Dissolved phosphorus concentrations in Cayuga Lake system and differences from two analytical protocols

ABSTRACT

Effler SW, Prestigiacomo AR, Hairston NG, Auer MT, Kuczynski A, Chapra SC.  2016. Dissolved phosphorus concentrations in Cayuga Lake system and differences from two analytical protocols. Lake Reserve Manage. 32:392–401.

Differences in the concentrations of dissolved forms of phosphorus (P) measured with 2 widely used spectral protocols were documented and evaluated for Cayuga Lake, New York, and 4 of its primary tributaries. The analysis focuses on 2 operationally defined forms of dissolved P, soluble reactive P (SRP) and soluble unreactive P (SUP), which together constitute dissolved P (TDP). Direct comparisons were based on analysis of the results from the 2 protocols of split samples of year-round deep water representative of the entire water column during turnover and the respective dependencies of tributary concentrations on stream flow. Although the TDP concentrations converged for the 2 protocols, there were systematic differences for the contributions of SRP versus SUP (i.e., one protocol yielding lower SRP but higher SUP). The interpretive implications of the differences in the operationally defined concentrations from the 2 analytical protocols were considered in the context of common limnological and bioavailable paradigms for these forms of P, and the needs and structures of mechanistic P-eutrophication models. The lower SRP analytical protocol was favored because of its greater consistencies with the independent bioavailability results, limnological paradigms for dissolved forms of P, and contemporary mechanistic modeling. The differences between the 2 protocols are particularly problematic where contemporary mechanistic models are to be applied, requiring compensating differences in kinetic representations, and likely structure, for cases of higher SRP datasets attributable to the protocol.

KEYWORDS:

• Dissolved phosphorus
• lakes
• phosphorus analyses
• phosphorus in models
• streams

Cultural eutrophication remains a major water quality and ecological issue for lake and reservoir management, with phosphorus (P), the limiting nutrient in most of these lacustrine waters (Schindler et al. 2008), the primary focus. Early studies often considered only concentrations of total P (TP) with respect to inputs, the in-lake pool, and management of P (Vollenweider 1976). Partitioning of multiple forms of P became important in the context of limnological understanding and management as the differences in their potential to support primary production were recognized (DePinto et al. 1981, Wetzel 2001, Reynolds and Davies 2001). P available to support algae and cyanobacteria production is described as bioavailable (DePinto et al. 1981, Auer et al. 1998).

The chemical composition of the multiple forms of P has not been amenable to routine determination (Hudson et al. 2000, Dodds 2003). Orthophosphate ion (PO^3-₄) provides the P necessary to support algal growth (Hudson et al. 2000, Wetzel 2001). Certain forms of dissolved organic P (DOP) can be converted to PO^3-₄ from enzymatic activity by phosphatases, especially under PO^3-₄-deficient conditions (Currie et al. 1986, Bentzen et al. 1992). Much of the terrigenous particulate P (PP) is not bioavailable (Prestigiacomo et al. 2016).

Spectrophotometric procedures continue to be the primary type of P analysis used to monitor fresh waters Three forms of P are commonly measured: TP, total dissolved P (TDP), and soluble reactive P (SRP). Two other forms are calculated from these 3 as residuals: particulate P (PP = TP − TDP) and soluble unreactive P (SUP = TDP − SRP). SUP is usually assumed to be composed primarily of forms of DOP (Baldwin 2013). The concentrations of these forms are “operationally defined,” reflecting the analytical protocols specified in a particular methodology. A paradigm has emerged for the differences in the bioavailability of the operationally defined forms of P delivered to lakes that partition TP. SRP has been considered to be completely available in the short-term (Auer et al. 1998, Prestigiacomo et al. 2016). By contrast, SUP has been described as less bioavailable (Auer et al. 1998, Prestigiacomo et al. 2016), becoming bioavailable over longer time scales through enzymatic and mineralization processes Currie et al. 1986, Bentzen et al. 1992). PP is the least bioavailable of these forms (Auer et al. 1998, Prestigiacomo et al. 2016).

The SRP fraction is often incorrectly described as a quantitative measure of PO^3-₄ (Dodds 2003). Radio-bioassay measurements for more than 40 years have established that PO^3-₄ represents only a small fraction (e.g., <1%) of SRP in productive layers of P-limited pelagic waters (Hudson et al. 2000, Hudson and Taylor 2005). Other components included in SRP are inorganic polyphosphates, the more reactive components of DOP (Dodds 2003), and reactive small (<0.45 µm) particles (Dodds 2003). The multiple analytical methods for SRP (6 in New York) are a potential source of variations in this operationally based concentration (Tarapchak et al. 1982, Tarapchak and Rubitschun 1981), with complicating implications for limnological interpretations and water quality modeling. Paired or even noncoincident results from multiple analytical protocols for P, particularly in combination with bioavailability assessments, are generally not available from contemporary monitoring programs.

The overarching goals of this analysis were to (1) document differences in concentrations of dissolved forms of P for a lake and its tributaries obtained with 2 widely used spectrophotometric protocols, (2) consider their consistencies with parallel bioavailability information for the fractions, and (3) consider the implications for limnological and water quality modeling analyses. This analysis is based on a rare combination (described subsequently) of monitoring and process information available for Cayuga Lake, New York, and its tributaries. It addresses the problems of limnological interpretation and modeling for lacustrine waters where fundamental protocol-based differences in concentrations of P fractions have been reported.

Study site

Cayuga Lake (42°41′30″N; 76°41′20″W) is the fourth easternmost of the 11 New York Finger Lakes (Fig. 1). It has the second largest volume (9.4 × 10⁹ m³) and surface area (172 km²) of the Finger Lakes, with mean and maximum depths of 55 m and 133 m, respectively. This lake has a monomictic stratification regime and a retention time of ∼10 years (Effler et al. 2010). Phytoplankton growth is P limited, and the lake is mesotrophic (Effler et al. 2010). The entire water column remains oxic throughout the year.

Figure 1. Cayuga Lake, the 11 Finger Lakes, and position in New York. Shown are 4 monitored tributaries, USGS gauges, pelagic site, lake source cooling (LSC) discharge, and shelf portion of the lake at its southern end.

Four of the primary tributaries to the lake were gauged for flow (Table 1); 3 enter the shallow southern end (“shelf”; Fig. 1) and one is positioned 10 km farther north, entering from the east (Fig. 1). Large fractions of the P loads received by these tributaries were delivered during runoff events (Prestigiacomo et al. 2016), which justifies a focus on characterizing these events. Agricultural land use varies, representing 22–68% of these watershed areas (Table 1). Cornell University, situated near the south end of the lake, installed a lake source cooling (LSC) facility in 2000 that withdraws water from a depth of 73 m (in the hypolimnion) to meet much of the institution's cooling needs and returns the spent cooling water (closed system, no P added within the facility) to the shelf.

Table 1. Specifications for inputs to Cayuga Lake monitored for forms of phosphorus (P); watershed areas, agricultural land use, samples analyzed according to the 2 protocols, and bioavailable fractions for the P forms as analyzed by the standard methods protocol.

				Samples		fBAP^e
Tributary/Input	USGS Gage	Watershed Area (km^2)	Land use Ag^a(%)	E^b	S^d	SRPS	SUPS	PPS	SUPS P (µg/L)^f	SRPS P (µg/L)^f
				02–13^c	2013
Fall Cr.	4234000	330.9	49%	54	97	0.96	0.75	0.13	8.7	10.8
Cayuga Inlet	4233255	240.8	56%	22	72	0.98	0.54	0.06	5.3	4.7
Salmon Cr.	423401815	233.8	68%	31	96	0.99	0.84	0.19	8.5	27.2
Six Mile Cr.	4233300	134.1	22%	56	79	0.99	0.56	0.08	4.8	10.3
LSC				48	48	1.00	0.08	0.14	2.7	8.9

^a Ag = agriculture.

^b number of samples collected and analyzed according to method (USEPA 1978).

^c for years 2002–2013.

^d number of samples collected and analyzed according to S method (Clesceri et al. 1998).

^e f_BAP – fraction of P form bioavailable, as reported by Prestigiacomo et al. (2016).

^f flow-weighted concentrations, from Prestigiacomo et al. (2016).

Materials and methods

Two widely used methodologies were applied to measure 3 forms of P (TP, TDP, and SRP): Standard Methods No. 4500 PE and 4500-P B(5) (Clesceri et al. 1998), hereafter called the S protocol, and USEPA No. 365.3 (USEPA 1983), hereafter called the E protocol. Both protocols are based on an ascorbic acid method (e.g., Murphy and Riley 1962). Ammonium molybdate and antimony potassium tartrate react in an acid solution with PO^3-₄ to form a heteropoly acid that is reduced to an intensely colored molybdenum blue by addition of ascorbic acid. The 2 protocols have certain operational differences (selected features in Table 2), detailed elsewhere (Clesceri et al. 1998, USEPA 1983). Results from the 2 methods are henceforth differentiated by subscripts “S” (e.g., SRP_S) and “E” (e.g., SRP_E), corresponding to the Standard Methods and EPA protocols, respectively.

Table 2. Comparison of selected analytical features for 2 methodologies for measurements of the concentration of phosphorus (P).

	Method
Analytical Step	SM 4500P-E	USEPA 365.3 (1983)
1. Sample volume	50 mL	50 mL
2. Reagent addition	simultaneously, as a “mixed”	sequential additions
	reagent
3. Reagents for color determination	5 N sulfuric acid (50 mL)	11 N sulfuric acid (1 mL)
	antimony potassium tartrate	ammonium molybdate-
	solution (5 mL)	antimonypotassium tartrate
	ammonium molybdate (15 mL)	(4 mL)
	ascorbic acid (30 mL)	ascorbic acid (2 mL)
4. Colorimetric reaction time	read samples within 10–30	read sample within 5 minutes
	minutes of “mixed reagent”	of addition of ascorbic acid
	addition
5. Wavelength	880 nm	660 nm or 880 nm

Paired weekly samples from the intake water at the LSC facility, representative of lake water at a depth of 73 m (hereafter “deep lake water”), were collected throughout 2013 for analysis of TP, TDP, and SRP according to both the S and E protocols, and carried out by 2 laboratories. The numbers of paired measurements of TP, TDP, and SRP were 48, 26, and 26 respectively; TDP_E discontinued after 26. Measurements of TDP_S, TP_S, and SRP_S were made for samples from a pelagic lake site (laboratory filtered chilled sample) during 1999−2006, biweekly (1 per 2 wk) for the April−October interval (Effler et al. 2010).

The samples supporting comparison of the 2 analytical protocols here for the 4 Cayuga Lake gauged tributaries (Table 1) were not paired but were collected at positions proximate to the mouths (Fig. 1). A long-term monitoring program has been conducted since 2002 adopting the E analytical protocol to measure TP_E and SRP_E (i.e., no TDP_E). These samples were collected manually on a temporally irregular basis in each year, targeting runoff events (up to 5 times per year) to reflect the effects of a range of stream flow conditions. A shorter-term, but more temporally intensive, sampling program was conducted over the April−October interval of 2013, with analyses of TP_S, TDP_S, and SRP_S (Prestigiacomo et al. 2016). This sampling program had 2 components: (1) biweekly (1 per 2 wk) manual collections, and (2) collections for 4 major runoff events with automated sampling equipment (e.g., multiple samples within days of events). The results from the 2 analytical protocols were compared in the common logarithmic concentration versus stream flow (Q, m³/s) format (e.g., Vogel et al. 2003). Because the samples used in the tributary protocols were not directly paired, tributary results were compared by analyzing the respective log-transformed flow-concentration relationships for the 2 protocols. The flow-concentration (log) slopes for the E and S relationships were tested using the homogeneity of slopes test (Statsoft 2003), and the intercepts were evaluated using an analysis of covariance (ANCOVA; Helsel and Hirsch 1992).

The bioavailability of SRP_S, SUP_S, and PP_S in the 4 gauged tributaries to the lake was assessed previously through bioassays on samples from 3 runoff events in 2013 with a P-starved green alga (Selenastrum capricornutum), as described by Prestigiacomo et al. (2016). This procedure provided measurements of the fraction of these forms taken up by the assay alga, defined as the bioavailable fraction, f_BAP (Table 1). The assays were additionally conducted for SRP_S and SUP_S for 2 samplings of the LSC facility discharge in 2014. (www.communityscience.org/database/monitoringregions/1; Prestigiacomo et al. 2016). These bioavailability results reported by Prestigiacomo et al. (2016) are used here as a basis to evaluate the character of differences in P concentrations reported for the S and E protocols. The paradigm of contrasting bioavailability of the forms of P was supported; SRP completely, SUP diminished, and PP very limited (Table 1; Auer et al. 1998, Prestigiacomo et al. 2016).

Results

Patterns for forms of P in deep and upper lake waters

Time series of TP, TDP, and SRP are presented for the paired hypolimnetic samples (LSC discharge) for the 2 analytical protocols for 2013 (Fig. 2; note measurements of TDP_E were discontinued at the end of June). Substantial temporal variation occurred for each of these forms according to both protocols. The temporal patterns of the 2 measurement methods for TP, TDP, and SRP were each significantly correlated (P < 0.05; r values of 0.85, 0.55 and 0.78, respectively). The most conspicuous differences in the patterns for the 2 protocols were that SRP_E exceeded SRP_S (Fig. 2b), and there were therefore greater differences in concentrations (SUP_S vs. SUP_E) between TDP_S and SRP_S (Fig. 2b and c) than between TDP_E and SRP_E (Fig. 2b and c). This second feature means that SUP_S was consistently higher than SUP_E in deep lake water.

Figure 2. Time series for the LSC discharge for 3 forms of phosphorus (P) for 2013, according to 2 analytical protocols, standard methods (S) protocol (subscript S), and EPA protocol (subscript E): (a) TP, (b) TDP, and (c) SRP.

Results from 2 analytical protocols were compared using the distributions of the ratios of the paired results for both the directly measured and calculated forms of P (Fig. 3). The 2 methods produced similar results for TP (i.e., TP_E:TP_s ≈ 1.0; Fig. 3a) and TDP (Fig. 3b); ∼70 and 75% of the observations of these forms of P were within 10% of each other. Accordingly, a general convergence was observed for the calculated PP (Fig. 3c; median ratio of 0.95). The modest contribution of PP to TP, together with the fact that it was calculated by difference (PP = TP − TDP), probably contributed to the PP ratio variance.

Figure 3. Representations of extent of closure of forms of P in the LSC discharge from split samples for the S and E protocols as distributions of ratio values: (a) TP, (b) TDP, (c) PP, (d) SRP, (e) SUP, and (f) the ratio SRP_E:TDP_S.

Values of SRP_E were systematically greater than SRP_S; the mean and median of SRP_E:SRP_S were 1.28 and 1.25 (Fig. 3d), respectively. SRP was the dominant form of P in these deep lake waters according to both protocols, although the contribution of SRP_E (88%) was greater than that of SRP_S (67%). These differences in SRP_E and SRPs, together with the near equivalence of TDP values from the 2 protocols, caused SUP_S to exceed SUP_E. According to the analytical results from the S protocol, SUP_S represented 23% of TP for the January−June interval and 21% for the entire year, substantially more than the 7% contribution indicated by the E protocol for the period January−June (TDP_E and SUP_E were not determined for July−December). The mean and median SUP_E:SUP_S values were 0.34 and 0.30 (Fig. 3e), respectively. The average values of the SRP_E:TDP_E and SRP_E: TDP_S (Fig. 3f) ratios were 0.94 and 0.96, indicating SRP_E approached TDP for deep lake water.

Seasonal patterns of TDP_S, SRP_S and SUP_S in the upper waters of Cayuga lake are presented as monthly mean values for the pelagic monitoring site for the April–October interval for 1999−2006 and 2013 (Fig. 4a). Despite interannual differences (1 standard deviation, vertical bars), recurring patterns were noted. The highest TDP_S and SRP concentrations were observed during turnover in April (Effler et al. 2010); both decreased, but particularly SRP_S. Although SUP_S was somewhat lower at turnover, it increased modestly by midsummer. SRP_S concentrations remain mostly <1 µg/L over the May–October interval. The measured SRP_S concentrations (as P) were often near a detection limit of 0.4 µg/L. SUP_S was the dominant component of TDP_S over the May–October interval; the average summertime SUP_S:TDP_S ratio value has been ∼0.8. The 8 April 2013 observations (Fig. 4b) corresponded to spring turnover, as shown by the near equivalence of the TDP_S and SRP_S measurements and SUP_S calculated values for samples from the upper waters and the LSC discharge (Fig. 4b). The substantial differences in the deep-water LSC E protocol values (SRP_E and SUP_E) from the reported S protocol observations for that day are shown. These differences would also have been observed in the upper lake waters on this day of turnover.

Figure 4. Lake P patterns and (a) time series of monthly average TDP_S, SRP_S, and SUP_S in the upper waters of Cayuga Lake for the April-October 1999–2006 interval and 2013; vertical bar corresponds to ±1 standard deviation, (b) comparison of TDP, SUP, and SRP concentrations in early April (during turnover) 2013 for deep wake Water and from the S and E protocols and upper lake water from S protocol.

Patterns for forms of P in tributaries

Comparisons of tributary P concentrations measured by the 2 analytical protocols were made using the commonly adopted log-transformation format (Vogel et al. 2003), log₁₀ concentration versus log₁₀ Q (daily average flow rate; Fig. 5). These comparisons are supported, despite differences in the temporal features of the associated monitoring programs, by a lack of any systematic changes in the concentration versus Q relationships over the long-term monitoring program (www.communityscience.org/database; Prestigiacomo et al. 2016). The comparisons presented here for the mouth of Fall Creek (Fig. 5) are generally representative for the 4 gauged tributaries (Prestigiacomo et al. 2016).

Figure 5. Phosphorus (P) concentration versus stream flow (Q) relationships for Fall Creek that compare results from the 2 analytical protocols: (a) dependencies of TP_S and TP_E on Q, (b) dependencies of SRP_S and SRP_E on Q, and (c) dependencies of TDP_S and SRP_E on Q.

The relationships between TP and Q (Fig. 5a) were similar for the longer-term TP_E and the 2013 TP_S datasets (P < 0.05; homogeneity of slopes test for slopes; ANCOVA for intercepts). The situation for SRP was strikingly different. SRP_E values shifted higher than the SRP_S values, with greater variation in the SRP_S dataset (Fig. 5b); the intercept of the power law relationship for SRP_E was significantly greater than for SRP_S (P < 0.001). By comparison, the SRP_E versus Q and TDP_S verus Q relationships, and thereby the concentrations, were similar (Fig. 5c), although the intercepts were significantly different (P = 0.003). Although SUP_E was not assessed in these streams (no TDP_E measurements), the analyses of the available information (Fig. 5) indicates it would have been low by comparison to SUP_S (e.g., note convergence of SRP_E and TDP_S relationships; Fig. 5c). This analysis of the datasets for the streams indicates that the same diverging characteristics in the concentrations of SRP and SUP determined by the 2 analytical protocols for deep lake waters were also manifested for the streams entering the lake (i.e., SRP_E > SRP_S, and SUP_E < SUP_S.

Discussion

Protocol-based differences in P concentrations

Analyses of split samples for deep Cayuga Lake water showed definitively that values of SRP_E were consistently greater than those for SRP_S and conversely that SUP_S was consistently greater than SUP_E (Fig. 3d and e). The differences in SUP were driven by the differences in SRP as TDP values from the 2 protocols converged. Analogous differences between the 2 SRP protocols for the lake's tributaries is supported by the significantly higher intercept of the SRP_E versus Q relationship compared with that for SRP_S versus Q (Fig. 5b). The case that SUP_S would have systematically exceeded SUP_E in the tributaries had TDP_E been measured is also strong, based on (1(1) ) the similarity of the SRP_E versus Q and TDP_S versus Q relationships for the tributaries (Fig. 5c), (2) the similarity of TDP_S and TDP_E for deep lake water samples, and (3) the equivalence of the results (Fig. 3a and b) for total unfiltered (TP) and filtered (TDP) P analyses for both the E and S protocols.

The lower values of SRP_S compared with SRP_E indicates more P-containing components other than ionic PO^3-₄ were converted (i.e., hydrolyzed) to PO^3-₄ in the E protocol than the S protocol. This paper does not present a position on which specific methodological differences between the 2 protocols (e.g., Table 2) caused the observed disparities in the SRP (and thereby SUP) results; that should be addressed elsewhere. Rather our focus was on the implications of such differences for limnological interpretation and utilizing contemporary mathematical models. It is noteworthy, however, that Tarapchak et al. (1982) reported that such method-specific differences in SRP results were not necessarily proportional to differences in the acid strength or exposure time of the protocols, but also varied with changes in the chemical composition of the SRP pool (e.g., sample origin). The differences in SRP_E and SUP_E from SRP_S and SUP_S probably reflect disparities in apportioning the diverse DOP pool between these 2 operationally defined forms of dissolved P (Baldwin 2013). In a recent review of organic P in the aquatic environment, Baldwin (2013) identified 5 important classes of dissolved organic P compounds: nucleic acids, other nucleotides, inositol phosphates, phospholipids, and phosphonates, some of which are known to have agricultural origins. SUP has often been equated to DOP, a position identified as flawed, in part because of the inclusion of certain of the forms of DOP in the SRP measurements (Baldwin 2013). Apparently more of the DOP pool was included within the SRP_E measurements than was included in SRP_S. The divergent SRP_S and SUP_S concentrations observed in Cayuga Lake and its tributaries, and elsewhere, represent a signature consistent with the widely held understanding of the metabolic sink effect of lacustrine primary production on DOP (Wetzel 2001) and the observed differences in the bioavailability of these 2 forms of P (Prestigiacomo et al. 2016).

Tributary levels of both SRP_S and SUP_S (represented as Q-weighted concentrations; Table 1) exceeded Cayuga Lake concentrations. The depletion of the SRP_S pool within the upper lake layers by early summer (Fig. 4a), despite tributary inputs, is consistent with the Reynolds (2006, p. 154) description of the SRP pool “as demonstrably bioexhaustible and freely bioavailable.” The higher SRP_S maintained in deep lake water relative to the upper waters (Fig. 2a) reflects the absence of photosynthesis-based depletion and likely some mobilization from the lake bottom during stratification (e.g., mussels; Prestigiacomo et al. 2016). Vertical oscillations of stratified layers from seiche activity (Effler et al. 2010) contribute to the irregular temporal dynamics in SRP_S (and SRP_E) and the other forms of P in these layers (Fig. 2). The higher SUP_S concentrations in the upper productive layers during summer compared with SRP_S (Fig. 4a) reflect the effects of (1) continuing tributary inputs of SUP_S, along with SRP_S (Prestigiacomo et al. 2016); (2) the diminished bioavailability of SUP_S relative to SRP_S (Table 1); and (3) the slower utilization kinetics of the bioavailable fraction of SUP_S (Auer et al. 1998). This dominance of SUP_S in the epilimnetic dissolved P pool in summer, as observed in Cayuga Lake (Fig. 4a), is widely observed (Fig. 4c; e.g., Wetzel 2001, Lampert and Sommer 2007). The dynamics of the epilimnetic SUP_S pool (Fig. 4a) can be viewed as reflecting temporal imbalances in sources (tributary inputs; Prestigiacomo et al. 2016) and sinks (enzymatic conversions of DOP toPO^3-₄; Currie et al. 1986, Bentzen et al. 1992). Decreases of SUP_S are observed when SRP_S is severely depleted, a temporal pattern attributed to the sustaining of phytoplankton growth by enzymatic conversions within the SUP_S pool (Connors et al. 1996 Effler et al. 2010).

The existence of a recalcitrant portion of SUP_S (P: 2–3 µg/L) in the lake was indicated by the return to the initial spring SUP_S concentration in the upper waters by fall (Fig. 4a), the temporal uniformity of SUP_S in the hypolimnion in 2013 (Fig. 2b and c), and the long-term uniformity of SUP_S in the hypolimnion since monitoring of deep lake water commenced in 1998. This position is supported by 2 features of the bioassay results: the incomplete bioavailability of SUP_S in the major tributaries, and the low bioavailability of SUP_S in the deep waters of the lake (f_BAP = 0.08 for LSC discharge; Table 1). The latter point is consistent with the extended exposure of the lake pool of SUP_S to enzymatic loss processes associated with the long water retention time in Cayuga Lake (Prestigiacomo et al. 2016). Estimates of the recalcitrant SUP_S concentrations of the tributaries (P: 1.4–2.5 µg/L), as the product of [1 − f_BAP] and SUP_S (i.e., the nonbioavailable fraction of SUP; Table 1), are generally consistent with the observed deep lake concentration (SUP_S = TDP_S − SRP_S; Fig. 2b and c). SUP_E underrepresents SUP.

Consistency with the bioavailability paradigm

The partitioning of the dissolved P pool by the S analytical protocol reasonably represents the distinct differences in bioavailability reported for SRP_S versus SUP_S (Table 1). Prestigiacomo et al. (2016) found both lake and tributary SRP_S to be nearly completely available on the short-term (Table 1), reflecting the utilization of PO^3-₄ and other forms of P that can be rapidly converted to PO^3-₄ (Baldwin 2013). SUP_S successfully differentiates those fractions of the dissolved P pool that either are not available (Table 1) or become available only over a longer time frame. SUP_S is imperfect as a measure of unavailable soluble P in Cayuga Lake (and no doubt in other lakes), however, because a uniform quantitative relationship with f_BAP does not prevail. The data for the 4 Cayuga Lake tributaries (Table 1) suggest a positive dependence of f_BAP on tributary SUP_S concentration.

Different conditions are suggested by the E protocol results. The observed approach of SRP_E to TDP (i.e., SUP_E approaches zero) in both deep lake waters and tributaries indicates that the E protocol yields greater inconsistencies with the bioavailability paradigm. Despite the general expectation in limnological studies that SRP represents P readily available for phytoplankton uptake, it seems SRP_E would not be completely available because a substantial portion of SUP_S (effectively included in SRP_E) is not available (Table 1). Moreover, this effective inclusion of SUP_S in SRP_E would preclude a differentiation of short-term and longer-term available fractions (within different kinetics) based on the E protocol. External loading estimates of SRP_E would overestimate the loading of dissolved bioavailable P (particularly short-term) to the lake (Prestigiacomo et al. 2016) from both the tributaries and the LSC facility. In studies where no TDP measurements are made (i.e., neither SUP_S nor SUP_E: not recommended), however, SRP_E would be a more useful measure than SRP_S because it would more closely approach the total load of dissolved bioavailable P.

Implications for mechanistic P-eutrophication modeling

Mechanistic mass balance P-phytoplankton (or eutrophication) models are widely used to guide management deliberations for systems with excessive eutrophication and are critical tools supporting P management analyses (Cooke et al. 2005). The “mechanistic” descriptor described earlier implicitly reflects an effort to utilize realistic process-based representations of lacustrine systems in model structure. A broad range of model complexity (e.g., number of processes and interactions considered) has been adopted in contemporary modeling efforts (Chapra 1997, Arhonditsis and Brett 2004, Robson 2014), most of which represent the patterns of the various operationally defined forms of P addressed in this paper. Despite myriad differences in structural detail, contemporary models generally share certain converging features, including representation of the multiple forms of P. Whether partitioning of the dissolved P pools of the tributaries and lake is based on the E or the S analytical protocol would have important implications for a Cayuga Lake modeling initiative, as well as for similar efforts for other systems.

The bioavailability of P in external loads is increasingly incorporated in mechanistic modeling initiatives (Effler et al. 2012) and has been adopted for a Cayuga Lake modeling effort to avoid the gross overestimates using TP loads instead (Prestigiacomo et al. 2016). Systematic errors in these estimates inevitably lead to compensating misrepresentation of in-lake source and/or sink processes for P as part of model calibration. The focus on control of external loads as the primary management option to abate cultural eutrophication (Chapra 1997, Cooke et al. 2005) further establishes the need for representative loading estimates that account for bioavailability. SRP_E data for the lake inputs are problematic in that regard.

Representing the extent of P-limitation of algal growth is a fundamental feature of these mechanistic models, often described by the Monod relationship (Chapra 1997): (1)

where G(N) is the fraction of the maximum specific growth rate (with a value of 1 corresponding to nutrient saturation); K_p is the half-saturation coefficient, a measure of the algal community's efficiency in acquiring P; and SRP is the concentration of this operationally defined form of P, assumed to be entirely available for algal uptake and growth (other forms of P are assumed to not be immediately available). Substantial nutrient (P) limitation (i.e., low G(N) levels) prevails in most north temperate lacustrine systems (for trophic states less than hypereutrophy) in summer (Reynolds 2006). Most of the K_p values adopted in contemporary models fall in the range of 0.5–1.5 µg/L (P as SRP; Connors et al. 1996, Doerr et al. 1998, Tomlinson et al. 2010). A threshold of SRP ≤1 µg/L has been used to identify conditions that correspond to distinct P limitation of algal growth (Effler et al. 2010), supported by accompanying increases in the activity of the enzyme alkaline phosphatase (indicator of P limitation; Connors et al. 1996). Reynolds (2006) described SRP concentrations of >3 µg/L as “scarcely” limiting, which is generally consistent with the above range of K_P values from the contemporary literature and the hyperbolic dependency of G(N) on SRP described in equation 1(1) . The indicated systematically higher levels of SRP_E compared to SRP_S, that doubtless extend into the upper productive layers (e.g., Fig. 3d and 4b), could be problematic for representing the extent of P-limitation in the upper productive layers because it would depict (e.g., K_p = 0.5−1.5 µg/L; equation 1(1) ) only a minor, instead of substantial, degree of limitation.

The SUP pool is represented as a separate state variable in most contemporary P-eutrophication models (Arhonditsis et al. 2006, Robson 2014), that together with SRP corresponds to the overall dissolved P pool. In general, SUP (or a fraction of SUP) is represented as a source of SRP. The kinetics of the conversion of DOP to SRP are generally specified to be slow relative to the uptake of SRP by phytoplankton (Doerr et al. 1998). Accordingly, the DOP pool can, along with other recycling pathways, contribute to sustained phytoplankton growth under severely depleted SRP conditions (Connors et al. 1996, Effler et al. 2010). The apportionment of TDP_S according to SRP_S and SUP_S can support such model representations of the P cycle. A shift to higher SRP and lower SUP levels that would occur using the E protocol would necessarily change this representation. Accordingly, in the context of contemporary P-eutrophication models (but not the older models; e.g., Vollenweider 1976), the selection between these 2 analytical protocols for P has the unfortunate likely effect of influencing the model structure choices that can be implemented and perhaps the associated level of realism represented.

The extent to which SRP_E and SUP_E datasets would influence the success of a P-eutrophication modeling initiative is uncertain. At the least, the differences between SRP_E and SRP_S reported here would require different values for model coefficients (model parameterization describing the P cycle), and perhaps differences in model structure, to achieve calibration using respective datasets from the 2 protocols. For example, use of an SRP_E dataset would require an increase in K_p (equation 1(1) ) to maintain appropriate P-limitation of algal growth and higher transformation rates of the (smaller) DOP (SUP_E) pool to SRP_E. Such analytical protocol-driven differences in model parameterization would translate to an undesirable increase in the degrees of freedom of related mechanistic modeling initiatives. Great care is needed to recognize which measurement protocols have been adopted in datasets to be used in contemporary P-eutrophication modeling initiatives and in related research to be drawn upon from the literature. Limnological and modeling expertise in P-eutrophication should be included in the data collection parts of the programs used for development and testing of related management modeling.

Acknowledgments

This is contribution no. 326 of the Upstate Freshwater Institute.

Funding

Funding for this work was provided by Cornell University.