Final Report: Measuring and Modeling the Source, Transport and Bioavailability of Phosphorus in Agricultural WatershedsEPA Grant Number: R830669
Title: Measuring and Modeling the Source, Transport and Bioavailability of Phosphorus in Agricultural Watersheds
Investigators: Lathrop, Richard C. , Armstrong, D. E. , Hoopes, John A. , Karthikeyan, K. G. , MacKay, David Scott , Nowak, Peter , Panuska, John C. , Penn, Michael R. , Potter, Kenneth W. , Wu, Chin H.
Institution: Wisconsin Department of Natural Resources , The State University of New York at Buffalo , University of Wisconsin Madison , University of Wisconsin - Platteville
EPA Project Officer: Packard, Benjamin H
Project Period: December 17, 2002 through December 16, 2005 (Extended to December 16, 2006)
Project Amount: $749,307
RFA: Nutrient Science for Improved Watershed Management (2002) RFA Text | Recipients Lists
Research Category: Water , Water and Watersheds
The overall research project objectives were to:
- Quantify effects of manure management and crop production systems on runoff phosphorus (P) losses, particularly related to the portion that is biologically available.
- Determine spatial patterns of sediment and associated P in streams.
- Determine in-stream fate and transport processes of P including bioavailable P (BAP).
- Evaluate and improve modeling tools used to assess P transport in agricultural watersheds over a wide range of spatial scales.
- Determine relation of P losses with the scale of animal operation.
- Integrate outreach into on-going research efforts.
We measured and modeled the sources, transport, and fate of phosphorus (P) with an emphasis on bioavailable P (BAP) in the mostly agricultural watershed of Lake Mendota near Madison, Wisconsin during 2002–2006. Our research focus was on the scale-dependent processes (especially related to the movement of soils and sediment) that link agricultural P sources to watershed export of BAP. Major emphasis was placed on both an upland research component and a stream research component that were combined to complete the linkage.
Agricultural management practices utilizing conservation tillage with three different crop management treatments—corn harvested for grain, corn harvested for silage, and corn harvested for silage with fall manure application—were evaluated on 146-m2 experimental field tracts at the University of Wisconsin Arlington Agricultural Research Station. This allowed edge-of-field runoff samples collected under natural rainfall conditions to be analyzed for each crop management treatment. Results included the size distribution of primary particles and soil aggregates in runoff, sediment P losses for five particle-size fractions, physical stability attributes of aggregates, and simple empirical equations predicting P losses for soils with similar textural properties. The corn grain treatment, with its higher residue levels, had significantly less runoff, less sediment and P losses, and greater aggregate stability. Approximately 57% of the P mass was associated with particles < 50 μm (silt-size) with no differences in relationships between the crop management treatments. Overall, four different aggregate size peaks were identified (5.4, 32, 160, and 570 μm). Size peaks with the largest frequency were 32 μm (34%) and 570 μm (31%); the median aggregate size of 32 μm compared favorably with the small aggregate size (30 μm) currently used in the Water Erosion Prediction Project (WEPP) and the Revised Universal Soil Loss Equation (RUSLE) soil loss models.
At the small subwatershed scale, we focused on internally-drained shallow depressions that interrupt surface runoff connectivity. Two out of five depressions located on a dairy farm in the North Fork Pheasant Branch subwatershed were intensively studied along with several erosional locations to elucidate sediment and P transport processes. While shallow depressions are common features in glaciated landscapes, the processes by which these depressions mediate P delivery are poorly understood. Given that such depressions are rarely identified by Digital Elevation Models (DEMs), these important landscape features are not usually modeled in agricultural management assessments. In our study we observed overtopping of depressions only during major runoff events or when crop management practices (e.g., manure application) altered surface soil infiltration rates. We also found significant differences in soil properties and P characteristics between erosional and depressional zones. Radioactive 137Cs fallout inventories indicated that shallow depressions functioned as colluvial sinks and depositional sites, with an accumulation of fine-textured sediments that were rich in P. While erodible soils exhibited greater potential for labile P losses than depressions, we found that labile P forms migrate downward into depressional zone soils, most likely through macropores created by worm holes. While more research is required to determine P delivery patterns from watersheds with depressions when larger storms cause spillage runoff conditions, our results suggest that agricultural catchments with depressions will export less P than watersheds without depressions.
The stream research component of our project focused on three main areas: (1) inventorying the amount of P in the sediment deposits of the drainage ditches, intermittent and perennially flowing stream channels, and major wetlands of the agricultural Dorn Creek watershed (part of the larger Lake Mendota watershed); (2) analyzing the forms of P (especially BAP) and associated physical and chemical characteristics of the sediment deposits; and (3) elucidating the hydrodynamic processes that dictate how much and what forms of sediment-associated P moves through the stream drainage network. The surficial sediment (3 to 5 cm thick) represented the most active layer susceptible to resuspension and transport. Sediment deposits in pools and other depositional locations in the stream channels were much thicker (as great as 1 meter) such that the topmost layer represented only about 5.2% and 4.4% of the total sediment and P inventories in Dorn Creek, respectively. The highest P concentrations were found in wetland areas, agricultural drainage ditches, and stream pools. The concentrations of BAP at many sites made up a large percentage of total P. Although total P in nondepositional zones was largely inorganic P (averaging 81%), the fraction represented by BAP was lower, averaging 39 % as compared to 53% for depositional sites. Concentrations of BAP in stream sediments were related closely to the concentrations of sediment total P (r2=0.97) and inorganic P (r2 = 0.98) and moderately to Fe (r2 = 0.71) and Al (r2 = 0.54). The close relation of inorganic P and BAP to sediment Fe and Al suggests that adsorption to Fe and Al oxyhydroxides may play a major role in controlling concentrations of inorganic P and BAP in stream sediments. Stream sediments play an important role in buffering stream water P concentrations as sediments deposited in Dorn Creek act as a sink for dissolved reactive P (DRP) during runoff events, while acting as a source for DRP during base-flow periods.
The natural radionuclides 7Be (53-day half life) and 210Pb (22-year half life) were used to profile the transport of “new sediment” to specific stream sites and the depth of mixing of surficial stream sediments during runoff events. We used a model based on 7Be/210Pb ratios in rainfall as compared to stream surficial sediments to estimate the average age and fraction of new sediment transported and deposited in the stream at various sites during storm events. The calculated age ranged from 100 to over 400 days. The fraction of surficial sediment comprised of new sediment ranged from nondetectable to about 25%. Changes due to diagenesis (i.e., conversion of apatite P or organic P to BAP, or conversion of BAP to other forms) were not apparent at most sites as trends in BAP concentrations with depth had modest variations. In contrast, concentrations of organic C, P, and total N (primarily organic) tended to decline with depth (time), indicating gradual mineralization of organic matter (OM) during sediment burial.
Our studies identified many important hydrodynamic processes affecting the fate of P in Dorn Creek. One important finding was that the critical shear stress generally increased with sediment depth indicating cohesive material at lower depths. A low critical shear stress was measured in the top 5 cm for a typical sediment core, suggesting that streambed sediments are susceptible to resuspension under 10–20 cm/s flow velocity. Thus, an increase in the shear stress above the measured critical shear value increases sediment transport due to resuspension of channel sediments. However, the large shear stresses only occur during the initial phase of the runoff event (rise of hydrograph). Because the rate of vertical downward erosion of the bed sediments requires high shear stresses to be sustained for long periods (more than a day) before the deeper sediments are exposed, the deeper sediments may not be resuspended even during very large events.
Our stream research conducted during 2006 also produced some important insights about the role of wetlands. All peak flows were attenuated in the upper Dorn Creek wetland, with the maximum attenuation corresponding to the intermediate events. The reduced attenuation of the larger events appeared to be due to filling of storage, either due to antecedent conditions or the event itself. In the case of sediment, the mass leaving the wetland in the two largest storms was twice the mass entering the wetland, which accounted for 96% of the exported sediment during the period of observation. The failure of the wetland to trap sediment was apparently due to the role of drainage ditches that trap sediment during the wetland-filling phase and release sediment during drainage. The export of sediment during the largest events appeared to be due to remobilization of sediment that had previously been deposited in the low-gradient channel during smaller events. Pertaining to P, the upper Dorn Creek wetland functioned as a sink for P during small events and a source during large events, with P inputs about equaling outputs for the entire study period. However, more DRP entered than left the wetland, as about 40% of the DRP was sequestered during the study period. This indicates that DRP in runoff was sorbed by wetland sediments, likely as water overflowed the stream banks to newly flooded areas.
Modeling tool assessments were also conducted as part of the project. One study focused on the Soil and Water Assessment Tool’s (SWAT) integration of the Modified Universal Soil Loss Equation (MUSLE) with hydrologic response units (HRU). A key question in scaling from fields to large watersheds is whether the spatial representation within the model affects the fluxes. Results showed that HRUs do not conserve sediment with scale of watershed representation. Instead, HRUs introduce almost half of the variability in sediment generation, which other researchers have previously attributed to input data aggregation. This occurs for two reasons. First, MUSLE defines a nonlinear relationship between sediment generation and HRU area, but the sediment load is scaled linearly from the HRU level to the subwatershed level. Second, HRUs aggregate land areas without regard for the surface connectivity assumptions, which are implicit in MUSLE. This means that runoff time-of-concentration is arbitrarily determined by HRU area rather than topographic controls on surface runoff. Thus, there is no physical connection between the sediment generated at each of the HRUs and the sub-basin total. These findings have direct implications for the interpretation of model output, and such conflicts can potentially mask the effect of different land use on soil erosion. We made recommendations to address these problems.
We also addressed the important question of whether watershed scale hydrological models can effectively represent the internal sinks (depressions) that are common in glaciated landscapes. The overall modeling approach employed a three-scale evaluation scheme from the small shallow depressions studied in our upland research, the small watershed scale represented by Dorn Creek, and the much larger Yahara River watershed (in the Mendota drainage basin) with the U.S. Geological Survey (USGS) monitoring data available. Recommendations included the use of reservoirs as a suitable proxy for internal sinks at the field scale where flow is intermittent in response to storm discharge. In small watersheds where flow is intermittent, such as the upper part of Dorn Creek, reservoirs may be appropriate. Where flow is perennial, reservoirs are not suitable proxies for sinks. In large watersheds Agricultural Policy/Environmental eXtender (APEX) is more appropriately parameterized to deposit sediment by slowing down discharge. Since this is hydraulically not equivalent to having standing pools at field scales, suitability of this approach needs to be evaluated. This highlights a fundamental scale limitation of distributed watershed models. With appropriate data APEX can form any size distributed land elements, such as individual fields, but these have to honor topographic boundaries (divides, channels) when using energy-based extensions of the Universal Soil Loss Equation (USLE). Future effort is needed in developing automated tools for defining such small scale distributed components.
Our outreach component includes a project Web Site (http://bse.wisc.edu/wi_nutsci_epa_stargrant Exit ) that contains all the individual research results and background information for the upland, stream, and modeling parts of the project.
- Field Plot Studies
- Determine the size distribution of primary particles and soil aggregates in runoff.
- Quantify the sediment total P mass distribution across five particle-size fractions.
- Investigate the physical stability of aggregates and the factors impacting aggregate stability in flow.
- Develop simple empirical equations to predict particulate P losses for soils with similar textural properties.
- Subwatershed Studies
The overall goal of the field plot studies was to investigate the influence of agricultural management practices (e.g., residue cover, tillage, manure management) on the delivery of P in particulate and dissolved forms in rainfall runoff. The specific research objectives were:
The investigation was conducted from 2003 to 2005 at the University of Wisconsin Arlington Agricultural Research Station on terraced fields (silt loam – Ripon series soils) planted in corn. Conservation tillage practices (fall chisel plowing followed by spring field cultivation) were conducted prior to planting. Three separate crop management treatments were tested: (1) corn harvested for grain (CG); (2) corn harvested for silage (CS); and (3) corn harvested for silage with fall manure application (SM). Of the three treatments, the CG treatment had the highest level of postharvest plant residue left on the soil surface. Liquid dairy manure containing sawdust bedding was applied to SM following harvest and incorporated by chisel plowing. Manure P application rates ranged between 33 and 53 kg P ha-1 in the 3 study years.
Edge-of-field runoff samples were collected under natural rainfall conditions for each crop management treatment in 146-m2 tracts with average slopes of 7.0–8.4%. A flow dividing bulk collection system was used to monitor a wide range of runoff events during the 3-year period. Routine chemical analyses of the runoff samples included total solids (TS) and volatile solids (VS); total phosphorus (TP), total dissolved phosphorus (TDP), and DRP as well as other parameters. In addition, the TP mass distribution was determined for five different particle size classes in runoff: < 2 μm (clay), 2–10 μm (fine silt), 10–50 μm (coarse silt), 50–500 μm (very fine to medium sand), and > 500 μm (coarse sand). Particle size distributions (PSD) and aggregate size distributions (ASD) of the runoff sediments were determined using gravity settling and laser diffraction methods. Particle size refers to a mixture of primary particles (i.e., sand, silt, and clay) plus agglomerates of primary particles, clay, and OM (aggregates). The ASD refers to aggregates only.
Runoff Volume. Median runoff coefficients for the three field management sites indicated that CG had a significantly lower median runoff coefficient (0.04) than CS (0.13) and SM (0.14), which were similar. The same relationship between sites was observed for the median runoff volume: CG (0.04) < CS (0.31) ~ SM (0.22). The lower median runoff coefficient for CG is most likely the result of the higher residue level (51% versus 10 to 15%) shielding the soil surface from raindrop impact thus preventing soil surface seal formation.
Sediment and Phosphorus Losses. For rainfall-runoff events, a combination of energy available in raindrop impact and shear stress in flowing water dictates the extent of soil detachment and sediment transport. Crop canopy and residue both play an important role in reducing rainfall erosivity by intercepting rainfall, retarding overland flow, and delaying surface crust development, thus allowing more opportunities for infiltration. The following discussion will consider these factors and their influence on sediment and P losses during the frost free (FF) period. All concentration values reported are event mean concentrations (EMC), unless stated otherwise.
The most frequently observed case during the FF period was that the specific water quality constituent values for CG were significantly less than those for CS and SM, which had statistically similar levels of solids and different forms of P. In general, the analysis indicated that the greatest difference between sites existed between the high-residue corn-grain site (CG) and the low-residue corn-silage sites (CS and SM). The presence of residue significantly reduced soil loss which, in turn, reduced TP export as supported by a strong relationship (r2 = 0.97) observed between TS and TP losses. The majority of the P loss was found to occur in the particulate-bound form. Source contributions to P export appeared different for the CG and corn-silage sites. The presence of crop residue also appeared to influence the form of P exported in runoff during the FF period. The median DRP/TDP ratio for CG (0.43) was significantly lower than those for CS (0.76) and SM (0.76). These results are indicative of a higher relative contribution from dissolved organic P (DOP = TDP–DRP) from CG. The higher organic P values could be attributed to CG crop residue decomposition.
Sediment and Phosphorus Mass by Gravity Settling. Sediment plays an important role in P losses from row-cropped agricultural systems. The transport potential of sediment particles is closely associated with sediment particle size and density. Fine-textured particles, such as those found in the clay and silt-sized classes, have higher specific surface area, thus providing increased binding capacity for contaminants and nutrients. Consequently, clay and silt-sized particles are enriched in P. Particle size, transport potential, and the increase in the proportion of fine particles are important factors impacting sediment and P delivery from agricultural fields.
Data for sediment mass (%), TP concentration (mg/kg), and the TP mass fraction (%) among different particle size classes found in runoff samples during the FF period were analyzed. An inverse relationship was observed between particle size and P concentration (in a specific size class), which is consistent with results reported in other studies. However, when the data were analyzed for individual treatment differences (i.e., between sites), the sediment TP mass by size fraction (% mass multiplied by TP concentration) did not differ among treatments across the different particle size classes.
The absence of a significant effect on the P-mass distribution in different size classes from differences in management (i.e., corn harvesting practices and/or manure management) suggests that this relationship may be unique for a given soil type, clay content, and mineralogy when a narrow range in soil OM exists. Approximately 57% of the P mass was associated with particles < 50 μm (silt-size). Having the TP mass-particle size relationship defined for a given set of soil conditions would allow determination of the P-mass distribution by particle size class directly from the sediment textural data. This observation has significant implications for improving P loss predictive capabilities of simple models.
Aggregate Stability. Aggregates are typically less dense and more P-enriched than primary particles making them important in particulate P delivery. Central to the question of aggregate size and P delivery is whether aggregates will remain intact or be dispersed during detachment and transport. The ability of aggregates to resist breakdown was characterized in this study by the dispersibility ratio where higher dispersibility ratio values were indicative of aggregates with lower stability. The analysis for differences among sites indicated that the median aggregate stability value for CG was greater than that of CS and SM, which were similar.
The higher aggregate stability of sediments from the CG treatment was likely the result of greater residue levels providing higher OM content and increased surface protection. Research by others suggests that higher OM levels provide additional substrate for micro-organisms to produce binding materials (polysaccharides) that would, in turn, increase aggregate stability. Possible reasons for a lack of this stabilizing effect due to manure addition at SM (manured site) include insufficient time and soil OM levels to support aggregate development, or, as others have observed, long-term manure addition does not necessarily change aggregate characteristics including stability levels.
Aggregate Size. Forty-four runoff samples were analyzed for aggregate size peaks determined by the difference between the dispersed and undispersed samples. The peaks were grouped into four size classes (5.4, 32, 160 and 570 μm). The aggregate size peaks that occurred with the greatest frequency were 32 μm (34%) and 570 μm (31%). The median aggregate size of 32 μm compared favorably with the small aggregate size (30 μm) previously identified by others and currently used in the WEPP and the RUSLE soil loss models. Two additional sizes (5.4 and 160 μm) were also identified during this investigation having a combined frequency of 35%.
Other. Regression equations that are potentially useful in management decision making were determined from the results of this study and are listed on the project Web Site and in individual publications.
Much has been written about P movement from nonpoint origins (e.g., animal feeding operations, farms, fields) through the watershed flow-system and finally to downstream lakes, rivers, and estuaries. However, areas receiving less attention are intermediate-scale fate-transport pathways linking P movement between edge-of-field and broader watershed scales. This component of our research project examined P fate and transport, as influenced by watershed features in the form of internally-drained depressional zones, characterized by colluvial sediment storage and interruption of surface runoff connectivity. Our two specific objectives were to:
- Characterize sediment movement patterns in a subwatershed characterized by colluvial sinks and internally-drained fields; and
- Examine the effects of the sediment movement patterns in this subwatershed on the potential for P mobilization and transport.
Field work for this study was conducted on a working dairy farm (300 animal equivalent units) in the North Fork Pheasant Branch subwatershed of Lake Mendota. The farm catchment is characterized by soils typical of the southern Wisconsin glaciated region with slopes ranging from 27% to absolutely flat in depressional areas. Two depressional zones were specifically investigated for this study, designated KA-1 (1.4 ha) and KA-3 (22.1 ha). The KA-1 depression overtops periodically into KA-3 during extreme runoff events. Despite the installation of a ditch and culverts to drain KA-3, this depressional area often remains waterlogged after heavy rainfall events.
Soil cores (0–70 cm) were taken at locations where signs of erosion or deposition could be visually noted. Incremental downcore soil samples were obtained by sectioning cores at 5-cm increments. Samples were then tested for particle size distribution, P levels and 137Cs radioactive fallout activity, which was used to quantify erosional or depositional magnitude. This is based on the premise that degraded locations will exhibit weak 137Cs inventories (relative to undisturbed locations) because of the continual removal of soil particles, while aggraded locations will possess higher 137Cs inventories. Soil P sorption analysis was conducted to quantify the capability of soils to retain or release P when exposed to a range of solution P concentrations. The P analysis suite tested soils for TP, water-extractable P (WEP) concentration and equilibrium P concentration at zero sorption (EPC0). WEP indicates that fraction of soil P most readily available to runoff. Though similar to WEP, EPC0 indicates the potential for soils to behave as P sources or sinks when interfacing with natural runoff conditions where DRP is usually greater than 0 mg L-1. Quantitatively, EPC0 is a good predictor of the dissolved P concentration at which the soil P level of a given sample equilibrates (no net adsorption by or desorption) with P in surrounding waters The EPC0 characterizes the ambient solution (or runoff) P concentration at which soils exhibit their maximum buffering capacity for P in the dissolved phase.
Evaluation of Site Depressional Zones Relative to the Larger Watershed. The potential for depressional zones to influence runoff, sediment, and associated constituent movement in the Lake Mendota watershed was examined with the ArcMap Hydrology Toolbox “Fill” function. This analysis illustrated that the farm catchment study area is one of several physically authentic sinks that would normally be ignored when delineating watersheds and developing conveyance networks for watershed modeling.
Sediment Movement Patterns at Erosional and Depressional Sites. Downcore soil 137Cs activities were, for the most part, consistent with DEM evaluations and field investigations conducted to identify eroded, depositional, and undisturbed locations. The 137Cs activities and inventories also reflected impacts of past agricultural land management practices (e.g., plowing) in the farm catchment. General findings were that the erosional and depressional zones were spatially distinguishable from 137Cs inventories of a nearby reference location (1936 gravesite). A second observation was that erosional locations exhibited weaker 137Cs inventories relative to the reference location while depressional zones had elevated 137Cs inventories, thus indicating that these latter areas functioned as colluvial sinks and deposition sites. Particle size analysis of undispersed soil samples also indicated greater accumulation of fine-textured sediments in depressions.Two other cores qualitatively described as “ridgetop” and “toeslope” exhibited inventories close to the reference. Finally, a third finding was that no statistical difference appeared between 137Cs inventories at KA-1 and KA-3, even though KA-3 is down-gradient of KA-1 and thus receives spillage from upslope KA-1 when it overtops during large runoff events. This pattern suggests immobilization of colluvium in the depressions.
Linking 137Cs Activities and Phosphorus Storage. Not surprisingly, depressional areas exhibited higher TP levels than erosional zones. Mean TP concentrations in plow-layer increments (0–30 cm) were also greater in depressions (921 mg kg-1) than in erosional locations (523 mg kg-1). Furthermore, mean TP levels and 137Cs inventories for the 0–30 cm increment were slightly correlated (r2 = 0.45). It is reasonable to suggest that TP surpluses in the depressions reflect sediment accumulation patterns identified by 137Cs methodology, since WEP constitutes a relatively small proportion of total soil P.
Phosphorus Sorption Characteristics in Erosional and Depressional Locations. Our assessment of P mobilization potential from subwatersheds characterized by depressional zones assumes that mobilized P will be delivered in spillage runoff primarily as DP. Particulate P (PP) associated with the most readily suspendable sediments (e.g., clay-sized particles, colloids) was considered as a secondary P transport mechanism. The sediment and TP immobilization patterns observed at the depositional coring sites appeared to support these assumptions. Levels for WEP ranged from 38.2–0.2 mg kg-1 and 14.9–1.0 mg kg-1 at erosional and depressional locations, respectively. Although WEP comprises a relatively small proportion of soil TP, others have reported that this highly labile fraction correlates with runoff DP losses more strongly than soil TP. Our farm catchment soils actually exhibited WEP levels several times larger than those measured in other studies.
Results indicate that P movement from erosional locations occurs as a runoff-driven process, while in depressional zones, downward migration and immobilization are the dominant fates affecting the DP and PP distributions, respectively. Once delivered to depressional zones, however, P fate and transport is generally affected by colluviation and downward WEP migration through macropores created by wormholes. Depressional zones appeared to act as efficient P sinks, resulting in P-enriched soil layers in deeper increments than erosional locations.
The long-term efficiency of these zones for preventing off-site P migration from this subwatershed was evaluated by examining the P sorption characteristics of soils from erosional and depressional locations. This examination indicated that soils at erosional locations exhibited strong potential for WEP migration from surficial increments (< 10 cm). Evaluations at depressional locations, however, suggested downcore migration of WEP into deeper increments (> 30 cm), in contrast with erosional soils where WEP levels were generally insignificant in subsurface increments. Though not specifically examined in this study, we hypothesized that elevated WEP concentrations in the subsurface of depressions resulted from macropore flow delivery processes, which may have been less influential in the coarser-textured erodible soils.
Together, the soil P sorption and radiometric analyses revealed three important patterns regarding P fate and transport in this internally-drained subwatershed. First, aside from supplying PP due to soil loss, erodible soils also exhibited greater potential for DP sourcing. Secondly, PP lost from erosional areas may be retained for long periods in depressional zones, as evidenced by greater 137Cs activities and TP levels (dominantly PP). Thirdly, the most labile P forms are at least partially mobile along transport pathways not related to sediment deposition or tillage patterns, instead migrating downward into the soil profile. Thus, in watersheds characterized by depressional morphology, the impact of upland DP and PP losses may be buffered by colluvial sediment accumulation and the potential for migration of labile P forms into the subsurface. Further research on cumulic soils is required, however, to determine the relative fractions of DP that remain available for mobilization within surficial deposited sediments or are transported downward through macropore pathways. Furthermore, although the P storage potential offered by depressions may serve to buffer offsite P migration during lesser storm events, more research is required to determine P delivery patterns from depressional watersheds when larger storms cause spillage runoff conditions. Our results suggest, however, that agricultural catchments with depressions will have less export of P than catchments without depressions.
- Sediment P Inventory
- P Dynamics
- Sediment Hydrodynamics
Streams in agricultural watersheds often accumulate large amounts of sediment and associated P, thus potentially reducing the amount of watershed P reaching lakes. Temporary storage in the stream not only influences the timing between loss of P from the watershed and delivery to the lake, but transformations can occur causing the chemical form or bioavailability of the P to be altered.
A comprehensive sediment and P inventory was performed on the Dorn Creek channel system including wetlands and agricultural drainage ditches. Using aerial photography, topographic information, and field verification, representative stream segments were attributed to each site. Transect profiles were analyzed that allowed for a quantification of sediment and P mass present in the entire Dorn Creek channel system. Additional transects were also sampled in other streams in the Lake Mendota watershed. From these analyses, GIS-based estimates of P and sediment mass present in the top 5 cm were generated for the stream channels and major in-line ditches (excluding major wetlands, estuaries, and minor upland ditches) throughout the Lake Mendota watershed. The 0–5 cm sediment layer was determined to represent the most active layer susceptible to resuspension and transport. However, sediment deposits in pools and other depositional locations in the stream channels were much thicker (as great as 1 meter) such that the topmost layer represented only about 5.2% and 4.4% of the total sediment and P inventories in Dorn Creek, respectively.
Large amounts of total P and BAP have accumulated in the sediments of Dorn Creek, but the amounts vary considerably with stream site. Inventories of BAP in the top 10 cm of sediment ranged from 0.1 to 14.5 mg P cm-2. Highest P concentrations were found in wetland areas and depositional areas (agricultural drainage ditches and stream pools). Total P concentrations in stream sediments at depositional sites (mean = 1920 mg/kg) were much higher than concentrations found in agricultural soils in our study and more similar to concentrations in surficial profundal sediments of Lake Mendota. Concentrations in the lower Dorn Creek wetlands (2500–5030 mg/kg) were similar to those found in the silt fraction of soils eroded from field sites in our study.
BAP (represented by the NaOH-extractable, molybdate reactive P fraction) ranged from 11 to 3020 mg/kg dry weight and constituted 7 to 76% of inorganic P (average 55%) in stream sediments. The concentrations of BAP at many sites made up a large percentage of total P, averaging 46% for all samples. Although total P in nondepositional zones was largely inorganic P, averaging 81%, the fraction represented by BAP was lower, averaging 39 % as compared to 53% for depositional sites.
A dramatic increase in sediment P was observed for median particle sizes less than ca. 30 microns found in depositional areas. In general, urban sites had greater particle size and lower P content, but this may be because the finer particles are transported more readily to the lake due to scour from urban storm flows. At several sites, sediment profiles suggest temporary deposition of fine-grained, high-P sediment above more permanent coarse-grained, relatively low-P sediments. At a transitional deposition site, the water content (percentage water) of the profile decreased with depth, but much more than that for which compaction can account. The high water content values in the surficial sediment (top 2 cm) are representative of those observed in depositional areas, and the values at greater depths are representative of high gradient scour zones. This surficial sediment likely represents highly mobile P that is transported downstream and potentially delivered to the lake. At one stream site where sediment cores were collected both in August 2005 and August 2006, P concentrations in the top 2 cm of sediment were 3 times higher in 2005 than the following year.
We investigated the accumulation and transformations of BAP as influenced by geochemical controls, mixing dynamics, sediment-water equilibrium processes, and stream site characteristics in the sediments of Dorn Creek. In addition, sediment source, transport, and mixing dynamics in Dorn Creek were examined using natural radionuclides 7Be (53-day half life) and 210Pb (22-year half life) to profile the transport of “new sediment” to specific stream sites and the depth of mixing of surficial stream sediments during runoff events.
Spatial Variability. TP and BAP concentrations and mass inventories varied widely between sites, indicating the importance of site-specific factors in controlling sediment and phosphorus retention and transport. Inventories were generally highest in sites classified as depositional and lowest at nondepositional or erosional sites. Longitudinal position along the stream channel was expected to be important due to progressive transport of fine-textured, BAP-rich sediments downstream. However, site-specific factors were more important, even though the most downstream wetland site exhibited the highest inventories and concentrations of BAP.
Classification according to depositional or nondepositional sites explained much of the difference between sites. In addition to differences in sediment depth, porosity, and percentage moisture, the sediments at depositional sites were characterized (mean values) by higher concentrations of total P, BAP, and organic C, lower organic C to total N, and organic C to organic P molar ratios, as compared to nondepositional sites.
The spatial pattern of total P and BAP accumulation in Dorn Creek sediments reflects the major role of site-specific factors, especially stream dynamics. However, geochemical factors also influenced the amounts of BAP accumulated within depositional zones and the potential for mobilization by transport of particulate or dissolved BAP. Concentrations of BAP in stream sediments were related closely to the concentrations of sediment total P (r2 = 0.97) and inorganic P (r2 = 0.98) and moderately to Fe (r2 = 0.71) and Al (r2 = 0.54). While the overall relations of TP, inorganic P, and BAP to Fe were relatively strong, sediments from some sites were clear outliers in this relationship, especially sediments from the downstream terminal wetland site. The close relation of inorganic P and BAP to sediment Fe and Al suggests that adsorption to Fe and Al oxyhydroxides may play a major role in controlling concentrations of inorganic P and BAP in stream sediments. Stream sediment TP was best correlated with percent water of the sediment for sites not including wetlands, which had higher P values. Sediment P was also well correlated with volatile solids (organic content). Surficial (0–1cm) sediment P concentrations indicated a wide range of concentrations in the Dorn Creek system.
Diagenesis. We examined sediment profiles for evidence of changes in P bioavailability during storage of sediment in the stream. Diagenesis involving inorganic P forms could alter BAP levels during sediment burial. Thus, changes in BAP concentrations with depth at a given site should reflect changes with time in either external loading or changes due to diagenesis, i.e., conversion of apatite P or organic P to BAP or conversion of BAP to other forms. In spite of modest variation, trends in BAP concentrations with depth were not apparent at most sites. In contrast, concentrations of organic C, P, and total N (primarily organic in nature) tended to decline with depth (time), indicating gradual mineralization of OM during sediment burial. Although production of BAP by mineralization was masked by the large amounts of BAP in the surficial sediments, these results show that temporary storage in stream sediments provides a good environment for breakdown of organic P derived from agricultural runoff.
BAP in Stream Waters and Sediment-Water Interactions. High concentrations of BAP (mostly DRP) were observed during peak flow in the early stages of runoff events. In addition, the proportions of organic and/or particulate P transported were highest in close proximity to the peak flow, suggesting potential sources of P to be agricultural in nature. Large variations in P concentrations were observed between events due to size and intensity of the events, but an even more pronounced concentration and time duration difference is observed between storm runoff events and snow melt runoff events. While rain runoff events show a rapid increase and decrease of P concentrations, runoff from snow melt maintained elevated stream water concentrations for longer time period. Despite occurring only a few times a year, snowmelt can have an important impact on the stream, often containing high percentages of DRP. Low concentrations in stream waters during base flow indicate that groundwater is not a major source of BAP. On an annual basis, over 90% of the total P transported by the stream occurred during storm runoff events.
Stream sediments may play an important role in buffering stream water P concentrations, sorbing or releasing P until an equilibrium concentration is reached (EPC0). When stream water P concentrations are higher than sediment EPC0 concentrations, the sediments will act as a P sink. If stream water P concentrations are below sediment EPC0 concentration, the sediments will act as a P source. In our study, EPC0 concentrations indicate that sediments deposited in Dorn Creek act as a sink for DRP during runoff events, while acting as a source for DRP during base-flow periods. This buffering effect is important due to the high DRP concentrations noted in stream waters during peak flow. DRP sorption during runoff events will reduce the amounts of P transported downstream. Although baseflow is not a large source of the BAP load carried by the stream, its importance lies in the realization that during baseflow, sediments will lose P in the form of DRP to the stream water. It is likely that response of the stream to decreasing inputs from the watershed would be partially off-set by augmentation of baseflow P concentrations by P release from the bed sediments.
Sediment Age and Mixing. Profiles of 7Be indicated that mixing during most events occurs within the upper 3 cm of surficial sediment. The lack of detectable 7Be activity below 3 cm indicates that deeper mixing occurs on time scales that are long relative to the half-life of 7Be (53 days). We used a model based on 7Be/210Pb ratios in rainfall as compared to stream surficial sediments to estimate the average age and fraction of new sediment transported and deposited in the stream at various sites during storm events. The calculated age ranged from 100 to over 400 days. The fraction of surficial sediment comprised of new sediment ranged from nondetectable to about 25%. Using this method, sediment age is influenced in part by the depth of soil erosion from the watershed. Soil below the soil surface contains low 7Be/210Pb ratios due to the more rapid decay of 7Be. Similarly, age is influenced by mixing of stream sediments. The age increases with the depth of mixing due to dilution of new watershed sediment by older stream sediment.
New sediment delivered from the watershed contains particles that have been “tagged” by 7Be and 210Pb in precipitation. The 7Be/210Pb ratio data indicated that new sediment is delivered to both depositional and nondepositional (temporary deposition) sites throughout the reach of upper Dorn Creek during events, although patterns vary among events. The tendency for greater mixing depth at depositional sites would lead to “older” surficial sediments at depositional sites. The mixing of surficial sediments during events enhances exchange of phosphate between bottom sediments and stream water through adsorption-desorption reactions through greater contact between sediments and stream water.
Although mixing occurred most frequently within the upper few centimeters of sediment, large events can cause deeper mixing. At site 6.38, the influence of deep mixing during a major event is apparent, as 7Be was detected 6.25 cm below the sediment-water interface. The higher P concentrations associated with the 7Be peak indicate that the newly introduced sediment was P-enriched soil. Based on 210Pb profiles, mixing of unconsolidated sediments over the upper 10 cm or more occurs on a time scale of years to decades, as patterns of decreasing 210Pb activity with depth were not apparent. Thus, mixing to these depths is rapid as compared to the half-life of 210Pb (22 years). We did observe patterns of declining organic concentrations of C, N, and P with depth at depositional sites, indicating mixing is not sufficiently rapid to obscure partial mineralization of organic matter during sediment burial.
Understanding P storage and transport in a fluvial system requires knowledge of sediment loads and patterns of sediment deposition and resuspension within a stream system. For example, the transport of P during high flow may be affected by sediment interactions with dissolved P, sediment residence times in bed storage, sediment particle size fractions, and sedigraph and hydrograph response to storm events. Furthermore, dead zones within the stream can affect patterns of sediment movement, acting as sinks at low flows and sources at high flows. This dual behavior could connect in-stream transient storage characteristics, which are related to the presence of dead zones adjacent to the flow areas, and sediment storage patterns. Wetlands, another unique feature of landscapes, also significantly influence the hydrology and water quality of surface and ground waters. Through the temporary storage of storm runoff, wetlands can reduce flood peaks, and retain and trap sediments and associated P nutrients. On the other hand, some observations show that wetlands can act as sediment and P sources during a large storm event. While most studies have qualitatively described the important roles of streams and wetlands in controlling P storage and transport, difficulties in measuring and quantifying the mobility of P remain.
Goals of this part of the project were to: (1) elucidate the hydrodynamic and hydraulic processes, including relationships and parameters, regulating the mobility of sediments and P in streams; and (2) quantify the hydrologic and water quality functions of wetlands in controlling stream sediments and P. Specific objectives were to:
- Determine the size of runoff events required to mobilize streambed sediments, and quantify the amounts of sediment in streams subject to short-term and long-term storage during occasional high-energy hydrologic events;
- Examine the effects of stream morphology on the dispersion and storage of dissolved and fine particulate materials in the stream and estimate the residence time for these fine-grained sediments related to BAP; and
- Elucidate the role of a wetland in controlling hydrologic processes and stream sediment dynamics under storm events of varying intensity and quantify the sediment budget of a wetland.
During the study period from 2003 to 2006, detailed measurements, including water level, flow velocity, sediment concentration, streambed sediment profile, sediment particle size and composition, and sediment transient storage, were conducted at nine study sites covering the lower Dorn Creek watershed. Reach A was located in the intermittent flow portion of the stream near stream km distance 11.0. Reach B was located downstream of the upper wetland area in a very flat portion of the stream near km distance 8.3. Reach C was located in a steep, boulder-filled section of the creek near km distance 7.1 and was the main study site of riffle-pool sequences. Reach D was located at the head of Dorn Creek’s downstream wetland near km distance 5.9 and had heavy sediment deposits. The slopes of Reaches A, B, C, and D are 0.0052, 0.0015, 0.013, and 0.0011, respectively.
Approximately every 2 weeks, cross-sectional channel profiles were monitored at the study sites to examine temporal changes in sediment storage in the stream channel. Overall the magnitude of variations in the sediment bed level was about 5 cm, which was in good agreement with our critical shear stress measurements (see below). Measurements of sediment depths showed that soft (unconsolidated) sediment deposits were primarily located in reaches with mild slope, and that steep areas stored little to no sediment. Bed sediment cores were also collected on a monthly basis within the four stream reaches and separated into four depths. Particle size and available phosphorus (AP) analysis showed that median AP values were highest near the bed surface. AP also increased with silt content, which is likely responsible for the majority of the suspended load carried in the stream.
For almost all sediment cores collected at all study sites, a loose silty sediment layer usually was observed at the top 5 cm. These results agree with literature values for noncohesive silt. The shear stress generally increased with sediment depth indicating cohesive material for lower depths. A low critical shear stress (i.e., 0.05 Pa) was measured in the top 5 cm for a typical sediment core, suggesting that streambed sediments are susceptible to resuspend under 10–20 cm/s flow velocity, which occurred frequently within the study year. The bed shear stress at site 8.25 is well correlated with sediment transport rate, further supporting the importance of bed shear stress induced by storm events to resuspend the streambed sediments.
Overall these results clearly show that an increase in the shear stress above the measured critical shear value produces a rise in the sediment transport due to resuspension of channel sediments. However, the large shear stresses only occur during the initial phase of the runoff event (rise of hydrograph). Because the rate of vertical downward erosion of the bed sediments requires high shear stresses to be sustained for long periods (more than a day) before the deeper sediments are exposed, the deeper sediments may not be resuspended even during very large events. Further analyses are planned to elucidate this important point.
Transient Storage Processes. Transient storage describes the temporary hydrologic retention of stream water apart from the main advection flow in the stream channel. This hydraulic storage increases the contact time of main-channel water with biogeochemically reactive sediments, and thus, increased transient storage is often presumed to increase nutrient retention in stream ecosystems. Transient storage can be a combination of hyporheic flow (within streambed sediments) and turbulent dead zones within the surface water, although most research to date has focused on hyporheic storage. Any flow obstruction in the stream (submerged vegetation, rocks, leaf packs, debris jams, etc.) contributes to channel roughness and subsequent flow resistance, thereby slowing the downstream passage of stream water.
In our study, solute transport processes were studied using injections of a tracer dye (Rhodamine WT). Ten dye experiments were conducted on three reaches—B, C, and D—with varying gradients. A one-dimensional hydrologic transport model was used to estimate transient storage area and exchange in streams. It was found that not only stream flow (advection) and mixing (dispersion) are important; in addition, streambed morphology (i.e., riffiles and pools, varying stream beds, and vegetations) can play a controlling role in the stream residence time of P. Reach D, where the most heavily vegetation occurs, had the highest transient storage but lower values for the exchange between main channels and stream banks, suggesting that vegetation may provide little mass exchange between the flow and storage area. Reaches B and C had smaller transient storage but larger exchange capability, which may be caused by irregularities in the channel morphology such as embayments, or in separated flow regions behind obstructions such as the boulder clusters in reach B. While these zones are small, a large amount of turbulent mixing takes place between water in the flow region and water in the storage area. Model results with transient storage were found to have 30% longer residence times in the study reach than those without transient storage.
Sediment and P Dynamics in the Upper Wetland. Wetlands are reputed to reduce peak flows and improve water quality by trapping sediment and P. However, there are relatively few studies that quantify these wetland services. We addressed this issue by monitoring streamflow, suspended sediment, total P, and DRP at automated stations upstream and downstream of the 45-ha upper wetland in Dorn Creek during the spring and summer of 2006. Ten events were observed during the observation period; the two largest events were 1- to 2-year storms. This channelized wetland is similar to many wetland systems in the glaciated portion of the United States.
All peak flows were attenuated by the wetland, with the maximum attenuation corresponding to the intermediate events. The reduced attenuation of the larger events appears to be due to filling of storage, either due to antecedent conditions or the event itself. In the case of sediment, the mass leaving the wetland in the two largest storms was twice the mass entering the wetland, which accounted for 96% of the exported sediment during the period of observation. The failure of the wetland to trap sediment was apparently due to the role of drainage ditches that trap sediment during the wetland-filling phase, and release sediment during drainage. The export of sediment during the largest events appeared to be due to remobilization of sediment that had previously been deposited in the low-gradient channel during smaller events. This hypothesis was supported by the finding that the estimated bed shear during large events exceeded laboratory measurements of the critical shear stress of bed sediment samples.
Pertaining to P, the upper Dorn Creek wetland functioned as a sink for P during small events and a source during large events with P inputs about equaling outputs for the entire study period. Although the peak concentration of P was higher at the upstream wetland site during the largest runoff event, the higher mass of P leaving the wetland was due to sustained high flow. The size of the runoff event also seemed to influence the composition of P being transported through the wetland. During large events, more sediment-bound P left the wetlands than entered, while the opposite was generally true during small events. However, more DRP entered than left the wetland as about 40% of the DRP was sequestered during the study period. This indicates that DRP in runoff was sorbed by wetland sediments, likely as water overflowed the stream banks to newly flooded areas.
Summary of Stream Research
Dorn Creek influences the composition and timing of sediment and P transported to downstream water bodies. Sediment and P tend to accumulate in the stream channel during small events but are then selectively transported downstream during larger events, although the deep sediment deposits found in the stream system suggest that more permanent deposition has occurred. However, there is a pattern of gradual “spiraling” of sediment downstream, with accumulation occurring in depositional zones, especially wetland areas. These depositional zones contain high concentrations of TP and BAP, and a high proportion of BAP. P concentrations were correlated to sediment water content, organic content, particle size, and Fe/Al concentrations, reflecting the importance of sediment physical and chemical properties and P dynamics (desorption, resuspension, and transport). Organic P is mineralized extensively during storage in depositional zones. Research conducted on the upper channelized wetland on Dorn Creek also indicates that these wetlands are not significantly reducing total P loads to downstream water bodies.
Sediment mixing influences P transport in Dorn Creek. Mixing may involve in-place disturbances as well as sediment resuspension and redeposition during events; 7Be analysis, repeated TP sampling over time, and sediment critical shear tests all suggest that mixing occurs primarily in the top 5 cm. Mixing of unconsolidated sediments during large events can reach depths of 10 cm or more within the timescale of Pb-210 decay. Taking into consideration the potential pool of TP and BAP and the depths of mixing, large amounts of P are likely to be mobilized in Dorn Creek during large runoff events. However, the large shear stresses only occur during the initial phase of the runoff event (rise of hydrograph). Because the rate of vertical downward erosion of the bed sediments requires high shear stresses to be sustained for long periods (more than a day) before the deeper sediments are exposed, the deeper sediments may not be resuspended, even during very large events. Further analyses are planned to elucidate this important point.
High concentrations of sediment and P (TP, BAP, and DRP) were measured in stream during runoff events. The high DRP concentrations were due mainly to the high concentrations in the agricultural runoff. Although some DRP was retained, apparently by sorption to stream/wetland sediments, much of the DRP was not attenuated by in-stream processes. Sediment P concentrations were highest in smaller particles, which are easily mobilized and transported during runoff events. Transient storage in both the sediments and during flow (due to advection and stream morphology) can alter the time it takes for sediment P to be transported in Dorn Creek, especially in the wetland areas.
Modeling Tools Assessment
The overall objective of the modeling component of the project was to evaluate and improve modeling tools used to assess P transport in agricultural watersheds over a wide range of spatial scales. One study was conducted with the SWAT model, but most of the modeling effort was a three-scale evaluation of APEX from field to large watersheds, examining the role of internal topographic sinks on sediment and phosphorus transport. Simulations integrated data collected by both field and stream research teams.
Assessment of SWAT. A key question in scaling from fields to large watersheds is whether the spatial representation within the model affects the fluxes. SWAT’s predictions of runoff and sediment yield respond differently to spatial data aggregation. We hypothesized that the aggregation effect observed is partly due to model structure, and hence the submodels for runoff and sediment yield respond differently to the same change in spatial representation. The study focused on SWAT’s integration of the MUSLE with HRU. The evaluation was conducted using 4 years of water and sediment discharge for the Pheasant Branch Watershed, and its North Fork and South Fork tributaries. Sediment generation was calculated for each HRU with the MUSLE equation.
Results showed that HRUs do not conserve sediment with scale of watershed representation. Instead, HRUs introduce almost half of the variability in sediment generation, which other researchers have previously attributed to input data aggregation. This occurs for two reasons. First, MUSLE defines a nonlinear relationship between sediment generation and HRU area, but the sediment load is scaled linearly from the HRU level to the subwatershed level. Second, HRUs aggregate land areas without regard for the surface connectivity assumptions, which are implicit in MUSLE. This means that runoff time-of-concentration is arbitrarily determined by HRU area rather than topographic controls on surface runoff. Thus, there is no physical connection between the sediment generated at each of the HRUs and the subbasin total. These findings have direct implications for the interpretation of model output, and such conflicts can potentially mask the effect of different land use on soil erosion.
We made the following recommendations for using SWAT and other MUSLE-based models, including APEX:
- Greater attention should be given to structuring the data inputs to match the underlying assumptions of submodels. In particular, we recommend that the MUSLE equation be used with small, homogeneous catchment areas;
- MUSLE and similar runoff-energy derivatives of USLE must be calculated within drainage areas that are enclosed by no-flux boundaries (i.e., divides, watershed boundaries) and have a single absorbing boundary or sink point (i.e., catchment outlet); and
- Although APEX explicitly defines connectivity between land surface patches, these should be topographically defined as in the previous recommendation.
Field-to-Large Watershed Simulations: Representing Internal Sinks. We hypothesized that small microtopographic depressions in aggregate represent potentially large sinks for sediment and P. Such sinks are prevalent within the Lake Mendota watershed due to its hummocky glacial topography. DEMs that are widely used in distributed hydrologic modeling generally view these interior depressions within catchments as errors in the DEM, even though they may be hydrologically significant features. Watershed models based on DEMs cannot directly resolve small-scale topographic depressions, which may act as sinks of solid/dissolved P. We asked the following questions: Can watershed-scale hydrological models effectively represent internal sinks? What impact do such sinks have on sediment and P delivery? Can watershed modeling tools effectively incorporate such features in large watershed analyses?
The overall modeling approach employed a three-scale evaluation scheme: (1) field scale, represented by two small source areas (on the order of 10-2 ha); (2) a small watershed scale represented by Dorn Creek (on the order of 104 ha); and (3) a large watershed, the Yahara River (on the order of 105 ha). We used, and when necessary modified, APEX (version 1310) for this study. The model evaluation activities related to two field studies within the project and new (USGS gauge no. 05427850) and existing (USGS gauge no. 05427718) gauge data on the Yahara River. First, at the field scale we evaluated APEX in the small catchment areas (KA-1 and KA-2) within the North Fork of Pheasant Branch. Second, for the small watershed scale we evaluated how APEX estimated sediment and P storage along the main channel of Dorn Creek. Third, we examined the implications of scaling up to the Yahara watershed, using relationships obtained in Dorn Creek.
Our working hypothesis was that sediment transport parameters could be used as proxies for the functioning of surface depressions to obtain the correct sediment response. An immediate problem with this is that DEM-derived parameters for modeling sediment transport cannot yield standing water, and clay-sized particles are not hydraulically-limited. The alternative was to explicitly prescribe depressions using proxy hydrologic objects (e.g., primary pools or reservoirs) that were explicitly inserted into the hydrologic flow paths.
For the two small catchments (KA-1 and KA-2), areas extracted from the DEM were 1.60 x 104 m2 and 5.10 x 104 m2, respectively. However, field-surveyed contributing areas to the monitoring stations were only 3.26 x 102 m2 and 1.10 x 102 m2. In both cases the actual contributing areas were smaller than predicted because of topographic sinks a short distance upstream of the monitoring stations. The over-estimated areas resulted in several orders of magnitude overestimation of predicted discharge and sediment yield. By incorporating reservoirs at the known coordinates of the topographic sinks, the order-of-magnitude of simulated fluxes generally matched that of the observations.
For the small watershed evaluation, we examined the potential of using reservoirs to represent sinks within Dorn Creek. The watershed was delineated and partitioned into 61 subbasins using a DEM gridded at 3-m resolution. The number of subbasins was chosen to maximize the number of channel segments, and hence resolution, along the main channel. Reservoirs were employed in a nested sense by adding them to the outlets of each of the subbasins along the main channel.
Three initial simulations were run: (1) with no reservoirs; (2) with reservoirs parameterized using the same value for the height parameter in all locations along the main channel; and (3) with reservoir heights optimized to variable levels based on measured stored sediment within each segment of the main channel. Simulations of Dorn Creek were run for 20 years, with sediment storage at end-of-simulation examined for comparison to measured storage.
The model with variable reservoir heights explained 99% of the variance in accumulated stored sediment, although it underpredicted the absolute differences among locations along the main channel (slope = 0.93). Since these were not calibrated as much for absolute value as for relative spatial patterns, this was considered a reasonable calibration for basing further simulations. Uniform height reservoir and no-reservoir models were poor predictors of sediment storage. Predictions for P generally follow the spatial trend when variable height reservoirs are employed, but there was no discernible relationship in P measurement between the no-reservoirs and the uniform height reservoirs. Pertaining to how each of the models performs with respect to measured sediment storage, without reservoirs a significant proportion of the main channel is acting as a sediment source rather than sink. We suppose that this could accurately reflect the true dynamics of some of the reaches of the system once they have reached capacity. Our model results suggest that most of the main channel is still aggrading.
The relationship between reservoir volume and stream slope developed in Dorn Creek did not scale up to the Yahara River. While reservoirs could be used to match modeled and measured sediment yield from the watershed, the same was not true for water discharge. In particular, low flows were shut down with any reservoirs present. Sediment yield and water discharge both show a highly nonlinear response to very small changes in reservoir height at low values of this parameter, and very little response at larger reservoir heights due to production of a series of standing pools in the channel. Even at relatively low heights (0.5 to 1.0 mm), total water discharge was substantially reduced, especially during low-to-moderate events. This was not remedied even when all reservoirs were moved into second-order basins (reservoirs cannot be placed in first-order basins where APEX has no true stream channel). Second-order and likely most of the first-order basins defined for Yahara in APEX have sufficient contributing areas to have sustained flow. We were able to correctly obtain water and sediment discharge from both Yahara gauges by adjusting maximum channel capacity. Thus, by scaling up to larger watersheds using larger subbasins, direct parameterization of microtopographic sinks is not possible and must instead be handled through parameterization of the soil erosion and channel transport routines. This is in contrast to KA-1 and KA-2, which are extreme examples of intermittent streams since they only overtopped their respective sinks during very large runoff events and when the soils were sealed following a manure application. Intermittent streams may have standing water during small storm events or shortly following larger storm events.
- Reservoirs are a suitable proxy for internal sinks at the field scale where flow is intermittent in response to storm discharge. In these regions microtopography enables standing water between events, and thus deposition of clay-size particles and attached P.
- In small watersheds where flow is intermittent, such as the upper part of Dorn Creek, reservoirs may be appropriate. Where flow is perennial, reservoirs are not suitable proxies for sinks. This highlights a scale limitation of distributed watershed models. While APEX can form any size of distributed land elements, such as individual fields, these have to honor topographic boundaries (divides, channels) when using energy-based extensions of USLE. Future effort is needed in developing automated tools for defining such small-scale distributed components.
- In large watersheds APEX is more appropriately parameterized to deposit sediment by slowing down discharge. Since this is hydraulically not equivalent to having standing pools at field scales, suitability of this approach needs to be evaluated. An initial understanding of hydrologic conditions throughout the watershed will help identify areas of intermittent flow.
- Some scientists argue that distributed models are overkill when trying to predict whole-watershed outputs. Our results to date support an argument in favor of distributed models for sediment and P predictions for Lake Mendota’s subwatersheds.
Research results will be disseminated using a variety of methods including the more traditional peer-reviewed journal publications and conference presentations as summarized in this report. In addition to the more traditional channels, a focused method will be implemented using direct contact with a target audience. The target audience for the project includes: watershed management researchers (i.e., USDA-ARS and university researchers), watershed managers (i.e., USDA-NRCS, model developers, and state and local natural resource management agencies).
An email list of individuals involved in watershed nutrient management activities will be developed based on publications, organizational affiliation, and other activities relating to watershed nutrient management. An introductory email summarizing the project will be sent along with the site URL. Web site activity will be monitored and tracked. The response to email distribution can be directly tracked, evaluated and modified as needed.
Journal Articles on this Report : 9 Displayed | Download in RIS Format
|Other project views:||All 50 publications||9 publications in selected types||All 9 journal articles|
||Cabot PE, Nowak P. Planned versus actual outcomes as a result of animal feeding operation decisions for managing phosphorus. Journal of Environmental Quality 2005;34(3):761-773.||
||Cabot PE, Karthikeyan KG, Miller PS, Nowak P. Sediment and phosphorus delivery from alfalfa swards. Transactions of the American Society of Agricultural and Biological Engineers (ASABE) 2006;49(2):375-388.||
||Cabot PE, Pierce FJ, Nowak P, Karthikeyan KG. Monitoring and predicting manure application rates using precision conservation technology. Journal of Soil and Water Conservation 2006;61(5):282-292.||
||Chen E, Mackay DS. Effects of distribution-based parameter aggregation on a spatially distributed agricultural nonpoint source pollution model. Journal of Hydrology 2004;295(1-4):211-224.||
||Hoffman AR, Armstrong DE, Lathrop RC, Penn MR. Characteristics and influence of phosphorus accumulated in the bed sediments of a stream located in an agricultural watershed. Aquatic Geochemistry 2009;15(3):371-389.||
||Lathrop RC. Perspectives on the eutrophication of the Yahara lakes. Lake and Reservoir Management 2007;23(4):345-365.||
||Panuska JC, Karthikeyan KG, Miller PS. Impact of surface roughness and crusting on particle size distribution of edge-of-field sediments. Geoderma 2008;145(3-4):315-324.||
||Panuska JC, Karthikeyan KG. Phosphorus and organic matter enrichment in snowmelt and rainfall-runoff from three corn management systems. Geoderma 2010;154(3-4):253-260.||
||Rogers JS, Potter KW, Hoffman AR, Hoopes JA, Wu CH, Armstrong DE. Hydrologic and water quality functions of a disturbed wetland in an agricultural setting. Journal of the American Water Resources Association 2009;45(3):628-640.||