Assessing the sustainability of harvest of Pterocarpus angolensis

in Rukwa Region, Tanzania

M.W. Schwartz
a,*
, T.M. Caro
b
, T. Banda-Sakala
c
a
Division of Environmental Science and Policy, University of California, One Shields Avenue, Davis, CA 95616, USA
b
Department of Wildlife, Fish and Conservation Biology, University of California, Davis, CA 95616, USA
c
Graduate Group in Ecology, University of California, Davis, CA 95616, USA
Received 5 July 2001
Abstract
Pterocarpus angolensis is being cut for construction and medicinal purposes through much of southern Africa. To assess the
sustainability of this offtake, we estimated current harvest practice, population size structure, growth rates and seed production
in protected and unprotected areas of Rukwa Region, western Tanzania. Results showed that adult P. angolensis are being
removed very rapidly given its slow growth rate as determined from tree ring data, and that regeneration both inside and outside
the protected area is very low. A simple model indicates that current harvest is unsustainable raising serious concerns about the
long-term viability of this important hardwood species in miombo woodland.
# 2002 Elsevier Science B.V. All rights reserved.
Keywords: Katavi National Park; Miombo; Pterocarpus angolensis; Tree harvest
1. Introduction
African miombo is a dominant vegetation type over
much of south central Africa. It covers 2.7 million km
2
(Millington et al., 1986) extending from Tanzania and
southern Zaire in the north to the northern provinces of
South Africa, and across the continent from Angola
through Zambia to Malawi and Mozambique (Chidu-
mayo, 1994). Over most of its range, mature undis-
turbed miombo is deciduous woodland, i.e. classied
with savanna ecosystems but grades into seasonal
closed dry forest (Frost, 1996). Miombo woodlands
are characteristically dominated by the genera
Brachystegia, Isoberlinia and Julbernadia, all of the
family Leguminosae (Chidumayo, 1994). One of these
hardwood tree species, Pterocarpus angolensis, is a
commercially important canopy tree that grows from
sea level in the Mozambican coast, up to 1650 m
above sea level (Jeffers and Boaler, 1966; Stahle
et al., 1999). In Tanzania, P. angolenesis wood is
widely utilized for furniture, veneer, carving and
general purpose timber (Monela et al., 1993). In South
Africa, the sap is used traditionally for treatment of
ringworms and ulcers (Van der Reit et al., 1998), eye
treatment against cataracts, malaria, blackwater fever
and skin inammation (Palgrave, 1981; Watt and
Breyer-Brandwijk, 1962), and urinary schistosomiasis
(Ndamba et al., 1994; Nyazema et al., 1994).
Van Daalen (1991a) observed that P. angolensis
regenerated sparsely in South Africa, while Boaler
(1966) estimated that only 2% of fruits produced by
this species germinated in the eld in Tanzania. With
Forest Ecology and Management 170 (2002) 259269
*
Corresponding author. Tel.: 1-530-752-0671;
fax: 1-530-752-3350.
E-mail address: mwschwartz@ucdavis.edu (M.W. Schwartz).
0378-1127/02/$ see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S0 3 7 8 - 1 1 2 7 ( 0 1 ) 0 0 7 7 4 - 5
observations of low recruitment success, the growing
stock of P. angolensis may be diminishing across its
distribution naturally. The growing stock is further
diminishing as logging pressure increases in an
attempt to satisfy an expanding market demand.
Documentation of cubic meters of wood volume
harvested is lacking, making it difcult to assess the
extent of the problem. However, the mining industry
has used timber and poles in Zambia (Chidumayo,
1993), and between 1937 and 1961 about 127,000 ha
of miombo were selectively harvested for these
purposes (Lees, 1962). In South Africa, a partial
estimate that does not include local timber use, values
P. angolensis harvest at over R5 million ($646,663)
(Van Daalen, 1991a). There are currently no data for
Tanzania.
In this paper, we survey populations of P. angolensis
inside and outside a protected area in western Tanza-
nia to predict the future population status for P.
angolensis in this location under current management
practices. We believe that this example is a good test
case of the likely outcome of logging since the meth-
ods of logging are inefcient and utilize little modern
technology. Trees are hand cut, rolled to sawpits, sawn
into logs and then removed by truck to a train depot
more 50 km distant. If under these labor intensive
logging practices, loggers can selectively cut P. ango-
lensis to economic extinction, then we presume this to
be the case wherever transportation corridors are
available. Specically, our goal is to estimate current
harvest patterns, population size structure, growth
rates, and seed production in our study site in order
to construct a model with which to predict future stand
population under current harvest regimes.
2. Methods
2.1. The study site
The study site was in and immediately south of
Katavi National Park (KNP), latitude 6845
0
7805
0
S,
longitude 30845
0
31825
0
E at the north end of Rukwa
Valley in Rukwa Region, western Tanzania. The area
is low elevation characterized by sandy soils and 600
1500 mm rainfall per year that falls between Novem-
ber and April. The study site consisted of two legally
designated areas (Fig. 1). KNP consists largely of
miombo woodland, a dry deciduous forest character-
ized by Acacia, Combretum, Grewia, Kigelia, Pter-
ocarpus and Terminalia tree species (Rodgers, 1996)
but also encompasses two seasonally inundated ood-
plains, Lakes Katavi and Chada (Caro, 1999a). No
temporary or permanent settlements are allowed aside
from Park headquarters and two outlying ranger posts;
no livestock, beekeeping, hunting, shing or timber
extractions are tolerated. Tanzania National Park ran-
gers conducting vehicle and foot patrols enforce these
laws.
The second area, Msaginia Forest Reserve (MFR),
adjacent and north east of the Park, supports miombo
woodland and is dominated by mixed stands of Bra-
chystegia spiciformis, B. boehmii, Afzelia quanzensis,
P. angolensis and Erythrophleum africanum. Settle-
ments and cattle grazing are forbidden in the Forest
Reserve but selective harvest of P. angolensis is
carried out under license. In practice, most tree cutters
spend months in the Forest Reserve and illegally hunt
mammals there. The study was conducted in August
and September 2000 during the middle of the dry
season.
2.2. Demographic estimates
2.2.1. Population density and size structure
In 50 m20 m (0.1 ha) plots, we recorded the
identity of all individual trees at 14 sites in KNP
and ve in MFR. In each plot, we measured the exact
diameter at breast height (dbh) of trees >2 cm dbh.
Separately, tree density estimates of P. angolensis in
MFR were estimated by walking three transects 1 km
in length and counting all live trees and cut stumps
25 m on either side of the observer (15 ha sampled).
The distribution of live and harvested P. angolensis
tree sizes in MFR were characterized by measuring all
trees N 87 encountered on sampling transects
conducted in the vicinity of on-going tree harvest.
Size was measured as dbh for all live trees; for cut
trees we measured the basal diameter, given that dbh
did not exist. These trees do not strongly buttress,
however, and basal diameter is only slightly larger
than dbh measurements. The size structure of trees in
KNP was measured from trees encountered in vegeta-
tion sampling plots as well as trees outside of plots that
were sampled in order to characterize seed production
by tree size.
260 M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269
2.2.2. Seed production
Two researchers counted ground and canopy
diaspores for each P. angolensis tree. Diaspores are
large (1015 cm diameter) and readily visible by their
prominent circular wing that contains a single central
seed. Diaspores were counted during the dry season
when the canopy was leaess, making visibility very
high, and before most had dispersed from trees. Bare
ground during the dry season made counting diaspores
that had fallen relatively easy. Despite a large wing,
most diaspores fall within 15 m of the parent tree so
we restricted our estimates to isolated trees so as to
reduce confusion with diaspores dispersed from
neighboring individuals. Despite these favorable
attributes, our estimates must be considered minimal
estimates of total seed production. For each tree, we
used the maximum value of diaspores counted on the
ground summed with the average value of diaspores
counted in the canopy. Our rationale for this is that all
canopy diaspores were apparent and could be counted.
Sources of error in canopy counts arise from uncertain
numbers of diaspores held in clusters on branches. As
a result, over-counting is possible. In contrast,
diaspores on the ground are not easily over-counted.
Fig. 1. Map of KNP and MFR showing the location of 14, 0.1 ha plots inside the Park and ve inside the Reserve (marked as squares). Legal
boundaries are dotted lines, dirt roads are dashed lines. The town of Sitalike and villages of Mirumba, Manga and Kibaoni are also shown.
Insert shows the location of KNP in Tanzania.
M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269 261
It is more likely that differences in ground counts
arose from the different observers encountering
patches missed by the other.
2.2.3. Tree growth rate
The phenology of P. angolensis is strongly
deciduous and is tightly synchronized with precipita-
tion: owering and leaf ush occur during August
December and leaves are shed in May and June. That
leaf fall begins early and is more complete and
prolonged than most other species affords P. ango-
lensis one of the strongest deciduous habits of trees in
the dry forests of south tropical Africa (Stahle et al.,
1999); annual rings are present on P. angolensis,
although they are often difcult to discern (Stahle
et al., 1999). We wanted to gain a rough estimate of the
time required for adult trees to pass fromone 5 cmsize
class to the next in order to facilitate a prediction of the
rate at which trees are currently being harvested to
extinction. As such, we present a simple estimate of
average ring width for trees in this region, but do not
attempt a full reconstruction of size by growth
patterns. Increment cores (5 mm) were collected from
10 adult trees, were mounted in wood, sanded smooth
and measured with a binocular microscope. Average
ring width values were used to estimate annual
diameter increment growth. We compare these growth
increments to published estimates of growth (Van
Daalen, 1991b; Van Daalen et al., 1992; Stahle et al.,
1999).
2.2.4. Tree harvest selection
We interviewed loggers in Kiswahili in order to
determine reasons for not cutting trees within size
classes of trees otherwise cut (>25 cm dbh). Primary
reasons given for not harvesting trees were bent trunks
and hollow trunks such that boards could not be cut
from the trunk. The larger a tree, the more likely it was
to be harvested as larger trunks produced more boards
per trunk. All boards were hand sawn in sawpits on
location by the individuals who harvested the trees.
2.3. Modeling the consequences of logging
To forecast the status of P. angolensis within the
MFR, we constructed a matrix projection model to
create scenarios for future forest stocking rates under
different harvest scenarios. Our model uses 11 size
classes of trees in 5 cmintervals from15 to 65 cmdbh,
encompassing all size classes currently observed
within MFR. Our model projects populations in time
steps that correspond to the length of time it takes for a
tree to pass from one 5 cm dbh size class to the next.
We estimate that these time steps to be approximately
50 years (see Section 3). Population projections pre-
sented here are deterministic and result in a single
forecast of future population. It is notable that we
observed no trees smaller than 15 cmdbh although it is
possible that these trees existed as one of our eld
recognition signs was seed in the canopy. While small
trees, lacking seeds, might have been easy to overlook,
the characteristic bark of P. angolensis made this
relatively unlikely.
The rst step of our model was to construct a stable
stage distribution (SSD) of a hypothetical population.
At SSD the proportion of individuals in each stage
class is constant. To do this, we used 90% survivorship
of adult trees per time step, roughly corresponding to
estimated tree mortality rates for P. angolensis (Des-
met et al., 1996) and typical of tree life table values
(Harcombe, 1987). For fecundity, we used our empiri-
cal estimate of the relative contributions of trees in
different size classes to total seed production and set
total seed production values to obtain a stable popula-
tion size l 1 at SSD. When l 1, the population
neither increases nor decreases in number. Projecting a
population using values at SSD is deterministic and
stationary.
To project future populations, we modeled a popu-
lation beginning at SSD. We then allow harvest of
trees based on our observed probabilities of trees in
different size classes being harvested. In other words,
we reduced survivorship of adult trees making survi-
vorship less likely with larger size, corresponding to
observations of the sizes of cut and uncut stems. In this
manner, the persistence likelihood of a small tree is
very high, as few of these are cut, and the likelihood of
persistence of a large tree is very low, as all of these
that we observed had been felled. Like the previous
model, this projection is deterministic, but now is not
stationary. We compared the current size structure of
the population to various time steps in the modeled
population with harvest.
We use two variations of this model to project future
populations under logging scenarios. These models
vary in their assumptions regarding logger behavior. In
262 M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269
our rst harvest, model we use the distribution of cut
and uncut trees within size classes as an estimate of the
likelihood of a tree surviving into the next size class.
In this model, the probability of tree mortality (harvest
plus natural mortality factors) is xed for the entire
projection (Table 1). In the second iteration, we allow
loggers to change their behavior such that they begin
harvest with the largest trees in the early time steps
and decrease the size of harvestable trees as they
deplete the larger size classes until they reach the
pattern of timber harvest currently observed. Inter-
views with loggers suggest that it is not feasible to cut
trees much smaller than they do at the present time.
Thus we xed logging behavior at current patterns and
did not allow them to continue to harvest even smaller
trees. In this sense, our model is conservative. Both of
these models result in nearly the same projections, so
we restrict our presentation to the conceptually sim-
pler rst model of constant logger behavior cutting
trees down to currently observed sizes. A third
scenario, where loggers continue to return to the forest
and harvest smaller and smaller trees results in a
simple and rapid extinction of the species: the result
is self-evident and does not require presentation.
We project the size class distribution of trees using
this model into the future assuming that a further
decrease in harvestable tree size is impossible. We
also present our forecast of tree population, along with
a prediction of former population size based on tree
densities. We used the most closely similar model time
step to the current distribution of trees in the different
size classes in order to estimate absolute tree density.
Based on current stand density, we backtracked model
time steps in order to estimate pre-logging tree den-
sities. These represent very rough estimates of past
tree density. We also projected future tree densities
based on current stand densities. These latter stand
density estimates are likely to be more robust.
Table 1
Transition matrix models used to: (a) establish an SSD; (b) simulate harvest at a constant level
a
Size class (cm dbh) 15 20 25 30 35 40 45 50 55 60 65
(a) SSD, no population growth
15 0 0 0.03 0.06 0.08 0.11 0.14 0.17 0.19 0.22
20 0.9 0
25 0.9 0
30 0.9 0
35 0.9 0
40 0.9 0
45 0.9 0
50 0.9 0
55 0.9 0
60 0.9 0
65 0.9 0
(b) Constant harvest schedule
15 0 0 0.03 0.06 0.08 0.11 0.14 0.17 0.19 0.22
20 0.9 0
25 0.83 0
30 0.81 0
35 0.5 0
40 0.25 0
45 0.25 0
50 0.2 0
55 0.2 0
60 0.2 0
65 0.2 0
a
In each case, the probability of remaining within a size class across time steps is 0, reecting a time step that is sufcient for trees to
increase by 5 cm diameter increment. Adult survival is high in the absence of logging. Recruitment values are apportioned by actual seed
production estimates and scaled so that the total population growth rate is 0 l 1:0.
M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269 263
3. Results
3.1. Current populations
The distribution of P. angolensis sizes show that
KNP trees are very much larger than cut or uncut trees
in MFR (Fig. 2). We attribute this rst to the fact that
selective logging in the Forest Reserve has been on-
going for 30 years and that large trees in MFR have
long since been felled. Second, the population within
KNP has had virtually no P. angolensis recruitment for
some period of time. As a result, there are virtually no
small trees within the Park. Note, however, that it is
not just small P. angolensis that are lacking in KNP:
there are few small trees of any species in the Park
compared to the MFR (Table 2).
The distribution of size classes of cut and uncut P.
angolensis trees in MFR demonstrates that loggers, as
expected, are cutting the largest of the remaining P.
angolensis. Nonetheless, the current round of selective
logging results in loggers leaving some seed produ-
cing trees. The tree density transects yielded a total
count of 171 trees in 15 ha surveyed (11.4 trees/ha).
Of these 171 trees, just 55 (32%) were alive, yielding
an estimate of 3.67 live trees/ha at the present time.
The relationship between tree size and seed
production was highly signicant (seed production
22:576dbh 525:17; r
2
0:5935; p < 0:0001,
Fig. 3). The majority of the error associated with this
estimate is associated with large trees; small trees
(<40 cm dbh) show considerably less variation in
Fig. 2. Size class distribution of trees in the MFR and KNP, disting-
uishing live and cut trees. Proportion of population, not density.
Table 2
Estimated mean number of trees per hectare (standard deviation) by
size class of all species within KNP (N 14 plots) compared to the
MFR (N 5 plots)
a
Location Size class (cm dbh)
25 510 1025 >25
KNP 7.4 (7.8) 6.8 (7.3) 7.3 (7.4) 7.1 (4.6)
MFR 20.8 (9.4) 34.2 (19.4) 26.0 (11.6) 6.4 (4.6)
t-Value 1.94 1.84 2.13 0.11
p-Value 0.069 0.083 0.047 0.91
a
Density estimates are based on trees sampled in 0.1 ha plots.
Differences in density within size class were tested using a t-test.
Fig. 3. The relationship between tree size and seed production for P. angolensis in and around KNP, Tanzania, y 22:576x
525:17; r
2
0:59; n 31; p < 0:001. The signicant positive relationship pertains with or without the inclusion of any or all of the three
outlying points represented by the largest trees sampled in this population.
264 M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269
seed production. The coefcient of variation in seed
production for trees <40 cm dbh was 0.242, roughly
half of that of the entire data set (0.475).
Based on our tree density estimate of 3.67 live trees/
ha and the distribution of size classes (Fig. 2), we
predict that the current forest produces 613 seeds/ha
(Table 3). Current seed production is estimated to be
15.6% of the seeds that would have been produced by
the trees currently represented in the forest as stumps
(Table 2), and equivalent to that predicted from a
single 50 cm dbh tree.
Our observed mean tree ring width was 0.49 mm
(Fig. 4). This estimated growth rate is slower than
that of Stahle et al. (1999), resulting in a predicted
residence time of approximately 50 years within each
5 cm size class. We use this estimate to calculate the
rate at which remaining trees will reach harvestable
size using our population model.
3.2. Population model
Our population projection model allows us to
accomplish three tasks. First, we compare the distri-
bution of trees (live plus recently cut) to the distribu-
tion of trees at SSD to assess the degree to which the
model appears to capture current logging behavior.
Second, we standardize SSD stand densities by the
current densities in order to estimate pre-logging
population size. Finally, we use additional time steps
to estimate the trajectory of the future population of P.
angolensis in our study region.
By assuming that prior to the current bout of cutting
in the area there was no anthropogenic loss of size
classes 46 (3045 cm trees), we can standardize SSD
densities in these size classes to equal observed den-
sities (Fig. 5). Observations of recently cut stumps
shows that current logging efforts are focused on the
3045 cm sized trees (Fig. 2). Most trees larger than
45 cm are already cut, thus the current population of
trees and stumps contains far fewer large trees than the
SSD predicts (Fig. 5). We expect this pattern, as we
know larger trees have been extracted from the stand.
We have no evidence that the population was at SSD
prior to commencement of logging practices in the
region in the 1970s; we simply chose this as starting
point. Nonetheless, it is reassuring to the veracity of
our model that the three size classes that are currently
being cut t the distribution of densities that we expect
based on SSD (Fig. 5).
We observe fewer small trees in the current stand
than in SSD(Fig. 5). The paucity of small trees may be
Table 3
Total estimated annual seed production from remaining live trees
within the MFR
Size class
(cm dbh)
Live
trees/ha
Estimated seed
production/tree
Seeds/ha
1519.9 0.26 0 0
2024.9 0.84 10.5 8.8
2529.9 1.35 142.4 192.5
3034.9 0.90 274.2 247.1
3539.9 0.13 406.0 52.3
4044.9 0.13 537.8 69.3
4549.9 0.06 669.7 43.1
Total 613.1
Fig. 4. The distribution of tree ring widths in millimeter based on a sample of 302 tree rings measured from 10 trees (dbh 4767 cm) from KNP.
M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269 265
explained by one of two reasons. Harvest of large trees
would dramatically reduce total stand seed production
(Table 3) and hence is likely to be reducing recruit-
ment. However, given that transition from one size
class to the next requires up to 50 years, based on tree
ring widths, we predict that all trees >15 cm dbh
predate the onset of logging. Thus, we hypothesize
that some secondary factor resulted in the reduction of
recruitment in P. angolensis prior to the onset of
logging in the past 30 years.
Mechanism aside, these data suggest that recent
recruitment in this forest falls far short of that required
for a sustainable population. This observation is con-
sistent with observations of tree recruitment immedi-
ately under existing trees and paired stumps, which
demonstrated no increased recruitment of any tree
species as a result of logging (Schwartz and Caro,
unpublished data).
The simulation projection of constant logging
efforts predicts a continued rapid decline of harvest-
able trees (Fig. 6). Standardizing model size distribu-
tions to current actual densities of live plus recently
cut trees in the 3045 cm size classes, we predict that
the density of trees >15 cm dbh at SSD was 27.8 trees/
ha. The logging model time step 4 predicts tree
densities of 9.4 trees/ha, close to the current densities
of live plus cut trees (11. 4 trees/ha). Model time step 6
predicts tree densities of 3.2 trees/ha; live tree density
is currently estimated at 3.7 trees/ha. Most of the
estimated 1.25 trees/ha large enough to cut at the
present time were not cut because of bends and splits
in trunks. Based on these results, we predict a total of
2.1 trees/ha maturing into a harvestable size class
during the next 2 time steps (100 years), or approxi-
mately 1 tree/ha each 50 years. Assuming continued
patterns of harvest and recruitment, this would leave
0.3 trees/ha if these trees were logged during this time.
Thus, the current stock of P. angolensis is nearly
completely depleted for the foreseeable future in this
stand.
Fig. 5. A comparison of actual tree density estimates based on live plus recently cut tree stumps in MFR to an SSD model of a generalized tree
model with high adult survivorship and a zero net population growth. Model results t tree size density distributions in size classes not readily
cut very well, but show that recruitment lags behind that needed to restock the current population.
Fig. 6. Model projections of the density of trees in the MFR
assuming constant harvest pressure. Time steps are the time it
requires for an individual tree to pass through a 5 cm size class.
This was estimated to be 50 years. Dark shaded regions indicate the
density of trees large enough to be harvested and light shaded
region are trees in smaller size classes.
266 M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269
4. Discussion
Our simple measurements of P. angolensis in the
Rukwa Region of western Tanzania have produced
three main ndings. First, in the MFR, harvest inten-
sity is leading to the rapid demise of P. angolensis and
evidence strongly suggests that this harvest rate is
unsustainable. Loggers have recently harvested a total
of 7.7 trees/ha, reducing population density from 11.4
to 3.7 trees/ha. Trees are left standing because they are
not of harvestable size yet. With slow growth rates and
a potential of just 2.1 trees/ha maturing into harvest-
able size over the next 100 years, we predict that there
currently remains far less than 1 tree/ha to harvest. We
cannot predict when economic extinction will occur
within this system, but we can say that the population
is heading toward economic extinction rapidly. At
present, trees are harvested by cutting trunks with
hand saws, then rolling them to saw pits and hand
sawing theminto boards. Given the weight of trunks, it
is not practical to roll a log long distances, nor dig a
saw pit for every tree harvested. Our model suggests
that P. angolensis may be commercially extinct when
this current cutting cycle is completed, at least in the
portion of the MFR where we conducted our study.
At present, we can predict two scenarios. First,
loggers will effectively harvest to economic extinc-
tion, leave the area and thus allowthe species to persist
at low density because it is not worth cutting an
occasional single tree or trees that are hollow or
unsuitable for harvest yet still produce seeds. Result-
ing saplings have a chance of entering adult age
classes in the absence of large herbivores. Alterna-
tively, loggers may shift harvest practice to other
sympatric species, such as Swartzia madagascarensis
(pers obs), and cut P. angolensis opportunistically as
encountered. The latter could result in complete extir-
pation of P. angolensis. Discussions with loggers
suggest that they believe that there is approximately
5 years to go before they will be forced to leave the
area, but that there are insufcient Swartzia in the
MFR to allow them to earn a living. Thus, if we had to
speculate, we would suggest that the rst hypothesis is
more likely.
A second conclusion that emerges from this work
is that populations of P. angolensis within KNP appear
to be suffering from a lack of recruitment relative
to populations outside the Park. As yet, we do not
understand why this is but our working hypothesis
is that large herbivores are preventing recruitment of
P. angolensis and many other trees into the adult
population. We reached this conclusion because we
noticed, in a parallel study, that recruitment was rapid
and vigorous outside the Park but very limited inside
despite both the areas being similarly subject to
repeated dry season res (Schwartz and Caro,
unpublished data). In contrast to areas outside, KNP
has signicantly greater densities of large ungulates
including elephant, giraffe, zebra, and impala all of
which browse regenerating trees (Caro, 1999b).
Herbivores live at low densities outside the Park as
a result of illegal hunting. In short, in an area of
miombo legally protected from tree harvest, and
where enforcement is diligent, there is an absence
of recruitment into the adult P. angolensis population.
In addition, there has not been compensatory recruit-
ment of trees in logged areas that would allow for
continued logging with increased canopy tree cycling,
or for recovery of the P. angolensis. The literature
suggests that P. angolensis might be characterized by
bouts of recruitment (e.g., Chidumayo, 1994). Given
that there is no seed dormancy, the potential to take
advantage of the right conditions for recruitment is
severely hindered by logging. A persistent seedling,
known in the literature as a suffrutex (Boaler and
Sciwale, 1966) provides the potential for advanced
recruitment in the forest. If there is a large population
of seedlings in the eld at present, these could form
the next generation of canopy trees. Most areas we
surveyed during the course of this study were burned,
and all P. angolensis had lost their leaves during our
study. As a result, we were not able to census seedlings
during the dry season. Thus, we cannot comment on
the role of seedlings for potential future recruitment in
these stands at this juncture. At the very least, recruit-
ment will be a slow process since few seed producing
trees are available.
In conclusion, we are concerned about loss of P.
angolensis in this Forest Reserve and others under
similar cutting pressure in the Rukwa Region of
Tanzania. We acknowledge that these Forest Reserves
provide a refuge for many forest resources at the
expense of a few harvested species: were Forest
Reserves not in place, many of the trees would already
have been removed completely for rewood. Thus, the
forest remains, at the expense of P. angolensis. We are
M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269 267
also concerned that the species is being lost fromKNP.
If it is a general phenomenon that P. angolensis cannot
regenerate in areas of high ungulate biomass, then it is
not safe within heavily protected National Parks in the
miombo region of southern Africa. In short, lack of
regeneration within heavily protected areas, unsus-
tainable harvest in lightly protected areas, and whole-
sale removal of forest in areas that enjoy no legal
protection paint a very disturbing picture for this,
and possibly other, hardwood species in this area of
Africa. Our study area, in western Tanzania is an
economically depressed region with little ability to
rapidly exploit forest resources: there are no chain
saws, few trucks and few loggers. If people here are
able to cut a hardwood species to economic extinction,
then we fear that hardwood populations are in
jeopardy in this part of Africa. Efforts to grow P.
angolensis commercially as a forestry species are
underway in South Africa but face problems of low
recruitment success and slow growth (Stahle et al.,
1999; Munyanziza and Oldeman, 1995).
To understand the impacts of logging on the global
distribution of P. angolensis, additional information is
required on the extent of forests under different levels
of protection, regional differences in the methods and
intensity of harvest, and the relationship between
transportation corridors and harvest in the different
countries within the species distribution. To under-
stand better the population dynamics of P. angolensis
within our particular study region, we need rst to
census seedlings and estimate the likelihood of
seedling recruitment into a juvenile tree. Second,
we need to comprehend the role of re management
and high herbivore densities on seedling recruitment
within the National Park. Given that this is likely to
be the sole refuge for these trees under intensifying
logging pressure, experiments need to be conducted as
to whether alternative re management regimes or
herbivore exclosures may facilitate healthy recruit-
ment within the Park.
Acknowledgements
We thank the Tanzania Wildlife Research Institute
and Tanzania National Parks for permission to conduct
research, the University of California Faculty
Research Grant Program for funding, the people of
Kibaoni for their hospitality, and two anonymous
reviewers for thoughtful comments on an earlier draft.
References
Boaler, S.B., 1966. The ecology of Pterocarpus angolensis DC. in
Tanzania. Ministry of Overseas Development, London. Over-
seas Research and Publication No. 12, 128 pp.
Boaler, S.B., Sciwale, K.C., 1966. Ecology of a miombo site,
LupaNorth Forest Reserve, Tanzania. III. Effects on the
vegetation of local cultivation practices. J. Ecol. 54, 577587.
Caro, T.M., 1999a. Abundance and distribution of mammals in
Katavi National Park, Tanzania. Afr. J. Ecol. 37, 305313.
Caro, T.M., 1999b. Densities of mammals in partially protected
areas: the Katavi ecosystem of western Tanzania. J. Appl. Ecol.
36, 205217.
Chidumayo, E.N., 1993. Responses of the Miombo to Harvesting:
Ecology and Management. Stockholm Environment Institute,
Stockholm.
Chidumayo, E.N., 1994. Miombo Ecology and Management: An
Introduction. Intermediate Technology Publications, London.
Desmet, P.G., Shackleton, C.M., Robinson, E.R., 1996. The
population dynamics and life-history attributes of a Pterocar-
pus angolensis DC. population in the Northern Province, South
Africa. S. Afr. J. Bot. 62, 160166.
Frost, P., 1996. The ecology of miombo forests. In: Campbell, B.
(Ed.), The Miombo Transition: Woodlands and Welfare in
Africa. Center for International Forestry Research, Bogor,
Indonesia, pp. 1157.
Harcombe, P.A., 1987. Tree life tables: simple birth, growth and
death data encapsulate life histories and ecological roles.
Bioscience 37, 557568.
Jeffers, J.N.R., Boaler, S.B., 1966. Ecology of a miombo site,
LupaNorth Forest Reserve, Tanzania. I. Weather and plant
growth 19621964. J. Ecol. 54, 447463.
Lees, H.M.N., 1962. Working plan for the forests supplying the
Copperbelt, Western Province. Government Printer, Lusaka,
159 pp.
Millington, A.C., Townsend, J.R.G., Saull, R.J., Kennedy, P.,
Prince, S.D., 1986. SADCC fuelwood project: biomass
assessment component. 2nd Interim Report, Munslow, 129 pp.
Monela, G.C., OKtingati, A., Kiwele, P.M., 1993. Socio-
economic aspects of charcoal consumption and environmental
consequences along the Dar es SalaasmMorogoro highway,
Tanzania. For. Ecol. Manage. 58, 249258.
Munyanziza, E., Oldeman, R.A.A., 1995. Pterocarpus angolensis
DC.: eld survival rates, growth, root pruning and fertilization
in the nursery. Fert. Res. 40, 235242.
Ndamba, N., Nyazema, N., Makaza, N., Kondera, K.C., 1994.
Traditional herbal remedies used for the treatment of urinary
schistosomiasis in Zimbabwe. J. Ethnopharmacol. 42, 125132.
Nyazema, N., Ndamba, N., Makaza, N., Kondera, K.C., 1994. The
doctrine of signatures or similitudes: a comparison efcacy of
praziquantel and traditional herbal remedies used for the
treatment of urinary schistosomiasis in Zimbabwe. Int. J.
Pharmacogn. 32, 142148.
268 M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269
Palgrave, K.C., 1981. Trees of Southern Africa. Struik Publishers
Ltd., Cape Town, 959 pp.
Rodgers, W.A., 1996. The miombo woodlands. In: McClanahan,
T.R., Young, T.P. (Eds.), East African Ecosystems and their
Conservation, Oxford University Press, New York, pp. 299
325.
Schwartz, M.W., Caro, T.M., unpublished data. Effect of selective
logging on tree and understory regeneration in miombo
woodland in western Tanzania. Afr. J. Ecol.
Stahle, D.W., Mushove, P.T., Cleaveland, M.K., Roig, F., Haynes,
G.A., 1999. Management implications of annual growth rings
in Pterocarpus angolesis from Zimbabwe. For. Ecol. Manage.
124, 217229.
Van Daalen, J.C., 1991a. Germination of Pterocarpus angolensis
seed. S. Afr. For. J. 158, 3336.
Van Daalen, J.C., 1991b. Forest growth: a 35-year southern cape
case study. S. Afr. For. J. 159, 110.
Van Daalen, J.C., Vogel, J.C., Malan, F.S., Fuls, A., 1992. Dating of
Pterocarpus angolensis trees. S. Afr. For. J. 162, 17.
Van der Reit, K., van Rensburg, L., De Sousa Correia, R.I., Mienie,
L.J., Kruger, G.H.J., 1998. Germination of Pterocarpus
angolensis DC. and evaluation of the possible antimicrobial
action of the phloem sap. S. Afr. J. Plant Soil 15, 141146.
Watt, J.M., Breyer-Brandwijk, M.G., 1962. The Medicinal and
Poisonous Plants of Southern and Eastern Africa. E&S
Livingstone Ltd., London.
M.W. Schwartz et al. / Forest Ecology and Management 170 (2002) 259269 269