• Keine Ergebnisse gefunden

Differential role of allochthonous and autochthonous carbon sources in the functioning of Neotropical seasonal wetlands

Luisa F. Vega, Marcelo Z. Moreiraand Karl M. Wantzen

In prep.

Abstract

Due to the over proportionally high ecological and economical value of wetlands, it is important to understand how the different carbon sources take part in the functioning of these systems. This study is specifically focused on the role of allochthonous and autochthonous carbon sources in seasonal shallow lakes (SSLs) from the Colombian Orinoco-Llanos. We conducted stable carbon isotope analysis of sediment, seston and most abundant plant and fish species. Samples were taken mostly after the season peak, during both flooded and non-flooded seasons in order to get the maximum seasonal effect on these wetland components.

Additionally, we analyzed flood pulse pattern together with depth at sampling points to obtain the sediment, seston and plant sample positions on the flood gradient. The δ13C values of allochthonous plants and sediment were mostly above δ13C = -20.5‰, the intermediate δ13C value between C4 (-13‰) and C3 (-28‰) plants. This suggests that sediment has predominately carbon with C4 plant signature. Contrary, the δ13C values of autochthonous plant assemblages, seston and fish were mostly under δ13C = -20.5‰, showing more influence of carbon with C3 plant signature. Some lower δ13C values of fish (-32.20, -32.43, -35.64 ‰) were attributed to phytoplankton consumption. We conclude that allochthonous and autochthonous carbon sources have different roles in the functioning of SSLs, being allochthonous carbon more important for carbon sequestration in sediment and autochthonous carbon for secondary production; both valuable wetlands environmental services.

Key words:C3 and C4 plants; stable isotope ratios; carbon storage; food web; Orinoco-Llanos

60

Introduction

The allochthonous and autochthonous inputs of energy and nutrients play important roles in aquatic ecosystems. Their importance varies according with the systems. In the case of rivers, they have given origin to different ecological concepts about their functioning, such as the River Continuum Concept, RCC (Vannote et al. 1980), the Flood Pulse Concept, FPC (Junk et al. 1989, Junk and Wantzen 2004) and the Riverine Productivity Model, RPM (Thorp and Delong 1994). The RCC proposes a gradual replacement from allochthonous to autochthonous carbon sources from small to medium river order and then a gradual return to allochthonous sources from medium to high river order. It emphasizes the importance allochthonous organic matter from high vegetated reaches of rivers (Vannote et al. 1980). The FPC postulates that in river-floodplain systems a large part of the production occurs in floodplains. During the flooded phase, the secondary production is based on drowned terrestrial organic material, highlighting the importance of floodplain-derived carbon and nutrients sources (Junk et al. 1989, Junk and Wantzen 2004). On other hand, the RPM give emphasis to the importance of autochthonous aquatic carbon sources from phytoplankton, benthic and aquatic plants, supporting food web in the main channel of the river (Thorp and Delong 1994). A later review of the above mentioned concepts and models is the Riverine Ecosystem Synthesis, RES (Thorp et al. 2006). It also suggests autochthonous algal production, as principal organic carbon sources in riverine systems (Thorp et al. 2006).

Although some of these concepts and models take into account floodplain wetlands from river-floodplain systems, we would like to explore to what extent they can be applied to tropical seasonal wetlands, relatively isolated from river main channel. This study is specifically focused on the role of allochthonous and autochthonous carbon sources to both carbon sequestration and secondary production, in seasonal shallow lakes (SSLs) from the savannahs of the Orinoco-Llanos, Colombia.

These SSLs provide good natural scenery to study the importance of allochthonous and autochthonous carbon sources in wetland functioning, as they are dominated by C3 plants (Vega et al. 2014 ), but located in the middle of savannahs covered by C4 grass (Vega et al.

unpublished manuscript). Most of the aquatic macrophytes are C3 plants, showing δ13C values around -28‰; while C4 plants, which are mainly the allochthonous vegetal production, have δ13C values, in average near -13‰ (O'Leary 1988). Thanks to the different carbon isotope

61

ratios of C3 and C4 plants, relative stability of these ratios through the food web and little carbon fractionation during organic matter decomposition (Fellerhoff et al. 2003, Fry 2006), we can track the pathways of these different carbon sources across SSL systems.

Among aquatic and terrestrial ecosystems, wetlands have been recognized to be highly productive, delivering a wide range of environmental services, from fish nurseries and production, to carbon sequestration. (Mitra et al. 2005, Barker and Maltby 2009, Keddy et al.

2009, Russi et al. 2013). Their carbon sources are essential to ensure these environmental services. Due to the over proportionally high ecological and economical value of wetlands, it is important to understand how the different carbon sources take part in the functioning of these systems. In a previous study in the same seasonal savannah wetlands, we found that carbon with C4 plant signature was significant higher in their sediments than in their plant assemblages, suggesting an important carbon contribution from allochthonous sources to sediment carbon storage (Vega et al. unpublished manuscript). This phenomenon gives rise to two alternative hypotheses: (i) the autochthonous C3 biomass production is less recalcitrant (lower C:N ratio) to decomposition and is therefore preferably incorporated into the food web compared with terrestrial allochthonous C4 biomass, so that the latter becomes in sediments in higher percentages than C3 biomass; or (ii) the allochthonous C4 biomass input exceeds that of autochthonous C3 biomass and therefore it overrides the C3 plant signature in sediment.

Methods

Study Areas

We performed our studies at a seasonal shallow lake (SSL) wetland named ‘El Venado’, situated at a geographic position of 4°51.497`N, 72°30.417`W (lake center). This wetland is located in the seasonal flooded savannahs of the Orinoco-Llanos. It belongs to the Cusiana River basin, at approximately 4.5 Km of the river main channel (Fig 1). This area has been described in a previous ecological study (Vega et al. 2014 ). These SSLs, denominated locally esteros, arise in savannahs depressions originated by paleo channels or subsidence zones.

Here soil level is lower and water saturation is higher than in other physiographic units of the savannah (Vega et al. unpublished manuscript). They are relatively isolated from the river main channel. Therefore their water level is principally influenced by local rains and

62

secondary by river flood pulse via back flooding from the river main stem, rather than from direct bank overflow. Nevertheless, they may be connected to the main channel, during large flood events of approximately decennial frequency or more.

Fig. 1 Location of the El Venado and other studied seasonal shallow lakes in the Cusiana river basin, Orinoco-Llanos.

The SSLs presents a flood seasonality related to the local rainfall seasonality (Fig. 2). Both flood and rains have similar duration, but flooded/non-flooded seasonal transition take place approximately from one to two months later than rainy/dry seasonal transition. The long-flooded season, from 9 to 10 months, allows C3 plants be dominant inside the SSL around the whole year, however during non-flooded season some C4 grass star to appear (Vega et al.

unpublished manuscript). The surroundings of the SSLs show more extensive drought phase, resulting in a dominance of C4 plants.

63 Fig. 2 Flood pulse, rainfall and temperature of the study area in the Orinoco-Llanos. Water level fluctuation of El Venado seasonal shallow lake, from August 2010 to July 2013. Rainfall and temperature (1993-2012) at the Tamarindo Climatological Station (05°01'N, 72°33'W) and the Aguazul Climatological Station (05°10'N, 72°33'W), respectively (IDEAM, Instituto de Hidrología, Meteorología y Estudios Ambientales – Institute of Hydrology, Meteorology and Environmental Studies), Colombia.

The fact that El Venado SSL is located in the middle of savannahs covered by C4 grass; it is dominated by C3 aquatic macrophytes and additionally it shows only little influence from river main stem; make it a good scenery to study carbon dynamics using stable isotope signatures.

64

Experimental Design and Sampling

We collected sediment, plant biomass, seston and fish samples, mostly after the season peak, during flooded and non-flooded seasons in order to get the maximum seasonal effect on these SSL components. The experimental design and sampling was conducted for the different SSL components in the following way:

Sediment and Plant Biomass

Sediment and plant biomass samples were collected from 4 to 10 November 2010 and from 24 to 25 February 2011, during flooded and non-flooded seasons, respectively. They were collected on two transects parallels to the longest axis of the SSL. Every 60 m, at geo-referenced points, we took sediment and plant biomass samples, giving a total of 19 sampling points. The samples were collected at the same points during both seasons and each sample was composed of three subsamples (see Vega et al. (2014) for a detailed explanation of transect design and sediment and biomass sampling).

In the Laboratory, plant biomass was washed, sorted into groups and species, identified and dried at 60°C to constant weight. Later during the same season, we collected three tissue samples (composed of at least three different individuals) per species of the most abundant plant species. The plant tissue was cleaned, dried and grounded to be used for stable isotope analysis (SIA). Similarly, sediment subsamples were mixed, cleaned of debris and dried at 60°C to constant weight. Dry sediment was ground and passed through a 125µm stainless steel sieve (see chapter 3 for a detailed explanation of laboratory processing of sediment and plant biomass).

Seston

In El Venado SSL, seston was collected by nine traps. Each trap consisted of 3 tubes (11.4 cm² base x 30 cm height) disposed at the same distance on the bottom of a cylinder plastic basket (706.86 cm² base) (Fig. 3). The mesh basket allows seston entrance, but protect tubes of been blocked by plants or disturbed by animals. The seston traps were settled on the SSL bottom (Fig. 3), located at three different depths (70, 110 and 150 cm). The transects were located perpendicular to the SSL margins.

65 Fig. 3 Photographs of seston trap design and underwater exposition.

During flooded season, the seston traps were exposed from 10 August to 18 November 2010.

During the non-flooded season, the traps were exposed until the pipes were cover by water.

Due to constant water evaporation in this season, the traps located at shallower depth (70 and 110 cm), were exposed for a shorter period (16 December 2010 to 5 February 2011) than the traps located deeper, at 150 cm (16 December 2010 to 30 March2011). After the exposition period, the traps were retrieved and the pipe contents were emptied into labeled flasks. The flasks were transported in a thermal box with ice to the Laboratory. There, the flask contents were transferred to aluminous trays and placed in a fridge for 24 hours to allow decantation of the seston. After that, supernatant water was removed and seston was dried (at 60°C to constant weight) and grounded.

Fish Assemblage

We assessed fish assemblages during flooded and non-flooded seasons in El Venado SSL and in other four similar SSLs (El Güio, Pequeño, Pica-Pica and Flor Amarillo) around it (Fig 1).

Like sediment and plant biomass, fish assemblages were evaluated after the season peaks. To study fish assemblages, we applied two different approaches: throw trap and trawl. The throw trap is especially used to sample fish in shallow vegetated habitats (Jordan et al. 1997). It consists of a cubic metallic frame of 1mside, enclosed by a nylon material, leaving opened its bottom and top (Fig. 4.d, Chapter 1). The throw trap was located every 60 m, next where sediment and plant biomass were sampled. The throw trap was placed first carefully above water on the area to be sampled, and then fast sank and pressed into the sediment. From the opened top, a cone fish net was passed into the throw trap to take out and count the caught individuals. The process with the cone fish net was repeated until we did not capture any more individuals for ten consecutive times (Fernandes et al. 2010). From the five SSLs, 91 m2 were

66

assessed during flooded season. As complement to the throw trap sampling we also used the trawl. It was 5 m long and 1.5 m high, with mesh size of 5 mm. The studied SSLs were densely covered by aquatic plants; therefore we passed the trawl in different areas where vegetation was not so dense or in open water, between 7-10 times per SSL. In non-flooded season, all throw trap sampling points were dry, therefore we used only trawl in the deepest SSL areas where water remained. Fish were identified if possible in situ and their abundance registered; in order to release them back into the lake. Some individual specimens were preserved in 10% formalin for posterior identification.

Once the fish species had been identified and quantified, we prepared five samples from one to three of the most abundant fish species per found guild (omnivorous, carnivorous and detritivorous) and per season, to carry out SIA. For species smaller than 40 mm we took the whole body. For bigger fish we took ca. 1cm3 of epaxial muscle. The fish samples were thoroughly washed with distillated water, dried in a ventilated oven at 60°C to constant weight, and grounded with a mortar and pestle.

Stable Isotope Analyses

The grounded plant and fish tissue, seston, and the fine portion of sediment were weighted and used to calculate the carbon and nitrogen ratios (δ13C and δ15N) with a continuous-flow isotope ratio mass spectrometer (ThermoScientific, model Delta Plus, Bremen, Germany coupled to a CarloErba Elemental Analizer CHN-1110, Milan, Italy), at the Isotope Ecology lab of the CENA-USP, Centro de Energia Nuclear na Agricultura -Nuclear Research Centre in Agriculture - Sao Paulo University, Piracicaba, Sao Paulo State, Brazil.

Flood Gradient, Rainfall and Temperature

In the Venado SSL, we recorded monthly the water level from August 2010 to July 2013, using a 2.30 cm long ruler installed in the middle of the lake. These data were used to estimate the daily and monthly water level fluctuation. Then, we computed it with the depth at the sampling events of each sampled point to obtain their number of flooded days/year and consequently, the flood gradient. Rainfall and temperature data (from 1993 to 2012) were taken from the Tamarindo Climatological Station (05°01'N, 72°33'W) and the Aguazul Climatological Station (05°10'N, 72°33'W), respectively; provided by IDEAM, Instituto de

67

Hidrología, Meteorología y Estudios Ambientales – Institute of Hydrology, Meteorology and Environmental Studies, Colombia.

Data Analysis

We compared δ13C values of the different studied SSL components: allochthonous plant species, sediment, autochthonous plant assemblages, seston and fish. As allochthonous plants, we took the δ13C values of Andropogon bicornis (δ13C=-13.1‰ (Vidotto et al. 2007)) and Axonopus purpusii (δ13C=-14.5‰ (2008)). These two perennial grasses are the most abundant species in the physiographic units surrounding SSLs (Vega et al. unpublished manuscript). To estimate the δ13C values of plant assemblages, we used δ13C values of the most abundant plant species (> 90 % of above ground green biomass) and some less abundant plant species to which their δ13C values were available from the SIA in this or in our previous study (see chapter 3 for a detailed explanation of δ13C value estimation of plant assemblages). As seston data are from El Venado SSL, we included here only the δ13C values of sediment and autochthonous plant assemblages from this lake, for a better understanding of the relations between these three SSL components. However, in our previous study (see chapter 3), we found that δ13C values of sediment and autochthonous plant assemblages have similar dynamics among other studied SSLs around El Venado. To extend our final conclusions, we included the δ13C values of fish from the five SSLs where fish assemblages were assessed (Fig. 1). The significance of the difference between-SSL component and season comparisons of δ13C values was determined applying a Wilcoxon test.

Additionally, a simple mixing model was used to estimate the percentage of carbon contribution from C3 and C4 plants to sediment, autochthonous plant assemblage, seston and fish. It was calculating according to the equation from Fry (2006):

where %CC3 is the percentage of carbon contribution from C3 plants, δ13Csample is δ13C value of a sample (i.e. sediment, autochthonous plant assemblage, seston or fish), δ13CC4 is the δ13C value from C4 plants (-13‰) and δ13CC3 is the δ13C value from C3 plants (-28‰).

Finally, we plotted the δ13C values of seston with paired (at a given site) δ13C values of plant assemblages and sediment regarding the number of flooded days/year at the sampling points

%CC3 = [(δ13Csample - δ13CC4) / (δ13C C3 - δ13CC4)]*100,

68

to observe the distribution of δ13C values of these three SSL components respecting their position in the flood gradient. All data analyses were conducted using the R program version 3.0.1 (R Core Team 2013).

Results

Autochthonous Plant Assemblages

In El Venado SSL, the most abundant plant species in terms of biomass percentage along flooded and non-flooded season were the following C3 plants: Luziola fragilis (65.77% and 64.85 %; δ13C = -28.34 ± 0.10 ‰ and -29.20 ± 0.08 ‰, respectively), Pontederia triflora (20.02 % and 15.51%; δ13C = -28.40 ± 0.10 ‰ and -28.49±0.13‰, respectively), P. subovata (4.34 % and 1.65%; δ13C = -29.16 ± 0.23‰ and -29.99 ± 0.21‰, respectively) and Eleocharis minima (2.75% and 2.24%; δ13C = - 20.27 ± 1.36 ‰, respectively). They counted for 92.87%

and 84.25 % of the total aboveground green biomass in the flooded and non-flooded season, respectively. During the non-flooded season, a C4 grass, Poaceae sp.1 (12.45%, δ13C = -13.27

± 0.04‰) started to cover some dried areas of the SSL.

Fish Assemblages

The most abundant fish species per guild among the SSLs in the flooded and non-flooded season, were the omnivorous: Hemigrammus marginatus (Sanabria 2004) 13C = -25.33 ± 0.29 ‰ and -29.08 ± 0.43 ‰, respectively), Bryconamericus cf.cismontanus (Taphorn 2003) 13C = -25.05 ± 0.84 ‰ and -23.97 ± 0.20 ‰, respectively) and Pyrrhulina lugubris (Taphorn 2003) (δ13C = -26.02 ± 1.01 ‰ and -23.25 ± 0.77 ‰, respectively); the carnivorous:

Eigenmannia virescens (Galvis et al. 2007) (δ13C = -28.74 ± 0.57 ‰ and -26.14 ± 0.44 ‰, respectively) and Hoplias malabaricus (Taphorn 2003) (δ13C = -24.78 ± 1.19 ‰ and -23.84 ± 0.15 ‰, respectively) and the detritivorous: Steindachnerina argentea (Taphorn 2003) (δ13C

= -32.20 ± 0.79 ‰ and -35.64 ± 0.35 ‰, respectively). During the flooded season were also abundant the carnivorous Charax gibbosus (Taphorn 2003) (δ13C = -28.25 ± 0.19‰) and the detritivorous Cyphocharax spilurus (Taphorn 2003) (δ13C = -32.43 ± 0.19‰) and in the non-flooded season, the carnivorous Gymnotus carapo (Galvis et al. 2007) (δ13C=- 23.80±0.85‰).

69

δ13C Values of SSL components

Fig. 4 δ13C values of allochthonous plant species (AL-PS), sediment (Sed), autochthonous plant assemblage (AU-PA), seston and fish. The median, quartiles, 1.5 interquartile range and outliers are indicated. The values above the boxplots are the number of data. The horizontal dash lines indicate -20.5‰, the intermediate δ13C value between C4 (-13‰) and C3 (-28‰) plants.

Table 1 Significance of the difference between-component and season comparisons (Wilcoxon test) of δ13C values from seasonal shallow lakes (SSL). AL-PS = allochthonous plants species, Sed = sediment, AU-PA = autochthonous plant assemblage. different, except for autochthonous plant assemblages and fish. During non-flooded season, the δ13C values of autochthonous plant assemblages, seston and fish were not significantly different (Fig. 4, Table 1). Comparing flooded and non-flooded seasons, autochthonous plant assemblages were not significantly different between seasons and it was just marginally

70

different from seston. Seston did not change among seasons and it was also similar to autochthonous plant assemblages and fish. Fish was significantly different among season, but similar to autochthonous plant assemblages and seston (Fig. 4, Table 1).

Although the δ13C mean values of sediment and allochthonous plants were significantly different (Table 1), they were more similar between them than with the other SSL components (Fig. 4). Contrary to the other SSL components, the δ13C values of allochthonous plant species and sediment were mostly above δ13C = -20.5‰, the intermediate δ13C value between C4 (-13‰) and C3 (-28‰) plants (Fig. 4). This suggests that sediment signature has a stronger impact from C4 plant carbon. The δ13C values of autochthonous plant assemblages, seston and fish were mostly under δ13C = -20.5‰, indicating a higher carbon content with C3 plant signature (Fig. 4).

According to the mixing model, the percentage of carbon contribution of C3 plants to each SSL component in flooded season was: sediment 25.12 % ± 1.84, autochthonous plant assemblages 98.88 % ± 0.93, seston 80.57 % ± 5.67 and fish 90.96 % ± 2.01. During non-flooded season, the percentage of carbon contribution from C3 plants to the SSL components was: sediment 34.15 % ± 3.54, autochthonous plant assemblages 81.19 % ± 5.36, seston 88.52

% ± 4.75 and fish 81.82 % ± 2.42.

Sediment, Seston and Plant Assemblage in the Flood Gradient

The δ13C values of sediment had similar trend among both season in regard to the flood gradient, declining when the number of flooded days/year increased (Fig. 5). The δ13C values of plant assemblages during flooded season were constantly low along the flood gradient, being more dominated by C3 plants. During the non-flooded season, the δ13C values of plant assemblages increased for flooded periods shorter than 265 days, showing influence of C4

plants. It is consistent with the C4 grasses arise in this season. For longer flooded periods than 265 days, the δ13C values of plant assemblages were again low, showing influence of C3

plants (see chapter 3 for detailed explanation of the break point calculation at 265 days flooded days/year). The δ13C values of seston showed an unclear pattern when flood was short. Carbon signature of C3 and C4 plants were found in the seston at the same position on the flood gradient in the two seasons (Fig. 5). The longer the flooding, the closer the carbon

plants (see chapter 3 for detailed explanation of the break point calculation at 265 days flooded days/year). The δ13C values of seston showed an unclear pattern when flood was short. Carbon signature of C3 and C4 plants were found in the seston at the same position on the flood gradient in the two seasons (Fig. 5). The longer the flooding, the closer the carbon