# 4 Analysis of Results

This chapter contains four separate analyses concerning the effect of:

• increased Area90 from FRA 1990 to FRA 2000 on forest cover change estimates (4.1)
• changed definitions of forest area and forest area change (4.2)
• regional adjustments of forest area change (4.3)
• improved data on forest area change (4.4)

## 4.1 Effect of increased Area90 on forest cover change estimates

When comparing the Area90 from FRA 1990 and FRA 2000 the difference is about 521 million ha. At the same time the study of the difference between forest area change from FRA 1990 to FRA 2000 indicates that the rate of change has decreased by about 3.7 million ha [from - 13.1 million ha (FRA1990) to - 9.4 million ha (FRA2000)]. In other words in spite of a total 1990 net forest area reported by FRA 2000 being about 15.1 % larger than the area reported by FRA 1990, a smaller area figure for annual net forest area loss reported by FRA 2000 is smaller than the equivalent FRA 1990 figure.

In table 4-01 below, the 11 countries that account for differences in Area90FRA1990 and Area90FRA2000 in excess of 10 million ha are considered with regards to what impact the increased forest area may have had on the assessment of AreachangeFRA2000.

#### Table 4-01. Estimate of impact of increased forest area on forest area change for 11 countries accounting for the main differences between Area90FRA1990 and Area90FRA2000

 Country data Estimate of effect of change due to Country 1990 forest area Area Area Change Area change different forest area considered FRA1990(000ha)( a ) FRA2000(000ha)( b ) Difference(000ha)c = (b - a) FRA 1990(000ha)( d ) FRA 2000(000ha)( e ) Difference(000ha)f = (e - d) Estimate (000ha)g = ( f/b) x c Other Former USSR 739,729 877,886 138,157 0.0 453.0 453.0 None 0.0 Australia 39,837 157,359 117,522 0.6 -282.0 -282.6 Negative - proportional -210.6 Sudan 43,179 71,216 28,037 -473.2 -959.0 -485.8 Negative - proportional -377.5 Dem. Rep. of the Congo 113,317 140,531 27,214 -729.4 -532.0 197.4 Negative - proportional -103.0 Kenya 1,305 18,027 16,722 -5.4 -93.0 -87.6 Negative - proportional -86.3 Mozambique 17,357 31,238 13,881 -134.0 -64.0 70.0 Negative - proportional -28.4 Zimbabwe 8,981 22,239 13,258 -59.6 -320.0 -260.4 Negative - proportional -190.8 Mexico 48,695 61,511 12,816 -672.7 -631.0 41.7 Negative - proportional -131.5 USA 209,573 222113 12,540 -316.5 388.0 704.5 None 0.0 Paraguay 12,868 24,602 11,734 -402.3 -123.0 279.3 Negative - proportional -58.7 China 133,799 145,417 11,618 739.8 1,806.0 1,066.2 None 0.0 Myanmar 29,091 39,588 10,497 -381.4 -517.0 -135.6 Negative - proportional -137.1 1,397,731 1,811,727 413,996 -2,434.1 -874.0 1,560.1 -1,323.9

Together the 11 countries account for 414 million ha (or 79.5 %) of the difference between Area90FRA1990 and Area90FRA2000. The countries however only account for 1,6 million ha (or 42%) of the total difference between AreachangeFRA1990 and AreachangeFRA2000.

For the above 11 main contributors to the increased Area90FRA2000, an attempt is made at assessing the area change on the additional areas that were included as forest for FRA 2000 compared with FRA 1990.

Three countries (Former USSR (excl Belarus and Ukraine), USA and China) had positive area changes during the 1990s. For Former USSR (excl Belarus and Ukraine) and USA it is assumed that the primary cause of this increase is natural regrowth on former agricultural land/grazing land. For China plantations establishment is reported as the main contributor to the increase. It is assumed that these positive area changes have occurred on areas that were not already forested in 1990. It therefore follows, that for these three countries the net effect of the larger Area90FRA2000 compared with Area90FRA1990 on forest cover change during the 1990s is zero.

Eight countries (Australia, Sudan, Democratic Republic of the Congo, Kenya, Mozambique, Zimbabwe, Mexico, Paraguay and Myanmar) had negative area changes during the 1990s. The assumed main driving force of the negative area changes is conversion of forested areas by various forms of agriculture including shifting cultivation and grazing. For the current comparison it is assumed that the negative change is distributed evenly on the national forest areas. The increase in Area90 from FRA 1990 to FRA 2000 for above mentioned 8 countries thereby contribute with a total of - 1,323,900 ha to the reported total annual forest cover change during the 1990s (column g in table 4-01).

If only the forest area of Area90FRA1990 was considered for the comparison of forest area change estimates, the global figures for AreachangeFRA2000 should therefore be adjusted upwards as some of the negative area changes reported for FRA 2000 occur on areas that were not registered as forest for FRA 1990. This would mean a lower annual global deforestation rate during the 1990s.

Vice versa using the above arguments in reverse, if AreaFRA1990 was derived using FRA 2000 terms and definitions, a negative adjustment of AreachangeFRA1990 would result.

In both cases the gap between AreachangeFRA1990 and AreachangeFRA2000 would increase indicating a larger positive development in annual forest area change from the 1980s to the 1990s if the same global forest areas were used as base for the estimates of forest cover change rates for the 1980s and the 1990s.

The above comparison has been based on the assumption that the negative area changes during the 1990s occur on areas defined as forest in 1990 and that the positive area changes during the 1990s occur on that were not defined as forests in 1990. The increase in Area90 by 521 million ha from FRA 1990 to FRA 2000 (mainly due to a broader definition of forest) will therefore result in a larger area subject to negative area changes during the 1990s. The initial forest area will not affect the positive area changes during the 1990s as they occur on non-forest areas.

The reported global deforestation rates depend on the total forest areas compared. The current estimate of an apparent slowdown in the rate of net forest area loss from the 1980s to the 1990s by 3.7 million ha would appear to be a conservative one, as it is based on different global forest definitions and thereby different forest areas from FRA 1990 and FRA 2000.

## 4.2 Effect of changed definitions on Forest Area and Forest Area Change

The 521 million ha increase in net forest area estimate from FRA 1990 to FRA 2000 is mainly due to change of FRA definitions with the largest impact on the reported areas of Other Former USSR and Australia. The change of definitions did not have a similar effect on differences in areas of annual net forest area change.

The change in definitions from FRA 1990 to FRA 2000 can have both a positive and a negative effect on the figure for net annual forest area change for the below main reasons.

_ Rubber plantations are considered forest plantations for FRA 2000. Rubber plantations established during the 1990's will be included as positive figures for AreaChangeFRA2000 while for FRA 1990 they were defined as non-forest plantations. Rubber plantations that have been converted to other land use during the 1990's will be included as negative figures in AreaChangeFRA2000. According to FAO (2001b) the average annual increase in the global area of rubber plantations between 1990 and 1997 was approximately 58,000 ha, which only explains a very small part (1.5%) of the 3.7 million ha difference between AreaChangeFRA2000 and AreaChangeFRA1990.

_ Due to change of definitions between the assessments, small scale forest tree plantations or natural regeneration on small areas established during the 1990's (between 0,5 ha and 100 ha) should be included in AreaChangeFRA2000 but the establishment of such plantations during the 1980's are not included in AreaChangeFRA1990. Vice versa conversion to other land use of small areas will figure as a negative net forest area change for FRA 2000 but not FRA 1990. As the area data of many of the country reports are based on remote sensing data with varying resolution, the degree to which small area changes were actually registered and reported for FRA 2000 is likely to be variable. The current data does not allow for evaluation of the effect of this change of definitions on AreaChangeFRA2000.

_ Changes in crown density and height criteria for forests in industrialized countries for FRA 2000 can in theory have a negative or a positive impact on the difference between AreaChangeFRA2000 and AreaChangeFRA1990. However the figures for AreaChangeFRA1990 for Industrialized countries used for the current working paper, include both forests and other wooded land (table 4 of FAO Forestry Paper 124). For the current comparison the effect on the net annual forest area change figures due to the change of stand density and height criteria between the two assessments is therefore not known.

Summing up FRA 2000 reports a larger global forest area than FRA 1990 primarily due to change of FRA definitions of forest in the industrialized countries between the two assessments.

The collective effect of change of definitions on annual net forest area change from FRA 1990 to FRA 2000 can have both positive and negative effects. The net effect is unknown but assumed to be marginal for the above string of reasons. With the exception of a small annual increase in area of rubber plantations it is however not possible to estimate the extent of the influence of changed definitions.

## 4.3 Effect of Adjustments on Forest Area Change.

In the analysis of the global figures for forest area and area change in Chapter 1 of the FRA 2000 Main Report FRA 2000 two major adjustments of the aggregated country data are made.

1. FRA 2000 assumes that on a global level, 70% of the reported new plantation establishments are successful. This implies that of the 4,457,900 ha reported established annually, 3,120,500 ha are assumed to be successful. The remaining 1,337,400 ha are considered as non-forest areas. The estimated 3.1 million ha established successfully annually is further broken down by FRA 2000 as follows: Tropical countries (afforestation 0.9 million ha, reforestation 1 million ha) Non-tropical countries (afforestation 0.7 million ha, reforestation 0.5 million ha). This break down is based on expert opinion of the proportion of afforestation and reforestation for the major plantation countries and by using the same relative figures throughout the tropical and the non tropical domains and assuming that no plantations were lost to other landuses or to natural forest.

2. Comparing the results of FRA 2000 remote sensing survey with the results of the FRA 2000 country data, a relatively good correlation was found for the change figures for Asia and Latin America. The large annual negative net forest area change according to country data for Africa however appears to be overestimated. The country data for Africa indicates an annual negative net forest area change of 5,262,000 ha (refer to table 3.1 and appendix II). The FRA 2000 remote sensing survey indicates an annual negative change in the range of 2,2 million ha/year, leading to possible overestimation of the negative net annual area change for Africa of up to 3 million ha. Some African countries have very weak country data and report very large negative net annual area changes e.g. Sudan and Zambia. The discrepancy in area change estimates between the country data for Africa and the findings of the FRA 2000 RSS for Africa was established at a 99% confidence level - making it necessary to adjust the figures. As large parts of the African forest resource is composed of dry forest, which is not readily detected on satellite imagery, the FRA 2000 RSS results were not adopted uncritically. For the adjustment it was assumed that the results of the national reports and the remote sensing survey were approximately equally reliable. With this assumption, the annual forest area change for Africa was approximately 3,7 million ha in the 1990s.

The two calibrations mentioned above have roughly the same magnitude, but as they were in opposite directions, the combined effect on the totals is minimal. For simplicity it is assumed that they are equal in magnitude but in opposite directions. The global net change rates as stated by summing the country data have therefore been adopted directly for FRA 2000. The adjustments however do have implications for the forest area change figures at regional level.

Table 4-02 below presents the adjustments in figures and figure 4-01 provides a graphic representation of the implications of the adjustments on the regional forest area change figures from FRA 1990 and FRA 2000.

With the above adjustments, the difference between the annual forest area changes during the 1980s and the 1990s decreases dramatically for Africa (due to adjustments for the apparent overestimation of deforestation as reported by the country data) and Asia (due to adjustments for plantation survival). For the other regions the difference is much less pronounced.

#### Table 4-02 Regional adjustment of country data due to plantations success rate and calibration due to Remote Sensing Survey for Africa

 Region (a)FRA 1990Forest areaannual change1980-1990(000 ha) (b)FRA 2000Forest areaannual change1990-2000(000 ha) (c = b-a)Diff, annualchangecountry data(000 ha) ( d )FRA 2000Annual plantationestablishment(000 ha) Adjustments (h = g+c)Adjusted Diff, annualchangecountry data(000 ha) (i=b+g)Adjusted annual net forest areachange1990-2000(000 ha) (e = 0.7 * d)Adjustment for assumed plantation successrate of 70%(000 ha) (f)RSSadjustment for Africa(000 ha) (g = e + f)Total Adjustment ofAnnualChange(000 ha) Africa -4,016 -5,262 -1,246 193.9 -58.2 1,337.4 1,279 34 -3,983 Asia -1,527 -650 877 3,500.2 -1,050.0 -1,050 -174 -1,700 Oceania -125 -364 -239 15.4 -4.6 -5 -243 -369 Europe 239 1,167 928 5.0 -1.5 -2 927 1,166 N + C America -1,491 -570 921 234.0 -70.2 -70 851 -640 South America -6,185 -3,711 2,474 509.4 -152.8 -153 2,321 -3,864 Total -13,105 -9,390 3,715 4,457.9 -1,337.4 1,337.4 0 3,715 -9,390

The adjustment of regional data is illustrated in figure 4-01, where the annual net forest area change during the 1980's according to FRA 1990 (column a in table 4-02) is plotted against the unadjusted and the adjusted annual net forest area change figures for the 1990's according to FRA 2000 (column b and i of table 4-02 respectively).

Figure 4-01: Regional adjustment of country data due to plantations success rate and calibration due to Remote Sensing Survey for Africa

The arrows in figure 4-01 indicate the mentioned adjustments for Africa and Asia. With these adjustments the annual net forest area change figures for Africa, Oceania and Asia assume almost unchanged rates of forest area decrease from the 1980 to the 1990's. For Europe (including Other Former USSR), North America (excl. Mexico) and most significantly for South America there is a positive development in forest area change rate from the 1980's to the 1990's.

With the performed adjustments to the regional data, the global difference of 3.7 million ha between the AreaChangeFRA1990 and AreaChangeFRA2000 owe almost entirely to the differences in the reported net forest area changes reported for Europe, North & Central America and in particular to South America. These regions are represented by the points for adjusted forest area change in figure 4-01 that are plotted furthest from the 45o line which indicates no difference between the change assessments of FRA 1990 to FRA 2000.

The figures from the RSS for forest cover change during the 1980s and the 1990s have been included in figure 4-01. The RSS figures apply only for the tropical domains of the three regions Africa, Asia and Latin America and are thereby not directly comparable with the regional figures based on country data. The proximity of the RSS points to the 45o diagonal is indicative of the fact that the FRA 2000 RSS did not find any statistically significant differences for the tropical regions between the forest area change rates from FRA 1990 to FRA 2000.

The inclusion of the RSS data points also illustrates a large difference between the country estimates and the RSS estimates for forest area change from FRA 1990 for Latin America and Africa as well as the above mentioned large difference between the FRA 2000 estimates for forest area change for Africa.

FRA 2000 estimates that globally -1.5 million ha of natural forest are converted annually to forest plantations. Along with an estimate of the global deforestation (to other land use) of -14.6 million ha, this leads to FRA 2000 arriving at an annual total loss of natural forest of -16.1 million ha (of which -15.2 million ha were in the tropical regions). An estimated natural expansion of 3.6 million ha (mainly in the non-tropical regions) reduces the net change in natural forests to -12.5 million ha (made up of a 1.7 million ha increase in the non-tropical regions and a decrease of -14.2 million ha in the tropical regions). Adding the adjusted figure for annual establishment of forest plantations of 3.1 million ha gives the total net annual forest area change of -9.4 million ha, the total for AreaChangeFRA2000.

FRA 2000 went to a greater level of detail than FRA 1990 in providing the figures broken down by process for forest area change and by tropical/non tropical domain. A stringent comparison between FRA 1990 and FRA 2000 is therefore not performed at this level. The FRA 2000 figures indicate that the total for AreaChangeFRA2000 -9.4 million ha is composed of a net deforestation rate of -14.6 million ha per year and a net annual expansion of 5.2 million ha per year. In addition FRA 2000 estimates an annual conversion from natural forest to plantations of 1.5 million ha. FRA 2000 indicates that AreaChangeFRA2000 was slightly lower than AreaChangeFRA1990 primarily due to a higher estimated rate of forest expansion during the 1990s. The FRA 2000 RSS indicated that direct conversions to permanent agriculture were more prominent than shifting agriculture in forest change processes.

## 4.4 Effect of Improved Data on Forest Area Change Estimates

There is higher confidence in the 1990-2000 forest area change estimates than earlier assessments. The 1990-2000 forest area change estimates are based on better documented information than earlier assessments and the global average reference year is 1994. FRA 1990 forest area change estimates and forest area 1990 were for the tropical countries generated by a forest area adjustment function using the at the time best available inventory baseline data which on average was referenced around 1980. FRA 2000 placed major emphasis on documenting the baseline data contained in the country reports. With reference to tables 4-04 and 4-05 below, the timeliness and the comparability of the baseline data has improved from FRA 1990 to FRA 2000. In spite of this improvement, fewer than half of all countries had time series information with high comparability between the observations and a high proportion of the developing countries relied on expert opinion for FRA 2000 area estimates.

The improvement of the baseline data between FRA 1990 and FRA 2000 has had the following two main effects on the forest area change assessments:

_ For Industrialized countries that lacked information on annual forest area change for FRA 1990 (thus registered as AreaChangeFRA1990 = 0 ha) the information was largely supplied for AreaChangeFRA2000.

_ For Developing countries AreaChangeFRA1990 was generated using a forest area adjustment function - for FRA 2000 the estimation of forest area change was based on time series, secondary source information or expert opinion.

The two are discussed in section 4.4.1 and 4.4.2 below

### 4.4.1 Non availability of AreaChangeFRA1990.

A number of industrialized countries reported AreaChangeFRA1990 as n.a. These countries are indicated in table 4-03

#### Table 4-03 Industrialized countries that reported AreaChangeFRA1990 as n.a.

 Country AreaChangeFRA 1990 Norway 31,000 Iceland 1,000 Israel 5,000 Italy 30,000 Other former USSR* 452,000 New Zealand 39,000 Total 558,000

* Defined below table 2-01

A country that should also be mentioned in this section is Australia. For Australia AreaChangeFRA1990 is reported as +600 ha (which includes forest and other wooded land) while for FRA 2000 it is reported as -282,000 ha. It is assumed that the change of definitions between the assessments does not fully explain the difference of - 282,600 ha between the FRA 1990 and FRA 2000 figures and that part of the explanation is improved country data (referenced 1992) becoming available between the assessments. AreaChangeFRA1990 for Australia is therefore for practical purposes in this discussion assumed to be n.a.

Forest area change data becoming available for the above countries from FRA 1990 to FRA 2000 may thus account for about 275,400 ha (or about 7.5 %) of the 3.7 million ha difference between the change assessments.

For Canada a non-significant (n.s.) annual forest cover change during the 1980s and the 1990s was reported for FRA 1990 and FRA 2000 respectively. According to the background information provided by UNECE/FAO 2000 the n.s. figure for FRA 2000 rests on an assumption that the national forest land base remains constant and that changes due to fires, pests harvesting etc, will be reflected in the ageclass distribution. UNECE 2000 states that change estimates were not possible to make for FRA 2000.

### 4.4.2 AreaChangeFRA1990 generated by Deforestation Model

FRA 1990 (FAO 1993 and FAO 1995b) summarizes the state of forest inventories of the developing countries as far as areas and change assessment as follows:

• Considerable variation with respect to completeness and quality of the information (for the tropics FRA 1990 data was best for Asia, followed by tropical America and finally Africa).
• There was considerable variation in the timeliness of the information, with data on average being 10 years old, thereby providing a potential source of bias.
• 21 tropical and 4 non-tropical countries, primarily in Asia & Pacific and South America had carried out more than one assessment, however the proper Continuous Forest Inventory (CFI) design had not been used for the change assessments.
• All countries (except 6 in Africa and 1 in Asia) possessed one estimate of forest area, and most tropical estimates were based on remote sensing. For non-tropical countries most estimates were based on expert estimates or official national statistics.
• Only few countries had reliable plantations estimates
• No countries had carried out a national forest inventory containing information that could be used to generate reliable estimates of woody biomass volume and change.

#### Table 4-04 State of Forest Inventory in the tropics at end 1990 (FRA 1990).

 Region Number of countries under assessment Number of countries with forest resources data at national level Forest Area information(number of assessments and reference years) Other topics covered No assess-ments One assessment More than one assessment Forest con-servation and management Forest Planta-tions Volume and biomass Forest harvesting and utilization before 1981 1981-90 Africa Tropical 40 3 23 12 2 4 8 2 4 Non-tropical 8 3 0 5 0 8 1 48 6 23 17 2 4 16 3 4 Asia & Pacific Tropical 17 0 1 6 10 10 8 7 7 Non-tropical 17 1 10 2 4 17 5 34 1 11 8 14 10 25 12 7 L. America & Caribbean Tropical 33 0 15 9 9 11 8 9 4 Non-tropical 3 0 1 2 0 3 0 36 0 16 11 9 11 11 9 4 Total Tropical 90 3 39 27 21 25 24 18 15 Total Non-tropical 28 4 11 9 4 28 6 Total 118 7 50 36 25 25 52 24 15

Source FAO 1993 and FAO 1995b

#### Table 4-05 Forest area information availability and quality by region (FRA 2000).

 Region Reference year for latest available area data (area weighted) No. of countries 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 Asia 1995 14 6 9 1 28 14 11 14 3 Europe 1997 3 38 32 6 N & C America 1995 15 2 11 22 9 21 1 Oceania 1992 12 5 6 13 5 1 South America 1991 4 10 11 3 8 2 1 World 1994 140 60 88 36 17

Source FAO 2001a

To address the above shortcomings and to bring the reported national forest area data to standard reporting years (1980 and 1990), a deforestation model (also called a forest area adjustment function) was developed for the developing countries. The model is described further in FAO Forestry Paper 112. The model was developed to describe deforestation as a location specific process driven by among others population pressure and environmental conditions. The model correlates forest area change in time with other variables including population density and population growth for the corresponding period, initial forest area and the ecological zone under consideration.

#### Deforestation Model Applied for DevelopingCountries FRA1990

The model expressed in the form of a differential equation has the form:

____________________

dY / dP = b1 * Yb2 - b3 * Y

____________________

where:

_ Y is the percentage of non-forested area in a subnational unit computed as:
Y = 100*(Total area - Forest Cover Area)/(Total Area)

_ P = ln(1+ Population density) with population density expressed in persons per square kilometer (ln 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, generating a relatively slow increase in deforestation during initial stages, faster rates during intermediate stages and a slow down during final stages. The model generated adjusted area data of the subnational units to the standard reference year 1990 and estimates of forest area change over the period 1981 to 1990. The most recent forest inventory data was used as baseline and forest area in 1980 and 1990 (standardized results) were computed according to one of the below 3 options:

_ Where reliable multi-date inventories were available. In this optimal case, the existing multidate information was used to calibrate the general model with the `local' b1 parameter and the resulting model was used to compute the standardized results.

_ Where reliable single data inventory was available standardized results were computed using the general model.

_ Where no reliable inventory was available, estimates of baseline forest area were extracted from calibrated vegetation maps and these were then used as input to the modelling procedure described in the above bullet.

Using the deforestation model, the figures for AreaChangeFRA1990 for the tropical countries were generated.

The deforestation model was not used for the assessment of AreaChangeFRA2000, as it was acknowledged that the correlation between demographic parameters and forest area change rates at national level was weak. With reference to table 4-05, time series of varying comparability were used for 140 reporting countries to generate AreaChangeFRA2000. However less than half the countries investigated had time series with high comparability between the observations. For countries that lacked time series information, AreaChangeFRA2000 was estimated using secondary information and/or expert opinion. Adjustment to the standard year 2000 was done, usually by means of linear extrapolation unless the data indicated other trends.

Although far from optimal, the baseline data used for estimating AreaChangeFRA2000 still represents an improvement from the data used for estimating AreaChangeFRA1990. As a result there is higher confidence in the 1990-2000 forest area change estimates than earlier assessments.

Stringent comparison of the forest area change data from the two assessment is however highly uncertain as AreaChangeFRA1990 and AreaChangeFRA2000 refer to different periods and are based on different methodologies (in the case of developing countries) and definitions (in the case of industrialized countries).

The change in methodology may in part explain the positive difference of 3.7 million ha in annual forest area change rates from FRA 1990 to FRA 2000. The model may for some countries have generated high estimates of net annual forest area loss during the 1980's. Distinguishing between how much of the 3.7 million ha difference between the change rates of the two assessments is due to a change of methodology, how much owes to other causes and how much reflects real changes is however not possible within the scope of the current working paper and with a reasonable degree of accuracy given the data of FRA 1990 and FRA 2000.