Field sites: soil characteristics, preparation, and layout. The field study was conducted at University of Wisconsin-Madison agricultural research stations at Hancock, Lancaster, and West Madison, Wis., during the summer of 2003. Prior to the study, each field was cropped in alfalfa. A wide-spectrum herbicide (RoundUp Ultra; Monsanto, St. Louis, Mo.) was applied to clear the field in the fall of 2002. The soil characteristics of each field site were determined by analyses at the University of Wisconsin-Madison Soil and Plant Analysis Laboratory and are shown in Table 1. The soil classifications were loamy sand (Hancock), silt loam (Lancaster), and silty clay loam (West Madison). Each field site consisted of a 4 × 11 grid of treatment plots that were 3.05 by 3.05 m (a total of 44 plots) with a 4.6-m alley on each side of each plot. In the spring of 2003, white clover was planted in the alleys to control weed growth and to minimize lateral soil transfer during rainstorms. Each treatment was randomly assigned to two plots at each site. The variables considered in treatments were the date of manure application; the time of planting; whether manure was immediately tilled into the soil, tilled into the soil immediately prior to planting, or not tilled into the soil; and whether the manure was amended with chopped oat straw. Although manure is typically tilled into the soil within a few days after application (and in some cases prompt tilling is required by environmental regulations), manure is sometimes left on top of the soil for lengthy periods before tilling, due to weather conditions or more pressing work demands. Amendment with straw is not a common intentional practice, but various amounts of straw may be present in bovine manure if straw is used as bedding. The purpose of adding straw in the present study was to possibly stimulate predatory protozoan activity. Descriptions of key treatments for each site are shown in Table 2.
Bovine manure application. Fresh manure (≤3 days old) was collected at the University of Wisconsin-Madison Dairy Cattle Instruction and Research Center from a herd of lactating Holstein cows fed a standard corn- and soy-based diet. In order to maximize the levels of indigenous E. coli, care was taken to obtain manure with low levels of bedding material. Manure was shoveled into 19.4-liter (5-gallon) buckets (approximately 16 kg per bucket), and the buckets were transported within 1.5 h to the field sites. The manure was applied at a rate of five buckets (80 kg) per treatment plot to simulate a typical manure application rate in Wisconsin, 67.2 metric tons (wet weight) per ha. To apply the manure, the buckets were emptied onto the treatment plot soil, and the manure was distributed evenly over the soil surface with a shovel. Some treatments included application of 0.5 bale (ca. 7 kg) of chopped oat straw to the treatment plot before manure addition. For these treatments, the straw was manually distributed over the soil surface, and this was followed by manure application and spreading. The plots treated with manure or with manure and straw were tilled with a shovel either immediately after application, immediately before planting, or not at all. Tilling with a shovel incorporated the manure to a depth of about 15 cm.
Vegetable production. Each treatment plot (other than controls) was manually planted with two side-by-side rows each of carrot (cultivar Short ‘n’ Sweet; W. Atlee Burpee & Co., Warminster, Pa.), radish (cultivar Cherry Belle; Burpee), and lettuce (cultivar Simpson Elite; Veseys Seeds, Ltd., Charlottetown, Prince Edward Island, Canada). There were two rows of each vegetable for a total of six rows; each row was about 2.7 m long and 46 cm from each neighboring row. At each site the rows were oriented north-south. The rates of planting were 2.0, 4.0, and 1.2 g of seeds per row for carrot, radish, and lettuce, respectively. Planting was done in May and June (Table 3). The vegetable plots on loamy sand were irrigated twice weekly from June through mid-September at a rate of 1.27 cm per irrigation. The total additional water was 40 cm, and the rainfall data and the daily high and low temperatures are shown in Tables 4 and 5, respectively. No irrigation was available for the silt loam, and vegetable growth ceased during a prolonged drought beginning in July and continuing for the remainder of the summer (Table 4). The silty clay loam was irrigated as needed, and the irrigation data and weather data are shown in Tables 4 and 5.
Soil sampling and analysis. Soil samples from each treatment plot were analyzed at biweekly intervals, beginning at the time of manure application. A composite of three soil core samples was collected from each treatment plot by using one sterile aluminum soil corer (Forestry Suppliers, Jackson, Miss.), one sterile tongue depressor to push the soil out of the corer, and one sterile stomacher filter bag (Nasco, Ft. Atkinson, Wis.). To determine the sampling location within each plot, the plot was divided into quadrants, a quadrant was randomly selected, and three samples were taken from randomly chosen locations within the quadrant. Soil samples were placed in shaded insulated coolers, transported to the laboratory, and refrigerated (5°C) until analysis. Each soil sample was weighed, and then 198 ml of Butterfield's phosphate diluent (BPD) (Nelson Jameson, Marshfield, Wis.) was added. The diluted sample was then manually shaken for 30 s, allowed to sit for 30 s, and then manually shaken for another 30 s. Decimal serial dilutions in BPD were prepared, and 1.0 ml of a dilution was plated on a 3M Petrifilm E. coli-coliform count plate (3M Microbiology Products, St. Paul, Minn.). The initial sample dilution (in the stomacher bag) was not plated because the soil color obscured colonies on the plate. The plates were incubated at 35°C for 48 h, and then presumptive E. coli colonies (blue with associated gas) were counted. By using the soil sample weight, dilution factor, and number of presumptive colonies, the log number of CFU per gram of soil was calculated for each treatment plot, and then the mean log number of CFU per gram for each treatment was calculated from values for the log number of CFU per gram for the duplicate treatment plots. When no colonies were detected for the least diluted sample plated, a log number of CFU per gram was calculated from an arbitrarily assigned value (0.5 colony).
When a soil sample yielded no presumptive E. coli colonies, an enrichment procedure was used for that treatment plot at the next sampling time. In this procedure, the initial dilution and mixing were performed by using nutrient broth (Difco, Becton Dickinson, Sparks, Md.) instead of BPD. For nonselective enrichment, the initial dilution was incubated for 24 h at 35°C. Then the initial dilution was gently mixed, and 1.0 ml was transferred to 9 ml of lauryl tryptose broth (Difco), which was vortex mixed and incubated at 45.5°C for 24 h as a selective enrichment step. Next, the contents of the selective enrichment tube were vortexed, and one loopful was streaked onto Levine eosin methylene blue agar (LEMB) (Difco). Each LEMB plate was incubated for 24 h at 35°C and observed for colonies with dark centers, with or without an associated metallic sheen. A positive result was recorded if one or more such colonies were present. For confirmation, presumptive colonies from Petrifilm E. coli-coliform count and LEMB plates were streaked onto brain heart infusion agar (Difco) and incubated 24 h at 35°C, and then an isolated colony was tested to determine its cell morphology, Gram reaction, oxidase reaction, and biochemical characteristics (API 20E kit; BioMerieux, Hazelwood, Mo.). For the first 3 months of the study, one colony from a given treatment at a given sampling time (e.g., one colony each from treatments 1, 2, 3, and 4 at the 11 April sampling) was tested to confirm the identity. After this, one colony was tested for each combination of manure application date and tilling (e.g., treatments 1 and 3 at an August sampling date). Colonies from treatments that received manure and straw were not tested after the first 3 months. Over the course of the study, 85% (271 of 319) of presumptive E. coli isolates were conclusively confirmed to be E. coli, while 8.5% were identified as doubtful E. coli; the remaining isolates were identified as various coliforms (Klebsiella, Citrobacter, Enterobacter), Kluyvera, Pasteurella, and Serratia. Because presumptive counts would always be greater than or equal to confirmed counts (thus overestimating E. coli survival), additional colonies were not tested when a colony was not confirmed to be E. coli.
Vegetable sampling and analysis. Vegetable samples were obtained at thinning and harvest. Vegetables were randomly selected from the plants thinned or harvested, and the edible portions (roots for radish and carrot and leaves for lettuce) were separated by using scissors or shears previously sprayed with 70% ethanol, placed in clean plastic sealable bags, and transported to the laboratory in insulated coolers. In the laboratory, a 25-g random subsample for a given treatment plot was weighed on cheesecloth. The cheesecloth and vegetables were placed into a colander that had been previously sprayed with 70% ethanol, and the vegetables were washed with cool running tap water (3.3 liters per min) for 30 s. After draining for 1 min, the vegetables were aseptically transferred to a sterile filter whirlpack bag for stomaching. Vegetables that were too large for stomaching were first cut into smaller pieces with a knife that had previously been sprayed with 70% ethanol. To the vegetables, 99 ml of nutrient broth was added, and the mixture was stomached on medium speed for 30 s. The samples were then analyzed by direct plating and enrichment by using the methods described above. The confirmation rate for presumptive E. coli isolates was 64% (145 of 225 isolates were confirmed to be E. coli); 20% of the isolates were identified as doubtful E. coli, 5% each were identified as Enterobacter and Klebsiella, and the remaining isolates were identified as Kluyvera, Citrobacter, Serratia, and Kleviscia.
Presentation of data and statistical analysis. All data were organized by treatment and are expressed below in terms of the number of days following manure application. Treatments were empirically compared in terms of first plot enrichment (the number of days after manure application when enrichment was first done for at least one plot [i.e., the preceding sample from that plot yielded no presumptive E. coli colonies as determined by direct plating]), second plot enrichment, and first negative enrichment (the number of days after manure application when a plot for a given treatment first yielded a negative result following enrichment). Analysis of variance (Minitab, release 12; Minitab, Inc., State College, Pa.) with a 5% significance level was used to compare soils in terms of mean first plot enrichment time, second plot enrichment time, and time until the first enrichment-negative result for soil and each vegetable. Similar analyses were done to compare treatments in which manure was tilled into the soil immediately to treatments with delayed tilling and treatments with no tilling. When no enrichment-negative result was obtained, 14 days was added to the time for the final enrichment-positive result, and this value was used in the statistical analyses.