Phosphorus Transport in the Mississippi Delta: Associations to Surface and Groundwater Interactions

: Groundwater (GW) in the Mississippi Delta has some of the highest phosphorus (P) concentrations measured in the U.S. Chemical data collected from GW and surface water (SW) sites were compared to understand factors affecting P concentrations. Spatial instability in Delta GWs indicates that P sources vary. High P measurements in shallow wells near rivers, in shallow nested wells compared to deeper nested wells, and P ﬂuctuations in wells over time suggest that the land surface may be a greater source of P in shallow groundwater than natural geological deposits. Widespread reducing conditions in shallow GW, long-term P applications to the land surface, and shallow wells being proximal to streams are possible covarying explanatory variables. Potential SW to GW pathways of P include leaching and preferential ﬂow paths; however, GW interactions with SW via irrigation, although unnatural, can result in P deposition on soils and later transport to SW or GW. GW tracer data indicate that irrigation return ﬂows can exceed natural baseﬂow discharge to some streams in late summer. Studies are needed to conﬁrm the degree that P is mobilized from soils and bed sediment to shallow GW and to determine how declines in GW levels resulting from irrigation affect ecological services in SW.


Introduction
Past water quality studies of phosphorus (P) in the Mississippi Delta, a 18,130 km 2 area in northwestern Mississippi, have indicated that concentrations in groundwater (GW) and surface water (SW) can be high relative to other parts of the United States [1][2][3][4]. Although the Mississippi River Valley alluvial aquifer (MRVA) which provides GW for the Delta has some of the highest P concentrations among U.S. principal aquifers [3], median total dissolved P (TDP) concentrations in GW samples from the Delta are well above the background concentration of 0.02 milligram per liter (mg/L) for the entire MRVA [4,5]. Less is known regarding how P concentrations in Delta SWs compare to SWs in other parts of the U.S. However, total phosphorus (TP) concentrations in streams in the Mississippi Embayment (an approximate 128,980 km 2 area in the six states of Arkansas, Kentucky, Louisiana, Mississippi, Missouri, and Tennessee, which includes and has in common with the Delta that land use is primarily for row crop agriculture) ranged from the 67th to 93rd percentile relative to other streams in the U.S. [6], and the water quality in some Delta waterbodies has been impaired because of nutrients [7,8].
The prominent direction of P movement and the degree of P transport between SW and GW sources has been a common debate in the Delta [4]. More specifically, high P concentrations in Delta GWs have resulted in hypotheses regarding the potential of geological deposits in GW (i.e., likely via estuarine deposition, [8]) to be sources of P to Delta SWs [4]. However, P is essential for agricultural production [9], and to sustain yields, P must be added to all soils when they are cropped heavily for long periods [8]. While   Water quality data collected from selected GW (i.e., wells) and SW sites (i.e., standing water in irrigated fields, drainage ditches, and small-and medium-sized rivers) in the Delta were compared to better understand P transport. Three primary approaches were involved in the evaluation process. Overall P concentrations measured by the U.S. Geological Survey (USGS) from Delta GWs and SWs were compared as a means of assessing the extent of P cycling and to potentially indicate sources and pathways of P. In addition, P concentrations in GW were compared to well depth and to the distance of the well from SW (i.e., streams or oxbow lakes). P concentrations were compared to well depth as a means of assessing possible P gradients (between the land surface and GW and vice versa), while P concentrations in wells were compared to the distance of the well from the nearest SW to determine if SW interactions with GW were affecting P concentrations in GW.
To supplement historical data, the study included three data collection phases that targeted different processes and questions related to dissolved P and TP concentrations in GW and SW. For Phase 1, conducted in 2014 and 2015, water samples were collected along an irrigation pathway to determine how water quality changed from (near) the well head, the field edge (i.e., field drains), and drainage canals that receive irrigation return water. The intent of Phase 2 was to determine how P concentrations varied by GW depth. P concentrations were compared among five groups of clustered wells at different Delta locations. Individual wells in each cluster had different depths; installation was completed in 2016 and 2017. For Phase 3, conducted in 2017, samples were collected at 10 paired well and stream sites to determine the degree of GW/SW interaction and, potentially, the direction of nutrient transport.

Sample Collection and Processing
Most of the SW and GW data collected from 2014 to 2017 were collected during summer and fall, which coincided with baseflow conditions. All samples (including historic sampling efforts) were collected following established USGS protocols for SW [39] and GW [40]. SW samples were collected from bridges or boats, or by wading. Depending on the amount of streamflow, SW water samples were collected by dipping a sample bottle into the stream or by equal width, depth-integration methods [39]. Field parameters were measured with calibrated water quality monitors. Samples were processed according to USGS protocols [41] and were stored on ice and shipped to the USGS National Water Quality Laboratory in Lakewood, Colorado, for analysis. Descriptions of the methods used to analyze water quality samples are provided in [42].
GW sample collection varied depending on well type and pumping status. Irrigation wells that were actively pumping were sampled by filling sample bottles directly from the irrigation well discharge. Irrigation wells that were not pumping and monitoring wells (not used for irrigation) were sampled with low-volume sampling pumps. Prior to collecting GW samples, three casing volumes were purged. Water quality samples for laboratory analysis were collected after field parameters (i.e., water temperature, pH, dissolved oxygen, and specific conductance) had stabilized. Field (parameter) measurements were recorded using electrodes placed in a flow cell chamber once GW had been pumped to the surface.

Field Sampling Overview for the Three Study Phases
Water quality data for the three study phases included field parameters, alkalinity, nutrient constituents, and major and minor ions (Table S1). Major and minor ions were analyzed because of their potential for indicating P pathways (i.e., storm runoff, GW and SW interactions) and P sources to streams (i.e., released irrigation water, storm runoff, or geologic deposits). Sampling details for the three phases of study were as follows.
Phase 1: The intent of Phase 1 was to determine how P concentrations varied from the well head to the receiving canal and how geochemical processes affected dissolved and TP concentrations along that pathway. Water quality samples were collected from three different points on the pathway: (1) the irrigation well (post-atmospheric exposure at the Water 2022, 14, 2925 6 of 29 well discharge pipes), (2) slotted-board risers that served as field drains for irrigated rice fields, and (3) in the SW canal downstream of irrigation return water. For the first round of Phase 1 sampling in 2014, 9 sets of water quality samples were collected on 9 dates at one well, the nearby field drain, and from an adjacent canal. For the second round of sampling in 2015, three additional clusters-three irrigation wells (depth of~30.5 m), associated field drains, and receiving canals-were sampled on three occasions from July to August.
Phase 2: The intent of Phase 2 was to compare P concentrations across different GW depths. In August 2016, 2-4 monitoring wells were installed (and sampled) at different depths (Table 1) but near to each of three permanent irrigation wells. In May 2017, 2 additional monitoring wells of different depths (Table 1) were installed nearby two different irrigation wells, and those two well clusters were sampled later that month. Sampling was conducted on the same date for 2 of 5 well clusters, and within the same month for 2 of the 3 remaining well clusters. For all five well clusters, monitoring wells were 2.54 cm (diameter) PVC pipes and screens were positioned at different depths below the land surface using direct push methods. Phase 3: Data collected in Phase 3 were gathered to evaluate the degree of exchange between GW and SW sources. On four occasions from March to September 2017, 10 stream and well pairs near each other were sampled on the same or subsequent days. In addition to comparisons of nutrient concentrations, major and minor ion concentrations and specific conductance values were compared at the 10 paired GW and SW locations to assess the degree of GW influence on SW (i.e., GW tracers). Concentrations of potential GW tracers identified with this process were compared to stream stage throughout the growing season and summer baseflow periods as a means of evaluating the timing and relative degrees of GW influence on SW, and, relatedly, the degree of P contributions from natural GW discharge and irrigation return flows.
Although the limited amount of available P analyses for Delta GWs had focused on TDP [4], TDP was inadvertently omitted from the nutrient analyses conducted on GW samples in Phases 1 and 2. TDP also was omitted from the SW analysis for Phase 3. Consequently, SRP was the primary dissolved P constituent evaluated for the 3 phases of study conducted between 2014 and 2017. TDP data were, however, available for historical sampling efforts prior to 2014. P mobility in soils and GW is influenced by reducing conditions perhaps more than any other environmental condition [43][44][45][46][47]. Selected GW and SW water quality data (i.e., dissolved oxygen, iron, and manganese) were evaluated for samples collected for the three study phases conducted from 2014 and 2017 to indicate the prevalence of reducing conditions.

Historical Data Evaluation
Past sampling conducted by USGS in the Delta for P has often been associated with GW and SW studies not related specifically to P; thus, analysis conducted for TP, SRP, and TDP has been sporadic in GW and SW. Consequently, P analysis for this study was limited to times and locations where P data were available.
Phosphorus concentrations measured in 108 Delta wells sampled between 1998 and 2017 (including data collected for the three phases of study, Table S2) were included for the GW analysis. Few Delta wells >36.6 m deep had P data, so analysis was limited to wells which were ≤36.6 m deep. P has been sampled by USGS much less frequently in GW than in SW. TP was measured in GW in only 46 samples, but analysis of SRP and TDP in GW was more common-195 and 150 samples, respectively. Unfortunately, 10 of those 150 TDP samples did not have associated depths.
TP, SRP, and TDP had been analyzed in 1672, 606, and 286 samples, respectively, at the 8 SW sites sampled in the Delta from 1990 to 2020 (Table S3). Much of the P data from Delta streams were collected during stormflow conditions (i.e., when P concentrations are generally much higher than baseflow conditions). It was important that stormflow data be adequately represented, however, because most of the annual load of P (and turbidity) is transported when rainfall results in storm runoff and increasing discharge [48].
Concentrations of SRP, TDP, and TP measured in historic SW and GW were compared to results of other water quality or ancillary variables to provide information regarding the timing and mechanisms associated to P transport in GW and SW. In situ turbidity data from the Bogue Phalia near Leland, Mississippi (a SW site in the Delta that USGS has sampled the largest number of times and one of the few sites where turbidity was routinely measured) between 2002 and 2020 were compared to TP concentrations in SW. Specific conductance values and SRP concentrations were also compared for the Bogue Phalia site between 2002 and 2020. Lastly, average monthly rainfall in Greenville, Mississippi from 1981 to 2010 and mean monthly discharge at the Bogue Phalia near Leland, Mississippi from October 1996 to February 2020 were compared to selected water quality data collected at the 8 SW monitoring sites.

Assessing Spatial and Temporal Variability of P in GW Samples
Two spatial aspects of this study involved comparing P concentrations to well (sample) depth and to distance of the sampled well from the nearest waterbody. Distances from wells to waterbodies were calculated using NHDPlus and the Near (Spatial Analysis) toolbox. Verification that the most appropriate stream location was selected by this approach was conducted using Google Earth.
A temporal assessment was conducted to evaluate TDP variability in individual wells with multiple samples collected over time. P data collected sporadically from GW between 1998 and 2017 (i.e., with small numbers of wells sampled in 1998 and 2008 and larger numbers of wells sampled in 2010, 2015-2017 (phases 1-3)) were used for this analysis. However, a trend analysis of P concentrations in GW over the long term was not possible because of the sporadic timing of sample collection.

Irrigation Return Flow Evaluation
Potential (conservative) groundwater tracers were evaluated in streams receiving irrigation return water during the growing season to determine the extent that GW interac- tions influence P transport and how SW quality is affected by irrigation return water. This evaluation required a detailed analysis of the timing and rates of irrigation return flows to Delta streams (i.e., the discharge of irrigation water from flooded fields to streams) which has not been previously conducted. In the absence of those data, in situ observations of irrigation return rates and specific conductance measurements in streams and over time in the MAP ecoregion (specifically for a wetland study involving irrigation return water in Arkansas [49] but also for other ecological studies conducted in the MAP in Arkansas, Louisiana, and Mississippi [23,50]) were used to provide a hypothetical curve for irrigation returns for the growing season (see Figure 2). Arkansas, Louisiana, and Mississippi [23,50]) were used to provide a hypothetical curve for irrigation returns for the growing season (see Figure 2).

Figure 2.
Average total irrigation applied in the Delta by crop and month (borrowed from [33]). The dashed line arrow was added to the original figure and represents a hypothetical irrigation return (runoff) rate for the primary irrigation period, which is based on observations of irrigation return rates and specific conductance measurements in the MAP ecoregion by the author throughout his career.

Water Quality Variation along an Irrigation Pathway (Phase 1)
Specific conductance and bicarbonate data collected from mid-July to mid-August in 2014 and 2015 at the well, field edge, and in the receiving canal indicate how water quality can vary over the irrigation path (Figure 3a,b). Concentrations for both constituents declined slightly from the well to the stream, indicating their potential as groundwater tracers. Although interesting with regard to geochemical changes that occurred across the irrigation path, data from Phase 1 were of little overall value for assessing P transport. Average total irrigation applied in the Delta by crop and month (borrowed from [33]). The dashed line arrow was added to the original figure and represents a hypothetical irrigation return (runoff) rate for the primary irrigation period, which is based on observations of irrigation return rates and specific conductance measurements in the MAP ecoregion by the author throughout his career.

Water Quality Variation along an Irrigation Pathway (Phase 1)
Specific conductance and bicarbonate data collected from mid-July to mid-August in 2014 and 2015 at the well, field edge, and in the receiving canal indicate how water quality can vary over the irrigation path (Figure 3a,b). Concentrations for both constituents declined slightly from the well to the stream, indicating their potential as groundwater tracers.
Water 2022, 14, x FOR PEER REVIEW 9 Arkansas, Louisiana, and Mississippi [23,50]) were used to provide a hypothetical for irrigation returns for the growing season (see Figure 2).

Water Quality Variation along an Irrigation Pathway (Phase 1)
Specific conductance and bicarbonate data collected from mid-July to mid-Aug 2014 and 2015 at the well, field edge, and in the receiving canal indicate how water q can vary over the irrigation path (Figure 3a,b). Concentrations for both constituen clined slightly from the well to the stream, indicating their potential as groundwate ers. Although interesting with regard to geochemical changes that occurred acro irrigation path, data from Phase 1 were of little overall value for assessing P tran Consequently, remaining analyses from Phase 1 are summarized as supplemental rial (see Document S1, which contains Table S4a,b, and Figure S1a  Although interesting with regard to geochemical changes that occurred across the irrigation path, data from Phase 1 were of little overall value for assessing P transport. Consequently, remaining analyses from Phase 1 are summarized as supplemental material (see Document S1, which contains Table S4a,b, and Figure S1a-d).

Clustered Well Depth Analysis 2016-2017 (Phase 2)
SRP and TP concentrations in the five well clusters sampled in 2016 and 2017 for Phase 2 were relatively high, exhibited a high degree of variability across well depth, and indicate that SRP can comprise a high percentage of TP in GW (Figure 4a,b). Maximum SRP and TP concentrations were multiple times those of minimum concentrations for wells in four of the five well clusters and generally decreased as depth increased. SRP and TP concentrations for the one set of nested wells that did not increase with depth were generally lower than for the four remaining well clusters.
Water 2022, 14, x FOR PEER REVIEW 10 of 33

Clustered Well Depth Analysis 2016-2017 (Phase 2)
SRP and TP concentrations in the five well clusters sampled in 2016 and 2017 for Phase 2 were relatively high, exhibited a high degree of variability across well depth, and indicate that SRP can comprise a high percentage of TP in GW (Figure 4a,b). Maximum SRP and TP concentrations were multiple times those of minimum concentrations for wells in four of the five well clusters and generally decreased as depth increased. SRP and TP concentrations for the one set of nested wells that did not increase with depth were generally lower than for the four remaining well clusters.

GW/SW Exchange 2017 (Phase 3)
Concentrations or values of several constituents, but particularly calcium (Ca), bicarbonate (HCO3 − ), and specific conductance, were higher in GW than in SW. Medians calculated for the three constituents from GW samples were 3 to 4 times higher and were statistically different than medians calculated for SW samples collected in 2017 (Figure 5a-c). Additionally, the three constituents were much higher for SW samples collected in the late July and early August timeframe when irrigation return rates were likely highest (see hypothetical return rates in Figure 2), compared to samples collected during other baseflow conditions (i.e., in late June and mid to late September) when less irrigation water would have been returning to streams (Figures 6a-c, S2a-c).
SRP concentrations measured at the 10 wells sampled in 2017 were also higher than in 10 nearby streams (Figure 5d, Tables 2 and 3). Ranges of SRP concentrations for GW samples (0.002-1.73 mg/L) were much wider than the range of concentrations for SW samples that were generally collected during baseflow conditions (0.03-0.66 mg/L), and median concentrations for GW were almost 3 times those of SW samples (0.28 to 0.1.0 mg/L, respectively).

GW/SW Exchange 2017 (Phase 3)
Concentrations or values of several constituents, but particularly calcium (Ca), bicarbonate (HCO 3 − ), and specific conductance, were higher in GW than in SW. Medians calculated for the three constituents from GW samples were 3 to 4 times higher and were statistically different than medians calculated for SW samples collected in 2017 (Figure 5a-c). Additionally, the three constituents were much higher for SW samples collected in the late July and early August timeframe when irrigation return rates were likely highest (see hypothetical return rates in Figure 2), compared to samples collected during other baseflow conditions (i.e., in late June and mid to late September) when less irrigation water would have been returning to streams (Figures 6a-c and S2a-c).
SRP concentrations measured at the 10 wells sampled in 2017 were also higher than in 10 nearby streams (Figure 5d, Tables 2 and 3). Ranges of SRP concentrations for GW samples (0.002-1.73 mg/L) were much wider than the range of concentrations for SW samples that were generally collected during baseflow conditions (0.03-0.66 mg/L), and median concentrations for GW were almost 3 times those of SW samples (0.28 to 0.1.0 mg/L, respectively). Mann-Whitney rank sum tests indicated that concentrations in GW and SW were statistically different (p < 0.05) for the first three constituents. Numbers represent the number of samples analyzed. (Bicarbonate concentrations are divided by 4 for scaling purposes. a A comparison of SRP and total dissolved phosphorus (TDP) concentrations in GW indicated that values for 11 SRP samples were compromised, so TDP concentrations were substituted. TDP was not analyzed in SW, so GW to SW comparisons were not possible.) Figure 6. (a-f). Plots of bicarbonate and calcium concentrations and specific conductance values for six surface water sites sampled during the general summer baseflow period in 2017. In all cases, the three constituents (groundwater tracers) were highest when highest amounts of irrigation water would be expected to be returning from fields to the streams (also see Figure 2). Bicarbonate data are missing for the Bogue Phalia on 26 September 2017. Mann-Whitney rank sum tests indicated that concentrations in GW and SW were statistically different (p < 0.05) for the first three constituents. Numbers represent the number of samples analyzed.
(Bicarbonate concentrations are divided by 4 for scaling purposes. a A comparison of SRP and total dissolved phosphorus (TDP) concentrations in GW indicated that values for 11 SRP samples were compromised, so TDP concentrations were substituted. TDP was not analyzed in SW, so GW to SW comparisons were not possible.) Figure 6.
(a-f). Plots of bicarbonate and calcium concentrations and specific conductance values for six surface water sites sampled during the general summer baseflow period in 2017. In all cases, the three constituents (groundwater tracers) were highest when highest amounts of irrigation water would be expected to be returning from fields to the streams (also see Figure 2). Bicarbonate data are missing for the Bogue Phalia on 26 September 2017. Figure 6.
(a-f). Plots of bicarbonate and calcium concentrations and specific conductance values for six surface water sites sampled during the general summer baseflow period in 2017. In all cases, the three constituents (groundwater tracers) were highest when highest amounts of irrigation water would be expected to be returning from fields to the streams (also see Figure 2). Bicarbonate data are missing for the Bogue Phalia on 26 September 2017.   Table 3. Selected water quality data and total phosphorus:soluble reactive phosphorus (TP:SRP) ratios for 10 surface water sites monitored on multiple dates in 2017.
Shaded rows indicate when specific conductance exceeded 350 µS (µS per centimeter at 25 degrees Celsius), which also coincided with times when highest amounts of irrigation water would be expected to be returning from fields to the streams (also see Figures 2 and 6a-c). All constituents except specific conductance and TP:SRP ratios are reported in milligrams per liter; NA, data missing or not collected; µS, microsiemens; total dissolved phosphorus was not measured in surface water samples.   Table 3). SRP generally comprised half or more of TP (≤2:1 TP:SRP ratio) in SW in late July and early August when specific conductance was >350 µS/cm but comprised half or less of TP at other times when specific conductance was <350 µS/cm (Figure 7, Table 3). Concentrations of constituents that typically vary by redox potential (i and unreduced environments) also differed between GW and SW sample trations were generally <1.0 mg/L in GW samples but ranged above 5 mg/L (Figure 8a). Iron and manganese, which have similar response times to red ments [30], were sometimes magnitudes higher in GW than in SW (Figure 8b manganese concentrations in the larger MRVA have been found to be high cipal aquifers in the U.S. [3]. Concentrations of constituents that typically vary by redox potential (i.e., in reduced and unreduced environments) also differed between GW and SW samples. DO concentrations were generally <1.0 mg/L in GW samples but ranged above 5 mg/L in SW samples (Figure 8a). Iron and manganese, which have similar response times to reduced environments [30], were sometimes magnitudes higher in GW than in SW (Figure 8b,c). Relatedly, manganese concentrations in the larger MRVA have been found to be highest of 29 principal aquifers in the U.S. [3].

Historical Water Quality Evaluation
Median monthly TP concentrations calculated with data collected from the eight Delta streams from 1996 to 2020 were generally highest in the winter and spring and lowest in late summer (Figure 9a). Seasonal fluctuations of turbidity were similar to those of TP (Figure 9a). A linear regression of turbidity and TP data collected from the Bogue Phalia near Leland, Mississippi from 2002 to 2020 (Figure 9b) demonstrates that a strong relation exists between the two constituents.
Unlike TP concentrations, median TDP and SRP concentrations and specific conductance values for the eight Delta streams were high in late summer compared to other times (Figure 10a-c). In contrast to TP and turbidity concentrations, which had a positive relation to rainfall and discharge (Figures 9a and 10d,e), SRP and TDP concentrations and specific conductance values had a negative relation to average monthly rainfall and discharge (Figure 10b-e). Because of the seasonality exhibited by TP and SRP, TP:SRP ratios at the eight streams were highest in early spring and lowest in late summer (Figure 10f).  Unlike TP concentrations, median TDP and SRP concentrations and specific conductance values for the eight Delta streams were high in late summer compared to other times (Figure 10a-c). In contrast to TP and turbidity concentrations, which had a positive relation to rainfall and discharge (Figures 9a and 10d,e), SRP and TDP concentrations and specific conductance values had a negative relation to average monthly rainfall and discharge (Figure 10b-e). Because of the seasonality exhibited by TP and SRP, TP:SRP ratios at the eight streams were highest in early spring and lowest in late summer (Figure 10f).

Comparison of Phosphorus Concentrations in GW and SW
TP concentrations at the eight SW sites were comparable to TDP concentrations in GW but were much higher than SRP concentrations in GW ( Figure 11). TP concentrations in SW ranged from 0.01 to 3.66 mg/L, with a median concentration of 0.4 mg/L. TDP concentrations in GW samples ranged from the laboratory reporting level (LRL) of 0.004 to 3.12 mg/L, with a median concentration of 0.28 mg/L. SRP concentrations in GW samples had a lower range, from the LRL of 0.004 to 1.73 mg/L, with a median concentration of 0.08 mg/L. (Figure 10a-c). In contrast to TP and turbidity concentrations, which had a pos tion to rainfall and discharge (Figures 9a and 10d,e), SRP and TDP concentra specific conductance values had a negative relation to average monthly rainfal charge (Figure 10b-e). Because of the seasonality exhibited by TP and SRP, TP:S at the eight streams were highest in early spring and lowest in late summer ( Fig   Figure 10. Median monthly total phosphorus (a) and soluble reactive phosphorus con (b), median monthly specific conductance values (c), and mean monthly TP:SRP ratio streams and small rivers in the Mississippi Delta from 1996 and 2020 compared to averag rainfall (e) at Greenville, Mississippi from 1981 to 2010 (NOAA National Climatic D 2021), and mean monthly discharge (f) at the Bogue Phalia near Leland, Mississippi (Oct February 2020).

Comparison of Phosphorus Concentrations in GW and SW
TP concentrations at the eight SW sites were comparable to TDP concent GW but were much higher than SRP concentrations in GW ( Figure 11). TP conc in SW ranged from 0.01 to 3.66 mg/L, with a median concentration of 0.4 m  SRP (the inorganic part of P) should comprise most of TDP (which includes both inorganic and organic dissolved forms), so SRP and TDP concentrations should be similar. A comparison of the 286 TDP samples analyzed in SW to paired SRP samples indicated a strong relation existed between the two constituents in SW (Pearson's R of 0.94; Figure  12a), but a similar analysis of 63 paired GW samples analyzed for SRP and TDP indicated that the relation between the two constituents in GW was much weaker than in SW (Pearson's R of 0.60; Figure 12b). A more in-depth analysis of SRP and TDP concentrations in those 63 GW samples revealed that in 44 instances, SRP concentrations were <66% of TDP  Figure 12a), but a similar analysis of 63 paired GW samples analyzed for SRP and TDP indicated that the relation between the two constituents in GW was much weaker than in SW (Pearson's R of 0.60; Figure 12b). A more in-depth analysis of SRP and TDP concentrations in those 63 GW samples revealed that in 44 instances, SRP concentrations were <66% of TDP concentrations. Because SRP concentrations measured in GW were sometimes erroneously low (likely because of P sorption to high concentrations of precipitating iron in Delta GWs and differences associated with sample processing methods, Jim Kingsbury, USGS, written comm., 6 June 2022), interpretations involving SRP data from GW were limited to samples that had the highest SRP concentrations and for times when TDP data were not available and SRP data were supported by TP data (e.g., Figure 4a,b).

Spatial Variability of P in GW
Wells with the highest SRP and TDP concentrations generally fell into two classes: relatively shallow (i.e., ≤15 m) wells that were located near small-to medium-sized rivers, and relatively deep (>21 m) wells that were at greater distances from large, deep rivers or their oxbow lakes (Figure 13a-d). Of the 38 well samples with TDP concentrations ≥75th percentile (0.84 mg/L), 24 samples were collected from relatively shallow wells located near small-to medium-sized streams. Fifteen of those twenty-four samples were collected from six shallow wells that were short distances away from the Sunflower River (near the towns of Sunflower, Merigold, Clarksdale, and Anguilla). Of the nine remaining samples, seven were collected from shallow wells near Steele Bayou (which flows near the Mississippi River and its oxbows), while two were collected near the Bogue Phalia (Table 4).

Spatial Variability of P in GW
Wells with the highest SRP and TDP concentrations generally fell into two classes: relatively shallow (i.e., ≤15 m) wells that were located near small-to medium-sized rivers, and relatively deep (>21 m) wells that were at greater distances from large, deep rivers or their oxbow lakes (Figure 13a-d). Of the 38 well samples with TDP concentrations ≥75th percentile (0.84 mg/L), 24 samples were collected from relatively shallow wells located near small-to medium-sized streams. Fifteen of those twenty-four samples were collected from six shallow wells that were short distances away from the Sunflower River (near the towns of Sunflower, Merigold, Clarksdale, and Anguilla). Of the nine remaining samples, seven were collected from shallow wells near Steele Bayou (which flows near the Mississippi River and its oxbows), while two were collected near the Bogue Phalia ( Table 4).
Four of the five wells with highest TDP concentrations measured were shallow (average depth of 10 m) and close (<50 m) to the Sunflower River ( Figure 14). A well with the three highest TDP concentrations measured (2.89-3.12 mg/L) was approximately 7 m deep and only 7 m from the Sunflower River in the city of Clarksdale (Table 4). A well that had the fifth highest TDP concentration of all wells sampled and the ninth highest overall concentration (1.3 mg/L) was located 138 km (straight distance) from the Clarksdale well near the town of Anguilla. The well near Anguilla was also on the Sunflower River and adjacent to a golf course. Table 4. Details for 38 well samples collected from the Mississippi Delta having total dissolved phosphorus (TDP) concentrations ≥75th percentile of 150 groundwater samples. Samples are sorted in descending order by TDP concentration (also see Figure 13a,b). mg/L, milligram per liter; NA, well depth not available.   Figure 13. Total dissolved and soluble reactive phosphorus concentrations compared to e distance of sampled wells to the nearest stream (a,c, respectively) and to well sampling de respectively). Horizontal dashed lines represent the 75th percentile concentrations for both uents. Gray boxes in the top left of each plot contain samples collected from shallow wells located near small-to medium-sized rivers. Boxes with diagonal lines on the right side of p tain samples collected from deep wells at greater distances from large rivers. SRP concen above the 75th percentile are shown to indicate GW samples where SRP concentrations we est; lower concentrations are not shown because some results were compromised.

USGS Station
Four of the five wells with highest TDP concentrations measured were shall erage depth of 10 m) and close (<50 m) to the Sunflower River ( Figure 14). A well w three highest TDP concentrations measured (2.89-3.12 mg/L) was approximately 7 and only 7 m from the Sunflower River in the city of Clarksdale (Table 4). A well t the fifth highest TDP concentration of all wells sampled and the ninth highest over centration (1.3 mg/L) was located 138 km (straight distance) from the Clarksdale w the town of Anguilla. The well near Anguilla was also on the Sunflower River an cent to a golf course. Of the thirteen samples with TDP concentrations ≥75th percentile collected from deep wells and generally near to large rivers, ten samples were collected from seven wells along the Mississippi River and three samples were collected from three different wells near the Yazoo River. The remaining (38th) sample with TDP concentrations ≥75th percentile was collected from a relatively deep well (>24 m) but near the Sunflower River in the upper part of the watershed (340413090340301, also see Table 4).

Temporal Variability of P in GW
TDP concentrations were moderately to highly unstable in some wells with multiple samples collected over time (Figure 15a-c) but were stable in other wells (Figure 15d-f). Wells with high TDP concentrations (>0.8 mg/L) exhibited more instability (Figure 15a-c) than wells with low TDP concentrations (Figure 15d), but instability was not evident in all wells with high concentrations (Figure 15e,f).  Water 2022, 14, x FOR PEER REVIEW 2 of 33 Figure 14. General proximity of 5 wells with total dissolved phosphorus concentrations that exceeded 1.2 mg/L to streams in the Mississippi Delta. Distances from wells to streams were calculated using NHDPlus and the Near Spatial Analysis toolbox.
Of the thirteen samples with TDP concentrations ≥75th percentile collected from deep wells and generally near to large rivers, ten samples were collected from seven wells along the Mississippi River and three samples were collected from three different wells near the Yazoo River. The remaining (38th) sample with TDP concentrations ≥75th percentile was collected from a relatively deep well (>24 m) but near the Sunflower River in the upper part of the watershed (340413090340301, also see Table 4). The well located in Clarksdale, Mississippi with the highest TDP concentrations (341210090343701) is also an example of a shallow well (~7 m deep) which had unstable TDP concentrations (Figure 15b). Three samples collected at this well from August 2010 to May 2011 ranged from 0.76 to 1.14 mg/L, but a sample collected between those three samples in November 2010 had the second highest TDP concentration measured for this study (3.06 mg/L). Furthermore, TDP concentrations measured in two samples collected in July and August 2017 were comparable to the 3.06 mg/L value measured in 2010, with the August sample having the highest concentration measured across all Delta wells evaluated for this study (3.12 mg/L; Table 4).

Temporal Variability of P in GW
TDP concentrations were moderately to highly unstable in some wells with multiple samples collected over time (Figure 15a-c) but were stable in other wells (Figure 15d-f). Wells with high TDP concentrations (>0.8 mg/L) exhibited more instability (Figure 15a-c) than wells with low TDP concentrations (Figure 15d), but instability was not evident in all wells with high concentrations (Figure 15e,f). Figure 15. Examples of Delta wells with stable and unstable TDP concentrations across time (15digit numbers are USGS well identifiers). TDP concentrations measured at three shallow wells located at different locations but adjacent to small-and medium-sized rivers varied over time (a-c). TDP concentrations were more stable over time in four wells at various locations (d), one shallow well located adjacent to a small river which was also near the Mississippi River (e), and a deep well located near the Mississippi River (f).
The well located in Clarksdale, Mississippi with the highest TDP concentrations (341210090343701) is also an example of a shallow well (~7 m deep) which had unstable TDP concentrations (Figure 15b). Three samples collected at this well from August 2010 to May 2011 ranged from 0.76 to 1.14 mg/L, but a sample collected between those three samples in November 2010 had the second highest TDP concentration measured for this study (3.06 mg/L). Furthermore, TDP concentrations measured in two samples collected in July and August 2017 were comparable to the 3.06 mg/L value measured in 2010, with the August sample having the highest concentration measured across all Delta wells evaluated for this study (3.12 mg/L; Table 4).
In contrast to that high degree of temporal instability, TDP concentrations at four wells that were 8.1-28.7 m deep had low TDP concentrations that were stable over time  In contrast to that high degree of temporal instability, TDP concentrations at four wells that were 8.1-28.7 m deep had low TDP concentrations that were stable over time (Figure 15d). Examples of deep wells with stable TDP concentrations include two wells located near the Mississippi River (USGS well numbers 330152090595603 and 335910090532901).

P Transport in SW
Consistent with other streams in the U.S. [48], P concentrations in SW are generally highest during stormflows when discharge is also high. Consequently, most of the annual P load in Delta streams is transported during winter and spring when storms are common. TP concentrations are typically high in stormflow because of fine sediment eroding from areas with legacy P [51] such as agriculture fields [18,19,52,53], previously deposited bed sediment [54], or stream banks [55]. Moreover, P loss may be higher in the Delta because soils are composed of large amounts of fine clays [52,56], resulting in Delta streams being more turbid than streams in other regions [23]. Similar to other areas, the strong association between TP and clay turbidity in this study suggests that turbidity could be used as a surrogate for predicting P loads in Delta streams [57][58][59].

GW Tracers in SW Indicate An Irrigation Signature
Although the majority of P is transported in Delta streams during stormflow conditions, P availability can be an environmental concern during low-flow summer conditions when most aquatic plant production occurs [60][61][62]. In streams in most areas, particularly those with little or no irrigation, natural baseflow contributions to SW normally increase through summer and into fall because GW comprises larger amounts of streamflow as discharge declines [63,64]. Consequently, baseflow contributions from GW have previously been considered the primary source of dissolved P to Delta streams during low-flow summer conditions [3,4]. Some water quality data collected in late summer for this study indicate, however, that in terms of P and other water quality constituents, irrigation return flows near the end of the growing season can be more ecologically significant than natural baseflow discharge. Higher concentrations of GW tracers were consistently measured in late July and early August when irrigation return rates are highest rather than in September when natural baseflow contributions to streamflow should be equally high or higher (Figure 6a-e). This observation is also supported by water quality data collected at two other times in this study. Rather than increasing upon field exposure, specific conductance and bicarbonate data collected at the well, field edge, and in receiving streams from mid-July to mid-August in 2014 and 2015 declined subtly across the irrigation path (i.e., from the time GW was exposed to the atmosphere to the time irrigation return water entered the stream, Figure 3a,b). Moreover, even though stream discharge for Delta streams and average rainfall are comparable in August and September (Figure 10d,e), median specific conductance values of all historic SW measurements evaluated were higher in August than in September (Figure 10c). Consequently, the data evaluated indicate that for the altered hydraulic setting common to the Delta during late summer conditions, P contributions from GW via irrigation return water exceed P contributions from natural baseflow contributions.

Effect of Reduced Conditions on P Transport
The general gradient of P concentrations across well depth in Phase 2 of this study suggests that the land surface may be a source of P and that leaching is occurring between Delta soils (including bed sediment) and shallow GW in some locations. Although previous studies indicate that reduced conditions increase P mobilization [46,47], another factor affecting P transport could be P availability on the land surface. Subsurface transport of P by leaching can exceed surface runoff where soil P content is high [53,65,66]. Although P transport is generally considered to be greater in porous rather than dense soils [65,67,68], P transport in some clay soils can exceed rates in coarse-textured sandy soils [68,69].
High degrees of irrigation in the Delta also may affect rates of P leaching and loss. Because of the geochemical reactions that occur when soils are flooded [30], Delta soils inundated with irrigation water for extended periods (e.g., rice field) would be expected to develop reduced conditions similar to those of natural wetlands. Draining and subsequent rewetting of enriched organic soils seem to increase soluble P flux [69]. P losses after flood irrigation can increase proportional to the P application rate [70], and P can be unavailable for plant uptake for several months after fields are drained [15]. Transport of dissolved P through flooded sediments over time can result in available sorption sites being saturated, which limits the sorption capacity of sediments and results in additional P loss [45,71].
In addition to diffuse movement by vertical leaching through shallow soils, some data evaluated indicate that lateral flow from rivers through preferential flow paths could be an important mode of P transport from SW to GW [13]. Furthermore, it seems that the highest P concentrations, which were measured in shallow wells, may be associated with P previously deposited or released on or near the land surface. Examples of related potential pathways and sources of P exposure via leaching and lateral riverine connections include P transport resulting from (1) P leaching through soils after heavy P applications or P storage (as fertilizer or litter), (2) municipal or rural septic influence (i.e., especially for wells in or near residential areas), and (3) deposition of P-laden sediments in streams and adjacent to preferential flow paths (i.e., coarse alluvial sands). It is also important to consider, however, that a possible alternative explanation for the occurrence of high P concentrations in GW near the land surface and shallow rivers could involve the presence of geologic P deposits in unexpected, shallow locations near small-and medium-sized Delta rivers, particularly the Sunflower River.

Well Depths and Proximity to Streams Provide Possible Covarying Explanations for P Instability in Delta GWs
Other than the geochemical processes involved in P transport into shallow GW through wet soils or stream bed sediments described above, the degree of hydraulic connection that wells have to rivers or remnant channels [35,72] may be the best explanation for why wells in close proximity to rivers seem to have variable P concentrations. Because rivers have (1) naturally cut channels through the MRVA confining layer in some locations, (2) historically meandered across natural floodplains, and (3) that larger alluvial soils (i.e., less clay) are deposited on and near river banks than at farther distances from streams [73,74], SW discharge to GW from rivers and oxbows can be laterally extensive [75][76][77]. Multiple GW modeling approaches have demonstrated the potential for GW recharge to be highest near geomorphic features that are coarser in grain size and near rivers [78]. Furthermore, recent resistivity and hydraulic conductance surveys conducted by USGS demonstrate that Delta river sections can have varying degrees of GW connection [79] [in press] ( Figure S3), which would result in horizontal and vertical conduits from the stream bed having different abilities to function as preferential flow paths for SW discharge to GW. Consequently, wells near to streams might be expected to have not only different degrees of hydraulic connections to SW but also variable P concentrations.
A point of emphasis relative to GW recharge is that SW discharge is highest during stormflow events. While P is much less conservative (i.e., persistent) in GW than many other chemical constituents (e.g., chlorides), when the contact time between percolating water and the soil particles is short or missing (as might be the case during storms and in some preferential flow paths [77]), P sorption capacity of soils can be reduced or nonexistent [16].
For the few deep wells where data were available across time, P concentrations seemed to be more stable compared to shallow wells. This observation can be interpreted in different ways. Given their greater depths and increased lateral discharge over smaller, shallower rivers, large, deep rivers are capable of transporting P to greater depths and distances than smaller rivers [80]. However, it is also possible that deep wells that were several kilometers from large rivers might be expected to have less hydraulic connection to SW compared to shallow wells near to shallow rivers. Thus, consistent P concentrations observed at some deep wells might imply a more consistent if not constant P source, such as would be the case if a geological deposit were nearby.

Study Implications and Directions of Future Research
Consistent with most areas, large amounts of P are transported in Delta streams during stormflows. Similar ranges of P constituent concentrations in GW and SW, however, suggest that a strong interaction (i.e., a high degree of P cycling) occurs between GW and SW in the Delta.
Recent conceptual models regarding P transport in the Delta have assumed that P sourced from GW is deposited on the land surface through the process of irrigation and is eventually transported to SW (James Rigby, USGS, written communication, 24 February 2022). Even so, the role of discharged surplus irrigation water (remaining in the field after evaporation, transpiration, and infiltration) to adjacent SWs does not seem to have been previously considered in regard to the P cycle. While the data evaluated obviously indicate that some P cycling occurs between SW and GW because of irrigation and suggest that irrigation return flows may be more ecologically significant than natural baseflow discharge in late summer, the high degree of spatial and temporal variability of P constituent concentrations measured across wells in the Delta also indicate that Delta GWs have different degrees and different pathways of P exposure.
Because it is generally accepted that shallow GWs are younger than deeper GWs [3], it was anticipated that if GW was the predominant source of P to SW, P concentrations would be higher in deep, older waters (where geological deposits might be likely to occur) than in shallower, younger waters. The data analyzed, however, suggest that P concentrations in shallow GW in the Delta are related more to SW sources than to deep geological deposits.
More specifically, (1) measurements of highest P concentrations occurred in shallow wells located near shallow rivers, (2) measurements of SRP and TP concentrations in shallow nested wells were generally higher than in deeper nested wells, and (3) P concentrations fluctuated in some shallow wells over time (possibly because of hydrological variation of adjacent SWs).
Although P concentrations in SW can be influenced by GW head pressure and SW proximity to GW [81] and multiple studies have found GW to be a source of P to SW [2,80,[82][83][84], SW has rarely been considered to be a source of P to GW [85]. Thus, the finding that GW concentrations proximal to Delta streams can be influenced by SW is atypical, and more data are needed to confirm this analysis. Explanations for these rather unique observations regarding P cycling from SW to GW may be related to distinct characteristics inherent to the Delta, namely, widespread reducing conditions, long-term and spatially variable P exposure to the land surface, and that shallow wells are frequently located next to streams.
While there is substantial literature documenting the conditions that have facilitated P leaching and transport in other areas, mechanisms for how P is transported from SW to GW in the Delta are poorly understood. Studies are needed that facilitate determinations for the degree of dissolved P mobilization upon its release from field soils or bed sediment under reducing conditions, when soils contain large amounts of clay, in areas of low and high conductive resistivity (i.e., varying preferential flow paths) near streams, and under different hydrological conditions. Given that most of the dissolved P and TP are transported in Delta streams during the winter and spring, frequent sampling of P in shallow GW could indicate how hydrology affects P transport.
Although the threat to future water availability has often been considered for the Delta and for the MAP [86], the ramifications of GW use on stream water quality has been considered much less frequently. Studies are needed in the Delta and throughout the MRVA to determine how irrigation practices and associated declines in GW levels are affecting streamflow, water quality, and associated ecological services.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/w14182925/s1, Document S1: Phase 1, Irrigation pathway study, 2014-2015. Figure S1: . Soluble reactive phosphorus (SRP) concentrations measured in well, field, and canal samples collected on three occasions in the 2015 irrigation season (a); SRP concentrations for each group of wells, field drains, and receiving canals sampled on three occasions in 2015 (b); total phosphorus (TP) concentrations measured in well, field, and canal samples collected on three occasions in the 2015 irrigation season (c); and TP concentrations measured for each group of wells, fields, and canals sampled in 2015 (d). Figure S2: Streambed hydraulic resistivity measurements in the Mississippi Delta (modified from [79]). Table S1: Constituents analyzed in groundwater and surface water samples evaluated from the Mississippi Delta. Table S2: Station identifiers and number of soluble reactive and total dissolved phosphorus samples collected from 108 wells located in the Mississippi Delta from 1998 to 2017 and analyzed for this study. Table S3: Information for surface water quality sites where sampling occurred and with historic data evaluated in this study. Table  S4: Soluble reactive phosphorus (SRP) and total phosphorus (TP) concentrations (in milligrams per liter) for samples collected from four irrigation clusters, each containing an irrigation well, associated field drain, and receiving canal site in 2014 (a) and 2015 (b). Reference [30] are cited in the supplementary materials.
Author Contributions: As the sole author, B.J. was responsible for data analysis and writing the article. The author has read and agreed to the published version of the manuscript.
Funding: The U.S. Corps of Engineers funded USGS data collection efforts for the three phases of study conducted between 2015 and 2017 and writing of this article (no grant number is available). Historical data collection efforts were funded by USGS and previous cooperators on various cooperative projects.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable. Data Availability Statement: All water quality data evaluated for this study are publicly available from the USGS National Water Information System database [87] and can be accessed using the USGS station numbers provided in various tables.