The impact of fire and forest conversion into savanna on the bird communities of west Madagascan dry forests


Pons, P.; Wendenburg, C.

Animal Conservation 8(2): 183-193

2005


We studied the influence of vegetation structure and tree phenology on bird communities along a gradient of tropical forest degradation in NW Madagascar. Birds were censused by point counts at Ankarafantsika, one of the largest remnants of the severely reduced dry deciduous forest. The original forest was dominated by foliage insectivores. A few years after wildfire, regrowth was dense, most forest bird species were still present and additional understorey species appeared. As a result, species richness and abundance per point count increased. In contrast, when forest was transformed into savanna, the avian assemblage became poor, dominated by granivores and aerial insectivores, with only seven species shared with the original forest. Foliage volume, grass volume and bare ground cover explained most of the bird community variation by means of canonical correspondence analysis. Birds tended to increase their habitat breadth along the forest-savanna gradient. An index of bird conservation value, including abundance, endemism and the threatened status of the species, was highest in burned forests (1.12), intermediate in unburned forests (1) and lowest in savanna (0.44). The results emphasise the urgent need to protect not only the undisturbed fragments, but also the burned dry forests, because of their high value for biodiversity.

Animal
Conservation
(2005)
8,
183-193
©
2005
The
Zoological
Society
of
London.
Printed
in
the
United
Kingdom
doi:10.1017/S1367943005001940
The
impact
of
fire
and
forest
conversion
into
savanna
on
the
bird
communities
of
West
Madagascan
dry
forests
Pere
Pons
l,
*
and
Clara
Wendenburg
2
1
Departament
de
Ciencies
Ambientals,
Universitat
de
Girona,
Campus
de
Montilivi,
E-17071
Girona,
Catalonia,
Spain
2
Carrer
les
Mallorquines,
83,
E-17410
Sils,
Catalonia,
Spain
(Received
15
March 2004;
accepted
30
September
2004)
Abstract
We
studied
the
influence
of
vegetation
structure
and
tree
phenology
on
bird
communities
along
a
gradient
of
tropical
forest
degradation
in
NW
Madagascar.
Birds
were
censused
by
point
counts
at
Ankarafantsika,
one
of
the
largest
remnants
of
the
severely
reduced
dry
deciduous
forest.
The
original
forest
was
dominated
by
foliage
insectivores.
A
few
years
after
wildfire,
regrowth
was
dense,
most
forest
bird
species
were
still
present
and
additional
understorey
species
appeared.
As
a
result,
species
richness
and
abundance
per
point
count
increased.
In
contrast,
when
forest
was
transformed
into
savanna,
the
avian
assemblage
became
poor,
dominated
by
granivores
and
aerial
insectivores,
with
only
seven
species
shared
with
the
original
forest.
Foliage
volume,
grass
volume
and
bare
ground
cover
explained
most
of
the
bird
community
variation
by
means
of
canonical
correspondence
analysis.
Birds
tended
to
increase
their
habitat
breadth
along
the
forest—savanna
gradient.
An
index
of
bird
conservation
value,
including
abundance,
endemism
and
the
threatened
status
of
the
species,
was
highest
in
burned
forests
(1.12),
intermediate
in
unburned forests
(1)
and
lowest
in
savanna
(0.44).
The
results
emphasise
the
urgent
need
to
protect
not
only
the
undisturbed
fragments,
but
also
the
burned
dry
forests,
because
of
their
high
value
for
biodiversity.
INTRODUCTION
The
combination
of
high
levels
of
endemism
and
diversity
and
high
rates
of
habitat
loss
make
Madagascar
one
of
the
world's
conservation
priorities,
perhaps
the
most
important
(Myers
et
al.,
2000).
The
dry
deciduous
forests
of
western
Madagascar
in
particular,
which,
by
1990,
had
been
reduced
to
approximately
3%
of
their
original
extent
(Smith,
1997),
are
among
the
most
threatened
biomes
in
the
world.
Forest
remnants
continue
to
be
destroyed
by
slash-and-burn
practices
aimed
at
increasing
land
for
stockbreeding
and
cultivation.
If
abandoned,
the
cleared
areas
do
not
appear
to
be
able
to
recover
the
original
cover
(Ganzhorn
&
Sorg,
1996).
The
process
of
forest
degradation
often
starts
with
the
removal
of
the
biggest
and
most
valuable
trees
for
timber.
Forest
destru-
ction
culminates
in
the
formation
of
a
savanna
or
secondary
grassland,
which
is
regularly
burned
to
promote
the
growth
of
grass
shoots
for
cattle
grazing.
Given
certain
meteorological
conditions
and
vegetation
structure,
grassland
fires
spread
into
the
forest
turning
into
wildfires.
After
a
single
fire
many
trees
can
survive
and
the
forest
regenerates
vigorously
(Albignac
et
al.,
1992).
Close
to
savannas,
however,
the
forest
is
repeatedly
burned,
*All
correspondence
to
P.
Pons.
Tel:
(34)
972418269;
Fax:
(34)
972418150;
E-mail:
pere.pons@udg.es
becomes
poorer
and
shows
a
sparser
canopy
dominated
by
fire-resistant
trees.
The
West
Madagascan
dry
forest
is
one
of
the
218
Endemic
Bird
Areas
(EBAs)
in
the
world
(Stattersfield
et
al.,
1998)
and
one
of
the
EBAs
granted
critical
conservation
priority,
in
urgent
need
of
protected
areas.
Knowledge
of
the
avian
assemblages
of
this
EBA
has
increased
during
the
last
decade
(c.f.
for
example,
Andrianarimisa,
1993;
Urano
et
al.,
1994;
Yamagishi,
Urano
&
Eguchi,
1995).
However,
studies
into
the
impact
on
the
bird
community
of
deforestation
in
this
area
have
never,
to
our
knowledge,
been
published.
Research
undertaken
in
other
tropical
forests
may
provide
some
guidance
on
the
potential
responses
of
the
western
Mala-
gasy
bird
communities
to
forest
degradation.
Different
approaches
have
been
used
to
this
end.
Predictive
models
of
species
extinction
due
to
deforestation
have
been
proposed
in
different
regions
based
on
species—area
curves
(Reid,
1992).
Nevertheless,
estimations
of
species
loss
in
relation
to
the
losses
in
forest
area
usually
do
not
match
real
extinction
figures
due
to
time
lags
in
the
response
to
fragmentation
(Brooks,
Pimm
&
Collar,
1997;
Brooks, Pimm
&
Oyugi,
1999b;
Brooks,
Tobias
&
Balmford,
1999c).
Moreover,
fieldwork
evidence
at
the
local
scale
is
still
poor
in
most
regions.
The
impact
of
deforestation
has
also
been
assessed
by
comparing
bird
communities
in
different
sized
forest
fragments
to
those
village
Marovoay
road
N-4
study
area
secondary
road
0
Forest
*
Reserve
*
Strict
Nature
Reserve
Ampijoroa
Station
*
ANKARAFANTSIKA
0
5
km
=MO
184
P.
PONS
AND
C.
WENDENBURG
Fig.
1.
Map
of
the
Ankarafantsika
region
showing
the
extent
of
the
reserves
and
the
location
of
the
study
areas
(stars),
which
contained
three
to
nine
sampling
sites
each.
Although
recent
cartography
of
burned
areas
and
savanna
was
not
available,
sampling
sites
were
interspersed
among
the
three
habitat
categories.
The
square
in
the
insert
shows
the
location
of
Ankarafantsika
within
Madagascar.
found
in
continuous
forests
(Stouffer
&
Bierregaard,
1995;
Jeffrey
&
Stouffer,
1999).
This
approach
includes
studies
in
East
Madagascan
humid
forests
(Raherilalao,
2001).
A
third
procedure
is
to
compare
the
avian
assemblage
of
undisturbed
forest
to
those
of
disturbed,
degraded
or
destroyed
forest
(Kofron
&
Chapman,
1995).
Whatever
the
approach
used,
it
appears that
most
forest
bird
species are
locally
extirpated
by
deforestation
(e.g.
70%
in
the
Upper
Guinea
rainforest:
Kofron
&
Chapman,
1995;
or
74%
of
terrestrial
insectivorous
in
the
Amazonian
rainforest:
Jeffrey
&
Stouffer,
1999).
Several
species,
including
endemics,
are
nevertheless
able
to
adapt
to
the
secondary
growth
that
follows
deforestation,
especially
in
dry
forests
(Brooks
et
al.,
1999a).
When
tropical
forests
are
not
destroyed
but
suffer
damage
from
single
understorey
wildfires,
the
consequence
to
the
bird
community
can
also
be
severe
(Barlow,
Haugaasen
&
Peres,
2002).
While
most
research
on
the
faunal
effects
of
defo-
restation
has
focused
on
forest
fragments,
our
aim
was
to
address
the
species-loss
problem
by
comparing original
continuous
forest
to
burned
areas
of
forest
and
to
the
surrounding
deforested
area.
By
means
of
sampling
birds
and
vegetation
in
one
of
the
biggest
remnants
of
western
dry
deciduous
forest
and
in
adjacent
savannas,
the
objectives
of
this
study
were
to
assess:
(1)
the
bird
community
patterns
in
undisturbed
and
disturbed
forests
and
in
savanna
and
their
relationship
with
habitat
structure,
(2)
the
loss
of
bird
species
when
forests
are
affected
by
wildfires
or
turned
into
savanna,
(3)
the
habitat
breadth
of
endemic
and
non-endemic
birds
in
the
gradient
of
forest
disturbance
and
(4)
the
conservation
value
of
the
bird
community
along
this
gradient.
METHODS
Study
region
The
Ankarafantsika
region
is
a
plateau
found
in
NW
Madagascar,
roughly
100
km
SE
of
Mahajanga.
The
region
hosts
two
adjacent
protected
areas,
a
forest
reserve
and
a
Strict
Nature
Reserve,
covering
some
140
000
ha
(Fig.
1).
The
headquarters
of
both
reserves
are
located
in
Ampijoroa
Forest
Station
(16°19'S,
46°45'E,
75
m
above
sea
level
(a.s.l.)).
A
National
Park
is
foreseen
for
the
region
in
the
near
future.
The
climate
is
monsoonal,
with
an
annual
rainfall
of
1507
mm
concentrated
in
the
November
to
March
period.
The
vegetation
consists
of
dry
deciduous
forests
and
savannas.
The
most
extensive
vegetation
type
in
the
area
is
plateau
forest
on
sandy
soil,
characterised
by
trees
of
the
genus
Dalbergia,
Com-
miphora,
Stereospermum,
Strychnos,
Hildegardia,
etc,
which
supports
a
rich
plant
community
(Albignac
et
al.,
1992).
Despite
being
under
legal
protection,
this
forest
suffers
from
several
forms
of
exploitation,
including
timber
extraction,
honey
and
plant
product
collection,
grazing
and
hunting.
Moreover,
wildfires
originating
from
adjacent
savannas
sometimes
spread
into
the
forest.
Forest
degradation
leads
to
a
savanna
dominated
by
the
grasses
Heteropogon
contortus
and
Hyparrhenia
rufa,
which
is
regularly
burned
to
improve
grazing.
Some
savannas
are
wooded
with
the
palms
Medemia
nobilis
and
Hyphaene
shatan
being
frequent,
together
with
Tamarindus
and
Zyziphus.
Sampling
design
Fourteen
study
areas
(Fig.
1),
consisting
of
three
to
nine
nearby
sampling
sites
each,
were
distributed
within
the
region.
This
totalled
73
sampling
sites
located
in
areas
of
homogeneous
vegetation
on
sandy
soil, according
to
three
broad
categories:
unburned
forest
(n
=
32),
burned
forest
(n
=
20)
and
savanna
(n
=
21).
The
extent
and
distribution
of
the
three
habitats
in
the
region
allowed
an
adequate
interspersion
of
sampling
sites
among
the
three
habitats
and
most
study
areas
contained
two
different
habitats.
Prior
to
site
selection
we
visually
confirmed
habitat
homogeneity
in
at
least
a
100-m
radius
in
forests
and
a
250-m
radius
in
savanna.
Unburned
forest
included
primary
as
well
as
lightly
exploited
or
once
disturbed
(more
than
15-year-old
evidence
of
fire)
plateau
forests.
Birds
and
forest
degradation
in
Madagascar
185
Table
1.
Variables
measured
in
the
field
and
combined
variables
used
in
the
Canonical
Correspondence
Analysis
(CCA)
Number
Initial
variables
CCA
variables
CODE
Foliage
cover
(%):
1
Layer
8-16
m
2
Layer
4-8
m
3
Layer
2-4
m
4
Layer
1-2
m
5
Layer
0.5-1
m
6
Layer
0.25-0.5
m
7
Layer
0-0.25
m
Grass
cover
(%):
8
Layer
1-2
m
9
Layer
0.5-1
m
10
Layer
0.25-0.5
m
11
Layer
0-0.25
m
12
Bare
ground
cover
(%)—
N
trees/100
m:
13
At
rest
14
With
flowers
15
With
leaves
16
Leaves
and
flowers
I
17
With
fruits
18
Leaves
and
fruits
19
Flowers
and
fruits
20
Leaves-flowers-fruits
21
Standing
dead
trees
I
22
Falling
snags
23
Canopy
height
(m)
24
Mean
dbh
(cm)
I
I
I
I
Log
(grass
volume)
Bare
ground
cover
N
trees
no
leaves
or
fruits
N
trees
leaves
no
fruits
N
trees
with
fruits
N
dead
trees
Log
(foliage
vol.)
high
layer
Log
(foliage
vol.)
low
layer
HIFO
LOFO
HERB
BARE
DECD
LEAV
FRUI
SNAG
Burned
forest
included
those
affected
by
recent
subcanopy
fires
(mostly
in
1987
and
in
1989),
which
behaved
as
crown
fires
locally
and
killed
part
of
the
trees.
All
burned
sites
were
therefore
located
in
an
initial
successional
stage
and
had
a
similar
vegetation
regrowth.
Savanna
included
fire-
dependent
secondary
grasslands
with
or
without
scattered
trees
and
palms.
Point
counts
for
birds
were
carried
out
at
each
sampling
site
in
which
24
environmental
variables
basically
habitat
structure
and
tree
phenology
(see
Table
1)
were
measured
to
interpret
the
patterns
of
avian
distribution
and
abundance.
Vegetation
structure
Vegetation
structure
was
measured
from
the
centre
of
the
sampling
site
within
an
area
of
about
1200
m
2
.
The
measured
area
was
enlarged
in
savanna
for
the
highest
vertical
layers,
in
order
to
take
the
uneven
distribution
of
trees
into
account.
We
estimated
foliage
and
grass
cover
separately
for
eight
vertical
layers
(0-0.25
m,
0.25-
0.5
m,
0.5-1
m,
1-2
m,
2-4
m,
4-8
m,
8-16
m,
16-32
m)
in
comparison
with
a
reference
chart
(Prodon
&
Lebreton,
1981).
In
order
to
assess
whether
tree
relative
density
and
phenology
could
influence
bird
species
composition
and
abundance
these
factors
were
studied
by
line
transects
(partially
following
Andrianarivo,
1993)
starting
at
the
centre
of
each
sampling
site
and
being
randomly
orientated.
All
trees
whose
ground
projection
intercepted
the
line
and
whose
stem
was
>
15
cm
girth
at
breast
height
(gbh),
i.e.
>
4.8
cm
diameter
at
breast
height
(dbh),
were
counted.
The
physiological
(living,
standing
dead
or
falling
snag)
and
phenological
condition
(presence/absence
of
leaves,
flowers
and
fruits)
of
these
trees
were
recorded
and
their
dbh
measured.
The
transect
length
was
30
m
in
forest
and
increased
to
100
m
in
savanna
to
account
for
the
variance
in
tree
density.
Bird
data
Bird
abundance
was
estimated
for
each
species
seen
or
heard
during
20-min
point
counts
with
unlimited
distance.
Counts
were
done
on
mornings
with
good
weather
conditions,
at
the
beginning
of
the
breeding
season,
from
18
October
to
16
November 1994.
The
detectability
bias
derived
from
sampling
different
vegetation
covers
(Bibby
&
Buckland,
1987)
was
considered
to
be
low
because
the
vast
majority
of
records
were
auditory.
However,
we
excluded
distant
vocalisations
when
possibly
emitted
from
a
different
habitat.
We
plotted
the
cumulative
number
of
point
counts
against
bird
species
richness
for
the
three
habitats
separately
to
evaluate
whether
our
sampling
effort
was
suitable
for
estimating
total
species
richness
(Walther
&
Martin,
2001).
The
sampling
effort
was
sufficient
in
savanna
and
unburned
forest.
However,
the
curve
for
burned
forest
did
not
reach,
although
nearly,
the
richness
asymptote.
Our
results
are
186
P.
PONS
AND
C.
WENDENBURG
therefore
conservative
regarding
species
number
in
burned
forest.
The
breeding
condition
of
birds
was
verified
by
the
observation
of
incubation
patches
in
birds
trapped
by
mist
nets
(from
9-17
November)
and
by
occasional
nest
finding.
Birds
were
classified
into
feeding
guilds
according
to
their
main
food
resource
and
foraging
technique
based
on
Langrand
(1990).
Seven
groups
were
distinguished:
frugivores,
granivores,
ground
insectivores,
canopy
and
understorey
insectivores,
aerial
insectivores,
ground
omnivores
and
carnivores.
Data
analysis
Canonical
Correspondence
Analysis
(CCA)
was
per-
formed
using
CANOCO-4
(ter
Braak
&
Smilauer,
1998)
to
relate
bird
community
composition
to
variation
in
the
environment.
The
species
matrix
consisted
of
52
bird
species
x
73
counts
and
the
environmental
matrix
of
24
variables
x
73
samples.
Due
to
the
relatively
high
num-
ber
of
variables
measured,
their
number
had
to
be
reduced
before
carrying
out
the
analysis
(ter
Braak,
1986;
Lebreton
et
al.,
1988).
Strongly
correlated
variables
were
removed
or
combined
into
eight
broader
variables
(Table
1).
Vertical
grass
cover
variables
were
combined
into
a
single
variable
for
grass
volume,
HERB,
calculated
as
the
sum
of
the
products
of
individual
cover
values
by
their
relative
height;
finally
this
was
log-transformed.
Similarly,
we
combined
vertical
foliage
cover
variables
into
a
low
layer
(0.25-2
m),
LOFO,
and
a
high
layer
(2-16
m),
HIFO,
logarithm
of
foliage
volume
variables.
Bare
ground
cover
remained
unchanged,
BARE.
Four
tree
categories
were
used
in
the
analysis:
number
of
fruiting
trees/100
m,
FRUI,
number
of
trees
with
leaves
but
no
fruits,
LEAV,
number
of
living
trees
without
leaves
and
fruits,
DECD,
and
number
of
dead
trees
and
snags,
SNAG
(Table
1).
The
significance
of
the
CCA
was
evaluated
using
Monte
Carlo
permutation
tests.
The
habitat
breadth
of
a
bird
species
on
the
main
gradient
was
then
defined
as
the
amount
of
dispersion
on
the
gradient
of
the
samples
in
which
the
species
occur
(Chessel,
Lebreton
&
Prodon,
1982).
Habitat
breadth
was
calculated
following
the
procedure
described
in
ter
Braak
&
Smilauer
(1998)
for
species
tolerance;
i.e.
the
standard
deviation
of
the
species
scores
on
the
main
gradient
(axis
1
of
CCA)
divided
by
(1
1/N
2
)
112
,
where
N2
is
the
effective
number
of
species
occurrences.
In
order
to
assess
the
conservation
value
of
the
three
habitat
types
for
birds
we
constructed
an
index
(CI:
Equation
1),
developed
from
Pons
et
al.
(2003a),
which
takes
into
account
the
status
and
abundance
of
the
species
recorded
in
the
point
counts
of
each
habitat.
Similar
indices
can
be
found
in
Spellerberg
(1992).
Abundance
was
log-transformed
to
balance
its
contribution
to
the
global
index
and
thus
avoid
a
disproportionate
weight
of
common
species.
The
specific
conservation
concern
(SPCC,)
value,
with
the
aim
of
reflecting
the
geographical
range
(sources:
Langrand,
1990;
Sinclair
&
Langrand,
1998)
and
threatened
status
(BirdLife
International,
2000,
updated
2004),
was
assigned
to
each
species
according
to
the
following
classification:
introduced
species
(IN)
=
1;
native
non-endemic
(NN)
=
1;
regional
endemic
(RE)
=
2;
Madagascar
endemic
(ME)
=
3;
Western
Madagascar
endemic
(WE)
=
6;
globally
vulnerable
western
endemic
(V-WE)
=
12;
globally
endangered
western
endemic
(E-WE)
=
24.
Regional
endemics
include
those
species
exclusive
of
Madagascar,
Comoro
and
Mascarene
islands.
CI
=
E
[
lo
g(
A
u
+
1)
x
SPCC
i
]lE
log(41
it
+
1)
(1)
i=i
i
=1
where
CI
=
conservation
value
index,
k
=
species
rich-
ness,
Au
=
mean
abundance
of
species
i
in
habitat
j
relative
to
10
point
counts,
Ai
t
=
mean
abundance
of
species
i
in
the
whole
sampling
relative
to
10
point
counts.
SPCCi
values
can
be
derived
from
the
species
status
given
in
Table
3,
below.
RESULTS
Vegetation
structure
The
mean
vegetation
structure
of
the
three
habitats
is
shown
in
Fig.
2.
Unburned
forests
had
a
modest
cover
in
the
lower
vertical
layers
and
a
peak
of
foliage
density
from
2-4
m.
Burned
forests
had
a
more
developed
undergrowth
reaching
a
peak
from
0.5-1
m.
Savannas
had
a
well-developed
grass
cover,
which
dominated
up
to
1
m
and
usually
a
scattered
tree
layer.
The
relative
tree
density
(trees
measuring
>
4.8
cm
dbh)
was:
18.8
7.1
standard
error
(s.e.))
intercepting
standing
trees
per
30
m
of
transect
in
unburned
forest,
8.9
trees
4.7
s.e.)
in
burned
forest
and
2.1
trees
(±4.1
s.e.)
in
savanna.
The
mean
dbh
of
standing
trees
was
9.7
cm
3.6
s.e.)
in
unburned
forest
transects
and
11.7
cm
4.9
s.e.)
in
burned
forest
transects.
Hence,
although
burned
forest
had
fewer
trees
(t
=
5.529,
P
<
.001)
they
were
larger
(almost
significantly:
t=
1.669,
P=
0.101)
than
in
unburned
forest,
owing
to
a
differential
age-related
mortality
due
to
fire.
The
proportion
of
living
trees,
standing
dead
trees
and
fallen
snags
was
94.1%,
1.8%
and
4.1%,
respectively,
in
unburned
forest
(N=
629).
The
corresponding
proportions
were
47.7%,
20.9%
and
31.4%
in
burned
forest
(N=
251).
These
proportions
were
significantly
different
between
habitats
(chi-square
=
253.4,
P
<
0.001).
Most
tree
species
are
deciduous
in
Ankarafantsika,
but
when
vegetation
was
surveyed
more
than
half
already
had
leaves
or
were
flowering
or
fruiting,
both
in
burned
and
unburned
forest
(Table
2).
Bird
community
Song
was
heard
for
all
territorial
songbirds.
Eleven
species,
out
of
13
analysed
species
captured
by
mist
nets,
showed
a
developed
incubation
patch.
Nests
containing
eggs
or
young
were
found
for
nine
species.
These
data
confirmed
that
sampling
was
performed
well
within
the
breeding
season.
A
total
of
1030
individuals
belonging
to
Birds
and
forest
degradation
in
Madagascar
187
Unburned
forest
(N=
32)
Burned
forest
(N=
20)
8-16
4-8
g
)
.,
2-4
1-2
0
0.5-1
&
0.25-0.5
>
0-0.25
E
8-16
4-8
>,
2-4
1-2
0
0.5-1
&
0.25-0.5
>
0-0.25
0
20
0
20
40
%
Cover
60
80
40
%
Cover
60
80
Savanna
(N=
21)
E
8-16
4-8
2-4
0
1-2
:
c
7
I
'
0.5-1
&
0.25-0.5
>
0-0.25
20
40
%
Cover
60
s0
Fig.
2.
Mean
vegetation
structure
of
point
counts
at
unburned
forest,
burned
forest
and
savanna.
Grey
=
grass
cover;
white
=
foliage
cover.
Bars
represent
standard
deviation
of
the
whole
cover
(grass
+
foliage).
Table
2.
Phenological
condition
of
trees
in
burned
and
unburned
forests
at
Ankarafantsika
from
mid-October
to
mid-November
1994
%
Unburned
forest
%
Burned
forest
(N
=
592
trees)
(N
=
119
trees)
Without
leaves
or
flowers
41.9
29.3
With
leaves
51.5
64.2
With
flowers
1.2
4.9
With
fruits
6.8
10.6
Percentages
added
together
exceed
100%
because
some
trees
belong
to
more
than
one
class
at
the
same
time.
52
species
(flight
observations
excluded)
was
recorded
in
point
counts
and
are
listed
in
Table
3.
Species
status
in
terms
of
endemicity
and
conservation
concern
as
well
as
total
abundance
and
relative
frequency
in
point
counts,
given
separately
for
the
three
habitats,
are
also
shown.
Surprisingly,
the
mean
species
richness,
abundance
and
Shannon
diversity
per
point
count
were
all
highest
in
burned
forest,
intermediate
in
unburned
forest
and
lowest
in
savanna
(Table
4).
Seven
of
the
52
bird
species
were
ubiquitous
(Fig.
3).
The
bulk
of
species
(n
=
21)
occurred
both
in
unburned
and
burned
forest.
Few
species
(n
=
3)
were
shared
by
savanna
and
burned
forest
and
none
by
savanna
and
unburned
forest.
Species
that
were
exclusive
to
one
habitat
were
a
majority
only
in
savanna
(11
out
of
21).
When
separated
into
feeding
guilds
(Table
3),
the
21
species
of
canopy
and
understorey
insectivores
con-
stituted
56.6%
of
the
overall
abundance.
The
abundance
of
the
seven
guilds
was
significantly
different
between
the
three
habitats
(chi-square
=
484.8,
P
<
0.001).
Canopy
and
understorey
insectivores
were
in
a
majority
both
in
unburned
and
burned
forest,
followed
by
frugivores
and
aerial
feeders
(Fig.
4).
Granivores
and
aerial
feeders,
however,
dominated
savanna,
followed
by
canopy
and
understorey
insectivores.
Bird
habitat
relationships
The
relationship
between
the
bird
species
and
the
set
of
environmental
variables
was
significant
both
for
the
first
axis
of
the
CCA
(Monte
Carlo
permutations
test,
F-ratio
=
12.41,
P
=
0.005)
and
for
the
overall
CCA
(F-ratio
=
3.20,
P
=
0.005).
Environmental
variables
explained
28.6%
of
the
total
variance
of
the
species
data,
as
interpreted
from
the
following
ratio:
sum
of
CCA
eigenvalues/sum
of
eigenvalues
of
the
unconstrained
correspondence
analysis
=
1.553/5.431
(Table
5).
The
first
axis
accounts
for
most
of
the
variation
in
the
species—environment
relationship
(56.8%).
This
axis
was
positively
correlated
to
the
logarithm
of
grass
volume
(HERB)
and
nega-
tively
to
the
logarithm
of
low
layer
foliage
volume
(LOFO),
the
logarithm
of
high
layer
foliage
volume
(HIFO)
and
the
bare
ground
cover
(BARE).
The
first
axis
therefore
represents
a
gradient
of
increasing
grass
cover
and
decreasing
bare
ground,
shrub
and
tree
cover,
from
unburned
forests
to
open
savannas
with
an
intermediate
position
for
burned
forests
and
wooded
savannas
(Fig.
5,
top).
The
second
axis
accounts
for
an
additional
14.6%
of
the
variation
in
the
species—environment
relation
(Table
5).
This
axis
shows
small
correlation
coefficients
with
the
eight
environmental
variables
and
has
a
mainly
spatial
basis.
Savanna
samples
are
separated
along
the
second
188
P.
PONS
AND
C.
WENDENBURG
Table
3.
Bird
species
status,
species
code,
total
abundance
(/V)
and
relative
frequency
(in
%)
in
point
counts
from
unburned
forest
(UF),
burned
forest
(BF)
and
savanna
(SV)
SPECIES
Code
Status
N
%
OF
%BF
%
SV
FG
Axisl
H
breadth
Bubulcus
ibis
BUIB
NN
11
0.0
0
14.3
GO
8
0.253
Milvus
aegyptius
MIAE
NN
2
0.0
0
9.5
CA
1
-
Aviceda
madagascariensis
AVMA
ME
1
0.0
5
0.0
CA
52
-
Polyboroides
radiatus
PORA
ME
1
0.0
5
0.0
CA
44
-
Buteo
brachypterus
BUBR
ME
9
3.1
35
0.0
CA
38
0.098
Falco
newtoni
FANE
RE
2
0.0
10
0.0
GI
29
-
Coturnix
delegorguei
CODE
NN
7
0.0
0
19.0
GG
2
0.449
Numida
meleagris
NUME
IN
2
3.1
5
0.0
GO
48
-
Mesitornis
variegate
MEVA
V-WE
10
15.6
10
0.0
GO
33
0.125
Turnix
nigricollis
TUNI
ME
2
0.0
10
0.0
GG
26
-
Pterocles
personates
PTPE
WE
7
0.0
0
19.0
GG
3
0.905
Streptopelia
picturata
STPI
ME
6
3.1
20
4.8
GG
16
0.417
Oena
capensis
OECA
NN
5
0.0
0
14.3
GG
6
0.474
Treron
australis
TRAU
ME
5
0.0
5
9.5
FR
15
0.969
Coracopsis
vasa
COVA
RE
31
53.1
45
0.0
FR
37
0.105
Coracopsis
nigra
CONI
RE
2
0.0
5
0.0
FR
14
-
Agapornis
cana
AGCA
ME
3
0.0
0
9.5
GG
11
-
Cuculus
rochii
CURO
ME
10
18.8
15
4.8
UI
22
0.393
Coua
coquereli
COCO
WE
14
18.8
25
0.0
GI
40
0.073
Coua
ruficeps
CORU
WE
7
15.6
0
0.0
GI
17
0.133
Coua
cristata
COCR
ME
39
53.1
60
0.0
UI
25
0.195
Centropus
toulou
CETO
RE
19
18.8
45
4.8
UI
18
0.313
Caprimulgus
madagascariensis
CPMA
RE
3
3.1
5
0.0
AI
42
-
Merops
superciliosus
MESU
NN
56
0.0
15
90.5
AI
9
0.880
Eurystomus
glaucurus
ERGL
NN
14
9.4
50
0.0
AI
28
0.284
Leptosomus
discolor
LEDI
RE
15
25.0
15
0.0
UI
41
0.109
Upupa
epops
UPEP
NN
10
18.8
15
0.0
GI
51
0.102
Mirafra
hova
MIHO
ME
41
0.0
0
76.2
GG
5
0.422
Motacilla
finviventris
MOFL
ME
3
0.0
0
9.5
GI
13
-
Coracina
cinerea
COCI
ME
14
18.8
25
0.0
UI
39
0.090
Bernieria
madagascariensis
BEMA
ME
22
31.3
20
0.0
UI
32
0.120
Hypsipetes
madagascariensis
HYMA
RE
98
81.3
90
19.0
FR
20
0.413
Copsychus
albospecularis
COAL
ME
26
28.1
45
0.0
UI
27
0.159
Nesillas
typica
NETY
RE
23
0.0
75
0.0
UI
30
0.190
Cisticola
cherina
CICH
RE
23
0.0
0
52.4
UI
4
0.430
Newtonia
brunneicauda
NEBR
ME
121
100.0
95
0.0
UI
31
0.160
Neomixis
tenella
NETE
ME
44
37.5
65
4.8
UI
21
0.275
Tersiphone
mutata
TEMU
RE
35
59.4
30
0.0
AI
36
0.113
Nectarinia
souimanga
NESO
RE
162
93.8
100
14.3
UI
23
0.312
Nectarinia
notata
NENO
RE
3
3.1
5
0.0
UI
50
-
Zosterops
maderaspatanus
ZOMA
RE
3
6.3
5
0.0
UI
24
0.250
Schetba
rufa
SCRU
ME
17
21.9
15
0.0
UI
46
0.075
Vanga
curvirostris
VACU
ME
6
15.6
0
0.0
UI
43
0.093
Xenopirostris
damii
XEDA
E-WE
16
28.1
25
0.0
UI
35
0.178
Fakulea
palliata
FAPA
ME
10
12.5
15
0.0
UI
34
0.126
Leptopterus
viridis
LEVI
ME
2
3.1
5
0.0
UI
49
-
Leptopterus
chabert
LECH
ME
7
0.0
20
0.0
UI
47
0.113
Cyanolanius
madagascarinus
CYMA
ME
1
3.1
0
0.0
UI
45
-
Dicrurus
forficatus
DIFO
RE
27
31.3 45
9.5
AI
19
0.307
Corvus
albus
CRAL
NN 3
0.0
0
14.3
GO
7
0.520
Foudia
madagascariensis
FOMA
ME
20
0.0
5
38.1
GG
12
0.718
Lonchura
nana
LONA
ME
10
0.0
0
14.3
GG
10
0.378
Range
Min:
1
0 0 0
-
1
0.073
Max:
162
100
100
90.5
-
52
0.969
Point
counts
for
each
habitat,
were:
UF,
n
=
32;
BF,
n
=
20;
SV,
n
=
21.
FG,
feeding
guild;
FR,
frugivores;
GG,
ground
granivores;
GI,
ground
insectivores;
UI,
canopy
and
understorey
insectivores;
AI,
aerial
insectivores;
GO,
ground
omnivores;
CA,
carnivores.
Axis
1
means
the
species
position
in
the
main
gradient,
increasing
from
open
savanna
to
forest.
H
breadth
is
the
habitat
breadth
on
the
first
axis
of
the
CCA
for
species
with
frequency
>
2
in
the
total
sample.
Status:
IN,
introduced
species;
NN,
native
non-endemic;
RE,
regional
endemic;
ME,
Madagascar
endemic;
WE,
Western
Madagascar
endemic;
V,
globally
vulnerable;
E,
globally
endangered
according
to
IUCN
Redlist.
Birds
and
forest
degradation
in
Madagascar
189
Table
4.
Bird
species
richness,
abundance
and
Shannon
diversity
(calculated
in
log
10)
in
point
counts
of
unburned
forest,
burned
forest
and
savanna
Ecological
parameters
Unburned
forest
Burned
forest
Savanna
(point
counts
mean
±
s.d.)
(rt
=
32)
(n
=
20)
(n
=
21)
Species
richness
8.3
+3.1
10.9
±
3.3
4.5
±
2.6
Abundance
13.8
±
4.5
18.9
±
6.6
10.0
±
5.5
Shannon
diversity
0.89
f
0.31
1.16
f
0.36
0.53
±
0.26
Table
5.
Summary
of
canonical
correspondence analysis
of
bird
abundance
from
point
counts
at
Ankarafantsika
with
environmental
variables
Axes
1
2
3
4
Eigenvalues
0.882
0.226
0.113
0.099
Species-environment
correlation
0.986
0.816
0.779
0.693
Cumulative
percentage
variance
of
species-environment
relation
56.8
71.4
78.7
85.0
Sum
of
unconstrained
eigenvalues
5.431
Sum
of
canonical
eigenvalues
1.553
Correlation
BARE-species'
axes
-
0.676
-
0.104
-
0.323
-
0.345
Correlation
HERB-species'
axes
0.926
0.273
0.004
0.035
Correlation
LOFO-species'
axes
-
0.949
0.192
0.073
-
0.013
Correlation
HIFO-species'
axes
-
0.845
-
0.003
-
0.267
0.183
Correlation
SNAG-species'
axes
-
0.311
-
0.021
0.297
-
0.202
Correlation
DECD-species'
axes
-
0.412
-
0.146
-
0.446
0.106
Correlation
LEAV-species'
axes
-
0.350
-
0.044
-
0.271
0.040
Correlation
FRUI-species'
axes
-
0.250
-
0.022
-
0.229
0.025
See
Table
1
for
the
meaning
of
the
variable
acronyms.
GO
CA
co
<9.
,
7
0
".%
o"
6
11
3
7
o
N
Al
GO
10%
2%
FR
1
3
%
FR
AI
1%
15%
16%
GG
10%
GI
2%
5%
GI
3
%
7
UI
T
'w
Unburned
UI
Burned
0
21
68%
forest
65%
forest
CA
3
1%
GO
FR
7
%
4%
Al
U
nburned
Forest
26%
GG
45%
Fig.
3.
Distribution
of
the
bird
species
richness
among
habitats.
UI
GI
16%
1%
Savanna
axis
after
their
location,
the
mean
distances
between
the
three
sites
being
between
9
and
16
km.
Finally,
unburned
and
burned
forests
are
also
separated
by
this
second
axis
but
to
a
lesser
extent;
burned
samples
having,
in
general,
more
positive
scores
(Fig.
5,
top).
Species
are
well
separated
on
the
biplot
of
the
first
two
axes
(Fig.
5,
bottom),
the
savanna
specialists
being
Coturnix
delegorguei,
Pterocles
personatus,
Cisticola
cherina,
Mirafra
hova,
Merops
superciliosus,
etc.
Fig.
4.
Proportion
ofbirds
from
the
seven
feeding
guilds
in
unburned
forest,
burned
forest
and
savanna.
FR,
frugivores;
GG,
ground
granivores;
GI,
ground
insectivores;
UI,
canopy
and
understorey
insectivores;
AI,
aerial
insectivores;
GO,
ground
omnivores;
CA,
carnivores.
Representative
species
of
unburned
forests
are
Mesitornis
variegata,
Tersiphone
mutata,
Leptosomus
discolor,
Coua
ruficeps
and
Bernieria
madagascariensis.
Those
DIFO
ERGL
XEDA
CURO
STPI
COCR
HYMA
VACU
NETY
NESO
UPEP
NETY
SCRUCOCI
NEBR
CPMA
COAL
NENu
FAPA
BEMA
COCOTEMUFANE
CORU
LECH
MEVA
LEVI
TUNI
MOFL
FOMA
AGCA
herb
LONA
CON
I
ZOMA
CETO
TRAU
lofo
hifo
41-
1TAg
bare
leaf
decd
CYMA
MESU
MIHO
CRAL
CICH
PTPE
BUIB
OECA
MIAE
CODE
P.
PONS
AND
C.
WENDENBURG
190
r
I
I
-
th
r
-,
,
-
D
.
e
"
,a
n
I
C
l i
i
•,,,
to
-
0
in
j
o
ru
0
0
I
4
herb
_
lofo
hifo
4—
bare
decd
iloP
_
0
_
0
0
. .
.
.
,
.
.
F2
F2
Fig.
5.
Biplots
of
the
first
two
axes
of
the
Canonical
Correspondence
Analysis
(CCA).
The
insert
squares
(broken
lines)
show
the
core
area
of
each
biplot;
the
largest
square
shows
this
area
enlarged.
Top:
samples
and
environmental
variables
(abbreviated
as
in
Table
1).
Symbols
are:
squares,
unburned
forests;
triangles,
burned
forests;
circles,
savanna
(open,
filled
and
mid-filled
circles
separate
the
savanna
samples
into
three
study
locations,
namely
Mahatazana,
Besaboabe
and
Ambitsika,
respectively).
Bottom:
species
(abbreviated
as
in
Table
3)
and
environmental
variables.
AVMA
is
hidden
under
SCRU,
PORA
under
CPMA,
BUBR
under
COCI,
NUME
under
LEVI,
COVA
under
TEMU
and
LEDI
under
COCO.
of
burned
forests
are
Centropus
toulou,
Eurystomus
account
for
less
than
8%
of
the
variation
in
the
species—
glaucurus,
Streptopelia
picturata,
Nesillas
typica
and
environment
relationship
(Table
5).
The
variable
most
Newtonia
brunneicauda.
The
third
and
successive
axes
correlated
to
the
third
axis
was
the
number
of
living
trees
F1
F1
Birds
and
forest
degradation
in
Madagascar
191
without
leaves
and
fruits
(DECD),
while
for
the
fourth
axis
it
was
the
bare
ground
cover
(BARE).
The
importance
of
tree
phenology
could
be
concealed
by
the
intensity
of
the
forest—savanna
gradient,
best
reflected
by
structural
variables.
In
order
to
avoid
this
effect,
the
CCA
was
repeated
excluding
savanna
samples
and
those
bird
species
occurring
only
in
savanna.
The
analysis
remained
significant
(Monte
Carlo
permutations
test,
F-ratio
=
1.388,
P
=
0.01),
but
the
only
variable
of
tree
condition
that
correlated
strongly
to
the
ordination
axes
was
the
number
of
dead
trees
and
snags
(SNAG,
r
=
0.560).
Bird
conservation
and
endemicity
Restricted
range
(W
Madagascar)
endemics
were
re-
presented
by
four
species
in
unburned
forest
(Mesitornis
variegata,
Xenopirostris
damii,
Coua
coquereli
and
Coua
ruficeps)
three
species
in
burned
forest
(Mesitornis
variegata,
Xenopirostris
damii
and
Coua
coquereli)
and
only
one
in
savanna
(Pterocles
personatus).
Three
additional
island
or
regional
endemics
of
global
conservation
concern
were
observed,
but
not
in
standard
censuses:
Ardea
humbloti,
Lophotibis
cristata
and
Philepitta
schlegeli.
The
impact
of
a
single
fire
on
forest
endemics
was
analysed
for
the
12
species
belonging
to
an
endemic
family
(Mesitornithidae,
Leptosomatidae
and
Vangidae)
or
subfamily
(Couinae)
of
the
Madagascar
region.
The
occurrence
of
Coua
ruficeps
and
Vanga
curvirostris
was
higher
in
unburned
than
in
burned
forests
over
the
total
of
52
point
counts
(Fisher
exact
test,
P
=
0.077,
for
both
species).
The
occurrence
of
Leptopterus
chabert
was
higher
in
burned
forests
(Fisher
exact
test,
P
=
0.018).
The
rest
of
the
species
showed
a
non-significant
pattern
in
relation
to
fire.
When
forest
is
transformed
into
savanna
the
conservation
index
(Equation
1)
decreases
from
79.9
to
35.5
(—
55.6%).
In
contrast,
this
index
increases
several
years
after
a
single
fire
to
reach
89.4
(+
11.9%).
Birds tend
to
increase
their
habitat
breadth
along
the
main
gradient
from
forest
to
savanna
(habitat
breadth
versus
scores
on
axis
1
of
CCA,
r
s
=
0.909,
P
<
0.001,
n
=
39
species
with
frequency
>
2
in
point
counts).
However,
there
are
no
significant
differences
between
Madagascar
endemics
(categories
WE
+
ME)
and
non-endemics
(categories
RE
+
NN
+
IN)
concerning
ecological
optimum
on
the
main
gradient
(median
score:
0.466
versus
0.423,
U=
271.5,
P
=
0.253,
n
=
52
species)
or
habitat
breadth
(median
breadth:
0.159
versus
0.307,
U
=
139,
P
=
0.174,
n
=
39
species).
DISCUSSION
Role
of
the
habitat
features
on
bird
community
We
have
analysed
the
gradient
of
forest
degradation
by
means
of
the
joint
sampling
of
vegetation
structure
and
bird
communities
and,
for
the
first
time
to
our
knowledge,
quantified
the
ecological
optimum
and
habitat
breadth
of
western
Madagascar
birds
along
this
gradient.
Because
of
the
difficulty
in
finding
information
on
fire
ageing
and
cartography
it
was
not
possible
to
reconstruct
the
post-fire
succession.
We
focused
therefore
on
three
distinct
habitats
(canopy-rich
forest,
understorey-dense
burned
forest
and
herbaceous
savanna),
despite
within-habitat
structural
variation
derived
from
local
differences
in
intensity
and
frequency
of
disturbances.
Such
a
categorical
approach
may
limit
the
comprehension
of
the
consequences
of
the
degradation
process.
However,
these
habitats
were
the
most
extensive
in
the
region
and
are
representative
of
extreme
disturbance
regimes.
Moreover,
species
richness
patterns
in
habitats
seem
to
be
not
scale
dependent,
since
both
point
counts
(small
scale)
and
the
whole
study
area
(large
scale)
give
the
same
consistent
results.
With
regard
to
the
diversity
of
structures,
it
is
not
surprising
that
40%
of
the
bird
species
recorded
in
point
counts
were
exclusive
to
one
habitat,
mostly
to
savanna.
Unburned
and
burned
forests,
however,
shared
another
40%
of
species.
Edge
effects
on
our
avian
results
can
be
ignored
because
of
the
extension
of
habitats
and
the
spatial
design
of
our
sampling.
The
structure
of
the
bird
community
was
related,
at
the
scale
of
our
study,
to
the
density
of
the
different
vegetation
layers
(grass,
shrub
and
tree
foliage)
and
to
the
bare
ground
cover.
These
explained
the
distribution
of
savanna
and
forest
specialists;
the
latter
separated
into
ground,
understorey
and
canopy
dwellers.
The
variables
relating
to
tree
condition
(phenology
and
mortality)
appeared
to
be
secondary,
even
when
the
CCA
was
repeated
with
just
the
forest
samples.
Thus,
only
the
density
of
trees
without
leaves
and
fruits,
higher
in
unburned
forests,
and
the
density
of
dead
trees
and
snags,
higher
in
burned
forest,
had
some
influence
on
avian
gradients.
This
contrasts
with
other
seasonal
tropical
forests
in
which
bird
abundance
appears
to
depend
on
the
irregular
spatial
distribution
of
plant
resources
(Poulin,
Lefebvre
&
McNeil,
1992,
1993).
In
general,
nectarivores
and
frugivores
are
the
more
nomadic
feeding
guilds,
spatially
associated
with
the
flowering
and
fruiting
activity
(Loiselle
&
Blake,
1991),
which
can
be
high
in
burned
forests.
Strict
nectar-eaters
were,
however,
absent
from
Ankarafantsika
and
fruit-
eaters
barely
reached
16%
of
the
bird
abundance
in
forests.
So
then,
the
scarcity
of
transient
species,
together
with
the
spatial
evenness
of
tree
seasonality,
helps
to
explain
why
the
occurrence
of
leaves,
flowers
and
fruits
in
trees
had
a
tiny
influence
on
bird
patterns.
The
impact
of
wildfire
As
explained
before,
species
richness
in
burned
forests
was
probably
underestimated.
In
spite
of
this,
burned
forest
contained
almost
all
of
the
bird
species
found
in
unburned
forest
plus
10
additional
species.
The
richness
of
burned
forests
does
not
arise
from
point
counts
being
distributed
in
covers
of
slightly
different
ages
because
mean
abundance
and
species
number
per
point
count
were
also
27%
and
24%
higher,
respectively,
after
wildfire.
The
vegetation
structure
was
also
relatively
similar
between
sites
in
burned
forests,
as
inferred
from
the
standard
deviation
bars
in
Fig.
2.
An
increased
diversity
is,
in
fact,
192
P.
PONS
AND
C.
WENDENBURG
not
exceptional
in
burned
forests
(Raphael,
Morrison
&
Yoder-Williams,
1987;
Woinarski,
1990),
especially
after
uneven
fires
that
leave
behind
heterogeneous
cover
(Prodon
&
Pons,
1993;
Schulte
&
Niemi,
1998),
but
may
conceal
potential
demographic
problems
for
certain
bird
species.
In
Madagascan
dry
forests
this
is
exemplified
by
the
low
productivity
and
group
size
of
Mesitornis
variegata
after
fire
(Hawkins,
1994),
slowing
down
the
recovery
of
the
population.
Other
moderate
disturbances,
such
as
low-intensity
logging,
were
positive
for lemur
abundance
(Ganzhorn,
1995)
and
neutral
for
bird
species
richness
and
abundance
(Hawkins
&
Wilme,
1996)
in
the
same
ecosystem.
Burned
forests
in
Ankarafantsika
corresponded
to
an
initial
stage
of
the
post-fire
succession
(5-7
years
in
general)
and
were
largely
the
result
of
subcanopy
wildfires
that
left
numerous
large
trees
alive.
This
habitat
shows
a
dense
regrowth
combined
with
a
sparse canopy
that
appears
suitable
both
for
shrub
dwelling
and
for
most
strictly
forest
birds,
including
canopy
and
several
ground-foragers.
As
a
result,
the
relative
abundance
of
feeding
guilds
did
not
vary
between
burned
and
unburned
forest.
The
forest
preference
of
endemic
bird
species
in
relation
to
fire
coincides
with
their
known
habitat
use.
Thus,
Coua
ruficeps
is
the
most
terrestrial
coua
in
Ankarafantsika
(Urano
et
al.,
1994)
and
occurred
exclusively
in
unburned
forest
in
which
the
ground
is
mostly
uncovered.
Vanga
curvirostris,
which
mainly
forages
in
tree
leaves,
twigs
and
branches
between
5
and
20
m
was
also
found
only
in
unburned
forests
(Yamagishi
&
Eguchi,
1996).
Finally,
the
selection
of
burned
forest
by
Leptopterus
chabert may
be
connected
to
its
foraging
activity
in
the
lower
layers
of
the
forest
(Yamagishi
&
Eguchi,
1996),
which
is
densely
grown
several
years
after
fire.
The
impact
of
forest
conversion
into
savanna
Forest
degradation
to
generate
secondary
grassland
impoverishes
the
bird
community.
More
than
three-
quarters
of
the
pool
of
31
species
found
in
unburned
forests
did
not
occur
in
savanna.
Opportunistic
and
open-
habitat
birds
replace
only
one
half
of
the
lost
forest
species.
Subsequently,
the
savanna
avian
assemblage
is
poor
and
mostly
banal,
with
the
single
restricted
range
endemic
Pterocles
personatus.
The
mean
abundance
and
species
richness
per
point
count
are
similarly
reduced.
Linked
to
the
drastic
change
in
habitat
structure,
the
proportion
of
feeding
guilds
is
strongly
modified.
The
forest
community
was
dominated
by
canopy
and
understorey
insectivores
and
by
frugivores,
mostly
foraging
in
branches
and
foliage.
When
degraded,
it
was
dominated
by
ground-foraging
granivores
and
insectivores
that
capture
prey
on
the
wing.
At
the
regional
scale,
deforestation
allows
some
forest
remnants
to
subsist
in
the
matrix
of
grazed
savanna.
As
is
known,
natural
fragments
suffer
from
a
gradual
loss
of
species
from
the
onset
of
fragmentation;
but
the
magnitude
and
timing
of
such
phenomena
have
not
yet
been
evaluated
in
West
Madagascan
dry
forest.
Once
savanna
is
established,
the
bird
community
does
not
seem
to
be
negatively
affected
by
burning
at
Ankarafantsika
(Pons,
Rakotobearison
&
Wendenburg,
2003b).
However,
burning
savanna
for
grazing
purposes
is
sometimes
the
cause
of
large
forest
fires.
Conservation
management
recommendations
Among
the
different
forms
of
dry
forest
exploitation,
the
conversion
into
pastureland
has
the
most
drastic
effect,
leading
to
extensive
forest
disappearance,
which
has
been
going
on
for
many
years
(Humbert,
1927).
Ankarafantsika
Reserve
suffers
more
subtle
exploitation.
Despite
the
legal
interdiction,
people
penetrate
the
forests
to
obtain
products
such
as
honey,
fruits and
tubers,
to
hunt
lemurs
and
other
vertebrates
or
to
cut
down
the
most
valuable
trees
and
palms
(Garcia
&
Goodman,
2003).
There
are
also
permanent
settlements
inside
the
reserve
around
which
small
areas
are
deforested
for
agriculture.
The
apparently
quick
recovery
of
the
forest
after
a
wildfire
means
threatened
endemics,
associated
with
a
high
avian
conservation
value,
can
be
maintained.
Specific
studies
are,
however,
needed
to
detect
whether
demographic
problems,
such
as
those
already
mentioned,
may
turn
recently
burned
forests
into
population
sinks.
In
contrast,
the
conservation
value
of
savannas
is
much
lower.
The
poorness
of
the
bird
assemblage,
specifically,
tends
to
support
the
view
of
a
human-mediated
expansion
of
savanna
in
Madagascar
from
a
small
original
area
(Langrand,
1990);
the
possibility
of
reconverting
savanna
into
forest
has
not
yet
been
evaluated
but
is
unlikely
on
human
timescales.
The
regeneration
of
burned
forests,
however,
seems
possible.
These
forests
merit
special
protection
because
they
can
occupy
large
areas
regionally,
serve
as
corridors
between
undisturbed
forests
and
hold
a
high
conservation
value.
Conservation
should
focus
on
avoiding
repeated
fires
and
timber
extraction
which
lead
to
further
degradation
and
on
allowing
the
natural
regeneration
to
take
place.
Modelling
the
effects
of
fragmentation
of
western
dry
forest
on
the
biota
should
be
a
main
goal
of
future
research.
Considering
that
Ankarafantsika
is
one
of
the
largest
remaining
tracts
of
this
forest
and
holds
the
main
world
population
of
threatened
endemics
such
as
Mesitornis
variegata
and
Xenopirostris
damii
(BirdLife
International,
2000),
its
careful
conservation
should
be
a
priority.
In
a
landscape
perspective,
the
connection
of
this
massif
to
smaller
surrounding
forests
could
allow
the
recovery
of
diversity
in
such
fragments.
Acknowledgements
Thanks
are
due
to
Germain
Rakotobearison,
Rosa
Wendenburg
and
Montse
Lopez
for
their
help
during
the
study
and
to
Brigitte
Poulin,
Frank
Hawkins
and
two
anonymous
referees
for
suggestions
on
a
previous
draft.
The
Direction
des
Eaux
et
Forets
(Antananarivo)
allowed
us
to
work
on
the
Strict
Reserve.
Laboratoire
Arago,
Bird
Exploration
Fund,
Rvd.
Streeter
C.S.
and
Gallina
Blanca
have
all
given
their
support.
Birds
and
forest
degradation
in
Madagascar
193
REFERENCES
Albignac,
R.,
Ramangason,
G.
S.,
Andriatsarafara,
S.,
Roger,
E.
&
Rakotofiringa,
S.
(1992).
Eco-developpement
des
communautes
rurales
pour
la
conservation
de
la
biodiversity.
Antananarivo:
Tsipika
Editeur.
Andrianarimisa,
A.
(1993).
Variation
structurale
de
la
communaute
avienne
d'une
foret
dense
seche
semi-caducifoliee
dans
l'Ouest
malgache.
In
Proceedings
of
the
VIII
Pan-African
Ornithological
Congress
Bujunbura:
187-191.
Annales
du
Musee
Royal
de
I
'
Afrique
Centrale
(Zoologie),
268.
Tervuren,
Belgium.
Andrianarivo,
J.
(1993).
Using
GIS
to
evaluate
the
crown-line
intersect
sampling
method
in
forest
survey.
For.
Ecol.
Mgmt.
59:
87-103.
Barlow,
J.,
Haugaasen,
T.
&
Peres,
C.
A.
(2002).
Effects
of
ground
fires
on
understorey
bird
assemblages
in
Amazonian
forests.
Biol.
Conserv.
105:
157-169.
Bibby,
C.
J.
&
Buckland,
S.
T.
(1987).
Bias
of
bird
census
results
due
to
detectability
varying
with
habitat.
Acta
Oecol.
8:
103-112.
BirdLife
International
(2000).
Threatened
birds
of
the
world.
Barcelona
and
Cambridge:
Lynx
Edicions
and
BirdLife
International.
Brooks,
T.
M.,
Pimm,
S.
L.
&
Collar,
N.
J.
(1997).
Deforestation
predicts
the
number
of
threatened
birds
in
Insular
Southeast
Asia.
Conserv.
Biol.
11:
382-394.
Brooks,
T.
M.,
Pimm,
S.
L.,
ICapos,
V
&
Ravilious,
C.
(1999a).
Threat
from
deforestation
to
montane
and
lowland
birds
and
mammals
in
insular
South-east
Asia.
J.
Anim.
Ecol.
68:
1061-1078.
Brooks,
T.
M.,
Pimm,
S.
L.
&
Oyugi,
J.
0.
(1999b).
Time
lag
between
deforestation
and
bird
extinction
in
tropical
forest
fragments.
Conserv.
Biol.
13:
1140-1150.
Brooks,
T.,
Tobias,
J.
&
Balmford,
A.
(1999c).
Deforestation
and
bird
extinctions
in
the
Atlantic
forest.
Anim.
Conserv.
2:
211-222.
Chessel,
D.,
Lebreton,
J.-D.
&
Prodon,
R.
(1982).
Mesures
symetriques
d'
amplitude
d'habitat
et
de
diversite
intra-echantillon
dans
un
tableau
especes-releves:
cas
d'un
gradient
simple.
C.
R.
Acad.
Sci.
Paris,
Serie
III
295:
83-88.
Ganzhom,
J.
U.
(1995).
Low-level
forest
disturbance
effects
on
primary
production,
leaf
chemistry,
and
lemur
populations.
Ecology
76:
2084-2096.
Ganzhom,
J.
U.
&
Sorg,
J.-P.
(Eds).
(1996).
Ecology
and
economy
of
a
tropical
dry
forest
in
Madagascar.
Primate
Rep.,
Spec.
Iss.
46-1:
1-382.
Garcia,
G.
&
Goodman,
S.
M.
(2003).
Hunting
of
protected
animals
in
the
Parc
National
d'Ankarafantsika,
north-western
Madagascar.
thyx
37:
115-118.
Hawkins,
A.
E
A.
(1994).
Conservation
status
and
regional
population
estimates
of
the
white-breasted
mesite
Mesitornis
variegata,
a
rare
Malagasy
endemic.
Bird
Conserv.
Int.
4:
279-303.
Hawkins,
A.
E
A.
&
Wilme,
L.
(1996).
Effects
oflogging
on
forest
birds.
In
Ecology
and
economy
of
a
tropical
dry
forest
in
Madagascar.
Ganzhorn,
J.
U.
&
Sorg,
J.-P.
(Eds).
Primate
Rep.,
Spec.
Iss.
46-1:
203-213.
Humbert,
H.
(1927).
La
destruction
d'une
fore
insulaire
par
le
feu.
Principaux
aspects
de
la
vegetation
a
Madagascar.
Mem.
Acad.
Malgache
5:
1-79.
Jeffrey,
J.
A.
&
Stouffer,
P.
C.
(1999).
Local
extinctions
of
terrestrial
insectivorous
birds
in
a
fragmented
landscape
near
Manaus,
Brazil.
Conserv.
Biol.
13:
1416-1423.
Kofron,
C.
P.
&
Chapman,
A.
(1995).
Deforestation
and
birds
species
composition
in
Liberia,
West
Africa.
Trop.
Zool.
8:
239-256.
Langrand,
0.
(1990).
Guide
to
the
birds
of
Madagascar.
New
Haven:
Yale
University
Press.
Lebreton,
J.-D.,
Chessel,
D.,
Prodon,
R.
&
Yoccoz,
N.
(1988).
L'analyse
des
relations
especes-milieu
par
l'analyse
canonique
des
corres-
pondances.
I.
Variables
de
milieu
quantitatives.
Acta
Oecol.
9:
53-67.
Loiselle,
B.
A.
&
Blake,
J.
G.
(1991).
Temporal
variation
in
birds
and
fruits
along
an
elevational
gradient
in
Costa
Rica.
Ecology
72:
180-
193.
Myers,
N.,
Mittermeier,
R.
A.,
Mittermeier,
C.
G.,
da
Fonseca,
G.
A.
B.
&
Kent,
J.
(2000).
Biodiversity
hotspots
for
conservation
priorities.
Nature
403:
853-858.
Pons,
P.,
Lambert,
B.,
Rigolot,
E.
&
Prodon,
R.
(2003a).
The
effects
of
grassland
management
using
fire
on
habitat
occupancy
and
conservation
of
birds
in
a
mosaic
landscape.
Biodiv.
Conserv.
12:
1843-1860.
Pons,
P.,
Rakotobearison,
G.
&
Wendenburg,
C.
(2003b).
Immediate
effects
of
a
fire
on
birds
and
vegetation
at
Ankarafantsika
Strict
Nature
Reserve,
NW
Madagascar.
Ostrich
74:
146-148.
Poulin,
B.,
Lefebvre,
G.
&
McNeil,
R.
(1992).
Tropical
avian
phenology
in
relation
to
abundance
and
exploitation
of
food
resources.
Ecology
73:
2295-2309.
Poulin,
B.,
Lefebvre,
G.
&
McNeil,
R.
(1993).
Variations
in
bird
abundance
in
tropical
arid
and
semi-arid
habitats.
Ibis
135:
432-
441.
Prodon,
R.
&
Lebreton,
J.-D.
(1981).
Breeding
avifauna
of
a
Medi-
terranean
succession:
the
holm
oak
and
cork
oak
series
in
the
eastern
Pyrenees,
1.
Analysis
and
modelling
of
the
structure
gradient.
Oikos
37:
21-38.
Prodon,
R.
&
Pons,
P.
(1993).
Postfire
bird
studies:
methods,
questions,
and
perspectives.
In
Fire
in
Mediterranean
Ecosystems:
332-343.
Trabaud,
L.
&
Prodon,
R.
(Eds).
Brussels:
Commission
of
the
European
Communities.
Raherilalao,
M.
J.
(2001).
Effets
de
la
fragmentation
de
la
foret
sur
les
oiseaux
autour
du
Parc
National
de
Ranomafana
(Madagascar).
Rev.
Ecol.
(Terre
Vie)
56:
389-406.
Raphael,
M.
G.,
Morrison,
M.
L.
&
Yoder-Williams,
M.
P.
(1987).
Breeding
bird
populations
during
twenty-five
years
of
postfire
succession
in
the
Sierra
Nevada.
Condor
89:
614-626.
Reid,
W.
V.
(1992).
How
many
species
will
there
be?
In
Tropical
deforestation
and
species
extinction:
55-73.Whitmore,
T.
C.
&
Sayer,
J.
A.
(Eds).
New
York:
Chapman
and
Hall.
Schulte,
L.
A.
&
Niemi,
G.
J.
(1998).
Bird
communities
of
early-
successional
burned
and
logged
forest.
J.
Wildl.
Mgmt.
62:
1418-
1429.
Sinclair,
I.
&
Langrand,
0.
(1998).
Birds
of
the
Indian
Ocean
Islands.
Cape
Town:
Struik
Publishers.
Smith,
A.
P.
(1997).
Deforestation,
fragmentation,
and
reserve
design
in
Western
Madagascar.
In
Tropical
forest
remnants,
ecology,
manage-
ment
and
conservation
of
fragmented
communities:
415-441.
Lawrence,
W.
&
Bierregaard,
0.
W.
(Eds).
Chicago:
University
of
Chicago
Press.
Spellerberg,
I.
E
(1992).
Evaluation
and
assessment
for
conservation.
London:
Chapman
&
Hall.
Stattersfield,
A.
J.,
Crosby,
M.
J.,
Long,
A.
J.
&
Wege,
D.
C.
(1998).
Endemic
bird
areas
of
the
world:
priorities
for
biodiversity
conservation.
Cambridge:
BirdLife
International.
Stouffer,
P.
C.
&
Bierregaard,
R.
0.
J.
(1995).
Use
of
Amazonian
forest
fragments
by
understory
insectivorous
birds.
Ecology
76:
2429-
2445.
ter
Braak,
C.
J.
E
(1986).
Canonical
correspondence
analysis:
a
new
eigenvector
technique
for
multivariate
direct
gradient
analysis.
Ecology
67:
1167-1179.
ter
Braak,
C.
J.
E
&
Smilauer,
P.
(1998).
Canoco
reference
manual
and
user's
guide
to
Canoco
for
Windows.
Ithaca,
NY:
Microcomputer
Power.
Urano,
E.,
Yamagishi,
S.,
Andrianarimisa,
A.
&
Andriatsarafara,
S.
(1994).
Different
habitat
use
among
three
sympatric
species
of
couas
Coua
cristata,
C.
coquereli
and
C.
rufzceps
in
western
Madagascar.
Ibis
136:
485-487.
Walther,
B.
A.
&
Martin,
J.
L.
(2001).
Species
richness
estimation
of
bird
communities:
how
to
control
for
sampling
effort?.
Ibis
143:
413-419.
Woinarski,
J.
C.
Z.
(1990).
Effects
of
fire
on
the
bird
communities
of
tropical
woodlands
and
open
forests
in
northern
Australia.
Aust.
J
Ecol.
15:
1-22.
Yamagishi,
S.
&
Eguchi,
K.
(1996).
Comparative
foraging
ecology
of
Madagascar
vangids
(Vangidae).
Ibis
138:
283-290.
Yamagishi,
S.,
Urano,
E.
&
Eguchi,
K.
(1995).
Group
composition
and
contributions
to
breeding
by
rufous
vangas
Schetba
rufa
in
Madagascar.
Ibis
137:
157-161.