. To support the country findings a systematic pan-tropical remote sensing survey provided estimates at the regional and pan-tropical levels. A detailed study of forest plantations was included in the analysis. A number of qualitative studies were carried out to enrich the knowledge on forest area change. This was the most comprehensive survey of forest area and area change at the global level to date, revealing extensive and detailed findings, but also considerable information gaps, particularly in Africa. The findings indicate that the world's forests covered 3 869 million hectares in 2000, about 30 percent of the world's land area. The net change in forest area was -9.4 million hectares per year, representing the difference between a deforestation rate of 14.6 million hectares per year of natural forests and an expansion of 5.2 million hectares per year of natural forests and forest plantations. In addition 1.5 million hectares per year of natural forests were converted to forest plantations. Most of the forest losses were in the tropics. The rate of net change was slightly lower in the 1990s compared to the 1980s, due to a higher estimated rate of forest expansion in the 1990s. A survey of scientific literature showed that the subject is rich in publications, but also that conclusions are not representative of all forests. Country studies point at land use rights as a common main determinant behind land use change (deforestation), and the remote sensing survey indicated that direct conversions of forests to permanent agriculture were more prominent than shifting agriculture in forest change processes.

INTRODUCTION

The core of FRA 2000 is the estimate of forest area and changes in forest area over time. The work has led to new knowledge about the dynamics of the world's forests. Forest area is an easily understood baseline parameter that provides the first indication of the relative importance of forests in a country or region. Estimates of forest area changes may also provide a clue to the demand for land for other uses and environmental pressures on forest ecosystems.

The main limitation of the emphasis on forest area is that this is not necessarily a good qualitative indicator of the health of a forest ecosystem. For example, environmental values such as critical areas for biological diversity may be concentrated in small and scattered areas or where forests are interwoven with other land uses. Such values may not be well assessed through spatial area-based classifications of disturbance. Social values may be derived from complex interactions and synergies between agriculture and forestry at the local level, and they may not be highly correlated to the absolute extent of the forest. Economic values are more dependent on variables such as productivity, volume, species composition, accessibility, demand, non-wood forest products and regulations than on the overall forest area and area change.

Nevertheless, forest area and area change remains a basic theme for global forest resources assessments. FRA 2000 has made considerable efforts to record and explain the extent of forests and the area transition to and from forests. An effort has been made to show clearly where the source data are weak and where the transformation into global classes was particularly difficult. In addition to the area estimates, qualitative studies of the change processes have been made through literature reviews.

This chapter synthesizes work that is described in more detail in other chapters and in the tables in the Appendix; a large number of FRA Working Papers; and conclusions reached at numerous workshops and meetings during the FRA 2000 process.

TERMS AND DEFINITIONS

Global assessments of forest resources coordinated by FAO have a long history. Over time, basic definitions have been developed that are generally accepted by participating countries and are well known to experts of forest inventories and assessments. It is nonetheless important to continue to develop and refine a core set of terms in light of changing technology and information requirements of countries, which are the primary users of the assessments. Common terms are needed to produce consistent and comparable estimates for each country. On the other hand, there is no single definition that fits well everywhere, owing to the greatly varying conditions throughout the world. Definitions of forests and forestry must extend over and take into consideration not only the climatic range from boreal to tropical ecological zones, but also economic variations from countries where recreational values rank among the highest priorities, to locations where collection of fuelwood and non-wood forest products is part of daily life. Global definitions by necessity involve compromises.

Figure 1-1. Schematic illustration of the two principal forest edges in the "forest" definition

International terms and definitions are not static, but follow the general development of international processes. For example, the importance of forests as carbon sinks was not widely discussed several decades ago, yet this issue is now at the top of the international political agenda. New terms are introduced as subjects enter into the international debate and old terms may need to be modified to better serve the current requirements for information. While this is a desirable evolution, it is also important to keep definitions consistent over time. Forests change relatively slowly, and it is necessary to compare estimates several decades apart to establish reliable trends. For this purpose, FRA 2000 has tried to maintain a globally homogeneous set of definitions that allows comparisons with earlier global forest resources assessments.

The most important exception to this rule was a change in the definition of "forest" for industrialized countries in order to adopt a single common definition for all countries. In FRA 1990, a 20 percent canopy cover threshold was used in countries with temperate and boreal forests for which data were collected by ECE/FAO. In contrast, 10 percent canopy cover was used as the minimum definition of forest in tropical countries. This distinction dated back many years, based on differences in forest inventories in different regions. During discussions of forest resource assessments in the Intergovernmental Panel on Forests (IPF), consensus emerged to use 10 percent minimum canopy cover as the common definition for forests in all countries in FRA 2000. As a result, FRA 2000 provided globally homogeneous data on forest cover. However, comparisons with earlier assessments required considerable work and extrapolation of previous data, particularly in dry subtropical zones and boreal zones where the extent of sparsely stocked forests with between 10 and 20 percent canopy cover is considerable.

The definitions used in FRA 2000 are found in Appendix 2. It is important to note that "forest" is defined both by the presence of trees and by the absence of other land uses regardless of the legal status of the land. In other words, "forest" is a combination of a land use classification and a land cover classification. There is some disagreement about this approach in the scientific community, but a majority of experts on forest assessments agree that such an approach is necessary if the FRA 2000 results are to be of optimal use to policy-makers. One practical outcome of this approach is that interpreters of remote sensing images must have knowledge of the situation on the ground; it is possible to interpret land cover from space, but it is not so simple to identify land use.

In estimating forest change, there are two principal kinds of forest edges. The first is the edge between a forest and areas where the climatic conditions are too harsh to support forest vegetation, such as the northern edge of a boreal forest or the edge between a forest and a desert. The second is the edge between a forest and areas where other land uses are practised, including agriculture, urban land use or infrastructure (Figure 1-1). The first edge is strictly derived from biophysical properties, whereas the second is subject to other considerations.

One question is, what degree of other land uses can be allowed without disqualifying the "forest" classification? Obviously some grazing can occur in what is called a forest, as well as collection of non-wood forest products. However, when other land uses dominate, the land use classification would not be forest. This raises questions about which products and purposes should be included in the "forest" classification. Stands of rubber trees and oil palms are included, whereas fruit orchards and agroforestry areas are not. National park areas are included, whereas urban parks are not. It is important to document how the definitions were applied in each case to enable future comparisons.

In addition to land use classification, area change processes were central to the assessment. These change processes were defined by seven terms for which the distinctions between terms were clear and the set of terms as a whole covered all possible changes. The seven identified change processes (Figure 1-2) can be grouped into land use changes (deforestation, afforestation, expansion of natural forests) and internal changes within the forest class (reforestation, regeneration of natural forests, degradation, improvement).

METHODS

Several different approaches were used to assess the extent of forest and its development over time. In Figure 1-3 the main processes and outputs are shown, together with a note on where further information can be found. For area statistics, FRA 2000 generated information at three scales - country (based on surveys of national inventory and mapping reports), region (FRA 2000 remote sensing survey) and world (FRA 2000 global mapping). For the estimates of area and area change, only country- and regional-level information was used, as the global forest map did not provide sufficient precision. The global-level information was used to derive relational data such as the distribution of forests by ecological zones.

Figure 1-2. Seven basic change processes for forests

In addition to the area statistics, narrative descriptions on woody vegetation and forest plantations were developed for many countries to accompany and enrich the area statistics. Descriptions of forest resources - including forest area and area change - are elaborated for all countries and presented for each subregion in Part II of this report. Finally, qualitative studies and literature reviews were carried out to deepen the knowledge on factors underlying forest changes.

FRA 2000 considered all available documents that contained primary country-level information on forest area, forest area change and forest plantations. Requests for information to all countries were followed up by in-country assignments in most developing countries, workshops and meetings involving more than 100 countries and a final validation of results by country correspondents.

In many countries, primary information on forest area was not available or was not reliable. Other countries lacked a time series of forest area information. In these instances, FRA 2000 had to rely on secondary information and/or expert estimates. Table 1-1 summarizes the information availability for forest area. The world average reference year for source data is 1994, with considerably older dates in some developing countries. A high proportion of developing countries had to rely on expert opinion for the latest area estimates. Furthermore, fewer than half of all countries have time series information with high compatibility between the observations.

Table 1-1. Forest area information availability and quality by region

Region	Reference year for latest available area data (area weighted)	No. of countries									References reviewed (No.)
		Source data for forest area estimate				Time series used for area change estimate		Time series compatibility
		Expert estimate	General mapping	Detailed mapping	Field survey	Yes	No	High	Medium	Low
Africa	1991	24	6	5	10	35	21	11	12	13	547
Asia	1995	14	6	9	1	28	14	11	14	3	284
Europe	1997				3	38		32	6		44
North and Central America	1995	15	2	11		22	9	21	1		304
Oceania	1992	12		5		6	13	5	1		85
South America	1991	4		10		11	3	8	2	1	280
World	1994					140	60	88	36	17	1 544

Source: Appendix 3, Table 2.
Note: Source data for industrialized countries were not classified. Number of reviewed references are sums from Table 2 in Appendix 3 and some references were therefore double counted. For industrialized countries, the reference UNECE/FAO (2000) was used.

Figure 1-3. Processes and outputs related to forest area and area change

Table 1-2. Forest area by region 2000

Region	Land area	Total forest (natural forests and forest plantations)				Natural forest	Forest plantation
	million ha	Million ha	% of land area	% of all forests	Net change 1990-2000 million ha/year	million ha	million ha
Africa	2 978	650	22	17	-5.3	642	8
Asia	3 085	548	18	14	-0.4	432	116
Europe	2 260	1 039	46	27	0.9	1 007	32
North and Central America	2 137	549	26	14	-0.6	532	18
Oceania	849	198	23	5	-0.4	194	3
South America	1 755	886	51	23	-3.7	875	10
WORLD TOTAL	13 064	3 869	30	100	-9.4	3 682	187

Note: Changes are the sums of reported changes by country.
Source: Appendix 3, Tables 3, 4 and 6.

Global maps of forest cover, ecological zones and protected areas were developed by using low-resolution satellite imagery and geographic information system (GIS) technology (see Chapter 47 and Figure 1-4). This technology was not used to estimate forest area or area change owing to limitations in the imagery. However, by overlaying country boundaries, country estimates were made for the proportional distribution of forests by ecological zone; protected forest areas; and the proportion of forests available for wood supply.

RESULTS

The tables in Appendix 3 display statistics for 213 countries and areas. In addition, the FAO Forestry Web site includes comprehensive country profiles where the results, methods and background material are presented in detail for each country (FAO 2001a). The global tables in Appendix 3 may also be downloaded from the Web site. Table 1-2 shows the distribution of forests by region and the estimated annual net change. Figure 1-5 shows the proportion of forest by country, and Figure 1-6 indicates where the net change rates are greatest.

Tables 3 and 4 in Appendix 3 display the estimates of forest area in 2000 and estimates of the change in forest area 1990-2000. Table 5 in Appendix 3 shows the distribution over global land use classes at the latest reference year. The FAO estimates are based on national reports and extrapolated to 2000. Countries validated the FAO estimates.

The FRA 2000 area and area change estimates were independent estimates; they were not based on results of earlier global assessments and were not based on models. Rather, they were estimates for each individual country based on the best available country data.

More than 600 forest types occurring in the national reports were transformed into global classes. In most cases, this reclassification was straightforward. Some classes, particularly in dry ecological zones, were difficult to transform as the results are sensitive to the method of analysis applied. Australia and Angola are examples where the forest edge towards drier woodlands was difficult to determine from national classifications. The reclassifications from national classes are documented in the FAO Forestry country profiles (FAO 2001a).

Forest plantation area (Table 1-2) was estimated separately. As plantations constitute only five percent of total forest, they are often misrepresented in national-level mapping surveys. More detailed reports on the plantation estate were analysed to provide a better picture of the resource. The separate results for plantations were incorporated in the overall national forest area estimates. One effect of this approach was that for some countries there was a discrepency between the national forest area data in which plantations were sometimes included as one mapping class, and the separate plantation area estimate (see Appendix 3, Table 6 for more detailed FRA 2000 results).

Table 1-3 shows the forest area estimates for tropical forests based on the pan-tropical remote sensing survey, including estimated standard errors (since resource limitations did not allow a 100 percent sample). Table 1-4 indicates the estimated deforestation and net change rates for the studied regions, with accompanying standard errors. It should be noted that no statistically significant difference between the periods 1980-1990 and 1990-2000 was observed for any region.

Table 1-5 compares the remote sensing survey findings with the country data obtained from national reports. There was a statistically significant difference in the estimates of forest area for each region, but this was a consistent discrepancy as the remote sensing survey showed a lower estimate for all regions. More interestingly, the change estimates from the remote sensing survey and the country data correspond well for Latin America and Asia, whereas for Africa the difference is very large. The likely reason was poor inventory information for many African countries and, as a consequence, an apparent exaggeration of deforestation for a few countries (for example, the Sudan and Zambia).

Table 1-3. Remote sensing survey: estimates of forest area by region and at pan-tropical level in 2000

Region	Forest area
	Million ha		%
	Estimate	SE	Estimate	SE
Africa	519	37	42	3
Latin America	780	49	63	4
Asia	272	23	45	4
Pan-tropical	1 571	66	51	2

Note: SE = Standard error of the mean. that the figures are related to the surveyed area, representing about 90 percent of the total forest land in the pan-tropical region. These estimates refer to the most inclusive definition of forest (f3) as defined in Chapter 46.

Table 1-4. Remote sensing survey: annual deforestation and net forest area changes during the period 1990-2000 by region and at pan-tropical level

Region	Annual deforestation million ha/year	Annual net forest area change million ha/year		Annual rate of net forest area change %/year
	Estimate	Estimate	SE	Estimate	SE
Africa	-2.3	-2.1	0.4	-0.34	0.06
Asia	-2.5	-2.3	0.6	-0.79	0.20
Latin America	-4.4	-4.2	1.1	-0.51	0.15
Pan-tropical	-9.2	-8.6	1.3	-0.52	0.08

Note: SE = Standard error of the mean.

The estimates of change in forest area are of potential significance to forest policy-makers, so they are described in more detail here. An increase or decrease in total forest area does not necessarily correspond to qualitative changes of the forest. FRA 2000 therefore attempted to identify the type of forest change - what is the new land use that replaced former forest land? Where there was an increase in forest land, what was the previous land use?

Figure 1-5. Proportion of forest by country (percent of land area)

Figure 1-6. Countries and forests with high rates of net forest area change 1990-2000

One important qualitative change is the conversion of natural forest to forest plantations. This change may have implications for biological diversity as well as the future productivity and use of the forest. Providing only one reference figure (e.g. either the overall net forest area change rate or the deforestation rate) would only give a partial picture of the forest area dynamics. It is therefore important to account for each of the change processes separately, to the extent possible.

Estimates were made for the conversion of natural forests to plantations for the major domains (tropics and non-tropics). Other qualitative changes (reforestation of forest plantations, regeneration of natural forests, forest degradation and forest improvement) are very important for forest policy development and forestry planning; however, the available statistics are not comprehensive for enough countries to make definitive estimates. The review of available studies and literature showed many statements that forests are being degraded; however, it was not possible to make objective estimates of the extent or severity of these changes for most countries because of data limitations. Qualitative changes are reported in FRA 2000 country reports and briefs, but it was not possible to derive globally valid statistics.

Table 1-5. Comparison of forest area and forest area change estimates from the remote sensing survey with country data

Region	Forest area 2000 million ha			Annual net forest area change million ha/year			Annual forest area change rate %/year
	Country data	Remote sensing survey	Significant difference	Country data	Remote- sensing survey	Significant difference	Country data	Remote- sensing survey	Significant difference
Africa	622	484	**	-5.2	-2.2	**	-0.77	-0.43	**
Asia	289	224	**	-2.4	-2.0	n.s.	-0.78	-0.84	n.s.
Latin America	892	767	*	-4.4	-4.1	n.s.	-0.45	-0.51	n.s.
Pan-tropical	1 803	1 475	***	-12.0	-8.3	**	-0.62	-0.54	n.s.

Note: Only the results from the countries included in the remote sensing survey were compiled to obtain the country data given in the table. The remote sensing estimates refer to the f2 definition of forest (see Chapter 47), that which most closely corresponds to the definition used in compiling the country data. The hypothesis tested in the table is that the country data value is the true value of the sampled population of the remote sensing survey. Level of significance of the difference between country data and remote sensing estimates: *** = 99.9 percent level of significance, ** = 99 percent level of significance, * = 95 percent level of significance, n.s. = not significant at the 95 percent level.

The FRA 2000 estimates of forest change therefore focused on the transformation to and from natural forests and forest plantations. Three separate studies were aggregated:

• area statistics of total forest area and area changes for each country;
• area statistics on forest plantation area and area changes for each country;
• results of the FRA 2000 remote sensing survey of area changes in the tropics.

These studies brought different strengths and weaknesses to synthetic analysis. Taking them together, and building on the strengths of each study, FAO developed reliable estimates for the tropics and non-tropics (Table 1-6, Figure 1-8). The steps and assumptions in this analysis are described below.

The starting point and base statistics were the estimated annual net change of forest area 1990-2000 for each country. For convenience, entire countries were assumed to lie either inside or outside the tropical domain. Tropical countries included the following subregions, as defined in this report: West, Central, East and Southern Africa (except South Africa), Central America and Mexico, Tropical South America, the Caribbean, South and Southeast Asia, and Other Oceania. The sum of net changes in the tropical and non-tropical domains appears in the last column in Table 1-6. Before this column was added to Table 1-6, however, two systematic errors in the material were considered.

First, when the results of the FRA 2000 remote sensing survey were compared with country data, a relatively good correlation was found for Asia and Latin America; but the negative area change for Africa appeared to be considerably overestimated in the national reports, possibly by as much as 3 million hectares annually for tropical Africa (-5.2 million hectares per year in the national reports and -2.2 million hectares per year in the remote sensing survey; see Table 1-5). It was known that country data were very weak for most African countries, and in some cases the reported change rate seemed very high (e.g. for the Sudan and Zambia). The discrepancy in area change estimates for Africa was secured at 99 percent confidence level, so it was necessary to make adjustments. Bearing in mind that large parts of Africa have dry forest types for which changes are not easily detected in satellite images, it is reasonable not to adopt the remote sensing survey results uncritically. It was assumed that the remote sensing survey and national report estimates were equally reliable for the region as a whole. Thus, the average of the remote sensing survey estimate and the national report estimate of forest area change to constitute a valid estimate for tropical Africa as a whole. This estimate of annual negative change for tropical Africa is about 1.5 million hectares lower than that of the national reports.

Table 1-6. Forest area changes 1990-2000 in tropical and non-tropical areas (million hectares per year)

Domain	Natural forest					Forest plantations			Total forest
	Losses			Gains	Net change	Gains		Net change	Net change
	Deforestation (to other land use)	Conversion to forest plantations	Total loss	Natural expansion		Conversion from natural forest (reforestation)	Affore-station
Tropical	-14.2	-1.0	-15.2	+1	-14.2	+1	+0.9	+1.9	-12.3
Non-tropical	-0.4	-0.5	-0.9	+2.6	+1.7	+0.5	+0.7	+1.2	+2.9
Global	-14.6	-1.5	-16.1	+3.6	-12.5	+1.5	+1.6	+3.1	-9.4

Figure 1-8. Forest area changes 1990-2000 (million hectares)

Second, the plantation establishment was considered to be exaggerated in the national reports in relation to the actual success rate for the 1990s as a whole. Based on many field studies, it was assumed that only 70 percent of the reported plantation establishment was successful. Overall, the successful plantation expansion rate was 1.4 million hectares less than what the country reports suggested. This adjustment of plantation estimates is consistent with past FAO analyses used since 1995 (FAO 1995).

It is noted that the above two calibrations were of roughly the same magnitude (-1.5 and -1.4) and as they were in different directions, the combined effect on the totals was minimal. For simplicity, it was therefore assumed that they were equal and that the net change rates in Appendix 3 could be adopted directly.

New forest plantations are established either on non-forest land (afforestation) or on land where they replace natural forest (reforestation). As these represent different change processes, it was important to quantify the proportion of each, for the tropical and non-tropical domains as a whole. This was done by expert opinion on the proportion of afforestation and reforestation for seven countries reporting major plantation establishments (80 percent of the world total) (Table 1-7). These proportions were considered valid throughout the respective domains and were extrapolated to the entire plantation establishment area. It was further assumed that no significant plantation areas were lost to other land uses or to natural forests. For tropical countries, the afforestation rate was estimated at 0.9 million hectares per year and the reforestation rate at 1 million hectares per year; for non-tropical countries the results were 0.7 and 0.5 million hectares per year respectively, (see Gains under Forest plantations in Table 1-6).

Table 1-7. Expert estimates on distribution of the reported plantation establishment over reforestation and afforestation for major plantation countries

Country	Domain	Reforestation as % of reported plantation establishment	Afforestation as % of reported plantation establishment
Argentina	Non-tropical	50	50
Brazil	Tropical	75	25
China	Non-tropical	40	60
India	Tropical	50	50
Indonesia	Tropical	90	10
Thailand	Tropical	25	75
United States	Non-tropical	0	100

Based on the above estimates for plantation change rates, the net changes of natural forest area were calculated in Table 1-6 as a net loss of 14.2 million hectares per year for the tropics and a net gain of 1.7 million hectares per year for the non-tropics.

The next step in the analysis was to distinguish between positive and negative changes within the natural forest. This was done in two different ways. For the non-tropics, the expansion of natural forests was calculated as the sum of all positive changes at the country level, an annual increase of 2.6 million hectares per year. This can be considered a very conservative estimate, as there may be local changes which are not reflected in country totals.

For the tropics, it was less relevant to use the country statistics, as the majority of countries had large negative net changes which would effectively disguise the expansion of natural forests. Instead, results from the remote sensing survey were used. The remote sensing survey indicated a total expansion of forests of 0.56 million hectares per year. It could be expected that the method underestimates expansion of forests, as this is a slow process which may be difficult to capture in remote sensing interpretation. Furthermore, the remote sensing survey had a minimum interpretation unit of 25 ha, which means that expansion of small-scale forest formations (0.5 to 25 ha) was not accounted for. In addition, a large proportion of land with (currently) low proportion of forest cover was not included in the sampled population. Finally, some tropical countries, notably Zimbabwe and Madagascar, were not included in the survey. These factors all suggested that the expansion in the tropics was higher than the estimated 0.56 million hectares per year. It was thus assumed that tropical natural forests expanded by about 1 million hectares per year (notwithstanding the substantial losses in other areas which are accounted for in the next paragraph).

Given the expansion rates for natural forests, and given that the conversion from natural forests into plantations was already known from the plantation analysis, the deforestation of natural forest could be calculated by subtracting these changes from the net change of natural forests. This gave an estimate of the global deforestation rate of about 14.6 million hectares per year, of which 14.2 million hectares per year occurred in the tropics.

These results (Table 1-6, Figure 1-8) represent the overall conclusions by FRA 2000 with respect to the change in forest area. These estimates were based on thorough analysis of three independent and original data sets (country data, forest plantation data and remote sensing survey data). An error estimate for the combined results is not possible to obtain, as the errors for national estimates are generally not known. However, building on the results of the remote sensing survey (about 15 percent sampling error on tropical forest area change estimates) and the study on reliability made by UNECE/FAO (2000), it can be concluded that the estimates have a high precision. More importantly, combining the three studies made it possible to eliminate some systematic errors in the material.

COMPARISON WITH EARLIER GLOBAL ASSESSMENTS

FRA 2000 was the first global forest assessment to use a common definition for all forests worldwide. Previous assessments used a minimum canopy cover threshold of 10 percent for developing countries and 20 percent for industrialized countries to define forests, based in part on past forest inventory practices in the two domains. When the results of FRA 1990 were reviewed, a number of experts suggested that the next global assessment should use a common forest definition for all regions. Following a consensus recommendation of the IPF, it was decided that FRA 2000 would use the 10 percent minimum canopy threshold for all countries. (That is, when observed from above, at least 10 percent of the land area is covered by forest canopy. "Other wooded land" has a canopy cover between 5 and 10 percent).

As mentioned above, the FRA 2000 area and change estimates were not based on results of earlier assessments. The data for state of forest resources from FRA 1990 were reviewed within the framework of the present assessment for purposes of comparison with the year 2000 data. To ensure comparability, the original 1990 data were adjusted, taking into consideration the following:

• availability of new national forest inventory data which improved the estimates for 1990;
• adjustment of existing 1990 data to the FRA 2000 definitions, and improved reclassification of national vegetation categories in accordance with these definitions;
• adjustment based on other new reliable data and information which had not been available in 1990;
• redefined political country boundaries.

The change in definition for non-tropical forests was the major reason that estimated global forest area for 2000 is 400 million hectares higher than the interim estimate for 1995 which used the FRA 1990 definition (FAO 1997). The effect was most significant for Australia and the Russian Federation. The estimate for Australia's forest area in 2000 was 155 million hectares, compared with 41 million hectares in 1995, in part because the 2000 estimate included large expanses of sparsely stocked forests with canopy cover between 10 and 20 percent that previously had been classified as other wooded land. For similar reasons, the estimate for the Russian Federation is 850 million hectares in 2000, compared with 764 million hectares in 1995.

Forest inventories conducted after 1990 resulted in different figures for a number of countries (including 47 developing countries) than were previously reported, and the inclusion of these results has also contributed to the higher estimate for 2000. In other countries (including 19 developing countries) a more detailed breakdown of forest classes in national inventory reports facilitated an improved reclassification of national results into FRA 2000 forest classes; the new estimates include as forest some areas previously classified as other wooded land. Further details are documented in a FRA Working Paper (in preparation) which reports on the results of an analysis of forest cover change estimates made in FRA 1990 (changes 1980-1990) and in FRA 2000 (changes 1990-2000).

It was difficult to create time series based directly on the forest area estimates in the different assessments because of variations in definitions and information quality as explained in the previous paragraphs. However, it was possible to compare the area change estimates for the 1980s and 1990s with due consideration to the effect of variations between assessment methodologies. The comparison showed that estimated net loss of forest (i.e. the balance of the loss of natural forest and the gain in forest area through afforestation and natural expansion of forest) was lower in the 1990s than in the 1980s. The net annual change in forest area was reported to be -9.4 million hectares for the 1990-2000 period (this report), -11.3 million hectares in the 1990-1995 period (FAO 1997) and -13.0 million hectares in 1980-1990 (FAO 1995b).

There is higher confidence in the 1990-2000 change estimates than in the earlier estimates. Nonetheless, if the effects of differences in definitions, methodologies and updating of national forest inventories are taken into consideration, some general conclusions can be made regarding deforestation over the past 20 years.

The change of forest definition for industrialized countries, while notably increasing global estimates of forest cover, did not greatly affect the estimated rate of change of global forest area. The change in definition had the greatest impact on the forest area of Australia and the Russian Federation, where conversions of forest to other land uses were relatively small on a global scale and thus did not significantly alter worldwide change rates. For most other industrialized countries, the revised 1990 national forest area figures (based on FRA 2000 definitions, methodologies and new data) showed a high degree of consistency and comparability with the 1990 figures of the previous two assessments. The three assessments used essentially the same definition for natural forest for developing countries. Although new estimates at the national level were not always comparable with earlier assessments, they did not significantly affect the estimates of global change rates. The new definition for plantations (which allowed the inclusion of rubber tree plantations) affected the forest area figure for a few tropical countries, but without significant effect on the world forest area change rate. It should also be noted that the three assessments used the same methodology to assess forest area change in the industrialized countries.

The findings of the FRA 2000 pan-tropical remote sensing survey supported the results of the country-based assessment. The survey indicated a net rate of change for tropical forests that was slightly lower in the 1990s than in the 1980s, but the difference was not statistically significant. The survey's findings on forest cover change in the 1980s and 1990s, which are completely compatible with one another, confirm a continued high rate of forest loss in the tropics during the 1990s. This result fits well with the results of the country assessment, as net gains in forest area are reported for the non-tropical countries as a whole while net losses are occurring in the tropics.

In conclusion, after analysis of the estimates of present and previous assessments, FRA 2000 pointed to a lower rate of net loss of forests worldwide in the 1990s than in the 1980s, owing mainly to a higher rate of natural expansion of forest area. At the same time, the worldwide loss of natural forests has continued at roughly comparable high levels over the past 20 years.

ON THE RELIABILITY OF THE FOREST AREA ESTIMATES

The nature of FRA 2000 statistics made it difficult to calculate confidence intervals for most estimates, with the exception of the remote sensing survey. At the country level very few countries, including developed countries, can derive statistically controlled confidence intervals for both forest area and forest area change. For some countries, the results were based on expert estimates using first-hand knowledge of the country but limited field data. For most countries, detailed field inventories provided reliable results, but often the results could not be compared with other inventories using comparable definitions.

For the survey of industrialized countries, UNECE/FAO (2000) addressed the precision issue in an attempt to estimate indirectly the standard error for some key variables including forest area but not area change. The main conclusion was that the precision was high, at ±3 percent "likely range" for forest area. The same general conclusion is valid for the global estimates in this report. In general, the country data quality is roughly comparable between the industrialized and developing countries. In both domains, highly reliable national forest inventories were uncommon and most country information on forest area was derived from aggregated land classifications. Furthermore, reclassification to global classes caused similar difficulties in both domains.

For the tropical domain, the remote sensing survey provided a unique possibility to calibrate for systematic errors in the country estimates. As discussed above, the extent of forests showed a good correlation. For area change, however, only two regions had a good match, and remote sensing estimates for Africa were greatly different from the aggregate country estimates. It was concluded that country data for Africa overestimate deforestation for the region as a whole, and a calibration was applied in the above global estimates (Table 1-6). Furthermore, systematic overestimations of plantation establishment were adjusted for, although this calibration was based on an expert estimate.

In conclusion, it appears that the precision at the global, tropical and non-tropical levels is good for estimates of area and area change, but that systematic errors may still distort the overall picture. Two major systematic errors were adjusted for as described above, but others may still be hidden in the material. For example, secondary forests in South America are often excluded from area change statistics and would contribute to a lower net change rate in this region if accounted for. As another example, the extent of forest plantations in Europe may be larger than reported, as the option "semi-natural" is not given in the global classification scheme.

RESULTS FROM QUALITATIVE STUDIES

The qualitative studies undertaken by FRA 2000 were extensively documented in FRA Working Papers (listed in Appendix 4 and available on the FAO Forestry Web site). It was generally not difficult to establish a good understanding of the important factors affecting land use change in a local context, where climatic, cultural, policy and economic parameters are reasonably constant. It was considerably more difficult to generalize the findings to an international level.

Although specific land use practices varied considerably among regions, a common finding was that rules that govern the right to use the land and its products tend to be correlated with the management of the land and the tendency of forest land to be converted to other uses. For example, rights to land use were often established after de facto conversion to agriculture, creating a strong incentive to encroach on forests.

Deforestation has been a popular research subject in the past decade. A survey of scientific papers (FAO 2000b) found over 1 200 published papers on tropical deforestation since 1980, of which 825 contained findings related to deforestation processes and were included in the analysis. While it is not possible to conclude that this bulk of research papers can describe all deforestation processes, nonetheless the material represents a significant input to international discussions and negotiations. It is therefore important to understand the extent to which this research is representative of the global situation.

The accumulated information on tropical deforestation studies showed in a recognizable pattern. Throughout the 1980s, as concern about deforestation grew, the number of publications increased - from 8 in 1980 to 41 in 1989. Since 1990 the rate of publications has remained relatively constant, between 45 and 60 publications per year. For purposes of discussion, the number of published papers is taken as an indicator, regardless of the area covered by the studies or the originality of their data or analysis.

Half of the studies were published since 1992. Almost one-third of the publications on tropical deforestation had no clear geographical reference point; they discussed a particular aspect of the problem in an abstract way or they undertook a global analysis of the problem. Slightly more than two-thirds of the studies had a clear geographical point of reference, but they were distributed unevenly across countries. As a generalization, easily accessed countries were represented more often than others.

The research methods also changed during the past two decades as summarized in Table 1-8. General studies drawing on secondary sources and first-hand accounts by field researchers predominate in the early publications about the problem. The methodological patterns changed from the 1980s to the 1990s. Funding for studies employing remote sensing and surveys have increased in frequency, while first-hand accounts of deforestation processes have declined in number. The number of studies based exclusively on secondary sources has also declined somewhat, although they continued to account for about 46 percent of the published research in the 1990s.

Table 1-8. Tropical deforestation studies in science journals, categorized by primary information source: trends over time

Data reference year	Remote sensing	Survey	Field observation	Secondary source	Total
	%	%	%	%	%	No. of studies
Pre-1980	8	8	39	46	100	88
1980s	8	5	30	57	100	276
1990s	17	15	20	47	100	332
Total	12	9	27	52	100	696

Source: FAO 2000b.
Note: Survey refers to household surveys and similar approaches.

The qualitative studies of forest change and deforestation carried out within the framework of FRA 2000, including detailed country studies, provided an interesting overview of the knowledge of forest change processes. The results provided useful insights for countries where studies were carried out, and some distinct geographic and temporal pan-tropical patterns emerged. However, perhaps a more important conclusion was that while many studies on changes in forest area were made over the past several decades, the studies were not well coordinated and were not necessarily representative of the global situation. It is therefore difficult for analysts to draw valid conclusions from the literature and to use existing results to develop policies that address forest change. The high proportion of studies that used secondary information indicated that the knowledge on forest dynamics may not be proportional to the number of published scientific papers.

A HISTORICAL PERSPECTIVE ON FOREST AREA AND DEFORESTATION

Clearing of forests to yield higher returns from land has a long history. Most studies estimate that about half of the Earth's land area was covered by forests 8 000 years ago, as opposed to 30 percent today (e.g. Ball 2001). Historically, deforestation has been much greater in temperate regions than in the tropics. Allowing that long-term changes in forest area are influenced by climatic fluctuations as well as by the actions of humans, the rate of deforestation since the introduction of agriculture might be estimated at about a quarter of a million hectares per year over the long term. However, much higher rates have been experienced in certain areas in the short term: for example, deforestation rates during the westward expansion in the United States in the late 1800s were roughly comparable to deforestation rates in the tropics today.

In this historical perspective it is obvious that increasing human population has been correlated with a negative impact on the extent of forests. Agriculture has expanded and replaced vast tracts of forests in all parts of the world to meet the demand for food and fibre. In some cases forests were removed primarily for wood products and the land was not reforested. Agricultural expansion has shifted between regions over time, following the general developments of civilizations, economies and increasing populations. It is still common in developing countries. The hypothesis that population growth per se drives deforestation through the demand for new agricultural land has also prevailed in many current papers and reports addressing deforestation. However, it has also been demonstrated in the United States and elsewhere in the twentieth century that population growth does not necessarily cause forest loss, especially if the rate of improvement in agricultural productivity is greater than the rate of population growth.

Table 1-9. Correlation coefficients ® between forest cover change rate and selected variables at country level

Variable	Population density	Population change rate	Population, rural proportion	GNP/caput	Forest area change rate
Population density		-0.09	0.00	0.12	-0.04
Population change rate	-0.09		0.31	-0.36	-0.26
Population, rural prop.	0.00	0.31		-0.59	-0.38
GNP/caput	0.12	-0.36	-0.59		0.21
Forest area change rate	-0.04	-0.26	-0.38	0.21

Source: Appendix 3.
Note: All data at national level and unweighted.

In recent decades, the rate of forest conversion has been particularly high in the tropics. FRA 2000 estimates tropical deforestation at 14.2 million hectares per year during 1990-2000, which means that almost 1 percent of the tropical forest is being lost each year. At the same time, the world's population has increased faster than ever before, but the direct link to deforestation and demand for agricultural land seems to have become less obvious. As the economies of most countries have grown, the relative importance of the agricultural sector has decreased. Most countries have experienced large-scale migration to cities. According to the United Nations Population Fund (UNFPA 2001), the population growth of urban areas greatly outpaces non-urban areas. Globally, only 13 percent of population growth is in rural areas, and rural populations are declining in most developed countries.

Table 1-9 shows relatively weak correlations at the national level between forest area change rate, demographic parameters and gross national product (GNP) per capita. The table suggests that the decision to abandon the population-driven deforestation model used in FRA 1990 was correct; but also that the forest change processes are too complex to be completely explained by any single indicator.

From the FRA 2000 findings, it appears that the expansion of agriculture was less prevalently associated with intensified shifting agriculture than with direct transformation of forest into permanent agriculture (or other land uses) at both large and small scales. This implies that economic and policy factors other than subsistence farming are more important in the deforestation processes.

On the positive side, many developing countries are trying to adopt policies to sustainably manage natural forests. For example, many countries are committed to monitoring progress towards sustainable forest management by national criteria and indicators. Numerous countries with substantial forest resources are trying to implement national forest programmes. This is a major development since the early 1990s and the United Nations Conference on Environment and Development (UNCED). Many industrialized countries actually experienced an increase in forest area in the 1990s, suggesting a positive link between development and the capability of a country to maintain or regain forest cover. In developed countries there is an increasing tendency for marginal lands to be valued more highly for forest goods and services than to be maintained for agriculture.

In conclusion, the changes in forest area observed for the period 1990-2000 were substantial, with a continued high rate of deforestation in the tropics. The net change, however, was lower than in the previous decade because of increased expansion of forests, primarily in non-tropical areas. The direct link to population growth and shifting cultivation earlier used to explain deforestation seems to be less valid in the most recent decade. However, demand for agricultural land remains the major driving force leading to deforestation. Factors related to land use rights seemed to determine forest area changes, as well as the general level of economic development, agricultural productivity and urbanization.

BIBLIOGRAPHY

Ball, J.B. 2001. Global forest resources; history and dynamics. The forests handbook. Oxford, Blackwell Science.

FAO. 1995a. Forest Resources Assessment 1990 - Tropical forest plantation resources. FAO Forestry Paper No. 128. Rome.

FAO. 1995b. Forest Resources Assessment 1990 - Global synthesis. FAO Forestry Paper No. 124. Rome.

FAO. 1997. State of the World's Forests 1997. Rome.

FAO. 2000a. On definitions of forest and forest change. FRA Working Paper No. 33. Rome.

FAO. 2000b. Tropical deforestation literature: geographical and historical patterns in the availability of information and the analysis of causes. FRA Working Paper No. 27. Rome.

FAO. 2001a. Forestry country profiles.
www.fao.org/forestry/fo/country/nav_world.jsp

FAO. 2001b. Forest Resources Assessment homepage.
www.fao.org/forestry/fo/fra/index.jsp

Sweden. National Board of Forestry (NBF). 2000. Statistical Yearbook of Forestry 2000. Jönköping, Sweden.
www.svo.se/statistik

Swedish University of Agricultural Sciences. 2000. Skogsdata 2000., Umeå, Sweden.

UNECE/FAO. 2000 Forest Resources of Europe, CIS, North America, Australia, Japan and New Zealand: contribution to the global Forest Resources Assessment 2000. Geneva Timber and Forest Study Papers No. 17. New York and Geneva, UN.
www.unece.org/trade/timber/fra/pdf/contents.htm

United Nations Population Fund (UNFPA). 2001. Demographic, economic and social indicators.
www.unfpa.org/swp/2000/english/indicators/ indicators2.html