You are on page 1of 6

Energy 47 (2012) 465e470

Contents lists available at SciVerse ScienceDirect

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

Extending the inputeoutput energy balance methodology in agriculture through


cluster analysis
Carlos Ricardo Bojacá a, *, Héctor Albeiro Casilimas b, Rodrigo Gil a, Eddie Schrevens c
a
Departamento de Ciencias Básicas, Facultad de Ciencias Naturales e Ingeniería, Universidad Jorge Tadeo Lozano, Bogotá, Colombia
b
Centro de Bio-Sistemas, Facultad de Ciencias Naturales e Ingeniería, Universidad Jorge Tadeo Lozano, Chía, Colombia
c
Department of Biosystems, Faculty of Bioscience Engineering, Katholieke Universiteit Leuven, Leuven, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: The inputeoutput balance methodology has been applied to characterize the energy balance of agri-
Received 2 May 2012 cultural systems. This study proposes to extend this methodology with the inclusion of multivariate
Received in revised form analysis to reveal particular patterns in the energy use of a system. The objective was to demonstrate the
31 August 2012
usefulness of multivariate exploratory techniques to analyze the variability found in a farming system
Accepted 22 September 2012
Available online 18 October 2012
and, establish efficiency categories that can be used to improve the energy balance of the system. To this
purpose an inputeoutput analysis was applied to the major greenhouse tomato production area in
Colombia. Individual energy profiles were built and the k-means clustering method was applied to the
Keywords:
Energy use
production factors. On average, the production system in the study zone consumes 141.8 GJ ha1 to
Regional analysis produce 96.4 GJ ha1, resulting in an energy efficiency of 0.68. With the k-means clustering analysis,
Greenhouse tomato three clusters of farmers were identified with energy efficiencies of 0.54, 0.67 and 0.78. The most energy
k-means clustering efficient cluster grouped 56.3% of the farmers. It is possible to optimize the production system by
Colombia improving the management practices of those with the lowest energy use efficiencies. Multivariate
analysis techniques demonstrated to be a complementary pathway to improve the energy efficiency of
a system.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction decades [3]. The energy used in agriculture concerns above all the
non-renewable energy resources and implies effects on the social,
The objective of this paper is to demonstrate the use of multi- economic and environmental dimensions of sustainability [4].
variate exploratory techniques to analyze the intrinsic variability of A widely applied method to assess the energy efficiency of
a farming system and by doing so, establish an energy efficiency agricultural systems is the energy inputeoutput analysis [5]. In this
classification that can be used to optimize the system from within. method, all agricultural inputs and production are multiplied by
To this purpose an inputeoutput analysis was applied to farmers conversion factors to approximate input and output energy [3]. The
located in one of the major greenhouse tomato production areas in method has been applied at global [3], national [6e8] and regional
Colombia. This case study also seeks to contribute in the analysis of [9e13] scales to analyze the energy use efficiency of several agri-
energy efficiency of agricultural systems in developing countries, cultural products. Through comparative studies, it has also been
where still exist a dearth of research in this subject [1]. used to assess the efficiency of different production systems such as
The agricultural sector has stepped up its energy consumption conventional vs organic [14e16] and open field vs greenhouse [17e
due to the increased use of external resources such as machinery, 19]. Once the inputs and outputs are transformed into energy units,
fertilizers and pesticides [2]. Globally, this increase in energy use indicators such as energy use efficiency, energy productivity,
has contributed to higher yields. While in developing countries the specific energy and net energy can be derived [20].
energy requirements per food calorie continue to rise, energy In regional studies, data collection for most of these inpute
intensities in developed countries decreased over the last three output analyses is done through personal interviews applied to
a sample of farmers [9e15,18,20]. How this sample was determined
is not mentioned in some cases [9,14,15] while others applied
* Corresponding author. Calle 22 No. 3-30, Módulo 15, Mezzanine, Bogotá,
different sample size calculations [10e13,18,20]. The factors of
Colombia. Tel.: þ57 1 8650218; fax: þ57 1 8650239. interest are climate, water availability, soil characteristics, and
E-mail address: carlos.bojaca@utadeo.edu.co (C.R. Bojacá). management practices all of which are highly variable and affect

0360-5442/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.energy.2012.09.051
466 C.R. Bojacá et al. / Energy 47 (2012) 465e470

the energy efficiency of the system [1]. In order to work with the energy use efficiency (EUE) and net energy (NE) indexes [20],
wide variability and obtain useful results, these factors are according to the following formulas:
assembled in a multivariate dataset, where an n  p matrix is 
composed by n observations (farmers) and p variables (agricultural Energy output MJ ha1
EUE ¼  (1)
inputs and crop yield). Energy input MJ ha1
Multivariate methods have been applied in various fields of
agricultural research, including plant community studies, house-    
hold assessment and regional studies [21]. However, the applica- NE MJ ha1 ¼ Energy output MJ ha1
tion of these methods to analyze the sustainability or the  
environmental impact of farming systems has been fairly limited  Energy input MJ ha1 (2)
[21,22]. Despite the power of these methods and the nature of the
data, they have not been used to analyze the energy efficiency of
a farming system. 2.2. k-means clustering
Within an agricultural region, many physical, chemical and
biological properties directly related to the production system Within the multivariate analysis methods, the k-means algo-
exhibit spatial variability, even at small distances [23]. Such varia- rithm can classify farmers in terms of their energy use. It is used to
tion, linked to the temporal variation, implies that different levels separate n individuals into k groups such that the sum of squares
of input factors will result in varying yields [9]. Moreover, in from the observations to the assigned cluster center is minimized
developing countries the limited access to technology, unstan- [31,32].
dardized management practices and restricted knowledge of the In order to implement k-means clustering for the energy anal-
production system increase the range of the productivity levels ysis in the study zone, we modified the initial procedure and the
within a farming system. The variability of a farming system can be energy use assessment was carried out on an individual basis
exploited to characterize farmers in terms of their energy efficiency. instead of an average basis. In this way, the energy profile for each
Such characterization can indicate pathways to optimize the energy farmer was calculated and by doing so, a 126 (growers)  9
efficiency of the system as a whole. (production factors as listed in Table 3) matrix configured a multi-
variate dataset. The variables in this dataset included the input
factors and the output as well. Prior to the application of the k-
2. Methods of analysis means clustering analysis it was necessary to standardize the
dataset due to the high variances among variables. Following the
2.1. Energy use assessment procedure suggested by [31], the range was used to standardize
each variable. Range standardization consisted in dividing each
In this work, we followed the process analysis method [24] variable by its range.
where material and energy flows are considered to trace all The next step for the k-means clustering analysis was the defi-
inputs and outputs into the production system [25]. Initially, the nition of the number of clusters to group the growers [33]. The
energy inputeoutput analysis was carried out following the stan- CalinskieHarabasz criterion [34], which returns its maximum value
dard approach where the production inputs and yield were aver-
aged over the entire dataset. Afterwards, average inputs and output
were transformed into energy units according to the energy Table 2
Average (Standard deviation) inputs and output amounts of the greenhouse
equivalents presented in Table 1. The energy content of the crop
tomato system components in the Alto Ricaurte province, Colombia.
residues retained on the field was not considered. Energy use
assessment for the farming system was estimated through the Energy source Quantity (unit ha1) Energy (MJ ha1) % of total
energy
input
Table 1 Human labor 10794.3  5343.7 21156.8  10473.7 14.92
Energy equivalent factors used to transform the inputs and the output yield of the Machinery
greenhouse tomato production system in the Alto Ricaurte province, Colombia. Tractor (h) 5.0  4.3 331.4  289 0.23
Irrigation pump (h) 1.3  1.2 87.9  78.7 0.06
Energy source Unit Energy equivalent Reference Diesel oil (l) 861.2  601.4 39098.1  27304.3 27.57
(MJ unit1) Infrastructure (kg)
Human labor h 1.96 [26] Steel 135.1  16.9 4458.8  558.1 3.14
Diesel oil l 45.4 [27] Wood 1615.4  266.4 1938.5  319.7 1.37
Infrastructure kg Polyethylene 660.0  74.4 6533.5  736.4 4.61
Steel 33 [28] Synthetic fiber 307.0  243.1 368.4  291.7 0.26
Wood 1.2 [28] PVC 151.8  40.1 1760.9  465.6 1.24
Polyethylene 9.9 [28] Fertilizers (kg)
Synthetic fiber 1.2 [28] N 423.1  201.8 25641.6  12228.7 18.08
PVC 11.6 P2O5 434.3  254.7 4821.1  2827.7 3.40
Fertilizers kg K2O 605.5  350.6 4056.6  2349.1 2.86
N 60.6 [29] Organic 9973.6  9425.4 2992.1  2827.6 2.11
P2O5 11.1 [29] fertilizers (kg)
K2O 6.7 [29] Pesticides (kg)
Organic fertilizers kg 0.3 [29] Fungicides 45.0  38.3 9710.1  8280.9 6.85
Pesticides kg Insecticides 31.3  31.1 3168.9  3151.3 2.23
Fungicides 216 [18] Plant material
Insecticides 101.2 [18] Seedlings (unit) 29917.1  5147.6 5983.4  1029.5 4.22
Plant material Water for 15373.2  10357.3 9685.1  6525.1 6.83
Seedlings unit 0.2 [9] irrigation (m3)
Water for irrigation m3 0.63 [30] Total energy input 141793.3  43665.7 100
Output kg Output
Tomato fruits 0.8 [13] Yield (kg ha1) 120530.8  49585.0 96424.6  39668.0
C.R. Bojacá et al. / Energy 47 (2012) 465e470 467

Table 3 In a person-to-person interview format, undergraduate intern-


Energy use assessment of the three clusters of growers found through the appli- ship students conducted the survey. Prior to fieldwork, the students
cation of the k-means classification method for the greenhouse tomato production
system in the Alto Ricaurte province, Colombia.
were trained on basic survey techniques, on how to adequately
apply the questionnaire and on how to properly fill in the form.
Energy source Cluster
Along with this, a pilot test was applied to growers of a different
(MJ ha1)
1 2 3 production region in order to check the consistency and clearness
Human labor 17659.5 23998.4 26627.6 of the survey instrument. Minor adjustments were done to the
Machinery 24635.1 65693.4 54749.7 survey form and recommendations were given to the students after
Infrastructure 14792.5 16135.9 14988
this preliminary work.
Fertilizers 32089.1 36605.3 38257.1
Organic fertilizers 3132.1 4342.5 1936.3
The interviewers went all over the study zone through the main
Pesticides 8242.6 10984.5 23367.0 and secondary road systems looking for farms actively engaged in
Plant material 5673.7 6534.1 6297.1 greenhouse tomato production. This strategy was adopted due to
Water for irrigation 8081.9 19753.9 7183.9 the absence of an updated agricultural census for the study zone.
Output 89607.7 124011.5 94489.3
After a brief introduction of the activity and the inquiry to partic-
EUE 0.78 0.67 0.54
NE 24698.6 60036.4 78917.8 ipate, the students applied the survey to individual farmers in an
average time of 1.5 h. During AugusteDecember 2010 a total of 171
surveys were conducted and compiled into a database. Surveys
having more than 20% of missing data were excluded from the
near the optimal number of clusters, was applied to this purpose.
analysis, resulting in a sample of 126 surveys. For the remaining
Next, the k-means algorithm was applied and the group averages
surveys, variables with missing data were completed using the
for the predefined number of clusters were computed. As a result,
nearest neighbors’ method [38].
the average energy use assessment per group was calculated as
well as the efficiency indexes defined in Eqs (1) and (2). The
4. Results
statistical analyses were performed using the statistical program-
ming software R [35].
4.1. Overview of the production system

3. Case study: greenhouse tomato from Colombia Greenhouse tomato production in the study zone, as in the rest
of the country, is carried out under simple wood made structures
3.1. Study zone covered with a single sheet of polyethylene. No active climate
control system is present and air exchange between the exterior
Tomato is among the most consumed horticultural commodities and the inside area is done through an open fixed ridge present in
in Colombia with a production of 519,843 tons (t) harvested in all the spans and by moveable curtains present in the greenhouse
14,128 ha during 2010 [36]. Tomato production is spread country- walls. According to the survey, the average greenhouse floor area
wide and the greenhouse production system is employed in 30% of was 2722 m2, with the greenhouses located in the range between
the cultivated area [37]. The Alto Ricaurte province lies in the 2078 and 2675 m above sea level.
Eastern Range of Colombia and is the most important greenhouse Growers rely mainly in their own experience to handle their
tomato production region in Colombia. This province comprises crop and, eventually; they can get some technical assistance by
a mountainous valley located in the Eastern range of the Colombian agronomists present in the area, but usually guided by commercial
Andes, in the central area of the department of Boyaca. Due to its interests. Many growers have an average of five years of experience
mild climate, the Alto Ricaurte province has seen a rapid expansion in growing tomatoes. Due to the climate of the zone and the
in tomato production area under protected conditions for the last practices applied, growers can plant two cycles of 26 weeks each,
decade. The overall impact of this agricultural activity is evident; per year. The entire production is sent to Bogota, the country’s
however, local institutions keep no reliable figures about it. capital, where the biggest market is located.
The technological level of the production system is low, while
3.2. Data acquisition mechanization is limited to soil preparation and most of the crop
management practices are done by hand. This is presented in
Data was collected through a semi-structured questionnaire Table 2 where the average amount of resources used to plant 1 ha of
consisting of 14 sections, including information concerning social, greenhouse tomatoes is shown. Fertilization in greenhouse toma-
economic, technical, and environmental topics. Farmers located in toes comprises an initial correction of the nutrients in the soil,
the municipalities of Sachica, Santa Sofia, Sutamarchan, Tinjaca and which is done by the application of solid fertilizers. Organic fertil-
Villa de Leyva answered the questionnaire. The overall objective of ization complements this initial solid correction and growers
this survey was to make a full characterization of the greenhouse mainly use chicken manure to this purpose. However, 28% of the
tomato growers residing in the study zone. The first section of the interviewees reported no use of organic fertilization at all. Due to
survey instrument included open questions related to his contact the nature of crop and the intensiveness of this production system,
details and family structure, followed by a set of sections with the permanent application of water and fertilizers through the
questions concerning the planting history and the management irrigation system is a must. Drip irrigation systems are commonly
practices applied. Technical questions included topics such as water used and growers must ensure the water availability all year round,
sources and use, fertigation system, nutrient supply, pest incidence mainly through reservoirs. Diesel pumps are used to operate the
and management, greenhouse characteristics and labor demand. irrigation systems, but 28% of the growers use gravity to run their
Considering the economic dimension, the survey asked for the final systems.
destination of the produce, transport and input costs, prices ob- Pest management in this production system relies entirely on
tained and business perception. Lastly, the survey included ques- the use of chemically synthesized pesticides to control a wide range
tions about the relationship of the grower with other actors of pest and diseases. Growers reported the use of 98 different
involved in the production chain such as local extension services, pesticides, of which 53% were fungicides, 45% were insecticides and
input suppliers, transporters and wholesalers. 2% were acaricides. On average, growers spray 5.8 commercial
468 C.R. Bojacá et al. / Energy 47 (2012) 465e470

pesticides during a growing season. The main targets of the pesti- clusters. This is in agreement with the results of the machinery
cides applied are whitefly (Trialeurodes vaporariorum), tomato leaf category where more operation hours and fuel are needed to irri-
miner (Tuta absoluta), tomato late blight (Phytophtora infestans) and gate the water applied. This cluster was the one with the highest
powdery mildew (Sphaeroteca pannosa). Locally, the implementa- total energy input with 184 GJ ha1 and at the same time the one
tion of sustainable management schemes such as integrated pest with the highest energy output. Cluster 2 surpassed by 38.3 and
management or good agricultural practices still in its infancy. 31.2%, the energy output of clusters 1 and 3, respectively.
As shown in Table 2, there is a great variability in the responses The machinery and fertilizers categories represented the high-
given by the growers. At regional level there is no standard est share of the energy input in cluster 3. In comparison with the
management plans for planting greenhouse tomatoes and growers other clusters the shares of energy use for infrastructure, organic
follow their own plan based on their own criteria. Another vari- fertilization, plant material and water were the lowest with 8.6, 1.1,
ability source for the results obtained in the survey was that 3.6 and 4.1%, respectively. The growers included in this cluster,
growers keep no record about the activities carried out during consumed the highest proportion of energy for pesticides with
a production cycle. In consequence, growers had to rely on their 13.5%. This result doubled the ones obtained for the other two
memory to answer the questionnaire giving some margin to inac- clusters. This cluster reported an intermediate energy input
curate answers. (173.4 GJ ha1) and output (94.5 GJ ha1) in comparison with the
other two clusters.
4.2. Average energy use assessment Considering the energy use assessment, the most energy effi-
cient growers were the ones belonging to cluster 1. This cluster
Following the conventional approach, the energy balance of the showed the highest EUE and NE indexes, by using the lowest input
production system was calculated based on the average resource energy and in consequence having the lowest yield levels. Cluster 3,
consumption and yield reported by the growers. The results for characterized by an output energy relatively close to the one
each production factor and their contribution to the total energy established for cluster 1, exhibited an extra energy input of 51.7% in
input are presented in Table 2. The materials included in the comparison to cluster 1, resulting in the less efficient cluster indi-
machinery and fertilizers categories accounted for approximately cated by the lowest EUE and NE indexes. In the case of cluster 3,
half of the total energy input. The highest energy share of all the each energy unit that exits the system needed almost twice the
inputs was the diesel used to operate the tractor and the irrigation amount of input energy to be produced. Despite of having the
pump. Next, N-based fertilizers and labor were the ones that highest energy input, the energy efficiency of cluster 2 was in the
contributed the most on the energy input. middle. The highest energy use was compensated by the highest
The energy consumption of materials and equipment used energy output of this cluster.
throughout several production cycles is small because energy is The graphical representation of the k-means clustering analysis
apportioned over the equipment lifespan. The energy used for pest is presented in Fig. 1. Through principal component analysis the
management represents 9.1% of total energy input. Energy standardized dataset was adapted and projected onto a two
consumption of the irrigation pump has the lowest share of all dimensional space. A biplot showing the first two principal
input factors. components of the reduced dataset is commonly used to exhibit the
The average production of fresh tomatoes was 120.5 t ha1, closeness among clusters. This is an indication of how different the
which represented an output energy of 96.4 GJ ha1. The energy clusters are by projecting those in the two dimensions considered.
embodied in the biomass residues was not considered since those Fig. 1 shows a clear separation of the observations included in
residues have no commercial value. The EUE index was 0.68 while cluster 2 in relation with the other two clusters. A slight overlap
the NE was 45368.7 MJ ha1. These results indicate a negative
energy balance implying that the amount of energy that enters the
system is higher than the amount represented by the edible
product.

4.3. Energy use assessment per cluster

The 126 growers were grouped into three clusters according to


the results given by the k-means algorithm. The multivariate
method was able to group 71, 20 and 35 growers into clusters 1, 2
and 3, respectively. Table 3 presents the energy profile for each
cluster based on the average input and output production factors.
Growers included in cluster 1 were characterized by a total energy
input of 114.3 GJ ha1, with the highest share represented by the
fertilizers category (28.1%). The lowest share was allocated to the
energy embodied in the organic fertilizers. Compared with the
others, the growers included in this cluster used the lowest amount
of energy but at the same time were the ones with the lowest
production.
Cluster 2 was characterized by a higher energy use from the
machinery category, which is mainly related to the use of fuel for
the machinery operation. This category represented 35.7% of the
total energy consumption while organic fertilization indicated the
lowest share with 2.4%. This cluster was also characterized by using
Fig. 1. Biplot of the k-means three-cluster solution for the standardized energy
the lowest energy share of labor, pesticides and fertilizers in production factors of the greenhouse tomato production system located in the Alto
comparison with the other two clusters. On the other hand, this Ricaurte province of Colombia. The legend of the figure includes a description of each
cluster rated the highest share of water use among the three cluster in terms of efficiency, input and output energy.
C.R. Bojacá et al. / Energy 47 (2012) 465e470 469

between cluster 1 and 3 is observed in the biplot, indicating a closer Acknowledgments


relationship between these two clusters. As explained before, this is
a consequence of lower energy inputs and closer energy output The present study funded by the Flemish Interuniversity Council
values for these two clusters. (VLIR, http://www.vliruos.be) as part of the own initiative project
ZEIN2009PR364 entitled ‘Multidisciplinary assessment of effi-
ciency and sustainability of smallholder-based tomato production
5. Discussion systems in Colombia, with a roadmap for change’.

The energy consumption for the production of greenhouse References


tomatoes in the Alto Ricaurte province is 20% lower than that of
a production system in the Bogota plateau [9]. Some of the differ- [1] Pelletier N, Audsley E, Brodt S, Garnett T, Henriksson P, Kendall A, et al. Energy
intensity of agriculture and food systems. Annual Review of Environment and
ence comes from differences in the data input. Nevertheless, both
Resources 2011;36:223e46.
studies indicated that one of the most energy consuming factors [2] Karkacier O, Gokalp Goktolga Z. Inputeoutput analysis of energy use in
was the application of chemical fertilizers. agriculture. Energy Conversion and Management 2005;46(9e10):1513e21.
In comparison with studies carried abroad, the energy use [3] Schneider U, Smith P. Energy intensities and greenhouse gas emission miti-
gation in global agriculture. Energy Efficiency 2009;2:195e206.
pattern showed in the present study is similar for studies that share [4] Pervanchon F, Bockstaller C, Girardin P. Assessment of energy use in arable
a comparable production system. In the present study, a total farming systems by means of an agro-ecological indicator: the energy indi-
energy consumption of 141.8 GJ ha1 was found, while in Turkey cator. Agricultural Systems 2002;72:149e72.
[5] Erdal G, Esengün K, Erdal H, Gündüz O. Energy use and economical analysis of
ranged from 90.3 to 126.3 GJ ha1 [13,39] and in Iran, greenhouse sugar beet production in Tokat province of Turkey. Energy 2007;32:35e41.
tomatoes consumed 131.6 GJ ha1 [40]. Considering the energy [6] Pimentel D. Food for thought: a review of the role of energy in current and
profile in developed countries, the total energy input in Spanish evolving agriculture. Critical Reviews in Plant Sciences 2011;30(1e2):35e44.
[7] Beheshti Tabar I, Keyhani A, Rafiee S. Energy balance in Iran’s agronomy
and Dutch greenhouses is 460.1 [41] and 13,000 GJ ha1 [42], (1990e2006). Renewable and Sustainable Energy Reviews 2010;14(2):
respectively. All the studies carried out in developing countries, 849e55.
including the present one, showed that the energy embodied in [8] Mushtaq S, Narayan Maraseni T, Maroulis J, Hafeez M. Energy and water
tradeoffs in enhancing food security: a selective international assessment.
diesel fuel and N-based fertilizers was the one with the highest
Energy Policy 2009;37(9):3635e44.
share among all the inputs considered. [9] Bojacá CR, Schrevens E. Energy assessment of peri-urban horticulture and its
According to the results the growers included in clusters 2 and 3, uncertainty: case study for Bogota, Colombia. Energy 2010;35(5):2109e18.
[10] Unakitan G, Hurma H, Yilmaz F. An analysis of energy use efficiency of canola
representing 43.7% of the surveyed growers, exhibited average EUE
production in Turkey. Energy 2010;35(9):3623e7.
values lower than the one obtained for cluster 1. This result indi- [11] Zangeneh M, Omid M, Akram A. A comparative study on energy use and cost
cates the possibility of optimize the energy use of those growers analysis of potato production under different farming technologies in
that are actually not reaching the EUE found in cluster 1. Improving Hamedan province of Iran. Energy 2010;35(7):2927e33.
[12] Kizilaslan H. Inputeoutput energy analysis of cherries production in Tokat
the situation of these less efficient growers can lead to an overall province of Turkey. Applied Energy 2009;86(7e8):1354e8.
improvement of the production system at regional level. Internally, [13] Hatirli SA, Ozkan B, Fert C. Energy inputs and crop yield relationships in
growers included in cluster 1 are representing the best energy greenhouse tomato production. Renewable Energy 2006;31(4):427e38.
[14] Liu Y, Langer V, Hogh-Jensen H, Egelyng H. Energy use in organic, green and
profile right now but this does not imply the best possible scenario. conventional pear producing systems-cases from China. Journal of Sustainable
For the case of study, it is necessary an overall optimization in the Agriculture 2010;34(6):630e46.
use of resources in order to reach a positive energy balance. [15] Alonso A, Guzman G. Comparison of the efficiency and use of energy in
organic and conventional farming in Spanish agricultural systems. Journal of
Improving the energy balance of the less efficient growers can be Sustainable Agriculture 2010;34(3):312e38.
seen as a complementary strategy to improve the overall energy [16] Deike S, Pallut B, Christen O. Investigations on the energy efficiency of organic
balance of the system. and integrated farming with specific emphasis on pesticide use intensity.
European Journal of Agronomy 2008;28(3):461e70.
[17] Liu Y, Hogh-Jensen H, Egelyng H, Langer V. Energy efficiency of organic pear
production in greenhouses in China. Renewable Agriculture and Food Systems
6. Conclusions 2010;25:196e203.
[18] Mohammadi A, Omid M. Economical analysis and relation between energy
inputs and yield of greenhouse cucumber production in Iran. Applied Energy
The applicability of multivariate data analysis techniques in the 2010;8791:191e6.
context of energy use in agriculture was investigated through a case [19] Ozkan B, Fert C, Karadeniz C. Energy and cost analysis for greenhouse and
study related to the greenhouse tomato production in Colombia. open-field grape production. Energy 2007;32(8):1500e4.
[20] Mohammadi A, Tabatabaeefar A, Shahin S, Rafiee S, Keyhani A. Energy use and
The actual production system generates a lower energy output than economical analysis of potato production in Iran a case study: Ardabil prov-
the one inputted, showing an inefficient model. The system mainly ince. Energy Conversion and Management 2008;49(12):3566e70.
relies in non-renewable energy sources such as fossil fuels and [21] Funes-Monzote FR, Monzote M, Lantinga EA, Ter Braak CFJ, Sánchez JE, Van
Keulen H. Agro-ecological indicators for dairy and mixed farming systems
nitrogen based fertilizers to achieve its actual yields. classification: identifying alternatives for the Cuban livestock sector. Journal
The k-means clustering analysis uncovered clusters of farmers of Sustainable Agriculture 2009;33(4):435e60.
who share similar management practices and energy profiles. This [22] Ottaviani D, Ji L, Pastore G. A multidimensional approach to understanding
agro-ecosystems: a case study in Hubei Province, China. Agricultural Systems
offers the possibility of optimizing the production of less efficient 2003;76:207e25.
growers by comparing the various energy inputs and outputs. This [23] Spavorek G, Scnug E. Soil tillage and precision agriculture: a theoretical study
would be a useful step to consider before the introduction of new for soil erosion control in Brazilian sugar cane production. Soil Tillage
Research 2001;61:47e54.
strategies for overall optimization of energy consumption of all [24] Jones MR. Analysis of the use of energy in agriculture e approaches and
growers. problems. Agricultural Systems 1989;29:339e55.
Multivariate analysis methods offer a straightforward and [25] Hülsbergen KJ, Feil B, Biermann S, Rathke GW, Kalk WD, Diepenbrock W.
A method of energy balancing in crop production and its application in
powerful set of procedures that were useful to gain insight into the
a long-term fertilizer trial. Agriculture, Ecosystems and Environment 2001;
data. The wealth of information given by these types of analysis 86:303e21.
suggest pathways to improve the inputeoutput energy balance [26] Singh H, Mishra D, Nahar NM. Energy use pattern in production agriculture of
methodology, especially for agricultural production systems where a typical village in Arid Zone India e Part I. Energy Conversion and
Management 2002;43:2275e86.
unstandardized practices and particular criteria widen the [27] Fluck RC. Energy in human labor. In: Fluck RC, editor. Energy in farm
productivity range in a given study area. production. Amsterdam: Elsevier; 1992. p. 31e7.
470 C.R. Bojacá et al. / Energy 47 (2012) 465e470

[28] Medina A, Cooman A, Parrado CA, Schrevens E. Evaluation of energy use and [36] Colombian Ministry of Agriculture and Rural Development. Statistics for the agri-
some environmental impacts for greenhouse tomato production in the high cultural sector. Available from: http://www.agronet.gov.co; 2011 (in Spanish).
altitude tropics. Acta Horticulturae 2006;718:415e21. [37] Miranda D, Fischer G, Barrientos JC, Carranza C, Rodríguez M, Lanchero O.
[29] Singh JM. On farm energy use pattern in different cropping systems in Har- Characterization of productive systems of tomato (Solanum lycopersicum L.) in
yana, India. M.Sc. thesis, International Institute of Management, University of producing zones of Colombia. Acta Horticulturae 2009;821:35e45.
Flensburg, Germany; 2002. [38] Torgo L. Data mining using R: learning with case studies. FL: CRC Press Boca
[30] Esengun K, Erdal G, Gündüz O, Erdal H. An economic analysis and energy use Raton; 2010. pp. 305.
in stake-tomato production in Tokat province of Turkey. Renewable Energy [39] Bayramoglu Z, Gundogmus E. The effect of EurepGAP standards on energy
2007;32:1873e81. input use: a comparative analysis between certified and uncertified green-
[31] Everitt B, Hothorn T. An introduction to applied multivariate analysis with R: house tomato producers in Turkey. Energy Conversion and Management
use R! New York: Springer; 2011. pp. 273. 2009;50:52e6.
[32] Wehrens R. Chemometrics with R: multivariate data analysis in the natural [40] Heidari MD, Omid M. Energy use patterns and econometric models of major
sciences and life sciences: use R! Springer-Verlag Berlin; 2011. pp. 285. greenhouse vegetable productions in Iran. Energy 2011;36(1):220e5.
[33] Steinley D. Stability analysis in k-means clustering. British Journal of Mathe- [41] Martínez-Blanco J, Muñoz P, Antón A, Rieradevall J. Assessment of tomato
matical and Statistical Psychology 2008;61:255e73. Mediterranean production in open-field and standard multi-tunnel green-
[34] Calinski T, Harabasz J. A dendrite method for cluster analysis. Communica- house, with compost or mineral fertilizers, from an agricultural and envi-
tions in Statistics 1974;3:1e27. ronmental standpoint. Journal of Cleaner Production 2011;19(9e10):985e97.
[35] R Development Core Team. R: a language and environment for statistical [42] de Gelder A, Poot EH, Dieleman JA, de Zwart HF. A concept for reduced energy
computing. Vienna, Austria: R Foundation for Statistical Computing, ISBN demand of greenhouses: the next generation greenhouse cultivation in the
3-900051-07-0; 2011. URL Available from: http://www.R-project.org; 2011. Netherlands. Acta Horticulturae 2012;952:539e44.

You might also like