PROBLEMS AND POTENTIAL ECONOMIC IMPACT OF PRECISION FARMING

Peter Wagner
Technische Universität München
Professur für Unternehemensforschung und Informationsmanagement
Alte Akademie 14
85350 Freising-Weihenstephan, Germany
phone +49-(0)8161-713406, fax +49-(0)8161-713408
e-mail: wagner@landw.uni-halle.de

ABSTRACT

Spatially-variable crop production, often known as precision farming, is being approached with a variety of methods which could be summarized as "mapping systems" and "sensor systems".

It is likely that future, successful implementation of precision farming will rely on a combination of the mapping approach and the sensor approach.

But, before this becomes to be true many problems have to be solved. Major problems to cope with are the accuracy of the GPS-signal, finding algorithms for processing and compressing data, developing tools for economic analysis in a sense of grid-based cost accounting, developing software for sensing weeds (digital image analysis) and, last but not least, combining the mapping and sensor approach to generate application maps.

However, the economic benefit of precision farming is still unknown. Positive effects may be caused by site-specific N management, adopted seed sowing rates and reduction of herbicide treatment. On the other hand, increasing yields up to three and more per cent are described. It has to be noted that the more heterogeneous a field appears, the more chances exist that precision farming will be profitable to the farmer.

INTRODUCTION

Precision farming - i.e. quantifying sowing, fertilizing and spraying according to soil variation and plant population. This requires the recording of even small spatial differences in the factors relevant to crop growth, such as the quality of soil, the availibitly of water and fertilizers, or crop yield, just to mention a few. The recording of these variables and the spatially differentiated use of these factors of production is performed by electronically guided machines and implements that receive the signals for exact in-field positioning from GPS-satelites. This allows greatly improved efficiency of the resources made use of, leads to reduced waste of inputs and, in addition, improves the adjustability of biological-technical systems.

Figure 1 shows that precision farming makes accurate information, precise work and local application possible. Sensors register information such as required working time, tractor hours, fuel consumption, yields, periods of waiting (downtime). Work precision is achieved by electronically controlling the implements. Depending on the given varying conditions, the working depth, the precise maintaining of the working width as well as the working speed and connecting turns of the implements can thus be regulated with accuracy even within small-sized grids. This implies a guarantee that the applied amounts of input correspond exactly to the amounts required. Thus seeds, fertilizers and pesticides are applied as wanted. In the same way, soil management can be regulated depending on small-scale spatial conditions. The same is true for local application. According to the variables measured by sensors, e.g. the current supply of nutrients and available soil moisture, and according to the reduction of nutrients in the previous year and the yield targets one can determine which inputs ought to be applied in which amounts and at what time. Depending on the position of the implement one could easily observe the established limitations e.g. for water protection areas.


Fig. 1: Possibilities and Potential of Precision Farming

All this is based on the fact that soils and growing conditions are subject to considerable variation even within very small plots.This finding is not knew. To the author's knowledge the first publication on precision farming is from 1929. There LINSLEY/BAUER (Compare figure 2) noted that "(T)he soils of this state, often within a field, vary wildly in their need for limestone" and "(I)t is important, therefore, that detailed tests be made of the field so that limestone may be applied according to the need for it." At that time, however, the possibilities of positioning and of electronic regulation and steering that we have today and will have in the near future did not exist. Only today's technology enables the realisation of what was clearly perceived 7o years ago.


Fig. 2: First Publication Concerning Precision Farming

PROBLEMS OF PRECISION FARMING

Precision farming is being talked about everywhere and many manufacturers make at least some efforts to offer information on the possible use of their machines for precision farming. However, quite a number of problems must still be solved before numerous farmers can practice precision farming as it is understood in this article. In the following we will discuss a choice of five problems , namely

Precision of Positioning

Different grades of precision are required for different tasks as the examples in figure 3 can show. For navigation, i.e. looking for fields in the case of co-operative machinery use, precision of +/- 1o meter is sufficient. For field work and the necessary procurement and documentation of information +/- 1 meter is required, whereby in particular a small working width as in combine harvesting can cause considerable problems and make greater precision desirable.


Fig. 3: Accuracy Needs of Precision Farming (Auernhammer, 1998)

If one considers automated vehicle guidance to ease driving, e.g. in case of connecting turns (whith a wide working width) or in combine harvesting, precision of +/- 1o centimeters is required. Precision must be even higher when e.g. implements for weed control are to be guided automatically.

As shown in figure 4 precision technology has developed rapidly in the last few years. In particular the transition from GPS to DGPS has led to some progress in this respect. The precision achievable today amounts to +/- 1-3 meters for DGPS in dependence on the distance from the station of reference.The greater the distance from the station the less accurate the measurement of the actual position is. Recently an additional problem has turned up more and more often: the nearer the field worked at is to the transmitter stations of mobile telephone networks, the more the signal will be distorted. In the immediate neighbourhood of such stations exact positioning is partially no longer possible.


Fig. 4: Analysis of Accuracy of GPS and DGPS (Auernhammer et al. 1998)

Figure 5 shows the difference between theoretical measurements and the actual readings for combine harvesting. At a driving speed of 1.8 meters per second, a working width of 5 meters, and measurements taken every second, considerable deviations can become apparent. Here algorithms suiting the required precision must be found to correct the overlapping and the gap of measurement cells.


Fig. 5: DGPS sensing of Combines (Auernhammer, 1998)

Combination of Mapping and of Sensor Systems

Basically two different approaches to precision farming are under discussion: The first is the mapping approach with mapping systems, the second is the sensor approach with sensor - ( real time) - systems.

The way in which mapping systems function is shown in figure 6. Past data is used in order to determine the necessary inputs for the current situation. Conclusions are drawn from the yields of former years and from the nutrients that are measured in the soil as to the amounts of fertilizers and seeds to be applied. In this way the application of herbicides can also be controlled. This approach is particularly well suited for low-yield areas, and thus for phosphate and potassium fertilizers, relatively constant weather conditions and exclusive cereal crop rotations. For high yield nitrogen fertilization the system reaches its limits, since nitrogen fertilization as a rule must be tuned to recent parameters rather than to past conditions.


Fig. 6: Mapping-Systems


Fig. 7: Real-time Systems


Fig. 8: Real-time-Fertilization

Real-time systems stand in contrast to the mapping systems. They apply inputs, especially nitrogen fertilizers, according to the needs of the plant population at the given moment. Sensors attached to the tractor give information on available soil moisture and the current plant N supply. The application of nitrogen fertilizer is based on such data. The principle of the approach is shown in figure 7. Real-time systems are suited for high yield areas with starkly varying weather conditions and are suitable for varied crop rotations, but basically exclusively for nitrogen fertilising. Figure 8 shows how one such system functions.

Both mapping and the real-time systems per se have advantages and disadvantages. Real progress can be expected by linking both systems together as shown in figure 9. By means of yield maps and soil nutrient maps this approach attempts to explore the yield potentials meter by meter and apply nitrogen fertilizers according to the yield potential and the current conditions. By this approach the current situation of the plant population can be optimally dealt with and ecological and economic limitations taken into consideration at the same time. The approach is suited for high yield regions with starkly changing weather conditions and for fertilizing with nitrogen, phosphate, and potassium. Thus the approach is highly suitable for regions in Europe. A large part of the problems connected with it have not been solved yet. On September 1, 1998, a team at Weihenstephan started an interdisciplinary research project with the aim of advancing this approach in the next 6 years. Particularly the development of sensors and models for the deduction of required amounts of fertilizer depending on the growth state of plants and the available soil moisture has just begun recently.


Fig. 9: Real-time Systems with Map-Overlay

Processing and Condensation of Accumulated Data

The probable volume of accumulated data in precision farming with automatic data collection per hectare, respectively a farm of 3oo hectares, is presented in figure 1o. A certain working width and a working speed were assumed for each of the following five activities: combine harvesting, tilling, sowing, spraying, and fertilizing. Working width and working speed of the activities define acrage performance. It is based on the assumption that measurements are taken each second. The number of data sets is computed by projection onto one hectare. The number of data attributes per set varies according to the measure taken. Each data set contains an identical header defining the type of data set, time, longitude, latitude and altitude and also allowing a statement on sensing quality. The number of further attributes dependends on the performed task aktivity. For example, in the case of cereal harvesting such attributes can be the actual cutting width, the operational position of the cutting unit, the distance covered, and also yield related data such as amount of yield, grain losses or grain moisture. If the implements are provided with appropriate sensors further attributes can be considered. The number of data sets per year and hectare as well as the number of attributes per year and hectare can be computed from the number of tasks per year. The last column in the figure shows the number of bytes per year and hectare based on a data length of 6 bytes per attribute. In this case 1.3 megabytes would result. The annual volume of data for a 3oo-hectare farm would thus amount to 4o2 megabytes or almost half a gigabyte.

Activity Working width (m) Working speed (m/sec) Area (Performance) (m2) Number of Datasets / Hectare and Task Number of Attributs / Dataset 1) Number of Attributs / Hectare and Task Number of Tasks / Year Number of Tasks / Year and Hectare Number of Attributes / Year and Hectare Bytes / Attribute Bytes / Year and Hectare
Grain Harvesting (Combine) 5 1,4 7 1.429 20 28.571 1 1.429 28.571 6 171.429
Tilling 3 2 6 1.667 25 41.667 2 3.333 83.333 6 500.000
Sowing 3 2 6 1.667 25 41.667 1 1.667 41.667 6 250.000
Spraying 15 2 30 333 35 11.667 3 1.000 35.000 6 210.000
Fertilizing 15 2 30 333 35 11.667 3 1.000 35.000 6 210.000
Total / ha  8.429 223.571   1.341.429
  (1,3 Mbyte)
For a 300 ha Farm  2.528.571 67.071.429   402.428.571
  (402 Mbyte)
1) Estimate, includes header with: Type of dataset; geographic longitude, latitude, altitude; sensing quality

Fig. 10: Data Volume through Precision Farming and Automated Data Logging (Estimated)

This volume of data is, of course, impracticable. So suitable algorithms for the condensation and the computing of data must be found to reduce the data volume noticeably. However it is still largely unknown which data are expendable.

Finding the Optimal Amount of Applied Input

With the systems available on the markets today the farmer alone decides how much input, e.g. nitrogen, will be applied to a site. This means that experience and intuition determine his decision. So far there are no reliably functioning models capable of computing the optimal amount of fertilizer with reference to the parameters of the soil and the plant population in small-sized grids. Such models are required, however, if we want to reduce the size of individually managed parts of fields and the more heterogeneous the known conditions of a location are. In some regions of Germany, for example, the quality of soil, the supply of nutrients and the availabilty of water vary greatly within a few meters.We can react to such variations only by managing the parts of fields on as small a scale as possible, for instance by keeping to the working width of 5 x 5 or 1o x 1o meters of combines.

The models to be developed must be capable both of taking into account data from the past, e.g. in order to establish the yield potential for one part of a field, and of processing current conditions such as the sensor probed nitrogen supply of the plants or the sensor registered water availability of the grid in question. Setting up such models is a component not to be underestimated in the acceptance of precion farming by the farmers themselves.

Models of Economic Analysis

Models for the economic analysis of precision farming do not exist yet. Commercial field record systems usually do not go beyond the gross margin level and so, at best, no more than the gross margin for a plot can be computed. Thus it is neither possible to calculate the gross margin for individual parts of a field, nor for homogeneous groups of such parts. From the point of view of farm management, one can conceive models within the framework of full cost accounting, which allow the analysis of any spatial part of a field. This is particularly important for the computation of the economic efficiency of precision farming. Whether such models necessary for the purposes of research will be put into practice is uncertain.

These considerations have shown that there is a long list of problems that must be solved before precision farming will be generally practiced.

THE ECONOMIC EFFICIENCY OF PRECISION FARMING

To run their farms on the principles of precision farming farmers must be willing to invest in technology and services. Basically there are three kinds of investment:

Some of the components are presently available, others will be available in the near future. As mentioned above it is the software based on the "application logic" of precision farming that is posing major problems.


Fig. 11: Determination of the Input Level for a Field (e.g. Fertilizer)

On the other hand there are the benefits of increasing yields and/or lowering the quantity of inputs.

The higher revenues and the potential of lowering costs ought to cover the extra cost of investment. The reasons for the additional profit and the cost reduction lie in the different approaches of undifferentiated (traditional) farm management and of precision farming. Figure 11 shows the undifferentiated approach of farm management concerning the application of fertilizers to one plot (or generally: the definition of the input level applied). The fertilizer application rate is determined equally for the whole field by the average yield expectation, here of 7o decitons per hectare, and the difference between the nitrogen available in the soil and the amount of nitrogen required by the yield expected.


Fig. 12: Determination of the Input Level for a Field (e.g. Fertilizer)

Figure 12 shows how the application rate of nitrogen for each grid is determined according to the approach of precision farming. The various parts of the field have different yield potentials, for each of which the locally required application rate of fertilizer can be determined. There are areas of higher and lower yield potential, which must be supplied with more or less nitrogen as required.


Fig. 13: Possible Effects of Precision Farming - Heterogeneous Location

Figure 13 presents possible effects of part-field related management for a heterogeneous site. The yield potential of this field varies from 56 decitons to 84 decitons/ha indicated by the line falling from left to right (dispersion of heterogenity +/- 2o% : 7o dt +/- 14 dt). Within the framework of yield mapping the site is divided into 5 equally sized classes of yield heterogenity each comprising 2o% of the whole site. This is indicated on the abscissa. So far the site (thick horizontal line) has been fertilized uniformly targeting 7o decitons/ha. In consequence on the one hand yield was forgone in high yield potential parts (Class I, II and 5o per cent of III) and on the other fertilizer was wasted in low yield potential parts ( 5o per cent of III, IV and V). By differentiated site management more yield can now be realised and waste of fertilizer reduced. In the present case the overall amount of fertilizer applied to the site remains the same since the smaller amount applied to parts of low yield potential is compensated for by the additional amount used for high yield areas. But this approach increases the overall efficiency of the applied inputs, here of nitrogen. These theoretically deduced findings correspond to praxis in many cases. On a 28-hectare site soil quality figures (best quality equals 100, lowest quality equals 0) of 19 - 7o were discovered (ALBERT, 1997). With Nmin probes on a neighbouring plot values of 21 -75 kg nitrogen were measured in one site (STENGER et al., 1993, p. 305). However there are also sites of lower heterogenity, i.e. comparatively homogeneous locations. The effects of precision farming on such locations are not so obvious. This can be seen in figure 14 by the example of a location having a minimum yield potential of 66.5 decitons/ha and a maximum of 73.5 decitons/ha. Here the potential yield increase is considerably lower as well as the savings of fertilizer on plots with a lower yield potential (dispersion of heterogenity +/- 5 % : 7o dt +/- 3,5 dt).


Fig. 14: Possible Effects of Precision Farming - Approximate Homogeneous Location

In practical farming one can observe that farmers do not apply fertilizer according to the mean yield potential of their fields but with regard to the yield potential of the best parts. The conditions in figure 15 are initially the same as in figure 13 with the exception of fertilizing towards 75 decitons/ha instead of 7o decitons. As one can see the orientation is towards the higher yield potential. This implies that the yield increase due to precison farming will be less in case of site specific management while there is a real saving of fertilizer. This comes closest to farming as practised today: The yield increases from documented experiments in most heterogeneous locations hardly ever reach more than 5 % in comparison to uniform management, while nitrogen efficiency could be improved (EHLERT/WAGNER, 1997, p. 27). That is to say that in major parts of the sites fertilizer was wasted. Precision farming shows beneficial results above all in production factor saving resp. in the higher efficiency of factor use. The yield increase that is aimed at will be rather low in comparison to uniform field management.


Fig. 15: Possible Effects of Precision Farming with High Past Fertilization Levels - Heterogeneous Location

Figure 16 shows the results of an analysis of literature dealing with the economic consequences of precision farming. Studies of herbicide cost reduction show that between 5o to 8o % of the costs for herbicides can be saved when treating only patches where weeds actually grow. The savings in terms of money depend greatly on the herbicide price, so that a generalization is not possible here.

Author research object results
Green et al.
(USA, 1997)
spatially variable herbicide
treatment in peanuts
up to 70 % less
herbicide use
Nordmeyer / Häusler / Niemann (Germany, 1997) spatially variable herbicide
treatment in cereal grains
up to 80 % of the area need
not be treated
Gerhards
(Germany, 1998)
spatially variable herbicide
treatment in cereals
40-50 % less
herbicide use
Harris
(England, 1997)
spatially variable sowing, fertilization and spraying in wheat and potatos EURO 50-60 economic
advantage / ha with wheat,
EURO 240-250 with potatos
Ostergaard
(Denmark, 1997)
spatially variable N,P,K and
lime application in cereals
$ 40-50 economic
advantage / ha
Schmerler/Jürschik
(Germany, 1997b)
spatially variable
N-fertilization in cereals
up to 3,9 dt/ha increased crop yield. On average 25 kg/ha less N with heterogeneous fields.
Results in more than EURO 25 economic advantage / ha.
Swinton/Ahmad
(USA, 1996)
spatially variable
N-fertilization in sugar beets
74 $/acre (~160 EURO/ha)
economic advantage and quality increases
Reetz/Fixen
(USA, 1995)
spatially variable
N-fertilization of all crops
on a farm
43 $/ha economic
advantage
Malzer et al.
(USA, 1996)
spatially variable
N-fertilization in corn
11-72 $/ha economic
advantage
Schmerler/Jürschik
(Germany, 1997a)
calculated costs for GPS use (machinery and labor) for a 2000 ha farm increased costs of
35-40 EURO/ha and year1)
Harris
(England, 1997)
calculated costs for GPS use (machinery equipment) for a 320 ha farm increased costs of
30-35 EURO/ha and year2)
1) incl. labor costs
2) without labor costs

Fig. 16: Economics of Precision Farming

Several studies have dealt with yield benefits and saving potentials of inputs as shown in the centre of the figure. SCHMERLER and JÜRSCHIK (1997 b, p. 995), for example, show yield increase (wheat) of not quite 4 decitons per hectare and an average reduction of 25 kg nitrogen/ha on heterogeneous sites. All in all this amounts to about EURO 5o in yield benefits and saving potentials per hectare. Other authors arrive at values of between EURO 5o and 6o with wheat or up to EURO 25o with potatoes (HARRIS, 1997, p. 953). With sugar beets SWINTON/AHMAD (1996, p. 1015) found yield and quality benefits as well as saving potentials as high as EURO 16o/ha.

On the other hand there are the costs of precision farming. For investments in the (imperfect) technology HARRIS (1997, p. 953) amount to EURO 3o to 35 per hectare for large farms (3oo ha) at present conditions. One can expect that these costs will decrease with the simultaneous improvement of the technology.

A final statement on the excellence of precision farming from the economic point of view cannot be made on the basis of the available studies. Generalizing, one can say, however that the higher the heterogenity of the farmed location the more obvious the economic benefits of precision farming are. Intensive work is being done on remedying the imperfections of technology as well as on the decison criteria, for instance, for the rate of nitrogen fertilizing. It must be emphasized that the results of research are undoubtedly in favour of site-specific management. In the discussion on the applicability of the new technology one ought to take into account that not only questions of the economic viability for the individual farm should be considered but also that precision farming allows great progress in the reduction of the pressure that agriculture exerts on the environment. This applies particularly to the reduction of nitrogen leaching and to herbicide savings.

There are additional effects of precision farming as well. Through automatised data acquisition the management of biological-technical systems will be as much improved as overall farm management. Significant progress in farm management can be expected. Simultaneously, so to say, as a spin-off, the farmer who has decided in favour of precision farming is provided with the documentation of his activities, which in Germany is required by fertilizing regulations and helps with applications for funds.

A further effect is that e.g. that soil compaction can be easily registered by sensors probing soil resistance in ploughing which allows immediate redress by on-the-spot reaction.

The technology of precision farming is also suited to make fleet management for optimization of co-operative machine use more efficient.

Precision farming is suitable not only for large farms but also for the management of minor fields in regions with small-scale agricultural structures. In this case the same crop can be grown beyond property borders on sites of various ownership, because the automated registration of data from input application and harvesting makes individual settlement of accounts for all owners possible. This needs the above mentioned site-specific cost accounting.

Last but not the least we should also think of the changing ways of land use of tomorrow - with the technology of precision farming farm robots seem to become more real.

Precision farming is far more than site specific application of production factors, it is economically and ecologically optimised farming that allows better management.

REFERENCES:

ALBERT, E. (1997): Sind für die Pflanzenernährung Teilflächen wie verschiedene Schläge zu betrachten? In: BUSCHE, Ph. (Ed.): Landwirtschaft braucht Zukunftstechnologien, DLG-Wintertagung 1997, Frankfurt. AUERNHAMMER, H. (1998): Precision Farming - Schlagwort oder nachhaltige Landwirtschaft von Morgen? Presentation at the Agricultural Seminar, 26.5.1998, Weihenstephan. EHLERT, D. and WAGNER, N. (1997): Schluß mit dem Gießkannenprinzip. In: DLZ-agrarmagazin, Sonderheft 10, p.24-27. GERHARDS, R. et al. (1997): Site Specific Weed Control in Winter Wheat. In: Journal of Agronomy & Crop Science 178, p.219-225. GREEN, H.M. et al. (1997): Precision Management of Spatially Variable Weeds. In: STAFFORD, J.V. (Ed.): Precision Agriculture 1997, Volume II: Technology, IT and Management, Oxford, p.983-989. HARRIS, D. (1997): Risk Management in Precision Farming. In: STAFFORD, J.V. (Ed.): Precision Agriculture 1997, Volume II: Technology, IT and Management, Oxford, p.949-956. MALZER, G.L. et al. (1996): Spatial Variability of Profitability in Site-Specific N Management. In: ROBERT, P.C., RUST, R.H. and LARSON, W.E. (Eds.): Proceedings of the Third International Conference on Precision Agriculture, Minneapolis, p.967-975. STERGAARD, H.S. (1997): Agronomic consequences of variable N fertilization. In: STAFFORD, J.V. (Ed.): Precision Agriculture 1997, Volume II: Technology, IT and Management, Oxford, p.315-320. REETZ, H.F. and FIXEN, P.E. (1995): Economic Analysis of Site-Specific Nutrient Management Systems. In: ROBERT, P.C., RUST, R.H. and LARSON, W.E. (Eds.): Proceedings of Site-Specific Management for Agricultural Systems, Second International Conference, Minneapolis, p.743-752. SCHMERLER, J. and JÜRSCHIK, P. (1997 a): Lohnt sich der Aufwand? In: DLG-Mitteilungen, H.6, p.49-51. SCHMERLER, J. and JÜRSCHIK, P. (1997 b): Technological and Economic Results of Precision Farming from a 7200 Hectares Farm in Eastern Germany. In: STAFFORD, J.V. (Ed.): Precision Agriculture 1997, Volume II: Technology, IT and Management, Oxford, p.991-995. STENGER, R., PRIESACK, E. and BESSE, F. (1993): Räumliche Variabilität von Nmin-Gehalten in Ackerflächen des FAM-Versuchsgutes Scheyern. In: REINER, L., GEIDEL, H. ans MANGSTEL, a. (Eds.): Expert-N und Wachstumsmodelle - Referate des Anwenderseminars im März 1993 in Weihenstephan. Agrarinformatik, Bd. 24, Stuttgart, p.301-309. SWINTON, S.M. and AHMAD, M. (1996): Returns to Farmer Investments in Precision Agriculture Equipment and Services. In: ROBERT, P.C., RUST, R.H. and LARSON, W.E. (Eds.): Proceedings of the Third International Conference on Precision Agriculture, Minneapolis, p.1009-1018.