Combining plant and bird data increases the accuracy of an

Accepted Manuscript
Title: Combining plant and bird data increases the accuracy of
an Index of Biotic Integrity to assess conservation levels of
tropical forest fragments
Author: Hugo Reis Medeiros Gabriela Menezes Bochio
Milton Cezar Ribeiro José Marcelo Torezan Luiz dos Anjos
PII:
DOI:
Reference:
S1617-1381(15)00009-6
http://dx.doi.org/doi:10.1016/j.jnc.2015.01.008
JNC 25402
To appear in:
Received date:
Revised date:
Accepted date:
19-6-2014
28-1-2015
29-1-2015
Please cite this article as: Medeiros, H. R., Bochio, G. M., Ribeiro, M. C., Torezan, J.
M., and dos Anjos, L.,Combining plant and bird data increases the accuracy of an Index
of Biotic Integrity to assess conservation levels of tropical forest fragments, Journal for
Nature Conservation (2015), http://dx.doi.org/10.1016/j.jnc.2015.01.008
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Combining plant and bird data increases the accuracy of an Index of Biotic
2
Integrity to assess conservation levels of tropical forest fragments
3
Hugo Reis Medeiros1, Gabriela Menezes Bochio1, Milton Cezar Ribeiro2, José Marcelo
4
Torezan3, and Luiz dos Anjos4
ip
t
1
5
1
7
Animal e Vegetal,Universidade Estadual de Londrina, Caixa Postal 6001, 86051-970
8
Londrina, Paraná, Brasil.
2
us
Programa de Pós-Graduação em Ciências Biológicas, Departamento de Biologia
Laboratório de Ecologia Espacial e Conservação - LEEC, Departamento de
an
9
cr
6
Ecologia,Universidade Estadual Paulista - UNESP, Caixa Postal 13506-900 Rio Claro,
11
São Paulo, Brasil.
12
3
13
Departamento de Biologia Animal e Vegetal, Universidade Estadual de Londrina, Caixa
14
Postal 6001, 86051-970 Londrina, Paraná, Brasil.
15
4
16
Vegetal, Universidade Estadual de Londrina, Caixa Postal 10011, 86051-970 Londrina,
17
Paraná, Brasil.
te
d
Laboratório de Biodiversidade e Restauração de Ecossistemas - LABRE,
Laboratório de Ornitologia e Bioacústica, Departamento de Biologia Animal e
Ac
ce
p
18
M
10
19
Abstract
20
Rapid ecological assessment methods, such as Rapid Ecological Assessments (REA)
21
and Indexes of Biotic Integrity (IBI) are useful tools for the selection of priority areas
22
for biodiversity conservation. However, the majority of rapid assessment methods are
23
based on data from a single taxonomic group; a multi-taxa index should provide a more
24
integrated evaluation of the response of a disturbed system. In this study, we propose a
Page 1 of 28
new, easy-to-follow, integrated Index of Biotic Integrity (IBIint) which combines plants
26
and birds to assess ecological integrity of tropical forest fragments. This integrated
27
index combines the information of two previously developed rapid assessment methods:
28
REA for plants and IBI for birds. These two indexes were built based on key vegetation
29
features and on levels of sensitivity to forest fragmentation of bird species. We applied
30
IBI, REA and the new IBIint indexes on the characterization of 10 forest fragments and
31
in a large continuous forest block (reference area). We also tested the correlation of the
32
proposed index (IBIint), REA and IBI with patch size, forest amount and connectivity at
33
four spatial scales (250, 500, 1000, 1500 meters). Our hypothesis was that IBIint would
34
be more correlated with landscape metrics than the REA and IBI. As expected, IBIint
35
was the more accurate index once it was explained by all landscape variables: area of
36
forest fragments; forest connectivity; and, percentage of forest cover at four spatial
37
scales. REA and IBI were explained only by one of those parameters. We conclude that
38
IBIint can be an excellent tool to aid conservationists and managers for defining
39
conservation strategies in scenarios with fast habitat loss.
cr
us
an
M
d
te
Ac
ce
p
40
ip
t
25
Rapid
assessment
methods;
Ecological
41
Keywords:
42
configuration; Avifauna, Vegetation structure, Bioindicators.
43
44
integrity,
Landscape
Introduction
45
Rapid assessment methods used for environmental evaluation, such as Rapid
46
Ecological Assessments (REA; Sobrevila & Bath, 1992) and Indexes of Biotic Integrity
47
(IBI; Karr, 1981), are important tools in aiding conservationists and decision makers in
48
the selection of conservation strategies (Pont et al., 2006). These techniques are used to
49
quickly assess the environmental value of different localities through evaluation of a
Page 2 of 28
finite and highly informative set of ecological indicators collected in the field (Karr,
51
1999; Sayre et al., 2000; Andreasen et al., 2001). Different taxonomic groups, such as
52
plants (Mack, 2004), fishes (Lyons et al., 2000), amphibians (Micacchion, 2002),
53
macroinvertebrates (Roque et al., 2010; Couceiro et al., 2012), birds (Glennon & Porter,
54
2005), as well as landscape structure (Barreto et al., 2010) have been used as indicators
55
of environmental integrity.
cr
ip
t
50
Indexes using plants or birds have been developed for a variety of habitats, mostly
57
found in temperate regions, including riparian (Bryce et al., 2002), wetlands (DeLuca et
58
al., 2004; Rooney & Bayley, 2012), arid (Allen, 2009), grasslands (Coppedge et al.,
59
2006), and forests ecosystems (O’Connell et al., 2007). However, the majority of rapid
60
assessment methods are based on data from a single taxonomic group (Diffendorfera et
61
al., 2007), although an index with more taxa should provide a more integrated
62
evaluation of the response of a disturbed system (Nieme et al., 2004).
d
M
an
us
56
In this study, we proposed a new integrated Index of Biotic Integrity (IBIint) that
64
combines plants and birds to assess ecological integrity of tropical forest fragments.
65
This integrated index includes the information of two previously developed rapid
66
assessment methods (REA for plants and IBI for birds). These two indexes were built
67
based on key vegetation features (REA; Medeiros & Torezan, 2013) and on levels of
68
sensitivity of bird species to forest fragmentation (IBI; Anjos et al., 2009). To test the
69
accuracy of the proposed index IBIint and both REA and IBI, we evaluated the relative
70
contribution of forest amount and landscape configuration to explaining the three
71
biodiversity indexes. Explanatory landscape structure variables included: (1) area of
72
forest fragments; (2) forest connectivity; and, (3) percentage of forest cover at four
73
spatial scales (250, 500, 1000, 1500 meters). It is known that the decrease of those three
74
variables influences species persistence, leading to a decline in the biological integrity
Ac
ce
p
te
63
Page 3 of 28
of a given forest fragment (Fahrig, 2003; Martensen et al., 2008; Martensen et al., 2012;
76
Silva et al., 2014). Our hypothesis is that this decline in the biological integrity could be
77
better captured by IBIint than the indexes based on only one single taxonomic group,
78
such as REA or IBI.
ip
t
75
Methods
81
Study area
us
80
cr
79
This study was carried out in a large forest block (187,000 ha) and in 10 forest
83
fragments, all located in Paraná State, southern Brazil. The predominant vegetation in
84
the region is the Seasonal Semideciduous Forest (SF), a type of Atlantic Forest that
85
occurs mainly in south and southeast Brazil. The SF is considered one of the most
86
threatened ecosystem types of the Atlantic Forest hotspot. Indeed, SF has been
87
systematically converted into monocultures of sugarcane, soybean and eucalyptus, as
88
well as pasture and cities; consequently, only 7% of its original cover still remains
89
(Ribeiro et al., 2009). This ecosystem is characterized by a partial deciduousness with
90
about 20 to 50% of trees losing their leaves during the dry season (Carvalho, 2003). The
91
canopy height reaches 8 to 15 m, though some emergent trees have been known to reach
92
20 m high. In addition to vines and epiphytes, a complex litter layer is also common in
93
SFs (Batalha, 1997). The SF also faces a large carbon loss (~60%) because of forest
94
fragmentation, when compared to control areas (i.e. patches > 10,000 ha; Putz et al.,
95
2014). The large forest block cited above is the Iguaçu National Park (hereafter IP),
96
which is the largest and best preserved remaining patch of SF for the entire Atlantic
97
Forest domain (Ribeiro et al., 2009). Therefore, we considered IP as the reference area
98
in this study. The IP and the set of 10 sampled forest fragments are located in the
99
extreme west and in the north of Paraná State, respectively (Fig. 1). The studied forest
Ac
ce
p
te
d
M
an
82
Page 4 of 28
fragments ranged from 60 to 876 ha and were located in landscape contexts with
101
different degrees of SF functional connectivity and forest cover (Table 1). Connectivity
102
is the capacity of the landscape to influence biological fluxes (Taylor et al., 1993), and
103
functional connectivity depends on the dispersal ability of fauna to access scattered
104
habitats or resources (Boscolo et al., 2008; Martensen et al., 2012). Please refer to the
105
landscape variables section for details about metric calculations.
cr
ip
t
100
Table 1. Location of 10 forest fragments and the reference area selected for
107
this study. FF = identification of forest fragments, Size (ha) = the size of forest
108
fragments in hectares (ha), FC (%) = amount of forest cover within the radius
109
of 1000 m, and Prox = proximity index, a proxy for functional connectivity.
Proxb
FC (%)
187,000
10,000
100.0
MG
650
88.1
44.8
MF
876
0.1
47.2
AT
85
9.0
22.9
EI
60
0.0
PI
74
9.6
PQ
542
25.6
CO
564
BU
288
DO
166
SH
85
113
114
25°09’12”S 53°50’15”W
23°26’46”S 51°14’46”W
23°09’37”S 50°34’00”W
M
23°20’41”S 51°08’23”W
29.2
22°46’49’’S 51°29’21’’W
9.5
23°30’05”S 51°04’39”W
3.9
40.0
23°28’12”S 51°02’50”W
5,903
39.5
23°24’19”S 51°19’31”W
0.0
0.0
23°18’05”S 50°59’11”W
9.0
21.9
23°24’38”S 51°14’09”W
d
23°15’21’’S 51°01’53”W
Ac
ce
p
110
111
112
a
Coordinates
0.0
te
IP
a
an
Size (ha)
FF
us
106
b
Reference area; Proximity index and forest cover used a search radius of
1000 meters
Vegetation structure
We characterized the vegetation using the REA method, which was previously
115
proposed by Medeiros & Torezan (2013). The REA is based on the plant community
116
structure and was developed to assess the vegetation quality of forest fragments. REA
117
comprises 10 ecological integrity variables, each of them rated from 1 (lowest integrity)
118
to 5 (highest integrity). The selected ecological variables were: 1 - number of standing
119
dead trees; 2 - presence and cover of exotic grasses; 3 - presence of other exotics; 4 -
120
abundance of vine tangles; 5 - types of eco-units present (light gaps with many vines,
Page 5 of 28
light gaps with few vines, low canopy up to 10 m tall, open canopy with up to 60 % of
122
lightness and closed canopy with up to 10 % of lightness); 6 – abundance of vascular
123
epiphytes (except Orchidaceae); 7 – abundance of fig trees (Ficus ssp, Moraceae); 8 –
124
abundance of Orchidaceae; 9 – abundance of Heart-palm trees (Euterpe edulis,
125
Arecaceae); and, 10 – abundance of Peroba Rosa trees (Aspidosperma polyneuron,
126
Apocynaceae). The selection of these ecological variables was based on ease of
127
observation in the field and ability to reflect vegetation structure. The ratings of the 10
128
variables were added together to give the REA values for the assessment. As each
129
variable rating ranged from 1 to 5, the possible REA rating is between 10 and 50. REA
130
scores ≤ 20 indicate very low integrity, ≥ 20 and ≤ 30 indicate low integrity, ≥ 30 and ≤
131
40 indicate regular integrity, ≥ 40 and ≤ 45 indicate good integrity and ≥ 45 indicate
132
excellent integrity.
M
an
us
cr
ip
t
121
The REA was applied to transects which were located in the same area of bird
134
censuses, equidistant from fragment borders and from other transects, and avoiding the
135
first 100 m from the edge in any direction. Each transect was 100 m long, 30 m wide
136
and contained three assessment points, 30 m apart. The number of transects in each
137
forest fragment varied according to patch size. Eight transects were surveyed in forest
138
fragments larger than 500 ha, six in those of 100 to 500 ha, and four in patches of less
139
than 100 ha. In the IP, eight transects were sampled in a pristine area, avoiding edges in
140
all directions (at least 1 km from any edge). The REA was performed from December
141
2011 to March 2012 in the forest fragments and in January 2013 for the IP. All
142
assessments were performed in the summer to avoid seasonal variations.
143
Bird survey
Ac
ce
p
te
d
133
144
To characterize the birds, we used the IBI that was previously proposed by Anjos
145
et al. (2009). The IBI is based on the recorded sensitivity of bird species to forest
Page 6 of 28
fragmentation in each studied site. We accomplished this using the levels of sensitivity
147
to forest fragmentation presented in Anjos (2006) and Anjos et al. (2011) and followed
148
the procedure presented in Anjos et al. (2009). Three levels of sensitivity were
149
considered: low; intermediate; and, high. Species with low or intermediate sensitivity
150
are those presented in Anjos (2006), and the high sensitivity follow the suggestion of
151
Anjos et al. (2011); the criteria to allocate a given bird species to its sensitivity level
152
was based on their occurrence and relative abundance in forest fragments. In all cases,
153
the species belongs to the same study region. We selected 10 species in each level of
154
sensitivity (total of 30 species; see Appendix A), choosing those most likely to be
155
recorded (i.e., with high detectability) during four consecutive days of survey, following
156
the recommendations of Bochio & Anjos (2012). We replaced Pteroglossus castanotis
157
by Pteroglossus aracari because the border of the former’s range lies adjacent to the IP
158
and, therefore, the population size of P. castanotis is expected to be very low in that
159
region.
te
d
M
an
us
cr
ip
t
146
To obtain the IBI values, we used the numbers of species recorded in each
161
sensitivity group multiplied by their respective weights. Three different weights were
162
used according to the level of sensitivity of bird species. The highest weight (3) was
163
attributed to bird species grouped with high sensitivity, while the species grouped with
164
medium sensitivity and low sensitivity received the weights (2) and (1) respectively.
165
Thereafter, the sums of the three multiplied values were divided by 60 so that the IBI
166
scores range from 0 to 1 (see Anjos et al., 2009).
Ac
ce
p
160
167
Then, to obtain a bird species list for each studied site, we performed bird surveys
168
on four consecutive days along a 1000 m transect that covered a standard area of
169
approximately 60 hectares. Surveys began at sunrise and finished four hours later,
170
totaling 16h of field observations per study site. Bird sampling was performed during
Page 7 of 28
the breeding season in all the selected fragments from October 2010 to January 2011
172
and from October 2011 to December 2011. In the IP, sampling was performed in
173
January 2013. Transects were walked with a constant speed and we only recorded the
174
bird species heard or seen (not their individual numbers) at any distance from the
175
transect.
176
The integrated index (IBIint)
cr
ip
t
171
To calculate the IBIint, different weights were used according to the sensitivity
178
levels of bird species and vegetation structure. We gave the number of high sensitivity
179
species and the REA value a weight of 3, by which we multiplied both values in the
180
formula. The number of bird species with medium sensitivity was multiplied by 2
181
(weight 2) while the number of bird species with low sensitivity was multiplied by 1
182
(weight 1). We varied the weight of REA in the index to ascertain the most appropriated
183
weight (1, 2 or 3), and only the IBIint based on REA with weight 3 was fully explained
184
by landscape metrics. Therefore, we opted to use weight 3 for the final version of the
185
index. Finally, the sum of the four multiplied values (three related to sensitivity level of
186
bird species and one related to REA values) was divided by 210, the sum of the
187
maximum value obtained by each of the four multiplied values as expressed by the
188
equation:
190
an
M
d
te
Ac
ce
p
189
us
177
Where HS is the number of birds species with high sensitivity recorded within a
191
given forest fragment, while MS and LS represent the number of bird species with
192
medium and low sensitivity, respectively. In each of the three bird sensitivity categories
193
the number of species recorded can vary from 0 to 10. The numbers 1, 2 and 3 represent
194
the multiplied weights of the bird sensitivity categories as well as of REA.
Page 8 of 28
The IBIint scores ranged from 0 to 1, and were classified as follow: values ≤ 0.25
196
indicate very low ecological integrity; values ≥ 0.26 and ≤ 0.50 indicate low integrity;
197
values ≥ 0.51 and ≤ 0.75 indicate intermediate ecological integrity; and values > 0.75
198
indicate high ecological integrity.
199
Landscape variables
ip
t
195
Landscape metrics were calculated using a binary forest vs. matrix map as input,
201
which was obtained from SOS Mata Atlântica (2008) and used by Ribeiro et al., (2009).
202
Forest cover and functional connectivity (Metzger et al., 2009; Martensen et al., 2012)
203
were estimated for search radii of 250, 500, 1000 and 1500 meters around a centroid
204
defined within the central area of the patches. Previous studies have indicated that these
205
radii are suitable to encompass the home range sizes of most understory birds (Develey
206
& Metzger, 2006; Boscolo & Metzger, 2009; 2011). Additionally, we used these four
207
scales to simulate species with varying dispersal ability. Functional connectivity, a
208
measure of landscape configuration, was estimated by a proximity index (Uezu et al.,
209
2008). This index is based on the size and proximity of all forest fragments situated
210
within a predetermined search radius of the focal patch (Gustafson & Parker, 1994). For
211
the IP reference site, we assumed 100% of forest cover and a proximity index higher
212
than the highest value among forest fragments at all spatial scales. Area and proximity
213
indexes were log-transformed in order to allow linear regression analysis. Landscape
214
metrics were calculated using ArcGIS (ESRI, 2005) and V-LATE extension (LARG,
215
2005).
216
Data analysis
Ac
ce
p
te
d
M
an
us
cr
200
217
A model selection approach was used to estimate the relative contribution of the
218
various competing, ecologically meaningful models (Burnham & Anderson, 2002). To
219
fit the models, we used Beta Regressions that are useful for models which the dependent
Page 9 of 28
variable is continuous and restricted to the standard unit interval (0, 1) (Ferrari &
221
Cribari-Neto, 2004). Beta regression naturally incorporates features such as
222
heteroskedasticity or skewness which are commonly observed in data such as rates or
223
proportions (Cribari-Neto & Zeileis, 2010). Furthermore, similarly to generalized linear
224
models (GLM), the inference of Beta regression is based on likelihood (Ibáñez et al.,
225
2011).
cr
ip
t
220
For each dependent variable (IBIint, REA and IBI), we analyzed three competing
227
models and a null model (Table 2). We performed a preliminary analysis on forest cover
228
to determine the best spatial scale (250, 500, 1000 or 1500 m) to explain the response
229
variable, which we then compared with other explanatory variables. The same was done
230
for connectivity. In our case, the null model represents the absence of effect (Martensen
231
et al., 2012).
M
an
us
226
Table 2. Competing models used to identify the
233
most accurate index to infer the ecological
234
integrity of Atlantic Forest hotspot fragments,
235
Brazil. Connectivity was based on Proximity
236
Index. Forest cover and connectivity were
237
calculated in four spatial scales (250, 500, 1000,
te
Ac
ce
p
238
d
232
1500
meters)
Models
Number of parameters
IBIint = Area
3
IBIint = Forest cover
3
IBIint = Connectivity
3
IBIint = null
3
REA = Area
3
REA = Forest cover
3
REA = Connectivity
3
REA = null
3
IBI = Area
3
IBI = Forest cover
3
IBI = Connectivity
3
IBI = null
3
239
Page 10 of 28
The best models were identified based on the Akaike Information Criterion (AIC;
241
Burnham & Anderson, 1998) with the small sample correction (AICc; Hurvich & Tsai,
242
1989). We also used the AICc weight (wAICc), which represents the weight of evidence
243
in favor of a given model being the best one (Burnham & Anderson, 2002). To further
244
guide model selection within the best concurrent models, we used the AICc, which is
245
the difference between the AICc of a considered model and the model with the lowest
246
AICc value (Martensen et al., 2008). Thus, models with wAICc ≥ 0.10 and AICc ≤ 2.5
247
were considered as equally plausible to explain the dependent variables. All analyses
248
were performed using R version 3.0.2 (R Development Core Team, 213). The “betareg”
249
package (Cribari-Neto & Zeileis, 2010) was used to fit beta regression models by using
250
the betareg function.
M
an
us
cr
ip
t
240
Since REA and IBI have already been biologically evaluated, we considered the
252
index with greater accuracy, which was explained by the larger number of explanatory
253
variables.
te
254
d
251
Results
256
Ecological integrity of forest fragments
257
Ac
ce
p
255
As we expected, the IP (reference area) obtained the highest IBIint values,
258
followed by the forest fragment MG. Among the 10 forest fragments, IBIint values
259
ranged from 0.39 to 0.95 (Table 3).
260
261
262
263
264
Page 11 of 28
Table 3. Estimated ecological integrity based on IBIint, REA and IBI values for the
266
10 forest fragments and the continuous area (reference area) within fragmented
267
landscapes of Atlantic Forest hotspot. The scores of IBIint and IBI ranged from 0 to
268
1, while the REA score ranged from 10 to 50. IBIint and IBI were classified as low,
269
intermediate and high ecological integrity, REA were classified as low, regular,
270
good and excellent integrity.
Integrity
REA
Integrity
IBI
Integrity
0.95
High
48
Excellent
0.93
High
MG
0.84
High
44
Good
0.75
High
MF
0.62
Intermediate
33
Regular
0.51
Intermediate
AT
0.50
Low
30
Low
0.23
EI
0.39
Low
23
Low
0.20
PI
0.47
Low
29
Low
PQ
0.59
Intermediate
28
Low
CO
0.58
Intermediate
27
Low
BU
0.78
High
44
DO
0.51
Intermediate
30
SH
0.70
Intermediate
35
272
Low
0.20
Low
0.67
Intermediate
0.57
Intermediate
Good
0.52
Intermediate
Low
0.28
Low
Regular
0.70
High
M
271
Low
us
IP
cr
IBIint
a
an
FF
ip
t
265
Index models
According to the best models, the IBIint was the more accurate index once it was
274
explained by a larger number of explanatory variables than REA and IBI at all four
275
spatial scales. IBIint was explained by all explanatory variables at 500 and 1000 meters
276
and by forest cover and connectivity at 250 and 1500 meters, while REA and IBI were
277
explained only by connectivity and patch size, respectively, at all spatial scales. We then
278
used the radius of 1000 meters to represent the general results that were similar among
279
the four scales (Table 4; Fig. 2).
te
Ac
ce
p
280
d
273
281
282
283
284
285
Page 12 of 28
Table 4. Best models to explain the dependent variables (Dep. Var.): REA, IBI and
287
their derived index IBIint. Pseudo R² indicates the goodness-of-fit of the estimated
288
models. AICc = Akaike Information Criterion with the small sample correction, AICc
289
= Delta value of AICc and wAICc = weight of evidence of the models. Connectivity
290
was based on proximity index. Lines highlighted in gray indicate the plausible models.
296
AICc
IBIint
IBIint
IBIint
IBIint
REA
REA
REA
REA
IBI
IBI
IBI
IBI
1
2
3
4
1
2
3
4
1
2
3
4
IBIint = Forest cover**
IBIint = Connectivity**
IBIint = Area**
IBIint = Null
REA = Connectivity **
REA = Forest cover
REA = Area
REA = Null
IBI = Area**
IBI = Forest cover
IBI = Connectivity
IBI = Null
0.69
0.68
0.62
0.06
0.75
0.64
0.46
0.17
0.64
0.46
0.37
0.00
0
0.2
2.3
10.3
0
4.8
8.9
13.9
0
3.7
5.9
11.6
us
an
M
wAICc
0.45
0.41
0.14
<0.001
0.90
0.08
0.01
<0.001
0.82
0.12
0.04
0.002
cr
Pseudo R²
** indicates the equally plausible models based on AICc and wAICc values. For
each index, models are ranked from best to worst according, to AICc and wAICc.
d
295
Model
te
294
Rank
Ac
ce
p
291
292
293
Dep. var.
ip
t
286
Page 13 of 28
ip
t
cr
us
an
M
d
te
Ac
ce
p
297
298
Fig. 2 Best supported models of the relation between the indexes and the three landscape explanatory
299
variables: area of forest fragments (ha), connectivity (proximity index) and forest cover (%). (A), (B) and
300
(C) the best models for IBIint (plants and birds), (D) the best model for REA (only plants) and (E) the
301
best model for IBI (only birds). wAICc = weight of evidence of the model.
302
303
304
305
Page 14 of 28
306
Discussion
IBIint was the more accurate index once it was explained by all landscape
308
variables, while REA and IBI were explained only by connectivity and fragment size,
309
respectively. In fact, habitat amount and landscape connectivity regulate important
310
multiscale ecological processes and patterns that were best reflected by IBIint. For
311
example, the habitat amount and connectivity are tightly related to colonization
312
dynamics, population size, local extinction and rescue effects (Brown & Kodrick-
313
Brown, 1977; Ricketts, 2001; Fahrig, 2003). Large fragments support large populations
314
which are less prone to local extinction (Temple & Cary, 1988; Jules, 1998).
an
us
cr
ip
t
307
REA results indicate that vegetation integrity increases with increasing
316
connectivity regardless of patch size and forest cover context. Connectivity may
317
increase vegetation resilience by enhancing seed dispersal and recolonization
318
opportunities for species with high and low dispersal capacities. Indeed, greater
319
proximity between habitat patches decreases the distances from propagule sources to
320
colonization targets (Lavorel, 1999; Cayuela et al., 2006; Minor et al., 2009).
321
Furthermore, these findings can be associated with anthropic impacts that do not depend
322
on landscape features and patch size. For instance, MF forest fragment (the largest
323
fragment with 876 ha), suffered large fires and intense logging in 1970s and 1980s and
324
achieved a lower value of REA than smaller fragments such as BU (288 ha) and SH (85
325
ha) which suffered fewer human disturbances. Therefore, the vegetation structure seems
326
to be more strongly moderated by the synergy between historical human impacts and
327
the current spatial arrangement of habitats than by habitat amount and patch size.
Ac
ce
p
te
d
M
315
328
IBI had correlation with fragment size, probably due to amount and quality of
329
food availability for birds. Uezu et al. (2005) noted that the abundance of frugivorous
330
bird species are affected more by fragment size than by connectivity, and the opposite is
Page 15 of 28
true for the insectivores. Martensen et al. (2008) demonstrated that both fragment size
332
and connectivity influenced differently functional groups of birds. In our 30 bird species
333
selection for IBI, we included both frugivorous and insectivorous birds. But it is
334
possible that another combination of species could indicate correlation with both
335
fragment size and connectivity. In our study, however, a balance between metrics
336
related to connectivity (REA) and fragment size (IBI) should improve balance for the
337
IBIint.
cr
ip
t
331
Here, we found that a multi-taxa approach improves the accuracy of rapid
339
assessment methods at identifying priority areas for biodiversity conservation in
340
fragmented landscapes such as our study region. Rapid assessment methods cannot, and
341
do not intend to replace long-term studies and scientific inventories, but they can be
342
useful tools to extend the geographical application of those studies and inventories with
343
a considerable reduction in time and cost (Abate, 1992; Sutula et al., 2006; Medeiros &
344
Torezan, 2013; Álvarez-Berastegui et al., 2014). These methods are especially relevant
345
in tropical regions such as the Brazilian Atlantic Forest, which has undergone a huge
346
habitat loss and is considered one of the most threatened biodiversity hotspots in the
347
world (Myers et al., 2000). Recent estimative reports that only 11.4% to 16% of the
348
Atlantic forest’s original cover remains, and are mainly distributed in isolated and small
349
fragments < 50 ha (Ribeiro et al., 2009). In this fast and wide habitat loss scenario, rapid
350
assessment methods based on multi-taxa indicators can be excellent tools to aid
351
conservationists and managers in defining conservation strategies at local and landscape
352
scales.
Ac
ce
p
te
d
M
an
us
338
353
354
355
Page 16 of 28
Acknowledgements
357
We are grateful to Odair C. Pavão for field assistance, Calebe P. Mendes and Juliana
358
Silveira dos Santos for statistical and GIS support. We thank the the government
359
agencies of environment protection (ICMbio and IAP) and the owners of the private
360
lands where some sites are located for permission to conduct our studies. We also thank
361
the editors and the two anonymous referees who helped us to improve substantially the
362
manuscript. CNPq (Brazilian Government Research Council) provided research grants
363
for Luiz dos Anjos (302694/2010-2), Milton Cezar Ribeiro (312045/2013-1), José
364
Marcelo D. Torezan (305854/2012-7) and a technical grant for O. C. Pavão
365
(503836/2010-9). Gabriela Menezes Bochio received a research grant from CAPES
366
(Coordination for the Improvement of Higher Education Personnel). MCR also thanks
367
the financial support by FAPESP (process 2013/50421-2).
M
an
us
cr
ip
t
356
d
368
References
370
Abate, T. (1992). Environmental Rapid-Assessment Programs Have Appeal and Critics.
Ac
ce
p
371
te
369
BioScience,42, 486-489.
372
Allen, C. D. (2009). Monitoring environmental impact in the upper sonoran lifestyle: a
373
new tool for rapid ecological assessment. Environmental Management, 43, 346–
374
356.
375
Álvarez-Berastegui, D., Amengual, J., Coll, J., Reñones, O., Moreno-Navas, J., &
376
Agardy, T. (2014). Multidisciplinary rapid assessment of coastal areas as a tool for
377
the design and management of marine protected areas. Journal for Nature
378
Conservation, 22, 1-14.
Page 17 of 28
379
Andreasen, J. K., O’Neill, V. R., Noss, R., & Slosser, N. C. (2001). Considerations for
380
the development of a terrestrial index of ecological integrity. Ecological Indicators,
381
1, 21–35.
ip
t
383
Anjos, L. (2006). Bird species sensitivity in a fragmented landscape of Atlantic forest in
southern Brazil. Biotropica, 38, 229–234.
cr
382
Anjos, L., Bochio, G. M., Campos, J. V., McCrate, G. B, & Palomino, F. (2009). Sobre
385
o uso de níveis de sensibilidade de aves à fragmentação florestal na avaliação da
386
Integridade Biótica: um estudo de caso no norte do Estado do Paraná, sul do Brasil.
387
Revista Brasileira de Ornitologia, 17, 28–36.
an
us
384
Anjos, L., Collins, D. C., Holt, R. D., Volpato, G. H., Mendonça, L. B., Lopes, E. V.,
389
Boçon, R., Bisheimer, M. V., Serafini, P. P., & Carvalho, J. (2011). Bird species
390
abundance–occupancy patterns and sensitivity to forest fragmentation: implications
391
for conservation in the Brazilian Atlantic forest. Biological Conservation, 144,
392
2213–2222.
Ac
ce
p
te
d
M
388
393
Barreto, L., Ribeiro, M.C., Veldkamp, A., Van Eupen, M., Kok, K., & Pontes, E.
394
(2010). Exploring effective conservation networks based on multi-scale planning
395
396
397
398
399
400
unit analysis. A case study of the Balsas sub-basin, Maranhão State, Brazil.
Ecological Indicators, 10, 1055-1063.
Batalha, M. A. (1997). Análise da vegetação da AIRE Cerrado Pé-de-Gigante (Santa
Rita do Passa Quatro, SP). USP-IB, São Paulo. MSc Dissertation.
Bochio, G. M., & Anjos, L. (2012). The importance of considering bird detectability for
assessing biological integrity. Natureza & Conservação, 10, 72–76.
Page 18 of 28
401
Boscolo, D., Candia-Gallardo, C., Awade, M., & Metzger, J. P. (2008). Importance of
402
inter-habitat gaps and stepping-stones for a bird species in the Atlantic Forest,
403
Brazil. Biotropica, 40, 273–276.
Boscolo, D., & Metzger, J. P. (2009). Is bird incidence in Atlantic forest fragments
405
influenced by landscape patterns at multiple scales? Landscape Ecology, 24, 907-
406
918.
409
410
cr
us
highly fragmented landscapes. Ecography, 34, 1018–1029.
an
408
Boscolo, D., & Metzer, J. P. (2011). Isolation determines patterns of species presence in
Brown, J. H., & Kodrick-Brown, A. (1977). Turnover rates in insular biogeography:
effect of immigration on extinction. Ecology, 58, 445–449.
M
407
ip
t
404
Bryce, S. A., Hughes, R. M., & Kaufmann, P. R. (2002). Development of a bird
412
integrity index: using bird assemblages as indicators of riparian condition.
413
Environmental Management, 30, 294–310.
415
te
Burnham, K. P., & Anderson, D. R. (1998). Model Selection and Inference: a Practical
Ac
ce
p
414
d
411
Information-Theoretic Approach. Springer-Verlag, New York.
416
Burnham, K. P., & Anderson, D. R. (2002). Model Selection and Multimodel Inference:
417
A Practical Information-Theoretical Approach. 2d ed. New York: Springer-Verlag.
418
419
Carvalho, P. E. R. (2003). Espécies arbóreas brasileiras. Embrapa Informações
Tecnológicas, Brasília, Distrito Federal, Brasil.
420
Cayuela, L., Benayas, J. M. R., & Echeverría, C. (2006). Clearance and fragmentation
421
of tropical montane forests in the Highlands of Chiapas, Mexico (1975–2000).
422
Forest Ecology and Management, 226(1), 208-218.
Page 19 of 28
423
Coppedge, B. R., Engle, D. M., Masters, R. E., & Gregory, M. S. (2006). Development
424
of a grassland integrity index based on breeding bird assemblages. Environmental
425
Monitoring and Assessment, 118, 125–145.
Couceiro, S. R. M., Hamada, N., Forsberg, B. R. T., Pimentel, P., & Luz, S. L. B.
427
(2012). A macroinvertebrate multimetric index to evaluate the biological condition
428
of streams in the Central Amazon region of Brazil. Ecological Indicators, 18, 118–
429
125.
cr
us
431
Cribari-Neto, F., & Zeileis, A. (2010). Beta Regression in R. Journal of Statistical
Software, 32(2), 1-24.
an
430
ip
t
426
DeLuca, W. V., Studds, C. E., Rockwood, L. L., & Marra, P. P. (2004). Influence of
433
land use on the integrity of marsh bird communities of Chesapeake Bay, USA
434
Wetlands, 24, 837–847.
d
M
432
Develey, P. F., & Metzger, J. P. (2006) Emerging threats to birds in Brazilian Atlantic
436
forests: the roles of forest loss and configuration in a severely fragmented
438
Ac
ce
p
437
te
435
ecosystem. In W. F., Laurance, & C. A. Peres (Eds) Emerging threats to tropical
forests. (pp.269–290). University of Chicago Press, Chicago.
439
Diffendorfera, J. E., Fleming, J. M., Duggana, J. M., Chapmana, R. E., Rahna, M. E.,
440
Mitrovicha, M. J., & Fisherb, R. N. (2007). Developing terrestrial, multi-taxon
441
442
443
444
indices of biological integrity: an example from coastal sage scrub. Biological
Conservation, 140, 130 –141.
ERSI (Environmental Systems Research Institute). 2005. ArcGIS Desktop: Release 8.
Environmental Systems Research Institute, Redlands, CA.
Page 20 of 28
446
447
448
Ferrari, S. L. P., & Cribari-Neto, F. (2004). Beta Regression for modeling rates and
proportions. Journal of Applied Statistics, 31(7), 799–815.
Fahrig, L. (2003). Effects of habitat fragmentation on biodiversity. Annual Reviews of
Ecology and Systematics, 34, 487-515.
ip
t
445
Glennon, M. J., & Porter, W. F. (2005). Effects of land use management on biotic
450
integrity: an investigation of bird communities. Biological Conservation, 126, 499–
451
511.
455
us
an
454
landscape design. Landscape Urban Plann, 29, 117-30.
Hurvich, C. M., & Tsai, C. L. (1989). Regression and time series model selection in
M
453
Gustafson, E. J., & Parker, G. R. (1994). Using an index of habitat patch proximity for
small samples. Biometrika, 76, 297-307.
d
452
cr
449
Ibáñez, M. V., Prades, M., & Simó, A. (2011). Modelling municipal waste separation
457
rates using generalized linear models and beta regression. Resources, Conservation
458
and Recycling, 55(12), 1129-1138.
460
461
462
463
464
Ac
ce
p
459
te
456
Jules, E. S. (1998). Habitat fragmentation and demographic change for a common plant:
trillium in old-growth forest. Ecology, 79, 1654-1656.
Karr, J. R. (1981). Assessment of biotic integrity using fish communities. Fisheries, 6,
21–27.
Karr, J. R. (1999). Defining and measuring river health. Freshwater Biology, 41, 221–
234.
Page 21 of 28
LARG (Landscape and Resource Management Research Group) (2005) V-LATE -
466
Vector-based
467
http://www.geo.sbg.ac.at/larg/vlate.htm
469
Analysis
Tools
Extension.
31.01.2007,
Lavorel, S. (1999). Ecological diversity and resilience of Mediterranean vegetation to
disturbance. Diversity and distributions, 5(12), 3-13.
cr
468
Landscape
ip
t
465
Lyons, J., Piette, R. R., & Niermeyer, K. W. (2000). Development, validation, and
471
application of a fish-based index of biotic integrity for Wisconsin's large warm
472
water rivers. Transactions of the American Fisheries Society, 130, 1077–1094.
473
Mack, J. J. (2004). Integrated wetland assessment program, part 4: vegetation index of
474
bioticintegrity (VIBI) and tiered aquatic life uses (TALUs) for Ohio wetlands, Ohio
475
EPA Technical report WET/ 2004-4, Ohio Protection Agency, Wetland ecology
476
group, Division of surface water, Columbus, Ohio.
d
M
an
us
470
Martensen, A. C., Pimentel, R. G., & Metzger, J. P. (2008). Relative effects of fragment
478
size and connectivity on bird community in the Atlantic Rain Forest: implications
Ac
ce
p
479
te
477
for conservation. Biological Conservation, 141, 2184–2192.
480
Martensen, A. C., Ribeiro, M. C., Banks-Leite, C., Prado, P. I., & Metzger, J. P. (2012).
481
Associations of forest cover, fragment area, and connectivity with Neotropical
482
483
understory bird species richness and abundance. Conservation Biology, 26, 1100–
1111.
484
Medeiros, H. R., & Torezan, J. M. (2013). Evaluating the ecological integrity of
485
Atlantic forest remnants by using rapid ecological assessment. Environmental
486
Monitoring and Assessment, 185, 4373-4382.
Page 22 of 28
Metzger, J. P., Martensen, A. C., Dixo, M., Bernacci, L. C., Ribeiro, M. C., Teixeira, A.
488
M. G, & Pardini, R., (2009). Time-lag in biological responses to landscape changes
489
in a highly dynamic Atlantic forest region. Biological Conservation, 142, 1166–
490
1177.
ip
t
487
Micacchion, M. (2002). Amphibian index of biotic intergrity (AmphIBI) for wetlands.
492
Final report to USEPA Grant No. CD985875-01. Ohio Environmental Protection
493
Agency. Wetland Ecology Group, Division of Surface Water, Columbus, Ohio
494
Minor, E. S., Tessel, S. M., Engelhardt, K. A., & Lookingbill, T. R. (2009). The role of
495
landscape connectivity in assembling exotic plant communities: a network analysis.
496
Ecology, 90(7), 1802-1809.
M
an
us
cr
491
Myers, N., Mittermeier, R. A., Mittermeier, C. G., Da Fonseca, G. A., & Kent, J.
498
(2000). Biodiversity hotspots for conservation priorities. Nature, 403, 853-858.
499
Niemi, G., Wardrop, D., Brooks, R., Anderson, S., Brady, V., Preal, H., Rakocinski, C.,
500
Brouwer, M., Levinson, B., & McDonald, M. (2004). Rationale for a new
502
te
Ac
ce
p
501
d
497
generation of indicators for coastal waters. Environmental Health Perspectives, 112,
979–986.
503
O’Connell, T. J., Bishop, A. J., Brooks, R. P. (2007). Sub-sampling data from the North
504
American breeding bird survey for application to the bird community index, an
505
indicator of ecological condition. Ecological Indicators, 7, 679–671.
506
Pont, D., Hugueny, B., Beier, U., Goffaux, D., Melcher, A., Noble, R., Rogers, C.,
507
Roset, N., & Scmutz, S. (2006). Assessing river biotic condition at a continental
508
scale: a European approach using functional metrics and fish assemblages. Journal
509
of Applied Ecology, 43, 70–80.
Page 23 of 28
510
Pütz, S., Groeneveld, J., Henle, K., Knogge, C., Martensen, A. C., Metz, M., Metzger, J.
511
P., Ribeiro, M. C., De Paula, M. D., Huth, A. (2014). Long-term carbon loss in
512
fragmented Neotropical forests. Nature Communications, 5, 5037.
ip
t
514
R Development Core Team. (2013). R: A language and environment for statistical
computing. RFoundation for Statistical Computing. Vienna, Austria.
cr
513
Ribeiro, M. C., Metzger, J. P., Martensen, A. C., & Hirota, M. M. (2009). The Brazilian
516
Atlantic forest: how much is left, and how is the remaining forest distributed?
517
Implications for conservation. Biological Conservation, 142, 1141–1153.
an
519
Ricketts, T. H. (2001). The matrix matters: effective isolation in fragmented landscapes.
The American Naturalist, 158, 87-99.
M
518
us
515
Rooney, R. C., & Bayley, S. E. (2012). Development and testing of an index of biotic
521
integrity based on submersed and floating vegetation and its application to assess
522
reclamation wetlands in Alberta’s oil sands area, Canada. Environmental
523
Monitoring and Assessment, 184, 749–761.
Ac
ce
p
te
d
520
524
Roque, F. O., Siqueira, T., Bini, L. M., Ribeiro, M. C., Tambosi, L. R., Ciocheti, G., &
525
Trivinho-strixino, S. (2010). Untangling associations between chironomid taxa in
526
527
Neotropical streams using local and landscape filters. Freshwater Biology, 55, 847865.
528
Sayre, R., Roca, E., Sedaghatkish, G., Young, B., Keel, S., Roca, R., & Sheppard, S.
529
(2000). Nature in focus: rapid ecological assessment. The Nature Conservancy,
530
Arlington, Virginia, USA.
Page 24 of 28
531
Silva, L. G., Ribeiro, M. C., Hasui, E., Costa, C.A., & Cunha, R. 2014. Patch size,
532
functional isolation, visibility and matrix permeability influences Neotropical
533
primate occurrence within highly fragmented landscapes. Plos One.
Sobrevila, C., & Bath, P. (1992). Evaluación ecológica rápida: un manual para
535
usuarios de América Latina y el Caribe. The Nature Conservancy, Arlington,
536
Virginia, USA.
cr
ip
t
534
SOS Mata Atlântica, Instituto Nacional de Pesquisas Espaciais. (2008). Atlas dos
538
remanescentes florestais da Mata Atlântica, período de 2000 a 2005.
539
<http://www.sosmatatlantica.org.br>.
an
us
537
Sutula, M. A., Stein, E. D., Collins, J. N., Fetscher, A. E., & Clark, R. (2006). A
541
practical guide for the development of a wetland assessment method: the California
542
experience. Journal of the American Water Resources Association, 42, 157–175.
543
Taylor, P. D., Fahrig, L., Henein, K., & Merriam, G. (1993). Connectivity is a vital
d
te
element of landscape structure. Oikos, 68, 571–573.
Ac
ce
p
544
M
540
545
Uezu, A., Metzger, J. P., & Vielliard, J. M. E. (2005). Effects of structural and
546
functional connectivity and patch size on the abundance of seven Atlantic Forest
547
bird species. Biological Conservation 123, 507–519.
548
Uezu, A., Beyer, D. D., & Metzger, J. P. (2008). Can agroforest woodlots work as
549
stepping stones for birds in the Atlantic Forest region? Biodiversity and
550
Conservation, 17, 1907–1922.
551
552
Page 25 of 28
553
554
555
ip
t
556
Appendix A
558
Bird species which were used for the calculation for the IBI in each forest fragment. The
559
species are arranged according to their levels of sensitivity to forest fragmentation.
us
cr
557
an
High Sensitivity
Tinamus solitarius; Micrastur semitorquatus; Triclaria malachitacea; Pionopsitta
M
pileata; Trogon rufus; Pteroglossus castanotis; Sirystes sibilator; Trichothraupis
te
Medium Sensitivity
d
melanops; Automolus leucophthalmus; Grallaria varia
Aratinga leucophthalma; Trogon surrucura; Baryphthengus ruficapillus; Pyrrhura
Ac
ce
p
frontalis; Selenidera maculirostris; Melanerpes flavifrons; Cacicus haemorrhous;
Sittasomus griseicapillus; Scytalopus indigoticus; Chiroxiphia caudata
Low Sensitivity
Crypturellus obsoletus; Pionus maximiliani; Dysithamnus mentalis; Pyriglena
leucoptera; Leptopogon amaurocephalus; Piaya cayana; Cyclarhis gujanensis;
Dendrocolaptes platyrostris; Basileuterus culicivorus; Basileuterus leucoblepharus.
560
561
Page 26 of 28
Figure
(D)
(B)
an
us
cr
(A)
(E)
ip
t
(C)
(F)
Ac
ce
pt
ed
M
Fig 1. Location of studied sites. (A) Location of Paraná State in southern Brazil. (B) Ecosystem domains of Paraná
State (gray scale) SF domain highlighted in white; the square 1 indicates the location of IP - reference area and the
square 2 indicates the location of the set of forest fragments. (C), (D), (E) and (F) are landscape sectors where the 10
studied forest fragments (highlighted in black) are situated.
Page 27 of 28
ip
t
cr
us
an
M
ed
ce
pt
Ac
Fig. 2 Best supported models of the relation between the indexes and the three landscape explanatory
variables: area of forest fragments (ha), connectivity (proximity index) and forest cover (%). (A), (B) and
(C) the best models for IBIint (plants and birds), (D) the best model for REA (only plants) and (E) the
best model for IBI (only birds). wAICc = weight of evidence of the model.
Page 28 of 28