Fertilizer is used extensively in Pakistan, and the amount used has risen from 2.6 million nutrient tonnes in 1999 to 3.7 million nutrient tonnes in 2009. The ratio between nitrogen, phosphorus, and potassium-based fertilizers (N:P:K ratio) averaged 1:0.28:0.01 from 1998 to 2007 (calculated from [14], p. 61). This is also supported by the Food and Agriculture Organization (FAO) [15]. Potash application has been low historically, and only two percent of farmers countrywide actually apply it [15]. This is further supported by scattered wheat and rice-specific figures supplied by FAO and International Fertilizer Industry Association (IFA) sources that show exceptionally low figures ranging from 0.7–3.5 kg ha⁻¹ on wheat and 0.2–1.0 kg ha⁻¹ on rice in various years since 1989 (with the exception of 1992–1993 where potash application was high at 11.3 kg ha⁻¹) [16-20].

Pesticide application on crops in Pakistan began to expand in the early 1980s, increasing from 3,500 tonnes in 1981 [21] and growing by a factor of 27 to 94,265 tonnes in 2007 ([9], p. 150). Much of Pakistan's pesticide was imported until the late 1990s, but now domestic production exceeds 60% of the amount used annually [21]. In Pakistan, most “pesticide” is insecticide applied largely to cotton and rice [22]. Wheat is the largest user of herbicide against grasses such as little seed canary grass (Phalaris minor) which, in Pakistan, is said to reduce wheat yields by 15–20% in the absence of herbicides [22]. Forty-eight percent of farmers believe that pesticide is a necessary input to increase crop yield [22]. However, the same survey shows that 97% of farmers believe pesticides are adulterated [22].

Tractors have become the dominant mode of traction power in agriculture and bullock-operated farms are on the decline [23]. The number of tractors being used in Pakistan increased by a factor of 85 between 1961 and 2007 [24]. Most of the country's wheat crop is threshed with machines and mechanical rice husking is also on the rise [23]. Farmers who do not either own or rent tractors are few and far between [25]. In addition, recent data shows that 90% of farms use tractors in contrast to only 17% in 1972 [21]. This figure climbs to 96% in The Punjab and Sindh regions of Pakistan [21].

Given Pakistan's dry climate and low average rainfall, irrigation has been a major part of agriculture in the region since 3,000 BC. Canal irrigation sources include glaciers in the north, snowmelt, and rainfall outside the Indus Plains [21]. Groundwater is an integral part of irrigation; the canal system is becoming more of a groundwater recharge mechanism than a water delivery system [21]. This is especially true of The Punjab province where recharge from canals is responsible for 80% of pumped groundwater [26]. On average, groundwater accounted for 36.7% of available irrigation water between 1999 and 2010 (calculated from [9], p. 138-139).

Energy analysis is not new to agriculture. The literature shows a vast quantity of research on the issue, especially for energy crops (ethanol) such as sugarcane and corn. Various writers including Shapouri and Salassi (2006) have performed energy analyses on agricultural systems [27]. The basic methodology is the same in most cases. What differs is the definition of “appropriate” input to the systems. Studies show EROI ratios for US corn-based ethanol both above and below one [28]. The reasons for these differences include the accounting of co-products, the exclusion of the embodied energy of equipment, and the use of internally-derived energy sources [28]. Defining study boundaries clearly is important and we define them for this study in Section 2.1.

Pakistan's agricultural sector shows an increasing dependence on fossil fuel inputs in the form of fertilizer, pesticide, and mechanization, a common trend worldwide. The 1960s and 1970s were characterized by massive yield increases [29] due to improved varieties, petroleum-based fertilizers and pesticides, irrigation, and diesel-driven tractors [30-32]. For example, nitrogen (N) fertilizer usage in Pakistan increased from an average of 36,000 tonnes in 1960–1963 to 326,000 tonnes in 1970–1973, to 876,000 tonnes in 1980–1983 [21]. The new crop varieties that were developed relied upon increasing amounts of fossil fuel inputs. Crosson and Brubaker (1982) stated with reference to new agricultural technologies that “Not only is the technology itself keyed to energy from fossil fuels, but the research establishment that developed the technology also is oriented to exploitation of this resource” [33]. The direct link between increased energy usage and increased agricultural output has been researched widely [34-37]. This notion has sparked debate on long-term yield viability and environmental degradation [38-40].

The relatively few studies on energy use in agriculture in Pakistan cover small districts using site-specific data. However, Jameel's (1982) study provides a detailed analysis of energy use in Pakistani agriculture. He found that fertilizer production accounted for 45% of all commercial energy supplied to the agricultural sector. Another 40% was used for irrigation and drainage [41].

A more recent Pakistani study examined energy use on sugar cane crops in Dera Ismail Khan District. It compared energy inputs to sugarcane yields and discovered that fertilizer and irrigation were the largest energy inputs. The results showed that energy consumption was higher on tractor-operated farms than bullock-operated farms by a factor of 1.2. The energy output-input ratios were marginally higher for bullock-operated farms [42]. Similarly, Khan and Singh (1996 and 1997) studied energy use on sugar cane and wheat in the same district [43,44]. However, all three analyses restricted the usage of “energy analysis” to direct use: human labor, animal power, diesel and electrical irrigation motors, and tractors. Furthermore, the premise of such studies was to compare energy usage and energy outputs between different categories of farms such as tractor-operated farms and bullock-operated farms. They found that per-hectare energy usage was the highest on farms using electrical or diesel-powered pumps for groundwater pumping. The output-input analysis showed that while crop yields were higher on irrigated farms, yield increases were not proportional to increases in energy input. This implies that farmers were investing large quantities of energy, but not harvesting proportional crop output. The reasons ranged from overwatering to excessive energy inputs to the point where they do not contribute positively to output. This is addressed in greater detail in section 4. Other studies in neighboring countries such as India [45-47] also quantified inputs (seed, fertilizer, pesticide) in energy terms. The purpose of such studies were generally the same, citing, for example, the need for an understanding of the fact that fertilizer and chemical pesticides are produced through fossil energy-intensive processes, and that perhaps yields could be increased or at least maintained through the judicious use of such inputs and the supplementation of inputs with farmyard manure [45].

To our knowledge, an energy analysis of the entire crop production system for the main crops grown in Pakistan has not been completed to date. The objective of this study therefore, is to perform an EROI analysis of the country's entire wheat and rice crops from 1999 to 2009 using extant secondary data, and to examine the size of the contribution of different inputs in relation to the output in total, and on a per-hectare basis. The purpose of this is to provide some insight on the trends in energy use and identify the main energy inputs. In addition, labor is represented unevenly in economic analyses because it is comparatively cheap in Asia and Pakistan. Considering it from an energy point of view removes this discrepancy. Furthermore, this analysis draws attention to the fact that fertilizer and pesticide are produced using highly energy intensive industrial processes. Price may not always reflect this adequately. This data can then be used to guide decisions about investments in the management of these cropping systems in a possibly energy-constrained future.

The concept of quantifying energy inputs and comparing them to energy outputs is rooted in an energy accounting method called “net energy analysis” (NEA). The central idea of NEA is that net energy is the gross energy output resulting from a given process minus the energy required to obtain it [the gross energy] [48]. The output must be greater than the input (sometimes termed “feedback”) in order to be energetically feasible. Energy return on investment, on the other hand, studies the energy output and inputs by means of a ratio. It is generally applied to the mine-mouth, wellhead, or farm-gate [49]. Thus, an EROI ratio of 100:1 means that 100 units of energy are produced for every one unit of energy invested to locate, extract, produce, and upgrade the energy source or product being studied. By the strict logic of a 1:1 EROI, foods such as beef, chicken, eggs, cauliflower, winter tomatoes, lettuce, and various seafoods could be considered unfeasible as they commonly require more energy to grow, harvest, or catch, and deliver to consumers than their own energy content [38]. Thus, it should be understood that everything cannot be evaluated simply in reference to a greater-than-one EROI without considering the importance of the quality of the output. Human beings make specific choices where affluence often allows nutrition and palatability to overshadow simple energy return. Understanding and being aware of the implications of these choices in terms of energy is important in understanding how resources are allocated and used.

The NEA concept has existed in the US since the 1950s [50]. Conducting NEAs is required by the Federal Nonnuclear Energy Research and Development Act of 1977 and The Energy Security Act of 1980 [51,52]. Net energy analysis is commonly used to evaluate the feasibility of energy projects such as electricity generation plants and various renewable energy technologies. This method of energy analysis is well-defined, and NEA offers a basis for reducing and conserving input energies, guarantees the chance to evaluate net energy yields independent of economic risk questions, and allows comparisons between the net energy yields of specific industrial plants and processes [53].

The value of NEA as a tool for resource-use analysis is derived chiefly from the fact that it assesses physical resources and is therefore resistant to market imperfections that may distort monetary data [54]. However, historical energy data is not always available, and in some situations, monetary data must be used to infer energy values. It is believed that NEAs reflect real value more accurately than monetary metrics because they use specific energy units rather than dollars which are affected by such variables as time, markets, changing tastes, living standards, and public policies [52]. The US Comptroller General acknowledged that economic analyses were required to evaluate energy projects, but that “Dollar measurements do not substitute for NEA because they are not based on explicit physical energy requirements and because of imperfections in the energy marketplace” [53]. In addition, NEAs can often be used to point out where improvements can be made in system operations [52].

Many neoclassical economists dispute the value of NEA, arguing that they do not contribute much more additional useful information than does a thorough economic analysis [54]. Most analyses of agricultural output are indeed done in monetary terms. The question then is, why use energy terms instead, when a wealth of economic tools already exist for studying the feasibility and profitability of an economic activity?

Our justification for performing an EROI analysis is two-fold. First, as Hall et al. (1986) point out, “energy is the ultimate limiting resource” [38]. While it is incorrect to say that everything can be reduced to energy, it is important to understand that all material and most nonmaterial resources have an associated energy cost. Second, energy flows are governed by the irrefutable laws of thermodynamics formulated by Joule, Clausius, and Thomson. This implies that the flows of energy through an agricultural system are subject to energy “losses” which must be accounted for. Monetary measures are not subject to these natural laws, and money, unlike energy, can be generated indefinitely. As mentioned earlier, they are influenced by social forces related to the economy such as inflation, public policy, and markets [52].

Conducting an EROI analysis requires a clear definition of the system being studied, in this case the agricultural systems of wheat and rice in Pakistan, and the energy flows that are being measured. The energy inputs considered are the fossil fuel inputs needed to produce wheat and rice and the output is the energy stored in the harvested crops at the farm-gate. Crop residues have varied uses as well—such as fodder and fuel—and it is possible that accounting for them would increase EROI figures. However, this analysis is concerned only with the edible (human) energy produced in these systems. It is noteworthy that crop residues such as wheat straw, cotton stems, sugar cane trash/tops and rice husks are not recycled in the soil in Pakistan where they could add to soil fertility [15]. However, we have not accounted for this as relevant secondary data is not available. The energy inputs quantified in this system are seed, fertilizer, pesticide, the energy to operate tractors and tube wells and their embodied energy, the energy used to transport fertilizer and pesticide locally, and energy invested in the form of human labor (Figure 1).

Inputs such as sunlight and rainfall are considered as natural inputs with no energy investment required by humans. Furthermore, as this is an energy analysis, there is the concern of resource depletion. The sun is considered an unlimited stock of energy whereas fossil fuels on earth are non-renewable, and therefore limited [55,56]. We excluded the sun for this reason. The output from the system is at the farm-gate, so post-harvest energy costs such as crop transport, storage, and processing were not included in this analysis [49]. Finally, environmental energy costs such as lost environmental services due to ecosystem degradation are beyond the ambit of this study, and while they are important for the long-term sustainability of these production systems, they were not included in this analysis.

Data on the energy inputs into the system are all expressed in terms of energy in joules (J). However, since input data for agriculture is seldom collected in terms of energy, we converted available data to energy units using established conversion factors, which are outlined below.

Tables 10 and 11 show all inputs to wheat and rice quantified in energy terms.

The energy content of seed for planting and as the product of the cropping system were taken to be the same, and used official production figures from [9] (pp. 6–7, pp. 16–17) to calculate the energy output of the crops.

All of the energy inputs were added up and compared against crop production or energy output to calculate the EROI using the formula “EROI = energy output/energy inputs.” [49].

Linear, quadratic, cubic, and quartic regressions were conducted for wheat and rice EROI over time (x = fiscal year, y = EROI) and for wheat and rice inputs per hectare against wheat and rice energy output per hectare (x = energy inputs per ha, y = energy output per hectare) using Analyze-it ® version 2.22 for Microsoft Excel ® [86].

The size of the energy input contribution of fertilizer and pesticide to both crops is noteworthy.

The EROI of wheat shows a gradual downward trend from 2000 to 2006.There was an overall decrease of 21.3% during that time period. After that, the trend is erratic registering alternative increases and decreases. The EROI of rice shows a decreasing trend until 2005 (with variability within that trend). Thereafter, it rises fairly rapidly, an increase of 56.2% between 2005 and 2009. Rice's EROI was consistently higher than wheat's in the same year. Wheat's EROI trend is fairly constant in that it hovers above or below the 3.0 mark throughout. Rice's performance is variable, but the general trend appears to be a decrease that lasts halfway through the study period, followed by a steady increase (Figure 2).

The linear and quadratic regressions for wheat EROI were significant. Cubic and quartic regressions were not (Table 12). For rice EROI, the quadratic, cubic, and quartic regressions were significant

(Table 13). However, the residual sum of squares values indicates that the quartic regression equation is the best fit for both wheat and rice EROI.

Both the rice and wheat crops follow a similar pattern of decreasing output at the beginning of the study period followed by a steady increase during the rest of the study period. Over the course of the study period, wheat output increased by 34.6% (31.9% if measured from 2002), and rice by 48.2% (80.3% if measured from 2002) (Figure 3).

The energy input trend to both crops follows a similar pattern. The inputs to both crops increased from 1999 to 2005 at different rates over the length of the study period. Inputs to wheat increased by 41.7% and by 41.0% for rice between 1999 and 2009. The sharpest increases in inputs to wheat occurred in 2005 and in 2004 to rice. It is noteworthy that after 2005, the inputs to wheat increased gradually, but for rice, decreased overall.

Presenting energy inputs and outputs on a per-hectare basis allows a direct comparison between wheat and rice. Wheat's per-hectare energy output was consistently greater than that of rice, on average by a factor of 1.1. Many of its inputs are also greater, including N fertilizer by an average factor of 2.5, seed by an average factor of 2.8, and tractor diesel by an average factor of 4.2. Conversely, rice had several inputs that were greater than corresponding ones on wheat. Rice is more insecticide intensive by an average factor of 44.6. It is also more labor intensive by an average factor of 2.1. Finally, as rice is more water-intensive than wheat, its tube well diesel figure was greater by an average factor of two.

Examining percentage changes in major inputs in relation to percentage changes in outputs on a per-hectare basis reveals some interesting patterns. It appears that increases in some major inputs do not necessarily translate into increased output. There are numerous instances where a sizable increase in N fertilizer had little, or a negative impact on wheat output. While this is not suggesting a correlation (or the lack thereof) between N fertilizer and output, it certainly merits discussion. Similar observations are made for rice. Large increases in insecticide (specifically 2002, 2004, and 2007) had little impact on output. Often times, large decreases in insecticide appeared to result in better output (2005, 2006, 2008, and 2009; Table 14).

The relationship between all per-hectare inputs and per-hectare output also shows some interesting patterns. The 11-year dataset shows that wheat yield is still increasing with rising inputs. Rice, however, shows a gradually declining trend (Figure 6).

For wheat per-hectare input against per-hectare output, the linear, quadratic, and quartic regressions were significant (Table 15). None of the regressions were significant in the case of rice (Table 16).