HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by van der Perk, M. Articles by Beven, K. J. Search for Related Content PubMed PubMed Citation Articles by van der Perk, M. Articles by Beven, K. J. Agricola Articles by van der Perk, M. Articles by Beven, K. J. Related Collections Watershed and Landscape Processes Phosphorus Spatial Distribution

TECHNICAL REPORTS

Landscape and Watershed Processes

Controls on Catchment-Scale Patterns of Phosphorus in Soil, Streambed Sediment, and Stream Water

^a Dep. of Physical Geography, Utrecht Univ., P.O. Box 80 115, 3508 TC Utrecht, The Netherlands
^b National Soil Resources Inst., Cranfield Univ., North Wyke Research Station, Okehampton, Devon EX20 2SB, UK
^c British Geological Survey, Keyworth, Nottingham NG12 5GG, UK
^d Cross Institute Programme for Sustainable Soil Function, Institute of Grassland and Environmental Research (IGER), North Wyke Research Station, Okehampton, Devon EX20 2SB, UK
^e Dep. of Environmental Science, Lancaster Univ., Lancaster, LA1 4YQ, UK

^* Corresponding author (m.vanderperk{at}geo.uu.nl)

Received for publication May 3, 2006.

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS SOIL PHOSPHORUS STREAMBED SEDIMENT PHOSPHORUS STREAM WATER PHOSPHORUS CONCLUSIONS REFERENCES

 
Many models of phosphorus (P) transfer at the catchment scale rely on input from generic databases including, amongst others, soil and land use maps. Spatially detailed geochemical data sets have the potential to improve the accuracy of the input parameters of catchment-scale nutrient transfer models. Furthermore, they enable the assessment of the utility of available, generic spatial data sets for the modeling and prediction of soil nutrient status and nutrient transfer at the catchment scale. This study aims to quantify the unique and joint contribution of soil and sediment properties, land cover, and point-source emissions to the spatial variation of P concentrations in soil, streambed sediments, and stream water at the scale of a medium-sized catchment. Soil parent material and soil chemical properties were identified as major factors controlling the catchment-scale spatial variation in soil total P and Olsen P concentrations. Soil type and land cover as derived from the generic spatial database explain 33.7% of the variation in soil total P concentrations and 17.4% of the variation in Olsen P concentrations. Streambed P concentrations are principally related to the major element concentrations in streambed sediment and P delivery from the hillslopes due to sediment erosion. During base flow conditions, the total phosphorus (<0.45 µm) concentrations in stream water are mainly controlled by the concentrations of P and the major elements in the streambed sediment.

Abbreviations: GIS, geographical information system • GPS, global positioning system • RP, reactive phosphorus • STW, sewage treatment work • TP, total phosphorus

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS SOIL PHOSPHORUS STREAMBED SEDIMENT PHOSPHORUS STREAM WATER PHOSPHORUS CONCLUSIONS REFERENCES

 
IN the past few decades, phosphorus (P) transfers in catchments have increasingly attracted the interest of earth and environmental scientists because of the potential for eutrophication of rivers, lakes, wetlands, estuaries, and coastal waters (Correll, 1998). It has long been recognized that P transport in and from catchments is controlled by climate, geology, topography, and anthropogenic influences, such as point-source discharges from sewage treatment works (STWs), industrial outfalls, and diffuse agricultural inputs (Dillon and Kirchner, 1975; Grobler and Silberbauer, 1985; McKee et al., 2001). As the human population is typically concentrated along rivers near the outlets of catchments, point-source inputs of P mainly affect higher order streams (e.g., Pieterse et al., 2003) and may contribute to more than 50% of the P export from populated catchments (e.g., Mourad et al., 2006). However, as P inputs from point sources have declined in many countries since the late 1980s, the scientific attention has focused on diffuse P losses from agricultural land.

Many studies have shown that enhanced dissolved reactive P (RP) runoff from agricultural soils is related to soil P status and the P buffering capacity of soils (Frossard et al., 2000; Hughes et al., 2000; Jordan et al., 2000; Quinton et al., 2003). Phosphorus is naturally present in primary soil minerals. In agricultural soils, large quantities of inorganic and organic P are added in the form of fertilizers and manures (Haygarth et al., 1998). Most of the inorganic P is fixed due to adsorption to soil sesquioxides (aluminum, iron, and manganese oxyhydroxides) and clay minerals or due to precipitation as hydroxyapatite (Ca₅(PO₄)₃OH). Consequently, a considerable proportion of P losses occur in particulate form. Quinton et al. (2003) demonstrated that soil P concentrations do not predict total P (TP) transport in overland flow very well, but they do when combined with suspended sediment concentrations in the overland flow. The extent to which P can reach the river network is closely related to topography and landscape connectivity (Helming et al., 2005). It is, therefore, plausible that areas of active soil erosion and near-stream areas contribute more to P transport from catchments than areas with low sediment transport rates further away from the river network. Gburek et al. (2000), for example, showed that in-stream P concentrations were more closely related to soils with high P concentrations in near-stream areas (within 60 m) than other areas of the catchment. However, the spatial delineation of the source areas of sediment and associated P during rainstorm events is known to be difficult as it relies on the assessment of factors such as the connectivity and sediment delivery from soils to the stream channel, which are highly variable in space and time (Beven et al., 2005; Page et al., 2005).

A considerable amount of sediment delivered to the stream network is temporarily stored in the channel bed (Owens et al., 2001). Consequently, the P content of streambed sediment may provide an integrated measure of upstream P loss from previous storm events (e.g., Jansson et al., 2000; Owens and Walling, 2002). Van der Perk et al. (2006) demonstrated that during base flow conditions, the P content of streambed sediments and other bed sediment geochemical properties exert an important influence on dissolved P concentrations in streams that have not been affected by point-source discharges. During high discharge conditions, much of the sediment and associated P is removed and transported further downstream. Therefore, the storage of P in bed sediment and the interactions between bed sediments and river water are important processes governing the P export from catchments and should be accounted for in catchment P budgets (Svendsen and Kronvang, 1993; House and Warwick, 1998; McDowell et al., 2001; Jarvie et al., 2005).

The majority of studies on P transfer have been performed at the plot scale. However, in recent years, various studies have attempted to scale observations of P transport from the plot scale to the catchment scale (e.g., Dougherty et al., 2004; Haygarth et al., 2005; Page et al., 2005). At the same time, various models have been developed and applied that predict long-term catchment P losses. These models range from simple conceptual models (e.g., PolFlow [De Wit, 2001] and the Phosphorus Indicators Tool [Heathwaite et al., 2003]) to more complex process-based models (e.g., CREAMS [Cooper et al., 1992] and SWAT2000 [Arnold and Fohrer, 2005]). The main input for these models is generally derived from readily available, generic spatial databases that contain, amongst others, soil maps, land use maps, and digital elevation models (DEMs). However, because of the complexity of catchment systems with respect to spatial variability of soil P status and hydrological processes, the spatial variation in the characteristics that control P loss is usually much greater than that represented by these data. This incomplete representation of these characteristics, together with our incomplete understanding of how they govern the transfer of P at the catchment scale, hampers the performance of these models.

In this study, we aim to quantify the unique and joint contribution of: (i) soil type, soil chemical properties, and land use, to the spatial variation of P concentrations in soil, (ii) soil P concentrations, sediment chemical properties, and point-source emissions to the spatial variation of P concentrations in streambed sediments, and (iii) sediment chemical properties, water chemistry, and point-source emissions to the spatial variation of P concentrations in stream water at the scale of a medium-sized catchment.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS SOIL PHOSPHORUS STREAMBED SEDIMENT PHOSPHORUS STREAM WATER PHOSPHORUS CONCLUSIONS REFERENCES

 
The Tamar catchment in southwest England was selected for this study because a spatially detailed geochemical data set has recently become available for it (Rawlins et al., 2003a). Such geochemical data sets have the potential to greatly enhance the input to catchment-scale nutrient transfer models. Moreover, they provide a means to assess the value and usefulness of available generic spatial data sets for modeling and prediction of soil nutrient status and nutrient transfer at the catchment scale. This is especially relevant for assessing the implications of the use of such generic spatial data for nutrient transfer modeling in data-poor catchments that are not extensively gauged or monitored.

Description of the Tamar Catchment
The Tamar catchment is situated in southwest England (50°30' N, 4°18' W) and covers an area of 976 km² with the River Tamar forming a natural county boundary between Devon and Cornwall (Fig. 1). The headwaters of the Tamar rise near the north coast of Devon and flow south along the western edge of Dartmoor before discharging into Plymouth Sound via the Tamar Estuary. The north of the catchment has an elevation of around 200 m above mean sea level (amsl), rising to the east (>500 m amsl on Dartmoor) and to the west (>300 m amsl on Bodmin Moor). Bedrock geology in the north of the catchment is predominantly interbedded sandstones and argillaceous (fine-grained) sedimentary rocks from the Carboniferous period. Further south, the Lower Carboniferous rocks are dominated by fine-grained sedimentary sequences and chert (crystalline silica). Outcrops of granite occur on the eastern (Dartmoor) and western (Bodmin Moor) sides of the catchment, and are also interspersed with outcrops of Lower Carboniferous and Devonian slates. According to the National Soil Map of England and Wales (NatMap), 23 soil associations can be identified in the Tamar catchment (Table 1). The soils in the Upper Tamar catchment are impermeable soils of the Bude Formation. South of Launceston, in the mid section, the soils (Crackington Formation) become more permeable. The upper reaches of the rivers Inny and Lyd are also on impermeable soils on northeastern Bodmin Moor or western Dartmoor, respectively.

View larger version (29K): [in this window] [in a new window]   Fig. 1. Location of the Tamar catchment.

Field Sampling and Laboratory Analysis
In September 2002, the British Geological Survey (BGS) performed a geochemical survey in the Tamar catchment (Rawlins et al., 2003a), which included the collection and analysis of between 460 and 495 samples of each of topsoil, stream water, and streambed sediment, which were distributed evenly over the catchment. The hydrological conditions in the Tamar catchment were relatively consistent throughout this month, with low flows ranging from 2.7 to 3.8 m³ s^–1 at the Gunnislake gauging station (mean flow 22.6 m³ s^–1). This suggests that variations in hydrological conditions were unlikely to have had a significant impact on the concentration of solutes between sites during the sampling period. However, there was a small increase in flow due to rainfall between 3–8 September with a peak discharge of 3.8 m³ s^–1. Flow subsequently declined to around 2.7 m³ s^–1. The samples of streambed sediment and stream water are, therefore, considered to be representative of low flow (base flow) conditions. The field and laboratory methods are summarized below; see Rawlins et al. (2003a) for a more detailed description of the analytical procedures and quality assurance and control.

Sites for the soil samples were selected from every second kilometer square of the British National Grid by random location within each square, subject to the avoidance of roads, tracks, railways, domestic and public gardens, and other seriously disturbed ground. The sample locations were determined using a global positioning system (GPS) and land use was recorded for each location. At each site, topsoil (0–15 cm) was taken from five holes augered at the corners and center of a square with a side of length 20 m with a hand auger and combined to form a bulked sample of approximately 1 kg.

Streambed sediment samples were collected from first- to fourth-order streams and the sample locations were evenly distributed across the catchment, recorded using a GPS. After removal of the oxidized surface layer, active sediments were wet-sieved through a 150-µm mesh yielding approximately 100 g of fine sediment, which was collected in a Kraft paper bag. As far as possible, streambed sediment samples were collected from active sediment, upstream of any potential source of local contamination, such as habitation, industrial activity, road, or track crossing.

All soil and sediment samples were dried, disaggregated, and passed through a 2-mm sieve. The <150-µm fraction of stream sediments were freeze-dried in the laboratory following initial air-drying. All samples were coned and quartered and a 50-g subsample ground in an agate planetary ball mill until 95% was <53 µm. The pulverized material was further subsampled to obtain portions for analysis.

Major, minor, and trace element determinations for soil and streambed sediment samples were performed by wavelength-dispersive X-ray fluorescence spectrometry (XRFS) (Ingham and Vrebos, 1994). A 12-g aliquot of milled material was mixed thoroughly with 3 g of binder for 3 min in an agate planetary ball mill. This mixture was then pressed into a 40-mm diam. pellet at 250 kN. The pellet was analyzed for Na, Mg, Al, Si, P, K, Ca, Ti, Mn, and Fe, and other trace elements using a Philips PW2400 sequential X-ray fluorescence spectrometer (Philips, Eindhoven, the Netherlands) fitted with rhodium-anode X-ray tubes (3 kW 60 kV). Element concentrations were expressed in g kg^–1 dry matter.

Olsen P was determined by extracting the P using a sodium bicarbonate solution (Olsen et al., 1954) and measuring the fraction using a phosphomolybdate method. The precise weight of approximately 2.5 g of air-dry soil was recorded and the soil was transferred to a glass bottle. Fifty milliliters of sodium bicarbonate solution was added and mechanically shaken for 20 min at 20°C. The solution was then filtered through a Whatman 125 filter paper. The concentration of P in the solution was determined using a flow injector analyzer (Tecator 5020, Method Application ASA 60-03/83; Foss UK Ltd., Warrington, UK). Olsen P concentrations were expressed in mg P kg^–1 dry soil.

Water samples were collected at the same locations as the streambed sediment samples. The water samples were collected before streambed sediment samples and slightly upstream of the stream-sediment sites to avoid contamination by disturbed sediment or pore water, and great care was taken during the sampling procedure to avoid any other contamination. All samples were filtered through 0.45-µm cellulose filters and collected in new, 30-mL polystyrene bottles, which had been rinsed with filtered water from the site before collection of the actual sample. The sample bottle containing the water sample for dissolved RP (<0.45) (terminology in accordance with Haygarth and Sharpley [2000]) analysis was immediately placed in a cool bag to maintain its temperature at 4°C until it could be placed in a refrigerator at the same temperature, before analysis. The samples were analyzed for electrical conductivity and pH within 10 h of sampling and for alkalinity within 24 h. All samples were analyzed for RP (<0.45) using the molybdate-blue method (Murphy and Riley, 1962). Although most sampling protocols prescribe that natural water samples should be analyzed for RP (<0.45) within 24 h (Haygarth et al., 1995), Gardolinski et al. (2001) showed that storage of natural water samples from the Tamar estuary at 4°C was effective in maintaining RP (<0.45) concentrations for up to 8 d. In this survey, the bottle containing the water sample for RP (<0.45) was kept at 4°C until analysis within 72 h of collection. The samples were also analyzed for Cl^–, SO₄^2–, and NO₃^– by ion chromatography (IC) (DX-600, Dionex Ltd., Camberley, Surrey, UK), and for Na⁺, K⁺, Ca²⁺, Mg²⁺ and TP (<0.45) by inductively coupled plasma–atomic emission spectrometry (ICP–AES) (ARL 3580, ARL, Luton, UK). Dissolved organic carbon (DOC) was determined using a Shimadzu TOC 5000 analyzer (Shimadzu Scientific Instruments, Columbia, Maryland, USA) for samples pretreated by the addition of a small volume of 10% HCl and purged with inert gas to remove any inorganic carbon. Total P concentrations were expressed in mg L^–1 and RP (<0.45) concentrations were expressed in µg L^–1.

Spatial Database
A digital spatial database, including a DEM, river network map, land cover map, and soil map was compiled for the Tamar catchment. Table 2 gives an overview of the sources and original resolution or map scale of these spatial data layers. The land cover map was reclassified into six land cover classes to enable comparison with the land use classes recorded by BGS during field sampling. The new land cover classes include bare soil, forest, unimproved grassland, improved grassland, arable land, and built-up areas. The BGS geochemical survey data for soil, stream sediment, and water were added to the spatial database as point data layers. In addition, South West Water–the company that supplies sewerage services for the Tamar catchment area–provided georeferenced information on point-source emissions (e.g., STWs). This information included the type of STW and the number of inhabitants connected.

All data layers were gridded to a 25-m resolution and converted to PCRaster, a raster-based GIS that includes a spatiotemporal modeling language (PCRaster, 2001), for further analysis and modeling. A first step encompassed the construction of DEM derivatives, such as local drainage direction, slope gradient, slope aspect, slope length, upstream area, and stream order, which were subsequently added to the spatial database.

The database was thoroughly checked for any spatial inconsistencies, which included the correction of the local drainage direction network to improve the hydrologic topology and ‘snapping’ the sample locations for stream water and streambed sediment and the point-source locations to the river network. To construct an improved local drainage network of the Tamar catchment, the gridded river network was ‘burned’ into the DEM, forcing the DEM-derived drainage network through the river network (see Renssen and Knoop, 2000). All stream water and streambed sediment sample locations and point-source locations were manually snapped to the nearest river. Locations that had to be moved more than 100 m to a nearest river were omitted from the database.

Statistical Analysis
The relationships of P concentrations in soil, streambed sediment, and stream water with potential explanatory variables contained in the spatial database were successively investigated using multiple linear regression. The P concentrations and explanatory variables were log-transformed by taking their natural logarithm if the regression residuals were not normally distributed or depended on the response variable (i.e., P concentration). Variables were only included in the regression equation if their regression coefficients were significant at the = 0.05 level and if the sign of the multiple linear regression coefficients could be physically justified. For example, this meant that in the case of regression between soil TP and major soil elements, the regression coefficients for Mn and Fe must be positive and the regression coefficient for Si must be negative.

The coefficients of determination (R²) from the regression analyses were used to partition the variation of the P concentration among the explanatory data sets. The procedure for variation partitioning includes the following steps (Legendre and Legendre, 1998):

Let vector y be the response variable (i.e., the P concentration), and matrices X and W contain the explanatory variables. Compute the multiple regression of y against X and W together. The R² represents the sum of the fractions of variations explained by explanatory variables [a + b + c] defined in Fig. 2. Fraction [d] represents the portion of unexplained variation and equals 1 – [a + b + c]. Alternatively, [d] can be determined by computing the variance of the regression residuals relative to the total variance of y.

Compute the regression of X and W separately. The R² between y and X represents the sum of the fractions of variation [a + b] and the R² between y and W represents [b + c]. Again, an alternative method to determine R² is to calculate 1 – the variance of the regression residuals relative to the total variance of y.

Fraction [b] is the intersection of the proportions of variation explained by X and W and can be obtained by subtraction: [b] = [a + b] + [b + c] – [a + b + c];

Fractions [a] and [c], which represent the unique contribution of X and W to the variation in y, can be obtained by subtraction: [a] = [a + b] – [b], and [c] = [b + c] – [b].

View larger version (14K): [in this window] [in a new window]   Fig. 2. Partitioning of the variation of a response variable y among two sets of explanatory variables. The entire square represents the total variation in y (adapted from Legendre and Legendre, 1998).

	SOIL PHOSPHORUS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS SOIL PHOSPHORUS STREAMBED SEDIMENT PHOSPHORUS STREAM WATER PHOSPHORUS CONCLUSIONS REFERENCES

 
Methods
Before regression analysis, an analysis of variance (ANOVA) was performed on the log-transformed soil TP and Olsen P concentrations to test whether the mean values for all of the soil classes and land cover classes are equal at = 0.05. Subsequently, the contributions of soil classes and land cover classes to the total variation in the log-transformed soil TP and Olsen P concentrations were examined using multiple linear regression with dummy variables. In this case, each soil type and land cover class comprised a dummy variable that has a value of 1 if the sample belongs to that class, otherwise a value of 0 was used. To enable the estimation of the regression coefficients independently, a dummy record was added with log-transformed P concentrations equal to the mean log-transformed P concentration of the soil class with the most observations (NatMap unit 0541j; N = 125) and the intercept was removed from the regression equation. Because we fixed the regression coefficient for soil class 0541j to its mean soil P concentration, the back-transformed regression coefficients for the soil classes represent approximations of the median values for these classes assuming a lognormal distribution of the soil P concentrations. The back-transformed regression coefficients for the land cover classes represent approximations of the median inflation factor (MIF) from the overall median P concentrations (i.e., median land cover class: overall median ratio). In addition, for each soil class a MIF relative to the overall median of the soil P concentrations was calculated. For each land cover class, a MIF relative to the median of the soil P concentrations in forests was calculated.

The coefficients of determination (explained variation) for soil class and land cover class were determined by computing the total variances of the log-transformed soil TP and Olsen P concentrations and the variances of the residuals. Next, the total variation was partitioned as described above.

To explore the geochemical controls on soil P concentrations, a multiple regression analysis between the log-transformed P concentrations and major soil elements (Si, Al, Fe, Mn, K, Ca, and Mg) was performed for each soil type with more than five observations under the hypothesis that the relationships between soil P and major soil elements differ between soil types. For Olsen P, the log-transformed soil TP concentration was also allowed to enter the regression equation as an explanatory variable. The explanatory variables were only included in the regression equation if they met the conditions described above. The total explained variation was determined by calculating the difference between the overall variance and the variance of the regression residuals. If the relationships were not significant or too few observations were available, the residuals were calculated by subtracting the soil class mean values from the observed values.

For further analysis of the effect of soil P on streambed sediment P, the log-transformed soil TP concentrations were interpolated using conditional simulation. First, we predicted the log-transformed TP concentrations based on soil type and land use using the regression model described above. Second, the variogram of the regression residuals was computed and the residuals were subsequently simulated 100 times at the centers of the 25 m by 25 m grid cells using the Gstat geostatistical package (Pebesma and Wesseling, 1998). For each simulation the log-transformed TP concentrations were calculated by adding the simulated regression residuals to the regression predictions. These log-transformed TP concentrations were then back-transformed to the original concentration scale and averaged for the 100 simulations, which were then mapped. For three soil types no samples were available, because these soil types cover only a small proportion of the Tamar catchment. For these soil types the average P concentration for the closest similar soil type was taken based on bedrock geology and soil texture as main criteria. Similarly, no soil samples were available from bare soil and built-up areas. For the purpose of interpolation, it was assumed that the mean soil TP concentrations for bare soil equals that of forest and the mean soil TP concentrations for built-up areas equals that of unimproved grassland. As with the unsampled soil types, the built-up areas and bare soil cover only a very small part of the catchment.

Results and Discussion
The results from the ANOVA showed that both soil type and land cover contribute significantly (p < 0.05) to the variation in soil P concentrations. This implied that both the soil classes and land cover classes could be included in the regression equation using dummy variables. Tables 3 and 4 show the regression coefficients and median inflation factors for the soil classes and land cover classes, respectively. Table 3 shows that the highest median soil TP and Olsen P concentrations occur in soil classes 0541j, 0541n, 0611a, 0611b, and 0611c, which represent loamy soils over Palaeozoic slaty mudstone and siltstone, and basic igneous rocks. The median soil P concentration is also high for the blanket peat soil (1013b), but this median value was based on only one sample. The lowest soil P concentrations are found in the loamy soils in Carboniferous sandstone and shale (soil class 0541h) and loamy and clayey stagnogley soils over Palaeozoic sandstone, slate, and mudstone (soil classes 0712d, 0712e, 0713b, and 0831a).

Table 5 and Fig. 3 show the variation partitioning of soil P. The information included in the soil (NatMap) and land cover (LCM1990) maps explain about 34% of the total variation in log-transformed soil TP and about 17% of the total variation in log-transformed soil Olsen P. Land cover accounts for only 4 to 5% of the variation in soil P. For soil TP, the contributions of land cover and soil class partly overlap, which implies, as noted above, that land cover and soil type are interrelated. As a result, the unique contribution of land cover could only be partly identified. The contributions of soil class and land cover to the variation in Olsen P are relatively independent, since the intersection [b] is relatively small. The small proportion of variation explained by the land cover classes may again be attributed to the broad land cover classes, which account poorly for temporal and local variations in land use and management, including livestock densities and fertilizer application rates. Various studies have shown that agricultural P inputs to soil increases the TP concentration in soil (Leinweber et al., 2002; Page et al., 2005). Accordingly, including information on land use and management could potentially enhance the proportion of explained variation of soil P, provided that this information is available at sufficient spatial and temporal detail. The lack of availability of such information across the catchment area, therefore, represents an important limitation of the present study.

View larger version (9K): [in this window] [in a new window]   Fig. 3. Partitioning of the variation of (a) log-transformed soil total P and (b) log-transformed soil Olsen P. The total surface areas of the big squares represent 100%.

Besides the TP concentrations, the Olsen P concentrations are associated with the same elements as for the TP concentrations, but the situation is the reverse. In general, the Olsen P concentrations are negatively correlated with the soil sesquioxides, and positively correlated with Si and Al. This is probably due to an increased adsorption capacity and, accordingly, decreased solubility and availability of soil P in the presence of soil sesquioxides. The presence of coarse grain size fractions increases the proportion of Olsen P relative to total soil P.

The fitted spherical semivariogram model of the residuals of the regression model with soil class and land cover class as dummy variables has a range (i.e., the distance beyond which the data are spatially uncorrelated) of 1725 m. The nugget variance (i.e., semivariance at zero lag) is 49% of the total variance (sill). Figure 4 shows the interpolated soil TP concentrations in the Tamar catchment. In general, the pattern of soil TP concentrations follows the spatial distribution of bedrock and soil types. The concentration of Olsen P in soil (not shown) shows a similar pattern. The soil P concentrations are generally lower in the northern part of the catchment, which is dominated by loamy and clayey soils over Carboniferous sandstone and mudstone. Soils in the southern part, which is dominated by Devonian mudstone and siltstone and basic igneous rock, have higher soil P concentrations. In particular, local ‘hotspots’ of soil TP occur in the southern part of the catchment. These hotspots may be attributed to local sources of P, for example due to enhanced rates of fertilizer application, or the occurrence of mine waste heaps, which occur throughout the area.

View larger version (64K): [in this window] [in a new window]   Fig. 4. Interpolated soil total P concentrations (g kg–1) in the Tamar catchment.

	STREAMBED SEDIMENT PHOSPHORUS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS SOIL PHOSPHORUS STREAMBED SEDIMENT PHOSPHORUS STREAM WATER PHOSPHORUS CONCLUSIONS REFERENCES

 
Methods
The contribution of soil P concentrations in the upstream areas, STW discharges, and major element chemistry of the streambed sediment (Si, Al, Fe, Mn, K, Ca, and Mg) to the variation in streambed sediment TP concentrations was analyzed using multiple regression. The contribution of soil P concentration to the variation in streambed sediment P was examined in two ways: (i) correlation with average interpolated soil TP concentrations in the area upstream from the streambed sediment sample locations and (ii) correlation with the TP concentrations in streambed sediment predicted from a sediment-associated P transport model that accounts for sediment erosion and deposition, and delivery to and transport through the river network. This model is described in more detail below. The variable that explained the variation in streambed P best was included in the subsequent variation partitioning analysis.

The contribution of the STW discharges was assessed by correlation with the estimated contribution of the STW discharge to the total water discharge at the sample locations. To calculate the total discharge per grid cell, we assumed that the total discharge (sum of STW discharges + base flow) at Gunnislake gauging station was equal to the mean monthly discharge for September 2002 (i.e., 3.0 m³ s^–1) and assumed that the base flow per unit area was constant over the catchment. The STW discharges were estimated by multiplying the number of people connected by a daily average water use of 0.150 m³ d^–1. Both the base flow and STW discharges were routed through the river network using the local drainage direction network map.

Phosphorus Transport Model
For modeling sediment-associated P transport and deposition on the hillslopes and in the river network, a GIS-based model was developed. This model estimates the relative contribution of each grid cell to sediment-associated P transport. In this way, the in-stream bed sediment TP concentration can be estimated as an average of the soil TP concentrations in the upstream grid cells weighted for the relative contribution of each grid cell to sediment-associated P transport.

The hillslope part of the model was based on a long-term rill erosion model described by Govers et al. (1993) and Van der Perk and Jetten (2006). A relative measure of potential soil erosion was derived using Eq. [1]:

[1]

where E_r = the relative potential erosion rate per unit area (kg m^–2 y^–1), f_soil = a soil factor (–), f_lc = a land cover factor (–), L = the slope length (m), S = the sine of the slope gradient, and n and m are exponents (–). Note that the model only estimates the relative soil erosion rate. The relative eroded mass of sediment-associated P was estimated by Eq. [2]:

[2]

where P_e = relative eroded mass of sediment-associated P per unit area (g m^–2 y^–1), f_enrich = a topsoil P enrichment factor (–), and P_soil = the soil TP concentration (g kg^–1). The sediment transport capacity (T_r) on a given location in the catchment was considered to be proportional to E_r (Govers et al., 1993):

[3]

where f_ld = a landscape diversity factor (–) and g_t = the transport capacity coefficient (–). The amount of eroded sediment and sediment-associated P per grid cell is transported downstream over the drainage network, as long as it does not exceed the transport capacity. The surplus is deposited. The TP concentration in the deposited sediment was calculated as an average of the upstream soil TP concentration (P_soil) weighted for the local relative erosion rates in the upstream grid cells (E_r).

The sediment-associated P that reaches the channel network in that manner was further routed through the river network. The fraction that is transported to the downstream grid cell (f_t) was calculated using Eq. [4]:

[4]

where X = the grid cell size (= 25 m) and = the distance over which the half of the sediment is supposed to be deposited (m). The TP concentration in the streambed sediment was calculated as an average of the upstream sediment-associated TP inputs from the hillslope weighted for the upstream sediment inputs and the transport fraction f_t.

Because the data available for this study did not allow the model to be calibrated, the model was parameterized using values from the literature. The starting point for the parameterization of the hillslope model was the model parameters presented by Van der Perk et al. (2002) for a small arable catchment with sandy loam soils (n = 0.8; m = 1.2; and g_t = 20).

The soil factor f_soil represents the sensitivity of soil types to soil erosion relative to a reference soil, which is a sandy loam soil in this case. To estimate the values of f_soil for the different soil type we used data reported by McHugh et al. (2002) and Wood et al. (2006) of estimated annual soil erosion for different UK soils with a 1 in 1 yr return period erosion event. The reported values were corrected for land cover (see below) and scaled relative to a sandy loam soil. The value for f_soil ranged from 0.33 for seasonally wet peat to loam soils (soil classes 0721a, b, d), to 0.67 for seasonally wet loam soils (soil class 0831a), 0.93 for clay soils (soil class 0421b) and seasonally wet deep clay soils (soil classes 0712d, e, 0713b), 1.0 for typical brown earths (0541h, j, k, n) and alluvial soils (soil classes 0561b, 0811b), 1.33 for brown podsolic soils (soil classes 0611a, b, c, 0612a, b, 0633) and peat to loam soils (soil classes 0651b, 0654a), and 1.41 for blanket peat (soil class 1013b).

The land cover factor f_lc represents sensitivity to soil erosion under different forms of land cover relative to a reference land cover type, which is arable land in this case. We estimated the values for f_lc using a method proposed by Walling and Zhang (2004) (see also McHugh et al., 2002). This method relates sediment transfer to Manning's roughness coefficient, which is tabulated for various land cover classes. Walling and Zhang (2004) used the parameter n^0.6 as a land cover factor. We used the parameter values reported in McHugh et al. (2002), but scaled them relative to arable land cover. The land cover factor f_lc is 0.18 for woodland, 0.39 for grassland, 0.34 to 0.52 for shrub heath, 0.66 for upland bogs, and 1.00 for arable land. The topsoil P enrichment factor f_enrich represents the enrichment in TP in the topsoil subject to erosion (usually 0–2 cm) relative to the soil TP concentrations in the top 15 cm. This parameter was estimated using data presented by Owens and Deeks (2004) and was set to 1.2 for grassland and 1.1 for arable land. This parameter may also represent the P enrichment due to the selective transport of fine sediments, which may contain more P than the coarse soil particle size fraction (Owens and Walling, 2002). However, because no quantitative information was available on selective transport at the catchment scale, this effect was left out of consideration. The landscape diversity factor f_ld takes into account the decrease in transport capacity at field boundaries. Since no quantitative information was available for the effect of field boundaries on sediment transport and deposition, as a best guess, this parameter was set to 0.5 for grid cells for which the downstream grid cell had a different land cover class (using the original LCM1990 map). The half-life distance was estimated to be 1000 m, which was also a best-guess estimate.

Results and Discussion
The average interpolated soil TP concentration upstream from the streambed sediment sample locations explained 16.9% of the variation in streambed TP concentrations, and the streambed TP concentrations predicted by the P transport model explained 21.0% of the variation in streambed TP concentrations. Because the P transport model yielded the highest R² for predicted streambed TP concentrations, these predictions were used in the subsequent variation partitioning analysis. It should be noted that the P transport model predicted TP concentrations based on interpolated soil TP concentrations, which were determined for a different particle size fraction (<2 mm) than the TP concentrations measured in the streambed sediment samples (<150 µm). The TP concentrations in the streambed sediment predicted by the P transport model are depicted in Fig. 5. In general, the sediment TP concentrations predicted by the P transport model display a similar general large-scale pattern as the soil TP concentrations with larger concentrations in the southern part of the Tamar catchment.

View larger version (40K): [in this window] [in a new window]   Fig. 5. Streambed sediment total P concentrations (g kg–1) predicted from soil total P concentrations using the P transport model.

[5]

where P_sed = the streambed TP concentration (g kg^–1), P_pred = the streambed TP concentration predicted by the P transport model (g kg^–1), Ca, K, Al, Fe = the respective concentrations of the major elements in the streambed (g kg^–1), and STWprop = the proportion of the discharge from STWs relative to the total discharge (–). Equation [5] shows that the streambed TP concentration is significantly positively related to the concentrations of Ca, K, and Fe, and significantly negatively related to Al. The K concentration is likely to be a proxy for the clay mineral content, though the relationship between K and clay content may vary considerably as the different lithologies could have some primary K-bearing minerals (e.g., K feldspar) in the <150-µm fraction. This suggests that streambed sediment P is partly controlled by adsorption to both sesquioxides and clay minerals, and by precipitation of apatite minerals (calcium phosphates). As with the soils, the Al concentration is likely to be positively correlated with the content of feldspars with few adsorption sites for P. These findings confirm the results from earlier studies on the relationships between sediment P and sediment physicochemical properties (House and Denison, 2002; Owens and Walling, 2002; Evans et al., 2004). These studies also found significant positive relationships between sediment P concentrations and sediment organic matter content. Organic matter content was not analyzed in the geochemical survey of the Tamar catchment (Rawlins et al., 2003a). It is, therefore, likely that, if organic matter content were measured, the proportion of explained variation in streambed TP concentrations would have been higher.

View larger version (16K): [in this window] [in a new window]   Fig. 6. Partitioning of the variation of streambed sediment P. The total surface area of the big square represents 100%.

	STREAM WATER PHOSPHORUS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS SOIL PHOSPHORUS STREAMBED SEDIMENT PHOSPHORUS STREAM WATER PHOSPHORUS CONCLUSIONS REFERENCES

Results and Discussion
The RP (<0.45) and TP (<0.45) concentrations were strongly correlated (r = 0.973) and the mean ratio RP (<0.45)/TP (<0.45) was 0.48 ± 0.34 (mean ± standard deviation) for samples with RP (<0.45) and TP (<0.45) concentrations above the detection limit (N = 368). No additional explanatory variable other than TP (<0.45) was found to be significantly related to RP (<0.45). Both the mean RP (<0.45) concentration and the mean TP (<0.45) concentration in streams with upstream STW discharges were found to be significantly higher ( = 0.05, p < 0.001) than in streams without upstream STW discharges. Table 8 shows the statistics of the RP (<0.45) and TP (<0.45) concentrations for streams with and without upstream STW discharges.

[6]

where TP = the TP (<0.45) concentration in stream water (mg L^–1), P_sed = the streambed TP concentration (g kg^–1), Al, Si, K = the respective concentrations of the major elements in the streambed sediment (g kg^–1), and STWprop = the proportion of the discharge from STWs relative to the total discharge (–). The regression equation shows that TP (<0.45) concentration in stream water is positively related to the P concentration in streambed sediment and to the proxies for minerals with limited adsorption capacity for P (Si and Al). The negative relationship between the TP (<0.45) concentration and the K concentration in streambed sediment suggests that adsorption to clay minerals significantly controls the TP (<0.45) concentrations in the stream water of the Tamar catchment. We could not establish a significant relationship between the log-transformed TP (<0.45) concentrations and the log-transformed mean soil TP concentration upstream from the stream water sample locations. This may be attributed to the variability of the particulate P transfer from soil to streambed sediment and the chemical processes controlling partitioning between P in the particulate and aqueous phases. Nevertheless, we found significant relations between soil P and streambed sediment P (see previous section), and subsequently between streambed sediment P and stream water P. Thus, the absence of a direct relationship also confirms the complexity of the interactions that govern the catchment-scale spatial variation of P concentrations in stream water.

Table 9 and Fig. 7 show the proportions of the variation explained by the different variables. Although the relationships with the variables are statistically significant, which was a condition to be included in the regression equation, the proportions of variation explained are modest, ranging from 3.7% for the sediment chemistry to 9.8% for the proportion of STW discharge. In this case, the regression equation including all significant explanatory variables explains a greater proportion (25.4%) of the total variation in stream water TP (<0.45) than the sum of the variation explained by the separate explanatory variables. The remaining variation that is not explained by the regression equation may largely be attributed to the spatial variability of reactive mineral forms of P in the streambed sediment, the temporal variability of stream water P concentrations and STW effluent discharges during the month of sampling, and other unknown local P sources, such as intermittent farm point-source inputs and septic tanks.

View larger version (19K): [in this window] [in a new window]   Fig. 7. Partitioning of the variation of stream water TP (<0.45). The shaded areas represent negative contributions to the variation. The total surface of the white area represents 100%.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS SOIL PHOSPHORUS STREAMBED SEDIMENT PHOSPHORUS STREAM WATER PHOSPHORUS CONCLUSIONS REFERENCES

Streambed P concentrations are correlated with major element concentrations probably through bulk mineralogy and adsorption on mineral surfaces. The P transfer model that accounts for erosion and deposition of sediment-associated P as a function of slope length, slope angle, soil type, and land cover predicts the streambed TP concentrations better than the average soil TP concentration in the area upstream from the streambed sediment sample locations. This is despite the fact that we only used values from the literature and best-guess estimates as input for this model and did not have data on local animal stocking densities and fertilizer inputs available across the catchment. Streambed P concentrations are principally related to the major element concentrations in streambed sediment and P delivery from the hillslopes due to soil erosion and sediment delivery. However, the contributions of the variation in major element concentrations and P delivery could not be uniquely identified, probably because both the P concentrations and the major element concentrations are both largely controlled by erosion inputs from upstream areas. In addition, inputs from STWs are a small but significant contribution to the total variation in streambed P concentrations.

Total P (<0.45) and RP (<0.45) concentrations in stream water during base flow conditions are closely correlated, and on average RP (<0.45) concentrations comprise about 50% of the TP (<0.45) concentration. During base flow conditions, the TP (<0.45) concentrations in stream water are mainly controlled by P concentrations and the concentrations of the major elements in the streambed sediment. Effluent discharges from STWs elevate the stream water P concentrations in the downstream river network and account for about 10% of the total variation in log-transformed TP (<0.45) concentrations in stream water. Although we could not establish a direct, significant relationship between the log-transformed TP (<0.45) concentrations and the log-transformed mean soil TP concentration upstream from the stream water sample locations, the results from this study demonstrated an indirect relationship via exchange with bed sediments, which proved to be related to soil P from spatially varying areas within the contributing catchment. The absence of a direct relationship underlines the complexity of the interactions between soil P and stream water P at the catchment scale. This complexity includes the variability of the particulate P transfer from soil to streambed sediment and the chemical processes controlling partitioning between P in the particulate and aqueous phases.

This study illustrates the complex cascade of P transfer from soil to streambed to stream water under base flow conditions. The P concentrations sampled at a site will always be a net balance between natural/anthropogenic/animal inputs, transport from upslope/upstream, and removal processes in solution and associated with sediment transport to downslope/downstream. This balance involves both long-term accumulation and storm (or succession of storm) effects on removal and transport to downslope/downstream sites–where the single biggest storm in the year (or several years) might dominate both removal and internal redistribution. Although many studies have shown that the majority of P transport takes place during short-term hydrologic events (e.g., Correll et al., 1999; Haygarth et al., 2004), the above findings demonstrate the importance of soil and sediment chemistry for controlling the catchment-scale spatial variation of P concentrations in soil, streambed sediment, and stream water, and hence the P export from catchments. Consequently, taking spatial variation of the soil and sediment chemistry into account in catchment-scale P transfer models could potentially enhance the model predictions. This information could be obtained from geochemical databases, such as the BGS data set that was used in this study. Sole reliance on spatial information from generic soil and land cover maps is of limited value as they accounted for only 30% of the total variation in soil TP concentrations. The proportion of variation explained could potentially be enhanced by including spatially detailed information on land use and management, including livestock distribution and fertilizer application rates. Despite the relatively small proportion of variation of P concentrations in soil explained by the soil classes, the soil map proved to be of great value as auxiliary information in the interpolation of soil P concentrations. Given that geochemical data are increasingly available for much of the UK, these data can serve as a major input to catchment-scale P loss models. The relationships found for the Tamar catchment would, however, have to be tested for catchments with different lithologies and land use types.

	ACKNOWLEDGMENTS

 
This work was undertaken as a subproject of the BBSRC Grant 89/MAF 12247–"Scale and uncertainty in modeling phosphorus transfer from agricultural grasslands to watercourses: Development of a catchment scale management tool". The authors are grateful to the Biotechnology and Biological Sciences Research Council (BBSRC) for providing the funding for this project. The authors would like to extend thanks to the project team including Trevor Page (Univ. of Lancaster), Patricia Butler and Adrian Joynes (IGER), and Gavin Wood (NSRI). We are also grateful to Sonia Thurley and Julian Greaves of the Environment Agency, and Tony Griffiths of South West Water. IGER is supported by the BBSRC. The British Geological Survey (BGS) wishes to acknowledge the Environment Agency (SW region) which contributed funding to the geochemical survey of the Tamar catchment. This paper is published with the permission of the Director of the BGS (NERC). Edzer Pebesma (Utrecht Univ.) is thanked for his geostatistical advice. The first author would like to acknowledge a NWO/British Council grant PPS 785 for a sabbatical at NSRI, North Wyke.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS SOIL PHOSPHORUS STREAMBED SEDIMENT PHOSPHORUS STREAM WATER PHOSPHORUS CONCLUSIONS REFERENCES

 

• Arnold, J.G., and N. Fohrer. 2005. SWAT2000: Current capabilities and research opportunities in applied watershed modeling. Hydrol. Processes 19:563–572.[CrossRef]
• Beven, K.J., L. Heathwaite, P.M. Haygarth, D. Walling, R. Brazier, and P. Withers. 2005. On the concept of delivery of sediment and nutrients to stream channels. Hydrol. Processes 19:551–556.[CrossRef]
• Cooper, A.B., C.M. Smith, and A.B. Bottcher. 1992. Predicting runoff of water, sediment, and nutrients from a New Zealand grazed pasture using CREAMS. Trans. ASAE 35:105–112.[Web of Science]
• Correll, D.L. 1998. The role of phosphorus in the eutrophication of receiving waters: A review. J. Environ. Qual. 27:261–266.[Web of Science]
• Correll, D.L., T.E. Jordan, and D.E. Weller. 1999. Transport of nitrogen and phosphorus from Rhode River watersheds during storm events. Water Resour. Res. 35:2513–2521.[CrossRef]
• De Wit, M.J.M. 2001. Nutrient fluxes at the river basin scale: I. The PolFlow model. Hydrol. Processes 15:743–759.[CrossRef]
• Dillon, P.J., and W.B. Kirchner. 1975. The effects of geology and land use on the export of phosphorus from watersheds. Water Res. 9:135–148.
• Dougherty, W.J., N.K. Fleming, J.W. Cox, and D.J. Chittleborough. 2004. Phosphorus transfer in surface runoff from intensive pasture systems at various scales. J. Environ. Qual. 33:1973–1988.[Abstract/Free Full Text]
• Evans, D.J., P.J. Johnes, and D.S. Lawrence. 2004. Physicochemical controls on phosphorus cycling in two lowland streams: II. The sediment phase. Sci. Total Environ. 329:165–182.[CrossRef][Medline]
• Frossard, E., L.M. Condron, A. Oberson, S. Sinaj, and J.C. Fardeau. 2000. Processes governing phosphorus availability in temperate soils. J. Environ. Qual. 29:15–23.[Abstract/Free Full Text]
• Fuller, R.M., G.B. Groom, and A.R. Jones. 1994. The land cover map of Great Britain: An automated classification of Landsat Thematic Mapper data. Photogr. Eng. Remote Sens. 60:553–562.
• Gardolinski, P., G. Hanrahan, E.P. Achterberg, M. Gledhill, A.D. Tappin, W.A. House, and P.J. Worsfold. 2001. Comparison of sample storage protocols for the determination of nutrients in natural waters. Water Res. 35:3670–3678.[Medline]
• Gburek, W.J., A.N. Sharpley, L. Heathwaite, and G.J. Folmar. 2000. Phosphorus management at the watershed scale: A modification of the phosphorus index. J. Environ. Qual. 29:130–144.[Abstract/Free Full Text]
• Govers, G., T.A. Quine, and D.E. Walling. 1993. The effects of water erosion and tillage movement on hillslope profile development: A comparison of field observations and model results. p. 285–300. In S. Wicherek (ed.) Farm land erosion in temperate plains and hills. Elsevier, Amsterdam, the Netherlands.
• Grobler, D.C., and M.J. Silberbauer. 1985. The combined effect of geology, phosphate sources, and runoff on phosphate export from drainage basins. Water Res. 19:975–981.
• Haygarth, P.M., C.D. Ashby, and S.C. Jarvis. 1995. Short-term changes in the molybdate reactive phosphorus of stored soil waters. J. Environ. Qual. 24:1133–1140.[Web of Science]
• Haygarth, P.M., S.C. Jarvis, P. Chapman, and R.V. Smith. 1998. Phosphorus budgets for two contrasting grassland farming systems in the UK. Soil Use Manage. 14:160–167.[CrossRef]
• Haygarth, P.M., and A.N. Sharpley. 2000. Terminology for phosphorus transfer. J. Environ. Qual. 29:10–15.[Abstract/Free Full Text]
• Haygarth, P.M., B.L. Turner, A. Fraser, S. Jarvis, T. Harrod, D. Nash, D. Halliwell, T. Page, and K. Beven. 2004. Temporal variability in phosphorus transfers: Classifying concentration-discharge event dynamics. Hydrol. Earth Syst. Sci. 8:88–97.
• Haygarth, P.M., F.L. Wood, A.L. Heathwaite, and P.B. Butler. 2005. Phosphorus dynamics observed through increasing scales in a nested headwater-to-river channel study. Sci. Total Environ. 344:83–106.[Medline]
• Heathwaite, A.L., A.I. Fraser, P.J. Johnes, M. Hutchins, E. Lord, and D. Butterfield. 2003. The Phosphorus Indicators Tool: A simple model of diffuse P loss from agricultural land to water. Soil Use Manage. 19:1–11.[CrossRef]
• Helming, K., A.-V. Auzet, and D. Favis-Mortlock. 2005. Soil erosion patterns: Evolution, spatio-temporal dynamics, and connectivity. Earth Surf. Processes Landforms 30:131–132.[CrossRef]
• House, W.A., and F.H. Denison. 2002. Total phosphorus content of river sediments in relationship to calcium, iron, and organic matter concentrations. Sci. Total Environ. 282–283:341–351.
• House, W.A., and M.S. Warwick. 1998. A mass-balance approach to quantifying the importance of in-stream processes during nutrient transport in a large river catchment. Sci. Total Environ. 210–211:139–152.
• Hughes, S., B. Reynolds, S.A. Bell, and C. Gardner. 2000. Simple phosphorus saturation index to estimate risk of dissolved P in runoff from arable soils. Soil Use Manage. 16:206–210.
• Ingham, M.N., and B.A.R. Vrebos. 1994. High productivity geochemical XRF analysis. Adv. X-Ray Anal. 37:717–724.
• Jansson, H., V. Mantylahti, A. Narvanen, and R. Uusitalo. 2000. Phosphorus content of ditch sediments as indicator of critical source areas. Agric. Food Sci. Finl. 9:217–221.
• Jarvie, H.P., M.D. Jürgens, R.J. Williams, C. Neal, J.J.L. Davies, C. Barrett, and J. White. 2005. Role of river bed sediments as sources and sinks of phosphorus across two major eutrophic river basins: The Hampshire Avon and Herefordshire Wye. J. Hydrol. 304:51–74.[CrossRef]
• Jordan, C., S.O. McGuckin, and R.V. Smith. 2000. Increased predicted losses of phosphorus to surface waters from soils with high Olsen P concentrations. Soil Use Manage. 16:27–35.
• Legendre, P., and L. Legendre. 1998. Numerical ecology. 2nd ed. Elsevier, Amsterdam, the Netherlands.
• Leinweber, P., B.L. Turner, and R. Meissner. 2002. Phosphorus. p. 29–55. In P.M. Haygarth and S.C. Jarvis (ed.) Agriculture, hydrology, and water quality. CAB International, Wallingford, UK.
• McDowell, R., A. Sharpley, and G. Folmar. 2001. Phosphorus export from an agricultural watershed: Linking source and transport mechanisms. J. Environ. Qual. 30:1587–1595.[Web of Science][Medline]
• McHugh, M., G. Wood, D. Walling, R. Morgan, Y. Zhang, S. Anthony, and M. Hutchins. 2002. Prediction of sediment delivery to watercourses from land; phase II. R&D Tech. Rep. No. P2-209. Environment Agency, Bristol, UK.
• McKee, L.J., B.D. Eyre, S. Hossain, and P.R. Pepperell. 2001. Influence of climate, geology, and humans on spatial and temporal nutrient geochemistry in the subtropical Richmond River catchment. Aust. J. Mar. Freshwater Res. 51:235–248.
• Mourad, D.S.J., M. van der Perk, and K. Piirimäe. 2006. Changes in nutrient emissions, fluxes, and retention in a North-Eastern European lowland drainage basin. Environ. Monit. Assess. 120:415–448.[CrossRef][Web of Science][Medline]
• Murphy, J., and J.P. Riley. 1962. A modified single solution for the determination of phosphate in natural waters. Anal. Chim. Acta 27:31–36.[Medline]
• Olsen, S.R., C.V. Cole, F.S. Watanbe, and L.A. Dean. 1954. Estimation of available phosphorus in soils by extraction with sodium bicarbonate. Circ. 939. USDA, Washington, DC.
• Owens, P.N., and L.K. Deeks. 2004. The role of soil phosphorus in controlling sediment-associated phosphorus transfers in river catchment. p. 444–450. In V. Golosov, V. Belyaev, and D.E. Walling (ed.) Sediment transfer through the fluvial system. IAHS Publ. 288. IAHS Press, Wallingford, UK.
• Owens, P.N., and D.E. Walling. 2002. The phosphorus content of fluvial sediment in rural and industrial river basins. Water Res. 36:685–707.[Medline]
• Owens, P.N., D.E. Walling, J. Carton, A.A. Meharg, J. Wright, and G.J.L. Leeks. 2001. Downstream changes in the transport and storage of sediment-associated contaminants (P, Cr, and PCBs) in agricultural and industrialized drainage basins. Sci. Total Environ. 266:177–186.[CrossRef][Medline]
• Page, T., P.M. Haygarth, K.J. Beven, A. Joynes, T. Butler, C. Keeler, J. Freer, P.N. Owens, and G.A. Wood. 2005. Spatial variability of soil phosphorus in relation to the topographic index and critical source areas: Sampling for assessing water quality. J. Environ. Qual. 34:2263–2277.[Abstract/Free Full Text]
• Pebesma, E.J., and C.G. Wesseling. 1998. Gstat: A program for geostatistical modeling, prediction, and simulation. Comput. Geosci. 24:17–31.[CrossRef]
• PCRaster. 2001. PCRaster manual [online]. Dep. of Physical Geography, Utrecht Univ., the Netherlands. Available at http://pcraster.geog.uu.nl/ (verified 17 Jan. 2007).
• Pieterse, N.M., W. Bleuten, and S.E. Jørgensen. 2003. Contribution of point sources and diffuse sources to nitrogen and phosphorus loads in lowland river tributaries. J. Hydrol. 271:213–225.[CrossRef]
• Preedy, N., K.B. McTiernan, R. Matthews, L. Heathwaite, and P.M. Haygarth. 2001. Rapid incidental phosphorus transfers from grassland. J. Environ. Qual. 30:2105–2112.[Abstract/Free Full Text]
• Quinton, J.N., P. Strauss, N. Miller, E. Azazoglu, M. Yli-Halla, and R. Uusitalo. 2003. The potential for soil P tests to predict phosphorus losses in overland flow. J. Plant Nutr. Soil Sci. 166:432–437.[CrossRef]
• Rawlins, B.G., K. O'Donnell, and M. Ingham. 2003a. Geochemical survey of the Tamar catchment (south-west England). Report CR/03/027. British Geological Survey, Keyworth, Nottingham, UK.
• Rawlins, B.G., R. Webster, and T.R. Lister. 2003b. The influence of parent material on topsoil geochemistry in eastern England. Earth Surf. Processes Landforms 28:1389–1409.[CrossRef]
• Renssen, H., and J.M. Knoop. 2000. A global river routing network for use in hydrological modeling. J. Hydrol. 230:230–243.[CrossRef]
• Svendsen, L.M., and B. Kronvang. 1993. Retention of nitrogen and phosphorus in a Danish lowland river system: Implications for the export from the watershed. Hydrobiologia 251:123–135.[CrossRef][Web of Science]
• Van der Perk, M., and V.G. Jetten. 2006. The use of a simple sediment budget model to estimate long-term contaminant export from small catchments. Geomorphology 79:3–12.
• Van der Perk, M., P.N. Owens, L.K. Deeks, and B.G. Rawlins. 2006. Sediment geochemical controls on in-stream phosphorus concentrations during base flow. Water Air Soil Pollut.-Focus 6:443–451.[CrossRef]
• Van der Perk, M., O. Slávik, and E. Fulajtár. 2002. Assessment of spatial variation of ¹³⁷Cs in small catchments. J. Environ. Qual. 31:1930–1939.[Abstract/Free Full Text]
• Walling, D.E., and Y. Zhang. 2004. Predicting slope-channel connectivity: A national scale approach. p. 107–114. In V. Golosov, V. Belyaev, and D.E. Walling (ed.) Sediment transfer through the fluvial system. IAHS Publ. 288. IAHS Press, Wallingford, UK.
• Wood, G.A., M. McHugh, R.P.C. Morgan, and A. Williamson. 2006. Estimating sediment generation from hillslopes in England and Wales: Development of a management planning tool. p. 217–227. In P.N. Owens and A.J. Collins (ed.) Soil erosion and sediment redistribution in river catchments: Measurement, modelling, and management. CAB International, Wallingford, UK.

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by van der Perk, M. Articles by Beven, K. J. Search for Related Content PubMed PubMed Citation Articles by van der Perk, M. Articles by Beven, K. J. Agricola Articles by van der Perk, M. Articles by Beven, K. J. Related Collections Watershed and Landscape Processes Phosphorus Spatial Distribution