Invasive floating macrophytes reduce greenhouse gas emissions from a small tropical lake

Floating macrophytes, including water hyacinth (Eichhornia crassipes), are dominant invasive organisms in tropical aquatic systems, and they may play an important role in modifying the gas exchange between water and the atmosphere. However, these systems are underrepresented in global datasets of greenhouse gas (GHG) emissions. This study investigated the carbon (C) turnover and GHG emissions from a small (0.6 km2) water-harvesting lake in South India and analysed the effect of floating macrophytes on these emissions. We measured carbon dioxide (CO2) and methane (CH4) emissions with gas chambers in the field as well as water C mineralization rates and physicochemical variables in both the open water and in water within stands of water hyacinths. The CO2 and CH4 emissions from areas covered by water hyacinths were reduced by 57% compared with that of open water. However, the C mineralization rates were not significantly different in the water between the two areas. We conclude that the increased invasion of water hyacinths and other floating macrophytes has the potential to change GHG emissions, a process that might be relevant in regional C budgets.

water hyacinth is known to change the physicochemical characteristics of water (e.g., the pH, alkalinity, dissolved oxygen (O 2 ) concentration and dissolved carbon dioxide (CO 2 ) concentration) 12 . For example, O 2 in the water can be diminished by emergent macrophytes, which limit pelagic and benthic photosynthesis through shading 13 and prohibit gas exchange and thus re-oxygenation from the atmosphere. Compared with the O 2 produced by submersed aquatic plants and phytoplankton, O 2 that is photosynthetically produced by emerged macrophytes is directly emitted into the atmosphere and does not contribute to aquatic O 2 concentrations 14 . C turnover underneath the water hyacinths can be further fuelled by root respiration and microbial activity in the water and sediments because of dissolved organic matter from root exudates and decaying plant litter [15][16][17] , which eventually increase CO 2 and CH 4 concentrations below these floating plants. Therefore, water hyacinths have considerable ecological impacts, which may confer unwanted economic effects 18 . However, most studies of water hyacinths have examined their effects on water quality and their dispersal spread or phytoremediation 19 (and references therein). The link between the invasion of water hyacinths and the emission of climate-relevant gases (CO 2 and CH 4 ) has not yet been explored.
Most freshwater systems are net greenhouse gas (GHG) emitters 20,21 . CO 2 and CH 4 are among the major gases impacting the atmospheric heat budget and contributing to global climate change. Consequently, investigations of GHG emissions and their influencing factors are of major importance for understanding current and predicting future climate conditions. Most GHG research in inland waters has been performed at temperate and boreal latitudes, whereas data from subtropical and tropical inland waters remain scarce 22 . However, the first upscaling approaches have ranked tropical and subtropical systems as major sources of GHG emissions 23,24 . In a comparative study of India's major inland water types, freshwater bodies were shown to emit large amounts of CO 2 and CH 4 into the atmosphere that corresponded to 42% of India's estimated land C sink 25 .
The aim of this study was to analyse and understand the impact of water hyacinths on water-column organic C mineralization and GHG (CO 2 and CH 4 ) emissions from a small, tropical water-harvesting lake in South India. We hypothesized that areas covered by water hyacinths will (1) have higher C mineralization rates and therefore lower O 2 and higher CH 4 and CO 2 concentrations but (2) decreased diffusive CO 2 and CH 4 emissions because of the lower gas exchange within the plants compared to that in open water.

Results
General lake characteristics. The water temperature of the lake did not drop below 20 °C during three field campaigns in February and November 2012 and March/April 2014 (Table 1), and the lake was slightly alkaline, with a pH between 7.6 and 8.6. In March/April 2014 the mean TIC and TOC concentrations were 7.92 ± 3.24 mmol L −1 and 13.5 ± 0.8 mg L −1 , respectively ( Table 1).
The population of water hyacinths on Lake Thimmapuram covered 12 to 55% of the surface area of the lake, with the maximum coverage of 55% reached in April 2014 ( Figs. 1 and 2). The dispersal of water hyacinths is strongly managed because the plants are harvested and used as fodder for cattle. Additionally, fishermen occasionally remove the majority of the plants to improve fishing efficiency. In addition to the direct measurements during our field campaign, we derived the coverage of water hyacinth for 2000-2003 and 2013-2014 from remote sensing data (Landsat 7 and 8, Fig. 2). The enhanced vegetation index (EVI) was used for the classification, which introduced a degree of uncertainty, particularly because mixed surface water and water hyacinth pixels are difficult to separate from dried lake bottom pixels (see details in Fig. 2). Water hyacinths could be detected in all available Landsat images, except on April, 4 th 2003, when a high percentage of uncertainty was encountered. However, from April, 7 th -10 th 2014 (Figs. 1 and 2), the two methods of water hyacinth coverage estimation, on-site GPS recording and remote sensing, produced similar results.  (Fig. 3d). The CO 2 fluxes from the lake areas with water hyacinths were less variable and ranged from 3.9 to 7.6 mmol m −2 h −1 , with a mean of 4.7 ± 1.2 mmol m −2 h −1 . The diffusive CH 4 fluxes were generally lower than the CO 2 fluxes and ranged from 2.3 to 190.7 μ mol m −2 h −1 in open water and from 6.5 to 71.3 μ mol m −2 h −1 between the hyacinths (Fig. 3e). The diffusive CO 2 and CH 4 emissions were significantly higher in open waters than in areas covered by water hyacinths (Table 2, p < 0.05). The distribution of CH 4 ebullition fluxes, however, was not significantly different between the two areas based on the Mann-Whitney U test (range from 0-6,813 μ mol m −2 h −1 ), although the total C emissions (CO 2 + CH 4 ) from areas covered by water hyacinths were 57% lower than that in open water (Fig. 4).
C mineralization rates in the water column ranged from 102.7 to 526.2 μ g C L −1 d −1 in open water and 138.4 to 599.1 μ g C L −1 d −1 under the water hyacinths, and the Mann-Whitney U test did not detect significant (p < 0.05) difference between the distributions of data (data not shown). The O 2 concentrations at the start of the water incubations from the vegetated areas were lower and reflected the conditions observed directly in the field (158 ± 57 μ mol L −1 in the water incubations from vegetated areas; 207 ± 29 μ mol L −1 in the water incubations from open water). Anoxic conditions were not observed in any of the water incubations, and such conditions would have diminished the mineralization rates.

Discussion
In Lake Thimmapuram, 0.48 to 1.03 million plants per hectare were counted, and their dry weight totalled 16.6 to 35.5 metric tons of dry weight per hectare. The abundance of water hyacinths in Lake Thimmapuram varied strongly between years (from 12 to 55%), although the lake was never completely covered (Figs. 1 and 2), which is presumably because of management by the local villagers and fishermen who depend on the lake for survival. Hyacinth mats can also disperse when there is enough wind, and such a dispersal has also been observed in strongly managed water bodies in the northern part of Bangalore City 9 .
The observed concentrations of O 2 and CH 4 in the surface waters of the areas covered by water hyacinths were significantly lower (22% and 26% lower, respectively) than the concentrations in the open water areas, whereas differences were not observed in the concentration of CO 2 (Figs 3a-c,4). Reduced O 2 concentrations and even anoxic conditions have also been observed in vegetated areas in other systems covered by water hyacinths 26 and other floating species 5,15 . As we did not measure differences in C mineralization in the water column itself, the reduced O 2 conditions could be attributed to higher respiration rates at the roots of the plants or in the sediment under water hyacinths. However, in Lake Thimmapuram, the O 2 content during the day below the vegetated areas was not completely depleted during our sampling campaign, which prevented anaerobic metabolism in the water column and thus affected the C turnover rates, the CO 2 and CH 4 concentrations. We did not measure the O 2 concentrations during the night when the potential for anoxia increases because of an absence of primary production caused by light limitations 27 . However, this potential remains speculative. In addition, CO 2 and CH 4 concentrations may also be higher at night.  Surprisingly, the CH 4 concentrations were lower in the areas covered by water hyacinths, although similar or even higher concentrations might be expected because of the lower O 2 concentrations and higher organic C content in the sediments. These conditions fuel methanogenesis, as observed in other studies of floating plants 6,17,28 . The lower surface CH 4 concentrations beneath the vegetation could be caused by CH 4 oxidizers living on the roots of the water hyacinths 29,30 . For example, Brix and co-authors 31 found that up to 76% of the CH 4 produced in the sediment was re-oxidized within the rhizosphere of Phragmites australis, which might explain the simultaneously lower concentrations of O 2 and CH 4 beneath the water hyacinth because O 2 is required for the aerobic oxidation of CH 4 . CO 2 is an end product of both aerobic and anaerobic respiration 32 . In Lake Thimmapuram, CO 2 concentrations were not significantly different between the water hyacinth and open water areas, suggesting that the metabolic rates were comparable. This assumption is supported by the similar aquatic C mineralization rates in both areas. A comparison between an area covered by yellow water lilies (Nuphar lutea) and an adjacent plant-free zone did not indicate significant differences in the water chemistry 28 , which is consistent with our results for CO 2 . However, the mean CO 2 concentrations tended to be slightly higher in the surface waters covered by water hyacinths in our study (Fig. 3b). In the central Amazon River and its floodplains, it has been shown that pCO 2 increased consistently from open water areas towards emergent plants including floating macrophytes 33 which is consistent with our results. The authors mainly attribute the increases in CO 2 to an increased supply with organic C from the litter fall and root exudation as well as a release of plant-respired CO 2 from the roots. However, this is uncoupled from O 2 consumption in the water column because O 2 is supplied from the atmosphere. This might explain why we observed different patterns in CO 2 and O 2 concentrations.
Furthermore, the differences in CO 2 concentrations may have been masked by the generally higher CO 2 concentrations compared with the O 2 and CH 4 concentrations (CO 2 concentrations were 2 and 200-300 times higher than the O 2 and CH 4 concentrations, respectively) and a high spatial heterogeneity. Nevertheless, the relatively small but significant differences (O 2 and CH 4 ) or lack (CO 2 ) of differences in the concentrations of O 2 , CH 4 and CO 2 between the water hyacinth-covered areas and open water might have resulted from the drift dynamics of the water hyacinth mats caused by changing wind directions during the day (personal observation) or lateral mixing of the water body driven by different heating and cooling and densities over the day and night cycle 34 . The drifting was also described by Abdel-Tawwab 35 , who only found a significant decrease in nutrient and O 2 concentrations and phytoplankton biomass in artificial fish ponds if the free-floating plant (Azolla pinnata) cover was greater than 50%, which hinders plant drift.
Generally, the water was supersaturated with CO 2 and CH 4 relative to the atmosphere, which led to a net emission of both gases across the air-water interface. By comparing the open water areas and the water hyacinth-covered areas, we found a significant reduction in diffusive C emissions between the covered areas and the open water ( Fig. 3d-f,4). CO 2 emissions could be further diminished in water hyacinth-covered areas because of CO 2 fixation through photosynthesis 5,6,28 . However, photosynthetic C fixation by water hyacinths was not quantified in this study. In boreal studies, vegetated littoral areas in aquatic systems have been shown to have the highest areal CH 4 emissions 36 , which are mostly generated through aerenchymal transport from the emergent  macrophytes rooting in sediments that connects the sediment directly to the atmosphere 36 . This mechanism was not relevant for the floating water hyacinths in Lake Thimmapuram, indicating that they must play a different role in the release of GHGs from aquatic systems. According to our hypothesis, the diffusive emissions of both CO 2 and CH 4 were reduced in the areas with water hyacinths. Differences in the surface water concentrations of CO 2 as a driver of diffusive fluxes can be excluded because differences were not observed in the CO 2 concentration between the open water and hyacinth-covered areas. Nevertheless, the emitted gases can be trapped inside the plant canopy, which results in

b), and CH4 (c) concentrations as well as CO2 (d) and CH4 fluxes as diffusion (e) and ebullition (f) in the open water (OW) and water hyacinth (WH) areas.
Boxplots indicate the medians, the 25th and 75th percentiles (boxes), the 5th and 95th percentiles (whiskers) and the mean values (black squares). Significant differences are denoted with asterisks. a decreased concentration gradient and thus a reduced diffusion. Furthermore, the gas transfer velocity between water and the atmosphere is positively related to the turbulence in the upper water column 37-39 and the concentration gradient between the media. Water hyacinths reduce the wind speed at the water surface by greatly increasing the roughness length (zone above the surface where the wind speed equals 0 m s −1 ) 40 . Thus, both the concentration gradient between the water and the atmosphere as well as the turbulence of the surface waters were reduced, leading to a reduced exchange of CO 2 and CH 4 across the air-water interface among water hyacinths. Similar mechanisms might be expected in other floating-leaved macrophyte communities, such as Lemna spp. or Trapa natans, which are often found in eutrophic lakes worldwide 4 .
In related studies comparing gas emissions from open water and macrophyte covered areas contradictory results were found. In a study in the Pantanal region, a higher emission of CH 4 from water hyacinth mats were detected 41 but other authors 42 found no differences in the Amazon floodplain between open waters, floating emergent macrophytes, and flooded forests. However, these authors did not determine CO 2 emissions and the gas fluxes were mainly dominated by CH 4 ebullition which we do not discuss further here. These different results highlight the demand for further studies to elucidate the role of floating macrophytes for GHG emissions.
We upscaled the CO 2 and CH 4 emissions from hourly to daily rates (multiplied by 24) to better compare them to other studies. Our CO 2 emissions with a mean CO 2 diffusion rate of 323.  25 . Those differences can be directly attributed to physical characteristics, such as turbulence, or indirectly to biogeochemical processes that are influenced by temperature as well as O 2 , C and nutrient concentrations 44,45 . Our results highlight the substantial GHG efflux potential of the analyzed lake type (manmade tanks and ponds), which belongs to the major of Tamil Nadu 46 .
Overall, the concentrations of O 2 and CH 4 as well as the C emissions from the areas covered by water hyacinths were reduced compared with that of open water. However, the CO 2 concentrations and water C mineralization rates were not significantly different between the two areas (Fig. 4). These results reveal that invasive water hyacinths can play an important role in biogeochemical processes as well as in the release of climate-relevant gases into the atmosphere. Floating macrophytes, especially invasive species, might therefore be considered as important regulators of gas exchange at the air-water interface, a process that might be central in regional C budgets.

Methods
Field campaign and study site description. The water body investigated in this study is Lake Thimmapuram (12.45°N, 78.22°E), which is located in South India (Tamil Nadu State) near the town of Krishnagiri (Fig. 1). The climate is typical of wet and dry tropical regions, with pronounced precipitation seasonality and minor temperature seasonality. The long-term mean annual precipitation in Krishnagiri is approximately 780 mm (measured at the nearby Krishnagiri Dam), and a primary rainy season occurs that is related to the southwest and northeast monsoons between August and November. The mean annual air temperature is 26.4 °C 47 . The lake is eutrophic and shallow (mean depth 1.5 m in March/April 2014) and serves as an irrigation reservoir for the surrounding arable land. The water level in the lake depends on the natural inflow during the monsoon season and the management of a cascade of upstream water-harvesting structures. Following the end of the rainy season, the lake receives additional inflow from December to approximately April via the Krishnagiri Dam (personal communication with dam management). Additional details on the study site can be found in Fiener and co-authors 48 .
The initial sampling campaigns were conducted in 2012 (Table 1), and an intensive sampling campaign was performed in March/April 2014, during which the rates of water-column organic C mineralization and GHG emissions were measured along with the physicochemical water variables (temperature, pH, O 2 , conductivity, total organic carbon (TOC), total inorganic carbon (TIC) and ammonium).
Multi-temporal observations of the lake's water hyacinth cover were performed by classifying 18 Landsat 7 and 8 scenes (Google Earth Engine) based on the enhanced vegetation index (EVI; 49 ) (Fig. 2). Instead of using the more common normalized differenced vegetation index (NDVI; 50 ), the EVI was used because of its reduced susceptibility to atmospheric influences and improved sensitivity in high biomass environments 51,52 . A simple but robust threshold approach was applied to the EVI product: surface water was classified by an EVI threshold <0.1, and water hyacinths were classified by an EVI > 0.3. EVI values between 0.1 and 0.3 were declared to be uncertain because separating a pixel containing both surface water and water hyacinths from the dried lake bottom was impossible. Water hyacinth coverage was also recorded by a Global Positioning System (GPS) from a boat, and the biomass inside a frame (1460 cm 2 ; total of six replicated samples on 9 and 11 April 2014) positioned on the water hyacinth meadows was sampled by hand. The plants were washed in situ, separated into emerged and submersed leaves plus roots, and desiccated at 70 °C until they reached a constant weight.
Physicochemical variables. O 2 , pH, conductivity (corrected to 25 °C), and temperature were measured with a YSI probe (YSI Inc., Yellow Springs, OH, USA). Gas samples for the analysis of dissolved CO 2 and CH 4 were obtained using the headspace extraction technique 53 . Water samples (20 mL) were collected from the surface waters (~10 cm depth) in glass vials equipped with septa, and the vials were immediately closed and kept gastight without a headspace. Subsequently, a 5 mL headspace was created with ambient air, the vials were vigorously shaken for 60 seconds, and 500 μ L gas samples were then collected from the headspace with a gastight syringe and manually injected into a closed loop between the gas inlet and the outlet of a Los Gatos GHG analyzer (Los Gatos Research Inc., Mountain View, CA, USA) to measure the CO 2 and CH 4 contents 54 . This method, first described by Baird and co-authors 55 , allows a fast on-site determination of CO 2 and CH 4 gas samples. The volume of the loop was 72.6 ± 2.2 mL and precision of measurements amounted to 3-5%. The samples used for the analysis of total inorganic carbon (TIC) were prepared following the same procedure as the CO 2 samples but with the addition of phosphoric acid (pH < 4) before shaking to outgas the inorganic carbonate species as CO 2 . The partial pressures of the gases were converted into concentrations in water (expressed as μ mol L −1 ) by using Henry's constant, the water temperature, and the measured gas partial pressures in the air (while accounting for the water volume and the headspace inside the bottle) 56 . Overall, 139 surface samples were collected at random locations across the lake over 13 days during the March/April 2014 sampling campaign and used to measure the concentrations of CO 2 and CH 4 . The analysis of the TOC from the surface waters (~10 cm depth) was performed using a TOC analyzer (Shimadzu Co., Kyoto, Japan) according to method 5310 57 .
Greenhouse gas emissions. The GHG flux (CO 2 and CH 4 ) across the water-atmosphere interface was measured with floating chambers that were gently deployed from a boat onto the water surface between water hyacinths and in open water areas to minimize artificial turbulence. Similar to the protocol described in McGinnis and co-authors 39 , the chambers were constructed of inverted non-transparent plastic buckets with a volume of 14.76 L and an area of 1,018 cm 2 . Some light could have penetrated through the plastic, however, this should not have changed the GHG emissions on these short timescales (20 min). A floating device composed of polyethylene was attached to the chambers, and approximately 2 cm of the chamber walls was allowed to submerge to ensure a gastight seal between the water surface and the chamber while minimizing the impact of the natural turbulence in the water column beneath the chamber 58 . Two gas ports (inlet and outlet) were fitted on top of each chamber and connected with 2 × 5 m-long gastight tubes (Tygon 2375) to a Los Gatos ultraportable GHG analyser. The internal pump circulated the air in the gas chamber through the GHG analyser at a rate of ~450 mL min −1 . The boat and the chambers were allowed to drift freely on the lake surface for 10-20 min per deployment, and the concentrations of CO 2 and CH 4 were measured every second, which allowed the changes in CO 2 /CH 4 to be tracked in situ. The concentrations of CH 4 and CO 2 inside the atmosphere of the chamber increased linearly over time under diffusional conditions, whereas the CH 4 concentrations increased abruptly when bubbling occurred. This process allowed us to separate the bubbling and the strict diffusional flux by the high sampling frequency enabled by the GHG analyser 59 . However, the short incubation time did not allow an accurate determination of CH 4 ebullition and is thus not further emphasized in the discussion. The water-atmosphere fluxes (J) of CO 2 and CH 4 (mmol m −2 h −1 and μ mol m −2 h −1 , respectively) were calculated from the slopes (s) of the linear regressions of the concentrations in the chamber versus time as follows: where V ch is the chamber volume, and A ch is the chamber area. The amount of gas released per bubbling event was determined by calculating a two-point regression from the concentrations in the chamber at the start of the bubbling event and after the bubbling event, when the CH 4 concentration in the chamber was well-mixed 59 . Fluxes were only measured during the day because local circumstances did not allow for night measurements. In total, 41 chamber measurements were performed at different locations on eight different days during the three-week sampling campaign.
Carbon mineralization. Water column C mineralization was determined using transparent acryl-glass incubation cores (length of 30 cm and inner diameter of 5.4 cm) that contained a septum in the tube wall for in situ O 2 measurements. The incubation containers were carefully filled with water collected at the water surface in the vegetated and open areas. We avoided collecting any plant remnants during the filling in the vegetated areas, which would have increased our mineralization rates. After applying an airtight seal to the containers, respiration was quantified for the water samples by O 2 depletion over 24 hours. The incubation cores were incubated at in situ temperatures in the dark. O 2 depletion was measured with a needle-type O 2 microsensor (Optode, PreSens, Regensburg, Germany) after the water column was mixed, and the amount of consumed O 2 was converted to μ g C L −1 d −1 using a conversion factor of one 60 . A more detailed description is given in Attermeyer and co-authors 61 .
Statistics. Because normal distributions were not observed for all of the parameters, we tested for differences in the chemical variables, CO 2 and CH 4 emissions, and water C mineralization under the water hyacinths and in open water using a non-parametric, two-sided Mann-Whitney U test 62 . To consider the temporal differences during the sampling periods, all of the values of each group (water hyacinths and open water) from different days of the sampling campaign in March/April 2014 were included. Differences in the distribution of the different groups were considered significant at p < 0.05. All of the values were expressed as the mean ± standard deviation, and all of the statistical analyses were performed with IBM SPSS Statistics 22 (IBM Corporation, Armonk, NY, USA).