1. Introduction
Energy consumption estimation in agriculture has been an essential tool in determining sustainable farming practices. The upsurge energy prices, strict environmental laws along with end-use energy policies increase the need for minimal and efficient energy consumption [
1,
2]. Energy use efficiency is seen as an important condition for sustainability of farming systems with the potential of financial savings, preservation of natural resources along with reduction in environmental impacts. It has been suggested that cost-efficient ways to save energy and related emissions can decrease one third of the global energy demand by 2050 [
3,
4,
5]. Globally increasing productivity and profitability ratios are the key concerns for farming systems and both depends on the magnitude of energy consumption. The energy used in agriculture including dairy farming systems depends on the amount of agricultural work performed, the land area used and the level of farm mechanization [
6,
7,
8].
Energy is a critical input and significant cost for dairy farming systems. Energy consumed in dairy systems can be classified into direct and indirect energy inputs. These energy inputs accounts for substantial direct and indirect fossil energy consumption, which produces carbon dioxide (CO
2) emissions on-farm and off-farm [
7,
9,
10]. Moreover, the development of energy efficient farming systems helps in reducing greenhouse gas emissions (such as CO
2) besides providing financial benefits to farmers [
11,
12]. To minimize the greenhouse gas (GHG) emissions requires a reduction in farm energy inputs (fossil fuels, fertilizer, etc.). This goal can be achieved in two ways: either through achieving a substantial increase in energy efficiency where the same output is produced with less energy input, or through using more sustainable energy sources such as solar, wind, biomass, etc. [
13]. The dairy industry is one of the most influential agricultural sectors of New Zealand’s economy and is responsible for 22.5% of NZ total greenhouse gas emissions. Recently, the New Zealand government approved a “Zero Carbon Bill” in response to their Paris Accord commitments, which sets new emission reduction targets for all industries including the dairy sector to reduce emissions (such as CO
2, N
2O) to net zero by 2050 [
14]. Under this situation, reducing the greenhouse gas emissions from NZ dairy farming systems has become a critical challenge for NZ dairy industry. Hence, it is necessary for NZ dairy farming systems to consider their energy expenditure and improve energy use efficiency for reduction in energy consumption and associated environmental emissions.
To estimate the efficiency of agricultural production systems, several parametric and non-parametric methods has been employed by researchers. For instance, a parametric technique the stochastic frontier production function (SFPF) employed for efficiency evaluation of crop production in Nigeria [
15]. A meta-regression analysis was applied in another study for efficiency evaluation of Spanish and English dairy farms [
16]. In New Zealand, stochastic frontier analysis (SFA) was employed to determine the efficiency of NZ dairy farms [
17,
18,
19]. Conversely, data envelopment analysis (non-parametric technique) based on mathematical programming which determines relative efficiency of a number of decision making units (DMUs) [
20]. Its application in agricultural systems has been recommended by many researchers, as it does not need any prior assumptions for the fundamental functional form among inputs and outputs [
21,
22,
23,
24,
25]. DEA allows to contemplate multiple inputs and outputs simultaneously, where each DMU efficiency is compared to that of an ideally efficient operating unit instead of average performer unit. Thus, enabling researchers to distinguish efficient DMUs from inefficient ones and detect the amount and sources of inefficiency for each inefficient DMU [
26]. For instance, Nassiri and Singh [
27] estimated efficiency of paddy crop farms in India through data envelopment analysis (DEA) approach. In Canada, Cloutier and Rowley [
28] compared efficiencies of 187 dairy farms between 1988–1989 and found larger farms were more efficient than the smaller ones. Barnes and Oglethorpe [
29] determined Scotland dairy farms technical, cost and scale efficiency and found low technical, cost and scale efficiencies, and thus recommended changes in farm size or scale. Jaforullah and Whiteman [
30] applied DEA on NZ dairy farms and found average scale efficiency around 94% with majority of dairy farms operating below the optimal scale. Based on same data set, further Jaforullah and Premachandra [
31] recognized that the technical efficiency of each dairy farm was sensitive to production frontier (such as SFA and DEA) selection. In another study, Wei [
32] determined the technical efficiency of NZ dairy farms through combined application of DEA and stochastic frontier analysis (SFA) for the season 2006–2007 and found average technical efficiency around 96% in SFA and 82% and 86% in DEA under constant and variable return to scale models, respectively.
Worldwide, several studies have evaluated energy efficiency of dairy farming systems. For example, in Konya Turkey Uzal [
8] compared energy efficiency of dairy farming systems with different housing structures (freestall, loose housing) and found the freestall dairy system to be more efficient. Meul, Nevens [
12] evaluated the changes in energy consumption efficiency of Flanders dairy farms and observed decreasing trend in energy use efficiency over the considered time frame due to increasing energy productivity. Sefeedpari [
33] applied the DEA technique to calculate the energy efficiency of Iranian dairy farms and found 51% of farmers efficiently using their energy inputs. Likewise, another Iranian study applied the DEA to determine the energy efficiency and energy saving targets for dairy farms and recognized feed intake and fossil fuels among the leading energy saving inputs [
22]. However, from a New Zealand perspective, several researchers have estimated energy consumption of pastoral dairy systems [
9,
34,
35,
36,
37], but very little consideration was given to energy efficiency except Wells [
9] and Podstolski [
36] who determined the overall energy ratio (OER) for NZ pastoral (PDFs) dairy system as an energy efficiency indicator. (Overall Energy Ratio (OER): is the ratio of total energy input to the total energy output of the product. This is inverse of energy efficiency and used as energy efficiency indicator).
The New Zealand dairy industry is renowned for its low input pastoral dairy farming system (PDFs). However, the intensification of this pastoral dairy system during the previous decades, as well as rising sustainability concerns due to the challenges of nutrient leaching and greenhouse gas emissions put NZ dairy systems under high scrutiny. One response to these challenges has been the introduction of the barn dairy system (BDFs) (also known as hybrid dairy system) into New Zealand, in which animal shelter (the barn facilities) is used in combination with pasture grazing for the purposes of reducing soil damage, animal lameness and environmental impacts [
10,
38]. Barn facility usage intensifies the system due to higher stocking rates and subsequently more energy consumption, to maintain and achieve financial and environmental benefits simultaneously [
39]. Under this situation, energy efficiency evaluation of contrasting dairy systems (PDFs versus BDFs) would be helpful to understand energy efficiency profile of NZ different dairy farming systems.
Therefore, the aim of this study was to evaluate energy efficiency of pastoral (PDFs) and barn (BDFs) dairy farming systems through application of data envelopment analysis approach. Further, benchmarking was performed to separate the efficient and inefficient dairy farms, and optimal energy consumption was determined for inefficient dairy farms in order to identify energy saving potential from different energy sources.
2. Materials and Methods
2.1. Data Collection and Processing
This study was carried out in the Canterbury province of New Zealand. In this study, 50 dairy farms were selected from Canterbury including 43 pastoral (PDFs) and 7 barn farms (BDFs). The primary data for the season 2016–2017 were collected from these dairy farmers through a survey questionnaire and face-to-face interview method. The questionnaire was developed to collect the information about various inputs including diesel, petrol, electricity, fertilizer, labour working hours, time usage of machinery, etc. This study only considered cradle-to-farm gate energy inputs that were used to produce milk up to the farm gate i.e., transport and post-processing components were not considered.
Each input recorded in the questionnaire was then converted into an energy equivalent by using their appropriate energy equivalent factors.
Table 1 shows the values of energy equivalents for inputs used in both PDFs and BDFs dairy systems. In this study, energy inputs comprised of fossil fuels, electricity, human labour, feed, fertilizer and machinery, while milk product was taken as output energy. The total energy consumption estimated was the sum of all the input multiplied with their suitable energy conversion coefficient [
40].
Energy inputs can be classified as direct and indirect inputs [
9]. In this study, direct energy encompassed diesel, petrol, electricity, human labour, while indirect energy involved fertilizer, imported feed supplements and machinery used in the dairy farming operations. In addition to energy efficiency of both dairy systems, energy indicators such as energy productivity (EP) and overall energy ratio (OER) were also determined through Equations (1) and (2) [
8,
12,
36,
44]:
where, ‘EP’ is energy productivity (tMS MJ
−1), ‘OER’ is the overall energy ratio “the ratio of total energy input to the total energy output of the product”. OER describes an inverse of energy efficiency, a higher OER means lower efficiency and vice versa.
2.2. Data Envelopment Analysis Approach
The data envelopment analysis (DEA) is a technique used for the assessment of non-parametric efficiency frontiers in multi-factor production analysis. DEA uses linear programming to form a non-parametric frontier above the data set, which serves as relative benchmark for evaluation of efficiency among other homogenous decision-making units (DMUs) under analysis [
45,
46]. Data envelopment analysis allows each DMU to select any combination of inputs and outputs to maximize its relative efficiency. The relative efficiency score of a decision-making unit (DMUs) is defined as a ratio of weighted sum of outputs to weighted sum of inputs. This relative efficiency score is a non-negative value based on the linear relationship between inputs and outputs [
47]. Assume ‘
n’ DMUs are to be assessed, each using different combination of ‘
’ outputs and ‘
’ inputs. The objective function of DMU ‘
d’ in the set of ‘
j’ DMUs (
j = 1, 2, 3, ...,
n) can be written as Equation (3):
Subject to 1, for j = 1, 2, 3, ..., n
ur and vs ≥ 0, r = 1, 2, 3, .…., p and s = 1, 2, 3, ..., q.
Whereas ‘y
rd’ is output amount (
r) produced by DMU ‘
d’, ‘x
sd’ is input amount (
s) consumed by DMU ‘
d’, ‘y
rj’ is output amount (
r) produced by DMU ‘
j’, ‘x
sj’ is input amount (
s) consumed by DMU ‘
j’ and ‘u
r’ and ‘v
s’ are the weight given to individual output and input [
48].
The two models CCR and BCC named after the authors Charnes, Cooper [
49] and Banker, Charnes [
50], respectively, are commonly used in DEA technique based on return to scale parameter. Charnes, Cooper [
49] introduced the CCR model based on the assumption of constant return to scale (CRS), which implies that an input increase will result in a proportional output increase. In CCR model, the efficiency frontier is a straight line which intersects the origin point and best performing unit(s) as shown in
Figure 1. The best performing unit is the one with the highest output to input ratio, in
Figure 1 this is P
2. This point thus serves as a reference DMU to all other units under investigation. The CCR model, allows the identification of inefficient DMUs with consideration of scale size. In CCR models, both technical and scales efficiencies are present, which are based on input/output arrangement (management techniques) and scale size. The efficiency measured under the CRS assumption named as technical efficiency.
Banker, Charnes [
50] presented the BCC model based on the assumption of variable returns to scale (VRS), which implies that an input increase will result in a non-proportional output increase. In BCC model, the efficiency frontier changed from a straight line to a convex structure. This convex combination of the efficient DMUs serves as reference point for other inefficient units. In
Figure 1, the BCC model shows more than one efficient DMU on the frontier line (P1, P2, P4, P5) using the same DMUs as in the CCR model. The BCC model has few advantages over the CCR model. The BCC model frontier envelops more data so more efficient units than CCR and the efficiency scores of BCC model are higher or equal to those of CCR as it connects the outer most DMUs (including the one determined efficient by CCR). Due to the presence of more than one efficient DMUs in the model, the inefficient units under BCC model get the opportunity to be compared with more appropriate efficient units [
47].
The pure technical efficiency (PTE) is defined as the technical efficiency of DMUs measured under variable return to scale assumption. The BCC model, also known as VRS model, gives the pure technical efficiency of DMUs without consideration of scale size. In simple words, the CCR model efficiency is the combination of technical efficiency (TE) and scale efficiency (SE), while the BCC model separates the TE and SE and measures pure technical efficiency (PTE).
Scale efficiency (SE) captures the effect of scale size on the efficiency of DMU and indicates that some portion of inefficiency belongs to the inappropriate size of a DMU. The efficiency score variation between CRS and VRS models is captured in scale efficiency. The relationship between technical (TE) ,pure technical efficiency (PTE) and scale efficiency (SE) can be explained as follows [
23,
27]:
In DEA application, the efficiency of a unit can be attained either by input or output orientation. In input orientation models, efficiency is attained by minimizing input usage while maintaining same output levels, whereas output orientation models focus on increasing output levels while maintaining same level of inputs. Here, an input-oriented DEA approach was adopted for the efficiency measurement of dairy farms. This orientation is considered more suitable for agriculture as farmers have more control over input usage compared to output, which are often influenced by exogenous factors (rain, soil structure, climate, etc.). Likewise, this orientation choice is in accordance with current situation of New Zealand dairy farming systems, where more focus is on efficient input usage (due to environmental issues) rather than production or yield increase. In this study, the decision-making units (DMUs) are the dairy farms (PDFs and BDFs), while direct and indirect farm inputs were considered as energy inputs in mega joule per hectare (MJha−1) and milk energy per hectare (MJha−1) was considered as the output energy for the individual DMU or dairy farm.
To measure the efficiencies of selected DMUs (dairy farms) based on CCR and BCC models, the Data Envelopment Analysis Program (DEAP) software version 2.1 was employed [
45,
51]. The focus was to determine the optimal energy input efficiency with consideration of the input/output management and scale size of different dairy farming systems (PDFs and BDFs), so further analysis was based on the CCR model.
The DEA divides the DMUs (dairy farms) into efficient and inefficient sets; the inefficient DMUs are ranked on their efficiency scores; while DEA lacks distinction between efficient DMUs. Thus, to rank efficient DMUs, a benchmarking method was employed, an efficient unit is ranked higher if chosen as relative peer by many inefficient DMUs, and frequently appears in the reference set.
4. Conclusions
Dairy farming systems with better energy efficiency would help to reduce energy costs and environmental footprints along with improving productivity and profitability of farming systems. The main purpose of this study was to evaluate energy efficiency of NZ contrasting dairy systems such as pastoral (PDFs) and barn (BDFs) and finding their optimal energy requirements through data envelopment analysis (DEA) approach. The average technical, pure technical and scale efficiencies of pastoral and barn dairy systems were found as 0.84, 0.90, 0.93 and 0.78, 0.84, 0.92, respectively, indicating that energy efficiency is slightly better in PDFs compared to BDFs system. Based on CCR and BCC models, 20 and 24 dairy farms respectively out of 50 selected farms were efficient, indicating that the majority of farms were not technically efficient due to using more energy inputs than required. The inefficient farmers need to pay attention towards their energy inputs (electricity, fertilizer and imported feed supplements), as they showed higher potential for energy savings. From systems perspective, when comparing actual and optimal energy use of pastoral (PDFs) and barn (BDFs) dairy systems, results shows that 23% and 35% energy can be saved in both dairy systems respectively, with optimal energy consumption. Thus, for energy efficiency improvement in both dairy systems, energy auditing and the use of more renewable energy sources along with application of precision agricultural technology were recommended.