Assessing a decade of phosphorus management in the Lake Mendota, Wisconsin watershed and scenarios for enhanced phosphorus management

Abstract

A phosphorus (P) budget was estimated for the watershed of Lake Mendota, Wisconsin, to assess the effects of nutrient management on P accumulation in the watershed soils. We estimated how nutrient management programs and legislation have affected the budget by comparing the budget for 2007 to a budget calculated for 1995, prior to implementation of the programs. Since 1995, inputs decreased from 1,310,000 to 853,000 kg P/yr (35% reduction) and accumulation decreased from 575,000 to 279,000 kg P/yr (51% reduction). Changes in P input and accumulation were attributed primarily to enhanced agricultural nutrient management, reduction in dairy cattle feed supplements and an urban P fertilizer ban. Four scenarios were investigated to determine potential impacts of additional nutrient management tactics on the watershed P budget and P loading to Lake Mendota. Elimination of chemical P fertilizer input has the greatest potential to reduce watershed P accumulation and establishment of riparian buffers has the greatest potential to prevent P loading to Lake Mendota.

Appendix

This appendix was created as a record of how the phosphorus budget for the Lake Mendota, WI watershed was derived. It is also a record of the numerous assumptions made in converting data available from various sources to watershed scale P imports and exports.

Converting county values to watershed values

In order to convert values for agricultural components of the budget which were only available to us at the county scale to the watershed scale, we divided the values by the area of agricultural land in the county to normalize to agricultural area. We then multiplied this by the agricultural area within the watershed which includes a small proportion of land in neighboring Columbia County. We assumed that agricultural practices in the Columbia County portion of the watershed were adequately represented by Dane County data.

e.g.

$$ \frac{DaneCty\,\,Cattle}{DaneCty\,AgArea\,}*Watershed\,AgArea = WatershedCattle $$

where “DaneCty AgArea” was 395,369 acres and “Watershed AgArea” was 80,475 acres based on the 2007 USDA NASS Cropland Data Layer (as cited within the main body of the paper). This resulted in a factor of 0.204 to multiply by Dane County values to obtain watershed scale estimates.

These assumptions differed slightly from the Bennett et al. (1999) approach, which assumed that the distribution of livestock and crops in the watershed was proportional to the distribution in the county. Our approach allowed for the distribution of agricultural land to be different within the watershed than for the county as a whole. The net effect of this difference is small. If we followed the Bennett et al. convention our factor above would be 0.220 (686/3113 km², or watershed area divided by county area).

Calculating exports

Dairy products

Milk production for Dane County was obtained from the USDA-NASS Quick Stats website for 2007. This value was scaled to the watershed as described above and converted to P using the weight-weight conversion of 0.0009 units P per unit milk.

Eggs

Our calculated mass of P exported in eggs differed by two orders of magnitude from the Bennett et al. value for 1995 (1999). The number of layers (hens of laying age) in the watershed was available from the USDA Census of Agriculture for 1992, 1997, 2002, and 2007. Of these four years, the 2007 value was the highest. By this metric alone, we would have expected to have higher P exported in eggs in 2007 than 1995, though Dane County data specifically for 1995 are not available through the USDA NASS online database. The number of layers in Dane County was 63,914 in 1992 and 75,052 in 2007.

We calculated egg P export as follows. The number of layers in Dane County was scaled to the watershed as described above. The number of eggs per layer per year was estimated for statewide data because county-specific data is not reported for years after 1992. For years 1981 through 1992 where data are available for both the state and county, there was no statistically significant difference between the mean for state and county numbers, though the year-to-year variance was greater for the county than the state. We used the mean value for the state as the most-likely estimate for eggs per layer per year (278). We bounded this mean value with a minimum, which was the historical low value for Dane County (215) and a maximum, which was the most likely value, plus approximately half of the historic range of the Dane County values in order to account for the increasing trend in egg yield (308). Data shown in Fig. 4.

Fig. 4

Egg yield for hens in Dane County and Wisconsin

To estimate mass of each egg, we bounded the calculation assuming that all eggs were “medium” in the minimum case, large/extra large in the most likely case, and “jumbo” in the maximum case according to the USDA grading standards for table eggs (USDA-AMS 2000). These corresponded to masses of 50, 60, and 71 grams per egg, respectively.

All eggs were assumed to have 0.0021 grams P per gram of egg (Lorimor et al. 2000).

Thus the minimum P exported in eggs was:

Number of layers * minimum yield * minimum egg size * proportion P, or (75,052 * 0.204) layers * 215 eggs per layer per year * 0.05 kg per egg * 0.0021 kg P per kg egg = 346 kg P per year. Likewise, the maximum value was (75,052 * 0.204) layers * 308 eggs per layer per year * 0.071 kg per egg * 0.0021 kg P per kg egg = 703 kg P per year.

Animal export

Cattle

The number of cattle sold in Dane County was reported in the Census of Agriculture for 2007. This is broken down into categories of calves (<500 lbs) and cattle (>500 lbs).

Calves sold for veal can be sold immediately after birth to be raised in a dedicated veal farm or can be raised on location to full slaughter weight. We bracketed the mass of calves sold using these two extremes, with the mean as the most likely case. In the minimum, we assumed all calves were sold (and exported from the watershed) soon after birth at a weight of 100 lbs. In the maximum, we assumed all calves were raised within the watershed to a veal slaughter weight of 300 lbs.

Cattle (>500 lbs) sold are likely a combination of mature Holstein cows, which are being culled from the herd along with beef cows, and dairy steers. Cattle raised for beef have an average slaughter weight of 1140 lbs (Greiner 2002), while mature Holstein cows in Wisconsin have an average weight of 1500 lbs (Hoffman et al. 1992). We used these two values as minimum and maximum values for cattle export with the mean as the most likely.

For the proportion of P in cattle, we used 0.007 (Lorimor et al. 2000).

Hogs and pigs

According to communication with area experts, there are no longer any commercial swine operations within the Lake Mendota watershed (P. Nowak, University of Wisconsin-Madison, personal communication). Therefore, our minimum and most likely estimates for Hogs and Pigs export is set at zero. Our maximum estimate is calculated as if the watershed has a proportional share of Dane County’s swine population.

The 2007 Census of Agriculture reported the total number of hogs and pigs sold from the county. In 1997 and years prior, this was split between feeder pigs (40–80 lbs) and non-feeder pigs. According to the data available, approximately 30% of the pigs sold in years with these data were feeder pigs. Therefore we estimated that 30% of the pigs sold were on average 60 lbs, while the remaining 70% were an average of 250 lbs.

Crop export

Corn, wheat, barley, oats, and rye for grain; soybeans (for beans); and tobacco

Total harvest for the crops above was obtained for Dane County from the 2007 Census of Agriculture. These were converted to watershed values as described above. Harvest was converted from yield units to P mass using conversion factors from Laboski et al. (2006, Table 4.3) for each crop.

Green peas and snap beans

Peas and beans were a minor component of the budget but were included to be consistent with the 1995 budget of Bennett et al. (1999).

County level harvest data for peas has not been reported since 2003. However acres planted in peas has continued to be reported. We used annual yields and acres planted from 1968 to 2003 to develop a linear model for yield (tons/acre) over that time period. Yield increased linearly from 1968 to 2003 (R² = 0.56). Predicted yield for 2003 was 2.45 tons per acre and as this was the last year data were available for Dane County yields. We did not extrapolate beyond 2003 and instead used the 2003 prediction as a conservative yield prediction for 2007. We did not use the actual 2003 yield as it appeared to be an outlier (Fig. 5). Interestingly, there has been a very steady decline in acres planted in peas in Dane County over the 35 year record, from 11,000 acres to approximately 500 acres.

Fig. 5

Yield of Green Peas (tons per acre) for Dane County, WI. Data were not reported after 2003

Similar to green peas, harvest data for snap beans has not been reported for the county level since 2001. Acres planted in beans was reported in the 2007 data, however. We used the linear prediction for 2001 yield as a conservative estimate for 2007 yield as in the case for green peas. For both peas and beans we converted from mass harvested to mass P using conversions found in Laboski et al. (2006); Fig. 6.

Fig. 6

Yield of snap beans (tons per acre) for Dane County, WI. Data were not reported after 2001

Forage crops

Forage crops were assumed to stay within the watershed as these tend not to be cash crops but rather are grown to support livestock on the farm (J.M. Powell, US Dairy Forage Research Center, University of Wisconsin, pers. comm.). Therefore these were not counted as exports from the watershed.

Export to Lake Mendota

Export to Lake Mendota was estimated to be the mean value of the Lake Mendota P load reported in Carpenter and Lathrop (2008). Minimum and maximum values were the 10th and 90th percentiles of the data, respectively.

Calculating imports

Feed supplements

Feed supplements were calculated based on a study of supplements to South Central Wisconsin dairy farms by J.M. Powell et al. (2002). Powell et al. reported the average mass of food consumed per cow per day, the percentage of that feed that is P, and the proportion of that feed that is from forage vs. mineral supplements. In personal communication with J.M. Powell, he suggested that amount of P supplements fed to cattle has decreased steadily over the last decade for reasons outlined in the main text of our paper (J.M. Powell, US Dairy Forage Research Center, University of Wisconsin, pers. comm.). We bounded our estimates of total feed supplements using a maximum value where all cows in the watershed receive supplements, a minimum where only lactating cows receive supplements, and a most likely value that is an average of the two.

Urban fertilizer

Our minimum estimate of urban fertilizer assumed that only lawns low in P were fertilized. To be in compliance with the Dane County P ordinance for lawns (Ord 80), a soil test demonstrating P-deficient soils needs to be conducted in order to legally apply chemical P to lawns. Since the ordinance went into effect, only 200 lawns have been tested in the watershed, and of those only 20% show low P. Thus of the approximately 200,000 lawns in the watershed, only 40 of them (0.02 percent) can legally apply P (D. Soldat, University of Wisconsin Soil Science Extension, personal communication). We therefore used zero as our minimum estimate for urban P fertilizer use. Maximum estimates assumed that 20% of the 200,000 lawns in the watershed were fertilized twice a year which is a baseline recommended rate of application (Soldat and Petrovic 2008). The most likely estimate was an average of the maximum and minimum estimates.

Agricultural fertilizer

The import of chemical P fertilizer is one of the biggest uncertainties of this budget due to the lack of data available publicly at the county or watershed scale.We know that recommended rates of P application have declined and that Nutrient Management Plans in the watershed have made farmers more aware of their P needs and management.

Our estimates of fertilizer imports are based on recommendations for each crop at each of three different soil P test levels from the University of Wisconsin Extension – Nutrient application guidelines for field, vegetable, and fruit crops in Wisconsin (Laboski et al. 2006). On average, Dane county soils fall into the “Excessively High” phosphorus category for which the recommended P application to soils is zero. An exception is reluctantly made in the guidelines for applying starter fertilizer for corn crops even in excessively high phosphorus soils. Our minimum values for P imports thus include zero imports for all crops with the exception of a low rate of starter fertilizer application for corn in the watershed.

Despite the recommendation of zero P fertilizer, we suspected that farmers apply some P fertilizer in hopes that it will provide a “bumper” crop (P. Nowak, University of Wisconsin-Madison, personal communication). Therefore our “most-likely” values for fertilizer application used the recommendations for “high” soil P instead of “excessively high” soil P. In the maximum case, we used recommendations for the “optimal” soil test P category which allows for P to be applied at approximately the same rate as it is removed in crops.

In all cases, the amount of chemical P imported was reduced by the amount of manure P in the watershed available to be spread and credited. For the crediting, we assumed 64% of the manure was able to be collected and that, 60% of the manure P was plant-available and therefore credited.

Manure production in the watershed

The 2007 Census of Agriculture categorized cattle and cows into 3 categories: beef cows, milk cows, and other cattle. The “other cattle” category included heifers (cows which have not yet calved), and all male cattle (beef steers, dairy steers, bulls, and calves). In order to use these data with the Midwest Plan Service manure P production rates, we needed to further divide these categories to match the MPS categories as follows.

Previous years of the agricultural census further divided the “other cattle” category between heifers and male cattle and we used these data to determine the proportion of this category which were heifers (54%) versus male cattle (46%).

Cattle categorized as “milk cows” were further divided between lactating (84%) and dry (16%) cows based upon the ratio found for south central Wisconsin dairy farms by Powell and others (2002). All cows in these two categories were assumed to be full grown, mature Holstein cows (1500 lbs).

Cattle in the “beef cows” category were all assumed to be full grown and were assumed to be the average size for mature beef cows (1139 lbs, Greiner 2002).

Cattle in the heifers category were divided between “young” and “mature” groups using mean data from south central Wisconsin herds (Powell et al, 2002). “Young” heifers are between 0 and 7 months old and account for 42% of heifers. “Mature” heifers are older than 7 months but not yet 24 months and account for 58% of heifers. We used a linear age-weight relationship (Hoffman 2006) and assumed a uniform distribution of ages within each of these two categories to determine the mean size of heifers in each.

Male cattle (beef and dairy) are primarily used for veal and beef production with a very small number kept as bulls. We assumed all cattle in this category were steers and raised until age 13 months at which point they are exported from the watershed. Within this category, we assumed a uniform age distribution such that the mean age was 6.5 months. This was potentially a high estimate as it assumes all males are raised to full size. More likely, because this is a predominately dairy watershed, a significant portion of these would be veal calves and the mean age would be much lower.

Manure production rates (lbs maure/lbs animal), as well as P concentrations in manure, were obtained from Lorimor et al (2000) for heifers, lactacting cows, dry cows, beef cows, and beef steers. We multiplied these rates by the numbers and mean size of cattle in each of these categories to estimate the total amount of manure produced in the watershed.We bounded our estimates for cattle by +/− 30% (Lorimor et al. 2000).

Chicken manure was estimated for layers and broilers using manure P production data from Lorimor et al. (2000).

Hog manure was set to zero for our minimum and most likely values for the same reasons as hog and pig export above. The maximum value assumed hogs in the watershed are proportional to the county as a whole. However we did not have age/size structure data for hogs and pigs. Here we assume the average hog is 150 lbs, half-way to market weight, and used the Lorimor (2000) values for manure P production (Table 4).

Table 4 Estimates of manure production within the watershed

A note on uncertainty in estimates

More uncertainty was associated with P inputs compared to outputs in the watershed. Uncertainties arose when estimating P fertilizer applied to agricultural soil because comprehensive and accurate fertilizer application and manure spreading rates in the watershed are not compiled or available publicly. Uncertainties in estimates of P fertilizer application to urban lawns also existed due to lack of reported fertilizer application rates on residential, commercial and industrial lawns. Estimates of animal feed supplements also introduced uncertainties in terms the quantity of supplements and to which types of cows the farmers are giving supplements to. Because of this lack of concrete data, we were forced to estimate based on recommended rates of fertilizer and feed supplement use combined with personal accounts and publications of scientists who investigate those areas.

Uncertainty in estimates also arises in converting county data to watershed data. This analysis assumes that agricultural practices are distributed uniformly across the agricultural land of the county. However, we have heard anecdotal evidence that the north-east part of the county may have more dairy operations, while the south west half of the county is more dominated by cash-grain operations (R. Lathrop, Wisconsin Department of Natural Resources, personal communication). If this is indeed the case, we may be overestimating agricultural fertilizer use while underestimating feed supplements, manure production, export of milk and cattle, etc.

Fewer uncertainties in outputs existed because production data and P content for crops are readily available in public databases. Uncertainties in outputs arose in estimates of P in livestock export because this is dependent on the size and type of animal exported. We attempted to capture the range of variability by examining exports at plausible extremes of animal sizes.

Nutrient management at the watershed scale would be greatly aided by aggregation of data at that scale rather than at political, county or state boundaries.

References not included in main text

Greiner SP (2002) Beef cattle breeds and biological types. Virginia Cooperative Extension Publication, pp 400–803.

Hoffman PC  (2006)  Strategies to reduce growth variance and improve feed efficiency in dairy heifer management.  Milk Production Reference Library. http://www.MilkProduction.com

Hoffman PC, Funk DA, Syverud TD (1992) Growth rates of Holstein replacement heifers in selected Wisconsin dairy herds. College of Agriculture and Life Sciences Research Report. R551. University of Wisconsin-Madison.