ASAE Conference Proceeding
This is not a peer-reviewed article.
Measurement and Prediction of Phosphorus Transport from Swine Manure at the Watershed Scale
T. J. Sauer, K. C. Sreematkandalam, U. S. Tim, D. E. James, and J. L. Hatfield
Pp. 535-541 in the Animal, Agricultural and Food Processing Wastes, Proceedings of the Ninth International Symposium, 11-14 October 2003 (Raleigh, North Carolina, USA), ed. Robert Burns. ,11 October 2003 . ASAE Pub #701P1203
Livestock production facilities are coming under increased scrutiny with regard to transport of phosphorus (P) from fields receiving animal manures. The objective of this study was to measure and simulate how swine manure management affects P export from a row crop watershed. Stream water sampling was conducted for one year at 14 locations within the Tipton Creek watershed in central Iowa. Data on soil P concentrations and manure production were also collected. Geographic Information System (GIS) data layers (topography, soils, land use, and land cover) were prepared to facilitate simulations using the Agricultural Non-Point Source model (AGNPS). The average dissolved reactive (DRP) and total phosphorus (TP) concentrations in stream water from 41 sampling dates were 0.14 and 0.21 mg L -1 , respectively. Total P export from the Tipton Creek watershed from April 1, 2000 to April 1, 2001 was estimated at 10.6 metric tons (Mg) for an average loss of 0.52 kg P ha -1 Y -1 . This loss represents approximately 2.8% of the estimated applied P (swine manure and commercial fertilizer). Four rainfall events during the 2000 growing season and snowmelt in March 2001 were responsible for 91% of the P transport from the watershed. Using animal inventory numbers and standard P excretion values, swine manure was estimated to supply approximately 35% of the land-applied P. AGNPS simulations of the 4 major storm events from 2000 were used to predict future P transport from an assumed 5% annual increase in swine production and subsequent increase in P application to soils of the watershed. These simulations indicated that, without any changes in current management , a 40% increase in P transport in such runoff events could occur in the next 5 years. However, a combination of diet manipulation and use of high available-P feedstuffs and/or phytase enzyme has been shown to reduce P excretion by as much as 50%. Broad adoption of these practices would likely prevent any increase in P production from swine facilities and potentially reduce overall manure P production.KEYWORDS. animal manure, phosphorus, water quality, modeling.
Livestock production facilities are coming under increased scrutiny with regard to runoff of phosphorus (P) from fields receiving animal manures. Much of the research dealing with P transport from swine manure has been on small-scale research plots (Gupta et al., 1997; Liu et al., 1997). Unlike research plots, watersheds are not uniform in their surface properties (slope, soil type, nutrient content, drainage, residue cover etc.) or management (crop rotation, tillage system, weed control, fertility program etc.). Simulation models are a tool that allow users to efficiently organize data by location within a watershed, compare simulations of nutrient transport with observations in the field, and predict how changes in application rates or management practices would impact water quality parameters. One model specifically designed for use in agricultural watersheds is the Agricultural Non-Point Source (AGNPS) model (Young et al., 1987). A Geographic Information System (GIS) interface for the AGNPS model allows spatial representations of landscape data to be used by the model (Jolly and Tim, 1994). Use of the ArcView ® -AGNPS model provides an opportunity to predict how manure and fertilizer management affect P transport to surface water.
Collect soil, water, and manure data and management information to assess the amount of P from swine manure that is transported out of a watershed with intensive swine production.
Compare predictions of P transport using the ArcView ® -AGNPS model with water quality data collected at the watershed outlet.
Predict future trends in P transport by extending current trends in swine production and soil P levels.
The Tipton Creek watershed is located in Hardin and Hamilton Counties in central Iowa, USA. The watershed has an area of 201 km 2 of which 84% is in row crop production (corn, Zea mays , L. and soybean, Glycine max ). In 2000, there were 26 confined animal feeding operations (swine) in the watershed, primarily concentrated in the western (headwaters) portion of the basin. These facilities currently produce 125,000+ animals for the slaughter market each year. Water sampling of Tipton Creek began on May 9, 2000. Fourteen stream sampling locations were chosen to provide a representative sampling of water quality within the watershed. Sample sites include locations upstream from swine production facilities, within the area of concentrated swine production, and downstream to the watershed outlet. Each location was visited every 1-2 weeks. A flow monitoring/storm water sampling station was installed at the watershed outlet. A rating curve was developed for this station to convert water depth measured with a pressure transducer and micrologger to discharge (m 3 day -1 ) that enabled P load estimates to be made. Discharge on days with missing data was estimated from flow at a U.S. Geological Survey (USGS) gaging station below the confluence of Tipton Creek and the Southfork of the Iowa River. Water samples were collected on 41 days during both baseflow and stormflow conditions. Over 350 samples were analyzed for total phosphorus (TP) and dissolved reactive phosphorus (DRP). Total P was determined by an acid persulfate digestion of unfiltered water samples. Samples for DRP analysis were filtered through a 0.45 µ m filter. Concentrations of P in both the acid digests and filtered samples were determined by continuous flow injection analysis.
Data on soil test P (Bray 1 extract, Bray and Kurtz, 1945) and fertilizer applications for over 2000 ha of cropland within the Tipton Creek watershed were provided by a local fertilizer dealer. These data indicated that most soils have low to optimum concentrations of Bray 1 P (~ 20 mg kg -1 ) and the average fertilizer rate was 21.1 kg P ha -1 of cropland. With 12,165 ha of corn in the watershed in 2000, this average rate of fertilizer P application would require 252 Mg of P fertilizer. The most recent (1999-2000 growing season) fertilizer sales data were obtained for Hardin and Hamilton counties (Iowa Department of Agriculture and Land Stewardship, 2000a, 2000b) and pro-rated for the Tipton Creek watershed to obtain another estimate of P fertilizer application. This method produced an estimate of 235 Mg of P fertilizer per year or 19.3 kg P ha -1 that was in good agreement (within 7%) of the estimate from the fertilizer application and soil test P data. Phosphorus removal in corn harvested for grain at yields typical of the area was estimated at ~25 kg P ha -1 (Pierzynski and Logan, 1993).
Phosphorus present in swine manure produced in the watershed was estimated from estimates of swine production and total P content in swine manure. Animal census numbers were obtained from the Iowa Department of Natural Resources and the 1997 Census of Agriculture (National Agricultural Statistics Service, 1999). Phosphorus concentration in the manure was estimated from ASAE Standards (1993). Total P present in the swine manure on an annual basis was estimated at 132 Mg, which corresponds to 10.8 kg P ha -1 in corn production in 2000. Spatial analysis indicated that >95% of the watershed area was within 3.2 km of a swine production facility. Since the estimated P fertilization rate was below the crop removal rate, and most corn fields were in close proximity to a swine production facility, it was assumed that the average P application rate for the watershed was the sum of the fertilizer and manure inputs (i.e. 21.1 + 10.8 = 31.9 kg P ha -1 ).
Input necessary to run the AGNPS model were collected and assimilated into ArcView ® GIS data layers (topography, soils, land use, and land cover). Four rainfall events generated runoff during the 2000 growing season. Rainfall depths and intensities for the four storms ranged from 20.3 to 37.6 mm and 4.6 to 81.3 mm h -1 . Each of these events was simulated with the AGNPS model. A P application rate of 33.6 kg ha -1 (30 lbs P A -1 ) was used for all corn acreage in the watershed. To project the effect of continued expansion of swine production in the watershed, simulations were completed using these storm characteristics for five years into the future by assuming that swine production will increase at a rate of 5% per year. Model parameters (soil P, P application rate, and amount of P extracted by runoff) were adjusted to simulate the impact of increased manure application rates on P concentrations and loads at the watershed outlet.
Water Quality Monitoring
Results of the water quality sampling are summarized in Figures 1 - 4. Figures 1 and 2 show the average concentration of DRP and TP for all samples collected at each site; Figs. 3 and 4 show the average concentrations for all sites by sampling date. Sites are numbered beginning at the headwaters of the watershed (TC1). Site SI1 is the USGS gaging station on the Southfork of the Iowa River approximately 3.2 km downstream from the lowest sampling site (gaging station) on Tipton Creek (TC25). The average DRP and TP concentrations were 0.14 and 0.21 mg L -1 , respectively. The ratio of DRP concentration to TP concentration averaged 0.66, indicating that 66% of the P in the stream water was in the dissolved form. However, since most samples were not collected during stormflow periods, this ratio is likely more representative of baseflow conditions and may not accurately reflect the dominant form of P transported on an annual time scale. Phosphorus transport from cultivated watersheds is often primarily associated with sediment produced by erosion during intense storm events. Sites TC5 and TC12 had the highest concentrations of both DRP and TP. Site TC5 is located just downstream from the highest concentration of swine production facilities in the watershed. Site TC12 is located downstream from a pasture grazed by beef cattle, which were not excluded from the stream.
Figure 1. Mean DRP at sampling locations in Tipton Creek. Sites are displayed beginning with TC-1, upstream from swine production facilities, to site SI-1 just below the confluence of Tipton Creek with the Southfork of the Iowa River. Site TC-5 is located downstream from an area with a high concentration of swine production facilities. Error bars represent one standard deviation from the mean.
Figure 2. Mean TP concentrations for the same sites as shown in Fig. 1.
Mean DRP and especially TP concentrations had greater concentrations during stormflow periods in May-July 2000 and snowmelt in March of 2001 (Figs. 3 and 4). Ranges of mean concentrations on individual sampling dates varied from 0.06 to 0.47 mg L -1 for DRP and 0.07 to 1.16 mg L -1 for TP. The peak concentration of DRP occurred on March 19, 2001 during snowmelt with other peaks on May 31 and June 14, 2000 following rainfall events. The peak TP concentration occurred on May 31, 2000, which followed a brief but intense rainfall of 20.3 mm in 15 min on May 30.
On a historic basis, total precipitation during the study period was near-normal. Precipitation at the National Weather Service station located in Eldora, IA, approximately 10 km NE of the watershed outlet, totaled 880 mm, which was only 28 mm above normal (National Oceanic and Atmospheric Administration, 2001). Precipitation patterns during the 2000 growing season, however, produced above-normal rainfall in May-July followed by below-normal rainfall in August-October. No runoff events were observed after July 2000 until snowmelt in March 2001 and all subsurface drain tiles in the watershed also ceased flowing.
Figure 3. Mean DRP concentrations for all sampling locations by sampling date. Error bars indicate one standard deviation.
Figure 4. Mean TP concentrations as for Fig. 3.
Discharge data and measured P concentrations were combined to estimate TP export from the Tipton Creek watershed from April 1, 2000 to April 1, 2001. The estimated TP export was 10.6 Mg or an average P loss of 0.52 kg ha -1 of watershed area. Four rainfall events (May 30, June 11, June 14, and July 10) during the 2000 growing season and snowmelt in March 2001 were responsible for the transport of >90% of the P from the watershed (Fig. 5). By contrast, P export during baseflow periods averaged only ~1 kg day -1 . The TP load was almost equally divided between DRP (5.5 Mg) and sediment P (5.1 Mg) contributions. Measured sediment export was 10,910 Mg, which represents soil loss equivalent to 0.54 Mg ha -1 for the entire watershed or 0.64 Mg ha -1 of cropland. It should be noted that rainfall patterns in 2000 produced fewer runoff events than would be expected in a year with more normal rainfall distribution. Thus sediment and P export values reported here may be less than the long-term average.
Figure 5. Total P load at the outlet of the Tipton Creek Watershed.
Water Quality Modeling
The AGNPS model was used to simulate runoff, erosion, and P transport from the four precipitation events that produced measurable stormflow at the watershed outlet. AGNPS predicted runoff volume within 10% of the measured runoff but predicted TP loads that were 1.9 times greater than measured. These trends can be attributed to the model consistently predicting greater DRP concentrations, which were on average 2.9 times greater than measured. Sensitivity analyses and parameter adjustments were unsuccessful in achieving closer agreement between measured and modeled DRP concentrations. Possible explanations for these findings include dilution of stormflow with tile drainage water containing less DRP, and retention of runoff P in riparian buffers. Both tile drainage and riparian buffers are common throughout the watershed, and their potential effects could not be simulated in AGNPS.
AGNPS simulations were completed using the precipitation event characteristics and surface conditions of the four runoff events from 2000 with an assumed 5% annual increase in swine production and concomitant increase in P application to the crop land soils of the watershed. Even though agreement between measured and predicted P losses for these events was not close, this exercise was intended to demonstrate relative differences brought about by changing soil P concentrations with increased swine production. Simulations of these four runoff events indicate that the level of increased production for 5 years, under the current management system , could result in a 40% increase in TP export from the Tipton Creek watershed. Increased swine production in the watershed should lead to reduced fertilizer P application rates. A holistic approach to manure P management that incorporates changes in animal diet and manure application rates scaled to reflect crop removal rates is most likely to successfully avoid excessive P buildup in soils of the watershed. A combination of swine diets with lower P concentrations and use of high available-P feedstuffs and/or phytase enzyme has already been utilized on a trial basis by swine producers in the watershed and has reduced P excretion by as much as 50%. Broader adoption of these practices over the next five years would likely prevent any increase in P production from swine facilities and could even reduce manure P production compared to current levels.
Recent scrutiny of the animal agriculture industry has focused on soil, water, and air quality impacts from livestock production. Although an estimated 125,000+ swine are produced annually in the Tipton Creek watershed, 65% of the land-applied P was applied as commercial fertilizer. A total of 252 Mg of fertilizer P and 132 Mg of manure P were applied, which is equivalent to 31.5 kg P ha -1 Y -1 on all fields in corn production. This application rate is near the estimated crop removal rate of 25 kg P ha -1 Y -1 . Total P runoff losses during the monitoring period were nearly equally divided between soluble and sediment-bound forms and were equivalent to 2.8% of the applied P. Stream water samples collected at one location near the area of concentrated swine production had greater concentrations of DRP and TP but no significant effect on overall stream water quality was apparent. As >90% of the P export occurred during four rainfall-induced runoff events and spring snowmelt, soil and water conservation measures that reduce runoff and erosion during large hydrologic events would have the greatest impact on reducing P export.
This project was funded in part by the National Pork Producers Council, on behalf of the National Pork Board. The assistance of Kevin Cole and Jeff Nichols (NSTL) with the installation and operation of the gaging station, Paul Doi, Anna Myhre, and Kelly Weichers (NSTL) in collecting water samples, Stan Kulow (Prairie Land Cooperative) in providing soil test P and fertilizer data, Marcus Mueller (Heartland Pork Enterprises, Inc.) for discussions on manure management practices, and Gary Hillmer and Marv Hoffman (USDA-NRCS) for information on cropping systems is appreciated.
ASAE Standards , 40 th ed. 1993. D384.1. Manure production and characteristics. St. Joseph, Mich.: ASAE.
Bray, R. H., and L. T. Kurtz. 1945. Determination of total, organic and available forms of phosphorus in soils. Soil Sci. 59:39-45.
Gupta, R. K., R. P. Rudra, W. T. Dickinson, and G. J. Wall. 1997. Surface water quality impacts of tillage practices under liquid swine manure application. J. Am. Water Resour. Assoc. 33:681-687.
Iowa Department of Agriculture and Land Stewardship. 2000a. Iowa fertilizer distribution, July 1, 1999 through December 31, 1999. Des Moines, IA.
Iowa Department of Agriculture and Land Stewardship. 2000b. Iowa fertilizer distribution, January 1, 2000 through June 30, 2000. Des Moines, IA.
Liu, F., C. C. Mitchell, D. T. Hill, J. W. Odom, and E. W. Rochester. 1997. Phosphorus recovery in surface runoff from swine lagoon effluent by overland flow. J. Environ. Qual. 26:995-1001.
National Agricultural Statistics Service. 1999. 1997 Census of Agriculture. U. S. Department of Agriculture, Washington, D.C.
National Oceanic and Atmospheric Administration, 2001. Climatological data – Iowa. National Climatic Data Center, Asheville, NC.
Pierzynski, G. M., and T. J. Logan. 1993. Crop, soil, and management effects on phosphorus soil test levels. J. Prod. Agric. 6:513-520.
Tim, U. S., and R. Jolly. 1994. Evaluating agricultural nonpoint-source pollution using integrated geographic information systems and hydrologic/water quality model. J. Environ. Qual. 23:25-35.
Young, R. A., C. A. Onstad, D. D. Bosch, and W. P. Anderson. 1987. AGNPS, agricultural non-point source pollution model. USDA-ARS, Morris, MN. Conserv. Res. Rep. 35.