Environmental Science and Pollution Research

, Volume 23, Issue 16, pp 16056–16066 | Cite as

Effects of forested floodplain soil properties on phosphorous concentrations in two Chesapeake Bay sub-watersheds, Virginia, USA

  • B. K. Odhiambo
  • M. C. Ricker
  • L. M. Le Blanc
  • K. A. Moxey
Research Article

Abstract

Aquatic ecosystems are known to undergo fluctuations in nutrient levels as a result of both natural and anthropogenic processes. Changes in both extrinsic and intrinsic fluvial dynamics necessitate constant monitoring as anthropogenic alterations exert new pressures to previously stable river basins. In this study, we analyzed stream water and riparian zone soil phosphorous (P) dynamics in two third-order sub-watersheds of the lower Chesapeake Bay in Virginia, USA. The Ni River is predominantly forested (70 % forested), and Sugarland Run is a more human impacted (>45 % impervious surfaces) sub-watershed located in the suburbs of Washington D.C. Total stream P concentrations were measured during both high and low flows and Mehlich-3 methods were used to evaluate potential P fluxes in riparian soils. The results show total stream P concentrations in Sugarland Run ranged from 0.002 to 0.20 ppm, with an average of 0.054 ppm. In contrast, the forested Ni River had typical stream P concentrations <0.01 ppm. Total soil P was significantly higher in the more urbanized Sugarland Run basin (23.8 ± 2.1 ppm) compared to the Ni River basin (16 ± 3.7 ppm). Average stream bank erosion rates and corresponding cut-bank P flux rates were estimated to be 7.98 cm year−1 and 361 kg P year−1 for Ni River and 9.84 cm year−1 and 11,600 kg P year−1 for Sugarland Run, respectively. The significantly higher values of total P in the stream water and floodplain cut-banks of Sugarland Run suggests erosion and resuspension of previously deposited legacy sediments is an important processes in this human-impacted basin.

Keywords

Stream phosphorus Riparian soils Legacy sediment Water quality 

Introduction

Anthropogenic alterations in many watersheds are associated with water quality declines of fluvial, lacustrine, and estuarine systems. Water quality degradation is becoming more prevalent and is of great concern to land and water resource managers globally (Hooda et al. 2000; Binkley et al. 2004; Lai and Lam 2008). Basin alterations, such as replacing stable forested landscapes with agricultural and urban land uses, typically results in increased fluxes of sediments, nutrients, and other contaminants. Phosphorous (P) is regarded as a limiting nutrient in freshwater aquatic systems and, hence, is typically associated with eutrophication in many cases (Dunne et al. 2006; Lai and Lam 2008; Young et al. 2013). The application of P fertilizers beyond plant requirements in many agricultural fields results in loss of excess P to adjacent water bodies, and this is a primary cause of eutrophication in many surface water bodies (Nair 2014). Floodplains and wetlands can alleviate these water quality issues by retaining sediments and nutrients adsorbed onto mineral and soil organic matter (SOM) surfaces (Darke and Walbridge 2000; Jones et al. 2001; Ricker and Lockaby 2014). However, in urbanizing areas, floodplains and riparian wetlands continue to decline because of progressive expansion of human infrastructure leading to increased impervious surfaces, runoff, and alluvial bank erosion (Ricker et al. 2008; Hupp et al. 2013; Bassi et al. 2014). In addition to external fluxes, excess runoff entering urban streams as a result of well-developed storm drain systems tends to translate into higher stream velocities leading to net erosion, incision, and re-suspension of stored stream sediments and nutrients. These changes in anthropogenically stressed watersheds are also accompanied by stream morphological alteration, typically in the form of lateral migration and less overall stream-floodplain hydrologic connectivity (Noe and Hupp 2005; Gellis et al. 2009; Hupp et al. 2013).

The historical changes in environmental conditions related to nutrient and sediment fluxes in the Chesapeake Bay, eastern USA, are well documented since the enactment of the Chesapeake Bay Program in 1983 (Phillips 2005; Ricker et al. 2008). However, progressive changes in both extrinsic and intrinsic processes within the Bay necessitate constant monitoring, as modern anthropogenic and climate alterations exert new pressures to the basin. Eutrophication is an ongoing problem within the Bay caused by high concentrations of inorganic nitrogen (N) and P that enhance growth of aquatic vegetation and other organisms (Cooper and Brush 1993; Søndergaard et al. 2003; Niemitz et al. 2013). It has been estimated that >US$3.5 billion was spent towards nutrient controls in the Bay’s watershed between 1985 and 1996 (Butt and Brown 2000). Although point sources, such as sewage discharges are continually being managed, newly identified sources like the remobilization of nutrient-rich legacy sediments that have been deposited from past agricultural activities suggests that the economic impacts of nutrients are not likely to change anytime soon as coastal zone populations and anthropogenic stresses on water ways continue to increase (Niemitz et al. 2013).

This study analyzed the concentrations and fluctuations of stream water total P to evaluate the role of floodplain soil P dynamics in two third-order tributaries of the lower Chesapeake Bay in central and northern Virginia (Fig. 1). The effectiveness of floodplain soils in regulating P fluxes to streams will vary from basin to basin due to differing soil properties and land use characteristics. Thus, determining current soil P concentrations and capacity to retain additional P inputs is essential to understand the environmental risk of future land use changes in these basins. The soil degree of P saturation (DPS) or P saturation ratio (PSR) has been commonly used in evaluating risk of P loss in relation to soil properties and land use (Nair et al. 2004; Dunne et al. 2006) and is thus used in this study to understand soil P dynamics at a sub-watershed scale. In addition, this study also analyzes the contribution of P fluxes from eroded stream bank sediment, a component that is typically missing in total fluvial P flux analyses.
Fig. 1

Location map showing the two study watersheds, Sugarland Run in northern Virginia and the Ni River in Central Virginia, USA. Hydric soils were classified according to the US NRCS database (Natural Resources Conservation 2015) and riparian forest cover was determined from land cover data obtained from earthexplorer.gov

Study sites description

The Ni River in central Virginia (38° 15′ 44.734″ N; 77° 37′ 38.33″ W) is a predominantly forested (70 % forested) sub-watershed of the greater York River system. The Ni River watershed occupies an area of approximately 66 km2, with a 166-ha reservoir located in the mid-reach of the basin that provides municipal drinking water for Spotsylvania County and the city of Fredericksburg, Virginia (VADGIF 2003). The basin is located approximately 81 km south of Washington D.C. in the Piedmont physiographic province of Virginia. According to the US Census Bureau, the study area is rapidly increasing in population, resulting in a 2.7 % bi-annual increase in total residents (USCB 2013). Despite this population increase, the watershed is mostly forested because much of the land is located within the Fredericksburg and Spotsylvania National Military Park, which maintains and protects the natural riparian zones. The Ni floodplain is dominated by wetlands that are bottomland hardwood forests with relatively short hydroperiods (Axt and Walbridge 1999). The predominant soils of the floodplain are classified as Fluvaquents and Udifluvents (Natural Resources Conservation Service 2015). The low-lying topography and rolling hills of the watershed form many natural wetlands that are potential sinks for eroding sediment and nutrients.

Sugarland Run flows from its headwaters in Fairfax County, Virginia into the Potomac River bordering Maryland. The Sugarland watershed is located in the suburbs of Washington D.C. (39° 3′ 31.438″ N; 77° 22′ 3.399″ W) and has undergone extensive deforestation for urban land uses (14 % forested, >45 % impervious surfaces). Sugarland Run basin occupies a total area of 58 km2 and has a regional population of approximately 1.1 million residents (USCB 2013). Sugarland Run is within the Piedmont physiographic region of Virginia, characterized by moderate to high gradient valleys, generally pool-riffle systems with cobble-gravel channel substrates (Hupp et al. 2013). The soils in this region are mapped as the Rowland series (fine-loamy, mixed, superactive, mesic Fluvaquentic Dystrudepts), Bowmansville series (fine-loamy, mixed, active, nonacid, mesic Fluventic Endoaquepts), and the Lindside-Kinkora-Delanco complex (Natural Resources Conservation Service 2015). Extensive land clearing for historical agriculture in the Sugarland basin resulted in upland erosion and sedimentation downstream. Floodplain legacy sediment deposits can be seen in the stream bank morphologies, where a precolonial gravel layer is overlain by at least 200 cm of fine-grained post-colonial alluvium (Hupp et al. 2013). Following the colonial agrarian era, a period of rapid urbanization lasted from the 1960s until the 1990s, before the Chesapeake Bay Preservation Act required the creation of Resource Protection Areas (RPAs). Rapid urbanization has contributed to the limited riparian buffer in many locations (Phillips 2005; Gellis et al. 2009).

Methods

Field sampling

Storm flow water samples were collected by WS750 auto samplers with external rain gauge sensors. One automatic water sampler was installed in each basin, near the mouth of Sugarland Run at station 1 and the lower most accessible location at station 4 in the Ni River basin (Fig. 1). The locations were chosen to capture representative samples of the total runoff during storm events in both basins. Low flow samples were also periodically collected manually at all stations in the two basins between January 2013 and March 2014.

Floodplain and riparian soil cores were collected using a 30-cm-long, 5-cm-diameter auger and slide-hammer between June 2013 and February 2014. The soil cores were taken from lateral transects perpendicular to the stream channel at distances from the bank of between 5 and 20 m. The majority of the cores were taken at the 5 and 10 m distance from the stream because of the limited and variable width of the active floodplain. A total of 29 soil cores per basin were sampled at the Ni River and Sugarland Run basins. The cores were subsampled in 10 cm increments before drying and sieving.

Laboratory analyses

Stream water P and soil water soluble P (WSP) were analyzed using the molybdenum blue reaction method detailed by Murphey and Riley (1962) and Hooda et al. (2000). For soil WSP analysis, a 1:10 soil-to-water ratio was shaken for 1 h, then centrifuged for 10 min at 4000 rpm before the molybdenum blue analysis. Samples were scanned using a 5-cm path length at 885.0 nm using a Perkin-Elmer UV–Vis Spectrophotometer. Standard calibration curves were then used to determine total P concentrations.

Sub-samples of soil core and stream bank soil samples were oven dried, disaggregated with mortar and pestle, sieved through a 2-mm screen, and then sent to A&L Laboratory in Richmond, Virginia for Mehlich-3 analysis. The soil samples were analyzed for total available P, and exchangeable cations including K, Mg, Ca, H+, Mn, Cu, B, Al, and Fe using the Mehlich-3 extraction method (Mehlich 1984; McDowell et al. 2003). Molar concentrations of Mehlich-3 extracted Al, Fe, and P were used to calculate the degree of P saturation (DPS) as described by Nair et al. (2004).
$$ \mathrm{D}\mathrm{P}\mathrm{S}\ \left(\%\right)=\frac{\left[\mathrm{P}\right]}{\left[\mathrm{F}\mathrm{e}+\mathrm{A}\mathrm{l}\right]} \times 100 $$

The remaining sub-samples were used to determine SOM content using loss on ignition technique as described by Young et al. (2013). Bulk density was calculated using oven-dried mass (105 °C) of soil samples of known volume (Blake and Hartge 1986).

The soil P storage capacity (SPSC) was calculated using the following after Nair (2014):
$$ \mathrm{SPSC} = \left(\mathrm{Threshold}\ \mathrm{P}\mathrm{S}\mathrm{R}\ \hbox{--}\ \mathrm{S}\mathrm{oil}\ \mathrm{P}\mathrm{S}\mathrm{R}\right) \times \left[\mathrm{F}\mathrm{e} + \mathrm{A}\mathrm{l}\right] \times 31\ \mathrm{mg}/\mathrm{kg} $$

Where threshold PSR is assumed as a constant of 0.1 and soil PSR is equal to DPS calculated above (Chakraborty et al. 2012). The soil SPSC value indicates whether a soil is a net sink (positive) or source (negative) of P.

Stream bank P fluxes

The total P contribution from stream bank erosion was estimated from soil P concentrations and net bank erosion rates. Bank erosion rates were estimated using standard erosion pin methods (Couper et al. 2002; Curran 2005; Julian and Torres 2006; Pope and Odhiambo 2014). These methods use a combination of pins consisting of rebar, approximately 1 cm in diameter and lengths of either 71 or 122 cm. The length of rebar chosen for each pin placement was determined based on the size of the stream bank, the soil texture, and depth to bedrock. Pins were placed approximately 30 cm apart on the bank face. Three sites were selected in each basin based on surrounding slope, soil erodibility, stream competence, and the nature of the stream bank (cut banks). The three sites in each basin had bank heights of >100 cm except for the lower Ni River site (close to the Ni reservoir deltaic zone) where the bank was only 50 cm in height. The changes in exposed pin lengths were monitored at 2-month intervals during the study period from January 2013 to March 2014. In addition, for the Ni basin, we relied on a combination of data from previous studies (Pope and Odhiambo 2014) and new sets of erosion data collected from all stations, plus a new lower station 4, between November 2009 and March 2014. Stream bank soil samples (for P analyses) were also collected at the pin locations based on soil horizon boundaries above the previously described pre-colonial gravel layers to represent resuspension of legacy sediment deposits. The volume and bulk density of the soil at the erosion pin stations were used to calculate the mean mass of material eroded per year. The total length of cut banks, width, and average bank heights affected by erosion were measured for the bank sediment loss volume calculations. The stream sinuosity and eroding unvegetated cut-bank lengths were obtained via orthoimagery and field measurements.

Statistical analyses

Descriptive statistical analyses (mean, standard errors) were calculated using SigmaPlot 12.5 (Systat Software Inc., San Jose, CA). Linear regression analyses were used to test relationships among two variables, such as total soil P and SOM or DPS and WSP. Comparison of mean soil properties (total SOM, Fe, Al, P) among study basins (Sugarland and Ni) were performed using paired Student’s t tests. All tests were considered statistically significant at α = 0.05.

Results and discussion

Stream water total phosphorous

The stream water total P results for both Sugarland Run and Ni River basins are shown in Fig. 2. The data shows that during low flow conditions Sugarland Run total P ranged from 0.002 to 0.11 ppm, with an average of 0.037 ppm. During low flow conditions, the upstream stations (3 and 4) of Sugarland Run had the highest average P concentrations of 0.063 and 0.065 ppm, respectively (Fig. 1), whereas station 1 located at the mouth of Sugarland Run had the lowest concentrations of total P during the study period, with an average of 0.027 ppm. The storm flow data at Sugarland ranged from 0.002 to 0.20 ppm with an average of 0.068 ppm (Fig. 2). In Ni River, the total P during low flow conditions ranged from 0 to 0.024 ppm with an average of 0.006 ppm, whereas high flow values were between 0 and 0.056 ppm with an average of 0.011 ppm. The highest P values in Ni River were recorded at the downstream station 4 site. The average stormflow sample total P from both Sugarland Run and the Ni River were significantly greater (p ≤ 0.05) than samples collected during low flow periods (Fig. 3). These data suggest that stream flow P concentrations are influenced by runoff in the two study basins.
Fig. 2

Discharge time series from closest stream gage stations (USGS 2014) and stream water total phosphorous concentrations during low and high flow conditions. The arrows indicate collections during flood events collected at station 1 of Sugarland Run and station 3 of the Ni River. Low flow samples were collected at all stations of the two basins throughout the study period

Fig. 3

Average total P concentrations (n = 25 per basin) in stream water during high (flood) and low flow sampling periods. Means within each stream with different letters are significantly different according to Student’s t tests

The total stream water P results in the two basins most likely reflect differences in the LULC characteristics. Sugarland basin is an urban stream with mostly residential housing and >45 % impervious surfaces and, thus, higher overall values of storm water P compared to the Ni basin. Urban streams have been shown to receive significant amount of P from municipal and industrial wastewaters in addition to non-point sources and thus tend to have P values comparable to agricultural dominated basins (Bassi et al. 2014; Duan et al. 2012). Because of the human developments, the urban basin also contains a well-developed storm drain system which minimizes the contact between soil surfaces and runoff, hence the minimal role of soils in decreasing P through adsorption onto mineral or organic surfaces during storm events. In contrast, the Ni basin contains >75 % forest cover and thus likely facilitates higher retention of P in the basin and lower overall values in the stream water (Fig. 2).

Floodplain soil properties

Table 1 summarizes the overall riparian/floodplain soil chemical characteristics and Mehlich-3 extracted total P from samples within the two basins. Soil pH averages ranged from 5.15 ± 0.15 in the subsoil to 5.69 ± 0.09 in the surface horizons of Sugarland, whereas in the Ni basin the floodplain soil pH ranged from 4.22 ± 0.07 in the subsoil to 4.45 ± 0.09 in the surface horizons. The average percent SOM in Ni basin was numerically lower near the stream (3.58 ± 0.48 %) versus further from the channel (6.67 ± 2.5 %), whereas in Sugarland percent SOM was higher near the stream (5.47 ± 1.3) compared to 3.61 ± 0.39 further from the channel. The more incised stream beds and reduced channel-floodplain connectivity associated with urban streams, like Sugarland Run, tend to minimize flooding occurrence, overbank sedimentation, and thus allows for accumulation of relatively more SOM closer to the active stream channel (Doll et al. 2002; Groffman et al. 2003; Pickett et al. 2008; Hardison et al. 2009). By contrast, the Ni River floodplain contained more mineral soil materials closer to the stream, which likely reflects frequent flooding and minimal buildup of SOM in the soils adjacent to the active channel. Overall, there was no significant difference in total SOM among the two study basins (Fig. 4).
Table 1

Mean soil properties (n = 10, ± 1 standard error) by depth and distance from stream channel in Sugarland Run and Ni River

  

Soil properties

P sorption sites

 

Depth

Distance

pH

SOM

CEC

Ca

Mg

Fe

Al

Total P

cm

m

 

%

cmol/kg

ppm

ppm

Sugarland Run

0–10

5

5.67 ± 0.09

4.88 ± 0.45

14.79 ± 1.0

1809 ± 147

244 ± 29

295 ± 20

550 ± 35

26.4 ± 3.1

10–20

5.48 ± 0.14

7.20 ± 3.4

14.10 ± 1.1

1585 ± 140

236 ± 40

272 ± 23

647 ± 31

19.5 ± 3.1

20–30

5.15 ± 0.15

4.25 ± 1.5

14.50 ± 1.5

1368 ± 150

203 ± 49

290 ± 26

696 ± 50

17.9 ± 1.4

Mean

 

5.44 ± 0.08

5.47 ± 1.3

14.46 ± 0.69

1592 ± 87

228 ± 22

285 ± 13

629 ± 24

21.3 ± 1.6

0–10

10

5.69 ± 0.09

5.59 ± 0.57

15.86 ± 0.80

2076 ± 110

200 ± 18

336 ± 35

568 ± 28

32.1 ± 2.3

10–20

5.49 ± 0.14

3.25 ± 0.21

14.13 ± 1.1

1746 ± 190

169 ± 21

264 ± 22

655 ± 29

21.5 ± 4.3

20–30

5.40 ± 0.17

3.06 ± 0.37

14.29 ± 1.4

1745 ± 240

158 ± 26

257 ± 25

697 ± 38

19.0 ± 2.8

Mean

 

5.53 ± 0.08

3.96 ± 0.33

14.76 ± 0.65

1856 ± 109

176 ± 13

285 ± 17

640 ± 21

23.4 ± 1.9

0–10

20

5.35 ± 0.14

5.25 ± 0.76

14.80 ± 1.2

1692 ± 119

167 ± 6.2

407 ± 50

545 ± 20

34.7 ± 5.4

10–20

5.48 ± 0.06

3.00 ± 0.32

13.52 ± 1.1

1663 ± 168

169 ± 7.2

285 ± 25

627 ± 19

24.8 ± 3.3

20–30

5.52 ± 0.09

2.58 ± 0.28

14.28 ± 1.7

1794 ± 262

169 ± 7.2

244 ± 18

662 ± 17

20.5 ± 2.2

Mean

 

5.45 ± 0.06

3.61 ± 0.39

14.20 ± 0.75

1717 ± 105

168 ± 3.7

312 ± 25

611 ± 16

26.7 ± 2.5

Ni River

0–10

5

4.45 ± 0.06

4.74 ± 0.23

6.52 ± 0.54

327 ± 38

93.7 ± 11

407 ± 29

638 ± 29

16.2 ± 1.9

10–20

4.22 ± 0.07

4.03 ± 1.3

4.70 ± 0.68

143 ± 21

59.1 ± 10

338 ± 26

776 ± 29

6.4 ± 1.2

20–30

4.24 ± 0.07

1.86 ± 0.07

4.47 ± 0.76

141 ± 21

60.9 ± 12

262 ± 19

780 ± 23

3.7 ± 0.9

Mean

 

4.30 ± 0.04

3.58 ± 0.48

5.23 ± 0.39

204 ± 20

71.2 ± 6.6

337 ± 16

731 ± 18

8.8 ± 1.1

0–10

10

4.45 ± 0.09

8.57 ± 3.1

6.42 ± 0.73

330 ± 71

92.2 ± 13

413 ± 36

653 ± 32

22.4 ± 2.8

10–20

4.31 ± 0.08

3.61 ± 0.73

4.65 ± 0.84

192 ± 53

62.0 ± 15

316 ± 29

739 ± 33

9.0 ± 2.8

20–30

4.28 ± 0.08

2.21 ± 0.28

4.43 ± 0.89

152 ± 47

66.0 ± 17

258 ± 29

770 ± 34

3.1 ± 0.8

Mean

 

4.34 ± 0.05

4.79 ± 1.1

5.16 ± 0.49

225 ± 35

73.4 ± 8.9

329 ± 21

721 ± 20

11.5 ± 1.9

0–10

20

4.10 ± 0.00

7.15 ± 1.7

5.10 ± 0.60

111 ± 1.0

63 ± 14

533 ± 140

729 ± 12

39.5 ± 21

10–20

4.25 ± 0.05

10.7 ± 7.4

3.90 ± 1.9

86 ± 31

64 ± 37

348 ± 44

753 ± 135

24.5 ± 14

20–30

4.95 ± 0.65

2.15 ± 1.2

5.30 ± 3.3

222 ± 170

239 ± 212

235 ± 70

669 ± 220

21.5 ± 17

Mean

 

4.43 ± 0.24

6.67 ± 2.5

4.77 ± 1.0

140 ± 51

122 ± 67

372 ± 69

717 ± 68

28.5 ± 8.4

Fig. 4

Mean values (n = 29 per basin, ± 1 standard error) of Fe, Al, P, and soil organic matter in Sugarland Run and Ni River floodplain soils. Means with different letters are significantly different according to paired Student’s t tests

Nearly all extractable soil elements in the two basins were higher in the surface horizons when compared to the subsoil samples (Table 1). Soil Fe and Al values were significantly greater in the Ni floodplain (Fig. 4), which suggests there are more available sites for P sorption within the Ni floodplain. Both Ca and Mg concentrations were significantly higher in Sugarland compared to the Ni River soils, with Ca showing the largest difference between the two basins, 1856 ± 109 versus 330 ± 71 ppm average values (Table 1). These findings are similar to other regional urban settings (Baltimore, Maryland) where soil pH and Ca concentrations are typically elevated due to leaching of cations from concrete infrastructure at watershed scales (Pouyat et al. 2007). Cation exchange capacity (CEC) was higher in the near-surface soil samples in both basins, which was correlated to higher SOM in the upper horizons. CEC was significantly higher in the Sugarland floodplain, more than twice that of the Ni River alluvial soils, which is likely a function of higher Ca and Mg concentrations (Table 1).

The Mehlich-3 extracted total P was significantly greater in the Sugarland Run floodplain soils compared to those from Ni River (Fig. 4). Total P in the Sugarland Run soils ranged from 17.9 ± 1.4 to 34.7 ± 5.4 ppm (Table 1), whereas samples from the Ni River floodplain ranged from as low as 3.1 ± 0.8 to anomalously higher values of 39.5 ± 21 ppm total P. The soil total P had no observable variations with distance from the stream in Sugarland, whereas in Ni there was an increase in total soil P with distance from the stream (Table 1). Longitudinally, soil P values were typically high in the lower reaches of Sugarland Run, whereas in the Ni basin the upper and middle headwater reaches had slightly higher values of soil total P than the lower wetland soils. Soil total P in both basins showed significant correlations with SOM (Fig. 5), with the strongest positive relationship evident in the surface horizons (<10 cm depth). Water-soluble P (WSP) was also greater in Sugarland Run compared to Ni River alluvial soils. WSP in Sugarland Run averaged 0.179 ppm (range from 0.025 to 1.23 ppm) whereas the Ni River basin soil WSP averaged 0.109 ppm (range from 0 (below detection limit) to 0.957 ppm).
Fig. 5

Soil organic matter (SOM) at various depths of the soil profile versus soil total P content. Samples with >20 % SOM were omitted because our primary focus is mineral riparian soils. Soil samples from the upper 10 cm showed the strongest relationship between organic matter and total P concentration. Although the Sugarland Run soils have a higher SOM and total P concentrations (Table 1), the Ni River has a stronger relationship (p < 0.001) between the two variables

Degree of phosphorous saturation (DPS) is commonly used to analyze P saturation levels in any given soil (Hooda et al. 2000; Nair et al. 2004; Dunne et al. 2006). Measurement of DPS relates P that is adsorbed in the soil to its adsorption capacity and is typically a function of extractable Fe and Al contents (Dunne et al. 2006). Mehlich-3 extracted Al, Fe, and P were used to calculate the soil DPS for both study basins (Fig. 6). None of the calculated values exceeded an empirical limit for P saturation >16 %, where soil P sorption becomes limited (Nair et al. 2004). Two soil samples from the mouth of Sugarland Run were greater than 10 % saturated with P (S4C10 and S3C10), whereas Ni River had one sample that exceeded 10 % (N21A1, located near the mouth of the river). Average Ni River floodplain soil DPS was 2.34 % and Sugarland Run samples averaged 5.37 %. Samples that had >20 % SOM (organic soil horizons) were not included in the analysis because DPS values are related to mineral soil constituents, not P bound to SOM. As expected, WSP and DPS were significantly correlated (p < 0.05, Fig. 6), demonstrating that higher P saturation in the soil contributes to increased P mobility. In addition, estimates of soil P storage capacity index (SPSC) for the two basins were highly positive, with Ni River soils having significantly (p < 0.001) greater SPSC (2959 ± 262) compared to Sugarland Run (1366 ± 402). These estimates suggest that the floodplains of both basins are strong sinks of P, but the Ni River soils have more available sites for P adsorption (Chakraborty et al. 2012; Nair 2014).
Fig. 6

Mehlich-3° of phosphorous saturation (DPS) versus water soluble P concentrations at different soil depths in Sugarland Run and the Ni River basins. The Ni River had a lower average DPS at 2.3 % and Sugarland Run average 5.4 %. Two samples from near the mouth of Sugarland and one sample from the Ni River exceeded the 10 % P saturation, yet none of the samples exceeded the suggested threshold of 16 % Mehlich-3 DPS for Atlantic Coastal Plain soils (Nair et al. 2004)

Overall, the stream water and alluvial soil sample results show that total stream P, WSP, and soil DPS are higher in the urban Sugarland Run basin compared to the predominantly forested Ni River. These differences are likely a result of both the current and historical LULC as well as differences in soil characteristics among the two basins. Although the Sugarland watershed is currently urban, most of the Chesapeake Bay drainage was dominated by agricultural land-uses from the colonial era (late 1600s) until the 1970s (Hupp et al. 2013). Therefore, P-rich legacy sediment deposits still occupy the floodplains of many regional stream systems. The significant role of LULC and soils in P dynamics is evident in the relationship between total P and the concentrations of Fe, Al, and SOM which are important controls in P retention.

Soil DPS estimates were a good predictor for WSP concentrations, but SOM content is also likely a factor in binding total P, although its importance varies (Darke and Walbridge 2000). The more human impacted Sugarland basin has a faster response time during storm events, there is a rapid increase in both discharge and WSP compared to the forested Ni basin. The limited lateral width of the Sugarland riparian forest may minimize the effectiveness of alluvial soils to reduce P fluxes to the stream. Sugarland floodplain soils had lower concentrations of Fe and Al and more bound-P, which likely results in less available sites for P adsorption. In contrast, the Ni River had significantly higher concentrations of soil Fe and Al, with less antecedent bound-P. The Ni River also has a large intact forested floodplain, which may also play a role in limiting total P flux in this system. These factors suggest that the Ni River watershed has a greater capacity for water quality improvement and adsorption of P in the future.

Stream bank erosion and phosphorus fluxes

The stream bank erosion results for both basins are shown in Table 2. The results show the changes in bank depth at different heights of the exposed face, the bottom height being the pin just below the normal summer stream flow levels, the middle, and top bank heights. As detailed by Pope and Odhiambo (2014), the average bank erosion rate at Ni River stations 2 and 3 was 6.93 and 10.27 cm year−1, respectively. These erosion rates were similar to the more recent measurements from this study. The lower station 4 (added in this study) had a relatively shallow bank height of approximately 50 cm with an average erosion rate of 6.13 cm year−1 based on two pins (Table 2). The average bank erosion in Sugarland stations 1, 2, and 3 were 9.35, 9.00, and 11.17 cm year−1, respectively. The rates were slightly higher in the upper slope faces at all three locations, with no measurable changes at the bottom pins. At all sites, bank morphology consisted of fine-grained eroding layers above basal gravel deposits. The thickness of eroding fine-grained alluvium ranged from 117 cm at station 3 in the upstream to 198 cm in station near the river mouth (Table 2). Bank erosion in both basins is dominated by diffuse particle erosion, with a few incidences of slope failures (slumps) in bank faces with pronounced variation in erosion rates. In addition, well-vegetated stream reaches contain plant roots that stabilize and slow erosion on the upper bank face, creating undercuts below the near surface (approximately upper 30 cm) which can lead to occasional bank failures. The average bank erosion from all the monitoring sites in each basin was 7.98 and 9.84 cm year−1 for Ni River and Sugarland Run, respectively.
Table 2

Summary of stream bank characteristics used to calculate total annual bank erosion rates and P fluxes for Sugarland Run and Ni River

Station

Bank height

Total phosphorous

Bank erosion

(cm)

(ppm)

(cm year−1)

Sugarland Run

St-1

30

17.0

14.05

112

17.0

14.00

198

17.0

0.00

St-2

15

22.0

15.00

74

28.0

12.00

120

11.0

0.00

St-3

23

17.0

20.00

74

19.0

13.50

117

22.0

0.00

Mean

85

18.9

9.84

Ni River

St-2

26

17.0

4.70

57

3.00

9.60

94

3.00

6.50

St-3

20

1.00

9.80

60

14.0

9.70

95

3.00

11.30

St-4

17

22.0

7.50

46

8.00

4.75

Mean

52

8.88

7.98

The rate of bank sediment fluxes for Ni River was previously estimated to be 518 Mg year−1 (Pope and Odhiambo 2014). Inclusion of new data from the lower station 4 in this study resulted in a slightly lower estimate of 407 Mg year−1, whereas stream bank sediment fluxes for Sugarland Run were estimated at 6141 Mg year−1, an order of magnitude higher than Ni fluxes due to greater total cut-bank lengths, higher average stream bank height, and faster bank erosion rates (Table 2). The average total soil P for Ni River and Sugarland River banks were 8.88 and 18.9 ppm, respectively. Therefore, estimates of total P flux rates from bank erosion were approximately 361 and 11,600 kg year−1 for Ni River and Sugarland Run, respectively. The stream water P contributions from bank erosion are mostly accounted for in competent systems (like Sugarland Run) where most sediments are transported in suspension and are captured during total P water sampling. However, in most forested or low competent streams (like Ni River), the majority of eroded sediments are transported as bed-load and a significant amount of total P from bank erosion may not be accounted for in P budgets that are calculated from suspended water samples.

Regardless of basin scales, there are site-specific factors that tend to influence the retention or release P from riparian floodplain soils to streams. These factors, among others, include basin LULC, flow paths, soil redox conditions, P loading rate, water residence time, sedimentation/resuspension, channel shape, and stream discharge/water levels (Darke and Walbridge 2000; McDowell et al. 2003; Binkley et al. 2004; Dunne et al. 2006; Williamson et al. 2014). However, it is also important to note that all of these factors are in some way related to LULC and soil characteristics. In urban and other human-modified basins, the prevalence of impervious surfaces and abundance of direct flow paths to streams can have both negative and positive impacts on stream P levels. In many urban areas with well-developed storm drain systems, water-soluble P has less contact or low residence time in the soil environment, lessening the role of floodplains and wetlands to remove nutrients from storm water. In addition, high discharge and velocity associated with urban runoff tend to enhance sediment re-suspension, stream bank erosion, and bed-incision. Thus, increased fluxes and transport of sediment-bound P can occur. Stream P dynamics can therefore vary in urban watersheds, depending on land management schemes including maintenance of riparian buffers, storm drain system networks, and the associated efficiency of runoff transfer to the local streams.

The results of this study show a link between LULC and total P in both stream water and floodplain soils, with the more urbanized Sugarland Run showing greater total P values than the forested Ni River (Figs. 3 and 4). These results are not unique in this region as modern environmental evolution of mid-Atlantic streams in both the Piedmont and Coastal Plain of the eastern USA have been shown to be linked to variability in land use, soils, and highly erodible valley fill legacy sediments (Walter and Merritts 2008; Niemitz et al. 2013; Schenk et al. 2013; Donovan et al. 2015; Smith and Wilcock 2015). Post-European settlement era sediments (legacy sediments) are often associated with past forest clearing, poor agricultural practices, and channel modifications. Legacy sediment deposits have been identified as a significant source of modern sediment, pollutant, and nutrient fluxes to surface water bodies in the greater Chesapeake Bay region (Noe and Hupp 2005; Hupp et al. 2013). The increased destabilization and erosion of post-colonial alluvium are exacerbated by modern land use management and urbanization that has shifted channel processes from aggrading to degrading (Schenk et al. 2013). The stream bank erosion results from both Sugarland and Ni floodplains are comparable to other regional estimates that range from 1.9 to 22.5 cm/year in Difficult Run (Hupp et al. 2013). Schenk and Hupp (2009) also estimated bank erosion rates of 2.9 to 16.3 cm/year for Little Conestoga Creek in Pennsylvania, and Donovan et al. (2015) reported average annual stream lateral migration rates range from 4 to 19 cm/year in Piedmont streams of Baltimore County, Maryland. The high bank erosion rates from these studies were attributed to erodible legacy sediments on the floodplains of most streams in this region. Recognition of these trends has resulted in the implementation of various watershed management practices in this region, such as the expansion of riparian buffers to curb P and sediment fluxes (Jones et al. 2001). However, in many cases, the intrinsic degradation associated with urbanization (high runoff, stream discharge) continues to partially negate these efforts.

Conclusions

Nutrient cycles and floodplain dynamics are complex processes controlled by a number of mechanisms that require an understanding of surficial processes, as well as internal stream dynamics and interactions. This study explored phosphorous dynamics in two varying floodplain environments to evaluate the potential sources of P to surface waters. The floodplain of the Ni River has large P sorption capacity due to limited disturbances of the forested floodplain and abundant soil Fe and Al which may minimize delivery of WSP to the stream. In contrast, anthropogenic alterations have reduced the floodplain width of Sugarland Run and the soils have significantly less available absorption sites for P. In addition, poor storm water management in the region has also increased delivery of sediment to the stream via increased net bank erosion, which likely increases total stream P flux by resuspending floodplain legacy sediments that were formerly a sink for soil-bound P.

Notes

Acknowledgement

This project was funded by the university of Mary Washington, Fredericksburg, Virginia, USA. The authors would also like to thank Taylor Coxon for her assistance in field sampling and laboratory analysis.

References

  1. Axt JR, Walbridge MR (1999) Phosphate removal capacity of palustrine forested wetlands and adjacent uplands in Virginia. Soil Sci Soc Am J 63:1019–1031CrossRefGoogle Scholar
  2. Bassi N, Kumar MD, Sharma A, Pardha-Saradhi P (2014) Status of wetlands in India: a review of extent, ecosystem benefits, threats and management strategies. J Hydrol: Reg Stud 2:1–19Google Scholar
  3. Binkley D, Ice GG, Kaye J, Williams CA (2004) Nitrogen and phosphorous concentration in forest streams of the United States. J Am Water Resour Assoc 40:1277–1291CrossRefGoogle Scholar
  4. Blake GR, Hartge K (1986) Bulk density. In: Klute A (ed) Methods of soil analysis. Part 1. Physical and mineralogical methods, 2nd edn. Agronomy monograph 9. ASA and SSSA, Wisconsin, pp 363–375Google Scholar
  5. Butt AJ, Brown BL (2000) The cost of nutrient reduction: a case study of Chesapeake Bay. Coast Manag 28:175–185CrossRefGoogle Scholar
  6. Chakraborty D, Nair VD, Harris WG, Rhue RD (2012) Environmentally relevant phosphorus retention capacity of sandy Coastal Plain soils. Soil Sci 177:701–707CrossRefGoogle Scholar
  7. Cooper SR, Brush GS (1993) A 2500-year history of anoxia and eutrophication in Chesapeake Bay. Estuaries 16:617–626CrossRefGoogle Scholar
  8. Couper P, Scott T, Maddock I (2002) Insights into river bank erosion processes derived from analysis of negative erosion-pin recordings: observations from three recent UK studies. Earth Surf Process Landf 27:59–79CrossRefGoogle Scholar
  9. Curran J (2005) Baseline channel morphology and bank erosion inventory of South Fork Campbell Creek at Campbell Tract, Anchorage, Alaska, 1999 and 2000 (U.S. Geological Survey Open-File Report 01–288)Google Scholar
  10. Darke AK, Walbridge MR (2000) Al and Fe biogeochemistry in a floodplain forest: implications for P retention. Biogeochemistry 51:1–32CrossRefGoogle Scholar
  11. Doll BA, Wise-Frederick DE, Buckner CM, Wilkerson SD, Harman WA, Smith RE, Spooner J (2002) Hydraulic geometry reletionships for urban streams throughout the Piedmont of North Carolina. J Am Water Resour Assoc 38(3):641–651Google Scholar
  12. Donovan M, Miller A, Baker M, Gellis A (2015) Sediment contributions from floodplains and legacy sediments to Piedmont streams of Baltimore County, Maryland. Geomorphology 235:88–105CrossRefGoogle Scholar
  13. Duan S, Kaushal SS, Groffman PM, Band LE, Belt KT (2012) Phosphorous export across an urban to rural gradient in the Chesapeake Bay watershed. J Geophys Res 117, GO1025. doi:10.1029/2011JG001782
  14. Dunne EJ, Reddy R, Clark MW (2006) Biogeochemical indices of phosphorous retention and release by wetland soils and adjacent stream sediments. Wetlands 26:1026–1041CrossRefGoogle Scholar
  15. Gellis AC, Hupp CR, Pavich MJ, Landwehr JM, Banks WSL, Hubbard BE, Langland MJ, Ritchie JC, Reuter JM (2009) Sources, transport, and storage of sediment in the Chesapeake Bay Watershed: U.S. Geological Survey Scientific Investigations Report 2008–5186., p 95Google Scholar
  16. Groffman PM, Bain DJ, Band LE, Belt KT, Brush GS, Grove JM, Pouyat RV, Yesilonis IC, Zipperer WC (2003) Down by the riverside: urban riparian ecology. Front Ecol Environ 1(6):315–321Google Scholar
  17. Hardison EC, O’Driscoll MA, Deloatch JP, Howard RJ, Brinson MM (2009) Urban land use, channel incision, and water table decline along coastal plain streams, North Carolina. J Am Water Resour Assoc 45(4):1032–1046Google Scholar
  18. Hooda PS, Rendell AR, Edwards AC, Withers PJA, Aitken MN (2000) Relating soil phosphorus indices to potential phosphorus release to water. J Environ Qual 29:1166–1171CrossRefGoogle Scholar
  19. Hupp CR, Noe GB, Schenk ER, Benthem AJ (2013) Recent and historic sediment dynamics along difficult run, a suburban Virginia Piedmont stream. Geomorphology 180–181:156–159CrossRefGoogle Scholar
  20. Jones KB, Neale AC, Nash MS, Van Remortel RD, Wickham JD, Riitters KH, O’Neill RV (2001) Predicting nutrient and sediment loadings to streams from landscape metrics: a multiple watershed study from the United States Mid-Atlantic Region. Landsc Ecol 16:301–312CrossRefGoogle Scholar
  21. Julian JP, Torres R (2006) Hydraulic erosion of cohesive riverbanks. Geomorphology 76:193–206CrossRefGoogle Scholar
  22. Lai DYF, Lam KC (2008) Phosphorus retention and release by sediments in the eutrophic Mai Po Marshes, Hong Kong. Mar Pollut Bull 57:349–356CrossRefGoogle Scholar
  23. McDowell RW, Sharpley AN, Folmar G (2003) Modification of phosphorous export from an eastern USA catchment by fluvial sediment and phosphorous input. Agric Ecosyst Environ 99:187–199CrossRefGoogle Scholar
  24. Mehlich A (1984) Mehlich-3 soil test extractant: a modification of mehlich-2 extractant. Commun Soil Sci Plant Anal 15:1409–1416CrossRefGoogle Scholar
  25. Murphey J, Riley JP (1962) A modified single solution method for the determination of phosphate in natural waters. Anal Chim Acta 27:31–36CrossRefGoogle Scholar
  26. Nair VD (2014) Soil phosphorous saturation ratio for risk assessment in land use systems. Frontiers in Environmental Science, Vol. 2, Art. 6. Doi 10.3389/fenvs.2014.00006
  27. Nair VD, Portier KM, Graetz DA, Walker ML (2004) An environmental threshold for degree of phosphorus saturation in sandy soils. J Environ Qual 33:107–113CrossRefGoogle Scholar
  28. Natural Resources Conservation Service (2015) United States Department of Agriculture. Web Soil Survey. http://websoilsurvey.nrcs.usda.gov/. Accessed May 2015
  29. Niemitz J, Haynes C, Lasher G (2013) Legacy sediments and historic land use: chemostratigraphic evidence for excess nutrient and heavy metal sources and remobilization. Geology 41:47–50CrossRefGoogle Scholar
  30. Noe GB, Hupp CR (2005) Carbon, nitrogen and phosphorous accumulation in floodplains of the Atlantic coastal plain rivers, USA. Ecol Appl 14:1178–1190CrossRefGoogle Scholar
  31. Phillips SW (2005) The U.S. Geological Survey Chesapeake Bay science plan, 2006-2011: U.S. Geological Survey Open File Report 2005-1440., p 53Google Scholar
  32. Pickett STA, Cadenasso ML, Grove JM, Groffman PM, Band LE, Boone CG, Burch WR, Grimmond CSB, Hom J, Jenkins JC, Law NL, Nilon CH, Pouyat RV, Szlavecz K, Warren PS, Wilson MA (2008) Beyond urban legends: an emerging framework of urban ecology, as illustrated by the Baltimore ecosystem study. BioScience 58(2):139–150Google Scholar
  33. Pope IC, Odhiambo BK (2014) Soil erosion and sediment fluxes analysis: a watershed study of the Ni Reservoir, Spotsylvania County, VA, USA. Environ Monit Assess 186:1719–1733CrossRefGoogle Scholar
  34. Pouyat RV, Yesilonis ID, Russell-Anelli J, Neerchal NK (2007) Soil chemical and physical properties that differentiate urban land-use and cover types. Soil Sci Soc Am J 71:1010–1019CrossRefGoogle Scholar
  35. Ricker MC, Lockaby BG (2014) Soil biogeochemical processes across a lateral toposequence in an old-growth floodplain forest. Soil Sci Soc Am J 78:2100–2111CrossRefGoogle Scholar
  36. Ricker MC, Odhiambo BK, Church JM (2008) Spatial analysis of soil erosion and sediment fluxes: a paired watershed study of two Rappahannock River tributaries, Stafford County, Virginia. Environ Manag 41:766–788CrossRefGoogle Scholar
  37. Schenk ER, Hupp CR (2009) Legacy effects of colonial millponds on floodplain sedimentation, bank erosion, and channel morphology, Mid-Atlantic, USA. J Am Water Resour Assoc 45:597–606CrossRefGoogle Scholar
  38. Schenk ER, Hupp CR, Gellis A, Noe G (2013) Developing a new stream metric for comparing stream function using a bank-floodplain sediment budget: a case study of three piedmont streams. Earth Surf Process Landf 38:771–784CrossRefGoogle Scholar
  39. Smith SMC, Wilcock PR (2015) Upland sediment supply and its relation to watershed sediment delivery in the contemporary mid-Atlantic piedmont (USA). Geomorphology 232:33–46CrossRefGoogle Scholar
  40. Søndergaard M, Jensen JP, Jeppesen E (2003) Role of sediment and internal loading of phosphorus in shallow lakes. Hydrobiologia 506(1–3):135–145CrossRefGoogle Scholar
  41. USCB (2013) United States Census Bureau. http://www.Ceasus.gov. Accessed December 2014.
  42. USGS (2014) USGS surface-water daily data for the nation. United States Geological Survey, Washington D.C, http://waterwatch.usgs.gov. Accessed August 2014Google Scholar
  43. VADGIF (2003) Ni Reservoir. Virginia Department of Game and Inland Fisheries., http://www.dgif.virginia.gov/fishing/waterbodies. Accessed August 2014Google Scholar
  44. Walter RC, Merritts DJ (2008) Natural streams and the legacy of water-powered mills. Science 319:299–304CrossRefGoogle Scholar
  45. Williamson TN, Christensen VG, Richardson WB, Frey JW, Gellis AC, Kieta KA, Fitzpatrick FA (2014) Stream sediment sources in Midwest agricultural basins with land retirement along channel. J Environ Qual 43:1624–1634CrossRefGoogle Scholar
  46. Young EO, Ross DS, Cade-Menun BJ, Liu CW (2013) Phosphorus speciation in riparian soils: a phosphorus-31 nuclear magnetic resonance spectroscopy and enzyme hydrolysis study. Soil Sci Soc Am J 77:1636–1647CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2016

Authors and Affiliations

  • B. K. Odhiambo
    • 1
  • M. C. Ricker
    • 2
  • L. M. Le Blanc
    • 3
  • K. A. Moxey
    • 4
  1. 1.Department of Earth and Environmental SciencesUniversity of Mary WashingtonFredericksburgUSA
  2. 2.Department of Environmental, Geographical and Geological SciencesBloomsburg UniversityBloomsburgUSA
  3. 3.Department of Geosciences, 210 Petrie HallAuburn UniversityAuburnUSA
  4. 4.Department of Geography, 216 Pearson HallUniversity of DelawareNewarkUSA

Personalised recommendations