Variability in chemistry of surface and soil waters of an evapotranspiration-dominated flood-pulsed wetland: solute processing in the Okavango Delta, Botswana

Keotshephile Mosimane1*, Eric Struyf2, Mangaliso J Gondwe1, Patrick Frings3, Dimitri van Pelt2, Piotr Wolski4, Jonas Schoelynck2 , Jörg Schaller5,6, Daniel J Conley3 and Mike Murray-Hudson1 1Okavango Research Institute, University of Botswana, Private Bag 285, Maun, Botswana 2ECOBE, Antwerp University, Universiteitsplein 1, Antwerp, Belgium 3Department of Geology, Lund University, Sölvegaten 12, Lund, Sweden 4Climate System Analysis Group, University of Cape Town, South Africa 5Institute of General Ecology and Environmental Protection, Technische Universität Dresden, D-01062, Dresden, Germany 6Environmental Geochemistry, Bayreuth Center for Ecology and Environmental Research (BayCEER), University Bayreuth, Universitätsstraße 30, 95447 Bayreuth, Germany


INTRODUCTION
Processes of solute transport and storage in wetlands are crucial for wetland functions and structure (Mitsch and Gosselink, 2007). The introduction of solutes into wetland systems occurs with the inflow of water as dissolved chemical species, or as a result of the dissolution of riverine or aeolian particulates (Garstang et al., 1998;Krah et al., 2004;Krah et al., 2006;Baker et al., 2009). Once in the system the solutes may undergo a series of transformations due to interactions between hydrological, biological and geochemical processes (Mitsch and Gosselink, 2007). Solute processing has important implications for biogeochemical processes, and thus for water quality functions. Water quality affects the ability of aquatic environments to sustain healthy ecosystems and thereby the provision of ecosystem services (Stark et al., 2000). Therefore, identifying factors that affect solute processing in a wetland system is fundamental for sustainable management of the resources to ensure sustained provision of ecosystem services.
In evapotranspiration-dominated wetland systems, the input and final fate of solutes is of critical ecological importance (Eugster and Jones, 1979;Eugster and Maglione, 1979;McCarthy and Ellery, 1998;Boettinger and Richardson, 2001). The progressive removal of water by evapotranspiration may ultimately cause solutes to reach saturation, resulting in their removal from solution through precipitation (Barbiero et al., 2002;Barnes et al., 2002;Furquim et al., 2004;Humphries et al., 2010;Humphries et al., 2011). This plays a major role in the chemical evolution of waters in such systems. For example, in the lower floodplain of the Mkuze wetland of South Africa, chemical precipitation under the influence of evapotranspiration was identified as the main factor responsible for the permanent removal of chemical solutes from solution (Humphries et al., 2010). Similar processes have been observed in the Florida Everglades (Wetzel et al., 2005).
In the Okavango Delta ( Fig. 1), approximately 98% of the water that enters the system (rainfall and catchment runoff) is lost to the atmosphere due to evapotranspiration throughout the year (Wilson and Dinçer, 1976;Gumbricht and McCarthy, 2003), leaving only about 2% of the mean inflow to be drained at the distal end (Gieske, 1996;. Despite such evaporative losses, the Delta remains a predominantly freshwater system. This is because solutes are sequestered beneath the islands by a process of unidirectional mass transfer from the floodplains to the numerous islands that are present in the Delta (McCarthy et al., 1993;McCarthy and Ellery, 1995). This process is driven by vegetation-induced local lowering of the water-   the Delta as a freshwater wetland system (Fig. 2) McCarthy et al., 1993;Bauer, 2004;Bauer et al., 2007;Ramberg and Wolski, 2008). From mass-balance calculations this results in the accumulation of about 360 000 t of dissolved solutes per year in the Okavango Delta . There are indications that differential partitioning of various solute species occurs in the transition between surface water and island groundwater (Mladenov et al., 2014;Struyf et al., 2015;, possibly as a result of pH or redox potential (Eh) changes. The processes dominating the cycling of solutes in surface-and soil waters of the Delta, before they eventually precipitate or are sequestered in highly saline groundwater beneath island centres, are not fully understood, however.

Figure 2
Main biogeochemical processes in water and islands of the Okavango Delta (Source: Ramberg and Wolski, 2008) none of these studies looked at concentrations in pore water. These previous studies, with the exception of those of Mackay et al. (2011) andMladenov et al. (2005), were either strictly localised or focused mainly along the Jao-Boro channel (Fig. 1). They found a gradient of increasing solute concentration in surface water from the Panhandle to the outlet of the Jao-Boro channel. Pronounced gradients in vegetation standing stock and hydroperiod (both in terms of duration and depth) exist at the landscape scale and at the sub-kilometre scale of the complex mosaic of channels, floodplains and islands that characterises the Delta. We hypothesise that these gradients may both result from, and influence, the distribution and biogeochemical behaviour of solutes.
To understand the early stages of solute cycling in the Delta we studied their spatial distribution and the associated processes in surface water and pore water. This paper provides a comparison of the distribution of chemical species, in soil pore water and surface water, along island-floodplain-channel gradients, in seasonally and permanently inundated habitats, and between major regions (the Panhandle, Guma, Xakanaxa, Nxaraga and outlets) in the Okavango Delta. We hypothesised that in the Panhandle and the permanent floodplains, solute processing is biologically dominated by floodplain emergent vegetation, whereas in the more distal parts of the Delta, it is driven by chemical precipitation through evapotranspiration.

Study area and site description
The Okavango Delta (Fig. 1), ~19˚ to 20˚ S and 22˚ to 24˚ E, is geomorphologically the terminal alluvial fan of the Okavango River in north-western Botswana. It experiences annual floodpulsing by Okavango River flows from seasonal rainfall in the Angolan highlands, 1 000 km to the north-west. Mean annual discharge into the Delta is approximately 9 x 10 9 m 3 while local summer rainfall contributes an additional 42% (about 6 × 10 9 m 3 per year) to the mean flow Gumbricht et al., 2004a;Gumbricht et al., 2004b;Ramberg and Wolski, 2008). The annual flood pulse that inundates the Delta causes fluctuations in the size of the inundated area, from 3 500 to 6 000 km 2 during low-flow seasons to ≤ 9 000-13 000 km 2 during highflow seasons (Gieske 1997;. Consequently, the Delta has been categorised into different eco-regions. These include the Panhandle (~800 km 2 ), permanent swamp (~2 500 km 2 ), seasonal floodplains (~3 300 km 2 ) and occasional floodplain (~7 100 km 2 ) (Gumbricht et al., 2004b).
Four study regions, namely, the Panhandle, Guma, Xakanaxa and Nxaraga (hereafter referred to as the system-scale longitudinal gradient), were selected for sampling (Fig. 1). These regions were chosen in order to cover different habitat types with different hydrological and vegetation gradients ( Table 1). The Panhandle, Guma and Xakanaxa are generally characterised by permanently flooded habitats but with different ranges of water fluctuation, while Nxaraga is characterised by seasonally inundated floodplains. South-east of the lower Panhandle the main channel (Okavango River) splits into three distributaries, the Thaoge to the west, the Jao-Boro flowing south-east and the Nqoga flowing eastwards. Guma is located within the Thaoge distributary system. Xakanaxa is part of the main distributary (Nqoga) just east of Chief 's Island (the Delta's largest island). Nxaraga makes up part of the Jao-Boro secondary distributary system, on the south-east tip of Chief 's Island. In all the regions except the Panhandle, transects spanning local gradients in hydrology and vegetation, from island centres across the floodplain and into the channel (hereafter referred to as local gradients), were established. A total of 10-11 sample points along each transect (1 transect per region, a total of 3 transects (n = 3)) were selected purposively to capture the variation apparent in the vegetation cover.

Sample collection
In September 2011, a total of 52 channel and floodplain surface water samples were collected from the Okavango River at Mohembo, Thaoge River at Guma, Moanatshira River at Xakanaxa and along the Boro River from Nxaraga to Maun (Fig. 1). Additional surface water samples were collected from river outlets near bridges at Maun, Shashe, Nxotega, Toteng, and Tsau (Fig. 1). All surface water samples were collected from 10 cm below the surface, avoiding any surface debris, and filtered through 0.45 µm nitrocellulose Chromafil syringe filters (A-45/25) into clean acid-washed polyethylene bottles. Water samples for metal analyses were acidified to pH < 2 with 1 N nitric acid. All samples were stored at < 4˚C, both in the field (using a portable car freezer) and in the environmental laboratory at the Okavango Research Institute (ORI), before transport for analysis at the University of Antwerp, Belgium.
At each sampling point, surface water temperature (˚C) and electrical conductivity (EC, µS·cm -1 ) were measured in situ using a Cond 3110/SET conductivity meter. Dissolved oxygen (DO, mg·L -1 ) was measured using an Oxi 3203/SET oxygen meter and pH was measured using a pH 3110/SET meter, all by Wissenschaftlich-Technische Werkstätten (WTW), Germany. All field equipment was calibrated before each set of data collection.
Soil water samples were obtained from surface soil (5 cm) samples collected along island-floodplain transects at Guma, Nxaraga and Xakanaxa (Fig. 1). A total of 3 surface soil samples (which were later pooled together) were collected from 0.25 m 2 plots at each sampling point along the three island-floodplain transects using Kopecky rings (5.1 cm height, 5 cm diameter, 100 cm 3 volume), following the method described by Javaux and Vanclooster (2006). All soil samples were stored at < 4˚C during sampling (using a portable car freezer) and frozen until further processing in the environmental laboratory at ORI. Soil water was extracted from the pooled soil samples by centrifuging thawed samples at 3 000 r·min -1 for 10 min. To minimise changes in redox potential (Eh), every attempt was made to limit air exposure during the handling and centrifugation of sediment samples. Soil samples were stored in sealed polyethylene bags to prevent water loss and limit air exposure. After centrifuging, the displaced water w as acidified and stored in air-tight scintillation bottles. All water samples were transported in a controlled environment at < 4˚C from Botswana to Belgium for analyses.

Sample analysis
All water samples were analysed colorimetrically for dissolved organic carbon (DOC), chloride (Cl), ammonium (NH 4 ), nitrate (NO 3 ), and nitrite (NO 2 ), and soluble reactive phosphorus (PO 4 ) using a SKALAR SA 5100 Segmented Flow Analyzer at the University of Antwerp, Belgium. For DOC determination, water samples were acidified with H 2 SO 4 and flushed with pure nitrogen gas to remove background CO 2 concentration before analysis with the SKALAR (phenolphthalein, detection at 550 nm).

Statistical analyses
All variables were subjected to normality tests using histograms, the Shapiro-Wilk normality test, and quantile-quantile (Q-Q) plots in the freeware package R version 2.15.0 (R Development Core Team, 2012), before analysis. The Levene test was employed to test the data for the assumption of homogeneity of variance. All of this was done to ensure that appropriate tests were used. Analysis of variance (ANOVA), at p ≤ 0.05, was used in the present study to test for significant differences of solute concentrations between regions and between channels and floodplains. The Tukey HSD post-hoc test was further used to determine the source of the significant differences by comparing pairs of regions and pairs along local gradients.

Surface water physicochemical parameters
A summary of in-situ physicochemical measurements for surface waters across the Delta (from the inlet to outlets) is given in Table 2. Generally, the surface waters in the sampled floodplains and channels in the Delta were slightly acidic to neutral (pH 6.2 to 6.8). The pH levels in floodplains and channels were generally similar at each study site (varying by a maximum pH of 0.2). In contrast, in the outlets the pH varied from around neutral to slightly alkaline. EC in the floodplains was lower at Guma than at Xakanaxa and Nxaraga (Table 2). EC in the channels increased from Mohembo (inlet) (45 ± 15 µS·cm -1 ) to the outlets (112 ± 36 µS·cm -1 ) ( Table 2).
Concentrations of dissolved inorganic nitrogen (NO 3 , NO 2 and NH 4 ), orthophosphate (PO 4 ), sulphate (SO 4 ) and chloride (Cl) in surface waters across the Delta were all below the detection limits (0.08 mg·L -1 for NH 4 , 0.02 mg·L -1 for PO 4 , 0.04 mg·L -1 for SO 4 , and 0.03 mg·L -1 for Cl) for the analytical method used. Soil water SO 4 was also undetectable. All solute concentrations in surface waters were below and/or within the threshold guidelines for natural riverine surface waters (Carr and Neary 2006) (Table 3).

Local gradients
Concentrations of chemical solutes (Na, Mn, Fe, Al, and DOC) in surface water showed pair-wise significant differences, at p ≤ 0.05, between channel and floodplain of the same region (i.e., along the local gradients), with higher concentrations in the floodplains than channels (Fig. 3). Na showed a significant channel-floodplain difference only in Guma (Fig. 3), with mean concentrations of 1.86 ± 0.41 and 2.36 ± 0.33 mg·L -1 in Guma channel and Guma floodplains, respectively. Fe and Al showed significant channel-floodplain lateral heterogeneity in surface water at Nxaraga only (Fig. 3). Fe concentrations in the floodplains ranged between 7.40 and 10.70 µg·L -1 while in the channel the range was between 2.40 and 9.50 µg·L -1 . The variation of Al concentrations in the floodplains was six-fold, whereas in the channels concentrations of Al ranged between 0.20 and 0.70 µg·L -1 . Significant differences between channel and floodplain at Xakanaxa were evident for Mn, at 3.62 ± 1.72 and 25.95 ± 15.32 µg·L -1 , respectively, and DOC, at 3.88 ± 0.56 and 6.02 ± 2.84 mg·L -1 , respectively. From Fig. 3, it can be seen that Xakanaxa and Nxaraga regions, in most cases, had higher concentrations of dissolved solutes than the Guma region. Fe and Mn were the only two chemical solutes that showed higher concentrations in the Guma region than in the Xakanaxa and Nxaraga regions.

System-scale longitudinal gradient
In Fig. 4, box plots are shown only for concentrations of solutes in surface water that exhibit significant differences, at p ≤ 0.05, between the selected regions and outlets. Ca, K, Mg and Na exhibited similar patterns throughout the Delta for surface waters. Their concentrations increased from the Panhandle to the outlets, with approximately a threefold increase in mean concentrations. The concentration of Fe generally decreased from the Panhandle to the downstream reaches, with a twofold increase at the outlets. Al followed the same pattern but concentrations at the outlets were an order of magnitude higher than elsewhere in the Delta. Mn decreased strongly from the Panhandle to Nxaraga, with almost undetectable concentrations in the outlets at 1.25 ± 0.40 µg·L -1 . DOC showed an increase in concentration with distance along the axis of the Delta, with a significant fivefold increase in the outlets. DSi concentrations decreased from the upper Panhandle to Guma, and then gradually increased from Guma to the downstream reaches, with highest concentrations at the outlets. The relationship between conductivity (μS·cm -1 ) and dissolved cation concentration (Ca, Na Mg, and K, mg L -1 ) show that Ca and Na have steeper slopes (0.08 as opposes to 0.02 and 0.04, respectively) (Fig. 5). The higher slopes for Ca and Na indicate that they remain in solution to a greater extent than Mg and K over the entire conductivity range (Barnes et al., 2002).  Soil water chemistry Figure 6 presents chemical solutes in water extracted from sediment samples from floodplains at Guma, Nxaraga and Xakanaxa study sites. The extraction of any usable volumes of soil pore water was impossible from the mainly dry island soil samples. In general, Guma had significantly lower solute concentrations than Nxaraga and Xakanaxa. Pairwise comparison of NOx, NH 4 , Cl, K, Mg, Na, and Zn only showed significant differences, at p ≤ 0.05, between Guma and Nxaraga, and Guma and Xakanaxa. On the other hand, the concentrations of Mn differed significantly between all the regions but were highest in Xakanaka. The concentrations of PO 4 and Ca only differed significantly between Guma and Nxaraga. Results for NOx, NH 4 + and PO 4 revealed that Guma had the lowest concentrations for all of the three parameters for soil water measurements. The highest concentration for NH 4 was registered at Xakanaxa, and increased on average tenfold from Guma to downstream reaches. Nxaraga registered the highest concentration of PO 4. Concentrations of DSi did not vary significantly between the study sites but appeared to increase from Guma to the downstream reaches of the Delta.

Local gradients
The results of this study indicate that the concentrations of solute in surface water were noticeably higher in floodplain surface waters than in channel waters. Similar findings have previously been reported by Cronberg et al. (1996b), who attributed the higher solute concentrations in floodplain surface waters to re-dissolution and mobilisation of these ions previously concentrated in the floodplain soils during flood recession and sediment desiccation. It can thus be suggested that floodplains constitute short-term storages of solutes. Significantly higher DOC in the Xakanaxa floodplain surface waters than in the channel suggest that the Xakanaxa floodplain may be a source of plant-derived DOC from floodplain soils and plant litter (Cronberg et al., 1996a;Mladenov et al., 2005), which was remobilised with the arrival of the annual flood. This is consistent with the results of Mladenov et al., (2005) and Gondwe and Masamba (2014), who found greater DOC concentrations in floodplains of the seasonal swamps than in nearby channels. Despite the observed higher concentrations in the floodplain waters than channel waters, most solute concentrations were not significantly different along the three channel-floodplain gradients studied. This suggests that lateral exchange is more important than vertical exchange, resulting in fast mixing of surface waters between channels and floodplains. This is consistent with the prevalent mode of surface watergroundwater interaction in the Okavango, where unidirectional infiltration of floodplain water into soil is present throughout the flooding season . This implies that mobilisation of chemicals from floodplains takes place through dissolution and diffusion, and perhaps by advective transport induced by micro-gradients within superficial soil layers rather than by advection from deeper layers of the soil.

System-scale longitudinal gradient
Concentrations of solutes in surface water would be expected to increase identically along the axis of the Delta as a result of the process of evapotranspiration, which dominates the water balance of this wetland. However, concentrations of various ions increase differently along the axis of the Delta. This implies that there must be some differentiation in the processes affecting individual ions.  The concentrations of major cations (Ca, K, Mg and Na) increased generally throughout the Delta for surface waters, similar to the trends reported by Sawula and Martins (1991) and Mackay et al. (2011). This correlates well with a general increase in conductivity along the Delta distributaries. Although the EC of the water increases progressively, the net proportion of each ion species contribution to EC changes along the gradient. In groundwater, these cations are usually associated with the reactions of the carbonate systems and also form compounds with silica (McCarthy and Ellery 1994). Here, evapoconcentration initially results in saturation with respect to calcium carbonate (calcite), while magnesium and silica are removed at a later stage of concentration by precipitation of sepiolite (Mg 2 Si 3 O 8 ·H 2 O) . Some of these cations, for example Ca, are essential for the growth and development of prokaryotes, fungi, and plants, and these organisms may sequester them in biomass. The order of solutes in concentration was Ca > Na > K > Mg, consistent with the findings of Sawula and Martins (1991) and Masamba and Muzila (2005). This was, however, contradictory to results of Mackay et al., (2011) who found a greater concentration of potassium than sodium. There has been little or no systematic change in water chemistry (major ions) over the years, with calcium the dominant cation in the Delta (Sawula and Martins, 1991;Mackay et al., 2011).
Surface water concentrations of DSi and K decreased between the Panhandle and Guma. These declines are likely due to the assimilation of solutes by emergent vegetation on the floodplains of the Panhandle and permanent swamps, and are an important process in the retention of nutrients . The vegetation in the perennial swamps (Panhandle and Guma) is dominated by extensive stands of the giant sedge Cyperus papyrus, the reed Phragmites australis and the grass Miscanthus junceus (Gumbricht et al., 2004b). Previous studies on Si cycling in aquatic vegetation worldwide have noted the importance of emergent species (amongst which are sedges and grasses) and submerged species as a sink for Si accumulating up to 27 mg·g -1 . Struyf et al., (2015)  shown that floodplain vegetation in Guma contained between 20 and 80 g·m −2 of Si in the biomass.
The concentrations of some surface water solutes (Fe, Al, Zn, Cu, and Mn) decreased from the Panhandle to downstream reaches of the Delta. These solutes (Fe, Zn and Mn) are all important nutrients for plants (especially Fe) and are generally redox sensitive, and hence are usually present only in small quantities in the waters of the Delta. They are thus at a premium and will be scavenged by emergent macrophytes from passing water. Furthermore, there may be additional removal of these ions by precipitation as insoluble metal oxide and/or by sorption onto organic matter or clays. Most of the stability fields of these solutes are highly sensitive to changes in pH (Reddy and Delaune 2008). In this case, however, there was insignificant variation in pH in the Delta surface waters between the regions. In the outlets, the slightly alkaline surface water pH might be expected to favour a degree of immobilisation of Fe, Zn, Mn, and Al, whose solubility increases with decreasing pH (Osman, 2014). Similar patterns were observed for Fe and Mn by Masamba and Muzila (2005) in the Delta. In contrast to our findings, they also found increasing Zn concentrations with increasing distance from the Panhandle, suggesting that dissolution effects were more dominant than deposition effects. The concentration peak observed in the outlets for Fe and Al suggests local introduction of these ions, possibly from natural origins, such as oxidised organic matter.
Solute distribution patterns in pore water were similar to those in surface water, but the concentrations were generally higher in the pore water. However, several biogeochemical processes affect the various solutes differently. For instance, the detection of nutrients (NH 4 , NO x , and PO 4 ) in pore waters, despite being below detectable levels in surface waters for the analytic method used, may be a result of processes of remobilisation and remineralisation of accumulated biological material in the near-surface sediments (Cohen 2003). In addition, higher concentrations of major cations (Na, Ca, Mg, and K) were observed in pore waters than in the overlying surface water, which may reflect the effects of evapoconcentration of these solutes, given that evapotranspiration is the major water loss pathway in the Okavango Delta (e.g. Humphries et al., 2011). During this study, floodplains were inundated and saturated with floodwaters, suggesting that the soils were likely reduced. This may consequently affect Fe and Mn concentrations in pore water; for example, under reducing conditions DOC has an important role in Fe dissolution and mobilisation (Mladenov et al., 2005). It should be emphasised that the lack of data on sediment or pore water physicochemical parameters, for example, pH and redox conditions, limited the determination of the most likely processes that affect pore water solute concentrations. Therefore, future studies on pore water biogeochemistry should incorporate sediment and pore water physiochemical parameters.

Processing of solutes in the Delta
Previous studies have shown that solutes are chemically precipitated or sequestered beneath the islands by a process of unidirectional mass transfer from the floodplains to the abundant islands that are present in the Delta vegetation McCarthy et al., 1993;McCarthy and Ellery, 1995;Bauer, 2004;Bauer et al., 2007;Ramberg and Wolski, 2008), a process driven partially by evapotranspiration by riparian woodland. This process is the mechanism responsible for the counter-intuitive small difference between influent and effluent concentrations of total dissolved solids.
In this study, however, the changes observed in the composition of the total solute load show that evapoconcentration in both floodplains and islands is not the sole process affecting surface water concentrations. For instance, while the surface water conductivity between the Panhandle and Guma increased by 83%, the concentration of K decreased by 77%. Similar patterns were observed for DSi. Given the low nutrient status of inflowing water (Cronberg et al., 1996a;West et al., 2015), and the extensive and dense stands of Cyperus papyrus and Phragmites australis vegetation in the Panhandle and permanent swamps Ellery et al., 2003), which are the first to make contact with the inflowing water, the question arises: could emergent vegetation in these areas preferentially remove K, and possibly other solutes, from solution? This study suggests that major nutrients (such as P, N, K, Fe, and DSi) brought in solution by surface water may be scavenged by macrophytes in the Panhandle and the permanent swamps, and cycled multiple times before being transported downstream to seasonal floodplains. This is supported by the results of Struyf et al. (2015), who showed that DSi trapping by floodplain vegetation results in a large temporary store of biogenic silica. Downstream of Guma (as the hydroperiod becomes more seasonal), the concentrations of all major cations, DSi, and DOC increased proportional to increasing conductivity, suggesting that processes operating on solutes were dominated by evapoconcentration and bulk sequestration in island soils and groundwater. Our study thus indicates that solute processing in the Panhandle and permanent swamp may be more biologically driven, while that in the seasonal eco-regions is more chemically driven. Trends of solutes in pore waters were similar to those of solutes in surface water but the concentrations were generally higher in the former, indicating that evapoconcentration is already taking place in the floodplains.

CONCLUSIONS
This study indicates that differential uptake of certain solutes needs to be added to the list of processes controlling the ionic concentration and composition of Delta water. The most likely driver for this differential uptake is the nutritional requirements of the extensive and highly productive stands of emergent macrophytes in the perennially flooded part of the Delta. In the more seasonally flooded lower and distal reaches, solute concentrations followed trajectories more typical of the overall solute load, which appear to be simply driven by the physical processes of evapotranspirative concentration, infiltration and sequestration in deep groundwater.
Further work should focus on the dynamics of relationships between chemical constituents of macrophytes of the perennially flooded parts of the Delta, and the composition of the water. This will elucidate the effect of biological processing, and also improve our understanding of potential responses of this vegetation to externally driven changes in nutrient load of influent water.