Sie sind auf Seite 1von 11

Hydrobiologia (2015) 750:245–255

DOI 10.1007/s10750-014-2040-0

TRENDS IN AQUATIC ECOLOGY

The importance of forest cover for fish richness


and abundance on the Amazon floodplain
Javior Lobón-Cerviá • Laura L. Hess •
John M. Melack • Carlos A. R. M. Araujo-Lima

Received: 6 June 2014 / Revised: 30 August 2014 / Accepted: 6 September 2014 / Published online: 25 September 2014
Ó Springer International Publishing Switzerland 2014

Abstract Flooded forest is one of the most important open water habitat were also related but the results
fish habitats in Neotropical rivers, and one that is were less consistent and apparently reliant on sam-
increasingly subjected to negative impacts from pling methodology. Our results suggest that conser-
logging, agriculture, and other human activities. The vation of the flooded forest is critical for the
purpose of our study was to test quantitatively whether maintenance of fish assemblages in the Amazon, and
fish richness and abundance in Amazonian floodplain that removal of flooded forest will reduce fish richness,
lakes are associated with the area of flooded forest. We fish abundance, and fisheries yield.
sampled fish and several other variables in 35 Ama-
zonian floodplain lakes during the high-water season. Keywords Amazon  Flooded forests  Fishes 
Our results highlighted that fish richness and abun- Fisheries  Synthetic aperture radar
dance were directly related to flooded forest, inversely
related to distance from the river, and influenced by
dissolved oxygen concentration\1 mg l-1. The same
result applied to fish richness and abundance landed by Introduction
fisheries. Other variables such as depth and area of
Fish have an important socio-economic role for human
communities living along tropical rivers and are a
Carlos A. R. M. Araujo-Lima—deceased. major protein source for these people (Daget, 1973;
Fabré & Alonso, 1998; Cerdeira & Ruffino, 2000;
Guest editor: Koen Martens / Emerging Trends in Aquatic Santos & Santos, 2005; Batista & Issac, 2012; Santos
Ecology
et al., 2014). Because the annual flood reduces the area
J. Lobón-Cerviá available to agriculture, local farmers turn to fisheries
Museo Nacional de Ciencias Naturales (CSIC), C/ Jose to obtain income during the high-water season. Over
Gutierrez Abascal, 2, 28006 Madrid, Spain the past 30 years, pressure on floodplain fish stocks in
the Amazon basin has increased owing to changes in
L. L. Hess  J. M. Melack (&)
Earth Research Institute, University of California, fishing technology, increased demand for fish pro-
Santa Barbara 93106, USA ducts, a shift in the local economy from farming to
e-mail: melack@bren.ucsb.edu fishing, and the expansion of cattle ranching on the
floodplain, with accompanying deforestation (McG-
C. A. R. M. Araujo-Lima
Instituto Nacional de Pesquisas da Amazônia, Manaus, rath et al., 1999; Castello et al., 2013). Over the past
Brazil 150 years, boom and bust cycles of rubber, cacao, and

123
246 Hydrobiologia (2015) 750:245–255

jute production have been associated with deforesta- higher density of algivorous fishes (Powers, 1984).
tion on the mainstem Amazon floodplain downstream Algivores and detritivores are abundant in Neotropical
of Manaus (WinklerPrins, 2006), and agriculture and rivers, and increasing primary production may allow
livestock activities continue to be important drivers of more species to coexist.
floodplain deforestation (Sheikh et al., 2006; Renó Efforts to evaluate relationships between attributes
et al., 2011). Negative impacts on fish habitat of both of fish assemblages and area of flooded forest have
large hydroelectric projects and smaller stream been limited by the difficulty of measuring the
impoundments are of increasing concern (Finer & vegetation cover of inundated forest in the humid
Jenkins, 2012; Macedo et al., 2013). Recent severe tropics. Synthetic aperture radar (SAR) imagery
droughts provide raise further issues for floodplain alleviates this limitation. This tool can recognize
habitats (Tomasella et al., 2013). forests and other types of vegetation and reveal water
Riparian forests are an important habitat for beneath the canopy, allowing reasonable estimation of
freshwater fish worldwide, because forests provide the amount of flooded forest, aquatic herbaceous
fish with both shelter and food including fruits, seeds, plants, and open water habitats in floodplains (Hess
and canopy insects (Goulding, 1980; Lowe-McCon- et al., 1995, 2003). However, habitat availability is
nell, 1987; Gerking, 1994). Seasonally inundated only one factor that may affect fish richness and
riparian forests, often called flooded forest or varzea abundance in floodplains. Junk et al. (1983) reported
in the Amazon basin, are extensive in tropical South that assemblage structure changed when dissolved
America and Africa, where they form one of the main oxygen availability was reduced. Relationships among
habitats on the floodplains (Mayaux et al., 2002; Junk depth, piscivores, and fish abundance have been
et al., 2010; Melack & Hess, 2010). reported in a few studies (Winemiller et al., 2000;
Despite ongoing deforestation in tropical areas, Súarez et al., 2001). Fisheries may also reduce the
little information exists concerning its impact on abundance of fish locally. Periodicity or variability of
tropical fish and fisheries. Several studies have inundation can influence fish species richness (Jardine
addressed the relation between forest cover and stream et al., in press) and influence the structure of fish
fishes in tropical areas (Burcham, 1988; Lyons et al., communities (Sousa & Freitas, 2008), while high
1995; Toham & Teugels, 1999; Bojsen & Barriga, water levels connect habitats leading to similarity in
2002). Similar studies have not been done for large biological communities (Thomaz et al., 2007).
rivers, even though destruction of habitats supporting The purpose of our study was to test whether fish
fish populations has been hypothesized to be an richness (i.e., number of species) and abundance are
important factor affecting the long-term health of associated with the availability of flooded forest
fisheries. during high-water season on the Amazon floodplain.
Studies of the relationship between forest cover and The local importance of fish, the availability of SAR
stream fish have reached inconsistent conclusions. In mosaics obtained at high water, and the extent and
Costa Rica, species richness was found to be higher in variability of flooded forest cover make the Amazon
pasture streams than in forest streams (Burcham, basin an appropriate region in which to evaluate this
1988), but such differences were not found in Mexico association. Fish abundance was analyzed as all
(Lyons et al., 1995) or Ecuador (Bojsen & Barriga, species, as those landed in the markets, and as
2002). In addition, Toham & Teugels (1999) and algivore–detritivore species. Our results provide evi-
Bojsen & Barriga (2002) found greater fish abundance dence for assessing the potentially deleterious effect of
in forest-covered streams of Mexico and Ecuador than floodplain deforestation on tropical fish and fisheries.
in deforested streams, but Lyons et al. (1995) reported
inverse relationships for streams in Mexico. A possi-
ble explanation for the inverse relationship between Methods
fish abundance and richness and the presence of
riparian forest is the higher availability of autochtho- The seasonal flood is one of the most important
nous food in deforested streams. Because open modulators of the Amazon landscape (Junk, 1997). In
streams receive more light and therefore have higher the central Amazon, the water level on the floodplain
algae growth than shaded streams, they tend to have a varies approximately 8–12 m annually. During the

123
Hydrobiologia (2015) 750:245–255 247

Fig. 1 Geographical 64º W 60º W


locations of the sampling
sites along the floodplain of
the central Solimões/
Amazon River. Numbers 1 2º S
3 4
refer to the lakes in Table 1 7 1
1 2526
2 5 8 Manaus 29
6 27 35 31 24 28
9 34 30 36
10 33
32
19 17 18
20 23 13 14
12 11
21 22 15 16 4º S

high-water period, most lakes are connected via producing species, and in some areas, emergent, small
channels to rivers and often to other lakes through trees such as Cecropia latiloba and Vitex cymosa.
the floodplain. The three main types of habitats in Macrophytes were mostly floating grasses.
floodplain lakes are herbaceous vegetation (aquatic Thirty-five lakes were selected for sampling along
macrophytes), riparian forest, and open water. Ripar- 1300 km of the Amazon River floodplain (Fig. 1;
ian forests border the lakes, and herbaceous plants are Table 1). Lakes were selected to include a range of
usually located between the riparian forest or shore sizes and types as well as a range of riparian forest
and the open water. The relative proportions of the cover. The lakes selected could be broadly classified
three habitats vary regionally and seasonally (Hess into three main types (Sippel et al., 1992): island lakes
et al., 2003). Further descriptions of Amazon flood- (12 lakes), round lakes (7 lakes), and dendritic lakes
plain lakes, their ecology and ecosystem processes are (16 lakes). Dendritic lakes included those with shapes
provided by Junk (1997), Melack & Forsberg (2001), partly dendritic and partly round. Lake areas varied
and Melack et al. (2009). from 0.01 to 18.9 km2, falling within the range of the
Classified mosaics of JERS-1 (Japanese Earth most common lake sizes on the Amazon floodplain
Resources Satellite-1) SAR images were used for the (Sippel et al., 1992).
selection of sampling sites and calculation of habitat Sampling site locations were measured with geo-
areas. The L-band SAR images were acquired during graphic positioning system (GPS) units in the field and
September and October 1995 (low-water stage) and located on the classified SAR mosaics. Habitat areas
during May and June 1996 (high-water stage), and were measured on classified mosaics within buffers
mosaicked to create geo-referenced digital datasets with radii of 1.9, 3.9, and 7.8 km (with corresponding
with a pixel size of 3 arc-seconds, approximately 93 m areas of 11.3, 47.8, and 191 km2), centered on the
(Chapman et al., 2002). After masking out upland fishing sites. The buffers will be referred to as the 2-,
areas not subject to inundation, the mosaics were 4-, and 8-km buffers. These radii were based on
classified into habitat types based on the radar estimates of maximum distance traveled by fish from
backscattering coefficient. The classified image was sampling sites, calculated using a mean fish body
field-checked in May 2000 in 12 areas with four or length of 18 cm, a swimming speed of 1 body
more contiguous pixels of one habitat type, and found length s-1, and periods of activity of 3, 6, and 12 h.
to be 75% accurate. Further details of SAR processing, Buffer limits were manually adjusted to avoid the
classification, and validation are provided by Hess main river channel and side channels. The area of
et al. (2002, 2003). habitat types in each buffer was calculated after
Habitats present at high-water stage were flooded correcting to an equal-area projection.
forest, non-flooded forest (on high levees), aquatic At each fishing site, dissolved oxygen concentra-
macrophytes, and open water. Riparian forests were tion (DO, mg l-1), water temperature (°C), and
densely covered by tall trees (height [ 20 m) with electrical conductivity (lS cm-1) were measured
water under the canopy, including many fruit- in situ with a polarographic oxygen electrode,

123
Table 1 Names, locations, and attributes of the 35 sampled lakes; numbers in brackets refer to the geographical location in Fig. 1
248

Name Lat Long Depth Cond DO T Houses D Area Ft Ft Ft W W W M M M


(°S) (°W) (m) (lS (mg l-1) (°C) (n) (km) (km2) 2 km 4 km 8 km 2 km 4 km 8 km 2 km 4 km 8 km

123
cm-1) (km2) (km2) (km2) (km2) (km2) (km2) (km2) (km2) (km2)

Unknown (1) 3.8953 62.5930 12.0 96 2.69 27.6 1 0.8 0.27 3.4 9.9 13.2 1.2 1.6 1.9 2.1 3.5 4.1
Ipixuna (2) 3.8553 63.8797 10.4 75 2.14 26.2 300 1.8 3.46 2.0 8.8 23.9 2.8 4.6 5.2 0.7 2.9 5.5
Cipotuba (3) 3.8140 62.3451 16.0 93 2.28 28.0 1 1.5 1.86 6.5 25.1 58.2 1.6 2.8 8.6 3.3 3.8 9.8
Anori (4) 3.7725 61.6997 9.8 78 0.53 26.6 1495 7.6 0.90 8.7 31.1 128.7 1.2 7.1 19.6 2.6 6.0 20.6
U.Grande (5) 3.9235 62.0302 12.0 90 3.15 28.3 8 3.1 6.83 5.3 26.8 103.5 4.8 10.0 14.8 2.4 3.9 9.3
S. Tomé (6) 3.9398 61.4812 11.5 90 0.68 27.3 6 2.6 3.27 4.6 23.2 82.0 4.8 7.8 8.3 3.1 8.9 29.7
Rasgado (7) 3.6060 61.1632 8.0 84 2.51 27.2 95 1.3 0.84 3.1 12.4 21.2 1.4 2.3 7.7 4.7 5.8 3.4
Jacaré (8) 3.6568 60.8067 9.5 39 1.23 27.8 45 6.3 1.36 3.4 18.1 86.3 2.4 7.9 12.3 3.5 8.6 16.3
Acurú (9) 3.6025 63.6795 12.0 29 1.33 27.0 52 4.3 0.00 3.0 22.8 85.3 1.1 2.2 77.4 0.6 3.2 5.1
Comprido (10) 3.7488 63.5458 11.4 68 0.49 26.7 1 4.9 2.35 4.7 27.8 126.1 3.8 10.8 23.1 4.0 6.8 18.3
Apaurá (11) 3.8987 63.4327 8.1 93 3.81 27.6 72 0.6 5.46 1.5 8.5 22.0 4.2 6.5 11.3 2.1 3.0 4.6
Ajurá (12) 3.9544 62.7659 9.4 87 1.67 27.8 3 1.9 1.55 4.5 15.4 42.3 1.1 2.3 3.3 0.6 1.6 3.0
Luis (13) 3.9073 62.8017 10.0 99 0.31 26.8 10 1.0 1.37 6.9 15.2 33.4 1.0 2.7 4.2 1.1 1.4 2.4
Tarará (14) 2.4225 66.5028 10.0 120 2.86 25.8 15 0.4 0.73 3.4 15.1 35.6 0.6 1.5 3.0 2.7 3.5 6.8
Campina (15) 2.6139 66.3472 7.0 91 0.22 25.6 1 1.8 0.05 4.6 23.2 65.5 1.8 2.5 4.9 6.1 9.2 18.0
Curimatá (16) 2.0378 66.3339 7.0 88 2.47 27.4 120 2.0 3.18 1.4 8.7 50.2 8.8 17.7 9.4 2.0 5.6 8.6
Mirita (17) 1.9794 66.0422 6.0 70 1.49 26.9 28 2.8 0.55 2.4 20.4 59.2 3.0 3.3 5.0 3.1 3.1 6.2
Sacambu (18) 2.5671 65.8360 7.0 48 1.37 26.8 1 3.7 2.08 3.4 29.2 93.5 6.0 8.0 12.6 3.1 9.3 18.9
Tamaniquá (19) 2.7311 65.7335 10.0 71 2.28 26.4 40 3.6 2.03 6.2 35.4 133.0 1.5 2.2 8.3 2.4 3.0 8.8
Malvado (20) 2.3311 66.3550 7.0 124 2.68 26.9 6 9.5 0.65 5.8 38.9 152.4 0.2 1.4 6.9 3.4 4.9 21.7
Uará (21) 2.7110 65.6036 8.0 25 1.94 26.9 42 4.8 0.90 3.7 17.1 53.0 1.1 1.7 2.9 0.5 1.9 4.4
Alvaraes (22) 3.2052 64.8340 11.0 78 3.57 27.5 1333 2.8 0.82 3.3 10.4 20.0 1.3 1.3 1.7 0.5 1.5 2.3
Estasio (23) 2.8083 57.2044 8.0 239 1.64 27.6 8 1.5 3.92 1.2 10.3 23.9 0.2 7.1 11.9 0.3 7.7 17.6
Cararaçú (24) 2.3646 57.5990 8.5 21 3.00 26.3 1 5.4 0.95 0.6 1.3 3.2 1.2 2.2 9.2 5.0 5.8 18.4
Terra preta (25) 2.4442 57.6381 8.0 75 1.40 27.6 1 4.8 5.52 0.1 0.3 7.3 2.8 6.1 9.8 2.9 9.0 15.8
Apunumá (26) 2.9149 58.1130 8.0 31 2.80 27.9 1 0.6 1.46 2.5 5.6 7.7 2.6 3.7 3.9 4.2 3.5 3.5
Piloto (27) 2.7798 57.0921 6.3 56 2.80 28.5 4 0.9 7.35 0.8 8.7 22.8 7.2 11.0 38.5 3.0 5.6 20.4
Arari (28) 2.5111 57.1956 6.5 47 1.91 27.8 2 1.4 8.74 3.3 5.1 24.3 0.2 10.3 12.9 9.0 10.6 17.2
Corócoró (29) 3.2276 58.6921 9.0 25 0.67 27.9 1 5.4 0.10 2.4 37.6 114.7 0.5 1.1 9.5 3.9 4.1 19.5
Moaná (30) 2.8690 57.5258 8.0 206 4.20 28.8 75 1.4 10.87 5.2 8.7 24.2 2.5 13.3 16.1 4.9 7.4 14.6
Taperebá (31) 3.3117 58.6618 10.2 25 1.62 28.2 10 2.4 5.09 3.0 18.6 40.9 2.6 7.1 12.4 5.4 13.2 27.6
Garças (32) 3.1795 57.6967 7.3 92 4.54 29.2 50 4.2 3.65 0.7 24.1 57.0 0.0 4.1 6.1 0.0 4.5 9.9
Paracuubá (33) 3.0644 59.2522 12.0 12 2.35 27.7 50 10.5 0.13 1.1 2.1 7.0 0.1 0.6 7.5 1.2 2.4 20.0
Hydrobiologia (2015) 750:245–255
Hydrobiologia (2015) 750:245–255 249

Cond electrical conductance, DO dissolved oxygen, T water temperature, Houses number of houses, D distance from river, Ft, W, and M the habitats flooded forest (Ft), open water (W), and macrophytes
thermistor, and conductivity sensor, respectively. DO

(km2)
8 km

26.7
21.0
was measured in the water column every meter. The
M

maximum depth (m) of lakes was determined by sonar.


Households around the lake were counted directly or
(km2)
4 km

8.9
9.8 by interviewing community leaders. Households were
M

used as an index of potential fishing pressure. Lake


area (km2) during low water and distance from the
(km2)
2 km

5.0
4.0
M

main river channel (km) were measured on the


mosaics.
(km2)
8 km

16.4
17.7

Fish were sampled with a set of 12 gill nets with


W

3–16 cm mesh sizes (between opposite knots). All


nets were 25 m long and 2.5 m high, except those with
(km2)
4 km

5.9
10.7

3 and 16 cm mesh, which were 1.9 and 3.2 m high,


W

respectively. A new set of nets was used per lake. To


(km2)
2 km

reduce seasonal variation of fish assemblages, we used


5.4
5.4
W

three independent boats and teams, each one sampling


one section of the river basin (i) 57.1 W–59.0 W; (ii)
(km2)
8 km

63.1
16.5

60.8 W–63.8 W; and (iii) 64.8 W–66.5 W. The sam-


Ft

pling season was chosen to capture the near maximum


species richness (Mérona & Bittencourt, 1993) and
(km2)
4 km

25.4
8.7

stability in water level. Due to a small longitudinal


Ft

flood lag, we started sampling in the west side of the


(km2)

basin.
2 km

2.1
2.1
Ft

Gill nets were set at the edge of the flooded forest at


(km2)

3.1 m depth [standard deviation (SD), 0.7], to reduce


Area

2.74
8.11

dolphin attacks and improve their performance. We


inspected the nets and measured DO and temperature
(km)

2.9
3.0
D

every 4 h during 24 h. Fish were packed after fixation


Houses

with 10% formalin. Fish identification, measuring and


(n)

weighing were conducted in the laboratory at the


2
5

Instituto Nacional de Pesquisas da Amazônia. Species


27.4
28.7
(°C)

landed by the commercial fleet were classified


T

following Ferreira et al. (1998). Catch was expressed


(mg l-1)

as 1000 m-2 day-1 (Catch Per Unit Effort, CPUE)


1.84
3.12
DO

and kg 1000 m-2 day-1 (Biomass Per Unit Effort,


cm-1)

BPUE) to allow comparison with other studies and to


Cond

38
113
(lS

correct the data from three lakes, where a few nets


were lost. CPUE and BPUE include juveniles and
(M) within the three buffers of 2, 4, and 8 km
Depth
(m)

7.0
5.2

adults.
We tested the effect of several independent vari-
57.9962
57.7464
Long
(°W)

ables on the species richness, CPUE, and BPUE of all


species, commercially important species, and detriti-
2.7775
3.0619

vores with a general linear model. Because we could


(°S)
Lat

not predict which buffer size would relate better to the


Table 1 continued

dependent variables, we built three models, one for


each buffer size. One categorical (dummy) variable
Buiuçú (35)
Arroz (34)

was added to the initial models: critical mean DO. This


variable was set to 1 when mean DO was above
Name

1 mg l-1 and to 0 when DO was lower than 1 mg l-1.

123
250 Hydrobiologia (2015) 750:245–255

This DO value is the critical concentration for many Table 2 Mean, SD and range of species richness, catch per
hypoxia sensitive species in the Amazon (Val & unit of effort in number and biomass for all species (t), landed
species (com), and detritivores (d) of 35 Amazonian floodplain
Almeida-Val, 1995). Heteroscedasticity was corrected
lakes
by log-transforming the dependent variables. To
develop the statistical model, we first used the partial Mean SD Range
residuals to evaluate the fitness of variables to the Species richness 41 13 15–80
model, and then fit the variables previously selected to CPUEt (fish/1000 m -2 -1
day ) 298 153 81–681
the model. BPUEt (kg/1000 m-2 day-1) 50 26 7–105
CPUEcom (fish/1000 m-2 day-1) 248 134 69–562
BPUEcom (kg/1000 m-2 day-1) 45 24 5–92
Results CPUEd (fish/1000 m-2 day-1) 69 91 1–367
BPUEd (kg/1000 m-2 day-1) 9 10 0.1–41
The maximum depth of the lakes sampled ranged from
5.2 to 16 m; their areas reduced during the low-water
season, but only a few lakes dried completely. we split the data into lakes with DO above and below
Electrical conductivity ranged from 12 to the critical level. The number of lakes with mean DO
239 lS cm-1, indicating a mixture of river and upland below the critical level was six, too few to allow model
runoff in the more dilute waters (Table 1). The lakes fitting.
have variable human populations (0–1445 house- With the dataset reduced to 29 observations, we
holds) on their periphery. tested the coefficient of the independent variable to
Most sampling sites (70%) were located within evaluate which explained the variability in species
4 km of the Amazon River main stem. Daily mean and richness and CPUE (Table 3). Two variables were
minimum DO of the water column varied from 0.2 to consistently related to species richness and CPUE of
4.5 and 0.0 to 2.7 mg l-1, respectively. In six lakes, total and of commercial species: distance from the
the mean DO was lower than the critical value river and area of flooded forest. Other variables such as
(1 mg l-1). Mean temperature of water columns depth, average DO, and open water area were signif-
ranged from 25.6 to 29.2°C. The relative area of icant in some of the regressions. Conductivity, number
habitats changed with the size of the buffer zones. of households, temperature, area of the lake during
Flooded forest was the largest habitat in the 4 and low-water season, and the area of aquatic macrophytes
8 km buffers, followed by herbaceous vegetation and were not significant in any of the multiple regressions.
open water. However, all habitats were equally The exception was CPUE and BPUE for detritivores;
extensive in the 2 km buffer. We collected 8122 fish regressions included only the distance from the river
belonging to 195 species (species list in Granado- and DO.
Lorencio et al., 2005; further information in Granado- The species richness, abundance, and biomass of all
Lorencio et al., 2007). The average and range of fishes and of commercially important fish increased
CPUE, BPUE, and species richness at sampling sites significantly with the amount of flooded forest in the
are given in Table 2. three buffer sizes and decreased with the distance of
The first step of the statistical analyses was to the sampling site from the main river channel
evaluate the correlation between the independent (Table 4). Total CPUE decreased with average DO
variables, which could lead to co-linearity. We found and increased with the area of open water habitat and
that in the 4- and 8-km buffers there were significant with maximum depth in some buffer sizes. In most
correlations between three variables: longitude, tem- models, the independent variables explained at least
perature, and area of the lake during low-water season. 50% of the variance in fish abundance, reaching nearly
We excluded temperature and longitude from the 80% in a few cases. The CPUE and BPUE of
analyses because they seemed less related to the detritivores were inversely related to the distance
dependent variables. Second, another exploratory from the main river and with DO.
analysis showed significant interactions between crit- In all cases, the residuals showed no consistent
ical DO and flooded forest area, open water area, and tendencies. Durbin-Watson statistics were not signif-
distance to the river. Therefore, as a first adjustment, icantly different than 2 and revealed no residual

123
Hydrobiologia (2015) 750:245–255 251

Table 3 Coefficients of multiple regressions for species richness and catch per unit of effort in number and biomass for total species
(t), landed species (com), and detritivores (d) versus seven environmental variables within three buffers
Variable Buffer Depth Cond DO T Houses Area D (km) Forest Water Macr
(km) (m) (lS cm-1) (mg l-1) (°C) (n) (km2) (km2) (km2) (km2)

Richness 2 0.03 0.37 0.53 0.03 0.11 0.31 2.94* 1.72 1.24 0.36
Richness 4 1.13 0.38 1.57 0.69 1.57 2.77* 2.10* 0.83 0.79
Richness 8 1.59 0.34 0.87 0.13 0.62 2.29* 1.96* 0.07 0.19
lnCPUEt 2 1.57 0.43 2.77* 0.35 0.05 1.29 3.19** 2.90** 3.07** 0.92
lnCPUEt 4 3.23** 0.21 2.57* 0.30 0.86 4.31*** 4.09*** 3.03** 0.24
lnCPUEt 8 2.45* 0.42 2.41* 0.01 1.68 3.88** 3.76** 1.03 0.39
lnBPUEt 2 0.30 0.39 0.39 1.18 1.17 0.22 3.15** 1.27 1.20 0.76
lnBPUEt 4 1.06 0.40 1.38 0.57 0.36 3.61** 2.77* 1.06 0.49
lnBPUEt 8 0.80 0.84 1.38 0.14 0.93 3.09** 3.14** 0.71** 1.29
lnCPUEc 2 2.11* 0.63 3.21** 0.56 0.23 1.01 3.64** 2.74* 4.51*** 0.25
lnCPUEc 4 2.75* 0.13 2.01 0.54 0.26 4.021** 2.59* 3.31** 0.49
lnCPUEc 8 1.92 0.11 2.20* 0.27 1.61 3.73** 2.89** 1.28 0.20
lnBPUEc 2 0.67 1.52 1.62 0.24 1.27 0.58 3.36** 2.57* 2.18* 1.29
lnBPUEc 4 1.12 0.82 1.14 0.25 0.42 3.49** 2.34* 1.02 0.58
lnBPUEc 8 0.81 0.84 1.17 0.20 1.06 2.80* 2.51* 0.79 1.25
lnCPUEd 2 1.32 0.04 2.98** 0.88 0.77 1.96 2.09* 0.25 0.80 1.48
lnCPUEd 4 1.40 0.61 1.78 0.56 0.05 2.70* 0.68 1.15 0.32
lnCPUEd 8 1.10 0.86 2.06* 0.46 1.20 2.18* 1.08 0.46 0.75
lnBPUEd 2 1.14 1.38 0.16 1.66 0.47 1.48 1.39 1.08 0.11 1.15
lnBPUEd 4 0.45 1.40 1.21 0.37 0.25 2.87* 0.52 0.45 0.71
lnBPUEd 8 0.35 1.68 1.23 1.01 0.88 2.12* 0.82 0.42 1.16
Number of observations = 29
Cond electrical conductance, DO dissolved oxygen, D distance from river, Macr macrophytes
Significant coefficients are indicated as * 0.05 \ P \ 0.01; ** 0.009 \ P \ 0.001; *** P \ 0.0001. Values of temperature (T) are
excluded from the analyses in buffers 4 and 8 km to avoid collinearity

correlations; autocorrelation was below 0.2. Toler- level (\1 mg l-1). These results apply also to the
ances were always above 0.7, suggesting that no multi- abundance of the fish species exploited by fisheries,
collinearities were occurring between independent suggesting a real dependence of the local fisheries on
variables. the flooded forest. Conversely, the abundance of
detritivorous fish was not associated with the amount
of flooded forest, but only with proximity to the main
Discussion river and DO.
Three variables were expected a priori to correlate
Fish richness and abundance in the floodplain of the with the amount of fish caught in the floodplain, but
Amazon River were associated with the amount of did not: number of people using the resources in the
flooded forest in the buffer areas. Four- and 8-km lakes, lake area during low water, and the amount of
buffers included most of the lakes and the associated aquatic macrophytes. Residents on the riverbanks fish
floodplain. Hence, we suggest that lakes with more intensively during the high-water season. They con-
flooded forest have higher fish abundance, either in sume large amounts of fish (300–500 g day-1) and
number or biomass, than lakes with less flooded forest. also sell part of the yield to earn income (Cerdeira
Other important factors were the proximity of the et al., 1997; Fabré & Alonso, 1998); therefore, lakes
sampling sites to the main river and the critical oxygen with dense population should have fewer fish.

123
252 Hydrobiologia (2015) 750:245–255

Table 4 Results of multiple linear regressions between species richness and catch per unit of effort in number and biomass for total
(t), landed (com), and detritivorous (d) fish versus nine environmental variables using three buffer sizes
Variable Buffer (km) Forest (km2) Water (km2) Dist. (km) DO (mg l-1) Depth (m) P r2

Richness 2 3.788*** NS -2.413** NS NS \0.001 0.50


Richness 4 0.483 NS -2.898 NS NS 0.002 0.38
Richness 8 0.14 NS -3.276 NS NS 0.002 0.39
lnCPUEt 2 0.140*** 0.080* -0.089*** -0.180** NS \0.001 0.70
lnCPUEt 4 0.022*** 0.054*** -0.106*** -0.176** 0.078** \0.001 0.79
lnCPUEt 8 0.007*** NS -0.149*** -0.154** NS \0.001 0.65
lnBPUEt 2 0.164*** NS -0.125*** NS NS \0.001 0.57
lnBPUEt 4 0.026** NS -0.151*** NS NS \0.001 0.53
lnBPUEt 8 0.009*** NS -0.177*** NS NS \0.001 0.61
lnCPUEc 2 0.137*** 0.104** -0.096*** -0.210** NS \0.001 0.78
lnCPUEc 4 0.020** 0.060** -0.102*** NS 0.077** \0.001 0.66
lnCPUEc 8 0.006** NS -0.157*** -0.194 NS \0.001 0.63
lnBPUEc 2 0.166** NS -0.135** NS NS \0.001 0.58
lnBPUEc 4 0.025** NS -0.16*** NS NS \0.001 0.51
lnBPUEc 8 0.008** NS -0.183*** NS NS \0.001 0.57
lnCPUEd 2–8 NS NS -0.286 -0.821 NS 0.003 0.35
lnBPUEd 2–8 NS NS -0.259 NS NS 0.023 0.18
Only significant independent variables in at least one regression are listed. Number of observations = 29; asterisks indicate the
probability of coefficients being equal to 0; *** P \ 0.001; ** 0.01 \ P [ 0.001; no asterisks = 0.05 \ P [ 0.01

Floodplain lakes are seasonal and water volume is assemblages in the Pantanal and Araguaia River
reduced during the dry season (Melack & Forsberg, (Brazil). Lake depth and open area had a weak effect
2001). If we consider that lake area approximates on the abundance of fish in our study; its correlation
relative lake volume, one would expect that lakes was inconsistent and was dependent on buffer size.
which vary less in area would be more favorable The lack of relationship between these variables
habitats than those where the area is reduced more and fish catch in the floodplains could be accounted for
during the dry season, and that habitat stability favors by sampling limitations (type II error) or could
high species richness and abundance. This effect has represent a real lack of effect. Since we studied 35
been reported in oxbow lakes of the Brazos River lakes, our statistical resolution was reasonable despite
(Winemiller et al., 2000) and, more generally, in a the variability in the fish catch, suggesting that the
comparison of Neotropical and Australia rivers (Jar- more probable hypothesis is that these variables had a
dine et al., in press). weak effect, if any, on fish richness and abundance.
Rodriguez & Lewis (1994), Bayley (1989), and Junk et al. (1983) and Val & Almeida-Val (1995)
Súarez et al. (2001) suggested that floodplain macro- argued that fish distribution is influenced by DO in
phytes and the abundance of piscivores are the main Amazonian floodplain lakes; our results agree. Lakes
factors structuring fish communities of the Amazon, with low DO (\1 mg l-1) did not fit well to the
Pantanal, and Orinoco. Neither factor was correlated regression models. However, the inverse relationship
with fish abundance in our study. Lake depth has often between DO and fish abundance in lakes with
been assigned as an important factor to explain fish intermediate DO was unexpected: we cannot explain
abundance. Winemiller et al. (2000) found a direct this correlation, but it suggests that fish prefer sites
relation between these two variables in oxbow lakes of with median DO.
the Brazos River (USA), and Súarez et al. (2001) and The most consistent variable, which, together with
Tejerina-Garro et al. (1998) reported depth as one of the amount of flooded forest, predicted the fish catch,
the main environmental structuring factors of fish was the distance of the sampling site from the river:

123
Hydrobiologia (2015) 750:245–255 253

species richness and abundance are greater near the As a consequence, any reduction of the flooded forest
river. Migratory species, which form an important for agriculture or cattle ranching will have an effect on
fraction of the assemblage, may prefer to disperse near the number and abundance of species, including those
the river along their major migration route. Possibly, exploited by fisheries. For example, the results of this
the area near the river is an ecotone having channel study predict that a reduction of 50% of the flooded
benthic species in addition to the floodplain species. forest of a lake, distant 1 km from the river, with
Fish richness and abundance consistently increased average DO equal to 2 mg l-1, average depth of 12 m,
with the amount of flooded forest surrounding the lake. and 5 km2 of open water, would cause a decrease of
The forest provides food for fish (Goulding et al., 25% in the catch by unit and 30% by weight.
1988) and this increases the overall food available for Furthermore, these areas are nurseries for the migra-
consumers. In addition, the higher fish abundance in tory species. Therefore, reduction in flooded forest
the forest attracts piscivores, which may find more habitat would have broad impacts since many of these
prey in this complex habitat. species are found as adults in nutrient-poor black
However, not all fish seem to depend on the forest. water and clear water rivers. Loss of flooded forest
The detritivores were not directly related to this would reduce local fish production and that in nutrient-
habitat, suggesting that they can find suitable detritus poor rivers because of their dependence on recruit-
in other habitats of the floodplain. This may conflict ment from the white water nurseries.
with the assertion in Goulding et al. (1988) that Our results suggest that common farming practices
dissolved organic carbon produced by the flooded on the floodplain may be in conflict with conserving
forest is one of the main sources for detritus. Algae fish abundance. The abundance of fish is higher in the
exudates and particulate organic matter exported by flooded forest nearest to the river, yet these areas are
the streams probably are more important for this the first to be cleared to create space for buildings,
trophic guild, as previously suggested by Araujo-Lima access ways, pastures, or gardens. Forest reserves,
et al. (1986). In addition, even for fish associated with when kept, are located mostly on the inland side of the
the flooded forest, other habitats may play a role properties, where fish tend to be less abundant.
during a part of the life cycle; for example, floating
meadows provide shelter for juveniles (Henderson & Acknowledgments This study was funded by the Spanish
CYTED program and additional funds from Brazil’s CNPq and
Robertson, 1999). In a study based on fish collections NASA. JERS-1 SAR imagery was provided through the Global
over 11 months on a seasonally inundated island in the Rain Forest Mapping Project of NASDA, the National Space
Solimões River with extensive deforestation, Corredor Development Agency of Japan (now JAXA, the Japan
(2004) found that the fish community structure was Aerospace Exploration Agency). Thanks to C. Granado-
Lorencio and A.Varella for comments on an earlier version of
similar in habitats with open water, flooded forests,
the manuscript, to Michael Goulding for recent references and
and floating macrophytes. Based on sampling another comments about migratory species, to C. da Silva, Yossa, F. Chu
inundated island in the Solimões River, Petry et al. Koo, J. Sanchéz-Botero, M. Picanço, J. Penna, W. Dias, and
(2003) reported that species richness was significantly J. Marinho for their assistance during fish sampling and to the
crew of boats Veloz-I and Cap. Dario.
higher in vegetated areas than in unvegetated areas,
and stands of floating grasses had highest richness.
Both Melack et al. (2009) and Monteiro (2011)
reported weak and complex relations between flooded References
areas and fish yields.
The dependence of fish on the flooded forest Araujo-Lima, C. A. R. M., B. R. Forsberg, R. Victoria & L.
elucidated in this study enhances the importance of Martinelli, 1986. Energy sources for detritivorous fishes in
the Amazon. Science 234: 1256–1258.
this habitat. What in the past was mostly a concern of Batista, V. S. & V. J. Issac, 2012. Peixes e Pesca no Solimões-
conservationists becomes a strategic resource for the Amazonas: Uma Avaliação Integrada. Instituto Brasileiro
whole human society of the Amazon, which is highly do Meio Ambiente e dos Recursos Naturais Renováveis,
dependent on local fisheries. The consumption of fish Brasilia: 276 pp.
Bayley, P. B., 1989. Aquatic environments in the Amazon basin,
in the region is among the highest in the world with an analysis of carbon sources, fish production and
(Cerdeira et al., 1997; Fabré & Alonso, 1998), and yield. Canadian Special Publication Fisheries & Aquatic
fisheries employ about 15% of the local populations. Sciences 106: 399–408.

123
254 Hydrobiologia (2015) 750:245–255

Bojsen, B. H. & R. Barriga, 2002. Effects of deforestation on In Padoch, C., J. M. Ayres, M. Pinedo-Vasquez & A.
fish community structure in Ecuadorian Amazon streams. Henderson (eds), Várzea: Diversity, Development, and
Freshwater Biology 47: 2246–2260. Conservation of Amazonia’s Whitewater Floodplains.
Burcham, J., 1988. Fish communities and environmental char- New York Botanical Garden Press, Bronx, NY: 45–58.
acteristics of two lowland streams in Costa Rica. Revista de Hess, L. L., J. M. Melack, J. M. Filoso & Y. Wang, 1995.
Biologia Tropical 36: 273–285. Delineation of inundated area and vegetation along the
Castello, L., D. G. McGrath, L. L. Hess, M. T. Coe, P. A. Le- Amazon floodplain with SIR-C synthetic aperture radar.
febvre, P. Petry, M. N. Macedo, V. F. Renó & C. C. A- IEEE Transactions on Geoscience and Remote Sensing 33:
rantes, 2013. The vulnerability of Amazon freshwater 896–904.
ecosystems. Conservation Letters 6: 217–229. Hess, L. L., E. M. L. M. Novo, D. M. Slaymaker, J. Holt, C.
Cerdeira, R. G. P. & M. L. Ruffino, 2000. Fish catches among Steffen, D. M. Valeriano, L. A. K. Mertes, T. Krug, J.
riverside communities around Lago Grande de Monte Al- M. Melack, M. Gastil, C. Holmes & C. Hayward, 2002.
egre, lower Amazon. Fisheries Management and Ecology Geocoded digital videography for validation of land cover
7: 355–374. mapping in the Amazon basin. International Journal of
Cerdeira, R. G. P., M. L. Ruffino & V. J. Isaac, 1997. Consumo Remote Sensing 7: 1527–1556.
de pescado e outros alimentos pela população ribeirinha do Hess, L. L., J. M. Melack, E. M. L. M. Novo, C. C. F. Barbosa &
Lago Grande de Monte Alegre. Acta Amazonica 27: M. Gastil, 2003. Dual-season mapping of wetland inun-
213–228. dation and vegetation for the central Amazon basin.
Chapman, B., P. Siqueira & A. Freeman, 2002. The JERS Remote Sensing of Environment 87: 404–428.
Amazon Multi-season Mapping Study (JAMMS): obser- Jardine, T. D., N. R. Bond, M. A. Burford, D. P. Ward, P.
vation strategies and data characteristics. International Bayliss, P. M. Davies, M. M. Douglas, S. K. Hamilton, M.
Journal of Remote Sensing 23: 1427–1446. J. Kennard, J. M. Melack, R. J. Naiman, N. E. Pettit, B.
Corredor, M. C. F. V., 2004. Influência das variaçðes temporais J. Pusey, D. M. Warfe & S. E. Bunn. Flood rhythm and
da disponibilidade relativa de habitats sobre a commun- ecosystem responses in tropical riverscapes. Ecology, in
idade de peixes em um lago de várzea da Amazônia Cen- press.
tral. M.Sc., Instituto Nacional de Pesquisas da Amazônia: Junk, W. J. (ed.), 1997. The Central Amazon Floodplain: ecol-
89 pp. ogy of a pulsing system. In Ecological Studies vol. 126.
Daget, J., 1973. La pêche dans le fleuve Niger. The African Springer, Berlin.
Journal of Tropical Biology and Fisheries, Special Issue II: Junk, W. J., G. M. Soares & F. M. Carvalho, 1983. Distribution
107–114. of fish species in a lake of the Amazon river floodplain near
de Mérona, B. & M. M. Bittencourt, 1993. Fish communities of Manaus (Lago Camaleão), with special reference to
the ‘‘Lago do Rei’’, a floodplain lake in the Central Ama- extreme oxygen conditions. Amazoniana 7: 397–431.
zon: a general description. Amazoniana 12: 415–442. Junk, W. J., M. Piedade, F. Wittmann, J. Schöngart & P. Parolin
Fabré, N. N. & J. C. Alonso, 1998. Recursos ı́cticos no alto (eds), 2010. Amazonian Floodplain Forests: Ecophysiol-
Amazonas: sua importância para as populações ribeirinhas. ogy, Ecology, Biodiversity and Sustainable Management.
Boletim do Museu Paraense Emı́lio Goeldi, Serie Zoologia Ecological Studies, Springer.
14: 19–55. Lowe-McConnell, R. H., 1987. Ecological Studies in Tropical
Ferreira, E. J. G., J. A. S. Zuanon & G. M. Santos, 1998. Peixes Fish Communities. Cambridge University Press,
comerciais do médio Amazonas: região de Santarém, Pará. Cambridge.
Edições. Instituto Brasileiro do Meio Ambiente e dos Re- Lyons, J., S. Navarro-Pérez, P. A. Cochran, E. Santana & M.
cursos Naturais Renováveis, Brası́lia. Guzmán-Arroyo, 1995. Index of biotic integrity based on
Finer, M. & C. N. Jenkins, 2012. Proliferation of hydroelectric fish assemblages for the conservation of streams and rivers
dams in the Andean Amazon and implications for Andes- in west-central Mexico. Conservation Biology 9: 569–584.
Amazon connectivity. PLoS One 7: e35126. Macedo, M. N., M. T. Coe, R. DeFries, M. Uriarte, P.
Gerking, S. D., 1994. Feeding Ecology of Fish. Academic Press, M. Brando, C. Neill & W. Walker, 2013. Land-use-driven
San Diego. stream warming in southeastern Amazonia. Philosophical
Goulding, M., 1980. The Fishes and the Forest. University of Transactions of the Royal Society B: Biological Sciences
California Press, Los Angeles. 368: 20120153.
Goulding, M., M. L. Carvalho & E. G. Ferreir, 1988. Rio Negro: Mayaux, P., G. F. De Grandi, Y. Rauste, M. Simard & S. Saa-
Rich Life in Poor Water. SPB Academic Publishing, The tchi, 2002. Large-scale vegetation maps derived from the
Hague. combined L-band GRFM and C-band CAMP wide area
Granado-Lorencio, C., C. R. M. Araujo-Lima & J. Lobón-Cer- radar mosaics of Central Africa. International Journal of
viá, 2005. Abundance – distribution relationships in fish Remote Sensing 23: 1261–1282.
assembly of the Amazonas floodplain lakes. Ecography 28: McGrath, D., F. de Castro, E. Câmara, C. Futemma, C. Padoch,
515–520. J. M. Ayres, M. Pinedo-Vasquez & A. Henderson, 1999.
Granado-Lorencio, C., J. Lobón-Cerviá & C. R. M. Araujo Community management of floodplain lakes and the sus-
Lima, 2007. Floodplain lake fish assemblages in the tainable development of Amazonian fisheries. In Padoch,
Amazon River: directions in conservation biology. Biodi- C., J. M. Ayres, M. Pinedo-Vasquez & A. Henderson (eds),
versity and Conservation 16: 679–692. Várzea: Diversity, Development, and Conservation of
Henderson, P. A. & B. A. Robertson, 1999. On structural Amazonia’s Whitewater Floodplains. New York Botanical
complexity and fish diversity in an Amazonian floodplain. Garden Press, Bronx, NY: 59–82.

123
Hydrobiologia (2015) 750:245–255 255

Melack, J. M. & B. R. Forsberg, 2001. Biogeochemistry of Sheikh, P. A., F. D. Merry & D. G. McGrath, 2006. Water
Amazon floodplain lakes and associated wetlands. In buffalo and cattle ranching in the lower Amazon basin:
McClain, M., R. Victoria & J. Richey (eds), The Biogeo- comparisons and conflicts. Agricultural Systems 87:
chemistry of the Amazon Basin and Its Role in a Changing 313–330.
World. Oxford University Press, Oxford: 235–274. Sippel, S. J., S. K. Hamilton & J. M. Melack, 1992. Inundation
Melack, J. M. & L. L. Hess, 2010. Remote sensing of the dis- area and morphometry of lakes on the Amazon river
tribution and extent of wetlands in the Amazon basin. In floodplain, Brazil. Archiv für Hydrobiologie 123: 385–400.
Junk, W. J., M. Piedade, F. Wittmann, J. Schöngart & P. Sousa, R. G. C. & C. E. de C. Freitas, 2008. The influence of
Parolin (eds), Amazonian Floodplain Forests: Ecophysi- flood pulse on fish communities of floodplain canals in the
ology, Ecology, Biodiversity andSsustainable Manage- middle Solimões River, Brazil. Neotropical Ichthyology 6:
ment. Springer, Berlin: 43–59. 249–255
Melack, J. M., E. M. L. M. Novo, B. R. Forsberg, M. T. F. Súarez, Y. R., M. Petrere Jr & A. C. Catella, 2001. Factors
Piedade & L. Maurice, 2009. Floodplain ecosystem pro- determining the structure of fish communities in Pantanal
cesses. In Gash, J., M. Keller & P. Silva-Dias (eds), lagoons (MS, Brazil). Fisheries Management and Ecology
Amazonia and Global Change. Geophysical Monograph 8: 173–186.
Series 186. American Geophysical Union, Washington: Tejerina-Garro, F. L., R. Fortin & M. A. Rodriguez, 1998. Fish
525–541. community structure in relation to environmental variation
Monteiro, R. M., 2011. Relação entre área inundada e o des- in floodplain lakes of the Araguaia River, Amazon Basin.
embarque pesqueiro na região de Tefé na Amazônia Cen- Environmental Biology of Fishes 51: 399–410.
tral. M.Sc., Instituto Nacional de Pesquisas da Amazônia: Thomaz, S. M., L. M. Bini & R. L. Bozelli, 2007. Floods
80 pp. increase similarity among aquatic habitats in river-flood-
Petry, P., P. B. Bayley & D. F. Markle, 2003. Relationships plain systems. Hydrobiologia 579: 1–13.
between fish assemblages, macrophytes and environmental Toham, A. K. & G. G. Teugels, 1999. First data on an Index of
gradients in the Amazon River floodplain. Journal of Fish Biotic Integrity (IBI) based on fish assemblages for the
Biology 63: 547–579. assessment of the impact of deforestation in a tropical West
Power, M. E., 1984. Grazing responses of tropical freshwater African river system. Hydrobiologia 397: 29–38.
fishes to different scales of variation in their food. In Zaret, Tomasella, J., P. F. Pinho, L. S. Borma, J. A. Marengo, C.
T. M. (ed.), Evolutionary Ecology of Neotropical Fresh- A. Nobre, O. R. F. O. Bittencourt, M. C. R. Prado, D.
water Fishes. Dr. W. Junk Publishers, The Hague: 25–37. A. Rodriguez & L. A. Cuartas, 2013. The droughts of 1997
Renó, V. F., E. M. L. M. Novo, C. Suemitsu, C. D. Rennó & T. and 2005 in Amazonia: floodplain hydrology and its
S. F. Silva, 2011. Assessment of deforestation in the lower potential ecological and human impacts. Climatic Change
Amazon floodplain using historical Landsat MSS/TM 116: 723–746.
imagery. Remote Sensing of Environment 115: Val, A. L. & V. M. F. Almeida-Val, 1995. Fishes of the Amazon
3446–3456. and Their Environment: Physiological and Biochemical
Rodriguez, M. A. & W. M. Lewis Jr, 1994. Regulation and Aspects. Springer, Berlin.
stability in fish assemblages of neotropical floodplain Winemiller, K. O., S. Tarim, D. Shormann & J. B. Cotner, 2000.
lakes. Oecologia 99: 166–180. Fish assemblage structure in relation to environmental
Santos, G. M. & A. C. M. Santos, 2005. Sustentabilidade da variation among Brazos River oxbow lakes. Transactions
pesca na Amazônia. Estudos Avançados 19: 165–182. of the American Fisheries Society 129: 451–468.
Santos, G. M., E. J. G. Ferreira & A. L. Val, 2014. Recursos WinklerPrins, A. M. G. A., 2006. Jute cultivation in the Lower
pesqueiros e sustentabilidade na Amazônica: Fatos e per- Amazon, 1940–1990: an ethnographic account from San-
spectivas. Heléia (Revista do Direito Ambiental da tarém, Pará, Brazil. Journal of Historical Geography 32:
Amazônia) 8: 43–77. 818–838.

123

Das könnte Ihnen auch gefallen