Methods

Temperature and relative humidity were recorded using Onset HOBO U23 Pro v2 temperature/relative humidity sensors in RS1 solar shields, which were attached to streetlight and utility poles in and around Madison, Wisconsin. The sensors were installed at a height of 3.5 meters, were oriented north, and automatically recorded instantaneous temperature and relative humidity every 15 minutes beginning in March 2012.

Methods

Soil moisture and temperature were collected using Decagon 5TM sensors at depths of 10, 35, and 65 cm at each site, and an additional deeper site (90 cm for W6, 110 cm for W7, and 125 cm for W8). Soil water potential was collected at 35 cm at each site using a Decagon MPS2 sensor. All data were collected at 15-minute resolution and stored in a Decagon EM-50G datalogger. Sensors were installed after planting (April-May) and removed prior to harvest (September-October) in 2012, 2013, and 2014. Installation was done by digging a soil pit adjacent to a planted strip and installing sensors into the undisturbed face, so that sensors were directly beneath plants.

Methods

Yield data was collected at the time of harvest following the 2012 and 2013 growing season using a John Deere 9660 combine equipped with a Greenstar yield monitoring system. Yield data was cleaned by removing any polygons collected during pre-harvest check strips, turn-around at the end of rows, any polygons less than 6 m2, and any polygons where yield exceeded the record reported yield for Dane County WI (327 bu ac-1). Yield was normalized by calculating z-scores (the number of standard deviations away from the mean) within each field and each year. 2012 data was resampled to the 2013 polygon boundaries by taking the mean of all 2012 polygons with their centroid within each 2013 polygon. For each polygon, 2013 interpolated groundwater metrics were extracted using ArcMAP 10.2 software. A full description of this methodology is contained in Zipper et al. (in review).

Methods

Except for the dates/times below, data is from the Arlington Automated Weather Observing Network station operated by UW Extension.- Precipitation measured on-site from 5/17/2013 18:00 to 11/30/2013 23:00 using an Onset HOBO RG-3 tipping bucket rain gauge mounted on a post off the south edge of the field at site WIBU-9, elevation 1.5 m.- Temperature & relative humidity measured on-site from 5/4/2012 13:00 to 12/31/2013 23:00 using an Onset HOBO Pro v2 temp/RH sensor mounted at an elevation of 3.5 m on the north side of a telephone pole on the west edge of the field.

Methods

Measurements were made using a Licor-brand LAI-2200 device at approximately weekly intervals during the 2012, 2013, and 2014 growing seasons. At each point, the measurements are the average LAI value calculated from 20 subcanopy measurements ranging from in-line with planted rows to the area between rows. In 2012 and 2013, measurements were taken exclusively during diffuse light conditions (sunrise, sunset, or full cloud cover). In 2014, measurements were taken during daylight hours and corrected for light scattering using the LAI-2200C correction described on their website. This means that 2014 data is less accurate than 2012-2013 data. Note that, for each growing season, a date is listed with 0 LAI – this corresponds to the date prior to emergence at each site.After installation of wells and soil monitoring gear in each of the 2012, 2013, and 2014 growing seasons, we used a Topcon GR5 RTK-GPS to collect the location of each relevant point within the field. In 2014, a gridded set of soil sampling points was also measured.

Methods

Each well consisted of a casing and a pressure transducer. Casings were made of 1.5 diameter PVC pipe. Each casing consisted of a screened interval, approximately 0.61 m in length, and continuous casing from the top of the screened interval to the land surface. Wells were installed via hand augering to greater than 1.5 meters below the water table at the time of installation. Where augering was impeded by subsurface rocks and gravel, steel drive point wells were used. Pea gravel was used to backfill the hole to the top of the screened interval, after which soil removed during the well installation process was re-packed to approximately the same density as prior to installation. A 5 cm layer of bentonite was installed at the land surface to prevent preferential flow down the borehole. After installation, wells were pumped to reduce the risk of fine sediment clogging the well screen, and then Onset HOBO U20 water level loggers were installed in each well. Of the 14 total wells, 9 wells were installed during the 2012 growing season (WIBU-1,2,3,5,7,8,9,10,11) and an additional 5 wells installed during the 2013 growing season (WIBU-4,12,13,14,15). Wells within the field were installed following planting and uninstalled prior to harvest for each growing season; the individual installation and start/end dates are evident in the data. Locations of individual wells are available in the Point Locations dataset. Note that some wells move between years due to uninstallation prior to harvest.

Methods

A gridded set of points distributed at appr. 30 m resolution over the Wibu field site were established during the 2014 growing season. At each of these points, a disturbed soil sample from a depth of 5 cm was collected and used to measure the soil organic content. At a randomly selected subset of these points, an additional undisturbed soil core was collected from a depth of 5-7.5 cm and used to measure porosity, bulk density, and organic content (measured by burning for 24 hrs at 440°C). Undisturbed soil cores were then split into 3 subsamples and a Beckman-Coulter LS-230 laser particle size analyzer was used to measure a continuous particle size distribution. The subsample with the median d50 particle size was used to calculate a variety of soil texture metrics, described in the data table below. These soil texture parameters were used as input to the Rosetta pedotransfer function (Schaap et al., 2001) to estimate soil water retention properties. At each of these points, the 2013 interpolated groundwater metrics and the yield data described in the Yield dataset were extracted using ArcMAP 10.2 software. A full description of this methodology is contained in Zipper et al. (in review).

Methods

We developed a database of water quality interventions by government agencies in the Yahara Watershed. Interventions were included if they were a) publicly funded and implemented, b) land-based, c) implemented within the 5-year period 2007 to 2012, and d) aimed to reduce nutrient (phosphorus and nitrogen) and sediment runoff to surface waters as a primary or secondary goal. Our criteria excluded interventions implemented directly in the water. They also excluded work by non-profit watershed groups and for-profit companies.Interventions were categorized by type of conservation tool: regulation or standard; incentive (grant and cost-share programs); direct management (including public management actions and engineered practices); and acquisition (land conserved through fee simple acquisition or conservation easement). Interventions were next categorized by which government level (or multiple levels) of government were involved in rulemaking and implementation (Table 1). We defined the rulemaking level of government as that which created the standard or wrote the law, and the implementing level of government as that which made field-level decisions, negotiated with landowners, and monitored practices. If the intervention was a grant given to private recipients, the implementing agencies were considered those that supervised grant implementation.We mapped policy interventions in ArcGIS (version 10.1). The goal of mapping was to determine the extent and overlap of interventions throughout the watershed and the agency responsible for establishing and implementing the policies. Public acquisitions of conservation land were mapped and categorized by the government level acquiring the parcel or parcels. Incentive programs – grants and cost-share – were mapped for the parcels where the incentive program was applied from 2007-2012. For federal Farm Bill Natural Resources Conservation Service (NRCS) conservation programs, for instance, the farm parcels of the cost-share recipients were mapped with publicly available data or by matching recipient names with parcel ownership records. Regulatory programs were mapped according to each statutes definition. For example, Wisconsins shoreland zoning ordinances were mapped as the area in the 300 meter buffer around rivers or streams and 1000 meter buffer of lakes or ponds, using the Wisconsin DNR water body base layer. Regulations were represented by specific permit area when permit data were available, such as farms with county winter manure-spreading permits.Regional water quality experts validated the interventions list and map. Reviewers included a regional planner, two municipal administrators, a commissioner on the County Lakes and Watershed Commission, a County water conservationist, a lawyer for an environmental non-profit, and the director of a Wisconsin soil and water conservation organization. Through this process we added several interventions and clarified the mapping rules. Analyses were conducted on 35 of 41 interventions that could be represented spatially through publicly available data. The most significant unmapped intervention was nutrient management planning, for which the County office did not have spatial data. We estimated the percent land area covered by each intervention by subwatershed. The Yahara Watershed was divided into 300 subwatersheds based on a recent modeling effort that delineated 200 subwatersheds in the upper Yahara Watershed (Montgomery Assoc., 2011) and our delineation of 100 comparably-sized subwatersheds based on a Digital Elevation Model in the lower Yahara Watershed. We then calculated the percentage of each subwatershed covered by each of the 35 interventions. The percentage of land covered by every intervention within a subwatershed was then summed to get a cumulative percent coverage. This ranged from 0 to a possible 3,500 for each subwatershed. The total percent intervention coverage is shown in heat maps depicting low to high policy coverage by subwatershed, created in ArcGIS.We categorized subwatersheds as urban or rural in order to compare coverage of interventions. Subwatersheds were classified as urban if the developed land cover classes were 50percent or more of land area, based on 2010 National Land Cover Data, which resulted in 83 urban (28percent) and 217 rural (72percent) subwatersheds. We conducted a Welch 2-sample t-test to determine whether cumulative percent area of interventions differed significantly for urban and rural subwatersheds. The untransformed cumulative percent area data were consistent with assumptions of normality and were not improved by an ArcSin transformation (sometimes used with percentage data), so we report the t-test with untransformed data.We compared intervention locations with total phosphorus yields (kilograms phosphorus per hectare per year) for the 200 subwatersheds modeled for the year 2008 with the Soil and Water Assessment Tool (SWAT). The Montgomery and Associates SWAT model is widely used by policymakers in the watershed. The subwatersheds with the highest nutrient yields are consistent with earlier models and measurements conducted for conservation planning (Lathrop, 2007).A Pearsons product-moment correlation matrix compared interventions with modeled phosphorus yield by subwatershed, calculated in R (version 3.0.1). We correlated phosphorus yields with cumulative percent intervention coverage by municipal, county, state, and federal governments in both their rulemaking and implementation capacities. We also compared the correlation of phosphorus yields with intervention coverage for each type of intervention tool. Interventions were also grouped by whether they targeted agricultural or nonagricultural activities. These correlations give a proxy measure of whether public interventions target areas of concern for watershed nutrient reduction.