The Great Restoration: The Potentials for Forest
Protection to 2050
Pilot study 1
Forest Cover Trends and Inventories
PRELIMINARY
DRAFT; DO NOT CITE OR QUOTE
Authors
Pekka Kauppi
(University of Helsinki, Department of Limnology and Environmental Protection,
pekka.kauppi@helsinki.fi)
Jari Liski
(European Forest Institute, Jari.Liski@efi.fi)
Ranga Myneni
(Boston University, rmyneni@crsa.bu.edu)
1 Introduction
We focus on the conversion of forest to non-forest, from non-forest to forest, and discuss long term changes of forest biomass. As increasing populations first develop, land has historically been cleared. However, as populations start stabilizing and get wealthier, as in the developed countries, cropland and other non-forest can be spared, and forests can expand (Waggoner 1994). Changes occurr also in forest practices.
Deforestation in rural societies often proceed in a way that the largest and most valuable trees are removed first. As the timber stock degrades, middle sized and small trees are harvested in the second phase. In this way , the tree cover gradually declines and finally disappears and the area is converted into non-forest. The conversion is rarely abrupt. Subsequent to an introduction of slash-and-burn cultivation, second-growth forest may occur. Permanet conversion to non-forest can require several rotations of slash-and-burn. Deforestation on a given location can last decades, in some cases even centuries.
The reversal of forest that is, conversion from non-forest to forest, may also take a long time. It takes decades for trees to grow. Only part of the reversal is a result of tree planting. Natural expansion is slower than plant introduction. Succecional processes are involved both in natural expansion and planting. A stand which has recently been planted on an open land is quite different from an old-growth forest even though the canopy can be fully closed. For wildlife or endangered species, such a yong stand may more resemble an open land than a true high forest.
With economic growth, the costs of labor increase and harvest oprations may turn expensive. Following examples refer to central and northern Europe. It was common in the middle of the 20th century , right after the Second World War, to collect wood from the forest in a practice called ‘thinning’. Small, suppressed trees were harvested while large dominating trees were left growing. This practice is still in use but has partly become uneconomic with increasing labor costs. In Europe in the early 20th century, cattle was grazing in the forest under the trees – a sort of ‘agroforestry’. With the phase-out of such traditional practices, forest biomass has tended to increase even if the area has remained the same. Changes can occur in forest biomass and structure even in the absence of a conversion of land cover.
An intervention which often has been neglected in forest analyses is the size selection of the trees utilized. In the deforestation phase, largest trees are removed first. Biomass may decline dramatically with no change in forest area. When large trees are no longer available, loggers hit the small trees. Eventually, the forest is lost. Conversely, in the initial phases of the reversal, small trees are left unharvested first. The harvest economy favors the utilization of the larger trees. Small trees can populate abondoned fields, forest openings, and the understorey of the existing high forsts. Eventually, in the long term, even the largest trees may no longer be needed from all the land previously used for timber production.
We conclude that the conversion of forest to non-forest, and back, is a multi-faceted process, not merely a geographic shift of the line between forest and non-forest. Young stand with low biomass is quite different from an old stand with high biomass. Biomass can change over time in a forest area, even though the forest area remains unchanged. It is interesting to analyse changes in both forest area and forest biomass.
Conclusion: We are interested in an analysis where both forest area and forest biomass are used as forest indicator. The issue is: How can we best monitor forest area and forest biomass at a global level. The FAO has done excellent work in this field for several decades. We examine the latest FAO data in order to gain insight as to the issues of deforestation and Great Reversal. We also analyse the different concepts used by ecologigists and forest inventory specialists. We also discuss new research possibilities for the future in 2000-2005. Our objective is to obtain guidelines for future scientific work regarding the monitoring of the global forest resources.
2 Forest area
In this section, we
report 1) two estimates for present forest area, one based on forest
inventories as reported in forest resource information collected together and
published by FAO and the other based on remote sensing and 2) recent trends in
the forest area based on the forest resource information. We begin by
introducing the concepts and definitions of forest and other wooded land in the
two approaches.
2.1 Definitions
of forest and other wooded land
a) FAO sources
In Forest Resources
Assessment 1990, FRA-1990, the definition of forest for temperate and boreal
forests was different from that for tropical forests (Table 1). The basic
difference was the requirement of crown cover, 20% for temperate and boreal
forests and 10% for tropical forests.
For Forest Resources Assessment
2000, FRA-2000, the definitions have been changed. The crown cover requirement
for temperate and boreal forests has beed decreased to 10%.
In addition to forest, FAO
statistics also account for other wooded land. It is more open area that does
not meet the requirements of forest (Table 1). Like the definition of forest,
the definition of other wooded land for temperate and boreal regions was
different from that for tropical regions, and the definitions have been changed
for the more recent FRA-2000.
b) concept in
remote sensing
Global Land Cover
Classification Approaches
Conventional
Approaches
Because of the
diversity of vegetation at a global scale, the accurate mapping and
representation of terrestrial vegetation has been a challenge for many years.
The compilation of reliable databases at global scales involves both the
generalization of vegetation types into a smaller set of critical attributes
and the development of means for measuring vegetation globally in a meaningful
timespan [Running et al. 1995].
Current global climate models,
however, rely on land-cover data sets which are typically derived from
pre-existing maps and atlases [Olson and Watts 1982; Matthews 1983; Wilson and
Henderson-Sellers 1985; Prentice et al. 1992]. This approach has a number of
limitations regarding model parameterization. First, the reference sources
themselves often represent a range of different scales, dates and
classification schemes, and the translation of mapping units into the
classification system and scale of interest may introduce significant new
errors. Second, some datasets are derived from maps of potential vegetation,
which is usually inferred from climate variables rather than the actual
vegetation type. A third limitation is that many datasets are static and are
therefore prone to the perpetuation of errors in the source from which they
were derived [Loveland et al. 1991; DeFries et al. 1995].
A good illustration of the problems
presented in this regard is given by Townshend et al. [1991], who compared
existing maps of global vegetation and showed that the estimates of vegetation
distribution from common sources varied consider ably. The lack of consistency
among the various map sources was attributed to both the vegetation
classification and resolutions used in spatial sampling. While such databases
have obvious limitations, they represent the state of the science for driving
large scale process models.
Remote Sensing-Based
Approaches
There is wide
consensus that remotely sensed data can provide an accurate and repeatable
means of land cover mapping and monitoring, especially with respect to areas
with rapidly changing landuse and land management activities [Running et al.
1994; Townshend et al. 1991]. In particular, remote sensing based approaches
make use of the distinct spectral reflectances from different land cover types
in association with the temporal variation of reflected radiation caused by the
phenological dynamics in vegetation [Loveland et al. 1991; Justice et al.
1985].
Most recent research on global land
cover classification has used satellite data collected by the Advanced Very
High Resolution Radiometer (AVHRR) instrument on board the National Oceanic and
Atmospheric Administration (NOAA) series of satellites [Justice et al. 1985;
Running et al. 1994]. The high temporal resolution of AVHRR data is desirable
for global land cover classification and allows repeated unobscured views on
land surface features [Townshend and Tucker 1984]. In order to reduce data
volumes, 10-day or monthly composited NDVI is commonly used as input to
classification algorithms [Holben 1986].
Surface temperature from NOAA/AVHRR,
used in conjunction with spectral vegetation indices (SVI), have been found to
be useful for the description and quantification of energy exchange processes
and absorption by plant canopies [Goward et al. 1994]. Satellite-derived land
surface temperatures are a function of the proportion of soil versus vegetation
in a pixel as
well as surface
wetness. Nemani et al. [1993] showed that under dry surface conditions, surface
temperature is linearly correlated with canopy density across different
vegetation types, whereas this relation is poorly defined over wet surfaces.
Furthermore, radiometric temperatures from space-borne sensors are complex
function of viewing geometry and illumination [Choudhury 1991].
Using AVHRR data, Loveland et al.
[1995] developed a land cover database using an unsupervised classification
algorithm in conjunction with extensive ancillary data. The unsupervised
classification yielded spectrally similar clusters of vegetation. Ancillary
data was then used to label those clusters. The final classification included
205 classes for North America, which may be collapsed into fewer and broader
set of classes in a straightforward manner. However, their algorithm involves
significant amounts of ancillary data and requires substantial manual post
processing.
Most current classification schemes
designed for application at continental to global scales are based on the
magnitude and temporal dynamic of spectral vegetation indices such as NDVI
[Justice et al. 1985; Loveland et al. 1991; Loveland et al. 1995; DeFries and
Townshend 1994]. More recently, Nemani and Running [1997] have demonstrated the
potential of a combination of both spectral vegetation indices (SVI) and
surface temperature observations. Their methodology is based on known energy
exchange processes rather than statistical associations of vegetation types and
spectral properties.
The use of additional information in
the training process, such as thermal bands or seasonal metrics has also been
suggested by DeFries et al. [1998]. The use of additional phenological metrics
provide little improvement in classification accuracy relative to using an
annual time series of NDVI data. Also, the use of geographic location as an
input feature yields substantially better accuracies than using only NDVI.
However, this does not reflect the true accuracies and can be explained by
interaction between the decision tree algorithm and the bias introduced by the
geographic distribution of training data.
Although the approaches described
above provide promising results, it must be noted that AVHRR data is limited in
several regards including a high level of atmospheric noise (especially in
channel 2), lack of onboard calibration, and only five spectral bands [Zhu and
Yang 1996; Cihlar et al. 1997; Moody and Strahler 1994]. As a consequence,
AVHRR data is insufficient to discriminate subtle differences among many
vegetation types. The MODIS instrument is expected to overcome these
limitations for global land cover classification. Specifically, it will provide
superior spectral and spatial resolution as well as better facilities for
atmospheric correction and instrument calibration. The specific properties of
the MODIS instrument are documented in Running et al. [1994].
Biome-Based
Classification
Climate and
biogeochemical models require accurate input and data on land cover [DeFries et
al. 1995]. For example, Running and Hunt [1993] introduced an ecosystem model
(BIOME-BGC) designed to capture the essential physio-morphological factors that
regulate energy exchange processes in vegetation. Within Biome-BGC, global
vegetation is represented by six different biome classes. The ecological
foundation for this classification approach was given in Running et al. [1995]
and the classification is based on three primary attributes of plant canopy
structure: (i) permanence of above ground biomass, (ii) leaf longevity and
(iii) leaf type or shape.
The first attribute, aboveground
biomass, discriminates between permanent respiring biomass, such as forests and
woody shrubs, and annual crops and grasses. It is an important determinant of
carbon cycles and is controlled primarily by climate. Leaf longevity, on the
other hand, separates evergreen from deciduous canopies and plays a major role
in carbon and energy exchange processes. Finally, the leaf type criteria
distinguishes broadleaf and needleleaf plants as well as grasses. It also
determines the radiation and gas exchange characteristics of canopies.
The combination of these three
criteria yields the following six biome classes: (1) evergreen needleleaf, (2)
evergreen broadleaf, (3) deciduous needleleaf, (4) deciduous broadleaf, (5)
broadleaf annual and (6) grasses. This classification scheme has three
advantages over earlier classification efforts. First, it uses only plant
attributes, therefore other variables, such as climate, are excluded from the
class definition. Second, it is tailored to the information content of remotely
sensed observations. Most importantly, it provides a relatively stable and
unambiguous classification scheme for the purpose of global biogeochemical
modeling [Nemani and Running 1997].
Nemani and Running [1997]
implemented this logic using a hierarchical classification structure based on
different thresholds for NDVI, surface temperature and their seasonality. However,
the choice of thresholds is somewhat arbitrary and estimation of the accuracy
and performance of this algorithm can only be done using pre-existing land
cover maps. A somewhat similar biome classification scheme based on canopy
architecture will be described in the next section in the context of radiative
transfer modeling of vegetation canopies.
Radiative Transfer
Modeling of Vegetation Canopies
Canopy radiative
transfer models (RTM) simulate radiation absorption and scattering in
vegetation canopies. A review of canopy radiative transfer models can be found
in Myneni et al. [1995]. Myneni et al. [1997] suggested an algorithm for the
estimation of LAI and FAPAR at a global scale using such models. A synergistic
algorithm for the estimation of vegetation
canopy LAI and FAPAR from MODIS and MISR data is described in Knyazikhin
et al. [1998].
The relationship between NDVI and
LAI/FAPAR has been established theoretically. However, the utility of this
relationship depends on the sensitivity of these variables to canopy
characteristics [Myneni et al. 1997]. While FAPAR exhibits a positive linear
relationship with increasing NDVI, LAI is curvi-linearly related and shows
saturation with increasing
NDVI (Figure 1). In order to estimate
LAI/FAPAR from remotely sensed data, canopy structural types must be defined
that exhibit different NDVI-LAI or FAPAR relations from one another. If the
canopy types have similar NDVI-LAI/FAPAR relations, information on land cover
is redundant for the estimation of LAI/FAPAR. Therefore many classification
schemes, which are based on ecological, botanical or functional metrics are not
necessarily suitable for LAI/FAPAR estimation.
The planned algorithm for the
retrieval of LAI and FAPAR from MODIS/MISR data is based on six distinct plant
structural types (biomes), which can be parameterized with variables that many
radiative transfer models employ [Knyazikhin et al. 1998].
This implies that a land cover
classification scheme that is compatible with radiative transfer and LAI/FAPAR
algorithms is needed. Myneni et al. [1997] define the following six biomes
based on their canopy structure, which invoke different radiative transfer
models to estimate LAI/FAPAR from remote sensing data.
Grasses and Cereal Crops (Biome 1):
This land cover type is characterized by vertical and lateral homogeneity, full
ground cover and plant height less than about a meter. The plants have erect
leaf inclination, no woody material, minimal leaf clumping and intermediate
soil brightness. Shrubs (Biome 2):
Unlike biome 1, canopies are laterally heterogeneous and show sparse to
intermediate vegetation ground cover (20-60 percent). The plants have small
leaves, woody material, and bright backgrounds. This land cover type is
typically found in semi-arid regions with extreme temperature regimes and poor
soils.
Broadleaf Crops (Biome 3): These
canopies are laterally heterogeneous and exhibit large variations in vegetation
ground cover, ranging from about 10 percent after planting to 100 percent at
full maturity. They are characterized by regular leaf spatial dispersion, a
high level of
photosynthetic
activity in both leaves and stems, and dark background soil.
Savannas (Biome 4): Savanna canopies
have two distinct vertical layers, an understory of grass (biome 1) and an
overstory of trees with about 20 percent ground cover. Savannas in the tropical
and sub-tropical regions are described as mixtures of broadleaf trees and warm
grasses, whereas in the cooler regimes of higher latitudes, they are
characterized as mixtures of cool grasses and needleleaf trees.
Broadleaf Forests (Biome 5):
Broadleaf forests are characterized by both vertical and horizontal
heterogeneity, i.e. high ground cover, green understory, mutual crown shadowing
and foliage clumping. Trunks and branches are included in the radiative
transfer models, which means that canopy structure and optical properties
differ spatially. Trunks are modeled as
erect structures and branches as randomly oriented.
Needleleaf Forests (Biome 6):
Needleleaf forests represent the most complex canopy structure. They are
characterized by needle clumping on shoots, shoot clumping in whorls, dark
vertical trunks, sparse green understory and mutual crown shad owing. Branches
are modeled as randomly oriented and trunks as erect structures. Needles are
assumed to be clumped in the shoots, and the shoots clumped in the crown space.
The definitions and properties of
the six biomes as they relate to radiative transfer are shown in table ?.
Tree-Based
Classification Algorithms
A suite of techniques
are currently used to classify remotely sensed data into classes of land cover.
Traditionally, the vast majority of land cover mapping approaches have used
parametric supervised classification algorithms or unsupervised classification
algorithms. The latter use clustering techniques to identify spectrally
distinct groups of data [Schoewengerdt 1997]. These techniques have generally
been used for high resolution imagery, such as Landsat or SPOT.
Global land cover classification efforts,
however, have mostly employed coarse resolution data from NOAA/AVHRR [DeFries
and Townshend 1994]. The literature provides various examples of global land
cover classification efforts. The more traditional approaches include
unsupervised clustering in conjunction with ancillary data and manual labeling
of clusters [Loveland et al. 1991], maximum likelihood classification [DeFries
and Townshend 1994], and simple classification logic based on structural and
biophysical parameters [Running et al. 1995].
More recent approaches include
applications of neural networks [Gopal and Woodcock 1996], including fuzzy
neural networks [Carpenter et al. 1992]. Neural networks can handle relatively
complex relations among the class properties, whereas traditional classification
algorithms are somewhat limited in their statistical and theoretical
sophistication. However, neural nets need an understanding of theory and a
parallel processor to run real-time. They may not be a viable solution to all
applications.
More recently, decision tree
algorithms have been used for the classification of global datasets with
promising results [Friedl and Brodley 1997; DeFries et al. 1998]. Decision tree
techniques have been used successfully for a wide spectrum of classification
problems in various fields [Safavian and Landgrebe 1991]. They are
computationally efficient and exible, and also have an intuitive simplicity.
They therefore have substantial advantages in remote sensing applications.
A decision tree is a classification
algorithm which recursively partitions the feature space of the data set into
increasingly homogeneous subsets based on a set of splitting rules. The tree
has a root, which represents the entire data set, a set of internal nodes
(splits), and a set of terminal nodes (leaves). The nodes represent subsets of
the data set, while the terminal nodes at the bottom of the tree represent the
predictions of the tree. Every node in the tree (except the terminal nodes) has
one parent node and two or more descendant nodes. Each observation is labeled
according to the majority class of the leaf in which it falls [Breiman et al.
1984].
Running et al. [1995] and Nemani and
Running [1997] applied a tree-based decision structure to a global data set of
NDVI values. The data set is both well understood and well behaved and the
classification tree was defined solely on analyst expertise, where the
threshold values are defined based on ecological knowledge. This algorithm,
however, is somewhat difficult to implement since significant spatial, temporal
and spectral variation make globally robust user defined threshold
specification almost impossible.
More commonly, tree-based algorithms
use statistical procedures, which estimate the classification rules from a
training sample. A classic example is the classification and regression tree
(CART) model described by [Breiman et al. 1984]. These algorithms combine the
advantages of statistically based techniques and learning algorithms, which
have their origin in the machine-learning and pattern-recognition communities.
Tree-based methods are supervised techniques and therefore a training set is
required from which the classes can be learned.
A critical step in the estimation of
a decision tree is to prune the tree back in order to avoid overfitting. By
convention a tree is constructed in such a way that all (or nearly all)
training samples are correctly classified, i.e. the training classification
accuracy is 100%. If the training data contains errors the tree will be
overfitted and will generate poor results when applied to unseen data.
2.2 Present
forest area
According to FAO
sources, the global forest area totalled 3529 million ha in 1990's (Fig. 2).
The area of temperate and boreal forests was estimated at 1494 million ha and
the area of tropical forests somewhat larger, 2035 million ha. The area of
other wooded land in temperate and boreal regions was estimated at 372 million
ha.
A remote sensing estimate, based on
radiative transfer biomes, for the global area of broadleaf and needleleaf
forests was 2802 million ha (Fig. 2, 3). This is 727 million ha smaller than
the global forest area according to FAO sources.
- reasons for the
smaller estimate
2.3 Change in
forest area
- FAO sources
The area of temperate
and boreal forest has been estimated to have increased by some 20 million ha
between 1980 and 1995 (State of the World's Forests 1997). Excluding former
USSR, this is a 2.7% increase in the forest area over the 15 year period, on
average approximately 0.2% annually (Fig. 4). The forest area increased,
because of afforestation and reforestation, including natural regrowth fo
forest on land abandoned by agriculture. The afforestation and reforestation
more than compensated the loss of forest due mainly to urbanization and
infrastructure development.
The area of tropical forests, on the
other hand, has been estimated to have decreased by some 200 million ha between
1980 and 1995 (State of the World's Forests 1997). This is a 9% decrease in the
area over the 15 year period, on average about 0.6% in a year (Fig. 4). During
a more recent period, between 1990 and 1995, loss of tropical forests has been
estimated at about 65 million ha. This suggests that the rate of decrease may
be slowing although, according to the FAO source, it is difficult to know if
this is a trend until more data is available.
- remote sensing
estimates
3 Biomass
In this section, we
report the present biomass and recent trends in the biomass of temperate and
boreal forests. We compare the rate of change in the biomass with the rate of
change in the forest area. We analyse trends in tree growth, harvesting levels
and stemwood volume by the main regions of temperate and boreal forests and in
more detail in the forests of Europe. We begin by introducing the concept of
biomass.
3.1 Concept of
biomass
- definition of
biomass in FAO sources
For the FAO Forest
Resources Assessment 2000 of temperate and boreal regions, woody biomass has
been defined as "The mass of the woody parts (wood, bark, branches, twigs,
stumps and roots) of trees, alive and dead, shrubs and bushes, measured to a
minimum diameter of 0 mm" (Forest Resources of Europe… in print). It
includes above-stump woody biomass, stumps and roots but excludes foliage.
The woody biomass is reported in
this report as the amount of carbon in the woody biomass. These carbon figures
are equal to half of the biomass, as they have been calculated by dividing the
biomass figures by two. Such conversion between carbon and biomass is common
practise in carbon studies and considered appropriate as the carbon
concentration varies only slightly between different trees and tree parts.
The present amount of woody biomass
was reported by most of the countries as a part of their recent forest
inventory reporting. For countries that did not report the amount, it was
calculated from growing stock. Growing stock is the stemwood volume of living
standing trees.
The rate of change in woody biomass
was not directly reported by the countries. It was calculated from the stemwood
balance as a difference between net annual increment and annual fellings. These
values, originally given as stemwood volumes in the forest statistics, were
converted to biomass using country-specific conversion factors.
3.2 Biomass in
temperate and boreal forests
- present biomass
The pool of carbon in the woody biomass of temperate and boreal region in 1990's was estimated at 88 Pg (Pg = 1015 g) (Table 3). As much as 47% of the biomass was found in CIS countries, 35% in North America, 10% in Europe and 9% in the other countries of the region. Considering individual countries, the largest pool was in the Russian Federation, 40 Pg or 45% of the total pool in the region, the second largest in the USA, 19 Pg or 21% of the total pool, and the third largest in Canada, 12 Pg or 14% of the total pool. The pool in these three countries represented 80 % of the total pool in the temperate and boreal forests.
Per unit land area, the carbon pool of woody biomass averaged 35 m.t./ha (Table 3). It was fairly similar in Europe, CIS countries and North America varying only from 39 to 44 m.t./ha between these regions. Considering individual countries, the carbon pool per unit area was highest, more than 100 m.t./ha, in four European countries, in Austria, Malta, Slovenia and Switzerland (Fig. 5). In the three countries containing the largest total pool, the carbon pool per unit area was largest the USA, 63 m.t./ha. In Russia it was 45 m.t./ha and in Canada 29 m.t./ha.
3.3. Change in
the biomass of temperate and boreal forests
- recent change
in biomass
- recent change
in biomass compared with the change in forest area
In all studied countries of the temperate and boreal region, net annual increment was larger than annual fellings and, consequently, the amount of woody biomass increased. In the whole region, the carbon pool of woody biomass increased 0.88 Pg/year (Table 3). Forests in CIS countries represented 51%, North America 30%, Europe 13% and the region of the other countries 7% of this increase. In individual countries, the increase rate of the carbon pool was largest in the Russian Federation, 0.43 Pg/year, second largest in the USA, 0.17 Pg/year, third largest in Canada, 0.09 Pg/year, and fourth largest in Australia 0.04 Pg/year. These four countries represented 83% of the total increase, of which the Russian Federation 49%, the USA 19%, Canada 11%, and Australia 5%.
The rate of increase in the carbon pool of woody biomass per unit area on forest and other wooded land averaged 0.35 m.t. carbon/ha/year (Table 3). It varied between the regions from 0.10 m.t. carbon/ha/year in the region of Australia, Japan and New Zealand to 0.52 m.t. carbon/ha/year in Europe. The increase rate was larger than 1 m.t. carbon/ha/year in seven European countries, in Austria, Germany, Hungary, the Netherlands, Romania, Slovakia and Slovenia (Fig. 6).
The annual increase in the carbon pool of woody biomass, 0.88 Pg/year, was 1.0% of the present pool of 88 Pg (Table 3). This annual increase percentage varied between the regions from 0.79% in the region of Australia, Japan and New Zealand to 1.3% in Europe. Among individual countries, the annual increase percentage was highest in European countries, 5.6%/year in Iceland, 4.8%/year in Israel, 4.0%/year in Ireland and 2.4%/year in Spain and Yugoslavia (Fig. 7). In the four countries where the increase rate was largest in absolute terms, the increase percentage was highest in the Russian Federation with 1.1%/year. In the USA it was 0.89 %/year, in Australia 0.79%/year and in Canada 0.78 %/year.
In all regions, the annual increase precentages in biomass were substantially higher than the increase percentages in forest area (Table 3). This indicates that the increase in the biomass is due mostly to a change in the status of the existing forests, regrowth and recovery from earlier harvesting and natural disturbances, and to a smaller extent to the expansion of the forest area.
- comparision of statistics for 1990's and 1980's
The increase of growing stock and, consequently, the increase of woody biomass appeared to be faster in the most recent forest statistics for 1990's than in the statistics for 1980's. The figures must be compared cautiously though, because the definition of forest and other wooded land has been changed (Table 1) and the data are not always available for exactly the same area. In the statistics for mid 1980's, the difference between net annual increment and annual fellings on 1834 Mha in the USSR, North America and Europe was 880 Mm3 (Fig. 8). In the present statistics for early and mid 1990's, the difference on 1899 Mha was more than twice as large, 2060 Mm3. In CIS countries, net annual increment had increased from 1020 to 1320 Mm3 but annual fellings dropped from 580 to 160 Mm3. In Europe and North America, annual fellings had increased from 1180 to 1350 Mm3 but net annual increment had increased even more, from 1620 to 2260 Mm3.
The figures of net annual increment have increased probably because of both a real increase in tree growth and improved increment estimates. The real increase in tree growth may be a result of environmental changes, such as nitrogen deposition, rising temperatures and atmospheric CO2 concentrations, efficient forest management in regions practising intensive forestry, and simply larger growing stock as a result of stock accumulation in the past. Measuring net annual increment is demanding, and publishing conservative increment estimates has been common in the past in order not to encourage or legitimize excessive fellings.
- forest area,
biomass, tree growth and fellings in Europe between 1950 and 2040
The trends in forest
area, biomass, tree growth and fellings were analysed for a longer term in
Europe (Fig. 9). According to forest statistics, the area of exploitable
forests has remained almost unchanged, about 133 million ha, between 1950 and
1990. According to a study of European future forest resources by FAO, the area
is expected increase only slightly from 1990 to 2040. On the other hand, the
estimate for the growing stock of trees, a proxy of biomass, has increased by
more than 40% from 13000 million m3 in 1950 to 18500 million m3 in 1990, and,
according to the FAO prediction, it is expected to increase further to 29400
million m3 in 2040. The accumulation of growing stock is a result of the trends
in tree growth and harvesting. Net annual increment was almost equal to annual
fellings in 1950's and 1960's, and, consequently, growing stock remained
unchanged. Since 1960's, net annual increment has exceeded annual fellings. In
1990's, the increment, 584 million m3, was 43% larger than annual fellings, 408
million m3. Such a trend is also predicted for the future.
4 Conclusions
At present it seems that forest area can be monitored from the satellites. However, forest biomass needs to be measured from the ground.
Both area and biomass are of interest. At present, combining satellite and ground measurements is the wise thing to do. Adopting a conservative approach, both ground and satellite measurements may be needed also in the future, at least for a while. With a more optimistic approach regarding stellite data, maybe the analysis of satellite data can do the whole thing. Even in that case parallel work with ground measurements is neccessary for some time in order to be sure - and to convince the sceptics - that that the new system will work.
References
Remote sensing
Breiman, L., Friedman, J., Olshen, R., and Stone, C. 1984. Classification and Regression Trees. Wadsworth International Group, Belmont, CA.
Carpenter, G., Grossberg, S., Markuxon, N., Reynolds, J., and Rosen, D. 1992. Fuzzy ARTMAP: A neural network architecture for incremental supervised learning of analog multidimensional maps. IEEE Transactions on Neural Networks 3, 698-713.
Choudhury, B. 1991. Multispectral satellite data in the context of land surface heat balance. Review of Geophysics 29, 217-236.corrections. Remote Sensing of Environment 60, 35-37. Cliff, A. and Ord, J. 1973. Spatial Autocorrelation. Pion Limited, London (England).
Cihlar, J., Ly, H., Chen, Z., Pokrant, H., and Huang, F. 1997. Multitemporal, multichannel AVHRR datasets for land biosphere studies-artifacts and corrections. Remote Sensing of Environment 60, 35-37.
DeFries, R., Field, C., Fung, I., Justice, C., Los, S., Matson, P. A., Matthews, E., Mooney, H., Potter, C., prentice, K., Sellers, P., Townshend, J., Tucker, C., Ustin, S., and Vitousek, P. 1995. Mapping the land surface for global atmosphere-biosphere models: Towards continuous distributions of vegetation's functional properties. Journal of Geophysical Research 100, D10, 20,867-20,882.
DeFries, R., Hansen, M., Townshend, J., and Sohlberg, R. 1998. Global land cover classification at 8Km spatial resolution: the use of training data derived from Landsat imagery in decision tree classifiers. International Journal of Remote Sensing 19, 3141-3168.
DeFries, R., Hansen, M., Townshend, J., and Sohlberg, R. 1998. Global land cover classification at 8Km spatial resolution: the use of training data derived from Landsat imagery in decision tree classifiers. International Journal of Remote Sensing 19, 3141-3168.
DeFries, R. and Townshend, J. 1994. NDVI-derived landcover classifications at a global scale. International Journal of Remote Sensing 15, 3567-3586.
Friedl, M. and Brodley, C. 1997. Decision tree classification of land cover from remotely sensed data. Remote Sensing of Environment 61, 399-409.
Gopal, S. and Woodcock, C. 1996. Remote sensing of forest change using artificial neural networks. IEEE Transactions on Geoscience and Remote Sensing 34, 2, 398-404.
Goward, S., Waring, R., Dye, D., and Yang, Y. 1994. Ecological remote sensing at OTTER: satellite macroscale observations. Ecological Applications 4, 322-343.
Holben, B. 1986. Characteristics of maximum-value composite images from temporal AVHRR data. International Journal of Remote Sensing 5, 145-160.
Justice, C., Townshend, J., Holben, B., and Tucker, C. 1985. Analysis of the phenology of global vegetation using meteorological satellite data. International Journal of Remote Sensing 6, 1271-1318.
Knyazikhin, Y., Martonchik, J., Myneni, R., Diner, D., and Running, S. 1998. Synergistic algorithm for estimating vegetation canopy leaf area index and fraction of absorbed photosynthetically active radiation from MODIS and MISR data. Journal of Geophysical Research 103, 32,257-32,279.
Loveland, T. R., Merchant, J. W., Brown, J. F., Ohlen, D. O., Reed, B. C., Olsen, P., and Hutchinson, J. 1995. Seasonal land cover of the United States. Annals of the Association of American Geographers 85, 2, 339-355.
Loveland, T. R., Merchant, J. W., Ohlen, D. O., and Brown, J. F. 1991. Development of a land cover characteristics data base for the conterminuous U.S. Photogrammetric Engineering and Remote Sensing 57, 1453-1463.
Matthews, E. 1983. Global vegetation and land use: New high resolution databases for climate studies. Journal of Climate and Applied Meteorology 22, 474-487.
Moody, A. and Strahler, A. 1994. Characteristics of composited AVHRR data and problems in their classification. International Journal of Remote Sensing 15(17), 3473-3491.
Myneni, R., Maggion, S., Iaquinta, J., Privette, J., Gobron, N., Pinty, B., Verstraete, M., Kimes, D., and Williams, D. 1995. Optical remote sensing of vegetation: Modeling, caveats and algorithms. Remote Sensing of Environment 51, 169-188.
Myneni, R., Nemani, R., and Running, S. 1997. Estimation of global leaf area index and absorbed par using radiative transfer models. IEEE Transactions on Geoscience and Remote Sensing 35, 1380-1393.
Nemani, R. and Running, S. 1997. Land Cover Characterization Using Multitemporal Red, Near-IR, And Thermal -IR Data From NOAA/AVHRR. Ecological Applications (7)1, 79-90.
Olson, J. and Watts, J. 1982. Major world ecosystem comlexes. Oak Ridge National Lab, Oak Ridge, TN.
Prentice, C., Cramer, W., Harrison, S., Leemans, R., Monserud, R., and Solomon, R. 1992. A global biome model based on plant physiology and dominance, soil properties and climate. Journal of Biogeography 19, 117-134.
Running, S. and Hunt, E. 1993. Generalization of a forest ecosystem process model for other biomes, BIOME-BGC, and an application for global-scale model. In: Scaling Physiological Processes: Leaf to Globe, Ehleringer, J. and Field, C.
Running, S., Justice, C., Salomonson, V., Hall, D., Barker, J., Kaufmann, Y., Strahler, A., Huete, A., Vanderbilt, J. M. V., Wan, Z., Teillet, P., and Carneggie, D. 1994. Terrestrial remote sensing science and algorithms planned for EOS/MODIS. International Journal of Remote Sensing 15, 3587-3620.
Running, S. W., Loveland, T. R., Pierce, L. L., Nemani, R. R., and Hunt Jr., E. R. 1995. A remote sensing based vegetation classification logic for global land cover analysis. Remote Sensing of Environment 51, 39-48.
Safavian, S. R. and Landgrebe, D. 1991. A survey of decision tree classifier methodology. IEEE Transactions on Systems, Man, and Cybernetics 21, 3, 660674.
Schoewengerdt, R. 1997. Remote Sensing, models and methods for image processing. Academic Press.
Townshend, J., Justice, C., Li, W., Gurney, C., and McManus, J. 1991. Global land cover classification by remote sensing: Present capabilities and future possibilities. Remote Sensing of Environment 35, 243-255.
Townshend, J. and Tucker, C. 1984. Objective assessment of Advanced Very High Resolution Radiometer data for land cover mapping. International Journal of Remote Sensing 5, 497-504.
Wilson, M. and Henderson-Sellers, A. 1985. A global archive of land cover and soils data for use in general circulation models. Journal of Climatology 5, 119-143.
Zhu, Z. and Yang, L. 1996. Characteristics of the 1-km AVHRR data set for North America. International Journal of Remote Sensing 17, 1915-1924.
Forest inventory information
State of the World's Forests 1997, FAO.
Forest Resources of Europe, CIS, North America, Australia, Japan and New Zealand (industrialized temperate/boreal countries), UN-ECE/FAO contribution to the Global Forest Resources Assessment 2000 (United Nations, New York, in print).
The Forest Resources of the Temperate Zones, The UN-ECE/FAO 1990 Forest Resource Assessment, Volume I, General Forest Resource Information (United Nations, New York, 1992); European Timber Trends and Propects to the year 2000 and beyond, Volume I, (United Nations, New York, 1986).
K. Kuusela, Forest Resources in Europe 1950-1990 (European Forest Institute Research Report 1, Cambridge University Press, New York, 1994).
Pajuoja, H. 1995. The outlook for the European forest resources and roundwood supply. ETTS V Working Paper. UN-ECE/FAO Geneva.
Table 1. FAO definitions for forest and other wooded land.
|
Country group |
Forest |
Other wooded land |
|
Developed countries of temperate and boreal regions (FRA-1990, State of World's Forests 1997) |
Land with tree crown cover (stand density) of more
than about 20 percent of the area. Continuous forest with trees usually
growing to more than about 7 m in height and able to produce wood. This
includes both closed forest formations where trees of various storeys and
undergrowth cover a high proportion of the ground, and open forest formations
with a continuous grass layer in which tree synusia cover at least 10 percent
of the ground. |
Land which has some forestry characteristics but is
not forest as defined above. It includes open woodland and scrub, shrub and
brushland (see below) whether or not used for pasture or range. It excludes
land occupied by ‘trees outside the forest’. |
|
Developed countries of temperate and boreal regions (FRA-2000, Forest Resources of Europe … in print) |
Land with tree crown cover (or equivalent stocking level) of more than 10 percent and area of more than 0.5 ha. The trees should be able to reach a minimum height of 5 m at maturity in situ. May consist either of closed forest formations where trees of various storreys and undergrowth cover a high proportion of the ground; or of open forest formations with a continuous vegetation cover in which tree crown cover exceeds 10 percent. Young natural stands and all plantations established for forestry purposes which have yet to reach a crown density of 10 percent or tree height of 5 m are included under forest, as are areas normally forming part of the forest area which are temporarily unstocked as a result of human intervention or natural causes but which are expected to revert to forest. |
Land either with a tree crown cover (or equivalent stocking level) of 5-10 percent of trees able to reach a height of 5 m at maturity in situ; or crown cover (or equivalent stocking level) of more than 10 percent of trees not able to reach a height of 5 m at maturity in situ (e.g. dwarf or stunted trees) and shrub or bush cover. |
|
Developing countries of tropical regions (FRA-1990, State of the World's Forests 1997) |
Ecosystem with a minimum of 10 percent
crown cover of trees and/or bamboos, generally associated with wild flora,
fauna and natural soil conditions, and not subject to agricultural practices.
The term forest is further subdivided, according to its origin, into two
categories: i) Natural forests: a subset of forests composed of tree species
known to be indigenous to the area; and ii) Plantation forests: established
artificially by afforestation on lands which previously did not carry forest
within living memory; established artificially by reforestation of land which
carried forest before, and involving the replacement of the indigenous
species by a new and essentially different species or genetic variety. |
Includes the following: i) Forest fallow, consisting
of all complexes of woody vegetation deriving from the clearing of natural
forest for shifting agriculture. It consists of a mosaic of various
successional phases and includes patches of uncleared forests and agriculture
fields which cannot be realistically segregated and accounted for area-wise,
especially from satellite imagery. Forest fallow is an intermediate class
between forest and non-forest land uses. Part of the area which is not under
cultivation may have the appearance of a secondary forest. Even the part
currently under cultivation sometimes has the appearance of forest due to the
presence of tree cover. Accurate separation between forest and forest fallow
may not always be possible. ii) Shrubs, referring to vegetation types where
the dominant woody elements are shrubs with more than 50 cm and less than 5 m
height on maturity. The height limits for trees and shrubs should be
interpreted with flexibility, as the minimum tree and maximum shrub heights
may vary between 5 and 7 m approximately. |
Table 2. Canopy structural attributes of global land covers from the viewpoint of
radiative transfer modeling (Myneni et al. 1997).
Table 3. The area, the change in the area, the woody biomass and the change in the
woody biomass of temperate and boreal forests.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Region |
Area of forest and other wooded land (Mha) a |
|
Carbon pool (Pg) a |
Carbon pool (tn hafowl-1) a |
|
Carbon sink (Pg year-1) a |
Carbon sink (tn hafowl-1 year-1)
a |
Carbon sink (% of the present pool year-1)
b |
|
Area of forest (Mha) b |
Change in the area of forest (% of the present area
year-1) b |
|
Europe |
215 |
|
8 |
39 |
|
0.11 |
0.52 |
1.3 |
|
153 |
0.3 |
|
CIS |
934 |
|
41 |
44 |
|
0.45 |
0.48 |
1.1 |
|
856 |
0.1 |
|
of which Russia |
887 |
|
40 |
45 |
|
0.43 |
0.48 |
1.1 |
|
817 |
n.a. |
|
North America |
716 |
|
31 |
43 |
|
0.27 |
0.38 |
0.9 |
|
461 |
0.2 |
|
of which Canada |
418 |
|
12 |
29 |
|
0.10 |
0.25 |
0.8 |
|
245 |
0.1 |
|
of which USA |
298 |
|
19 |
63 |
|
0.17 |
0.56 |
0.9 |
|
217 |
0.3 |
|
Australia, Japan and New Zealand |
613 |
|
8 |
13 |
|
0.06 |
0.10 |
0.8 |
|
74 |
0.1 |
|
of which Australia |
578 |
|
5 |
9 |
|
0.04 |
0.07 |
0.8 |
|
41 |
n.s. |
|
Total |
2477 |
|
88 |
35 |
|
0.88 |
0.35 |
1.0 |
|
1544 |
0.1 |
a Forest Resources of Europe … in print
b State of the World's Forests 1997
Fig. 1. Relationships of NDVI/LAI and NDVI/FAPAR: Results for broadleaf forests and needleleaf forests from prototyping efforts with POLDER data (Zhang et al., BU MODIS/MISR LAI/FAPAR team at Boston University).
Fig. 2. Global forest area by regions in 1990's. For each region, the column on
the left hand side represents the estimate according to FAO sources (State of
the World's Forests 1997, Forest Resources of Europe … in print) and the colums
on the right hand side the estimate according to a remote sensing approach. The
FAO estimates are divided to forest and other wooded land and the remote
sensing estimates to broadleaf and needleleaf forest and savanas.
Fig. 3. Multisource-Mapping of Structural Vegetation Types (Biomes). The six biome classes are reproduced poorly
in this draft scan: red="savannas", dark green="broadleaf forests",
light green="needleleaf forests".