The synergistic effect of elevated CO₂ and phosphorus on reservoir eutrophication

ABSTRACT

Winston B, Scott JT, Pollock E. 2016. The synergistic effect of elevated CO₂ and phosphorus on reservoir eutrophication. Lake Reserve Manage. 32:373–385.

Increased delivery of carbon (C) and phosphorus (P) as a result of climate change is predicted to significantly increase algal biomass and alter algal nutrient content. We conducted a 2-phased experiment to determine the following: (1) whether carbon dioxide partial pressure (PCO₂) in the water column of reservoirs dropped low enough to limit algal biomass, (2) whether increasing the PCO₂ in the water column could alleviate C limitation and increase algal biomass, and (3) whether the simultaneous increase in PCO₂ and P alter algal nutrient content. In the first phase we reconstructed PCO₂ in the water column of several northwest Arkansas reservoirs on timescales ranging from days to decades. During fall, winter, and spring, PCO₂ was greater than atmospheric levels, suggesting the reservoirs were sources of C released to the atmosphere. During summer, PCO₂ was less than atmospheric levels, suggesting that reservoir PCO₂ fell low enough to potentially limit algal biomass. In the second phase, we conducted an enrichment experiment along a PCO₂ gradient representative of preindustrial (250 µatm), current (400 µatm), and 2050 projected levels from the Intergovernmental Panel on Climate Change (550 µatm) along a P gradient. Our results showed that relative to preindustrial and current levels, elevated PCO₂ had no significant effect on algal biomass and nutrient content. Instead, biomass increased significantly with increasing levels of P, supporting predictions that even under elevated PCO₂ and global climate change scenarios, P might still be the major nutrient limiting algal growth in these reservoirs.

KEYWORDS:

• Carbon dioxide
• eutrophication
• global climate change
• lakes
• nutrients
• stoichiometry

Before the 1970s, carbon (C) was perceived to be the primary nutrient limiting algal biomass. Numerous studies from the 1970s through the 1980s showed consistent relationships between phosphorus (P) and algal biomass (Schindler 1977, Hecky and Kilham 1988, Jones and Knowlton 1993) and relegated C to a minor role. To date, most studies investigating algal limitation focus on P, nitrogen (N), or combinations of both (Schindler 1977, Dzialowski et al. 2005 Scott and McCarthy 2010) with little to no emphasis on C.

Despite the overwhelming evidence supporting N- and P-driven limitation, C limitation is still possible under certain scenarios. For example, in softwater lakes with low or intermediate dissolved inorganic C (DIC), photosynthetic activity decreased when carbon dioxide partial pressure (PCO₂) declined (Hein 1997). The study concluded that activity declined because the demand during photosynthesis exceeded the supply through diffusion and reaeration. Jansson et al. (2012) reported phytoplankton primary production in boreal lakes 10 times greater at elevated PCO₂ compared to ambient conditions. Under experimental conditions, Urabe et al. (2003) reported increased algal growth with elevated PCO₂ and low nutrient levels. These results suggested that C limitation might be more common than currently investigated.

If C is limiting algal biomass in northwest Arkansas (NWA) reservoirs, activities that directly or indirectly increase C may alleviate limitation and increase biomass in these reservoirs. Extreme precipitation events are predicted to occur more frequently globally (IPCC 2007, 2014), which can increase the supply of allochthonous C and provide excessive nutrients (eutrophication) to NWA reservoirs. Winston et al. (2014) showed nutrient concentration at least doubled during heavy rainfall periods in a NWA reservoir.

The combined effects of increased C and nutrients are projected to alter algal nutrient content and the structure of aquatic food webs (Van de Waal et al. 2010) because phytoplankton with a high C-nutrient content is poor-quality food for most zooplankton species (Sterner and Elser 2002). An aquatic community dominated by phytoplankton with a high C-nutrient content might favor less nutrient-demanding zooplankton species, which in turn will effectively alter the structure and function of the aquatic food web. From a water management perspective, the consequences of elevated C and increased nutrients could be significant. Elevated C could result in increased algal biomass (Urabe et al. 2003, Jansson et al. 2012), which in turn could lead to increased episodes of harmful algal blooms (Paerl and Huisman 2009), taste and odor episodes (Winston et al. 2014), and the formation of disinfection byproducts (DBPs; Mash et al. 2014). These negative effects would likely cause massive disruption of water supplies and recreational uses of reservoirs (Paerl and Huisman 2009, Wagner and Adrian 2009).

We conducted a 2-phased experiment to determine (1(1) ) whether PCO₂ in the water column of reservoirs dropped low enough to limit algal biomass, (2(2) ) whether increasing the PCO₂ in the water column could alleviate C limitation and increase biomass, and (3) whether the simultaneous increase in PCO₂ and P would alter algal nutrient content. Algal biomass was expressed as chlorophyll a (Chl-a) and particulate carbon (PC) and algal nutrient content as seston C:P, C:N, and N:P. These proxies were chosen to compare with previous studies, some of which showed increased algal biomass and altered algal nutrient content (Urabe et al. 2003, Verschoor et al. 2013), and some of which showed contrasting results (Burkhardt et al. 1999, Gervais and Riebesell 2001).

Site description

The research was conducted in 5 lakes in the NWA area. Beaver Lake, Lake Brittany, Lake Rayburn, and Lake Norwood were used to reconstruct PCO₂, and Lake Fayetteville was used to validate the model used in the reconstruction. All lakes were located in the Springfield Plateau, composed of karstified limestone and dolostone mantled by soil and insoluble regolith. The plateau is deeply dissected with well-developed dendritic drainage underlain by Paleozoic strata from the Ordovician through Mississippian periods (Croneis 1930, Haley et al. 1993).

Beaver Lake (94°12′N, 36°18″W) is a warm monomictic lake with a surface area of 113 km² and an average depth of 18 m. The reservoir was built between 1960 and 1966 on the upper portions of the White River for flood control, hydropower generation, and water supply. The watershed drainage area is 3087 km², with 57% forest, 32% agriculture, and 5% urban (Galloway and Green 2007). The trophic status of Beaver Lake is variable and ranges from eutrophic in the riverine portions to oligotrophic near the dam. Average annual total N (TN), total P (TP), Chl-a, and pH are 800 µg/L, 5 µg/L, 5 µg/L, and 8.53, respectively (Table 1).

Table 1. Select reservoir characteristics for Beaver Lake and Lakes Brittany, Norwood, Rayburn, and Fayetteville. Data from Ben Thompson's master's thesis and Bryant Baker's master's thesis, and from the Arkansas Department of Environmental Quality.

Reservoirs	Surface area km^2	TN (ug/L)	TP (ug/L)	Chl-a (ug/L)	pH
Beaver Lake	113	800	5	5	8.53
Lake Brittany	<0.2	400	15	6	7.88
Lake Norwood	<0.2	650	20	15	8.14
Lake Rayburn	<0.2	700	32	19	8.37
Lake Fayetteville	<0.2	690	92	38	8.24

Lakes Brittany (36°28′08″N, 94°12′04″W), Norwood (36°28′43.01″N, 94°14′48.62″W), and Rayburn (36°27′43″N, 94°14′21″W) are warm monomictic water bodies, similar in size (<1 km²), depth (24 m), watershed dimensions, and land-use characteristics. The reservoirs were built in the 1970s, predominantly for recreational purposes. Forests account for the majority of the land use (64–77%) followed by urban (14–25%), which is mostly single-family homes and roadways. Lake Brittany had the lowest average annual TN, TP, Chl-a, and pH of all the smaller lakes (Table 1), with annual averages of 400 µg/L, 15 µg/L, 6 µg/L, and 7.88, respectively.

Lake Fayetteville (36°13′56″N, 94°13′9″E) was built in 1949 as a reservoir for the city of Fayetteville, Arkansas, but now is used for recreational purposes. The reservoir is also warm and monomictic, with a surface area <1 km² (Scott and Grantz 2013) and elevated levels of TN, TP (ADEQ 2004), and seston Chl-a (Table 1).

Materials and methods

To assess C limitation of algal biomass, we reconstructed water-column PCO₂ over long-term, short-term, and daily scales for 4 lakes in NWA. Carbon limitation was defined as water-column PCO₂ less than atmospheric PCO₂, which averaged 397 µatm in 2013 measured at the Mauna Loa Observatory.

Long-term and short term PCO₂ reconstruction

All PCO₂ reconstructions are based on C chemistry equations from Stumm and Morgan (1996) and Henry's Law.

Stumm and Morgan (1996): (1) where [CO₂] is the concentration of carbon dioxide in moles per liter, [HCO₃⁻] is the concentration of bicarbonate in moles per liter, K_a is the acid dissociation constant, and pH is the negative log of the hydrogen ion concentration.

Henry's Law: (2) where K_H is Henry's constant (29.41 L atm/mol) at standard temperature and pressure. Our longest PCO₂ reconstruction was for Beaver Lake based on a historical dataset from 1986 to 2010. Long-term pH, temperature, and bicarbonate concentration data were obtained from Beaver Water District, the regional water management authority.

Short-term PCO₂ reconstructions were from data compiled over a 2-year period from in situ monitoring of physical water quality parameters at Lakes Brittany, Norwood, and Rayburn. A HACH datasonde recorded pH, temperature, and dissolved oxygen (DO) at 15-min intervals. Every 2 weeks, the probe was retrieved and data downloaded. The probe was also recalibrated and DO sensors were checked and replaced as necessary. There was no significant fouling of the sensors; therefore, DO corrections were not required.

Daily reconstructions were based on data retrieved from Beaver Lake on the 22 June, 22 July, and 22 August (JJA) 2011 and 2012. These months were chosen because they generally coincided with warm temperatures, stable water column, and increased algal biomass in the reservoir. To calculate daily fluctuations in PCO₂, we first developed a linear relationship between pH and DO as follows. First, pH, temperature, and DO were measured at 15-min intervals using a HACH datasonde for 1 week at a depth of 0.5 m from the surface. The probe was returned to the lab, data downloaded, and the linear relationship between pH and DO assessed. Second, we obtained 5-min interval DO and temperature data measured by the US Geological Survey Lake Diagnostic System (LDS, http://waterdata.usgs.gov/ar/nwis) for 2011 and 2012. Briefly, the LDS consists of a string of temperature and DO sensors at 0.5 m depths from the surface to the bottom of the lake. Daily DO from the 0.5 m sensor was then used to estimate daily pH from the linear relationship obtained during the week-long sampling. Daily fluctuations in PCO₂ were determined using equations 1(1) and 2(2) .

Model verification: diurnal PCO₂ and δ¹³C direct measurements

To verify the model used for PCO₂ calculations, we performed direct measurements of PCO₂ and δ¹³C diurnally during August 2012. Samples were collected from Lake Fayetteville in acid- washed and rinsed bottles with no headspace, crimped, and placed on ice. Lake Fayetteville was chosen to maximize chances of obtaining drastic swings in PCO₂ because of its eutrophic nature. Temperature, pH, and DO were also measured at sample collection using a Yellow Springs 600 XLM multiparameter datasonde. PCO₂ and δ¹³C were measured on a Picarro CO₂/CH₄ analyzer (G2201-I) at the University of Arkansas Stable Isotope Lab (UASIL). Instrument precision for CO₂ was 0.16‰. Isotopic standards ranged from −3.64 to −44.47‰, obtained from Oztech Trading Company, Dallas, Texas. The final C isotope values are expressed relative to V-PBD standard. The quantitative standards for CO₂ were from Scott Specialty Gases (Aire Liquide) with a balance of air. Additionally, we estimated PCO₂ from the measured pH during sample collection using the same method for the long-term data. Estimated PCO₂ was then compared with the instrumentally measured PCO₂ using regression to determine the reliability of the model.

Algal response to elevated PCO₂

To determine if elevated PCO₂ and P increased biomass and altered algal nutrient content, we conducted a semicontinuous enrichment experiment at PCO₂ representative of preindustrial, present-day, and 2050 projected levels from the Intergovernmental Panel on Climate Change across a P gradient. We used preindustrial PCO₂ (250 µatm) to represent the control and PCO₂ of 550 µatm to represent elevated PCO₂ conditions. This experiment was conducted for 13 days during summer 2012.

We harvested 70 L of water from the Beaver Lake transition zone, which was placed on ice and transported back to the lab. The transition zone is the area between the riverine and lacustrine zones and is characterized by decreased turbidity and increased phytoplankton productivity relative to the riverine zone (Wetzel 2001). We performed the enrichment experiment for the transition zone for several reasons. First, Beaver Water District, the regional domestic water supplier, has been drawing water from the transition zone since the 1970s and had the longest available dataset for the PCO₂ reconstructions (1986–2010). Second, the lake undergoes annual taste and odor episodes specifically in the transition zone. Our experiment is part of the research designed to better understand conditions leading to taste and odor episodes in the transition zone.

In the lab, the sample was subdivided over a 63 µm mesh to remove large zooplankton and placed into 3 acid-washed 20 L aquatainers. To buffer against pH changes, 5.5 g of CaCO₃/L were added to each aquatainer. Each aquatainer was aerated with CO₂ at 250, 400, and 550 µatm for 4 h to allow equilibration. From each aquatainer, 18 acid-washed 1 L cubitainers were filled with 0.95 L of the aerated water. Each cubitainer was amended with 1000 µg/L N as potassium nitrate (KNO₃) to prevent N limitation. The cubitainers were subdivided into threes and amended with P as disodium hydrogen phosphate (Na₂HPO₄) to achieve final concentrations of 5, 10, 25, 50, 75, and 100 µg/L, respectively (initial P concentration 5 µg/L). At the end of the setup, we had 54 cubitainers: 18 cubitainers at each PCO₂ level (250, 400, and 550 µatm) along an enrichment gradient of 5–100 µg/L P.

After N and P amendments, cubitainers were placed in a 30 C circulating water bath under artificial lighting. Light exposure was measured to be 500 µmol photons/m² s² and controlled by a timer on a 12 h on/off cycle. The conditions were designed to mimic in situ conditions at Beaver Lake during summer, ranging from 400 to 500 µmol photons/m² s² at the surface. The cubitainers were shaken manually each day to prevent particles from attaching to the sides. On days 4, 7, 10, and 13, ∼0.7 L of water was collected from each cubitainer to determine particulate C, N, and P content and Chl-a. Each cubitainer was renewed with the same volume of sterilized Beaver Lake water at the respective CO₂ concentration and initial concentrations of N and P. Following the experiment, samples were filtered onto Whatman GFF filters for phytoplankton PC, particulate N (PN), Chl-a, and particulate P (PP). Filtrate was also retained for dissolved nutrient analysis.

Algal P content was determined using the ascorbic acid method following a persulfate digestion (APHA 2005). Briefly, the filters were transferred to vials using 7.5 mL of Nanopure water with 7.5 mL of persulfate digestion solution (20g K₂S₂O₈/L) added to each vial. The vials were weighed and then autoclaved at 121 C for 1 h and allowed to cool overnight before being reweighed to calculate the percent recovery to account for water loss. After the digestion, P was determined according to the ascorbic acid method on a Turner Design Trilogy fluorometer with a spectrophotometric adapter. Algal C and N were determined using a Thermo Flash 2000 combustion elemental analyzer. Peach leaf standards were included with the P analyses and peach leaf and aspartic acid standards with the C and N analyses. From these data we calculated molar C:P, C:N, and N:P ratios for all treatments on all sample days.

Chl-a was measured to estimate algal biomass according to Standard Methods, with modifications (APHA 2005). Briefly, each filter was transferred into a 15 mL test tube containing 7 mL of 90% acetone solution. The samples were placed in a dark freezer for 24 h to further enhance complete pigment extraction. In a dark room, 3 mL of each sample extract was then transferred into VWR disposable test tubes and analyzed on a Turner Design Fluorometer. To adjust for the chlorophyll degradation product pheophytin, following the first measurement, 0.1 mL of 0.1 M HCl was added to each sample and measured again after 90 s.

Data analysis

All data^{Figure 1}were analyzed using SYSTAT 13 (Systat Software Inc, Chicago, IL). Simple linear regression was used to assess the relationship between modeled and measured PCO₂. To assess the impact of elevated PCO₂ and P on algal biomass and nutrient content, response variables were first averaged over the 13 d period, checked for normality and homoscedasticity, and transformed when appropriate. Simple linear regression was used to determine the relationship between algal biomass (PC, Chl-a) and P. We compared the slopes of the regression lines between biomass and P using analysis of covariance (ANCOVA) to determine if the rate of biomass increase was influenced by the covariate (PCO₂ level). A 2-way ANOVA was performed to determine if mean Chl-a, PC, C:N, C:P, and N:P ratios differed significantly across PCO₂ levels and to determine the synergistic effects of PCO₂ and P on algal biomass. When significant differences were obtained, post-hoc comparisons were performed with the Tukey's HSD test. Significance was set at P < 0.05.

Results

Carbon dioxide (CO₂) drawdown on decadal to daily scale and relationship to biomass

Beaver Reservoir was the oldest of the 4 lakes sampled as part of this project, with data representing 25 years (1986–2011) of reconstructed PCO₂ (Fig. 1a). Reconstructed PCO₂ ranged from 10 to 6800 μatm and averaged 1070 ± 46 μatm over the entire period of record. The 25-year summer average was 319 ± 49 μatm. Water column PCO₂ was greater than atmospheric PCO₂ in ∼79% of the reconstructed values and less than atmospheric PCO₂ in 21% of the values. Low water column PCO₂ occurred almost exclusively during summer. Chl-a values were low and ranged between 1 and 15 µg/L. Generally, higher Chl-a occurred during summer and coincided with low PCO₂ values (Fig. 1a).

Figure 1. Reconstructed carbon dioxide partial pressures (PCO₂) for (a) Beaver Lake (1966–2010) and Lakes (b) Brittany, (c) Norwood, and (d) Rayburn (2011 and 2012). Note the different scale on (a). The dotted line represents the current global PCO₂ level.

Lakes Brittany, Norwood, and Rayburn illustrate the results of PCO₂ reconstructed between 2011 and 2012. Among these lakes, Lake Brittany had the greatest average PCO₂ of 989 ± 50 μatm. Reconstructed PCO₂ ranged from 63 to 4336 μatm with a median value of 504 μatm. Values were generally greater than atmospheric PCO₂ in fall, winter, and spring and less than atmospheric PCO₂ during summer (Fig. 1b). Lake Norwood had an average PCO₂ of 528 ± 26 μatm. Reconstructed PCO₂ values ranged from 32 to 3388 μatm with a median value of 273 μatm (Fig. 1c). Lake Brittany had water column PCO₂ greater than atmospheric levels in fall, winter, and spring and less than atmospheric PCO₂ during summer. Lake Rayburn also had higher PCO₂ values during fall, winter, and spring and lowest values during summer (Fig. 1d). Comparatively, Lake Rayburn had the lowest PCO₂ of the younger lakes and averaged 405 ± 28 μatm. Reconstructed PCO₂ ranged from 13 to 3780 μatm with a median value of 127 μatm. Values then increased from 19:00 h onward. Chl-a concentrations were lowest at Lake Brittany and ranged from 2 to 10 µg/L (Fig. 1b). At Lakes Norwood and Rayburn Chl-a ranged from 2.5 to 35 µg/L and 8 to 40 µg/L, respectively. Similar to Beaver Lake, Chl-a concentrations were highest during summer and generally coincided with low PCO₂ values (Fig. 1c and d). For both 2011 and 2012, daily reconstructed PCO₂ was typically highest from 23:00 to 07:00 h and lowest from 12:00 to 18:00 h. (Fig. 2^{Table 2}).

Figure 2. Reconstruction of carbon dioxide partial pressures (PCO₂) during June, July, and August in (a–c) 2011 and (d–f) 2012 at Beaver Lake, northwest Arkansas. The dotted line represents the current global PCO₂ level.

For the 2011 reconstructions, June PCO₂ ranged from 40 to 209 μatm, July from 18 to 86 μatm, and August from 14 to 66 μatm (Table 2). Mean PCO₂ was lowest in August (35 ± 1 μatm) and highest in June (99 ± 3 μatm). For the 2012 reconstructions, June PCO₂ ranged from 8 to 339 μatm, July from 32 to 195 μatm, and August from 131 to 361 μatm (Table 2). Mean PCO₂ was lowest in July (74 ± 2 μatm) and highest in August (228 ± 4 μatm).

Table 2. Summary of CO₂ partial pressure (PCO₂) for 22 June, 22 July, and 22 August 2011 and 2012.

	Carbon dioxide partial pressure (PCO2 µatm)
Month	Maximum	Minimum	Median	Mean
22 June 2011	209	40	89	99 ± 3
22 July 2011	86	18	34	44 ± 1
22 August 2011	66	14	37	35 ± 1
22 June 2012	339	8	55	76 ± 4
22 July 2012	195	32	58	74 ± 2
22 August 2012	361	131	203	228 ± 4

Diurnal PCO₂ and δ¹³C direct measurements

The diurnal PCO₂ measurements were similar in trend to the PCO₂ reconstructed for JJA 2011 and 2012 in that maximum and minimum PCO₂ had a predictable diurnal variation. At the beginning of the sampling, measured PCO₂ was 180 μatm at 20:00 h, which was less than half of atmospheric PCO₂ (Fig. 3a). By 06:00 h the following day, measured PCO₂ increased to a maximum value of 800 μatm, approximately double atmospheric PCO₂. The PCO₂ then fell to 150 μatm around midday and remained at that level until the end of the sampling at 19:00 h (Fig. 3a). The δ¹³C exhibited a trend opposite that of measured PCO₂ (Fig. 3a). Minimum δ¹³C of −12‰ occurred at 06:00 h when PCO₂ was at a maximum (Fig. 3a). The highest δ¹³C of 6‰ was measured at 19:00 h when PCO₂ was at a minimum (Fig. 3a).

Figure 3. Changes in δ¹³C and measured PCO₂ over (a) a 24-hour period at Lake Fayetteville, northwest Arkansas, and (b) relationship between modeled and measured PCO₂ for Lake Fayetteville, northwest Arkansas. The dotted line represents the current atmospheric PCO₂ level.

Modeled PCO₂ values were typically lower than measured PCO₂ values (Fig. 3a). Underestimation seemed to be greater at lower PCO₂. For example, at the start of the sample collection, measured PCO₂ was 180 μatm compared to the modeled value of 62 μatm (Fig. 3a). Modeled and measured values were almost identical at 06:00 and 09:00 h, at 800 and 700 μatm, respectively. Despite the underestimation, modeled PCO₂ was significantly correlated to measured PCO₂ (r² = 0.99, P < 0.001; Fig. 3b).

Algal response to elevated PCO₂ and P

To determine if Chl-a and seston PC were good proxies for algal biomass, we plotted a regression-based relationship between the 2 variables (Fig. 4). PC and Chl-a were significantly correlated to each other at every level of PCO₂ (P < 0.001; Fig. 4). The strength of the relationship, expressed by the r² value, was strongest at PCO₂ values of 400 and 550 µatm (r² = 0.88 and 0.81, respectively) and weakest at 250 µatm (r² = 0.54).

Figure 4. The relationship between chlorophyll a (Chl-a) and particulate carbon (PC) at (a) 250 µatm, (b) 400 µatm, and (c) 550 µatm.

Algal PC ranged from a low of 2 mg/L to a maximum of 8 mg/L across all PCO₂ treatments (Fig. 5a–c). Lowest PC was measured in cubitainers that received no P treatments, whereas highest PC occurred at P treatments of 100 μg/L. PC was positively related to P concentration irrespective of PCO₂ level (Fig. 5a–c). The greatest increases in PC generally occurred when P increased from 10 to 25 μg/L and from 25 to 50 μg/L. Doubling P approximately doubled PC, but the effect was not observed when P increased from 50 to 100 μg/L (Fig. 5a–c). There was a significant response in mean PC to P treatment (F_4,10 = 91, P < 0.001). Tukey's post-hoc test values were used to summarize P comparisons (Table 3).

Figure 5. Algal biomass response reported as (a, b, and c) particulate carbon (PC) and (d, e, and f) chlorophyll-a to different levels of PCO₂ along a phosphorus concentration gradient. The r² values represent the linear fit with P concentration as the x-variable.

Table 3. Pairwise comparison of phosphorus treatment and corresponding p-values for particulate carbon (PC), chlorophyll a (Chl-a) and carbon to phosphorus (C:P), carbon to nitrogen (C:N), and nitrogen to phosphorus (N:P) ratios. Results were from a 2-way ANOVA, but there were no significant differences in PCO₂, so only P treatment values are shown. Values <0.05 represent significant differences.

P treatment comparison	PC	Chl-a	C:P	C:N	N:P
(ug/L)	p-value
100 vs. 5	<0.001	<0.001	<0.001	NS	<0.001
100 vs. 10	<0.001	<0.001	<0.001	NS	<0.001
100 vs. 25	<0.001	0.002	<0.001	0.005	<0.001
100 vs. 50	NS	NS	0.021	0.018	NS
75 vs. 5	<0.001	0.002	<0.001	NS	<0.001
75 vs. 10	0.001	0.012	<0.001	NS	<0.001
75 vs. 25	<0.001	NS	<0.001	0.03	0.002
50 vs. 5	NS	NS	<0.001	NS	<0.001
50 vs. 10	0.004	NS	<0.001	NS	<0.001
50 vs. 25	0.002	NS	<0.001	NS	NS
25 vs. 5	<0.001	NS	<0.001	NS	<0.001
25 vs. 10	<0.001	NS	<0.001	0.033	NS

Increased PCO₂ had a small but measurable effect on PC concentration. For example, PC at 25 μg/L P and 250 μatm PCO₂ was ∼4 mg/L compared to 6 mg/L at 550 μatm PCO₂ (Fig. 5a and c), but the increase was not significant (F_2,13 = 0.06, P = 0.94).

Chl-a concentration ranged from 2 to 15 μg/L across all PCO₂ levels (Fig. 5d–f). Similar to PC, the lowest Chl-a was measured in no P treatments, whereas highest Chl-a was measured at P levels of 100 μg/L (Fig. 5d and f). Doubling P concentration did not double Chl-a as observed with PC. Chl-a was positively correlated to P levels and increased significantly as P level increased (Fig. 5d and f). Mean Chl-a was significantly different across P treatments (F_4,10 = 15, P < 0.001). Tukey's post-hoc test results were used to summarize P comparisons (Table 3). Increased PCO₂ had no measurable effect on Chl-a concentration. Similar to PC, mean Chl-a was not significantly different across PCO₂ levels (F_2,12 = 0.03, P = 0.97).

Effect of PCO₂ and P on algal nutrient content

There were no significant differences in algal C:P, C:N, and N:P ratios at individual levels of P for PCO₂ treatment (P = 0.069, 0.640, and 0.170 respectively), but significant differences were observed with increasing P levels. For example, between 50 and 100 μg/L, the C:P ratio declined significantly as P increased (Fig. 6; F_2,2 = 110, P < 0.001). Tukey's post-hoc test results were used to summarize the pairwise comparisons of P for the C:P ratio (Table 3). The C:P ratio indicated P limitation at P concentrations between 5 and 50 μg/L P. Algal C:N increased significantly from about 10 at 5 μg/L P to 33 at 100 μg/L P (Fig. 6b). Tukey's post-hoc test results were used to summarize the pairwise comparisons of P for the C:N ratio (Table 3). The C:N ratio suggested N limitation at P concentrations >10 μg/L P (Fig. 6b). The N:P ratio declined significantly with increasing P concentration (Fig. 6c; F_2,5 = 52, P = 0.001). The N:P ratio suggested P limitation at P concentrations <25 μg/L P. Tukey's post-hoc test results were used to summarize the pairwise comparisons of P for the N:P ratio (Table 3).

Figure 6. Algal carbon to phosphorus (C:P), carbon to nitrogen (C:N), and nitrogen to phosphorus (N:P) molar ratio across a phosphorus gradient at carbon dioxide partial pressures (PCO₂) of 250 µatm, 400 µatm and 550 µatm.

Discussion

Results from our study suggest elevated PCO₂ is not likely to increase algal growth by alleviating algal C limitation or alter carbon:nutrient ratios in NWA reservoirs. This result was contradictory to our expectations because the field-based PCO₂ reconstructions and δ¹³C analysis suggested inorganic carbon limitation during summer. For the 4 lakes where PCO₂ was reconstructed, summer concentrations were consistently below atmospheric PCO₂. The decline in PCO₂ was simultaneous with Chl-a increase, suggesting active CO₂ drawdown during primary production to potentially limiting levels. During fall, winter, and spring, reconstructed PCO₂ was generally greater than atmospheric PCO₂.

The mean PCO₂ calculated for Beaver Lake (1070 µatm) in this study is consistent with the mean value for global lakes (1036 µatm) reported by Cole et al. (1994) and for about a 1000 lakes across the United States (1043 µatm) reported by McDonald et al. (2013). Lakes Brittany, Norwood, and Rayburn were all less than the global average, possibly due to their smaller watershed sizes and lower inputs of allochthonous C from their tributaries (McDonald et al. 2013). Nevertheless, the temporal variations in PCO₂ were similar to that of the larger Beaver Lake. Our results suggest that NWA reservoirs are sources of CO₂ during fall, winter, and spring, but during summer, the reverse occurs, and the reservoirs become a sink for CO₂. Because the number of lakes similar in size to Lakes Brittany, Norwood, and Rayburn are more globally abundant, low summer PCO₂ is possibly much more widespread than currently known. Our study further demonstrated that long-term studies are needed to understand the seasonal role individual lakes play in CO₂ dynamics given their importance in regulating regional climate and global C cycling (Alin and Johnson 2007, Battin et al. 2008, Balmer and Downing 2011).

Interpretations from our analysis of PCO₂ limitation depended on the validity of the PCO₂ reconstructions. Algal abundance through CO₂ uptake and pH could both lead to lower PCO₂ values. Quiñones-Rivera et al. (2015) showed that when lake water pH fell below 8.6 in Lake Diefenbaker, Canada, CO₂ was released. The pH in all the NWA lakes was <8.6, but, interestingly, Lake Brittany had the lowest pH and also the highest annual average PCO₂. This finding suggests that degassing due to pH was not a factor in this study, and variations in reconstructed PCO₂ were reflective of water column PCO₂.

We used 2 techniques to validate the model: direct measurement of PCO₂ in the water column and measurement of δ¹³C-CO₂. The strong agreement between the measured and modeled values suggests that the model accurately predicted water column PCO₂ based on the input values. Values of δ¹³C have been used to track source and transformation of C in numerous studies (Friedli et al. 1986, Clark and Fritz 1997, Karlsson et al. 2007, Winston et al. 2014). Typically, as CO₂ is utilized for primary production, concentrations will decrease, whereas the δ¹³C-CO₂ will increase due to the preferential incorporation of the lighter C (¹²C) into biomass. Based on previous work, atmospherically derived CO₂ has a δ¹³C range from −6 to 0‰, respiration from −28 to −31‰, and ground water from −5 to −25‰ (Friedli et al. 1986, Clark and Fritz 1997, Karlsson et al. 2007).

Variations in δ¹³C during the study corroborated the reconstructions, and the range in δ¹³C values indicated potential C sources. During the night, water-column PCO₂ was almost twice atmospheric PCO₂ with a concomitant decrease in δ¹³C. Unexpectedly, the δ¹³C values were higher than respiration-derived δ¹³C and ranged from −12 to 7‰, reflecting a C source dominated by atmospheric and ground water inputs. In a recent study of 1000 lakes, 12% had C for primary production subsidized with hydrologic DIC inputs (McDonald et al. 2013). The results suggest that NWA reservoirs maybe supplemented by hydrologic C inputs given the vast array of caves, tributaries, and preferential flow paths present in the watershed.

During the day, as PCO₂ in the water column decreased from supersaturated to undersaturated levels, δ¹³C increased. Values ranged from −3 to 0‰ at 12:00 h and then increased to 7‰ at 18:00 h. Values of δ¹³C >0‰ suggested that utilization of CO₂ during primary production exceeded supply of CO₂ into the water column (Falkowski and Raven 1997) leaving the residual pool enriched in δ¹³C. This result was consistent with conclusions from Shapiro (1973) and Barnese and Schelske (1994), who suggested that atmospheric CO₂ solution may not fully supply phytoplankton carbon demand during primary production.

Chl-a and PC are both used as proxies for phytoplankton biomass. Chl-a is more frequently used because it is easier to measure (Felip and Catalan 2000), but it may underestimate biomass due to species-specific differences. By comparison, PC may overestimate biomass by incorporating detritus, primarily from recently deceased planktonic organisms. The relationship between the 2 variables was significantly correlated at all PCO₂ levels (P < 0.001), including at 250 μatm where the strength of the relationship seemed weaker (r² = 0.54). Overall, the 2 variables suggest they were good surrogates of each other and potentially biomass, particularly at the higher PCO₂ levels.

We observed no significant differences in biomass or ratios of C:P, C:N, C:P with elevated PCO₂. These results contradicted studies by (Hein 1997), Urabe et al. (2003), and Jansson et al. (2012) but were similar to results obtained by Verschoor et al. (2013). Algal community composition and nutrient limitation might account for the differences.

Phytoplankton taxa differ in their ability to uptake CO₂ and in their growth response to elevated PCO₂ (Rost et al. 2003, Riebesell 2004). Cyanobacteria tend to grow faster at lower PCO₂, whereas chlorophytes grow faster at higher PCO₂ (Birmingham and Colman 1979, Low-Décarie et al. 2011). Additionally, the competitive response of cyanobacteria decreases as PCO₂ increases, whereas that of chlorophytes steadily increases (Low-Décarie et al. 2011). Diatom growth responses are intermediate and tend to show little change with elevated PCO₂ (Low-Décarie et al. 2011). Although the algal community was not assessed microscopically, previous work suggests the community would have been dominated by diatoms at the time of sample collection (Winston et al. 2014). Diatom indifference toward elevated PCO₂ (Low-Décarie et al. 2011) might account for the lack of significant relationships among PCO₂, biomass, and nutrient content, suggesting that responses to elevated PCO₂ might be dependent on the in situ algal community and species-specific CO₂ affinities (Gradoville et al. 2014).

Verschoor et al. (2013) suggests the lack of a significant increase in biomass was because elevated PCO₂ relieved C limitation and drove phytoplankton toward nutrient limitation. Nutrient limitation was also likely in this study. Using limitation lines of Healey and Hendzel (1979), C:P and N:P ratios indicated P limitation between 5 and 50 µg/L, whereas the C:N ratio indicated N limitation at P >10 µg/L. PCO₂-driven limitation, however, does not seem to be the mechanism driving limitation in this study. Parallel line analysis indicates that the slope of the regression between biomass and P level were not significantly different among PCO₂ treatments.

Micronutrient co-limitation, specifically iron (Fe), likely played a role in dampening the increases in biomass. North et al. (2007) showed that the addition of Fe reduces N limitation by allowing algal communities to uptake more nitrate. Because nitrate was the source of N in this study, uptake was possibly limited by low Fe concentration (North et al. 2007) because micronutrients were not part of the enrichment regime. This could further explain why N limitation occurred above 10 µg/L P, despite enriching cubitainers with 1000 µg/L N.

The most significant increases in biomass, C:N, and N:P occurred with elevated P, an expected finding because P, or combinations of N and P, have been shown to control phytoplankton biomass (Schindler 1977, Dzialowski et al. 2005, Scott and McCarthy 2010). Thus, even under conditions of elevated PCO₂, nutrients provide primary control on phytoplankton biomass. Several researchers have suggested that sound nutrient management remains the most feasible method for curtailing eutrophication symptoms (Conley et al. 2009, Esten and Wagner 2010, Wagner 2010) under global climate change scenarios (Paerl and Paul 2012).

Conclusion

Overall, our results suggest that the simultaneous increase in C and P in NWA reservoirs is not likely to increase algal biomass and alter algal nutrient content any more than by increasing P alone. Our results showed significant changes in biomass and nutrient content only with increasing P. We agree with recommendations from several researchers that sound nutrient management remains the best method for curtailing eutrophication symptoms under climate change scenarios. The study also provided a cost-effective way to obtain long-term data on C dynamics. Given the abundance of reservoirs globally, this approach could provide insight into CO₂ fluctuations on a more refined spatial and temporal scale and help better constrain the modern global carbon budget.

Funding

Special thanks to the Arkansas Water Resources Center for funding this project, USGS 104 B.

Acknowledgments

Special thanks to the Scott Biogeochemistry and University of Arkansas Stable Isotope Labs (UASIL) for analyses, and Aaron Jackson, Hal Halvorson, Amie West, Erin Scott, and Maysill Pascal for comments on the manuscript. Special thanks to Dr. Douglas Lepley for his invaluable contribution to the manuscript.