Actual versus environmentally recommended fertilizer application rates: Implications for water quality and policy

Agriculture, Ecosystems and Environment 240 (2017) 109–120

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment journal homepage: www.elsevier.com/locate/agee

Actual versus environmentally recommended fertilizer application rates: Implications for water quality and policy Jennifer E. Lesliea , Alfons Weersinka,* , Wanhong Yangb , Glenn Foxa a b

Dept of Food, Agricultural and Resource Economics, University of Guelph, Canada Dept of Geography, University of Guelph, Canada

A R T I C L E I N F O

Article history: Received 25 November 2016 Received in revised form 7 February 2017 Accepted 9 February 2017 Available online xxx Keywords: Fertilizer application rate Nitrogen Phosphorus Excess nutrient application

A B S T R A C T

Excessive application of crop nutrients has been identified as a threat to surface water quality in many jurisdictions. The Western Basin of Lake Erie Collaborative Agreement commits the governments of Michigan, Ohio and Ontario to reduce phosphorus entering the Lake Erie’s western basin by 40% by 2025/ 2026 from 2008 levels by, among other things, reducing fertilizer use in agriculture. The International Joint Commission (2014) estimates that agriculture accounts for 44% of total phosphorus loadings to Lake Erie. Our study uses a unique micro level data set of 16 farms over 6 years that allows us to examine 397 individual nutrient application choices at the field and farm level. If efforts to reduce excessive application of nutrients are to be successful at aggregate level, they need to be informed by an understanding of how farmers make nutrient application decisions within existing production systems. The study aims to enhance our understanding by determining whether farmers applying nutrients to maximize yields, to maximize net returns or to meet environmental targets, and whether overapplication depends on factors such as farm size, crop type, manure use, and type of nutrient. We compare actual nutrient application rates with site specific rates intended to minimize excess nutrient application and we regress nutrient application levels against potential explanatory variables including farm size, crop rotation practices, and application of livestock. The data were collected from farmers in the Gully Creek watershed in Ontario. We found that excess nutrient applications, as a percentage of the total nutrient applications, are much higher for phosphorus than for nitrogen and higher for wheat than for corn. While most of the farmers in the study are not required to comply with provincial nutrient management regulations, many of them apply fertilizer at rates close to that recommended by those regulations and some at rates significantly less. While most farms are applying fertilizer, particularly nitrogen, at rates close to the minimum crop requirements, nevertheless, a few farms apply much more phosphorous than recommended. Policy and research efforts should be directed toward targeting these individuals that appear to be the primary contributor to the nutrient loading issue. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Globally, agriculture accounts for approximately 70% of global fresh water withdraws (Calzadilla and Rehdanza, 2010; ShtullTrauring et al., 2016). The impacts of agriculture on water has become the focus of study due to the high volume of use, changes in weather patterns and the associated environmental impacts (Parris, 2011; Shtull-Trauring et al., 2016). The impact of excess application of agricultural nutrients on water quality in the Great Lakes region has become an area of concern for federal and state/

* Corresponding author. E-mail addresses: [email protected] (J.E. Leslie), [email protected] (A. Weersink), [email protected] (W. Yang), [email protected] (G. Fox). http://dx.doi.org/10.1016/j.agee.2017.02.009 0167-8809/© 2017 Elsevier B.V. All rights reserved.

provincial governments surrounding these waters (Dupont, 2010). The Great Lakes region has been subjected to various environmental regulation and incentive programs focusing on water quality since the algae blooms in the late 1960s and early 1970s. The initial policy efforts focused on identifiable point sources of nutrient pollutants, specifically municipal sewage treatment and industrial sources (International Joint Commission, 2014). These programs, along with the development of profitable conservation tillage technology, appeared to be addressing the issue of pollution in the Great Lakes region until the algae blooms began to reappear in the mid 2000s (International Joint Commission, 2014). The International Joint Commission (2014) estimates that agriculture accounts for 44% of total phosphorus loadings to Lake Erie. Commercial field crop farming requires the application of

110

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

fertilizers to provide adequate nutrients for profitable crop production (Barker and Pilbeam, 2006; Blanco and Lal, 2008; Karlen et al., 2004). However, the application of fertilizer in excess of the physiological requirements of the crops can adversely affect crop production. Bernstein et al. (2005) found that excess application of nitrate (NO
3 ), the primary form of nitrogen in commercial fertilizer that is readily available for plant uptake, adversely affected quality, growth and production of crops. Research has also found that the form of the applied nitrogen
fertilizer (organic nitrogen (NHþ 4 ) or mineral nitrogen (NO3 )) can decrease crop quality as well as lead to environmental externalities (Elmann et al., 2016; Engels et al., 2012; Liu et al., 2014). The International Joint Commission (2014) has concluded that dissolved reactive phosphorus is the growth limiting nutrient in the case of algal blooms in the Great Lakes as nitrogen has been found to be consistently available in sufficient concentrations in the water body. The Western Basin of Lake Erie Collaborative Agreement, signed by Michigan, Ohio, and Ontario, commits these governments to reduce phosphorus entering the Lake Erie’s western basin by 40% by 2025/2026 from 2008 levels by, among other things, reducing fertilizer use in agriculture (Government of Ontario, 2015). Reducing nutrient loadings requires an understanding of the rates at which farmers apply fertilizer. Previous research indicates that farmers use numerous sources of information to inform fertilizer application decisions, including fertilizer dealers, seed company agronomists as well as personal experience (Stuart et al., 2014; Osmond et al., 2015). Information and recommendations originating from university scientists and government extension agents was found to be distrusted by farmers (Stuart et al., 2014; Osmond et al., 2015). In addition to the sources of information, the objectives behind the decision can also vary from maximization of profit to minimization of excess loadings while meeting crop requirements. The latter can be estimated through a program like the Nutrient Management (NMAN) Software Program developed to assist Ontario farmers in managing nutrients at the farm and field scale, specifically targeting the nutrient capacity of soils (OMAFRA, 2015; Walters et al., 2013). The purpose of this paper is to compare farmers’ actual application rates of nitrogen and phosphorus to the recommended NMAN rates as well as to rates suggested by other agronomic and economic decision criteria. The paper begins with a description of the unique data set used in the study. It includes data on 396 actual application rates on individual fields from 16 farmers in a Great Lakes watershed in Ontario along with the rates recommended by NMAN to limit the application of nutrients to the rate of crop removal. We then derive alternative fertilizer application rates based on objectives such as maximization of yield and gross margin. We then compare yields and gross margins across the various rates and examine the effects of farm attributes on nutrient application decisions. If a farmer’s actual application rates of fertilizer are greater than the recommended NMAN rates, this could contribute to excess nutrient loadings in surface water. Evidence of a departure from the recommended NMAN rate could be attributed to the farmer using alternative decision making criteria, such as yield maximization or gross margin maximization. Understanding why farmers might apply more than the recommended NMAN rates can help in the design of policies to achieve the desired reduction in phosphorus loadings under the Western Basin of Lake Erie Collaborative Agreement. In contrast, if farmers’ actual application rates of fertilizer are less than or equal to the recommended NMAN rates, then further investigation as to the source of the increased nutrient loadings in the Great Lakes region is required to direct future water quality policy.

2. Material and methods 2.1. Data Data were obtained from a farm survey conducted in the Gully Creek watershed in Ontario. Gully Creek is a 14.3 km2 (1430 ha) sub-watershed of the Bayfield North watershed in Huron County that drains into Lake Huron. Lake Huron flows into Lake St. Clair which empties into the Western Basin of Lake Erie. Approximately 70% (994 ha) of the watershed’s total area is allocated to agriculture with corn, soybeans and winter wheat the primary field crops (Oginskyy, 2014). The soils are largely Podsolic and Luvisolic clay loams (Oginskyy, 2014). Under the Ontario Ministry of Agriculture, Food and Rural Affairs’ (OMAFRA) Watershed Based BMP Evaluation (WBBE) program, the Ausable Bayfield Conservation Authority (ABCA) conducted a land management survey in March 2011. Information was collected on soil nutrient levels, seeding dates and rates, tillage practices, organic and inorganic fertilizer decisions (rate, date, method) and crop yields. Farmers were asked to report historical land management practices and yields for 2008, 2009 and 2010, and to project expected land management practices for 2011, 2012 and 2013 (Simmons et al., 2013). The historical and projected survey data were verified with field observations from the OMAFRA Agricultural Resource Inventory (AgRI) (Simmons et al., 2013). The projected survey data (2011–2013) were further verified by windshield surveys conducted by the ABCA, confirming that the projected crop type matched the actual crop in the field. Surveys were completed by 16 farmers, who farmed 643.5 ha in the watershed. The largest farm cropped 208 ha in the watershed and the smallest farm in the survey cropped 2.11 ha (see Table 1). More than half of the farms in the survey applied some livestock manure, specifically poultry manure. Most farmers in the Gully Creek watershed follow a three-year corn-soybean-wheat rotation. The annual area planted for each crop on the surveyed farms is illustrated in Fig. 1. The share among the three major crops is split approximately equally with some annual variations depending on prices. For example, soybean area rose to over 300 ha in 2010 while winter wheat area fell to less than half this amount due to relative increase in soybean prices. Other crops, including edible beans and hay, represent approximately 5% of the planted area in the watershed. Three-quarters of the surveyed farms use at least one beneficial management practice (BMP) (see Table 1). In Ontario, many crop farmers use a combination of tillage systems in which corn is tilled conventionally and conservation tillage used on soybeans and winter wheat. Some form of commercial chemical fertilizer was applied on all fields. Manure was applied to 57 of the 396 individual field observations across the six years of data. Combining the chemical fertilizer application rate and the estimated available nutrients from manure, the actual annual fertilizer application rate was calculated for each individual field. Testing manure for nutrient content was not part of the data collection for this study. We assumed that the nutrient content of poultry manure is 7.1 kg of nitrogen per MT and 6.3 kg of phosphorus per MT in calculating the actual rate, (NA), which we compared to rates derived from alternative decision models. 2.2. Nutrient rate decision rules We compared the actual fertilizer application rate (NA) obtained from the farmer survey to rates derived under three decision rules: (1) the voluntary adoption of the NMA rate (NMAN rate), (2) maximizing yield and (3) maximizing gross margin. The basis for the decision rules behind these rates is discussed below.

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

111

Table 1 Farm Descriptive Statistics. Farm

Number of Fields

Farm Area (ha)

Crop Rotation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

3 2 1 5 13 3 9 3 1 3 2 5 2 2 5 7

24.12 41.91 11.46 85.45 207.66 11.92 57.80 40.31 19.85 6.15 17.41 10.44 2.11 31.11 25.15 28.83

C/S/WW C/S/WW C/S C/S/WW C/S/WW C/S/WW C/S/WW/O C/S/WW C/S/WW C/S/WW C/S/WW C/S/O C/S/WW C/S/WW C/S/WW O

Total Average

66 4.125

643.05 40.19

a

Manure Applicationb

BMPc

Yes Yes No Yes Yes Yes No Yes Yes Yes No No No Yes Yes No

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No No Yes No

a Crops in the rotation are corn (C), soybeans (S), winter wheat (WW) and other (O), which includes edible beans, forages, barely, hay and land temporarily removed from production. The order of crops indicates the ordering in the rotation. b Yes indicates manure was applied to at least one of the fields within the farm over the time period from 2008 to 2010. c Yes indicates a Beneficial Management Practice (BMP), such as conservation tillage, soil testing, shelter belts, fragile land retirement, buffer strips and berms were used on the farm through the period of observations.

350

Area (hectares)

300 250 200 150 100 50 0

2008

2009

2010

2011

2012

2013

Year Corn

Winter Wheat

Soybeans

Other

Fig. 1. Annual Area Allocated to Crop Production in Gully Creek Watershed, 2008–2013.

2.2.1. NMAN rate (NE) The rate that considers the environmental implications of fertilizer application is denoted as NE. In Ontario, NE is assumed to be the recommended NMAN rate as it aims to decrease excess nutrient application, which is calculated using the NMAN software package, designed and maintained by OMAFRA (McKague, 2016). Field specific management plans from the NMAN software produce recommended rates for nitrogen and phosphorus fertilizer required to attain a specified yield. Factors such as soil type, soil fertility tests, crops planted in the previous year, including the use of a cover crop, and tillage practices are considered in the NMAN calculation. The NMAN software uses a mass balance system to account for nutrient inputs (such as manure and inorganic fertilizers) and nutrient losses (such as plant uptake, runoff and infiltration) (OMAFRA, 2015; Walters et al., 2013). Nutrient management regulations based on the NMAN rate currently only

apply to livestock operations with 300 or more Nutrient Units1 (NU), expanding or new livestock facilities, or farms located within 100 m of a municipal well. NMAN is currently voluntary for farmers who do not meet the aforementioned criteria (Walters et al., 2013). None of the farmers surveyed were required to use NMAN due to the limited size of the farms with livestock operations. The NMAN rate minimizes excess application levels such that the nutrient requirement required to attain the target yield are met

1 One nutrient unit is defined as the number of livestock that produce manure, a fertilizer replacement value, of either 43 kg of nitrogen or 55 kg of phosphate, which ever one is lower, on an annual basis. For example, a dairy herd of 70 cows, including heifers and calves, produces 127 NU per year. For multiple livestock types on a farm, multiple calculations of NU are required. The factors required for these calculations are available from OMAFRA and are an average of the nutrient content of the manure

112

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

but not exceeded. We interpret the NMAN rate as the environmental rate (NE) as it is designed to minimize excess application of nutrients, decreasing the risk of nutrient flows off farm and into the surrounding aquatic environment. If the actual application rate (NA) is greater than the environmental rate (NE), the excess application of fertilizer may lead to nutrients entering the surrounding environment. 2.2.2. Yield maximizing rate (NY) The yield maximizing rate (NY) is defined as the nutrient application rate that allows the crop to reach its maximum physical output per hectare in that cropping year. Determining the fertilizer rate that maximizes yield involves first specifying how application rates (N) influence crop yield (Y) expressed through a yield response function such as Y = f (N)

(1)

where Y is crop output per hectare per year, N is the quantity of nutrient applied per hectare per year, and f(N) is the total product curve. The marginal physical product of a nutrient can be found by taking the first derivative of the total product or yield response function given in equation (1). Setting this derivative to zero and solving for N, will give the level of N application that maximizes yield per hectare, NY where

ðf ðNYmax Þ ¼0 ðN

ð2Þ

This yield maximizing rate (NY) will differ for nitrogen and phosphorus and depends on empirically determined yield response functions specific to the nutrient and crop type. 2.2.3. Gross margin maximizing rate (Np) The gross margin per land unit (p) is given by

p ¼ Py  Y
PN  N

ð3Þ

s:t: N > 0 and Y ¼ f ðNÞ

ð4Þ

The assumed winter wheat yield response function from OMAFRA (2015) is YW = 54.99 + 0.5963  N
0.0019  N2

where YW is annual wheat yield, measured in bu/ac, and N is the annual nitrogen application rate, measured in lbs/ac. These values were converted into metric units for comparison in the results section. Soybeans typically do not require the application of nitrogen fertilizer to meet yield expectations due to the nitrogen fixation properties of the crop. Yield response functions for phosphorus were unavailable for the crops and within the region of the case study. Therefore, we were unable to calculate the yield maximizing and gross margin rate for phosphorus fertilizer in winter wheat, corn and soybeans. For phosphorous, we were only able to compare the actual application rate for phosphorus to the NMAN rate for phosphorus. 2.3. Empirical comparisons We hypothesized a ranking of input use among the three decision rules as NY > Np > NE

@f ðNpmax Þ ¼ PN Np wherePy  @N

ð5Þ

Maximization of the gross margin rate occurs where the value of an additional unit of fertilizer (left hand side of (5)) is equal to the incremental cost of the fertilizer or PN (Rajsic et al., 2009). Annual crop prices (Py) are from the historical compilations as reported by OMAFRA (2016) and the quarterly prices of urea and monoammonium phosphate (MAP) from McEwan (2008–2013) were converted to annual averages to get the respective fertilizer prices (PN) for nitrogen and phosphorus. The yield and gross margin decision rules required a crop specific yield response function to fertilizer (Y = f(N)). The following corn yield response function to nitrogen was adapted from Rajsic and Weersink (2008), YC = 1431 + 48.33  N
0.1364  N2

(6)

where YC is the annual corn yield, measured in kg/ha, and N is the annual nitrogen application rate measured in kg/ha.

(8)

The yield maximizing rate (NY) will be greater than the gross margin maximizing rate (Np) by definition unless the fertilizer price is zero, in which case the decision rules for gross margin maximization and yield maximization will result in the same application rate. The NMAN rate (NE) is expected to be less than the gross margin maximizing rate since the former limits nutrient loadings caused by over-application, whereas the later rate does not. We compared the means of the actual rate for each crop to each of three alternative rates, using a paired t-test under the assumption that the distribution of the variables is normal. Our first hypothesis is that the actual rate is greater than the NMAN rate NA > NE

where Py is the crop price per unit of output, Y, PN is the unit price of fertilizer and N is the nutrient application rate. The application rate that maximizes gross margin with respect to a single nutrient, gross margin (Np) is found by substituting the yield response function (4) into the gross margin (3), taking the derivative of the gross margin function with respect to the quantity of nutrient N, setting this first order condition to zero and solving for the corresponding N (Nicholson, 2005):

(7)

(9)

If this is the case, then excess nutrient application may occur if the NMAN rate is in fact the rate that eliminates excess nutrient application. On the other hand, levels below the NMAN rate (NA < NE) would suggest that, at least in the local study area, farmers are not applying excess nutrients. If there is no statistically significant difference between the actual rate and the gross margin maximizing rate, but both are above the NMAN rate, NA = Np > NE

(10)

then financial incentives might be required to reduce the rate to incentivize farmers to apply nutrients at the NMAN rate. On the other hand, if the actual rate is equal to the yield maximizing rate, which is higher than both the gross margin maximizing rate and the NMAN rate, that is NA = NY > Np > NE

(11)

then outreach programs to farmers could focus on increasing financial returns as well as improving environmental performance through reducing application rates. 2.4. Factors affecting over-Application We used an ordinary least squares (OLS) regression analysis to examine the influence of farm characteristics on farmers’ decisions to apply fertilizer in excess of crop requirements. We estimated four separate regressions: (1) excess nitrogen on corn, (2) excess nitrogen on wheat, (3) excess phosphorus on corn, and (4) excess

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

phosphorus on wheat. The dependent variable in each case is the difference between the actual application rate and the recommended NMAN rate (NA
NE) on each farm field for the nutrient and crop being analyzed. We used Stata to run the regression, and tested for heteroskedasticity and multicollinearity in each of the four scenarios. The explanatory variables are listed in Table 2 along with a definition and summary statistics by nutrient and crop. We had no prior expectations on the signs of the coefficients on farm size, field size, crop rotation and last year’s crop. We hypothesized that an increase in yield is correlated to an increase in the difference between the actual application rate and the recommended NMAN rate. We also hypothesize that an increase in the difference between NA
NE for one nutrient (i.e. phosphorus) will lead to an increase in the difference between the actual application rate and the NMAN rate for the other nutrient (i.e. nitrogen) as the decision to over apply one nutrient may lead to excess application of the other nutrient. 2.5. Nutrient applications at watershed-Level We calculated the total and excess amount of nitrogen and phosphorus applied to the two crops (corn and winter wheat) at the watershed scale in order to determine the aggregate impacts of the individual fertilizer application decisions. Gross nutrient application per year is the total weight of nutrients applied in the watershed as given by: TNj ¼

66 X NAij  Aij

ð12Þ

i¼1

where TNj is the total application of the nutrient on crop j for the watershed in kg, NAij is the actual nutrient application rate for crop j on field i (i = 1,2, . . . , 66), and Aij is the area of crop field i for crop j. TNj is calculated for each nutrient (nitrogen and phosphorus) on corn and soybeans. The net or excess nutrient application is also calculated for each nutrient and each crop at the watershed level. The excess is determined by comparing the actual to the NMAN rate for each

113

field according to ENj ¼

66 66 66   X X X NAij  Aij
NEij  Aij ¼ NAij
NEij  Aij i¼1

i¼1

ð13Þ

i¼1

where ENj is the excess application of a nutrient on crop j for the watershed in kg, and NEij is the NMAN nutrient application rate field i for crop j. Using the annual gross (equation (12)) and net nutrient applications (equation (13)) at the watershed scale, we also calculated the percentage of excess nitrogen and phosphorus applied in each year for both corn and winter wheat. This percentage illustrates the compounding effects of individual excess nutrient application decisions at the watershed scale. 3. Results 3.1. Actual application rates (NA) versus NMAN application rates (NE) The annual actual nutrient application rates (NA) and the NMAN recommended rates (NE) for nitrogen and phosphorus on corn, winter wheat and soybeans are listed by year in Table 3. The Table also reports the number of fields in production of the specified crop in each year and the range of application rates. The averages reported in Tables 3 and 4 are area weighted averages. For example, the actual application rate of nitrogen on corn in 2008 of 188.5 kg/ha is a weighted average of the application across the 27 fields planted to corn in that year with the weights of an individual application based on the size of the field relative to the total amount of corn planted, Xn N  Ai i¼1 Ai NA ¼ ð14Þ X n A i¼1 i where NA is area weighted average actual application rate, NAiis the actual application rate of the nutrient on field i, n is the number of fields in production of either corn or winter wheat, and Ai is the size of field i. Before proceeding with further analysis, we examined the validity of including the predicted nutrient application rates

Table 2 Summary Statistics of Variables in Regression for Over-Application (NA
NE).

Variable

Description

Corn Mean

Dependent Variable (NA
NE)i Difference between actual nutrient application rate (NA) and NMAN rate (NE) for nutrient i = nitrogen N? (N) and phosphorus (P) in kg/ha/yr
1.63 P? 13.91 Explanatory Variables Farm Characteristics Farm size Overall size of farm within the watershed (ha) Field size Total Average field size (ha) BMPs Crop Dummy variable
0 for a two crop rotation and 1 for two or more crop rotation rotation Dummy variable- 0 for no manure and 1 with manure applied Manure (NA
NE)j See above but for corresponding nutrient Economics PN/Pcrop Nitrogen price to crop price ($/t) PP/Pcrop Phosphorus price to crop price ($/t) Yield Actual (or target) crop yield (Metric Ton/ha)

Winter Wheat Min

Max

Mean

Min

Max


93.00 164.00 7.89


105.00 57


1.00

76.00

7.47


9.00

39.00

110.02 10.55

334.88 93.44

11.06 0.09

132.79 13.50

334.88 93.44

19.85 0.44

0.82

0

1

0.95

0

1

0.06

0

1

0.06

0

1

290.89 206.87 7.80 5.34 10.81 12.80

379.48 254.62 202.81 12.16 7.10 5.73 8.80 5.84 7.40

375.82 13.15 4.20

114

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

Table 3 Summary Statistics of Actual (NA) and NMAN (NE) Nutrient Application Rates of Nitrogen and Phosphorus on Corn, Winter Wheat and Soybeans. Fertilizer

Corn Year

NA

NE

a b c

Winter Wheat a

n

b

Avg

c

Soybeans

Min

Max

n

Avg

Min

Max

n

Avg

Min

Max

Nitrogen

2008 2009 2010 2011 2012 2013

27 15 20 17 23 23

188.5 188.6 182.7 182.1 173.5 178.3

122 150 112 112 112 123

209 308 227 209 209 227

10 12 15 16 15 11

129.3 116.7 109.8 0.8 123.9 121.2

112 112 0 31 112 73

160 133 171 135 135 146

21 32 22 25 21 25

8.4 3.0 2.0 1.6 3.2 0.5

0 0 0 0 0 0

67 12 10 6 40 22

Phosphorus

2008 2009 2010 2011 2012 2013

27 15 20 17 23 23

47.2 39.8 35.1 14.7 26.1 33.8

14 14 0 0 0 0

135 91 54 135 54 54

10 12 15 16 15 11

2.4 2.3 11.2 9.5 10.9 14.1

0 0 0 0 0 0

27 15 39 32 39 39

21 32 22 25 21 25

7.5 4.9 3.9 2.6 5.0 0.7

0 0 0 0 0 0

40 20 40 13 24 13

Nitrogen

2008 2009 2010 2011 2012 2013

27 15 20 17 23 23

185.6 171.4 180.7 164.1 174.5 179.4

122 122 142 122 122 122

220 208 215 206 205 204

10 12 15 16 15 11

128.9 98.2 96.1 107.5 104.9 121.2

111 76 89 91 99 88

136 116 144 125 122 136

21 32 22 25 21 25

6.3 0.0 0.0 0.0 0.7 0.0

0 0 0 0 0 0

36 0 0 0 24 22

Phosphorus

2008 2009 2010 2011 2012 2013

27 15 20 17 23 23

27.7 26.8 24.2 22.2 17.5 22.1

0 9 0 0 0 0

59 59 54 59 54 54

10 12 15 16 15 11

0.4 1.0 1.1 1.4 1.0 0.0

0 0 0 0 0 0

26 9 29 20 20 11

21 32 22 25 21 25

3.1 0.3 0.0 0.8 0.6 0.0

0 0 0 0 0 0

17 13 0 13 13 11

2008–2010 are actual historical observations and 2011–2013 are values survey respondents were asked to project. Number of fields in production of the specified crop in a given year. Area weighted average of the actual application rate of on a field basis in the given year.

(2011–2013 observations) in the empirical analysis. The minimum and maximum values of NA represent the rates applied by an individual farmer on one of their fields in the stated year. The projected set of extreme individual values of NA for 2011–2013 is similar to the set of actual rates for 2008–2010. For example, the projected minimum (maximum) actual application rates of nitrogen on corn range from 112 to 123 (209–227) while the actual minimum values ranged from 112 to 150 (209–308) (Table 3). In addition, there is no statistically significant difference between the historical observations (2008–2010) and the predicted observations (2011–2013) in terms of either the mean or variance for the actual application rates of nitrogen and phosphorus. Therefore, we treat the values reported by farmers for future years (2011–2013) at the time of the survey in the same way as the values that they report for past years (2008–2010) in the remaining analysis. The average actual application rates of nitrogen are consistent with agronomic expectations in that relative rates across the crops are consistent with the nutrient demands of the crops. For example, the average actual application of nitrogen in 2009 for corn is 188.6 kg/ha, which is greater than the average rate for winter wheat (116.7 kg/ha/) and soybeans (3.0 kg/ha). The minimal rates of nitrogen fertilizer applied on soybeans are expected, given the ability of the crop to form a symbiosis with bacteria that fixes nitrogen from the atmosphere. The application rates for nitrogen (average, minimum and maximum) tend to be highest across the crops in 2008 and 2009, which may be due to the significant price jump that had just occurred for these crops, particularly corn and soybeans. The average actual application rates of phosphorus across crops also follow expected agronomic trends across the six years of observations. For example, in 2010, the average actual application rate of phosphorus in corn production (35.1 kg/ha/yr) is greater than the average actual rate of phosphorus in winter wheat

(11.2 kg/ha/yr) and soybean (3.9 kg/ha/yr) production. The range of values of the actual application rate of phosphorus varies more than that of nitrogen across crops and years. Of particular concern are the instances where the phosphorus applied by a farmer is much higher than the average rate applied by other farmers in that year. In contrast, the individual application rates of nitrogen are not significantly greater than the average. The maximum actual application rate in four of the six years is less than 30% higher than the average actual application rate. For example, the maximum and average actual application rates in 2008 were 209 kg/ha/yr and 188.5 kg/ha/yr respectively. In the case of phosphorus, the average rate is often much smaller than the maximum rate, differing between 75% and 270% across the six years of data, suggesting the possibility of over-application. The large range in phosphorus applications across farmers is highlighted further below. The recommended NMAN rates for each nutrient also fit into the range of expected agronomic values across crops. For example, the average NMAN application rate for nitrogen is 180.7 kg/ha/yr on corn, 96.1 kg/ha/yr on winter wheat and 0 kg/ha/yr on soybeans (see Table 3). Recommended phosphorus rates are significantly less than nitrogen rates across crops and are minimal on most winter wheat and soybean fields in the survey. The average actual application rates tend to be higher than the average NMAN rates across the six years with the differences greatest for phosphorus. For example, the average actual phosphorus application rate of 47.2 kg/ha/yr on corn in 2008 was 70% greater than the average rate of 27.7 kg/ha/yr recommended by NMAN. While NA was generally greater than NE on average for each year, an exception was 2011 when the NE for phosphorus on corn was 22.2 kg/ha/yr and the average actual rate was 14.7 kg/ha/yr. While the minimum rates of NA and NE for the nutrients are similar across the years, the higher average is partially due to the

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120 Table 4 Comparisons of Area Weighted Average Farm Level Actual Nutrient Application Rates (NA) with NMAN (NE), Profit Maximizing Rate (Np) and Yield Maximizing Rate (NY) (kg/ha/yr) for Corn Production in the Gully Creek Watershed from 2008 to 2013. Farm

Nitrogen a

Phosphorus NA

NA
NE

207.07

37.94 (0.000)b

42.20 (0.000)

29.38 (0.166)

54

0 (0.500)

2

196.09

0.74 (0.249)

32.95 (0.288)

18.40 (0.484)

37.40

0.00 (0.196)

3

150.00

0.00 (0.211)


14.52 (0.010)


27.79 (0.000)

45.00

36.00 (0.000)

4

205.83

12.93 (0.024)

41.00 (0.000)

28.15 (0.000)

36.55

0.00 (0.173)

5

191.43

30.65 (0.000)

26.82 (0.000)

13.75 (0.002)

53.01

34.80 (0.000)

6

190.27

30.70 (0.001)

25.93 (0.000)

12.58 (0.000)

21.41

0 (0.178)

7

167.92


26.70 (0.000)

3.49 (0.017)


9.76 (0.036)

20.36


11.42 (0.000)

Farm


53.92 (0.000)


12.77 (0.000)


27.71 (0.000)

33.00

0.21 (0.000)

1

8

150.00

NA
NE

NA
Np

NA
Ny

(Table 4). There is a high level of variability in the actual nitrogen application rates on corn in the survey. Five farmers applied in excess of the NMAN rate by up to 37.94 kg/ha/yr and eight farmers applied up to 53.92 kg/ha/yr below the recommended NMAN rate. Of the 16 farmers surveyed, only three applied at a rate not statistically different from the recommended NMAN rate for nitrogen on corn. In the case of phosphorus in corn production, the average actual rate was 13.5 kg/ha/yr greater than the recommended NMAN rate, which is consistent with our hypothesis (equation (8)). None of the farmers applied phosphorus below NE and more than half applied at rates statistically greater than NE. Nine applied phosphorus up to 36 kg/ha/yr above the recommended NMAN rate of 9 kg/ha/yr. The average actual application rate of nitrogen on wheat across the 13 farmers growing wheat (as 3 farmers did not grow winter

1

NA

115

Table 5 Comparisons of Area Weighted Average Farm Level Actual Nutrient Application Rates (NA) with NMAN (NE), Profit Maximizing Rate (Np) and Yield Maximizing Rate (NY) (kg/ha/yr) for Winter Wheat Production in the Gully Creek Watershed from 2008 to 2013. Nitrogen

Phosphorus

NA

NA
NE

NA
Npmax

NA
Nymax

NA

NA
NE

112.00a

17.00 (0.124)

11.53 (0.013)


60.64 (0.000)

0

0 –

b

9

179.00


27.50 (0.017)

14.36 (0.077)

1.16 (0.043)

30.59

0.00 (0.250)

2

136.23

7.66 (0.108)

85.68 (0.039)


36.40 (0.071)

13.17

0.00 (0.196)

10

183.01


23.60 (0.000)

19.76 (0.001)

5.32 (0.014)

19


5.26 (0.187)

3

–

–

–

–

–

–

4

109.20

157.00


42.68 (0.003)


8.88 (0.007)


20.71 (0.000)

23.54

14.46 (0.001)

17.23 (0.001)

35.08 (0.003)


63.44 (0.000)

0.00

11

0.00 –

5

129.02

94.28


31.62 (0.000)


32.19 (0.000)

94.28 (0.000)

22.23

15.33 (0.000)

16.32 (0.000)

54.83 (0.000)


43.62 (0.000)

29.10

12

29.10 (0.000)

6

86.51

144.00


58.00 (0.000)


22.32 (0.000)


33.73 (0.000)

22.00

13.00 (0.000)

10.50 (0.093)

7.99 (0.270)


86.13 (0.060)

0.00

13

0 –

7

112.00

63.43


47.95 (0.007)


100.64 (0.000)


114.25 (0.000)

–

–

9.76 (0.004)

38.25 (0.007)


60.64 (0.000)

0.00

14


0.64 (0.041)

8

112.00

179.37

12.21 (0.190)

14.73 (0.465)

1.68 (0.260)

50.01

21.77 (0.030)


10.33 (0.000)

56.47 (0.016)


60.64 (0.000)

0.00

15

0.00 –

9

74.67

164.00

35.47 (0.000)


0.58 (0.287)


13.70 (0.000)

14.00

5.00 (0.000)

11.33 (0.092)

4.02 (0.324)


97.97 (0.060)

0.00

16

0.00 –

10

110.40

168.26


1.36 (0.349)

3.78 (0.150)


9.42 (0.005)

32.19

13.50 (0.000)

2.63 (0.300)

43.06 (0.383)


62.24 (0.036)

11.00

Average


0.68 (0.196)

11

113.13

10.72 (0.091)

22.96 (0.079)


59.50 (0.000)

0


7.79 (0.091)

12

–

–

–

–

–

–

13

112.00

11.00 (0.000)

15.57 (0.010)


60.64 (0.000)

0.00

0.00 –

14

80.59


29.82 (0.023)


11.12 (0.168)


92.05 (0.000)

6.00

0.40 (0.173)

15

50.06


11.14 (0.009)


22.29 (0.001)


122.58 (0.0001)

0.00

0.00 –

16

–

–

–

–

–

–

Average

107.59

7.89 (0.027)

26.20 (0.000)


65.05 (0.000)

9.92

7.47 (0.000)

a

The area weighted average of the farmers’ actual application rate (kg/ha/yr) across the six years of observations. b p-value for the t-test comparing difference in average individual application rates.

higher maximum actual rates. For example, the maximum phosphorus rate recommended by NMAN for an individual farmer was 59 kg/ha/yr in 2008 whereas the actual maximum was 135 kg/ ha/yr. However, the maximum NA and NE rates for phosphorus on corn are very similar in half of the years. Similarly, the maximum actual and recommended nitrogen rates are generally close in value but, when NA is greater than NE, it is significantly greater. For example, the maximum NMAN rate for nitrogen on corn was 208 kg/ha/yr in 2009 but the maximum actual rate for corn in that year was 308 kg/ha/yr. The average NMAN rate of nitrogen on corn is 1.36 kg/ha/yr more than the average actual application but the difference according to the paired t-test is not statistically significant

a The area weighted average of the farmers’ actual application rate (kg/ha/yr) across the six years of observations. b p-value for the t-test comparing difference in average individual application rates.

116

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

wheat over the period of analysis) is 7.89 kg/ha/yr higher than the NMAN rate, as hypothesized (Table 5). While the paired t-test indicated the average difference is statistically significant across all farms, there is a large variation in the relationship between NA and NE across the individual farms. Seven applied nitrogen at rates statistically significantly greater than the NMAN rate, three applied below the NMAN rate, and NA and NE are not statistically different for the remaining three farmers (Table 5). The average actual phosphorus application rate on wheat was also statistically higher than the recommended NMAN rate. The average 7.47 kg/ha/yr difference across all farms, however, hides significant differences across individual farms (Table 5). There was no significant difference in NA and NE for phosphorus on wheat for all farmers apart from one. This farmer grew an average of 25 ha of winter wheat in 2010, 2012 and 2013 and applied 29 kg/ha/yr more phosphorus than recommended by NMAN. Thus, a single farmer is responsible for the over-application of phosphorus, exceeding the NMAN rate on winter wheat. The NA and NE rates for nitrogen and phosphorus application rates on corn (Fig. 2) and winter wheat (Fig. 3) are generally similar with a few exceptions. In the case of nitrogen on corn, most farms are applying between 150 kg/ha/yr to 200 kg/ha/yr and the NMAN rates tend to be in the same range. Data points are distributed around the 45 line where 45 line NE is equal to NA with only a few outliers (one farm was applying over 300 kg/ha/yr in one year

despite NE of just over 200 kg/ha/yr, but were a few cases of no nitrogen application despite the NE recommendations of over 150 kg/ha/yr) (Fig. 2). In contrast, data points for phosphorus on corn are generally on or above the 45 line (lower panel, Fig. 2). There are several cases in which the actual application of phosphorus is significantly higher than the NMAN rate. Similar results are observed for winter wheat (see Fig. 3). As was the case with corn, nitrogen data points in the upper panel of the Figure cluster around the 45 line (NE = NA) with a few farms applying significantly less than their NE level. In the lower panel of Fig. 3, phosphorus application rates generally lie above the 45 line with some high actual application rates occurring even when the recommended rate is zero. Recall that information on the yield response to phosphorus, in the year of fertilizer application, is not available for these crops in Ontario, and it is possible that farmers, without the ability to optimized application rates apply this nutrient prophylactically. 3.2. NA versus rates maximizing gross margin (Np) and yield (NY) The difference between actual application rates and the recommended rates could be due to decision rules other than minimizing excess nutrient loadings. The difference between NA and the nitrogen application rate to maximize yield (NY) and gross margin (Np) are calculated according to equation (2) and equation

Fig. 2. Actual Nutrient Application Rate (NA) vs. NMAN Application Rate (NE) for Nitrogen (A) and Phosphorus (B) on Corn (The 45 line represents where the NA is equal to NE).

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

117

Fig. 3. Actual Nutrient Application Rate (NA) vs. NMAN Application Rate (NE) for Nitrogen (A) and Phosphorus (B) on Winter Wheat (The 45 line represents where the NA is equal to NE).

(5) respectively for both corn and winter wheat. Note there are no available yield response functions for the effect of phosphorus on these two crops so the analysis is only conducted for nitrogen applications. The average actual nitrogen application rate on corn is 3.78 kg/ ha/yr larger than Np but the difference is not statistically significant, which is consistent with the hypothesis that most farmers choose rates to generate the highest net returns (Table 3). However, there are large variations in the average differences across the 16 farms over the period of analysis. The actual rate is statistically significantly larger than Np on seven farms, significantly less on six farms, and no significant difference is evident on the remaining three farms. While there is also a large range in the individual farm nitrogen application rates for wheat (Table 4), NA is generally larger than Np on the majority of farms and is only statistically significantly less on one operation. On average across all farms, NA is 26.2 kg/ha/yr greater than Np for nitrogen on wheat. The yield maximizing rate (NY) for nitrogen is higher than the gross margin maximizing rate as expected (see equation (8)), since the latter accounts for the costs of the fertilizer. On average, NA for nitrogen is 9.42 kg/ha/yr less than NY on corn and 65.05 kg/ha/yr less than NY on wheat. The statistically significant difference between than NA and NY is also true for each of the 16 farms within the watershed for wheat. However, while the difference between NA and NY for nitrogen on corn becomes more negative than the difference between NA and Np as expected, there are still five

individual farms for which NA is statistically higher than NY. The high rates not only exceed the recommended NMAN rate but are above levels necessary to maximize yield. 3.3. Factors affecting application rate differences The results of the four OLS regressions estimated to explain the factors affecting the difference between actual application rate and the recommended NMAN rate at the field scale are summarized in Table 6. Given the panel nature of the data, the adjusted R-square values suggest the regressions explain the variation in NA – NE reasonably well. The coefficients for farm size are significant and positive across both crops and nutrients, suggesting that larger farms are more likely to apply nutrients at a rate greater than the recommended NMAN rate. On average, an increase in farm size by one hectare results in an approximate 0.2 kg/ha/yr increase in the actual rate compared to the NMAN rate. With more area to fertilize and the potentially increasing heterogeneity in the fertility of that larger land base, the average level of fertilizer applied increases as farmers are less likely to target rates to individual locations. Increases in field size reduce NA relative to NE but the difference is statistically significant only for phosphorus on wheat. The result suggests increasing returns to identifying differences in spatial fertility across a field. Together, the estimated coefficients suggest that increasing the potential net returns to variable rate

118

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

Table 6 Regression Results of Factors Affecting Differences between Actual and NMAN Nutrient Application Rates (NA
NE) for Nitrogen and Phosphorus on Corn and Winter Wheat Production. Explanatory Variable

Corn

a

Farm Size Field Size Crop Rotation

Phosphorus (P)

Nitrogen (P)

Phosphorus (P)

0.236*** c (0.039) b
0.017 (0.258)
30.700*** (7.884)

0.107*** (0.015)
0.136 (0.110)
2.660 (4.537)
6.848* (3.228) –

0.183*** (0.064)
0.036 (0.195)
12.919 (8.457)
28.308** (11.297)
0.583 (0.380) –

0.124*** (0.007)
0.137** (0.054)
0.942 (1.707)
1.265 (3.032) –

22.087*** (7.437)
0.196 (0.229) –

Manure (NA

NE)- P (NA

NE)- N

Constant


2.144 (1.071) 0.055 (0.037)
15.643*** (2.203) 160.645*** (26.562)

# of observations R-squared Adj R-squared

131 0.62 0.60

PN/Pcrop PP/Pcrop Yield

a b c

Winter Wheat

Nitrogen (N)


0.009 (0.035) 0.001 (0.014) 0.766* (0.394)
0.186 (0.988) 2.963 (11.564) 131 0.57 0.55

0.087 (0.053)
2.744** (1.270)
12.093** (5.372) 71.114** (32.158)


0.006 (0.011)
0.007 (0.007)
0.010 (0.246) 2.290** (0.961)
17.151*** (6.575)

85 0.50 0.45

85 0.90 0.89

Variable definitions in Table 2. Standard errors are in parentheses. Variable has a statistically significant effect at 1% (***), 5% (**), and 10% (*).

application will lower the actual application rate relative to the NMAN rate. The use of a crop rotation more complex than a simple cornsoybean rotation also reduces NA relative to NE but the result is only statistically significant for nitrogen on corn. The next more complex rotation involves the incorporation of winter wheat, and this third crop in the rotation enhances organic matter and nitrogen availability, particularly if a cover crop is also planted with the wheat. While NMAN accounts for this in its recommendation, the farmers using a more complex rotation are also aware of the

fertility enhancement and reduce NA more than if they were using a simple rotation. Using manure as a nutrient source has a statistically significant positive (negative) effect on the actual rate of nitrogen applied on corn (winter wheat) relative to the recommended NMAN rate. Recall that data on the actual nutrient content of manure were not collected in the survey and we had to rely on assumed values. Applying inorganic fertilizer reduces NA relative to NE for phosphorus on both crops, but the absolute impact is smaller than for nitrogen. The result is due to the relative nutrient content

Table 7 Total and Excess Nutrient Applications for Corn Production in the Gully Creek Watershed on an Annual Basis, and Total Across the Six Years of Observations. Crop

Year

Total Nitrogena (kg)

Excess Nitrogenb (kg)

% of Excess to Total

Corn

2008 2009 2010 2011 2012 2013 Total d

45,391.79 50,617.51 38,845.66 42,198.48 38,800.25 39,740.68 255,594.36

697.35 4,625.97 426.30 4,187.40
227.53
228.56 9,480.93

Wheat

2008 2009 2010 2011 2012 2013 Total e

31,105.62 16,054.85 15,967.15 21,329.62 23,882.43 26,282.67 13,4622.34

Both

Total

f

390,216.70

a b c d e f

c

Total Phosphorusa (Kg)

Excess Phosphorusb (Kg)

% of Excess to Total

1.54% 9.14% 1.10% 9.92%
0.59%
0.58% 3.71%

11,354.67 10,673.25 7,458.10 10,245.41 5,830.54 7,526.98 53,088.94

4,681.69 3,469.23 2,305.37 5,100.19 1,922.61 26,03.92 20,083.01

41.23% 32.50% 30.91% 49.78% 32.97% 34.59% 37.83%

97.23 2,555.72 1,992.71 1,048.21 3,650.48 1,633.08 10,977.41

0.31% 15.92% 12.48% 4.91% 15.29% 6.21% 8.15%

3,672.87 314.38 2,095.55 1,792.52 2,109.20 3,054.42 13,038.93

2,585.54 178.85 1,931.22 1,522.90 1,915.10 3,049.45 11,183.05

70.40% 56.89% 92.16% 84.96% 90.80% 99.84% 85.77%

20,458.34

5.24%

66,127.87

31,266.06

47.28%

The total nutrient application (TN), measured in kilograms, in the watershed (see equation (12)). The excess nutrient application (EN), measured in kilograms, in the watershed (see equation (13)). % = (EN/TN)*100. The total nutrient application across the six years of observations for corn production. The total nutrient application across the six years of observations for wheat production. The total nutrient application across the six years of observations for both corn and winter wheat production.

c

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

of manure and the crop nutrient requirements. Weersink et al. (2004) found the phosphorus standard more restrictive than a nitrogen standard for swine farmers using manure as a fertilizer source. The coefficient for target yield is statistically significant and negative for nitrogen in both corn and winter wheat production. The estimated coefficients suggest that a one MT/ha/yr increase in the target yield reduces excess nitrogen application by 15 (12) kg/ ha/yr for corn (winter wheat). The higher the target yield, the greater the use of the available nutrients by the crops and the lower will the excess nutrient loading. The effect is much smaller for phosphorus. 3.4. Watershed impacts of application rate differences The effect of the actual nitrogen and phosphorus fertilizer application decisions onto corn and winter wheat on excess nutrient applications from the Gully Creek watershed are calculated per equation (13). Total nutrient application is the total weight of nutrients applied in the watershed and excess nutrient application is the total nutrient application less the crop requirements as defined under the NMAN rate. Corn accounts for approximately two-thirds of the total nitrogen application and 80% of the total phosphorus application in the watershed. The share of corn on the total nutrient application is expected given the highest relative amount of fertilizer is applied to corn, and corn is planted on one-third of the crop area in the watershed. Excess nutrient applications differ significantly across crop and nutrient (see Table 7). Across the six years of observations, the excess nitrogen application on corn and winter wheat at the watershed level is only 5.2% of the total nitrogen application. In the case of corn, the excess nitrogen application was minimal for four of the six years and represented around 9.5% of the total nitrogen application in 2009 and 2011. On the other hand, while the excess nitrogen application on corn is small in percentage terms, it accounts for close to half of the 20,458 kg of excess nitrogen application over the six years. Although the total nitrogen application from wheat is less than half of the total for corn, the excess nitrogen application is approximately 1500 kg greater (10,977 kg vs. 9481 kg). The difference is due to the number of farms in which NA was greater than NE for nitrogen on wheat (see Fig. 3 and Table 4). Total phosphorus application (66,128 kg) is significantly less than total nitrogen application (390,217 kg) for the watershed due to the lower per hectare application rates for phosphorus than nitrogen on corn and soybeans (See Table 3 and 4). However, average aggregate excess phosphorus application is over 47% of the total application of this nutrient for these two crops. Total excess application of phosphorus at 31,266 kg is approximately 50% greater than the excess nitrogen application at 20,458 kg over the 6 years of data. Approximately two-thirds of the excess phosphorus application is from corn (20,083 kg) and this amount represents around 38% of the total phosphorus application for corn. While the total phosphorus application for wheat (13,083 kg) is around onequarter of the total phosphorus application in the watershed, the 11,183 kg of excess application is one-third of the total excess application. The excess phosphorus application in winter wheat production is 86% of the total phosphorus application on wheat. 4. Conclusions The regions surrounding Lake Erie have pledged to reduce phosphorus loadings by 40% over the next decade. Given that agriculture accounts for nearly half of total phosphorus loadings into Lake Erie, the desired reduction in loadings will require an understanding of how fertilizer application levels can be reduced

119

by farmers. This paper has compared farmers’ actual nutrient application rate decisions to the rates that would minimize excess loadings while meeting crop requirements referred to as the NMAN standard in Ontario. Actual application rates similar to or less than NMNA would suggest that total cropland area would have to be reduced in order to meet the target. However, actual rates greater than NMAN would imply that farmers’ decisions about the amount applied per hectare could be modified as the means to achieve the objective. If this is the case, then understanding the factors influencing the reasons for the excess nutrient application rate can guide policies to lower those rates. Our analysis shows that farmers’ decisions to apply in excess of the recommended NMAN rate are linked to nutrient and crop type. Nitrogen application rates are highest on corn but the difference between actual and NMAN rates are statistically insignificant as some farmers apply less than the crop requirements. In the case of wheat, nitrogen application rates are statistically higher than NMAN rates. The pattern for phosphorus differs from nitrogen for both corn and winter wheat. Actual phosphorus application rates are rarely less than NMAN rates and are frequently substantially higher. At a watershed level, excess nutrient applications as a percentage of total nutrient applications are much higher for phosphorus than nitrogen and higher for wheat than for corn. Since we observed considerable variability across farms, we examined the factors contributing to the deviations between actual application rates and NMAN application rates. Using decision rules such as profit maximization and yield maximization result in higher rates than NMAN but do not explain the outliers that apply significantly more than the NMAN rates. Excess nutrient applications increase with farm size and decrease with field size. Using beneficial management practices, such as a complex rotation or manure, lowers the NMAN rate but also tends to lower the difference between the actual and this recommended rate, suggesting farmers using these BMPs are aware of their fertility benefits. Finally, increases in target yield reduce the difference between actual and recommended rates, particularly for nitrogen. Within the Gully Creek watershed, none of the 16 farms in the data set were regulated under the Nutrient Management Act (NMA), and consequently none were required to submit Nutrient Management Plans and determine the associated NMAN rate. Despite not being a requirement, many of the farmers apply their fertilizer at rates close to that recommended by NMAN and some at rates significantly less. While the majority of farms are applying fertilizer, particularly nitrogen, at rates close to the minimum crop requirements, there are a few farms that apply much more than recommended. Policy and research efforts should be directed toward targeting these individuals that appear to be the primary contributor to the nutrient loading issue. Access to a unique micro level data set enabled us to understand nutrient application decisions at the farm and field level in a way that is not possible with more aggregate approaches. Our results provide confirmation of those aggregate level estimates in that we find strong evidence of application of phosphorus in excess of estimated plant nutrient requirements. The absence of publically available agronomic information on the yield response of field crops to phosphorus not only limits our analysis, but is a practical constraint on farmers who might want to optimize application of this nutrient. It is possible that the observed differences between actual application rates and NMAN recommended rates represent cases of farmers applying phosphorus out of risk aversion. In addition, we did not have data on the actual nutrient content of manure applied to the farms in this study. This also, in addition to being a limitation to research, is a practical problem for farmers who want to optimize their overall nutrient application strategy. Improving access to timely, inexpensive and accurate information

120

J.E. Leslie et al. / Agriculture, Ecosystems and Environment 240 (2017) 109–120

on the nutrient content of manure may be an important policy lever to reduce excess phosphorus use. Acknowledgments This research is supported by Agricultural and Rural Policy Theme of OMAFRA-University of Guelph Partnership Research Program. We would like to thank Mari Veliz, Abigail Gutteridge, Tracey McPherson, and Brynn Upsdell Wright of Ausable Bayfield Conservation Authority and Kevin McKague of OMAFRA for providing the farm survey data collected through OMAFRA’s Watershed Based BMP evaluation (WBBE) program. References Barker, A.V., Pilbeam, D.J., 2006. Handbook of Plant Nutrition. Taylor & Francis. CRC Press, Boca Raton, FL, USA. Bernstein, N., Ioffe, M., Bruner, M., Nishri, Y., Luria, G., Dori, I., Matan, E., PhilosophHadas, S., Umiel, N., Hagiladi, A., 2005. Effects of supplied nitrogen form and quantity on growth and postharvest quality of Ranunculus asiaticus flowers. J. Hort. Sci. 40 (6), 1879–1886. Blanco, H., Lal, R., 2008. Principles of Soil Conservation and Management. Springer, Dordrecht ISBN: 978-90-481-8529-0. Calzadilla, A., Rehdanza, K., Tol, R.S.J., 2010. The economic impact of more sustainable water use in agriculture: a computable general equilibrium analysis. J. Hydrol. 384, 292–305. doi:http://dx.doi.org/10.1016/j. jhydrol.2009.12.012. Dupont, D.P., 2010. “Cost-Sharing Incentive Programs for source water protection: The Grand River’s Rural Water Quality Program”. Can. J. Agric. Econ. 58 (1), 481– 496. Elmann, A., Garra, A., Alkalai-Tuvia, S., Fallik, E., 2016. Influence of organic and mineral-based conventional fertilization practices on nutrient levels, antiproliferative activities and quality of sweet red peppers following cold storage. Isr. J. Plant Sci. 63 (1), 51–57. doi:http://dx.doi.org/10.1080/ 07929978.2016.1151286. Engels, C., Kirkby, E., White, P., 2012. Mineral nutrition, yield and source-sink relationships, Marschner’s Mineral Nutrition of Higher Plants. 3rd edition Elsevier Science Ltd.. Government of Ontario, 2015. Blue-green AlgaeOntario Ministry of Environment and Climate Change. . Retrieved on October 18, 2015 from http://www.ontario. ca/page/blue-green-algae. International Joint Commission, 2014. A Balanced Diet for Lake Erie: Reducing Phosphorus Loadings and Harmful Algal Blooms. Report of the Lake Erie Ecosystem Priority. . Karlen, D.L., Andrews, S.S., Wienhold, B.J., 2004. Soil quality, fertility and health
historical context, status and perspectives. In: Schjonning, P., Elmholt, S.,

Christensen, B.T. (Eds.), Managing Soil Quality Chall in Mod Agric. CABI Publishing, Wallingford UK. Liu, C., Sung, Y., Chen, B., Lai, H., 2014. Effects of nitrogen fertilizers on the growth and nitrate content of lettuce (Lactuca sativa L.). Int. J. Environ. Res. Public Health 11 (4), 4427–4440. McEwan, K., (2008–2013). Ontario farm input monitoring project, Survey #1-4 University of Guelph, Ridgetown Campus. Economics and Business Section. McKague, K., 2016. Personal Email Correspondence Re: Interpreting the NMAN output from AgriSuite. Received on March 30th, 2016. Nicholson, W., 2005. Microeconomic Theory: Basic Principles and Extensions, 9th ed. Thomson South-Western, Mason, OH. OMAFRA, 2015. Ontario’s Agricultural Planning Tools Suite (AgriSuite). . Retrieved on October 30th, 2015 from http://www.omafra.gov.on.ca/english/nm/nman/ agrisuite.htm. OMAFRA, 2016. Historical Provincial Estimates by Crop, 1981–2014 (Metric Units). . Retrieved on January 15th, 2016 from http://www.omafra.gov.on.ca/english/ stats/crops/estimate_hist_metric.htm. Oginskyy, A.M., 2014. Integrated Hydrologic-economic Optimization Modeling for Watershed Evaluation of Agricultural BMPs and Policies PhD Thesis. University of Guelph. Osmond, D.L., Hoag, D.L.K., Luloff, A.E., Meals, D.W., Neas, K., 2015. Farmers’ use of nutrient management: lessons from watershed studies. J. Environ. Qual. 44 (2), 382–390. Parris, K., 2011. Impact of agriculture on water pollution in OECD countries: recent trends and future prospects. Int. J. Water Resour. Dev. 27 (1), 33–52. doi:http:// dx.doi.org/10.1080/07900627.2010.531898. Rajsic, P., Weersink, A., 2008. Do farmers waste fertilizer? A comparison of ex post optimal nitrogen rates and ex ante recommendations by model, site and year. Agric. Syst. 97, 56–67. Rajsic, P., Weersink, A., Gandorfer, M., 2009. Risk and nitrogen application levels. Can. J. Agric. Econ. 52 (2), 223–239. Shtull-Trauring, E., Aviani, I., Avisar, D., Bernstein, N., 2016. Integrating high resolution water footprint and GIS for promoting water efficiency in the agricultural sector: a case study of plantation crops in the Jordan Valley. Front. Plant Sci. 7, 1877. doi:http://dx.doi.org/10.3389/fpls.2016.01877. Simmons, J., Wright, B.U., Velize, M., McKague, K., 2013. A Synthesis Report of the Watershed Based Best Management Practices Evaluation, HuronAusable Bayfield Conservation Authority & OMAFRA. . Retrieved on April 15th, 2016 from http://www.abca.on.ca/downloads/WBBE-Huron-Synthesis-2013-09-16.pdf. Stuart, D., Schewe, R.L., McDermott, M., 2014. Reducing nitrogen fertilizer application as a climate change mitigation strategy: understanding farmer decision-making and potential barriers to change in the US. Land Use Policy 36, 210–218. Walters, D.F., Smolikova-Wachowiak, R., Wachowiak, M., Shrubsole, D., Malczewski, J., 2013. Ontario's nutrient calculator: overview and focus on sensitivity analysis. J. Agric. Sci. 5 (11), 189–200. Weersink, A., DeVos, G., Stonehouse, P., 2004. Farm returns and land price effects from environmental standards and stocking density restrictions. Agric. Resour. Econ. Rev. 33 (2), 272–281.