Edge-of-field monitoring was conducted by the University of Wisconsin Discovery Farms program between 2004 and 2019 across Wisconsin and Minnesota. Twenty-two farm fields were monitored across 125 site-years and 1339 individual runoff events. For each runoff event, total runoff and total phosphorus (TP) and dissolved P (DP) (both as flow-weighted concentration and load) were measured following Discovery Farms and USGS protocols. For each runoff event, some variables changed event to event, while others did not. Some variables never changed per site (e.g., texture, soil test P), while others changed annually (crop, previous crop, tillage) or changed seasonally [e.g,. soil condition (frozen vs. non-frozen) or soil surface condition (actively growing crop, crop residue, or no residue)]. Many variables are provided that change by event, most notably time since manure application, rainfall, and time since last tillage operation. Collectively, this dataset provides as much information as possible to assess the influence of inherent soil factors, farmer management, and weather on P loss.

Site Descriptions

Discovery Farms Wisconsin and Discovery Farms Minnesota, in partnership with the U.S. Geological Survey, have collected 125 site-years (2004 - 2019) of event-based EOF surface runoff data from 22 fields that contain a total of 1339 individual runoff events. Sixteen of the fields are in Wisconsin (n = 803 individual events), and six are in Minnesota (n = 536 events) (Figure 1). Across all sites, the dominant crops were corn, soybean, and alfalfa. The drainage area of these sites varied from 2.4 to 17.0 ha with an average of 7.9 ha. Most of these sites were located in fields with slopes greater than 3% (19 sites), with only three sites having slopes less than 3%. Soil texture was characterized as primarily silt loam (17 sites) according to the National Cooperative Soil Survey (websoilsurvey.cs.egov.usda.gov), with other textures being loam (2 sites), clay loam (1 site), silty clay loam (1 site), and sandy clay loam (1 site).

Soil Analysis and Runoff Sampling

Water quantity and quality data were collected at the edge of each field. Runoff events were defined as the time at which rainfall or snowmelt-induced runoff began until the runoff ceased (Stuntebeck et al., 2008). Runoff (water quantity) was collected using H or trapezoidal flumes with plywood enclosures and wooden wingwalls in waterways or points of concentrated flow near the fields (USGS, 2016). Non-submersible pressure transducers, along with nitrogen bubbler systems tracked the stage within the flumes. Water quality samples were collected through automated time-based ISCO samplers and stored in a refrigerated enclosure (Stuntebeck et al., 2008; Rassmussen and Matteson, 2011). Up to 24 samples per day were collected from the ISCO samplers which were combined into one flow-weighted composite sample. These samples were retrieved within 24 hours of the discharge and sent to University of Wisconsin-Stevens Point Water and Environmental Analysis Laboratory for analysis. The TP was analyzed following persulfate digestion (APHA, 2017) and DP was analyzed following filtration with a 0.45 mm filter; P concentration was determined using the ascorbic acid method (Murphy and Riley, 1962). Event runoff volume was multiplied by the concentration to obtain TP and DP loads for each collection time. If a single runoff event occurred over the course of two sampling days (which would include two separate TP and DP samples), the load for each day was summed and then divided by the total runoff across the sampling periods to calculate the flow-weighted mean concentration (FWMC). Thus, all event concentrations are referred to as FWMC based on composite sampling within individual or across multiple sampling periods. All water collection, handling, and storage, as well as precipitation and soil temperature measurements are described in Stuntebeck et al., 2008 and Rassmussen and Matteson, 2011. Soil sampling for soil test phosphorus (STP) was conducted on each field at the onset of runoff monitoring at 0-5 and 0-15 cm depths and was determined using Bray-P1 and Olsen methodology (Frank et al., 2012).

Dataset Construction

To investigate the mechanisms governing edge-of-field runoff P losses, the dataset compiled for this study included event response variables (TP and DP loads and TP and DP FWMC), runoff amount (mm) and type, weather data, agronomic management practices, and site characteristics that were identified as having the potential to influence TP and DP loss. The runoff event type was categorized as either rainfall runoff on non-frozen soil, rainfall runoff on frozen soil, or snowmelt runoff. The soil condition was considered frozen if the soil temperature was below 0°C at any point during a day. The weather data includes individual rainfall event variables were identified for each runoff event including total precipitation, storm duration, average intensity, maximum intensity (as 5-, 10-, 30- and 60-minute maximum intensities), and antecedent rainfall one, two, and three days prior to the runoff event. Agronomic management was information collected from each farmer and includes if the site was tile drained or not (Y/N), the type of tillage used [no-till (NT), reduced tillage (RT) (which includes vertical or strip tillage or tillage every other year), or conventional tillage (CT) (where at least one tillage pass occurred each year)], if the site had manure application during the monitoring time period (Y/N), and the active crop grown or most recently harvested (corn, alfalfa, corn & alfalfa, soybean, or pea). We also created a variable of ‘time since manure’ which was the days between a manure application and the runoff event. Site characteristics include soil hydrologic group (SHG), soil condition, surface condition. 0-5 cm STP, and 0-15 cm STP. Soil hydrologic group (A, B, C, and D) was obtained from Web Soil Survey (websoilsurvey.nrcs.usda.gov) and indicates the runoff potential of the soil and integrates effects of saturated hydraulic conductivity (to the least transmissive layer), depth to water-impermeable layer and depth to water table (NRCS, 2009). The soil condition is if the soil was frozen or not at the time of the runoff event. The surface condition variable described the state of the field at the time of the runoff event and include: (1) an actively growing crop (crop), (2) crop residue (no actively growing crop or a cover crop) (residue), and (3) no cover (no cover). Surface condition was designated as ‘crop’ during the time between canopy closure of the current crop, and a subsequent tillage pass or the harvest of corn silage as this represents high potential for the canopy to reduce raindrop impact and subsequent soil displacement. For corn, the estimated canopy closure date was June 20th, for soy and pea this date was July 15th, and for alfalfa this date was June 20th. When there were multiple crops planted at the site, ‘crop’ would start when the crop planted on over 50% of the site reached canopy closure. The residue category represented time periods when plant matter was left on the field without any tillage occurring, or when cover crops were actively growing or recently terminated. The no cover category represented when the field was fallow, either following a tillage pass or the harvest of corn silage.

It is important to note that note all input variables in the model reflect the same number of runoff events. For example, five of the variables did not change over time for each site: SHG, tillage, manure application, 0-5 cm STP, and 0-15 cm STP, while four of the variables varied on an annual or within year basis: crop, surface condition, soil condition, and event type. All other variables were continuous variables that changed event by event.