Rising input costs and increasing societal demand for sustainability, call for significant improvements in resource use efficiency in agriculture. Fortunately, in many cases resources such as fertilizers, pesticides, irrigation water, land, and labor can be used with much greater efficiency than currently found by measuring and responding to spatial and temporal variability. Precision Agriculture (PA) is the scientific domain that deals with management of spatial and temporal variability to improve economic returns and reduce environmental impact. For farmers, PA is expected to lead to an increase in profitability; for society, PA is expected to lead to increased sustainability [1
PA techniques can be classified as recording, reacting, or guidance techniques [4
]. A recording technique measures the status of crop, soil, weeds, pests, or diseases, typically using sensors on mobile equipment. Examples are soil scans, canopy reflectance measurements, weed detection using cameras, yield monitors on grain combines or potato harvesters, and sensors mounted on satellites, airplanes and Unmanned Aerial Vehicles (UAVs) [5
]. Typical examples of reacting techniques are variable rate application of fertilizers and pesticides, usually on the basis of prepared maps, but this can also be on-the-go in response to a sensor signal. The most ubiquitous guidance technique is Global Navigation Satellite System (GNSS) based auto-steering for tractors and other machinery which reduces overlap between tractor passes and reduces operator fatigue [7
]. Guidance techniques are also the foundation of Controlled Traffic Farming (CTF).
Recording techniques and reacting techniques as described above must be linked by a data processing step in which an action is defined based on the new data, previously recorded data, data from other sources (weather), scientific knowledge (mechanistic models), practical knowledge (empirical models), and farmer preferences. Typically, a Farm Management Information System (FMIS) is used to handle the data [8
]. The most advanced FMIS include algorithms to make decisions. For example, Akkerweb (www.akkerweb.nl
), which evolved out of a decision support system for control of plant parasitic nematodes [9
], has more than ten recommendation apps [11
]. A recent review of data recording and processing for weed control and crop protection is available [13
PA can be used to make actions spatially precise. Conventionally, fertilizer or pesticide is applied uniformly on the entire field. Against this baseline, PA can increase fertilizer efficiency by measuring crop nutrient status and adjusting the fertilizer rate accordingly. Also, PA can increase herbicide use efficiency by adjusting the herbicide application rate to the observed weed density. The scale at which measurements are made and at which actions are implemented is important [14
]: in general, the reduction in fertilizer or pesticide use will be greater as the area for which decisions are made decreases. Thus, recording and reacting on 5 × 5 m2
will often result in greater reductions than recording and reacting on 50 × 50 m2
but working on a smaller scale will require a larger investment in machines [15
PA can also be used to make actions temporally precise. In this mode, a decision is made when (if at all) to perform an operation, for example to control weeds [16
], to control late blight [18
], or to control plant parasitic nematodes [9
PA is thought to benefit both the farmer and society. For farmers, PA is attractive primarily if it provides a positive return on investment. Switching from conventional agriculture to PA requires investments: in technology (sensors, software, applicators), in knowledge (to operate the new machinery), and time (to learn a different way of working, to build relationships with a new set of service providers). The return on investment may come from several directions. Typically, PA will result in a reduction in input use. For expensive inputs such as certain pesticides, the associated cost savings may provide a sufficient return on investment. This will not be the case for cheap inputs such as nitrogen (N) fertilizer. Here, PA may be worthwhile for a farmer because it makes it easier to comply with laws that regulate fertilizer use. PA may lead to yield increase and thus higher revenue, for example because a reduced herbicide application causes less damage to the crop. PA may increase revenue through quality increase, for example when variable-rate planting or variable-rate application of N in potatoes leads to more uniform tuber size distribution or higher specific gravity of tubers.
For society, PA is attractive because it is assumed to increase the sustainability of farming. This is a reasonable assumption because in many cases a reduction in the use of inputs will lead to a reduction in environmental impact, but this is not necessarily the case. As an example, let us consider weed control. Weed control using an herbicide has a negative impact on the environment. This impact may be avoided by using close-to-crop mechanical weeding guided by computer vision. However, CO2 emission of such a system is higher because mechanical weeding requires more energy than is needed to produce and apply an herbicide, especially if two or more weeding passes are required. Thus the question arises whether sustainability is improved by accepting higher CO2 emissions in order to achieve a reduction in herbicide use.
The Brundtland Commission provided a widely accepted, qualitative definition of sustainability when it stated that “sustainable development is development that meets the needs of the present without compromising the ability of future generations to meet their own needs” [19
]. Many other definitions can be found [20
]. These definitions have as a common thread that three categories of aspects are considered, namely economics (“profit”), environment (“planet”), and well-being of humans (“people”), the so-called Triple Bottom Line [25
]. Thus, sustainability measurement requires that trade-offs between categories are quantified, something that is not normally in the mind set of people [27
Life Cycle Analysis (LCA) is one of the concrete methodologies that recognizes the Triple Bottom Line and it is widely used to assess sustainability in concrete cases [28
]. LCA measures performance in impact categories such as global warming, acidification, eco-toxicity, but it stops short of integrating these separate measures into one composite indicator. Such an overall measure of sustainability is provided by indicator-based methods [31
], e.g., Multi Criteria Analysis (MCA) [33
], Total Factor Productivity (TFP) [34
] and indexes of sustainability [38
]. In this paper we focus on the adjusted (or social) profit method [40
The objective of this paper is to examine several PA practices in potato and olive production in terms of how much they increase profitability and sustainability.
Results for potato are shown in Table 6
and Table S1
. Emission of GHG for potatoes according to our calculations is 4329 kg CO2
-eq for a production of 53,560 kg of potatoes, which is 83 kg t−1
; this is comparable with the 77 kg t−1
reported elsewhere [73
The externality pesticide toxicity is reduced most by using variable rate application of fungicides for late blight control: this reduces the environmental cost from 82 € ha−1
to 73 € ha−1
(an 11% reduction; see Table 7
). VRA for soil herbicides (SH) reduces this externality from 82 € ha−1
to 76 € ha−1
(a 7% reduction) and using variable rate application of potato haulm killing herbicides (PHK) reduces this externality from 82 € ha−1
to 77 € ha−1
(5% reduction). Using variable rate application for all three (SH, PHK, and LB) has of course the largest effect: it reduces the externality pesticide toxicity from 82 € ha−1
to 63 € ha−1
The externality GHG emissions is reduced by variable rate application for sidedress N (SN) from 136 € ha−1
to 124 € ha−1
(a 9% reduction; see Table 7
). VRA for SN also reduces eutrophication by 17%.
Profit is increased by variable rate application of herbicides (11%) but the other PA techniques have just a small or even a slightly negative effect on profit. When all four PA techniques are applied together, profit is increased by 13%.
Social profit is affected in the same way as profit, but the effects are slightly larger because the increase in profit and the decrease in environmental cost operate in the same direction. Thus, as shown in Table 7
, social profit is increased by 13% if VRA for SH is used and by 26% when all four techniques are used together. Profit, externalities and social profit summarized in Figure 2
Results for olive are detailed in Table 8
and Table S2
and summarized in Figure 3
. PA in olives results in substantial reductions in P, K, and lime use. Fertilizer is relatively cheap, however, and the soil samples needed for variable rate application are relatively expensive, which when taken together means that profit is increased only slightly (compare SN in potatoes). Likewise, the impact on social profit is small.
Our results show that in potato production, PA can lead to a reduction in pesticide use of 23% (expressed as EIQ) and a reduction of nitrogen fertilizer use of 15%. In olive production, the use of potassium fertilizer is reduced by 31%, of phosphate fertilizer by 59%, and of lime by 86%. A significant reduction in environmental impact is associated with this reduction in input use.
Farmers incur some costs when they use PA but profitability increases nevertheless. The numbers are different for each PA technique. In the case of expensive inputs (some pesticides), reduction in input use results quickly in cost reduction, but in the case of less expensive inputs (nitrogen fertilizer), the cost savings are approximately equal to the extra cost to the farmer. In this case there is no or just a small economic benefit to the farmer. However, overall the conclusion is that PA increases profitability as well as sustainability.
Our quantification of the contribution of PA to sustainability started with a detailed listing of inputs, outputs, and externalities. Many methods to measure sustainability start with this step. This first step provides valuable insight into the effect of using PA but it leaves unanswered the kind of question that was posed in the Introduction section, namely do we increase sustainability when pesticide use is reduced by a certain amount and fuel use is increased by a certain other amount? In order to answer this and similar questions, ultimately sustainability must be quantified using hard measurements which can be subjected to rigorous analysis [74
]. In this paper we attempt to progress towards such analysis by making explicit the trade-off between incompatible quantities. We use shadow prices to express all quantities using a common denominator. This enabled us to subtract both conventional costs and (external) environmental costs from revenue and thus calculate social profit as an aggregated measure of sustainability.
It can be argued that economic valuation of important environmental impacts such as loss of biodiversity and pollution of natural resources is not meaningful. However, it is clear that our society needs to do a better job of protecting intangible values and common resources while at the same time continuing to create economic value. Human beings are very familiar with making trade-offs and economic value plays a role, if often on a subconscious level. We argue that economic valuation can be a tool to help make decisions that take into account previously invisible goods and services such as biodiversity and other forms of natural capital.
A more serious drawback of our method is perhaps that we have not taken into account temporal effects: a full analysis of sustainability should include temporal effects [75
]. In this paper we feel justified in leaving out temporal aspects because we are interested in relative sustainability. That is, we attempt to answer the question whether using PA is “more” sustainable than not using PA. We do not make a statement about whether either production system could be continued indefinitely [19
]. If the goal is to make statements about longer time horizons, many more aspects need to be included in the analysis. For example, changes in rainfall patterns due to climate change are expected to increase nutrient leaching and eutrophication [77
Farmers make decisions at the level of the farm. An important aspect is how different crops are grown in a rotation. Our analysis at the level of a crop of potatoes does not address all issues important to a farmer. However, potato is the most profitable crop in The Netherlands and profitability of the farm will be determined in large part by the performance of the potato crop. Olives are a permanent crop and then there is no rotation.
The size of a farm is an important determinant for some per-ha prices and therefore for the profitability of PA technologies. In earlier work it was calculated that an arable farm in the Netherlands must be 60–125 ha or larger, depending on the crops grown, in order for the investments in PA technology to pay off [53
], In this paper we assumed 100 ha of potatoes to calculate the price for some technologies. However, in potato production combining all four PA technologies we assumed that all sensor measurements are bought as a service at a fixed price per ha—then, the size of the farm does not matter.
There is of course both uncertainty and variability of prices. For example, the price of potatoes of 0.16 € kg−1 will not be realized in all years; likewise, the prices of inputs may vary over time. Taken together, the profitability of an individual farm may differ from our calculations.
Our shadow prices for pollution are such that the effect on sustainability as measured with social profit is not very large. This gives rise to the feeling that there is a discrepancy between the intuitive importance of reducing pesticide use and the magnitude of the calculated effect on sustainability. This may be due at least in part to how we calculate the environmental cost of using pesticides. EIQ is probably the best measure available [78
]. However, it provides an underestimation because it does not take into account the external costs associated with acute and chronic pesticide poisoning of humans, the long term effects on the environment, and synergetic and multiplicative effects of the use of pesticides. There is insufficient data available to take these mechanisms into account.
Some olive orchards in Greece may be located close to high-value natural areas. If so, the shadow prices that we used may underestimate the true cost.
Policy-makers are increasingly using economic valuation when making decisions and people are increasingly receptive to the idea that the costs and benefits of a production activity can be considered in monetary terms [79
]. Nevertheless, a main weakness of economic valuation lies in the complexity of the methodologies and the modelling techniques. Also, the adoption of multiple assumptions often reduces credibility of the estimates. Finally, the limited knowledge about the environmental and socio-economic consequences caused by some externalities and about the time-spans at which they operate (e.g., future impacts of pesticides), results in the exclusion of external costs that cannot be accounted for in the valuation exercise.
Moreover, to which extent the future external costs should be discounted to estimate an optimal intergenerational shadow price, is still an issue which is currently subject to lively debate. This has not to do with the choice of the economic valuation method but with ethical considerations.
We have mentioned four PA technologies in potato production but additional technologies are in the pipeline. Variable rate planting of potatoes in response to soil texture is a promising technique [80
] even though the mechanism through which it can increase yield is only partly understood [81
] but there are indications that by applying a variable planting rate the yield of marketable tubers may be increased by 5%. Also in the pipeline is that the resolution at which measurements and variable rate applications are made will increase and this will result in greater savings. Currently, measurements in potato production are typically made at 30 × 30 m2
but with ever-decreasing cost of sensors and application technology, in the future a resolution of 1 × 1 m2
seems within reach. It has been calculated that in the case of potato haulm killing, increasing the resolution from 30 × 30 m2
to just 15 × 15 m2
will reduce herbicide use from 2.2 to 1.7 L ha−1
] which is a reduction of 60% relative to conventional application. Similarly, for soil herbicides a reduction of 50% and for late blight a reduction of 40% seems possible (relative to conventional application). For sidedress N a further reduction in N use is not possible, but it is expected that real-time soil NO3
measurements using new sensors will allow further reductions in N use.