According to the predictions, the world's population will rise to 9.5 billion in 2050 (Rahimizadeh et al. 2008). As a result of the increase in world population and increasing demand for food, exploitation of energy resources is growing (Rahimizadeh et al. 2008; Kraatz, 2012). In addition, the estimated world fossil fuel reserve depletion times for oil, gas and coal are 2044, 2046 and 2116, respectively (Shafiee and Topal, 2009). However, the indiscriminate use of fossil energy sources will have lasting adverse effects on the environment (Pishgar-Komleh et al. 2013). However, the agricultural sector to provide adequate food for the growing population, it is dependent on non-renewable energy sources such as fossil fuels (Bakhoda et al. 2012; Nikkhah et al. 2015; Khojastehpour et al. 2015). On the other hand, it should not be ignored that agricultural sector is both a producer and consumer of energy as well as it can increase or reduce the environmental impacts (Liu et al. 2010; Najafi et al. 2011; Royan et al. 2012; Mohammadi et al. 2013; Rajaeifar et al. 2014). The livestock industry is identified as one of the most important sources of protein and food supply for a growing population. The milk production of Iran in 2000 was about 5 million tonnes that up to 2012 reached over 10 million tonnes (FAO, 2012). In recent years, there has been a special attention devoted to the importance of energy consumption in livestock sector. This part has received a relatively high share of energy consumption for raw milk and animal feed preparation processing compared to the other agricultural substructures (O’Mara, 2011; Gerber et al. 2011). There are some studies on the energy analysis of dairy farms. Maysami et al. (2013) analyzed the energy consumption of dairy farms in northern west of Iran and announced that the total energy input was 5800 MJ per ton of milk and energy use efficiency was 2.6. However, they did not consider the modelling and optimization of energy. Sefeed Pari et al. (2012) studied energy indices and greenhouse gas emissions on dairy farms in Tehran province of Iran. They claimed that the total of energy inputs and energy efficiency per ton of milk were 7809 MJ and 1.15, respectively. Divya et al. (2012) explored energy use pattern of dairy farms in India. They showed that the energy efficiency was 0.04 and the total energy inputs were equal to 9189.30 MJ per ton of milk. They claimed that the animal feed and human labor contributed most to energy consumption. The literature review illustrated that many researchers have reported the valuable application of data envelopment analysis ((DEA) approach in energy management of agricultural production (Mohammadi et al. 2013; Mousavi-Avval et al. 2011; Pahlavan et al. 2012). The DEA approach was used to the optimization of energy for milk production in Greece. Energy use efficiency was reported 0.30 MJ per liter. The results of variable returns to scale (VRS) illustrated that from the total of 165 producers, 17 units were purely technically efficient in energy consumption (Theodoridis and Psychoudakis, 2005). In a similar study on the optimization of energy inputs of Australian dairy units, the results showed that from the total of 60 farms, 51 farms were purely technically efficient (Carter, 2000). Other investigations to determine the energy use pattern of dairy farms has been carried out in worldwide (Zucchetto and Bickle, 1984; Stokes et al. 2007; Veysset et al. 2010; Brinker and Laurent, 2011). A literature review showed that there is a high amount of energy in milk production. While researches who evaluated energy pattern in milk production units have reproted a degree of inefficiency, the authors could not find any documented publication on optimization of energy inputs for milk production using DEA approach and sensitivity analysis. Therefore, the aim of this study is to optimize the use of energy inputs for milk production in Guilan province of Iran through integration DEA approach and sensitivity analysis model.
MATERIALS AND METHODS
Studied location and survey
This study was carried out on industrial dairy farms of Guilan during agricultural year of 2012-2013. Guilan is located in the north of Iran on the south of Caspian Sea, within 36˚ 34׳ and 38˚ 27׳ north latitude and 48˚ 53׳ and 50˚ 34׳ east longitude. Guilan province covers an area of 14711 km2, and has a population of about 2.5 million people (Statistical Centre of Iran, 2014). In order to calculate energy optimization and sensitivity analysis, frontier analyst 4.2 and JMP8 software were used. There were 180 industrial farms of cow breeding in Guilan at the time of the study, 129 of them were dairy farms. Due to high prices of inputs such as animal feed, many production units were inactive. This study ranged from 20 to 200 cow heads with the average of 54.5 heads. Hence, 44 active farms were selected that their related information is presented in Table 1. In this study, the inputs included human labor, animal feed, fossil fuels, electricity and machinery and the milk yield served as the output. Energy equivalents are shown in Table 2, were used to estimate the inputs and output energy.
In this study, Cobb–Douglas (CD) model was used to determine the effect of energy inputs on milk yield in Guilan province of Iran(Kuswardhani et al. 2013). The general form of the model is presented in equation 1. Taking the logarithm of both sides of the equation and locating five inputs into equation 1, equation comes in the form of equation 3. In this formula a0 and ei are the constant and coefficient error, respectively. In addition, the returns to scale (RTS) of production was computed. RTS refer to changes in output, subsequent to a proportional change in all inputs (where all inputs increase by a constant factor; Royan et al. 2012).
Lnyi= a0 + αj ln(xij) + ei (i=1, 2,…,n)
Lnyi= a0 + α1lnx1 + α2lnx2 + α3lnx3 + α4lnx4 + α5lnx5 + ei
In order to determine the sensitivity of energy inputs for milk production marginal physical productivity (MPP) approach was used which shows the change value for an increase unit in one of the energy inputs, if the other factors of production assumed to be constant. The MPP is calculated by the equation 4:
MPPxj= ((GM(Y)/GM(Xij)) × αij
MPPxj: marginal physical productivity per unit of Jth inputs.
αij: input regression coefficient.
GM (Y): geometric mean of per ton of milk production.
GM (Xij): geometric mean of energy input (Royan et al. 2012).
Table 1 Characteristics of industrial dairy farms in Guilan province, Iran
Data envelopment analysis (DEA)
To evaluate the efficiency of dairy farms, there are two approaches which can be divided into parametric and non-parametric methods. In the parametric approach, through using statistical and econometric methods, the production function is estimated and then by using this function, efficiency of units is calculated. The second group includes nonparametric methods which do not require the methods of mathematical functions distribution. One of the most important non-parametric models is the DEA approach, which is a linear programming model that measures the relative efficiency of decision making units (DMU) (Charnes et al. 1984). The DEA approach has different models, but the most significant of which is as follows:
Constant returns to scale (CRS)
The CRS approach is the first model of DEA approach proposed by Charnes et al. (1984). In this model, when a unit changes in the inputs, outputs are altered with a fixed ratio (increasing or decreasing), thus this model is called the constant return to scale. For example, if the inputs are doubled, the outputs are doubled consequently. Provided that the outputs have a more than double increase and less than double increase, it is assumed that their efficiency would be increased and decreased, respectively. The slope of the production function is constant in this model (Charnes et al. 1984). The details of the model presented in the equation 5.
Variable return to scale (VRS)
This model was used when the same scaling, which was more and less than the maximum amount of each observed inputs and outputs, should not be plausible. In this model, with changing a unit in the inputs; the outputs change with different ratios. These changes can be increasing or decreasing. In this model, the slope of the production function varies (Banker et al. 1984). The details of the model presented in the equation 6.
> ε, Ur > ε free
Ep: efficiency rate of the ith.
Ur: input weight.
Yrp,: output value rth for DMUp.
Vi: output weight.
Xip: input value ith for DMUp.
Yrj: output value rth for DMUj.
Xij: Input value i th for DMUj.
j= 1, 2,…,n.
w: free indication.
Table 2 Energy equivalents of inputs and output in industrial milk production in Guilan province, Iran
DM: dry matter.
a: Economic life of machine (year).
The variety of energy efficiency with DEA approach
Three efficiency types are calculated by using data envelopment namely: TE, PTE and SE. To calculate the TE; the CRS model was used and also to calculate PTE; the VRS model was employed. As shown in equation 7, the SE is obtained by dividing TE by PTE.
Scale efficiency (SE)= TE / PTE
In addition, the actual energy consumption and the optimal energy were obtained using the DEA approach. The two before-mentioned factors were employed to obtain capabilities of saving energy in the industrial dairies in Guilan province. Hence, the percentage of saving energy for each making decision unit (j) was evaluated using the DEA approach in equation 8 (Hu and Kao, 2007).
ESTRj (%)= (EST/AEI) × 100
j: decision making unit.
AEI: actual energy consumption.
EST: saving energy.
The EST was obtained from the difference between the actual and optimal energy per unit.
RESULTS AND DISCUSSION
The inputs-output energy analysis
The detailed description of energy flow for milk production in Guilan province of Iran is illustrated in Table 3. The total animal feed energy was found to be 6180.08 MJ per ton of milk. The share of animal feed energy use in total energy inputs was 82%. It indicated that animal feed energy had the most significant contribution in milk production between all inputs. In most studies, for example dairy farms in Tehran, north western provinces of Iran and India, this input has been the most consumed input in terms of energy consumption (Divya et al. 2012; Maysami et al. 2013). After the animal feed input, fossil fuels equal to 947.87 MJ per ton of milk were recognized i.e. the second source with high energy consumption, therefore that this input accounted for approximately 13 percent of total energy consumption of milk production in Guilan province of Iran. The value of fuel energy consumption for milk production in Tehranprovince of Iran was reported 1272.91 MJ. In this case, the amount of fuel energy consumption for one ton production of milk was less than that of milk production in Tehran province, Iran. The reason was that the dairy farms in Guilan used more vacuum pump dairy with electricity energy instead of fossil fuels (diesel and natural gas). More fossil fuels such as diesel fuel consumed by tractors to produce forage in farms and also to carry out processing of animal diet, crushing and mixing of all kinds of forage. The electricity input was the third source of energy consumption in Guilan dairy farms, with a share of 4% of the total energy consumption and 298.19 MJ per ton of milk which was calculated. The amount of electricity energy for milk production in Tehranprovince of Iran was reported as 249.88 MJ per ton of milk. In the present study, the most of electricity consumption was allocated to milking systems, heating water for washing milk pipes, milk cooling and followed by lighting units in milking parlors as the second factor.
Table 3 The total of inputs and output energy consumption per ton of milk produced in Guilan dairy farms
In another similar study in Ireland, the input of electricity with share of 60% of the total energy consumption had the highest contribution to milking machines and cooling systems units. They also claimed that in order to save electricity in Ireland, especially in the case of milking, the use of milking vacuum pumps equipped with variable speed drive (VSD) was proposed to reduce energy costs to 50%. In addition, to manage the electricity consumption of dairies in the milk cooling units, heat exchanger which acts as a cooling system was proposed. Moreover, a suitable solution in this region to reduce the energy consumption for the heating water; the heat returning milk from cooling systems were suggested (Upton et al. 2010; Upton et al. 2013). Machinery and human labor inputs in the production of milk in Guilan province of Iran had the lowest energy consumption as well.
The results ofthesensitivityanalysis
The results of the CD model to determine the effect of energy inputs on milk yield are shown in Table 4. The effect of energy inputs includes human labor, fossil fuels and animal feed on milk yield were positive and the effect of the electricity and machinery on milk yield was negative. The maximum of regression coefficient (0.82) between the inputs of energy allocated to animal feed was more than that of the other inputs. In addition, the effect of this input on milk yield was significant at 5 % level (P-value<0.001). The second input affecting milk yield was the fossil fuels (diesel and natural gas) with regression coefficient equal to 0.18. Human labor was the third input affecting the milk yield (Table 4). The results of the sensitivity analysis showed that with the increase of one MJ of energy inputs including human labor, fossil fuels and animal feed, yield increases 3.14, 1.36 and 0.95 kg respectively and with an increase of one MJ of electricity and machinery equal to -1.03 and -7.16 kg respectively, yield for per ton of milk production was decreased.
In this study, the CRS and VRS models being input-oriented were applied for evaluating TE, PTE and SE of milk production in Guilan province of Iran. Table 5 shows the types of efficiency and the returns to scale for each industrial milk production farms. On the other hand, Table 5 is a supplement of Table 6 that the averages of TE, PTE and SE on Guilan dairy farms were computed as 44.6, 74.48 and 0.53, respectively. In a similar study, Uzmay et al. (2009) showed that the average of TE and PTE were 0.52 and 0.62 respectively, for milk production in Turkey. In a study carried out in Australia, the results showed that the TE, PTE were 0.95 and 0.94, respectively, that indicated a good management in energy inputs of dairy farms of this country, especially with using the principles of the diet formulation in animal feed (Carter, 2000). Also, in this study the “returns to scale” factor was determined for CRS and VRS models (Table 5). Determination of this factor is obtained by output weight. If it is less than zero, the “returns to scale” factor will be increasing. If it is more than zero, the “returns to scale” factor will be decreasing. If equal to zero, the output will be with no change. Increasing returns to scale (IRS) can not reduce the production scale unit, but it can be increased to infinity. The ratio of output to input for each point on the efficient frontier in terms of inputs will be decreasing. In fact, increase in outputs is almost equal to changes in inputs (Ghojabeige et al. 2009). The results of the models (VRS and CRS) are shown in Figure 1. Based on the CRS model, only 7 farms out of 44 industrial farms were efficient, whereas the rest was on the inefficient frontier. The farms of 1, 2, 4, 6, 16, 21 and 25 were fully functional, which means that these farms in both scale model were efficient. In a study conducted in Australia with regard to the appropriate functionality in the energy efficiency of milk production of dairy about 51 farms, out of 60 farms were efficient frontiers (Carter, 2000). The results of optimization of energy inputs for dairy farms in Greece revealed that from the total of 167 dairy farms considered for the analysis, 17 farms were purely technically efficient in energy consumption from the different sources (Theodoridis and Psychoudakis, 2005).
Table 4 Estimating the effect of energy inputs on milk yield
MPP: marginal physical productivity.
Table 5 Technical, pure technical, scale efficiency and returns to scale for the production of ton of milk based on CRS and VRS input oriented model
SE: scale efficiency; TE: technical efficiency; PTE: pure technical efficiency; DMU: decision making units; IRS: increase return to scale; CRS: constant returns to scale and DRS: decrease return to scale.
Table 6 Average and standard deviation for variety of efficiency with CRS and VRS in Guilan province, Iran
SE: scale efficiency; TE: technical efficiency and PTE: pure technical efficiency.
Besides, in Pennsylvania state the energy efficiency was examined using a DEA approach that about 30% of the dairy farms were efficient (Stokes et al. 2007(. The concept of actual energy is the energy that is actually used in the industrial dairy farms and also the optimum energy is determined using the DEA approach or with covering the data and comparing them with the efficient frontier recommen dation for the dairy managers to improve the efficiency in decision making farms. The average of the actual and of the optimal energy and of the saving energy for inefficient farms with the VRS model was equal to 7853.16, 4000.77 and 3852.39 MJ, respectively, and also the saving energy rate was acquired as much as 47.74% for per ton of milk production (Figure 2).
Figure 1 The number of functional units based on CRS and VRS model
Figure 2 Average of total actual, optimal and saving energy for inefficient units with VRS model
Figure 3 Average of total actual, optimal and saving energy for each of the inputs with VRS model
Figure 4 Distribution of saving energy from different inputs per ton of milk in Guilan province
Furthermore, the average amount of actual energy for each input, including animal feed, fossil fuels, electricity, machinery, human labor, were equal to 6180.08, 953.96, 298.19, 42.07 and 68.55, respectively. The average amount of optimal energy for each input was equal to 3937.43, 70.03, 56.53, 3.59 and 38.86 MJ per ton of milk, respectively, so that the amount of saving energy based on equation 8 from the difference between actual energy and optimal energy can be achieved (Figure 3). As seen in the Figure 4, the total of saving energy in Guilan industrial dairies was achieved as 3852.39 MJ per ton of milk production. The most energy consumption for saving was allocated to the animal feed input (65%). One of the fundamental problems in Guilan dairy farms was the irrational consumption and the incorrect diet program for supplying the food for animals. A regular diet program turned to improve stability and increase energy efficiency in milk production. In addition to employing the experts in diet formulation, one of the technologies currently used in the management of diet is the automatic feeding stations, which the animal feeding rate is variable based on the animal weight and age (Fujiwara and Rushen, 2014). Following by animal feed input in terms of possibility of saving energy, fossil fuels and electricity inputs were with a share of 26% and 7% respectively. In order to optimize the energy consumption of these inputs; using a heat exchanger to milk cooling unit, vacuum pumps with feature of VSD and using heat milk in the milk cooling for heating water are appropriate methods (Ubbels and Bouman, 1979; Upton et al. 2010; Upton et al. 2013). On the other hand, one of the methods of replacing the fossil fuels is the usage of renewable energy systems such as biogas technology in large numbers of dairy farms (Cornejo and Wilkie, 2010; Hosseini et al. 2013). But, accepting and employing this technology requires further studies in the dairy industrial farms of Guilan province.
We investigated the energy consumption of milk production in Guilan province of Iran. The results showed that the highest share of energy consumption belonged to the animal feed followed by fossil fuels. The result of DEA approach based on VRS model indicated that about 47.74% of total energy input could be saved without reducing the milk yield through following recommendations resulted from this study. Moreover, sensitivity analysis results showed that with one MJ extra use of human labor, fossil fuels and animal feed energy, it would lead to an additional increase in milk yield. Overall, to optimize the energy consumption of milk production in Guilan province of Iran, new optimized solutions such as technology of feed management, employing the experts in diet formulation and using a heat exchanger to milk cooling units are suggested.
The research leading to these results has received funding from the Ferdowsi University of Mashhad, Iran under grant agreement number 30927, is gratefully acknowledged.