Amounts, forms, and management of nitrogen and phosphorus export from agricultural peatlands

Peatlands provide a setting that is well suited for cranberry agriculture in the Northeastern United States. However, misconceptions exist about the amounts and forms of nitrogen (N) and phosphorus (P) export from cranberry farms. In this study, we report inorganic and organic forms of N and P export from five peatlands cultivated for cranberry production in southeastern, Massachusetts, United States. We then compare N loading rates among cranberry farms in southeastern Massachusetts, row crop farms in the Midwestern United States, and uncultivated peatlands in the United States and United Kingdom. Based on a fluvial mass balance analysis, we find that nonriparian cranberry farms export 2.56 kg of P ha−1 year−1of total P and 12.1 kg of N ha−1 year−1of total N. Total N export from riparian or “flow through” farms is two times higher than nonriparian farms due to less retention of N fertilizer in the vadose zone of riparian farms. Gross total N export from riparian and nonriparian cranberry farms consists of 35% particulate organic N, 26% dissolved organic N, 31% ammonium (NH4+), and 8% nitrate (NO3−). The low proportions of NO3− export (13% of total dissolved N [TDN]) for cranberry farms differ from NO3− export for row crop farms (75% of TDN; p < .001) but not for uncultivated peatlands (17% of TDN; p = .61). Despite being highly modified by fertilizers and artificial drainage, low NO3− export (2.2 kg of N ha−1 year−1) from cranberry farms is consistent with field measurements of rapid N turnover in uncultivated peatlands. This finding suggests that state‐funded wetland restoration efforts to restore denitrification in retired cranberry farms may be limited by NO3− rather than soil moisture or organic matter.

The mechanisms that control nutrient transfers from cranberry farms to surface water have not been rigorously evaluated, but some general trends have emerged to describe the wide variability in nutrient losses from cranberry farms. Cranberry farms can be classified based on their connection to surface water: riparian ("flow through") farms, which Hoekstra, Neill, and Kennedy (2020) described as having a large central channel (usually meandering) with farm units on either side of the channel, and nonriparian farms, which are relatively isolated from surface water and include upland farms. Riparian cranberry farms export up to three times the N and P from nonriparian cranberry farms (MA DEP, 2009b;Howes et al., 2014), although continuously flowing nonriparian farms appear to export similar amounts of N compared with riparian farms (Howes & Teal, 1995;Neill, Jakuba, Kennedy, & DeMoranville, 2017). The seasonal timing of nutrient export from cranberry farms also varies between riparian and nonriparian farms: Howes and Teal (1995) observed high N and P losses from a riparian cranberry farm that were attributed to the manipulation of floodwaters, while Neill et al. (2017) reported 55% to 77% of gross N and P export from three nonriparian farms during nonflood periods. In a multi-year study, Kennedy et al. (2018a) observed elevated N and P export from a nonriparian cranberry farm when extreme summer rainfall (111 mm day −1 ) coincided with fertilizer application.
As is the case for most peatlands, N export from cranberry farms is governed by the abiotic and biotic processes that produce, consume, and transport N in the unsaturated zone. Cranberry farms are perhaps best described as "cultivated peatlands" (Howes & Teal, 1995;Kennedy et al., 2018b), which consist of anthropogenic sand overlying peat. The sand layer is generally acidic (mean pH = 4.5; Davenport et al., 2003) Ballentine et al., 2017). By comparison, uncultivated peatlands are also acidic (pH < 5) and exhibit relatively high rates of biological uptake of inorganic N (Westbrook & Devito, 2004). As a consequence, N export from peatlands is generally high in organic N relative to inorganic N (Hemond, 1980(Hemond, , 1983Verry & Timmons, 1982;Worrall, Clay, & Burt, 2012;Hill et al., 2016).
To better understand the processes of nutrient transport in cultivated peatlands, we quantified the amounts and forms of N and P export from a large, 20.2-ha cranberry farm in Plymouth, Massachusetts, which was implicated in a 2010 P total maximum daily load (MA DEP 2009a). We then compiled N loading rates from empirical budget studies for cranberry farms, as well as for croplands and peatlands. We hypothesize that the major N inputs to cranberry farms are similar to those for row crop farms, but that the forms of N outputs differ between these two types of agriculture. In terms of nutrient management, the objective of this study was to refine cranberry farm N and P loading rates and by doing so, elucidate the processes that supply, store, and transport N and P in cranberry farms.

| Study area
The study area is located in the upper Buttermilk Bay watershed on a commercial cranberry farm 13 km south of Plymouth (Howes et al., 2014; Figure 1). Local climate is typical of New England with mild summers (mean July temperature of 22 C), cold winters (mean January temperature of −2 C), and moderate precipitation ( Caron et al., 2017;Kennedy et al. 2017a; Figure 1).
There is considerable interest in defining cranberry farms based on their hydrologic and edaphic properties (Hoekstra et al., 2020;Howes et al., 2014). Currently, cranberry farms are classified as wetland or upland farms using soil maps (Kennedy et al., 2018b) and as riparian or nonriparian farms using remote sensing techniques (Hoekstra et al., 2020). Within wetland farms, subclasses include marshes and fens or bogs, collectively referred to as peatlands (Mitsch & Gosselink, 2007  Wareham, Massachusetts. Errors in annual precipitation data were about 5%, on average, but as high as 20% seasonally (Legates & DeLiberty, 1993;Winter, 1981 Heiner and Vermeyen, 2012), whereas that in the winter flood was estimated as the coefficient of variation about the 3-year mean winter flood (±39%; Table 2).

| Hydrometric measurements
Hydrologic outputs included fluxes to surface water, groundwater, and the atmosphere. We used a propeller-type flow meter (McCrometer, Helmet, California) to measure surface water discharges, which were recorded, on average, every 4 days. A notable exception was in late October, when the farm was flooded for the cranberry harvest. Discharge flows were recorded less frequently during the harvest when the pump was off but nearly daily as floodwaters were pumped to the upland leaching field (e.g., Kennedy et al., 2015).

| Sample collection
During the study, the main sources for irrigation water were a lake (i.e., White Island Pond), a retention pond, and a shallow well, whereas the sole source for floodwater was the lake (Figure 1). We collected monthly samples from the lake and retention pond by manually dipping a sample bottle at roughly the depth of the pump intake screen.
For floodwater, we collected three lake water samples during the harvest flood and another three samples prior to the winter flood, based on the relatively low nutrient variation of the incoming floodwaters (Kennedy, 2019;Kennedy et al., 2015). We collected samples of surface water discharge as surface (<15-cm depth) grab samples on a weekly basis, with more frequent sampling during flood discharges (66 samples in total).

| Analytical methods
We collected filtered and unfiltered water samples in 60-ml polyethylene bottles using a plastic syringe and pre-rinsed (0.45 μm) cellulose membrane filters. Total P (TP) and total N (TN) were determined on unfiltered samples, whereas total dissolved P (TDP),

| Data analysis
We quantified annual groundwater exchanges with the cranberry farm as the residual term in a water budget equation for a wetland (Kennedy, 2015;Mitsch & Gosselink, 2007): where annual volumes of surface water discharge (Q sw ), ET, precipitation (Q p ), irrigation (Q i ), floodwater (Q f ), and storage change (ΔS) were used to solve for net groundwater exchanges (Q gw ). Methods for Q sw , Q i , and Q f were described in the previous section, as were daily mean values of Q p and ET, which were aggregated and multiplied by the farm area (20.2 ha) to determine volumetric fluxes (Table 1). Annual ΔS was calculated as in Kennedy et al. (2018a), and because it was generally low (<5%), was set to zero for water budget calculations.
Assuming steady state with respect to changes in nutrient storage and negligible nutrient loss from the underlying peat layer, a mass balance equation was applied to calculate net TP and TN fluxes from the cranberry farm. We express the nutrient load in surface water discharge (F sw ) as a function of nutrient inputs to the farm: where F fert , F f , and F i are nutrient inputs of fertilizer, floodwater, and irrigation, respectively, and the term c y x is the fraction of the nutrient input x that is not lost at the surface (y = s) and in the vadose zone (y = vz) of the cranberry farm (i.e., the amount of F fert , F f , F i , and F atm that contributes to F sw ). Groundwater flows (Q gw ) and nutrient fluxes (F gw ) may represent an input or output, yet we show in Section 3.1 that Q gw is a minor (3%) output from the cranberry farm. As such, F gw was set to zero in Equation (3).
In the case of TP, we assumed c s f c vz f = c s i c vz i = c s atm c vz atm = 1, as in previous studies (Neill et al., 2017;Kennedy et al., 2018a) & Valiela, 1995;Valiela et al., 1997) and assumed negligible retention of particulate organic N (PON; c s f,i c vz f,i = 1). Using these nutrient loss coefficients and the relative amounts of DON, DIN, and PON in floodwater and irrigation (Neill et al., 2017; this study), we calculated mean values of c s f c vz f = 0:65 and c s i c vz i = 0:78 for TN. Assuming c vz fert = c vz atm , as in Valiela et al. (1997), we solved for site-specific values of c vz atm and c vz fert in Equation 3. The net flux term, F net , which represents N fertilizer export from the cranberry farm, was then calculated as We calculated values of F x as the product of volumetric discharge (Q x ) and nutrient concentration (C):  were particularly sensitive to Q gw for the cranberry farm. Although incomplete release of the winter flood seems possible, it is more likely that positive differences in Q gw + ΔS for April and May were due to regional patterns of elevated spring groundwater discharge to the cranberry farm (Li, Rodell, & Famiglietti, 2015). In contrast, the highly negative difference in Q gw + ΔS for October was consistent with harvest floodwaters as sources of groundwater recharge (Kennedy, 2015;Kennedy, 2019;Masterson, Carlson, & Walter, 2009).

| Nutrient concentrations
Nitrogen concentrations in hydrologic inputs were generally lower than those in surface water discharge, with the exception of particulate forms of N and P (    Phosphorus concentrations in hydrological inputs were highly variable with 5-17 times higher TP concentrations in irrigation water than incoming floodwaters (Table 4). TP concentrations of hydrological inputs were predominantly composed of organic P (85-95%, DOP + TPP), with small but significant levels of DRP (5-15%). Mean TP and DRP concentrations of surface water discharge were 2 and 7 times higher than hydrological inputs, respectively. These results aligned with those from focused investigations of the winter and harvest floods, which showed higher TP concentrations in surface water discharge compared with incoming floodwaters (Kennedy, 2019;Kennedy et al., 2015).
Seasonally, the highest concentrations of P, particularly DRP, were observed in autumn and associated with the cranberry harvest ( Figure 3). The pattern of increasing concentration of DRP with discharge of the harvest flood was consistent with the physical leaching of P released from anoxic sediments to soil pore water, which is a well-documented mechanism in both wetland and agricultural environments (Patrick & Mahapatra, 1968 (Hemond, 1980). Monthly values are given except for the winter months from December to March et al., 2015). Summer fertilizer additions also appeared to increase P concentrations in surface water (Figure 3). For instance, the July fifth "spike" in TP concentration was composed of 36% DRP, 35% TPP, and 29% DOP. These forms of P are somewhat at odds with the application of inorganic P fertilizers, possibly due to rapid sorption and biological uptake of phosphate in agricultural soils (Jarvie et al., 2005;McDowell, Sharpley, & Folmar, 2003). Nevertheless, summer peaks of P and N concentrations in surface water occurred when fertilizers were commonly applied to the cranberry farm (Figure 3).

| Nutrient loads
Annual TN loads in hydrological inputs were 101.6, 30.7, 23.8, and Within season, N forms in surface water loads were highly variable, with DON and NH 4 + collectively representing 82% of summer and autumn TN export and DON and PON accounting for about half of winter TN export ( Figure 5).
Seasonal patterns in TN export could be broadly tied to climate variation and regional trends in agricultural management ( Figure 6).
We observed a generally weak relationship between seasonal TN export and surface water discharge, with the highest value of TN export occurring in the driest season (R 2 = .13, p = .64; N = 4). However, when seasonal values of TN export were regressed against the respective TN concentrations, a more significant linear relationship was observed (R 2 = .77; p = .12; N = 4). A first-order interpretation of these results is that seasonal differences in TN export are tied to increases in N pools that are susceptible to transport. Increases in more mobile forms of N may be related to warmer summer temperatures that facilitate N transformation processes in the unsaturated zone, particularly mineralization of organic N to NH 4 + (Davenport & DeMoranville, 2004). Given the timing of fertilizer applications in July, which alone accounted for 32% of the annual TN load export, fertilizer management was also likely a source of elevated N loss  Irrigation  16  91  15  199  164  468  39  79  136  254   Harvest flood  2  102  13  61  29  206  2  5  9  15   Winter flood  3  31  22  88  23  164  2  45  7  54   Surface water  66  252  31  224  91  598  96  96  62  254 Abbreviations: DON, dissolved organic N; DOP, dissolved organic P; DRP, dissolved reactive P; N, sample size; n.a., not applicable; n.d., not determined; PON, particulate organic N; TN, total N; TP, total P; TPP, total particulate P. a Calculated as the modelled atmospheric deposition (see text) divided by the measured annual precipitation.
F I G U R E 5 Nutrient forms of hydrological inputs (precipitation, winter flood, harvest flood, and irrigation) and outputs (surface water) for the cranberry farm. Total P maximum daily load target shown (dotted line). DON, dissolved organic N; DOP, dissolved organic P; DRP, dissolved reactive P; PON, particulate organic N; TDP, total dissolved P; TP, total P; TPP, total particulate P ( Figure 6). High autumn N export was consistent with seasonal N losses from a riparian cranberry farm (Figure 6), which were linked to enhanced foliar leaching, disturbance, and anoxic release of N during the harvest flood in October (Howes & Teal, 1995). Notably, 38% of the annual DON export occurred in October, possibly as a consequence of the cranberry harvest.
Hydrological inputs of TP were predominantly derived from irrigation water, which accounted for 72% of all P inputs to the cranberry farm. In contrast, Kennedy et al. (2018a) observed that P inputs were mostly derived from incoming floodwaters, which were transferred from adjacent cranberry fields rather than a lake. In total, hydrological inputs added 28.26 kg of P year −1 to the cranberry farm as 46% TPP, 29% DOP, and 25% DRP. Surface water export of 89.64 kg of P year −1 was about three times higher than hydrological inputs of TP, indicating leaching of soil P to surface water. Of this amount, 39% was DOP, 37% was DRP, and 24% was TPP. In general, monthly variation in TP export from the cranberry farm closely mirrored that for TN, with the exception of the months of July and October (Figure 7), which may be related to higher fertilizer additions of N compared with P in July and greater anaerobic dissolution of P than N compounds in October (Table 1). Thus, the factors that contributed to TN export generally appeared to have a similar effect on TP, with the exceptions of fertilizer application and the harvest flood ( Figure 6).  (Table 5), which is consistent with field observations of low soil NO 3 − in cranberry farms (Stackpoole et al., 2008;Ballentine et al., 2017). In most croplands, NO 3 − is the dominant form of TDN export (Pellerin et al., 2006), in part because it is the most mobile form of N in the environment (Böhlke, 2002). In a study of four agricultural watersheds within the Mississippi River Basin, mean TDN export was 75% NO 3 − , 22% DON, and 3% NH 4 + (Goolsby & Battaglin, 2001), which differed from cranberry farms for NO 3 − and NH 4 + but not for DON ( Figure 8). In the Midwest, about 90% of synthetic fertilizers are applied as NH 4 + (Cao, Lu, & Yu, 2018), indicating high rates of nitrification relative to NH 4 + transport through the unsaturated zone (Böhlke, 2002).

| Forms of dissolved nitrogen export
In contrast, low rates of net nitrification have been observed in peatlands (Devito, Westbrook, & Schiff, 1999;Hayden & Ross, 2005;Roswall & Granhall, 1980), at least partly due to rapid uptake of NO 3 − (Westbrook & Devito, 2004). For example, mean TDN export was 58% DON, 25% NH 4 + , and 17% NO 3 − for peatlands in the United States and United Kingdom (Verry & Timmons, 1982;Hemond, 1980Hemond, , 1983Worrall et al., 2012;Hill et al., 2016), suggesting minimal production or rapid consumption of NO 3 − F I G U R E 6 Monthly variation in the forms of N and P export from the cranberry farm (1 May 2017 to 30 April 2018). DON, dissolved organic N; DOP, dissolved organic P; DRP, dissolved reactive P; PON, particulate organic N; TPP, total particulate P F I G U R E 7 Relationship between monthly total N (TN) and P (TP) export from the cranberry farm excluding July outlier, which was determined as values of Cook's distance, D, greater than 4/n = 0.33 (Cook, 1977) compared with row crop farms such as wheat, corn, and soybeans ( Figure 8). Although N transformation processes are still poorly understood in cranberry farms, our results suggest that they may resemble those in peatlands, despite the management effects of fertilization and artificial drainage.

| Management implications
Accurate estimates of N loading rates from cranberry agriculture are essential for comprehensive N management plans to protect coastal water quality in southeastern Massachusetts (Howes et al., 2014;Williamson et al., 2017). We calculated TN values of F net that ranged from 6.8 to 23.9 kg of N ha −1 year −1 (mean = 14.0 kg of N ha −1 year −1 ; N = 6), or were equal to 23.9 kg of N ha −1 year −1 for riparian cranberry farms and 12.1 kg of N ha −1 year −1 for nonriparian cranberry farms (mean; N = 5). We found that the two highest values of TN export were related to less retention of N fertilizer in the vadose zone of riparian cranberry farms or nonriparian cranberry farms that receive high groundwater inputs ( F I G U R E 8 Forms of TDN export from cranberry farms, row crop farms, and natural peatlands. Literature data were used to calculate mean values for cranberry farms (N = 5; Howes & Teal, 1995;Neill et al., 2017; this study), peatlands (N = 5; Verry & Timmons, 1982;Hemond, 1980Hemond, , 1983Worrall et al., 2012;Hill et al., 2016), and row crop farms (N = 4; Goolsby & Battaglin, 2001). The peatland soil flux was used from Worrall et al. 2012, and negligible particulate N was assumed for data from Verry and Timmons 1982, Hemond 1980,1983, and Hill et al. 2016. Concentration data (1980-1998 was used from Goolsby and Battaglin, 2001; all other data were in units of kg ha −1 yr −1 . DON, dissolved organic N; TDN, total dissolved N especially those developed for the major agricultural watersheds in southeastern Massachusetts (i.e., Sippican River, Weweantic River, and Wareham River), classify cranberry farm N loading rates based on riparian and nonriparian hydrology (Hoekstra et al., 2020) and the extent of groundwater exchanges in nonriparian farms (Neill et al., 2017).
Unlike most agricultural crops, cranberry farms export high amounts of DON and NH 4 + relative to NO 3 − (Figure 8). Recently, a state-funded programme was established to support wetland restoration of retired cranberry farms (MA DER, 2019). The programme was motivated, in part, by the potential to enhance denitrification through wetland restoration of retired cranberry farms (Ballentine et al., 2017). However, the amount of NO 3 − removed through wetland restoration of retired cranberry farms has not been quantified.
As a first-order estimate, we calculated gross NO 3 − export from five cranberry farms (i.e., F sw in Equation 4), with values ranging from 0.2 to 6.2 kg of N ha −1 year −1 (mean ± SD = 2.2 ± 2.5 kg of N ha −1 year −1 ; Neill et al., 2017;Howes & Teal, 1995; Table 5). Although the microbial capacity to consume NO 3 − may limit the amount of NO 3 − available for denitrification (Devito and Westbrook, 2004;Stark & Hart, 1997), as suggested by low NO 3 − export, some cranberry farms represent seepage faces for high-NO 3 − groundwater inflows, particularly riparian cranberry farms located in low-lying areas of watersheds where septic effluent is a major N input to groundwater (e.g., Coonamessett Bogs on Cape Cod). As such, statefunded wetland restoration projects should target retired riparian cranberry farms that receive high groundwater inputs of NO 3 − , which appears to be the principal limiting factor of denitrification in these peatland systems.
In addition to coastal watershed N management plans, P regulations have been implemented to decrease agricultural P loadings to White Island Pond by 86% (MA DEP 2009a), with our estimate of For the month of June, we measured TP export of 8.8-kg P from the cranberry farm ( Figure 6). Although this does not account for thẽ 6-hr June discharge to the lake, it does suggest that most June discharges from the cranberry farm are below the TMDL allocation of 10-kg P year −1 (MA DEP 2009a), including those that occurred in 2014 and 2017.
Wet (flood) harvesting is practiced by~90% of cranberry growers in Massachusetts (DeMoranville, 2008), but is also a significant source of P loss from cranberry farms. In this study, we estimate that wet harvesting contributes up to 32% of the annual TP load in surface water. Although growers are unlikely to replace wet harvesting, cranberry production is uniquely adapted to remedial strategies along the lines of containment, treatment, and trapping (Aslan & Cakici, 2007;Blowes, Robertson, Ptacek, & Merkley, 1994). Given that about two thirds of the harvest flood P loss was in the form of organic P (Figure 5), surface applications of Psorbing materials that promote flocculation, such as aluminium sulfate, are well suited to sequester P from the harvest flood (Kennedy et al., 2017b). In addition, the iron-rich soils of cranberry farms may function as natural filter beds, depending on the P concentration and infiltration rate of the harvest flood (Kennedy, 2019).

| CONCLUSIONS
We find that cranberry farms export significantly less NO 3 − than would be expected from agricultural crops that receive regular additions of N fertilizers. In uncultivated peatlands, low NO 3 − export has been connected to rapid biological consumption of NO 3 − and NH 4 + (Westbrook & Devito, 2004), which would also serve to reduce nitrification rates. We believe similar soil biogeochemical processes may occur in cranberry farms, but detailed studies on N cycling in cranberry farms are lacking. Field measurements of gross nitrification and mineralization rates will fill a critical gap in our understanding of the fate of N in cranberry farms. Hoekstra improved this paper.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.