Determining efficiency of energy input for silage corn production: An econometric approach

Determining efficiency of energy input for silage corn production: An econometric approach

Energy 93 (2015) 2166e2174 Contents lists available at ScienceDirect Energy journal homepage: www.elsevier.com/locate/energy Determining efficiency ...

471KB Sizes 0 Downloads 1 Views

Energy 93 (2015) 2166e2174

Contents lists available at ScienceDirect

Energy journal homepage: www.elsevier.com/locate/energy

Determining efficiency of energy input for silage corn production: An econometric approach Ehsan Houshyar a, *, Hamid Reza Zareifard b, Philipp Grundmann c, Pete Smith d a

Department of Farm Machinery Mechanics, Jahrom University, PO BOX 74135-111, Jahrom, Iran Department of Statistics, Jahrom University, PO BOX 74135-111, Jahrom, Iran c Leibniz-Institute for Agricultural Engineering Potsdam-Bornim e.V., Technology Assessment and Substance Cycles, Max-Eyth-Allee 100, 14469, Potsdam, Germany d Institute of Biological and Environmental Sciences, University of Aberdeen, 23 St Machar Drive, Aberdeen, AB24 3UU, UK b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 28 January 2015 Received in revised form 4 September 2015 Accepted 5 September 2015 Available online 19 November 2015

This study was undertaken to analyze the energy consumption patterns of silage corn production, the corresponding GHG emissions, the relationships between energy inputs and outputs, and the sensitivity of yield-to-energy inputs, using the CobbeDouglass econometric model and MPP (Marginal Physical Productivity) in the Fars province of southwest Iran. Although the average amount of inputs and outputs were analyzed, 20% of the maximum and minimum values were also reported as cluster 1 (C1) and cluster 2 (C2) farmers. The results showed that around 45e68 GJ/ha energy was needed to produce 67 e85 ton ha1 of silage corn. According to the MPP, the most effective inputs on the yield were human power, chemicals and seed energy inputs, since the yield had the highest sensitivity to these three inputs. Three energy input scenarios were proposed based on the average, minimum and maximum energy consumptions; i.e. LEI (Low Energy Input), MEI (Medium Energy Input) and HEI (High Energy Input) scenarios. The lowest energy and yield were consumed and produced in the LEI, respectively. The output einput energy ratio, energy productivity and kg yield per kg CO2, were highest in the LEI, although higher energy and yield were used and produced in the MEI and HEI, respectively. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Energy use efficiency Econometric models MPP (Marginal Physical Productivity)

1. Introduction Different sources of energy are used in agriculture to produce food for the increasing population [1,2]. Energy is applied directly, mainly through diesel engines in farm operations and for pumping water. The indirect energy, which is not applied within the farm gate, are consumed in the formulation, storage and distribution of farm inputs, such as fertilizers and chemicals [3,4]. Agriculture can be regarded as both an energy consumer and producer. Higher yields per unit area are obtained in modern agriculture largely through the external supply of energy, especially fossil fuels [5,6]. In the coming decades, the world will face the challenge of increasing energy efficiency and saving fossil fuels to reduce the negative impact of burning fossil fuels on the environment [79]. The gradual reduction of the availability of fossil fuels, and their increasing costs, will inevitably create problems for global

* Corresponding author. E-mail address: [email protected] (E. Houshyar). http://dx.doi.org/10.1016/j.energy.2015.09.105 0360-5442/© 2015 Elsevier Ltd. All rights reserved.

agriculture. It is possible to reduce the amount of fossil fuels used in agriculture by various methods that aim to conserve resources as well as use more effective cultivation techniques [10]. Renewable sources of energy such as solar, wind, geothermal and biofuels have a high potential of reducing dependency on fossil fuels and emitting lower GHGs [1114]. In addition to economic benefits, there are social and environmental advantages to reducing energy consumption, such as preserving fossil fuel supply and minimizing global warming. The limited availability of energy, particularly from fossil fuels, has led researchers to assess the energy-use-efficiency of different crops in different regions [1521]. Energy-use analysis is an important step in making appropriate energy policies [22,23]. A more efficient use of energy inputs will result in lower GHG emissions and environmental footprints [24,25]. It is essential to determine how different inputs affect crop growth, especially in agronomy. Finding the relationships between energy input and yield is necessary in order to focus on the most effective inputs to use and, consequently, to reduce waste. Furthermore, the sensitivity of crop output to each energy input can be determined by suitable econometric models [1,26].

E. Houshyar et al. / Energy 93 (2015) 2166e2174

Corn (for grain and silage) is a strategic crop in Iran as a main source of feed in poultry production and animal husbandry. The Fars province in southwest Iran is the top corn producer in the country [27]. Accordingly, two main objectives of this study were: 1) to evaluate the energy efficiency of silage corn production, and 2) to determine a suitable econometric model to estimate the appropriate energy input for silage corn production. To meet these objectives, the efficiency of energy use and related equivalent carbon emission were investigated by energy and environmental indices (See Section 2.1. for the methodology and Sections 3.1 and 3.2 for the results). Next, the suitable econometric model was estimated based on the energy inputs and output (see Section 2.2) by applying different statistical analyses (see Section 2.3). In the econometric analysis, the well-known CobbeDouglas production function was used to determine the relationships between energy inputs and silage corn yield. Afterwards, the MPP (Marginal Physical Productivity) measure was employed to determine the sensitivity of yield to each energy input, which in turn is helpful for determining which energy inputs are most effectively increased or decreased (see Section 2.2 for the methodology and Sections 3.3 and 3.4 for the results). Finally, three scenarios for using energy were proposed according to current energy inputs and outputs in Section 3.5. 2. Methodology This study was carried out to assess appropriate energy inputs in silage corn production and the sensitivity of yield to energy inputs in the Fars province. The province is located within 27 030 and 31ο 400 north latitude and 50 360 and 55 350 east longitude. The farmers were selected through a simple random sampling without replacement. The desired sample size of farms was calculated by Eq. (1) [28]: n ¼ (N  Z2  p  q)/(N  d2 þ Z2  p  q)

(1)

where: - n is the required sample size; - N is the number of holdings in target population; - Z is the reliability coefficient (1.96 which represents the 95% reliability); - p is equal to 0.5; - q is equal to 0.5; - d is the precision (xX) which is equal to 0.07 in this study.

2.1. Analyses of energy use and related CO2 emissions An eight-page questionnaire was designed to gather data related to silage corn production. The amount of each input, including fertilizers, chemicals, human power, diesel fuel for farm operations and water pumping, water and seed was calculated per hectare and multiplied by its energy equivalent (Table 1) to account for the total energy use in MJ/ha. The amount of indirect energy input in the manufacturing of farm machines was estimated by Eq. (2) [33]:

 EG  MJ ha1 TW

E is the cumulative energy demand for machinery, MJ kg1; G is the total weight of the specific machine, kg; T is the life time of machinery until replacement is required, h. W is the Effective field capacity in ha h1.

Common energy indices were employed to assess efficiency of energy inputs [30]:

Energy ratio ¼

energy output ðMJ=haÞ ðdimensionlessÞ energy inputðMJ=haÞ

Specific energy ¼

energy input ðMJ=haÞ ðMJ=kgÞ yield output ðkg=haÞ

Energy productivity ¼

yield output ðkg=haÞ ðkg=MJÞ energy input ðMJ=haÞ

Net energy gain ¼ energy output  energy input ðMJ=haÞ

(3)

(4)

(5)

(6)

These indices express the efficiency of energy used in a system. However, they cannot reveal the environmental risks of energy inputs. The equivalent CO2 from each energy input was calculated using available data given in Table 1. The amount of yield per kg emitted CO2 was determined as follows:

Y ðIGHG Þ ¼ Pn

j¼1

CI

ðdimensionlessÞ

(7)

where: -

(IGHG)A is the index of average kg yield per kg CO2; Y is the yield, kg ha1; Pn 1 j¼1 CI is the total CO2 from energy inputs, kg ha ; j ¼ 1, 2,…, n is the number of energy inputs.

Since the index shows kg yield per kg CO2, different groups of farmers or systems can be easily compared with regard to environmental risks of energy inputs. This index can be considered as corresponding to the energy productivity index, but also contains the environmental risk of production.

2.2. Econometric model of silage corn production

Accordingly, 194 farmers were chosen from a total of 2780.

ME ¼

-

2167

(2)

where: - ME is the indirect energy used in the farm machinery manufacturing, MJ ha1;

The relationship between energy inputs and silage corn yield was investigated using the CobbeDouglass production function. The CobbeDouglas and Translog are two common econometric models used to estimate the relationships between inputs and output [37]. The CobbeDouglas model in the power form is [26]:

Yi ¼ a

n Y

f

xij j eui

j¼1

where: -

yi is the yield of ith farmer; a is the constant term; xij is the vector of inputs; fj is the coefficients of inputs; eui is the error term; i ¼ 1, 2,…, K is the number of farmers; j ¼ 1, 2,…, n is the number of inputs.

(8)

2168

E. Houshyar et al. / Energy 93 (2015) 2166e2174

Table 1 Applied energy equivalents of inputs and corresponding CO2 emission in agricultural production. Input (unit)

Energy equivalent (MJ/unit)

Reference

Emission factor (kg CO2e/unit)

Reference

Machinery (kg) Diesel (L) Liquid chemical (L) Granular chemical (kg) Human power (h) Nitrogen (kg) Phosphorus (kg) Potassium (kg) Zinc sulphate (kg) Corn seed (kg) Water (m3) Silage corn yield

62.7 56.3 102 120 1.96 78.1 12.44 11.15 20.9 100 0.63 8

[15] [15] [29] [29] [30] [31] [32] [32] [15] [33] [34] [34]

0.072 kg/MJ 5.07 kg/L e 0.09 kg/MJ 25.3 kg/kg e 0.25 kg/MJ e 0.7 kg/man-h e 0.36 kg/MJ 3 kg/kg e 0.038 kg/MJ e e e 0.17 kg/m30.27 kg/MJ 0.17 kg/m30.27 kg/MJa e

[35] [13] [13] e [13] [13] e e e e [36] e

a

425.1 kg CE/ha for 25 cm of irrigation (2500 m3).

Converting the power to linear form gives the following function:

ln yi ¼ a þ

n X

  fj ln xij þ ei

(9)

j¼1

2.3. Statistical analysis

The relationship of seven energy inputs and the silage corn yield was investigated by Eq. (10): Econometric model: ln yi ¼ a þ f1 ln(x1) þ f2 ln(x2) þ f3 ln(x3) þ f4 ln(x4) þ f5 ln(x5) þ f6 ln(x6) þ f7 ln(x7) þ ei ln yi f1 x1 f2 x2 f3 x3 f4 x4 f5 x5 f6 x6 f7 x7 ei (10) where: - a is the constant term; - yi is the silage corn yield; - x1 , x2 , x3 , x4 , x5 , x6 and x7 are machinery, diesel fuel, fertilizers, chemicals, water, seed and human power energy inputs, respectively. Different regression methods were tested to find a suitable regression model, as described in the next section. The CobbeDouglas production function is useful because the economic status of production, the elasticity of each energy input, and the return to scale can be derived from the model. Economists have defined three production stages based on fj . Economically, the production is on the first, second and third stage if fj > 1; 0  fj  1 and fj < 0, respectively. The second stage of production is economically optimized, indicating that the input is used appropriately. The third stage of production shows that the input is applied in excess and should be reduced. The first stage shows that higher inputs are required to reach higher yield. P The sum of elasticity ( nj¼1 fj ) gives information about the returns to scale. The return to scale is increasing, constant and P P P decreasing if nj¼1 fj > 1, nj¼1 fj ¼ 1 and nj¼1 fj < 1, respectively. To find the sensitivity of output to energy inputs, the Marginal Physical Productivity (MPP) equation was employed [38]:

2.3.1. Testing model adequacy Every regression model is fitted based on several implicit assumptions. After fitting a regression model, it is important to determine whether all the necessary model assumptions are valid. Gross violations of the assumptions may yield an unstable model with totally different outputs and opposite conclusions for a special sample. The major assumptions of regression analysis in the current study are as follows:  The relationship between the response and the regressors is linear.  The variance of error terms is assumed to be equal. If the assumption of homoskedastic variance is not fulfilled, the OLS estimators are no longer minimum variance estimators [39].  The errors are normally and independently distributed with mean zero and variance s2 . With this normality assumption, the OLS estimators are minimum-variance estimators in the entire class of unbiased estimators, whether linear or not. In short, they are the best unbiased estimators [39,40].

(11)

The RESET (Ramsey's Regression Specification Error Test) was applied to test the functional form of the model [41]. The null hypothesis was that the functional form of the model is correct, while the OLS estimators will be biased and inconsistent if this hypothesis is rejected. The functional form of the model was confirmed by the RESET. In order to test heteroskedasticity (unequal variance), a number of alternative tests are available. The heteroskedasticity was investigated by the White test [42]. The White test revealed no heteroskedasticity in the model. Finally, the JarqueeBera statistic was used for testing residuals normality. The JarqueeBera statistic showed that the residuals of the model were distributed normally. Additionally, this result was supported by the KolmogoroveSmirnov test (Significant ¼ 0.62). The histogram of residuals in Fig. 1 shows that the data have a symmetrical distribution.

MPPxj is the marginal physical productivity of jth energy input; MðYÞ is the mean of yield; MðxijÞ is the mean of jth energy input; fij is the elasticity of jth energy input.

2.3.2. Multi-collinearity The use and interpretation of a multiple regression model often either explicitly or implicitly depends on the estimates of the individual regression coefficients. Among the considerations in the use of regression analysis, multi-collinearity problem can seriously disturb the fitting of least-squares. More precisely, the regressors

MPPxj ¼

MðYÞ  fij MðxijÞ

where: -

The MPP indicates the change in total output from a unit change on jth energy input while all the other inputs are fixed. Thus, an energy input with higher MPP would be considered critical, since yield is more sensitive to it.

E. Houshyar et al. / Energy 93 (2015) 2166e2174

2169

multi-collinearity between xj and any subset of the other K  1 regressors, then the value of R2j will be close to unity. Subsequently, the variances of some of the estimated regression coefficients can become very large, leading to unstable and potentially misleading estimates of the regression equation. Several techniques have been proposed for detecting multicollinearity. A very simple measure of multi-collinearity is inspection of the off-diagonal elements rij in X0 X. If regressors xi and xj are nearly linearly dependent, then rij  will be near unity. The X0 X matrix correlation for the variables of this study is:

1 0:85 0:86 X 0 X ¼ 0:83 0:78 0:79 0:76 Fig. 1. The histogram of residuals.

are orthogonal if there is no linear relationship between them. When the regressors are orthogonal, the inferences are more valid, although in most applications of regression, the regressors are not totally orthogonal. When there are near-linear dependencies among the regressors, the problem of multi-collinearity would be existed. Sometimes the lack of orthogonality is not serious. However, in some situations the regressors are linearly related in such a way that inference can be misleading or erroneous. Subsequently, it is important to have low collinearity among regressors for OLS since the coefficients given by the model cannot be reliable where regressors are highly collinear, even if a good prediction is obtained [39,43,44]. In other words, the separate influence of inputs on the output cannot be fully understood in spite of good prediction [45,46]. To demonstrate some of the effects of the presence of multi-collinearity, consider the usual multiple linear regression model:

y ¼ Xb þ ε;

  ε  Nn 0; s2 In

(12)

where y is an n  1 vector of responses, X ¼ ½x1 ; /; xK  is an n  K matrix of the regressor variables in which xj contains the n levels of the jth regressor variable, b is a K  1 vector of unknown constants, and ε is an n  1 vector of random errors. It is reasonable to assume that the regressor variables and the response are centralized and scaled to unit lengths; i.e. each variable is centralized by subtracting the mean of the variable and dividing by the square root of the corrected sum of squares for that variable. Consequently, X0 X is a K  K matrix of correlations between the regressors and X0 y is a K  1 vector of correlations between the regressors and the response. The multi-collinearity can be formally defined in terms of the linear dependence of the X columns. The vectors x1 ; x2 ; /; xK are linearly dependent if there is PK a set of constantst1 ; t2 ; /; tK , not all zero, while j¼1 tj xj is approximately zero. In this case, there will be a near-linear dependency in X0 X and the problem of multi-collinearity still existed. It can be shown that the diagonal elements of the matrix C ¼ ðX0 X Þ1 are:

Cjj ¼

1 ; 1  R2j

j ¼ 1; /; K

(13)

where R2j is the coefficient of multiple determinations from the regression of xj on the remaining K  1 regressors. If there is strong

0:85 1 0:82 0:78 0:77 0:72 0:68

0:86 0:82 1 0:84 0:82 0:79 0:74

0:83 0:78 0:84 1 0:78 0:76 0:74

0:78 0:77 0:82 0:78 1 0:70 0:67

0:79 0:72 0:79 0:79 0:79 1 0:67

0:76 0:68 0:74 0:74 0:67 0:67 1

Inspection of the correlation matrix indicates that there is nearlinear dependency between the regressors. It was also observed that the diagonal elements of the C ¼ ðX0 X Þ1 matrix are very useful in detecting multi-collinearity. Marquardt [47] mentioned the diagonal elements of the C as the VIFs (Variance Inflation Factors). The VIF for each term in the model measures the combined effect of the dependences among the regressors on the variance of that term. One or more high VIFs indicate multi-collinearity. If any of the VIFs exceeds 5 or 10, it is an indication that the associated regression coefficients are poorly estimated, due to multicollinearity. The VIFs of the variables diesel fuel, machinery, fertilizers, chemicals, water, seed and human power are 6.31, 4.30, 6.29, 4.49, 3.55, 3.15, 2.63, respectively. The correlation matrix and amounts of VIFs both clearly reveal that near-linear dependencies and multi-collinearity problems exist. Several techniques have been proposed for dealing with the problems caused by multi-collinearity. The general approaches include additional data collection [48,49]. The additional data should be collected in such a way that reduces the multicollinearity of the current data. Nonetheless, collecting additional data is not always possible due mainly to economic constraints or because the process being studied is no longer available for sampling. Variable elimination is another widely used approach in combating multi-collinearity. In other words, if high collinearity is found, one of the inputs with linear relationship can be removed from the model and new models can be provided [50]. Although variable elimination is often an effective technique, it may not provide a satisfactory solution if the regressors, dropped from the model, have significant explanatory power relative to the response y. In other words, eliminating regressors to reduce multicollinearity may damage the prediction profitability of the model [51]. As discussed in the previous section, all the coefficients in the CobbeDouglass production function are important since they show the elasticity of each input. Thus, it is essential to use some estimation methods other than OLS which reveal the coefficients reliably in addition to making good predictions. To solve the problem, PCR (Principle Component Regression), ridge regression and lasso regression are often proposed as methods that are widely used in many practical data analyses. They provide biased coefficient estimators with a relatively smaller variation than the variance of the least squares estimator [52]. The PCR is a regression analysis technique that is based on principal components corresponding with K inputs x1 ; /; xK . The PCR approach combats multi-collinearity by using less than the full set of principal components in the model [45]. Numerous methods for selecting a subset of components in the PCR approach are

2170

E. Houshyar et al. / Energy 93 (2015) 2166e2174

available [49,52]. Although ridge and lasso methods are principally different, both models shrink the regression coefficients by imposing a penalty on their size. Ridge regression, which was originally described by Refs. [53,54]; shrinks the coefficients of the principal components, with the shrinking being more dependent on the size of the corresponding eigenvalue. In other words, the ridge coefficients minimize a penalized residual sum of squares. All the three methods were considered in this study and the best was chosen. 2.3.3. Choosing the best model The result of the OLS and ridge regression models revealed that the machinery is not a significant corresponding variable to the yield while the significance of this input was 0.63 and 0.56, based on the OLS and ridge regression, respectively. It should be noted that all the other variables had significant contribution to the yield. The estimation of the regression coefficients based on the OLS, PCR and ridge regression are presented in Table 2. All the models were developed by cross-validation, because the simplest and most widely used method for estimating prediction error is probably cross-validation which also provides high precision results [55]. To compare the predictive performance of the models, the PRESS (Predicted Residual Sum of Squares) statistic was used, which is a form of cross-validation and one of the most popular techniques in this field [5658]. The PRESS statistic is defined as:

PRESS ¼

n X

2

ðyi  b y i Þ

(14)

i¼1

where b y i is the predicted value obtained from a model fitting while the ith observation is taken. PRESS is generally regarded as a measure of how well a regression model can predict some new data when a model with a small value of PRESS is desired. The value of this criterion for OLS, ridge regression and PCR, was determined as 0.587, 0.074, 0.072, respectively, revealing that the ridge regression and PCR perform better than the OLS method. Although both the ridge regression and PCR were considered, based on the reasons discussed, the ridge regression was preferably applied in this study. 3. Results and discussion 3.1. Energy use in the silage corn farming system The result showed that around 55,000 MJ ha1 energy were used to produce 73,000 kg ha1 silage corn (Table 3). The farmers in the study area consumed around 18,000 MJ ha1 less energy than farmers in Tehran province [59]. The four main energy inputs are fertilizers, water, diesel fuel and seed, which consumed around 65%, 13%, 12% and 5% of the total energy inputs, respectively. The

maximum and minimum values of energy consumptions belong to 20% of the best and the worst farmers and are also tabulated in two clusters as the C1 and C2 groups. The C1 farmers consume energy and produce silage corn by using 68,000 MJ ha1 of energy for yields of 85,000 kg ha1, while the C2 farmers use energy and produce silage corn at rates of 45,000 MJ ha1 and 68,000 kg ha1, respectively. Four energy inputs are consumed significantly higher by C1 farmers: nitrogen, water, diesel fuel and machinery. The investigation showed that two main tillage systems are applied by the farmers and that these led to different fuel and machinery energy inputs. About 65% of the farmers use mouldboard ploughing (20e25 cm), disking (only once or twice), land levelling and planting, whereas only two machines are used by the other farmers, namely, the combined chisel plough (35e40 cm tilling depth) and the row corn planter. The latter is called a ‘reduced’ tillage system since less machinery is applied compared to the conventional system [60]. Moving toward conservation tillage systems can significantly reduce energy input. In comparison to Turkish farmers, who use minimum tillage, the farmers in the study area have an approximately 0.80 kg MJ1 lower yield [61]. Although analyzing the costs connected with timeline was not the purpose of this study, most farmers complained that they could not find suitable farm machinery and implements at the time needed, which resulted in dry and uneven soil after farm operations, especially after ploughing. It is recommended that the area be supplied with enough and suitable farm machinery and implements. Currently, furrow irrigation systems are employed on about 85% of farms, using a well for each 5 ha. Although modern irrigation systems, such as sprinklers, are used on C2 farms, the farmers have some difficulties when they use the furrow irrigation system with reduced tillage, due mainly to more crop residues and deeper soil tilling. It seems that incomplete irrigation on such farms is also a reason for the lower yield. A comprehensive investigation, nevertheless, is recommended that future studies determine the amount of water needed for each tillage system in order to obtain the highest possible yield. Although different types of macro- and micro-element fertilizers are used, the share of nitrogen in energy inputs is noticeably higher (95%) than the total fertilizer energy input. This is critical since nitrogenous fertilizers used in the study area are very mobile and more energy intensive in their production than other fertilizers, and hence, the suboptimal use and losses reduce the sustainability of silage corn production. Around 18,600 MJ ha1 higher nitrogen energy input is used in C1. In spite of the higher yield in this group, the exact effect of nitrogen on the yield is not clear, and therefore, it is not possible to determine whether higher fertilizer consumption is reasonable. Unfortunately, soil sampling is usually not made by the farmers because of financial constraints or lack of awareness of the benefits of regular soil sampling. Similar ranges of fertilizers are applied in reduced and conventional tillage systems, although reduced tillage leads to lower fuel and machinery energy inputs. This is critical since farmers use different fertilizers without

Table 2 The coefficients of regressors. Model

OLS

Ridge regression

PCR

Unstandardized

Standardized

Unstandardized

Standardized

Unstandardized

Standardized

Constant x*1 x*2 x*3 x*4 x*5 x*6

3.331 0.325 0.112 0.106 0.127 0.453 0.091

e 0.222 0.209 0.173 0.125 0.224 0.106

3.36 0.322 0.111 0.106 0.128 0.450 0.092

e 0.22 0.208 0.174 0.126 0.222 0.107

3.66 0.284 0.118 0.119 0.129 0.436 0.091

e 0.194 0.221 0.195 0.127 0.215 0.106

E. Houshyar et al. / Energy 93 (2015) 2166e2174

2171

Table 3 Energy used in, and carbon emission from, silage corn farms.

1 e Machinery 2 e Diesel fuel 3 e Fertilizers 3-1 e Nitrogen 3-2 e Phosphorous 3-3 e Potassium 3-4 e Other (zinc sulphate, iron, etc.) 4 e Chemicals 5 e Water 6 e Seed 7 e Human Power

Average energy used (MJ/ha)

Maximum energy used (MJ/ha)

Minimum energy used (MJ/ha)

Average carbon released (kg/ha)

Maximum carbon released (kg/ha)

Minimum carbon released (kg/ha)

1289.74 6314.13 35615.70 33834.92 837.32 587.30 356.16

1540.30 7220.88 46837.82 45758.23 476.21 215.86 387.52

1226.83 5992.07 28130.89 27176.69 485.13 235.29 233.78

92.86 568.27 e 1353.40 e e e

110.90 649.88 e 1779.84 e e e

88.33 539.29 e 1068.97 e e e

468.29 7395.13 3149.59 260.69

586.32 8666.52 3439.29 307.52

378.94 6600.10 3027.03 233.78

117.07 1996.69 850.39 93.85

146.58 2339.96 928.61 110.71

94.74 1782.03 817.30 84.16

considering the type of tillage they use. Most agricultural and economic experts believe that extensional classes and some laws to restrict fertilizer use are necessary to prevent farmers from using extra fertilizers. Furthermore, mechanized, precision spreading of chemical pesticides and fertilizers are used by only 17% of the farmers, especially on larger farms, while hand application, practiced on most farms, may lead to high losses of these inputs. Hence, suitable fertilizer applicators and alternative ways to enhance soil fertility, such as by planting green crops, could reduce the amount of chemical fertilizers applied in the study area. The specific energy index shows that about 0.75 MJ energy is applied in the form of fertilizer for each kg of silage corn (Table 4). The outputeinput energy ratio is around five times that of farmers in the Tehran province [58], which indicates that energy efficiency in silage corn production on farms in the Fars province is more efficient than in the Tehran province. Although the silage corn output is lower in the C2 than the C1 group (by 17,000 kg ha1), outputeinput energy ratio, energy productivity and specific energy indices are higher in C2. Nonetheless, the amount of net energy gain is higher in C1. It is worth noting that although different quantities of energy inputs are used in C1 and C2, the share of energy from each input in the total energy input is similar between the groups. This indicates that farmers have similar preferences for input-use in both classes (Fig. 2). All groups use a higher amount of fertilizers, water, seed and diesel fuel energy than other energy inputs. 3.2. Equivalent CO2 emission from silage corn farms The average CO2e emission from silage corn production in the surveyed farms in the study area is 5000 kg ha1, while 1500 kg ha1 more CO2e is emitted from C1 farms (Table 4). It

should be noted that CO2e emission from corn seed was assumed to be equal to water since no related information were found. The highest CO2 is emitted from water (40%), nitrogen (26%), seed (17%) and diesel fuel (9%). The (IGHG)A shows that C2 farmers produce approximately 1.12 kg ha1 higher kg silage corn per kg CO2.. Accordingly, the specific energy index and (IGHG)A reveal that C2 farmers consume and produce lower energy and CO2e, respectively.

3.3. Appropriate energy consumption The result of the CobbeDouglas econometric model shows that six energy inputs have significant effects on silage corn yield (Table 5): diesel fuel, fertilizers, chemicals, water, seed and human power. The high value of R2 of 0.93 for the model indicates high contributions of energy variables to silage corn yield. All the coefficients are between 0 and 1, which shows that they are used in the second stage of production (the economically optimum stage). Accordingly, the inputs are used properly and significant changes in applied values are not necessary. As described before, if the coefficient is negative, it shows that the input is applied in excess. The highest positive elasticity belongs to seed, followed by diesel fuel, by 0.45 and 0.322, which shows that a 10% increase in seed quantity and diesel fuel used for specific cropping operations may increase the yield by 4.5% and 3.22%. The sum of significant elasticity of energy inputs is 1.209, which shows an increasing return to scale. This value implies that a 10% increase in all the significant energy inputs would lead to 12.09% increase in silage corn yield. However, this kind of enhancement in yield may be restricted by photosynthetic/biophysical limitations.

Table 4 Energy and CO2 emission indices.

Energy use

Carbon release

a b

Indices

Average

Maximum

Minimum

Yield (kg/ha) Total energy input Total energy output Direct energya Indirect energyb Output-input energy ratio Energy Productivity (kg/MJ) Specific energy (MJ/kg) Net energy gain (MJ/ha) Total carbon emission (kg/ha) Carbon from direct energy Carbon from indirect energy (IGHG)A

72924.16 54493.27 583393.30 13969.95 40523.32 10.71 1.34 0.75 528,900 5072.52 2658.81 2413.72 14.38

85,200 68598.65 681,600 16194.92 52403.73 11.91 1.49 0.81 613001.4 6066.47 3100.55 2965.93 14.04

67862.16 45589.64 542897.3 12825.95 32763.69 9.94 1.24 0.67 497307.7 4474.81 2405.47 2069.34 15.16

Includes diesel fuel, human power and water. Includes machinery, chemicals, fertilizers and seed.

2172

E. Houshyar et al. / Energy 93 (2015) 2166e2174

Fig. 2. Percentage of energy inputs.

Table 5 The result of CobbeDouglas econometric model based on the Ridge regression model. Econometric model: Ln Yi ¼ a þ a1Ln(X1) þ a2Ln(X2) þ a3Ln(X3) þ a4Ln(X4) þ a5Ln(X5) þ a6Ln(X6) þ a7Ln(X7) þ ei

Total including 194 farmers

a ¼ 3.36 1 e Machinery (X1) 2 e Diesel fuel (X2) 3 e Fertilizers (X3) 4 e Chemicals (X4) 5 e Water (X5) 6 e Seed (X6) 7 e Human Power (X7) Return to scale R2

Coefficient (a)

Probability level

MPP

e e 0.322 0.111 0.106 0.128 0.45 0.092 1.209 0.93

e e 0.2784 0.0000A 0.0000A 0.0000A 0.0363B 0.0000A e e

e e 3.719 0.239 16.507 1.262 10.419 25.736 e e

A and B show significance at 0.01 and 0.05 probability levels, respectively.

Although the elasticity of energy inputs show the corresponding changes on the yield, the sensitivity of yield to each energy input is critical; this helps farmers to focus on the most effective energy inputs.

3.4. The sensitivity of silage corn yield to energy inputs The MPP shows the sensitivity of output to energy inputs. The MPP shows that human power, chemicals and seed energy inputs show the highest sensitivity, namely, of 25.7, 16.5 and 10.4, respectively (Table 5). The sensitivity of 25.7, for instance, displays that 1 MJ ha1 more energy input for water would lead to a yield increase of 25.7 kg ha1 silage corn. Accordingly, these three production factors should be in focus when targeting an increase in energy efficiency in silage corn production. Ref. [26] suggested that the MPP should be strongly considered, alongside the significance of energy elasticity, particularly when the energy resources are limited. A low MPP suggests that the increase (or loss) in yield is not noticeable. The MPP of 0.239 for fertilizers, for instance, indicates that a 1 MJ ha1 higher fertilizer energy input would lead to an additional yield of only 0.239 kg ha1 silage corn, which is negligible. Accordingly, the focus should be on the highest MPPs to obtain the highest yield. Additionally, this focus would be more environmentally friendly, since less carbon would be released. As shown before, the highest carbon was emitted from water and nitrogen fertilizer, while these two production factors have the lowest MPPs. Thus, it is very beneficial for both the farmers (higher yield and income) and the environment (lower carbon release) if MPPs of energy inputs are considered when deciding on the use of production factors.

It is worth noting that no obligatory reductions in energy inputs are necessary, because no negative MPP has been found. 3.5. Different scenarios to use energy According to the results presented in Section 3.1, the average maximum and minimum values of inputs and outputs are very different between the studied farmers' categories. Accordingly, three production scenarios are proposed for further studying the efficiency of energy use in silage corn production as: a: LEI (Low Energy Input) scenario corresponding to minimum energy use; b: MEI (Medium Energy Input) scenario corresponding to average energy use; c: HEI (High Energy Input) scenario corresponding to maximum energy use. In the MEI (Medium Energy Input) scenario, 73 ton ha1 of silage corn are produced out of 55 GJ ha1 of energy used in the form of production inputs. In the MEI scenario, the silage corn yield is 12 ton ha1 lower than in the HEI (High Energy Input) scenario. Compared to the LEI scenario, energy consumption for silage corn production in the MEI scenario is around 10 GJ ha1 higher. Nonetheless, the expected kg yield per GJ energy (energy productivity) is higher in the MEI scenario than in the HEI and LEI scenarios, and the CO2 emissions are lower in the MEI scenario than in the other two scenarios. In the LEI scenario, energy inputs and silage corn yields are lower compared to the MEI and HEI scenarios. However, energy productivity and outputeinput energy ratio are higher in the LEI

E. Houshyar et al. / Energy 93 (2015) 2166e2174

scenario than in the MEI and HEI scenarios. These findings suggest that the LEI scenario represents a cropping system in which the least energy is used, and which is more competitive in areas with environmental concerns, high energy prices and energy scarcity. Nevertheless, it should be noted that silage corn production in the LEI scenario is also reduced by around 7500 kg ha1 compared to the MEI and 17,000 kg ha1 compared to the HEI scenario. Consequently, the cropping system depicted by the LEI scenario may only be adequate for farms that do not depend on high levels of silage corn production per unit of farm area to sustain their animal husbandry activities. Conversely, the HEI scenario would be suggested for situations where higher silage corn production is essential, even with high levels of energy use. Although the highest range of energy input is recorded in the HEI scenario, it is advisable that any increase in energy inputs should take into consideration the econometric model results for preventing an inefficient utilization of energy. Furthermore, this scenario could be even more justified in cases were renewable energy is available abundantly and at low costs, while offering a solution for environmental and economic concerns about non-renewable energy utilization. It is worth noting that the proposed scenarios are derived from the cropping practices currently undertaken on the farms surveyed. However, a limitation of the practical implementation based on the results of the analysis is that it is not clear which scenario is the optimum one. Although the energy indices are higher in the LEI scenario, it cannot be said that this is the optimal cropping system for the farms in the study area. Consequently, it is strongly recommended that the exact factors determining the optimum energy use in each scenario should be investigated in future studies. 4. Conclusions This study determined that around 45,000 to 68,000 MJ ha1 were consumed in the Fars province to produce 67,000 to 85,000 kg ha1 silage corn. The average CO2e emissions were 4400e6000 kg ha1. The four main contributors to 95% of energy inputs were fertilizers, water, diesel fuel for farm operations and water pumping and seed. It is suggested that the use of precision farming fertilizer applicators, green crops (like alfalfa) and conservation tillage practices are expedient in reducing energy consumption and CO2e emissions, due to fertilizers, machinery and fuel use for silage corn production in Fars province. Different regression models, i.e. OLS, PCR and ridge regression, were tested to obtain a suitable econometric model. The econometric model showed that seed and diesel fuel energy inputs had the highest positive elasticity. Despite lower elasticity of human power and chemical energy, the MPP displayed that the yield had the highest positive sensitivity to these energy inputs. We recommend as a strategy for counteracting energy scarcity and energy related environmental concerns optimizing the application of production factors with a high MPP when targeting yield increases, and optimizing the application of production factors with a low MPPs for targeting energy inputs reductions. Three energy input scenarios were proposed based on the average, the highest, and the lowest energy used in the studied farms. The lowest energy input and the highest energy productivity are obtained in the LEI (low energy input) scenario. The result revealed how different scenarios can be performed in different situations of energy availability, which potentially can help to manage economic and environmental concerns. Further, valuable insights may be obtained from studying optimal energy input levels for silage corn production in future studies.

2173

References [1] Hatirli SA, Ozkan B, Fert C. An econometric analysis of energy input-output in Turkish agriculture. Renew Sustain Energy Rev 2005;9:608e23. [2] Pimentel D, Pimentel M. Global environmental resources versus world population growth. Ecol Econ 2006;59:195e8. [3] Fluck RC, Baird CD. Agricultural energetics. West Port, Connecticut: AVI Publishing Company, Inc.; 1980. [4] Dalgaard T, Halberg N, Porter JR. A model for fossil energy use in Danish agriculture used to compare organic and conventional farming. Agric Ecosyst Environ 2001;87:51e65. [5] Chen G, Baillie C. Development of a framework and tool to asses on-farm energy uses of cotton production. Energy Convers Manage 2009;50:1256e63. [6] Bardi U, El Asmar T, Lavacchi A. Turning electricity into food: the role of renewable energy in the future of agriculture. J Clean Prod 2013;53:224e31. [7] Monteny GJ, Groenestein CM, Hilhorst MA. Interactions and coupling between emissions of methane and nitrous oxide from animal husbandry. Nutr Cycl Agroecosyst 2001;60:123e32. [8] Van Lerland EC, Lansink AO. Economics of sustainable energy in agriculture. Kluwer Academic Publisher; 2003. [9] IEA. Energy technology perspectives 2012, pathways to a clean energy system. Available from:. 2012. www.iea.org/textbase/npsum/ETP2012SUM.pdf. [10] Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, et al. IPCC: climate change 2007: the physical science basis. Contribution of Working Group I to the Fourth Assessment, Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press; 2007. p. 996. [11] Barbir F, Ulgiati S. Sustainable energy production and consumption: benefits, strategies and environmental costing. Springer; 2007. [12] Dai D, Hu Z, Pu G, Li H, Wang C. Energy efficiency and potentials of cassava fuel ethanol in Guangxi region of China. Energy Convers Manage 2006;47: 1686e99. [13] Nguyen TLT, Hermansen JE. System expansion for handling co-products in LCA of sugar cane bio-energy systems: GHG consequences of using molasses for ethanol production. Appl Energy 2012;89:254e61. [14] Blancard S, Martin E. Energy efficiency measurement in agriculture with imprecise energy content information. Energy Policy 2014;66:198e208. [15] Verma SR. Energy in production agriculture and food processing. In: Proceedings of the National Conference held at the Punjab Agricultural University, Ludhana; 1987. [16] Canakci M, Topakci M, Akinci I, Ozmerzi A. Energy use pattern of some field crops and vegetable production: case study for Antalya Region, Turkey. Energy Convers Manage 2005;46:655e66. [17] Jørgensen U, Dalgaard T, Kristensen ES. Biomass energy in organic farmingthe potential role of short rotation coppice. Biomass Bioenergy 2005;28: 237e48. [18] Karkacier O, Goktolga ZG. Inputeoutput analysis of energy use in agriculture. Energy Convers Manage 2005;46:1513e21. [19] Chamsing A, Salokhe VM, Singh G. Energy consumption analysis for selected crops in different regions of Thailand. Agric Eng Int CIGR Ejournal 2006;VIII: 1e18. [20] Alluvione F, Moretti B, Sacco D, Grignani C. EUE (energy use efficiency) of cropping systems for a sustainable agriculture. Energy 2011;36(7): 4468e81. [21] Houshyar E, Azadi H, Almassi M, Sheikh Davoodi MJ, Witlox F. Sustainable and efficient energy consumption of corn production in Southwest Iran: combination of multi-fuzzy and DEA modeling. Energy 2012;44:672e81. [22] Wang X, Shen J, Zhang W. Energy evaluation of agricultural sustainability of Northwest China before and after the grain-for-green policy. Energy Policy 2014;67:508e16. [23] Yu W, Jensen HG. China's agricultural policy transition: impacts of recent reforms and future scenarios. J Agric Econ 2010;61(2):343e68. [24] Smith P, Martino D, Cai Z, Gwary D, Janzen H, Kumar P, et al. Policy and technological constraints to implementation of greenhouse gas mitigation options in agriculture. Agric Ecosyst Environ 2007;118(1e4):6e28. [25] Gasparatos A. Resource consumption in Japanese agriculture and its link to food security. Energy Policy 2011;39(3):1101e12. [26] Singh G, Singh S, Singh J. Optimization of energy inputs for wheat crop in Punjab. Energy Convers Manage 2004;45:453e65. [27] Ministry of Jihad-e-Agriculture. Annual agricultural statistics. Department of Agronomy; www.maj.ir/, [accessed January 2013]. [28] Mousavi-Avval SH, Rafiee S, Mohammadi A. Optimization of energy consumption and input costs for apple production in Iran using data envelopment analysis. Energy 2011;36(2):909e16. [29] Chaudhary VP, Gangwar B, Pandey DK. Auditing of energy use and output of different cropping systems in India. Agric Eng Int CIGR Ejournal 2006;VIII: 1e13. [30] Mani I, Kumar P, Panwar JS, Kant K. Variation in energy consumption in production of wheat-maize with varying altitudes in hilly regions of Himachal Pradesh, India. Energy 2007;32:2336e9. [31] Kitani O. CIGR handbook of agricultural engineering. Energy and biomass engineering, vol. 5. St Joseph, MI: ASAE publication; 1999. [32] Shrestha DS. Energy use efficiency indicator for agriculture. 1998. http:// www.usaskca/agriculture/caedac/PDF/mcrae.PDF.

2174

E. Houshyar et al. / Energy 93 (2015) 2166e2174

[33] Ozkan B, Kurklu A, Akcaoz H. An input-output energy analysis in greenhouse vegetable production: a case study for Antalya region of Turkey. Biomass Bioenergy 2003;26:89e95. [34] Yaldiz O, Ozturk H, Zeren Y, Bascetincelik A. Energy usage in production of field crops in Turkey. In: 5th Int. Cong. on Mechanization and energy Use in Agriculture. Kusadasi, Turkey; 11e14 Oct 1993. [35] Dyer JA, Kulshreshtha SN, McConkey BG, Desjardins RL. An assessment of fossil fuel energy use and CO2 emissions from farm field operations using a regional level crop and land use database for Canada. Energy 2010;35(5): 2261e9. [36] Lal R. Carbon emission from farm operations. Environ Int 2004;30:981e90. [37] Coelli TJ, Rao DSP, O'Donnell CJ, Battese G. An introduction to efficiency and productivity analysis. New York: Springer; 2005. [38] Rafiee S, Mousavi Avval SH, Mohammadi A. Modeling and sensitivity analysis of energy inputs for apple production in Iran. Energy 2010;35(8):3301e6. [39] Gujarati DN. Basic econometrics. 4th ed. The McGrawHill Companies; 2004. [40] Rao CR. Linear statistical inference and its applications. New York: John Wiley & Sons; 1965. [41] Ramsey JB. Tests for specification errors in classical linear least squares regression analysis. J R Stat Soc 1969;31(2):350e71. [42] White H. Asymptotic theory for econometricians. Orlando: Academic Press; 1984. [43] Gunst RF. Regression analysis with multicollinear predictor variables: definition, detection and effects. Commun Stat Theor Methods 1983;12:2217e60. [44] Gunst RF, Mason RL. Some considerations in the evaluation of alternative prediction equations. Technometrics 1979;21:55e63. [45] Hastie T, Tibshirani R, Friedman J. The elements of statistical learning: data mining, inference, and prediction. Springer; 2009. [46] Rao CR, Toutenburg H. Linear models: least squares and alternatives. Springer; 1999. 427 p.

[47] Marquardt DW. Generalized inverses, ridge regression, biased linear estimation, and nonlinear estimation. Technometrics 1970;12:591e612. [48] Farrar DE, Glauber RR. Multicollinearity in regression analysis: the problem revisited. Rev Econ Stat 1967;49:92e107. [49] Silvey SD. Multicollinearity and imprecise estimation. J R Stat Soc 1969;31: 539e52. [50] Draper NR, Smith H. Applied regression analysis. 3rd ed. New York: Wiley; 1998. [51] Montgomery DC, Peck EA, Vining GG. Introduction to linear regression analysis. 5th ed. New York: Wiley; 2012. [52] Jolliffe IT. Principal component analysis. New York: Springer-Verlag; 1986. [53] Hoerl AE, Kennard RW. Ridge regression: biased estimation for nonorthogonal problems. Technometrics 1970;12:55e67. [54] Hoerl AE, Kennard RW. Ridge regression: applications to nonorthogonal problems. Technometrics 1970;12:69e82. [55] Seber GAF, Lee AJ. Linear regression analysis. Wiley Publication; 2003. [56] Allen DM. Mean square error of prediction as a criterion for selecting variables. Technometrics 1971;13:469e75. [57] Allen DM. The relationship between variable selection and data augmentation and a method for prediction. Technometrics 1974;16:125e7. [58] Stone M. Cross-validating choice and assessment of statistical predictions (with discussion). J R Stat Soc 1974;36:111e47. [59] Pishgar Komleh SH, Keyhani A, Rafiee Sh, Sefeedpary P. Energy use and economic analysis of corn silage production under three cultivated area levels in Tehran province of Iran. Energy 2011;36:3335e41.  [60] Sarauskis E, Buragiene_ S, Masilionyte_ L, Romaneckas K, Avi zienyte_ D, Sakalauskas A. Energy balance, costs and CO2 analysis of tillage technologies in maize cultivation. Original Research Article Energy 2014;69:227e35. [61] Barut ZB, Ertekin C, Karaagac HA. Tillage effects on energy use for corn silage in Mediterranean Coastal of Turkey. Energy 2011;36(9):5466e75.