Winter Phosphorus Release from Cover Crops and Linkages with Runoff Chemistry

Cover crops (CC) have both agronomic and environmental benefits but also have the potential to increase losses of dissolved reactive P after freeze–thaw cycles (FTC). This field study, conducted over one nongrowing season (NGS) in Ontario, Canada, characterized water-extractable P (WEP) content in different CC species and compared observed changes in plant WEP content with changes in P content in soil, surface runoff, and shallow groundwater (5–25 cm). Five plots (0.4 ha) of cereal rye (Secale cereal L.), oilseed radish (Raphanus sativus L. var. oleoferus Metzg Stokes), oat (Avena sativa L.), and hairy vetch (Vicia villosa Roth) were established after winter wheat (Triticum aestivum L.) harvest. Throughout the NGS (October–April), CC shoot tissues and surface soil were routinely sampled for WEP analyses, and groundwater and runoff water samples were collected after rain and snowmelt. Responses to FTC varied among CC species, with P released from frost-intolerant species but not frost-tolerant species. Although CC released P, the top 5 cm of soil contained greater WEP than plants at all times, and the changing WEP content in CC over the NGS was not reflected in soil or water P concentrations. These results suggest that the degree of frost exposure should be considered in the selection of CC species in cold regions; however, in temperate regions with snow cover that insulates the soil surface from heavy frost, P release from vegetation may not lead to increased P loss in runoff. Winter Phosphorus Release from Cover Crops and Linkages with Runoff Chemistry

America.In a recent laboratory study, Cober et al. (2018) found that P release from CC was significantly greater after harsh frost than after a light frost and suggested that the risk of winter P loss from CC may be reduced in regions that do not experience harsh winter frosts or that have significant snowpack to insulate plants.This is supported by the work of Roberson et al. (2007) and Øgaard (2015), who observed greater P loss in laboratory trials compared with CC exposed to FTC in situ.To better inform regional management recommendations, an improved understanding of winter P release from CC under field NGS conditions in temperate climates is needed.
A limited number of studies have quantified the impacts of vegetation P loss on P loads in runoff.In the Northern Great Plains of North America, Liu et al. (2014b) observed elevated P loss from fields with CC, primarily in overland flow during the spring snowmelt period.In the more temperate Great Lakes region, Lozier et al. (2017) found that edge-of-field P losses were small despite a large P supply in CC, demonstrating the potential for the retention of vegetation P by soils.The potential for extracted P to be held within soils has been well demonstrated, particularly when leachate is able to infiltrate into the soil matrix, as opposed to moving overland (Sharpley, 1995;Bechmann et al., 2005;Riddle and Bergstrom, 2013).Thus, P export may be minimized if leaching from frozen CC occurs before the soil is frozen, permitting infiltration, or if P in overland flow has sufficient contact with surface soil, allowing adsorption to occur (Roberson et al., 2007;Lozier et al., 2017).To evaluate the efficacy of CC as a BMP for the mitigation of P loss, the simultaneous examination of plant and soil water-extractable P (WEP) contents and P concentrations in surface and subsurface runoff in a field setting is needed.
The objectives of this study were (i) to determine how WEP and plant total P (TP) contents in the field change throughout the NGS and if this varies among CC species; and (ii) to examine P concentrations in soil, surface runoff, soil water (leachate), and shallow groundwater throughout the NGS and determine if changes in water chemistry coincide with changes in the vegetation WEP concentrations.

Study Site and Experimental Treatments
This study was conducted in a field located in Ontario, Canada (43°30¢30¢¢ N, 80°28¢12¢¢ W).Climate in the region is classified as warm-summer, humid continental (Dfb, Köppen Classification), with the mean annual maxima in July (20°C) and mean annual minima in January (−7°C).The region receives ?780 mm of precipitation annually, including 160 cm as snowfall between November and March (ECCC, 2018).The field is under a corn (Zea mays L.)-soybean [Glycine max (L.) Merr.]winter wheat (Triticum aestivum L.) rotation on gray brown luvisols (light Burford loam [fine-silty, mixed, superactive, thermic Typic Haplustepts]-London loam [imperfectly drained loam with sandy layers in the A and B horizons, and calcareous C horizons with gravel or stony lenses between the B and C horizons] series) (Presant and Wicklund, 1971) with a slope of 4%.Concentrations of Olsen P in the top 15 cm of soil were 26.5 ± 8.3 mg kg −1 .The field is generally no till, although periodic shallow vertical tillage (5 cm) is done when crop residues are heavy.
The field was not tilled prior to the planting of the CC.Fertilizer was applied via subsurface placement with winter wheat (10 kg N, 58 kg P 2 O 5 , and 67 kg K 2 O ha −1 ) in fall 2016, with a broadcast application of 134 kg N ha −1 in spring 2016.No fertilizer was applied prior to the initiation of the study in August 2017.
The ?5-ha field contained five plots, each ?0.4 ha in size (Supplemental Fig. S1).The CC plots were planted on 9 Aug. 2017, after winter wheat harvest.The plots consisted of cereal rye (Secale cereal L.) seeded at 67 kg ha −1 , oilseed radish (Raphanus sativus L. var.oleoferus Metzg Stokes) seeded at 8 kg ha −1 , oat (Avena sativa L.) seeded at 45 kg ha −1 , hairy vetch (Vicia villosa Roth) seeded at 22 kg ha −1 , and a control plot that contained five strips of volunteer wheat growth, which is typical of wheat production in North America.By late October, the volunteer wheat covered roughly 30% of the plot area.

Plant and Soil Sample Collection and Processing
Plant biomass for each CC was measured once during the field season (12 Oct. 2016).This date was chosen to achieve maximize growth prior to a killing frost.Three 0.25-m 2 quadrats were clipped from each plot.Only plant shoot tissue was collected, with the exception of oilseed radish, for which protruding root tissue was also collected.Oilseed radish roots were weighed separately from the shoot tissue.A subset of each sample was dried at 100°C for 24 h to determine water content, and all biomass results are presented as dry weight.
Plant and soil samples were collected at 2-wk intervals between September and December, and approximately monthly once the snow pack developed (December-April).No soil samples were collected during the January sampling campaign due to frozen ground that prevented sample collection.Plant samples were collected by clipping the entire shoot tissue at the surface, with the exception of oilseed radish, which was separated by shoot and root tissue.For each sampling event, 10 to 20 individual plants were clipped throughout the plot and batched together to homogenize samples before extraction.Plant samples were kept in coolers with ice packs and extracted within 12 h of collection.Shoot tissues were clipped into ?4-cm-longpieces to fit into sample containers.Oilseed radish root samples were treated differently because of the substantial size; cylindrical pieces of root (?1 cm in length) were cut and extracted individually.Living volunteer wheat from the control plot was collected and extracted, but no wheat residue from the previous crop was collected.Approximately 5 g of field-moist material was shaken in 50 mL deionized water for 1 h at 100 rpm, after which leachate was immediately gravity filtered through Whatman No. 42 filter papers (pore size = 2.5 mm) and promptly frozen for the subsequent analysis of WEP.A small subset of tissue from each CC species and sampling event was dried at 100°C for 24 h to determine water and TP content.
Three soil samples (0-5 cm), each consisting of three composite soil cores, were collected at random locations across each plot (located in close proximity to where CC had been sampled) on each sampling date for the determination of WEP concentrations.Soil samples were oven dried at 30°C before extraction.Following a slightly modified method from Kovar and Pierzynski (2009), 50 mL of deionized water was added to 10 g of dried soil and extracted using the same methods used for the plant material.Additional soil samples were taken in October 2017 (when mature CC are typically exposed to frost) for the determination of bulk density using standard techniques (12-cm-diam.cores to 5-cm depth).In April 2017 (just prior to CC termination), six composite soil samples were collected (15 cores per sample) from 0to 15-cm depth, and their Olsen P content (the standard soil test for the region given the high-pH soils) was analyzed using sodium bicarbonate extraction and ammonium molybdate colorimetric methods (Olsen et al., 1954;A & L Canada Laboratories).

Environmental Data and Water Sample Collection and Processing
Probes were deployed 20 Sept. 2016 in the field to monitor soil temperature and air temperature (1 m above the surface) (Decagon Devices 5TMSM/Temp sensors with an Onset EM-50 data logger).Temperature probes were installed in each of the five plots at 5-cm soil depth, with two additional probes installed at 1-and 5-cm depth in one plot (Supplemental Fig. S1).Estimates of precipitation and snow cover were taken from a nearby monitoring station (9.5 km from the site, Climate Identifier 6149389; ECCC, 2018) as a result of complications with the onsite gauge.
Water sample collection units were installed to capture surface runoff and shallow groundwater samples (5-to 25-cm depth) for comparison with spatial patterns in soil and vegetation P concentrations.Overland flow was collected using the approach of Daniels and Gilliam (1996) and Dunn et al. (2011), using weirs that directed flow into partially buried 1-L Nalgene bottles equipped with ping-pong balls to seal lid openings (3 cm) when bottles were full.Five overland flow collectors were installed in each CC plot.Because the field was sloped, and surface runoff typically does not occur at the plot scale on sloping ground, surface runoff volume was not quantified.Cover crop plots were set up in the field to take advantage of the field specific topography, which directed runoff toward a low point that ran across the center of the field.This permitted the collection of overland flow that was unique to the plots and minimized opportunities for runoff to flow between adjacent plots.A limitation of this site-specific topography, however, was that during some small events that generated overland flow, unique samples could not be collected within individual plots because water was restricted to the central zone, which received runoff from all plots.Shallow groundwater was sampled at five nests of piezometers (screen depths of 5-15 cm, 15-25 cm per nest) in each CC plot.Piezometers were constructed from acrylonitrile butadiene styrene (ABS) pipe (3.8cm i.d.) triple wrapped with fiberglass screening.Each piezometer had a nonperforated 200-mL reservoir below the screening depth to collect samples during periods when the water table rose and fell rapidly between field visits.Piezometers were purged two to three times before samples were collected using a peristaltic pump to minimize the impacts of disturbance from well installation on water samples.In the control plot, all piezometers were installed in areas where no volunteer wheat was present.
All water samples were collected within 24 h of a storm event ending, and after sample collection, each collector was purged.A storm was deemed to have occurred when rainfall produced a hydrologic response and water samples could be collected from the piezometers or surface runoff collectors (in general, a minimum threshold of 5 mm was required during the NGS).A 50-mL subsample of each sample was filtered through a 0.45-mm cellulose acetate filter within 24 h of collection.Filtered samples were immediately frozen prior to analysis for DRP.In the surface runoff samples, an unfiltered 50-mL subsample was acidified with H 2 SO 4 (to 0.2% final H 2 SO 4 concentration) prior to digestion for TP analysis.

Laboratory Analysis
For the determination of plant TP content, plant digestions were completed using H 2 SO 4 -H 2 O 2 -Li 2 SO 4-Se methods (Parkinson and Allen, 1975).For the determination of TP in overland flow, water samples were digested using acid persulfate digestions and an autoclave (EPA/600/R-93/100, Method 365.1;Jeffries et al., 1979).Phosphorus concentrations from plant and soil extractions, plant digestions, and filtered water samples were analyzed in the Biogeochemistry Laboratory at the University of Waterloo using ammonium molybdate-ascorbic acid colorimetric methods (Bran-Luebbe AutoAnalyzer III system, Seal Analytical, Methods G-103-93 [DRP] and G-188-097 [TP]).

Statistical Analysis
A series of two-way ANOVA tests were done to determine if plant or soil WEP concentrations varied with CC species and/ or time.Where data were not normally distributed, data were log-transformed or nonparametric equivalent tests (Scheirer-Ray-Hare test) were used.Where significant interactions existed between the two factors, data were split and one-way ANOVAs were run on each level of the factors.A Pearson product moment correlation was used to compare median DRP concentrations in water samples with median plant and soil WEP from samples taken prior to each water sampling event to evaluate the potential relationship between each of the potential P concentrations and runoff P concentrations.Medians were used because the soil and plant samples were composites.All statistical tests were done with an a value of 0.05 to determine significance.

Weather Conditions over the Study Period
Weather conditions over the study period were divided into distinct periods based on air temperatures (early, mid, late; Fig. 1).Little precipitation preceded the "early" period (October-December), which was characterized by dry conditions and generally small precipitation events (with the exception of a single large rain event in November).Although temperatures fell below freezing during this period, temperatures remained close to zero with the exception of two nights when temperatures fell between −5 and −10°C in October and November.The "mid" period commenced with the accumulation of a snowpack (mid-December).Throughout this period, air temperatures dropped as low as −20°C, but surface soils (1-cm depth) were insulated from freezing temperatures until a January thaw melted the insulating snowpack.Freeze-thaw cycles that ranged from −15 to 5°C (with most FTC fluctuating close to 0°C), occurred throughout the mid period, and overland flow and subsurface drainage were generated on several occasions between January and March.The "late" period occurred at the end of the NGS (March-April) and was dominated by warm temperatures with low-intensity frosts (−2 to 10°C), with one period of colder frost (−10°C).Despite >17 cm of snow falling in March, warmer temperatures prevented the accumulation of snowpack, exposing CC to high levels of rainfall, snowmelt, and warm daily temperatures.Several runoff events occurred during the late period.During the study period, ?600 mm of precipitation fell (160 mm as snow water equivalent), 100 mm more than the 30-yr average of 500 mm of precipitation (1981-2010;ECCC, 2018); however, the snowfall was consistent with what is typical for the region.

Phosphorus Release from Different Cover Crop Species
Before exposure to FTC, there was little difference in WEP availability among the CC species, and median WEP pools for all species were <0.25 kg ha −1 (Fig. 2).After exposure to varying intensities of frost (and overall air temperature fluctuations), differences in peak WEP concentrations were observed among CC species with regard to the amount and seasonal timing of P release.
Water-extractable P concentrations differed with CC species in all three periods of the study (p < 0.05, Supplemental Table S1b).The greatest WEP concentrations were observed in the root tissue of the oilseed radish, with slightly smaller WEP concentrations in the radish shoots and oat (Fig. 2).In contrast, WEP concentrations in the volunteer wheat, vetch, and rye did not increase after FTC exposure.These differences were driven by the combination of the range in WEP concentrations among species and the large range in plant biomass (>1000 kg ha −1 dry) among the CC plots (Table 1).Species with greater biomass also had the greatest WEP release.
The timing of changes in WEP concentrations also differed among species following weather patterns (Fig. 1).Although the three frost-tolerant species (cereal rye, hairy vetch, and volunteer wheat) all maintained relatively low WEP concentrations throughout the NGS in comparison with other CC species (Fig. 2), WEP concentrations in both volunteer wheat (p = 0.002) and cereal rye (p < 0.001) declined significantly over the three periods after repeated exposure to FTC, whereas hairy vetch did not show a significant change in WEP pools (Supplemental Table S1c).In contrast, the two frost-intolerant species, oat and oilseed radish, demonstrated different patterns in WEP release.Oat was strongly affected by the early autumn FTC (Fig. 2), where median WEP concentrations increased by a factor of five between sampling campaigns (no precipitation fell between these occasions).After these early FTC, the oat CC was frost killed and began to decompose over the remainder of the NGS, resulting in consistently elevated WEP concentrations.Water-extractable P concentrations in oilseed radish differed significantly among periods of the NGS in both shoot (p < 0.001) and root (p = 0.013) tissue (Supplemental Table S1c).Similar to oat, the greatest WEP availability from radish shoot tissue was experienced in the early period, following light FTC (greater than −10°C) that doubled after the plants were frost killed, whereas the root tissue remained unaffected (Fig. 2).However, WEP concentrations in the oilseed radish roots increased considerably during the mid period after heavy FTC (−10 to −20°C).Water-extractable P concentrations in the radish roots were smaller in subsequent sampling campaigns, although they remained above the fall baseline.
Water-extractable P concentrations were <5% of the total vegetation P pool for hairy vetch, cereal rye, winter wheat, and oilseed radish shoot samples (Table 1).Concentrations in oat (5-10%) and oilseed radish roots (>20%) were a greater proportion of total vegetation P.Although WEP availability varied temporally in some species, the TP content of the CC species did not vary over time, with the exception of oat, which declined (data not shown).

Phosphorus Release from Soils and Phosphorus Concentrations in Leachate
Median soil WEP pools in the top 5 cm typically ranged between 0.5 and 2 kg ha −1 (Fig. 3), and maximums ranged from 1 to 3.5 kg ha −1 (Table 1).The soil WEP pool was considerably greater than the vegetation WEP pool (Fig. 2 and 3; Table 1), and only the oilseed radish released P amounts comparable with the lower end of the soil range (Table 1).Soil WEP pools within the CC plots and the control were similar in magnitude.Significant differences existed among the CC plots (F = 3.97, p = 0.001; Supplemental Table S1d), but not between periods.Greater soil WEP pools were observed in the oat, rye, and vetch plots than in the control and oilseed radish plots (representative data shown, Fig. 3), but significant differences were only found between the oat plot and the control and radish plots.It should be noted that the oat, rye, and vetch plots were located at lower topographic positions in the field than the control and radish plots (Supplemental Fig. S1).Although soil WEP concentrations varied throughout the NGS, they did not differ significantly with period (Supplemental Table S1d) and they did not relate to seasonal changes in plant WEP concentrations in the CC (Fig. 2 and 3).Even the peak WEP concentrations in the oilseed radish root in midwinter did not produce elevated WEP concentrations in the top 5 cm of soil (Fig. 2 and 3).
Water samples collected from piezometers (5-15 and 15-25 cm) showed baseline DRP concentrations were generally in the range of 0.01 to 0.3 mg L −1 .Dissolved reactive P varied between the depths (5-15 vs. 15-25 cm), with smaller concentrations observed in deeper soil (Fig. 3).Concentrations of DRP in shallow groundwater exhibited temporal variability, with the greatest concentrations occurring during the initial wetting phase of the early period (Fig. 3), following very dry soil conditions after wheat harvest in 2016.The earliest piezometer samples were collected after a 22 mm rainfall event (Fig. 1) and resulted in concentrations >0.5 mg DRP L −1 in shallow groundwater, with several samples as high as 1 to 3 mg L −1 .Although DRP concentrations were elevated in some plots in the mid and late periods, no concentrations reached the initial range.Elevated concentrations were not consistent in timing nor magnitude with CC WEP release (Fig. 3), except the oat plot, which had greater DRP concentrations throughout the study period (Fig. 3).Dissolved reactive P concentrations shallow groundwater were compared with plant and soil WEP pools measured prior to the rain event to determine if DRP concentrations in runoff reflected either P source, and neither potential source of P was significantly correlated (Supplemental Fig. S2).
Despite the number of rainfall and snowmelt events over the NGS, overland flow seldom occurred (only on four occasions), and overland flow did not occur consistently across the plots, primarily due to field microtopography.In several of the events that generated minimal amounts of overland flow, runoff drained down the middle of the field along the intersection of the CC plots (Supplemental Fig. S1).Overland flow collectors were placed away from this zone to avoid the potential of runoff mixing from multiple CC plots, and consequently, water did not reach all collection reservoirs during small events and samples were limited.However, during the larger events, which typically account for the majority of P loss in surface runoff (Van Esbroeck et al., 2017), samples were captured within the collectors at the plots.Concentrations of P in overland flow varied considerably, with concentrations ranging from 0.025 to >1 mg DRP L −1 and 0.01 to 9 mg TP L −1 .Overland flow DRP concentrations were  not significantly correlated with plant WEP, but a strong significant correlation was found with soil WEP (t = 2.500, df = 7, p = 0.041, r = 0.687; Supplemental Fig. S2).Concentrations of TP were generally the smallest in the control and oilseed radish plots, which were located in the upslope topographic positions, whereas oat, rye, rye radish, and vetch, which were located in the lower slope positions, had more elevated TP concentrations (data not shown).No significant correlation was found between plant WEP and overland flow TP concentrations (Supplemental Fig. S2); however, TP concentrations were significantly correlated with surface soil WEP (t = 2.989, df = 7, p = 0.020, r = 0.749).In the control and radish plots, TP was predominantly in the particulate form, with DRP consisting 19 and 12% of TP (median of all concentration ratios), respectively, whereas the oat and rye plots were more evenly split between DRP and particulate P, with DRP consisting of 48% of TP in both plots.

Differences in Phosphorus Release among Cover Crop Species
The observed differences in P release among CC species were consistent with previous work, which suggests that some species may be more susceptible to the impacts of FTC than others (Miller et al., 1994;Sturite et al., 2007;Liu et al., 2013;Øgaard, 2015;Lozier and Macrae, 2017).Observations of smaller WEP concentrations from more frost-tolerant species and larger WEP concentrations from frost-intolerant species were consistent with previous studies (Øgaard, 2015;Lozier and Macrae, 2017;Cober et al., 2018).Leachate concentrations were also in a range similar to those reported elsewhere in the literature.For example, Riddle and Bergstrom (2013) reported ranges of roughly 0.1 to 1.5 kg ha −1 in leachate from various CC species.Much of the existing research on mechanisms of P retention in plants was done on nonagricultural species in colder regions: thus, it would be difficult to identify the precise mechanisms used by these CC species (Morton, 1977;Chapin, 1980;Chapin and Kedrowski, 1983).However, the low P release and decline in P availability after early NGS FTC suggest that frost-tolerant CC may be better options for managing P release in cold regions.
For the frost-intolerant species, the potential for P release after FTC was identified previously.Numerous laboratory studies have demonstrated great potential for P leaching from oilseed radish (Liu et al., 2014a;Miller et al., 1994;Øgaard, 2015;Cober et al., 2018), or oat (Lozier and Macrae, 2017;Cober et al., 2018), which is consistent with this field study.However, the magnitude of WEP released was much smaller (three times less) in this field setting compared with what has been observed in the laboratory (Cober et al., 2018).This difference among the frost-intolerant species is likely due to the senescence and desiccation of the oat tissue that occurred within a week of the first killing frost, which Cober et al. (2018) noted did not occur in the laboratory experiment.Similar to oat, Elliott (2013) found that living winter wheat released significantly more P than dried wheat stubble.
The high concentrations of WEP released from oilseed radish and oat suggest that these species may be less desirable choices as CC, particularly in fields where overland flow and ponding occur regularly or P loss is a concern.The sizeable concentration of WEP in radish tissues (>1100 mg kg −1 ) presents a great risk for P release given the sizeable biomass of oilseed radish produced (Table 1).Indeed, with only emergent portions of root sampled for biomass, the radish root has more potential to act as a P source.However, it should be noted that the greater WEP release from frost-intolerant species was likely affected by the use of autumn biomass values, as biomass was not resampled in spring.Biomass in the oat and oilseed radish plots visually degraded by spring, and both species likely released a greater total amount of P through this degradation compared with other species.However, the plot-scale magnitude of WEP available from these plants was likely smaller in the late period than what was reported here, because of the inflated biomass value compared with unaffected CC species.This is unlikely to affect the peak points of WEP release, such as the January spike in oilseed radish root that occurred concurrently with the start of biomass degradation.

Relationships between the Cover Crop Phosphorus Content, Soil Phosphorus Supply, and Phosphorus Concentrations in Water
The observed concentrations of soil WEP were toward the low range compared with other studies that examined the impacts of FTC on P loss from plants (Bechmann et al., 2005) or DRP leaching from soil (Pote et al., 1996).The release of P by CC was not apparent in soil P concentrations after storm events (potential leaching opportunities).Indeed, the differences in WEP concentration between CC sampling events were so minimal that any impact of leached CC P was not apparent within the natural variability of the soil P concentrations.Although there were small but significant differences in soil WEP between the oat and the control and radish plots due to within-field variability (F = 3.97, p = 0.001), the range of soil WEP content was similar between the control and most CC plots before and during exposure to FTC.In addition, some of the variability in soil P concentrations may have been related to the topography of the field (e.g., greater soil P in the oat and hairy vetch plots, which were located in the downslope topographic position; Supplemental Fig. S1), as topography has been shown to influence soil P (Moore et al., 1993;Kozar et al., 2002;Adhikari et al., 2018).Although small increases in soil test P concentrations and the occurrence of runoff were apparent in the low slope positions of the field, they do not appear to have affected the results of the study, which demonstrated that P release from CC does not appear to increase soil WEP concentrations.
The low concentrations of P in piezometer samples (Fig. 3) suggest that either very little P was leached from the CC, the movement through the soil matrix enabled adsorption to the soil, or both.The smaller DRP concentrations in the deeper piezometers (15-25 cm) relative to shallower piezometers (5-15 cm) indicates that P release from plants or surface soils is likely buffered by the subsoil (Holtan et al., 1988;Riddle and Bergstrom, 2013).Given that tile depth in southern Ontario is typically ?60 to 90 cm (Van Esbroeck et al., 2017), the concentrations observed in piezometers are not directly comparable with potential edge-of-field loss from tiles, especially given that plant WEP was not correlated with piezometer DRP.Given the high soil test P for this field site (26.5 ± 8.3 mg kg −1 ), there is potential for other fields with lower soil test P to exhibit even greater buffering of P in leachate.
The occurrence of high TP concentrations in overland flow samples (e.g., 1-2 mg L −1 ) supports existing literature that has shown the potential for overland flow to transport large amounts of P directly from fields (Sharpley and Kleinman, 1998;McDowell et al., 2001;Gentry et al., 2007;Macrae et al., 2007).Compared with the control, the ratio of DRP to TP was greater in all CC plots except oilseed radish.The greater proportion of particulate P concentrations from the control and oilseed radish plots was related to the degree of ground cover during the mid period, during which the two plots had the least residue cover (no CC planted and rapid decomposition, respectively), and thus soils appeared to be more exposed to the forces that increase particulate P export (Sharpley and Smith, 1991;Sharpley and Kleinman, 1998).
Although some contribution of DRP from CC leachate should be expected, no consistent trend was observed among the different CC plots.For example, although the radish root had a much greater WEP pool than oat (Fig. 2), the oat plot had more DRP in overland flow than the radish plot.In contrast, the greater TP concentrations in surface runoff from the rye plot showed little contribution from the plant leaching (in agreement with the poor correlations with plant WEP, Supplemental Fig. S2) and instead suggest a stronger contribution from soil P. The lack of the expected trend between runoff P and CC WEP may be due to temporal differences in P supply and transport timing, as different CC species have been shown to release P under different FTC conditions (Øgaard, 2015).For example, the greatest oilseed radish WEP concentration was observed after the first overland flow event had occurred, and oilseed radish WEP declined by the second overland flow event.Similarly, Roberson et al. (2007) expected to see a significant increase in runoff P concentrations from CC contributions but suggested that the natural variability in the field system, FTC timing, and precipitation may have reduced the potential for P export from the field.The results of the current study indicate that irrespective of CC species, CC have little impact on P concentrations in surface or subsurface runoff, and that the chemistry of runoff is more related to soil P content and buffering capacity.However, the conclusions that can be drawn from this analysis are limited by the small number of runoff events that occurred through the NGS, and the fact that overland flow at this site was generated by rainfall and snowmelt on partially frozen soil rather than surface ponding, as prolonged contact between ponding water and vegetation can increase P leaching (Lozier and Macrae, 2017).
More field trials are needed to better quantify the potential impact of overland flow on P loss from CC, and these should consider the impact of different soils, variation in soil test P, and the impact of slope.However, it is also important to note that despite the potential for overland flow to move DRP, environmental managers should consider the other beneficial properties of CC, such as reducing erosion and increasing soil organic C (Kleinman et al., 2005;Kaspar and Singer, 2011).

Conclusions
The extent of P leaching and export from plants during the winter period depends on regional climate conditions, such as the timing and intensity of freeze-thaw cycling relative to precipitation, as well as field characteristics.This study has shown that CC were not a greater source of P export than the soil under climate conditions typical of the NGS in the Great Lakes region of North America.Frost-intolerant CC demonstrated potential for greater P release than frost-tolerant species.Phosphorus from surface soils may have a larger contribution to overland flow concentrations than most CC.
Planting CC should remain an agricultural BMP in the Great Lakes region, as the small quantities of P released by the plants do not appear to lead to elevated P in surface or subsurface runoff in a field setting.Species choice should be dependent on site-specific characteristics and climate; however, frost-tolerant species are recommended for cold regions where P loss is a concern.More plot-and field-scale studies are needed to evaluate the efficacy of CC in other regions where plants may be exposed to more intensive freezing and consistently frozen soils.

Supplemental Material
The supplemental material includes two figures: a map detailing the location of the CC plots and sampling equipment, and the correlations between P sources (plant WEP and surface soil WEP) and DRP concentrations in overland, piezometer, and lysimeter samples.The material also includes a table containing the results of the ANOVA tests.
Sciences and Engineering Research Council (Macrae-DG).Vito Lam is acknowledged for laboratory and field assistance.

Fig. 1 .
Fig. 1.Weather data from the field season showing (a) air temperature 1 m above the surface (gray line) and soil temperatures 1 cm below the surface (black line), and (b) precipitation (rain + snow water equivalent, bars) and snow cover (dashed line) throughout the 2016-2017 nongrowing season.Sampling events were split into three periods of the nongrowing season, delineated from weather conditions.

Fig. 2 .
Fig. 2. Median (point) and 95% confidence intervals (bars) for water-extractable P (WEP) pools from cover crop samples of various species collected from a field site throughout the 2016-2017 nongrowing season.Sampling events have been split by periods of the nongrowing season by weather conditions (shown in Fig. 1).Note the different scaling used for oilseed radish root.

Fig. 3 .
Fig. 3. Median (point) and 95% confidence intervals (bars) for (a, c, e) water-extractable P (WEP) concentrations released from soil and (b, d, f) dissolved reactive P (DRP) concentrations from water collected in piezometers from plots of (a-b) hairy vetch, (c-d) oat, and (e-f) oilseed radish.Data from shallow (5-15 cm) piezometers are shown in black, and data from deeper piezometers (15-25 cm) are shown in gray.Sampling events have been split by periods of the nongrowing season by weather conditions (shown in Fig. 1).

Table 1 . Plant and soil properties of cover crop plots located in Ontario, Canada, from the 2016-2017 nongrowing season. All values shown are averages ± SD, with the exception of plant total P concentrations from the 12 Oct. 2016 sampling, which were not replicated. Plant biomass was sampled 12 Oct., 2016, soil for Olsen P was sampled 12 Apr. 2017, and soil bulk density was sampled in October 2017.
98 ± 1.32 (18 Jan.) † WEP, water-extractable P.