J. Dairy Sci. 85:1218–1226

American Dairy Science Association, 2002.
Dairy Herd Management Practices that Impact
Nitrogen Utilization Efficiency1
J. S. Jonker,*,2 R. A. Kohn,* and J. High
Department of Animal and Avian Sciences
University of Maryland, College Park, 20742
Lancaster Dairy Herd Improvement Association
Manheim, Pennsylvania, PA 17545

ABSTRACT

Improving the efficiency of feed N utilization by dairy cattle is the most effective means to reduce nutrient losses from dairy farms. The objectives of this study were to quantify the impact of different management strategies on the efficiency of feedNutilization for dairy farms in the Chesapeake Bay Drainage Basin. A confi- dential mail survey was completed in December 1998 by 454 dairy farmers in PA, MD, VA, WV, and DE. Nitrogen intake, urinary and fecal N, and efficiency of feed N utilization was estimated from survey data and milk analysis for each herd. Average efficiency of feed Nutilization for milk production by lactating dairy cows (N in milk/N in feed ? 100) was 28.4% (SD = 3.9). On average, farmers fed 6.6% more N than recommended by the National Research Council, resulting in a 16% increase in urinary N and a 2.7% increase in fecal N. Use of monthly milk yield and component testing, administration of bovine somatotropin (bST), and extending photoperiod with artificial light each increased efficiency of feed N utilization by 4.2 to 6.9%, while use of a complete feed decreased efficiency by 5.6%. Increased frequency of ration balancing and more frequent forage nutrient testing were associated with higher milk production, but not increased N utilization efficiency. Feeding protein closer to recommendations and increasing production per cow both contributed to improving efficiency of feed N utilization. (Key words: nitrogen pollution, milk urea nitrogen, dairy cattle protein requirements) Abbreviation key: MUN = milk urea N, 3? milking = three-times daily milking.

Introduction

Nitrogen losses from agriculture to water resources present a major environmental challenge for the Chesapeake Bay Drainage Basin (Thomann et al., 1994). Dairy farming is a large agricultural enterprise in the region, making dairy farms a major contributor to the nonpoint N loading of the bay. Kohn et al. (1997) used a simple mathematical model to evaluate which management practices had the greatest impact on reducing N losses from the farm: dairy herd feeding and management, soil and crop management, or manure storage and handling. This model suggests that improving herd management is the most effective means to reduce nutrient losses to the environment. Improving herd nutrient utilization efficiency by 50% was calculated to reduce nitrogen losses to water by up to 40%, but improving manure utilization efficiency by 100% only reduced N losses to water by 10 to 14%. Other authors determined the effect of several management practices, such as animal grouping (St-Pierre and Thraen, 2001), use of bST, milking three times daily (3? milking) or artificial lighting (Dunlap et al., 2000), on nutrient utilization efficiency and nutrient excretion in research dairy herds. However, the variation in herd nutrient utilization efficiency on commercial dairy farms is not known. Jonker et al. (1998) developed and evaluated a model to estimate N excretion, N intake, and N utilization efficiency for lactating dairy cows. The model requires knowledge of milk production per cow, milk protein percentage, and milk urea N (MUN). The first objective of this study was to determine current N utilization efficiency of dairy herds in the Chesapeake Bay Drainage Basin. The second objective was to identify factors that contribute to variation from herd to herd in nitrogen utilization efficiency. With a better understanding of current management practices and their effect on potential N loading to the environment, opportunities to improve overall management may be identified.

Materials and Methods

A confidential mail survey was conducted in December 1998 with the Maryland and Virginia Milk Produc ers Cooperative (West Reston, VA). An introductory letter was mailed 1 wk prior to the survey, and a reminder letter was sent 1 wk after the survey. The cooperative had 1156 members located throughout most of the Chesapeake Bay Drainage Basin, including Delaware (n = 23), Maryland (n = 432), Pennsylvania (n = 519), Virginia (n = 172), and West Virginia (n = 18). Participants were offered monthly bulk tank milk analysis of MUN for 6 mo as an incentive to return the survey. The survey included information on dairy herd characteristics, milk production, crop production, feed inputs, management characteristics, and MUN knowledge and use. Herd characteristics included information regarding breed, number and distribution by parity of milking animals, and number and age distribution of replacement heifers. Milk production included volume and compositional data. Crop production and management included types and acreage of crops grown and use of a nutrient management plan (NMP). Feed inputs included types of feeds routinely fed and frequency of ration balancing and nutrient composition testing. Management characteristics indicated the use of various technologies (bST, increased milking frequency, etc.).

MUN Sampling and Analysis

Bulk tank MUN analyses were performed monthly for 6mo for dairy farms from the Maryland and Virginia Milk Producers Cooperative (n = 1156) from December 1998 through May 1999. Only the December samples are used in the current paper because other results were affected by our correspondence with farmers. Milk samples were collected weekly from Environmental Systems Services (College Park, MD) after routine milk component analyses were performed for cooperative members. One sample for MUN analysis was analyzed per herd per month. The fresh milk samples were treated with an antimicrobial preservative (Broad Spectrum Microtabs II, D & F Control Systems, San Ramon, CA). The milk samples were then shipped to Lancaster DHIA (Manheim, PA) for MUN analysis using the Bently Chemspec autoanalyzer (Chaska, MN).

Modeling and Data Analysis

The mean and standard deviation in N feeding parameters were calculated based on model predictions (Table 1) from the survey data and December milk analysis. Nitrogen intake, urinary and fecal N, and N utilization efficiency were determined for each herd using the model of Jonker et al. (1998), except prediction of urinary N was equal to 0.0259 times body weight times MUN, as recommended by Kauffman and St-Pierre (2001). Average BW for the cows in each herd was predicted as the weighted mean for all cows, where each cow’s body weight was assigned according to breed as follows: Holstein and Brown Swiss, 600 kg; Ayrshire and Guernsey, 500 kg; and Jersey, Milking Shorthorn, and Dutch Belted, 400 kg. Estimates of body weights were made based on DHIA data summarized by Dunlap et al. (2000). Crossbred animals were assumed to weigh the average of the breeds crossed. Crude protein requirements were determined using the National Research Council (NRC, 1989) recommendations for dairy cattle, assuming a one-group TMR was fed (Jonker et al., 1999). The protein required was assumed to be that needed by the 83rd percentile cow with respect to protein requirements for the entire milking herd (Stallings and McGilliard, 1983). Excess N feeding was determined as the difference between observed N intake, estimated using the model described previously (Table 1), and that predicted to be required. Thus, negative values represented underfeeding and decreased the average estimate of overfeeding. The accuracy ofNfeeding was calculated by taking the absolute value of observed minus required N, so that the average represents both overfeeding and underfeeding. Statistical analyses were performed using the software package JMP (SAS, 1995). Herds were excluded from the analysis whenever incomplete survey answers resulted in missing data for a variable in the model to predict N intake or utilization efficiency. Treatment means for each observed or calculated value were compared using ANOVA for discrete variables or regression for continuous variables. Observations were excluded when either X or Y values were missing (i.e., missing data were not estimated by the model). When more than two discrete variables were compared (e.g., frequency of diet formulation), the Tukey-Kramer t-test was used to compare each pair. The environmental and economic impact of overfeeding dairy herds was estimated based on summarized results. Excess N fed in the watershed was calculated by multiplying the number of cows in the watershed during the study (n = 758,347 [United States Department of Agriculture, 1998]) by the fraction of farms overfeeding N and the average excess N per overfed cow. The N losses to water resources that result from overfeeding were calculated by accounting for losses from manure during storage and application. The total excess N fed was assumed to be excreted into manure. Assuming that 25% of manure N excreted eventually becomes available to crops, excess feed N was multiplied by 0.75 to estimate manure N losses (Kohn et al., 1997). There would be additional losses of N from the production of crops. However, imported soybean meal was assumed to provide the excess feed N, so the N losses would not have occurred in this watershed. The cost of feeding excess N was estimated assuming that soybean meal (44%) could be replaced by corn grain to decrease N content. The 5-yr average prices (1996 to 2000) for soybean meal ($0.210/kg) and corn grain ($0.097/kg) were used (Bridge Information Systems, Inc., 2000).

JONKER ET AL.

Table 1. Prediction equations.

RESULTS AND DISCUSSION

A total of 472 dairy farmers responded to the survey, for a 40.8% rate of return. However, nine farms stopped shipping milk shortly after completing the survey and were therefore excluded, and 91 farms were excluded because of incomplete data (usually rolling herd average or milk protein percentage was missing). For the final data, the largest number of surveys were from Pennsylvania (n = 165), followed by Maryland (n = 139), Virginia (n = 56), West Virginia (n = 6), and Delaware (n = 6). A large range in farm size and production was represented in the survey (Table 2). Average FCM was 28.3 kg/d per cow (SD = 4.2) with 3.74% (SD = 0.24) fat and 3.25% (SD = 0.15) true protein. The average farm surveyed had 109 cows (93 milking and 16 dry cows)and 86 replacement heifers (Table 2). Several farms reported not raising any replacements. Nearly every farm (>98%) reported having Holstein cows. Jersey cattle were the second most predominant breed—reported on 11.7% of the farms—and made up 3.7% of all dairy cattle. Other breeds represented less than 1% of total dairy cattle.

The farmers participating in the program generally appeared to represent the range of farmers in the Chesapeake Bay Drainage Basin. The average milk production reported by participants was 28.3 kg/d per cow, compared to the average of 29 kg/d per cow reported for Lancaster DHIA members between July 1996 and April 1998 (Dunlap et al., 2000). Herd distribution (Table 2) was also similar to results reported for those records. The mean MUN for participating farms was 12.8 mg/dl (Table 3), compared to 12.4 mg/dl for all farms in the cooperative (Jonker et al., 2002). Higher MUN may have resulted from a tendency of participants to have higher milk production or to feed higher CP diets than nonparticipants, but the error imposed by nonrandom participation of farmers compared to all cooperative members would be 3.2% of MUN.

The model of Jonker et al. (1998) enables calculation of the variance in N utilization efficiency for a large number of herds in the field. The mean and standard deviation in N utilization parameters for lactating cows across all herds are given in Table 3. These calculations do not include dry cows or heifers, which would otherwise add to excreted N and decrease N utilization effi- ciency for the herd. Observed parameters differed significantly from recommended levels for all parameters. Observed MUN was 12.7 mg/dl, but feeding according to NRC (1989) and allowing for variation within the herd by feeding the 83rd percentile cow would have resulted in a MUN of 11.0 mg/dl.

As with any measurement, the variance can be attributed to both errors in measurement and true variance within the population. The root mean square prediction error (RMSPE) for the model used in this analysis was 16.9% of mean urinary N prediction (Jonker et al., 1998). A similar prediction error for the current study would result in prediction error accounting for 40% (100 ? RMSPE2/SD2) of the total variance (SD2) among farms reported in Table 3. The RMSPE for prediction of N utilization efficiency was 11% of prediction (Jonker et al., 1998). A similar prediction error in the present study would explain 64% of total variance in utilization efficiency among farms. Most of the model prediction error used for these calculations was associated with lab and cow variation (Jonker et al., 1998), and these would be reduced by using a single lab that uses wet chemistry and bulk tank samples representing an average of 109 cows. Therefore, we do not have adequate data to accurately estimate model prediction error under the circumstances in which the model was used in the present study. Nonetheless, the model prediction errors reported previously provide an upper limit of model prediction error.

Table 2. Milk production and distribution of cows from surveyed

farms.

Source: Feedstuffs Magazine
Author: Michael Howie