Computers and Electronics in Agriculture 73 (2010) 4455

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture

journal homepage: www.elsevier.com/locate/compag

A user-centric approach for information modelling in arable farming

C.G. Sorensen a, , L. Pesonen b , S. Fountas c , P. Suomi b , D. Bochtis a , P. Bildse a , S.M. Pedersen d
a

rhus University, Faculty of Agricultural Sciences, Department of Agricultural Engineering, Research Centre Foulum, Blichers Alle 20, 8830 Tjele, Denmark MTT Agrifood Research Finland, Vakolantie 55, 03400 Vihti, Finland c Center for Research and Technology, Institute of Technology and Management of Agricultural Ecosystems, Technology Park of Thessaly, 1st Industrial Area, GR 385 00, Volos, Greece d Institute of Food and Resource Economics, University of Copenhagen, Rolighedsvej 25, 1958 Frederiksberg C, Denmark
b

a r t i c l e

i n f o

a b s t r a c t
Agriculture and farmers face a great challenge in effectively manage information both internally and externally in order to improve the economic and operational efciency of operations, reduce environmental impact and comply with various documentation requirements. As a part of meeting this challenge, the ow of information between decisions processes dened as realizing a decision must be analyzed and modelled as a prerequisite for the subsequent design, construction and implementation of information systems. This paper denes the actors, their role and communication specics associated with the various decision and control processes in farmers information management. Core-task analysis and core task demands from earlier research are utilised as premises for the modelling of information ow from the farmers point of view. A user-friendly generic FMIS design reference model is the primary objective for the study in which planning, execution and evaluation measures have been incorporated. A user-centric approach to model the information ows for targeted eld operations is presented. The information models are centred around the farmer as the principal decision maker and involves external entities as well as mobile unit entities as the main information producers. This is a detailed approach to information modelling that will enable the generation of a Farm Management Information System in crop production. 2010 Elsevier B.V. All rights reserved.

Article history: Received 21 August 2009 Received in revised form 30 March 2010 Accepted 8 April 2010 Keywords: Precision agriculture Fertilisation Core-task analysis FMIS DFD models

1. Introduction The analysis of decision processes, as well as information modeling for eld operations is not a new approach. Decision-making is an important aspect in farm management and has been studied by numerous studies (e.g. Anderson et al., 1980; Van Elderen and Kroeze, 1994). The reasoning of why there is a need to analyze decision-making has been addressed by Gladwin (1989), who argued that the benet is to know and understand why a specic group of people acts as they do. This will enable researchers to provide the farmers with supporting knowledge and tools as a way to enhance decision-making at specic stages of the process. In agriculture though, farmers, in general, both generate and execute any plan made, and their decision process associated with the planning remains very much implicit and internal (Srensen, 2000) and often make decisions based on their intuition and not using formalized planning tools. That is contradictory to the industry, where there is a long tradition for explicit planning comprising formalised documents passed down to the shop oor by the management section

Corresponding author. Tel.: +45 89991930. E-mail address: Claus.Soerensen@agrsci.dk (C.G. Sorensen). 0168-1699/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.compag.2010.04.003

for implementation (Chary, 2006). The efforts aimed at developing agricultural planning support must be targeted at externalising and formalising the farmers planning effort. Kay and Edwards (1999) discussed the unique attributes that make farm business complex in comparison to the industry, such as the biological processes, the xed supply of land, the small size, weather forecast and the perfect competition (Runge, 2006). They argued that the systematic analysis of decision-making process would not necessarily lead to perfect decisions, but would help a farm manager act in a logical and organized manner when confronted with choices. The US North-Central Regional Research team in Farm Information Systems (2000) categorized the farmers into two groups: information hogs seeking and using large amount of information and seat of the pants where personal intellect and intuition are the main drivers in decision-making. They observed that seat of the pants farmers most likely utilize information in ways not fully understood by researchers or advisors. It is perhaps useful to recognize intuition as a complex result of a given farmers unique experience and familiarity with his/her farm. Much of the information exists as tacit knowledge of the farmer, but in order to specify all the elements it is necessary to explicitly specify the detailed information ows for individual planning tasks.

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

45

With the advent of Precision Agriculture (PA) technologies farmers acquire a vast amount of data and they face even bigger problems on how to effectively utilize them to make better decisions (Auernhammer, 2001). PA aims to change the focus of agricultural production from quantity to quality and sustainability (Jensen et al., 2000). By generic denition, PA refers to agricultural techniques that increase the number of (correct) decisions per unit area of land per unit of time, with associated net benets (McBratney et al., 2005). When practising PA, a farmer manages crop production inputs (seed, fertiliser, lime, pesticides, etc.) on a site-specic basis to increase prots and crop quality, but also to reduce waste and maintain environmental quality. In order to make precise decisions in different phases of the farming process, the farmer therefore needs to analyse information from different vast and dispersed located information sources. Management of the information and decision-making is the core issue for the farmer in successful PA, not the data acquisition process. Fountas et al. (2006) have decomposed the decision-making process for farmers using PA into twenty-one decision analysis factors. These factors were assembled into a data ow diagram describing the main information processes and ows. The diagrams analyzed the information ows for eld operations taking a general approach to represent the transformation from gathered data into information and then decisions. Nash et al. (2009) modeled all range of data ows covering the broad spectrum of PA practices into one very large diagram showing the interrelations, while also including the modeled data-streams for specic PA practices, such as management zones, yield mapping or exploitation of remote sensing data. The required information modelling can be fullled through concentrated efforts aimed at extracting domain knowledge and deriving information ows at various planning and process levels. This effort demands considerable research and development, which is the case in terms of incorporating user preferences and requirements. The tendency to use a more user-centric approach in developing new technologies has gained considerable appeal (e.g. Akao and Mazur, 2003; Norros, 2004). The core-task analysis (CTA) is a user-centric methodology, which was initially developed in Governmental Technical Research Centre of Finland (VTT: Valtion Teknillinen Tutkimuskeskus) (Norros, 2004). It is a functional modelling technique that informs system modellers of the aims, intrinsic constraints and user practices in the work under study A user-centric approach assumes that the users ideas and requirements reactions concerning the specic characteristics of the designed technology are integrated in the subsequent design. When end-users and other actors in the value chain are involved into the design and development process from its early stage, the system becomes more realistic to realise and build in real world, and it readily meets most of the user requirements. Since the adoption of Decision Support Systems (DSS) and Farm Management Information Systems (FMIS) within PA has been disappointingly low (Rosskopf and Wagner, 2003; McBratney et al., 2005; Parker, 2005), this kind of user-centric development method is expected to improve the acceptance of the new technology in the markets and among end-users. This leads to smaller risk associated with introducing new FMIS in the farm business domain (Norros et al., 2009), which was communicated in a research project, InfoXT1 user-centric mobile information management in automated plant production running from 2006 to 2008 in Scandinavia (Pesonen et al., 2008). Here, the applicability of using a user-centric approach to develop an information system for mobile work units was indicated.

The aim of this paper is to dene the actors, their role and communication specics associated with the various decision and control processes in farmers information management. The coretask analysis involving farmer interviews and derived Core-Task demands is the basic framework for the pursued approach. A userfriendly generic FMIS design reference model is the primary target for the study where both planning, execution and evaluation measures are incorporated. The generic design is intended to support and guide the actual implementation of a specic FMIS in terms of capability, invoking of information and communication technologies, etc. The design reference model is developed in the early phases of system design and the detailed decomposition and component construction can be derived from this model. This study was part of an on-going EU research project FutureFarm.2 FutureFarm has dened aims at meeting the challenges of the farm of tomorrow by integrating Farm Management Information Systems (FMIS) to support real-time management decisions and compliance to standards. 2. Farm management and eld operations The agricultural production processes within arable farming involve transformation processes that are realised by biological processes (e.g. crop biomass growth) taking place in the course of the growing season. The processes are regarded as an autonomous system, which is basically independent of decisions made by the farmer. In contrast to this, the intervention realised by labour and machinery during the plant growing process is dependent on decisions made by the farmer and termed an operation. An formal denition of an operation is given by Van Elderen (1977), who states that an operation is a technical coherent combination of treatments by which at a certain time a characteristic change of condition of an object (a eld, a building, an equipment, a crop) is observed, realised or prevented. This denition extends operations beyond those for crop production to supporting enterprise functions like maintenance, repairs, etc. An operation is generally seen as the link between some resources (e.g. labour and machinery), some materials processed, and some material produced (e.g. harvested crops, repaired machine, etc.). The decomposition of information processes attributed to the planning and execution of eld operations is based on the management functions ranging from strategical to operational planning, execution control and evaluation, and a number of underlying processes and sub-processes. All planning levels have to be included in a generic FMIS as it is necessary to know what kind of information the system has to be able to handle. Farmers cannot have separate systems for each management level. All levels utilise/need data/information produced in the other levels. The integration of all planning levels is pivotal to the usefulness of the FMIS. Fig. 1 outlines the basic management processes which are identied within the agricultural plant production cycle for both manned and unmanned machinery items. Plan generation and execution must be linked in a system monitoring effects of actions, unexpected events and any new information that can attribute to a validation, a renement, or a reconsideration of the plan. Plans must be presented conditionally, so that supplementary knowledge from observations, farm databases, sensors, etc., can be incorporated in order to revise plans. It should be noted, that although that the concept of farm databases is an important issue in the modelling of a FMIS, it is not within the scope of the paper. The pursued concept, in principle, does not make any difference between information in a database

www.mtt..infoxt.

www.futurefarm.eu.

46

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

Fig. 1. Information and planning activities in agricultural operations management with the identication of the revised task formulation to be invoked in the case of autonomous vehicles (adapted from Goense and Hofstee, 1994).

and the information present in the memory of the farmer. It is all available information which can be drawn upon and where the distribution of the available information between a physical database and the memory of the farmer depends on the degree of automated decision-making and as such, the degree of explicitly formulated information storage. 3. Methodology The information modelling used in this study utilises the usercentric research and designing method, the so-called core-task analysis (CTA) method, developed in earlier work carried out in Scandinavia (Norros, 2004; Nurkka et al. 2007; Pesonen et al., 2008). The method employs ISO 13407 standard (ISO 13407, 1999, Human-centred design processes for interactive systems) aiming at good systems usability by integrating the end-users to the development process from the beginning. The method consists of seven phases (Fig. 2) and it is based on core-task analysis (CTA) of farm work. The phases of the method are shown in Fig. 2. CTA is a functional modelling technique that informs system modellers of the aims, intrinsic constraints and user practices in the work under study. In the method, it is seen that a result-oriented and meaningful human-environment interaction is a functional system in which the environment provides possibilities (affordances) that human actors learn to grasp (prehensilities) (Norros and Savioja, 2006).

Fig. 2. Seven phases of the core-task analysis-based user demand modelling system (Nurkka et al., 2007).

Affordances and prehensilities of a particular domain and the aimed results constitute the core-task of that activity. The modelling of the core-task gives the potential for action in the particular context. The CTA method was developed further and applied in a study that aimed to improve information management of high quality cereal (malt barley) production (Nurkka et al. 2007). The CTA method attained the form where it consists of two parts: research and design (Fig. 2). The methodology employed in this study, utilises basic work of phases 14 of Fig. 2 for targeted farm operations. Science-based modelling of a core task was the rst phase of the modelling relying on expert knowledge of the use of PA practices within crop production. Scientic and professional literature, four expert interviews and several workshops with the research group provided the data for the modelling. The aim was to indicate the content of the relevant and important information regarding the decision-making for the targeted eld operations under study. This information comprised the information used in planning, execution and evaluation, together with the core-task demands. The core-task demands dene how farmers use skills, knowledge and collaboration to control dynamicity, uncertainty and complexity of crop production work (Nurkka and Norros cit. Pesonen et al., 2008). Analysis of orientation (phase 2) was carried out in Finland among 11 interviewed producers to identify those farmers who are keen on constantly improving eld processes to gain good product quality and environmental friendly methods, in other words the farmers who possess PA orientation (Nurkka et al., 2007). In the third phase, as part of the practice-based modelling, individual farmer interviews were carried out in a half-day workshop organized at four farms in Finland (Nurkka et al., 2007) and at ve farms in Sweden (Olsson and Rydberg, 2008). During the workshops, the farmers practices were simulated by inviting the farmers to draw their own conceptual models of the growing process based on the experiences of the preceding farming period. The model was to portray the actions they made and tools they used during the process. They were encouraged to make as many comments as possible on the content, origin and type of the information they used and how they used it in different phases of the process. The aim of this phase was to understand what the obstacles are in the use of the systems to support PA and what the actual needs for the farmers are regarding FMIS.

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

47

Fig. 3. Analysis and modelling approach.

In the phase 4 (Integrated Information Modelling), all the data from the preceding modelling were applied. The work was done in workshops comprising agronomists, agrotechnology scientists and farmers. This phase was creative and iterative combining research and practice, and the rst drafts of information models were designed as a result. Ideas about the possible technological advances were taken into account. The modelling was an iterative process having current farm technology as a foundation, but also looking for the possibilities to utilise efciently advanced technologies like ISOBUS and wireless communication. However, at all times, the information ow models were created and revised from the farmers (end-users) point of view.
Table 1 Planning levels and aggregated information ows in eld operations. Planning level Strategic planning or design of the production system: Design of production system for a period of 15 years or 2 or more cropping cyclesspecically the labour/machinery system and selection of types of crops Tactical planning: Setting up a production plan for a period of 12 years or 12 cropping cycles. Information required

A detailed structuring and formalisation of physical entities and the information, which surrounds the planning and control of efcient mobile working units is a decisive prerequisite for the development of comprehensive and effective ICT-system for task management on the farm. The basic idea was to capture the highlevel planning and control activities, which take place in a targeted production section, and represent explicitly the domain knowledge in terms of domain entities and their relationships, (Fig. 3), and based on the basic CTA approach and work presented above The actors are dened as information operatives or as the entities which are capable of storing or processing information by way of the explicitly dened decision processes. The dened actors

Information provided Number and dimensions of machines Machine capacity Labour requirement Crops selected Crop plan Machinery replacement Fertiliser/chemical application plans Maintenance plans Labour budget (peak loads) Required/optional operations Operations urgency Operations specications Work plan for planned operations indicating: Starting time Duration Work-sets required

Possible production levels and price developments Operations demands Possible work methods Available machinery on the market Costs Strategic plan Availability of land, buildings and equipment External/internal standards

Operational planning: Determining activities in the coming cropping cycle, i.e. within the coming season

Scheduling: Work scheduling setting up formulations of jobs. Planning the implementation of work in the short-term.

Tactical production plan Internal/external standards Maintenance plan for land, buildings and equipment Required operations Urgency of operations Soil and crop status Weather forecast Workability criteria Availability of labour and equipment Operations specications Equipment breakdowns Unavailable material Change in soil, crop or weather conditions Priority changes Work time elements (effective time, ancillary time, preparation time, disturbance time, etc.) on work-sets Operations specications

Task formulation: Handling tasks concerning inspection of formulated tasks

Deviation from plans/schedules

Execution: Controlling tasks, and work-sets performancetask control and operation control

Realised work time Realised capacity Set point values for implement Deviations from planned tasks

Evaluation: Comparing planned and actual executed tasks

Realised work time Realised capacity Realised yield Documentation information

48

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

include farm managers, machine controllers, external services, etc. A close coupling of the actors with the decision processes exists as the actors are the executers of these processes in terms of deciding which actions to take in relation to the execution of activities dened as operations (goal-specic work as required by the relevant production system) and tasks (the physical implementation of the operation in terns of resources). The decision processes are inuenced by a number of factors including strategies (e.g. the farmers preferences for a specic production form), triggers (e.g. weather conditions determining the planning and initiation of eld operations), and timing (e.g. the degree of timecritical decision-making, where the operational decision-making is more time-critical than strategic decision-making). The information used in the decision process is the required information for making a rational decision, whereas the information produced by the decision process comprises the planning, guidance and control information used for actual implementing the specic decision. The consecutive decision processes were organised into a diagram with a sequential timespan so that the decision process together with data transfer with different actors form an information ow through the different decision levels; strategic, tactical, operational, execution and evaluation. The farm specic data are stored in a database from which the data are available to a specic decision process whenever needed. It is to be noticed that the timespan will vary according to the decision level as, for example,

a decision process is faster in execution than in strategic planning. The outcome of this modelling is a devised information ow diagram (IFD). The aim of this user-centric IFD is to introduce the actors and their roles as needed in farmers decision-making associated the core tasks. As a result, it was easy to involve the farmers in developing and testing the model. Farmers were able to judge whether the different details of the information ow in the diagram are correct and the proposed system usable for them. In this study, the information model was developed further from that reported in earlier studies. Here, six eld operations were analyzed and modelled as IFD: tillage, seeding, fertilising, spraying, irrigation and harvesting. These are the main eld operations in crop production and have been as the focal point in the development of the FMIS in the FutureFarm project (Chatzinikos et al., 2009). By decomposing the activities inherent in the six eld operations, the information, processes and actors involved in the decision-making were recorded. The earlier modelling work was evaluated, supplemented and expanded to cover the core task of European farm work by FutureFarm project researchers. Whilst the main focus in previous work was in information management of mobile work units, the main focus here was on farm planning and real-time assistance. Thus, the automated compliance to standards and checking of plans was added as a technological advance in the modelling. The information models produced were cross-validated with the information modelling produced by Fountas et al. (2006) with its independent data set.

Fig. 4. Legend describing used symbols and identied actors in information ow modelling.

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

49

4. Results From the six eld operations analyzed in the FutureFarm project, the information model for the fertilisation case is presented in this paper. Table 1 sets the framework for the different decision levels together with the main parts of the information ows required by the decision process or produced by the decision process. By specifying in detail the information provided and the information required for the information handling processes, the design and functionalities of the individual information system elements can be derived. This has involved explicitly specifying tacit knowledge of the farmer as way to extend the FMIS design into automated decision-making. The presentation of the results include a generic schematic illustrates the symbols and the terms used in the information models. Next, the information models for the ve levels of decisions are presented, that is strategic, tactical, operational, execution and evaluation. Fig. 4 shows the scope of the information modelling for the fertiliser case in terms of identied actors to be included in the system. These actors include external entities outside the farm, the farmer as the prime decision maker and the mobile unit entities involved in actual carrying out of the planned tasks, such as the task controller, implement electronic control unit (ECU) and user interfaces like the Virtual Terminal (VT) on the tractor. These mobile unit entities, which are grouped using the dashed line, can be from one to many as during a eld operation more than one fertilizer could be

utilized. Also, the actor identication is applicable to both manned and unmanned mobile unit congurations. Specically, external entities include the market receiving the produce from the farm production system, the advisory system providing advise to the farmer upon requests, the government system setting the rules and provisions for the farm production, the weather service providing weather information either generally published or dedicated weather information upon request, the agricultural service companies like a machine contractor providing the service of executing specic work tasks, and the technology providers like agricultural machine dealers. As compared to the external entities, the internal entities include the farmer as the prime executor of decisions, the farm database as the container of farm relevant information, and the mobile unit entities comprising individual elements of the work machines on the farm. These latter elements involve information and communication measures as well as principal traction units like the tractor. Following the ISOBUS concept, the virtual terminal function as an interface to the ISOBUS compatible implements while the task controller is the prime communication device between the mobile unit and the management system. As the data acquisition part, the internal sensors provide information for the control of the planned task while the external sensors provide information about the eld, environmental indicators, etc. As part of the strategic decision level, Fig. 5 focuses on the longterm potential production dimensions (vision) of the farm and involves a static description or analysis of the whole farm plan-

Fig. 5. First part of strategic planning focusing on the farm development. The information transfer arrows to and from decision process are in a sequence.

50

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

Fig. 6. Strategic planning focusing on the choice of technology according to the relevant standards and management practices.

ning integrating the specics of fertilising. External provisions like environmental requirements and standards for the application of fertiliser will inuence the crop rotation scheme, the necessary technical capabilities of the equipment for fertiliser application, etc. Next, the fertilising strategy in terms of selected fertiliser types, strategies for application, etc. is determined based on an overall evaluation of the external requirements, the prices of production factors and sales products (Fig. 6). The selection of fertiliser equipment and the machinery size or capacity planning concerns both a qualitative and a quantitative selection of machinery items as related to the demand. The optimisation can be done by determining (1) demands put up by the operation to be performed, (2) availability of equipment on the market, (3) possible working methods, (4) dimensions and capacity, (5) costs, and (6) use of own machinery or contractors. Many types of models supporting this optimisation have been launched, from simple deterministic models (see for instance Hunt, 1983) to more complex simulation and linear programming models (e.g. Sgaard and Srensen, 2004). The models involve the interaction between the labour and machinery system and the biological and meteorological system involving crops, soil, weather conditions, etc. The optimisation is done while considering constraints like available labour and machinery, timeliness functions and workability. The strategical planning of machinery needs is strongly interconnected with the operational planning. If this connection is not taken into consideration, a strategically chosen plan could turn out to be non-executable, because it would produce a non-workable schedule.

The determination of the necessary planning information in the subsequent planning levels is considered an integral part of the strategic selection of the fertilising technology. This involves determining the scope of the information and the adherent information technology measures to be available as part of the tactical and operational planning of the fertiliser application. The tactical planning as depicted in Fig. 7 for the fertilising case concerns the determination of the quantity and timing of the operational activity in the medium planning range. The tactical planning involves an all-year setting up of the fertiliser plans and operational plans for the input of labour and machinery. Normative models for the labour and machinery input given a specic crop plan and a specic machinery inventory have been developed (e.g. Srensen and Nielsen, 2005; Achten, 1997). The main decision involve selecting the fertilising features of the application equipment in terms updated information from the technology provider, new adjustment features, etc. as well as acquiring planning information about available fertiliser types, recommendations on fertiliser use in the light of prevailing weather conditions, available external services which will be requested for service, etc. The input of labour and machinery will be constrained within the development of crop and weather, and consequently, peak load periods will be unavoidable (Nielsen and Srensen, 1994). To reduce the organisational impact of these peak load periods, crops requiring different periods for treatment have to be selected. However, one thing, which reduces the impact of peak load periods signicantly, is that the labour force in agriculture is rather ex-

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

51

Fig. 7. Tactical planning focusing on in-season fertilising features and information acquisition for the fertilisation planning according to the strategy.

ible when it comes to complying with the extra workload in these periods. The operational planning outlined in Fig. 8 for the fertilising case concerns determining when a work operation will have to be executed in order to obtain a maximised work quality and maximum labour and machinery input effectiveness. Concerning the specic task of timing work operations, detailed job scheduling comes into play. In an agricultural context, scheduling is dened as determining the time, when various operations are to be performed. Availability of time, labour and machinery supply, job priorities and crop requirements are some important factors (ASAE, 1974). Work scheduling is the formulation of jobs based on required operations. Jobs can be scheduled when soil, crop and weather conditions are within certain limits. Planned jobs form the basis for task formulation (or implementation). Task formulation involves the actual specications of work-sets performing the tasks. The main decisions involve selecting the nal data acquisition services that is required. That together with a subsequent direct inspection of the eld conditions will determine the actual layout of the execution plan containing fertilising schedules, application rates, etc. The execution of the planned job in the operational planning phase focus on nal scheduling, task control and documentation of realised work (Fig. 9). Task control concerns the work efciency of work-sets (measured by time elements and compared with standard data). Operation control concerns the transformation of operation specications into set points for various parts

of the equipment being used. Device control involves comparing machine operation (e.g. work quality) with planned specications. The main decision process in this phase involve the inspection and controlling by the farm manager of the fertilising task by receiving information about the on-going task execution and take corrective actions if needed. Also, directly on the executing machine, a more or less automated decision process is running taking process information and converting it into actionable data for on-line corrective measures. The work performance is recorded and the documentation data is stored to the farm database. The nal step in planning and control cycle for the fertilising operation involves the evaluation of the executed operation (Fig. 10). The documented eld data need processing and aggregation in order to be useful in the evaluation. The key point is a comparison between the planned operation and the actual executed operation. The result of this comparison will be integrated in the subsequent planning cycle and will enable the manager to adapt to the actual or unexpected conditions on the farm. The feedback increases the farmers situation awareness and readiness to act in unexpected situations. The farmer gains new knowledge and skills which he/she can utilise to improve the performance in the subsequent planning cycle. This phase involve 4 main decision processes: (a) data processing for documentation, (b) compliance with standards check, (c) summarising fertiliser performance, and (d) comparison with target. External requirements for documentation as well as internal

52

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

Fig. 8. Operational planning focusing on the nal formulation of the execution plan. The planning utilises farm and site-specic information, historical weather information and updated local crop and norm information.

management practises will determine the extent of data processing before presenting the nal operations documentation data. The nal processed documentation data will be used for checking whether the actual operations execution have complied with the required standards. For internal operations planning, the realised fertiliser application performance is derived and compared with the planned operations performance. The latter comparison will guide the planning of the crop cycle. 5. Discussion The presented information ow models are aimed to function as the basis for Farm Management Information System design. It is able to derive functional features needed in the FMIS from the model diagrams. The Information Flow Diagrams (IFDs) describes the system interface features of such a FMIS which gives support to farmers core task. The IFDs can be utilised in FMIS design in a stage where improvements to the old system are considered, such as outlining business specications in the Structured Systems Analysis and Design Methodology (SSADM) for Designing Information Systems (Skidmore, 1997). The information modelling approach applied here differs signicantly from other information modelling attempts in complex domains. Fountas et al. (2006) presented an approach, where decision processes and information ows were modelled on the strategic, tactical, and operational level but not for the full range

of planning levels including execution level and evaluation. Nash et al. (2009) focussed solely on the data ows and not on a representation of the different actors and the sources of information ows. In earlier Scandinavian research (Nurkka and Norros, Pesonen et al., 2008), six core-task demands were identied: (1) Flexibility of acting interprets how skills are used to control dynamicity, (2) Readiness to actexible orientation how skills enhance controlling uncertainty, (3) Interpretativeness of acting portrays how knowledge is used to control uncertainty, (4) Conceptual mastery of farming how to control complexity, (5) Shared situation awareness describes how collaboration is used to control complexity, and (6) Optimizing resources to control dynamicity, respectively. It is proposed that when the designed and built technological system supports core-task demands, it will be useful and thus acceptable for the user. PA increases the complexity of farming by increasing details in planning and execution. As a consequence, the need of external services will increase (Olsson and Rydberg, 2008). Also, correct timing of eld work is important, which emphasis the meaning of farmers skills. Feedback via evaluation process is essential to develop appropriate skills. PA is information intensive farming and therefore easy and adequate availability of information is essential. Another aspect is the strong connection of the presented information model with decision support systems that are based on algorithmic optimisation approaches. The following two examples

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

53

Fig. 9. Execution focusing on nal scheduling, task control and documentation of realised work.

are presented in order to discuss this connection. As a rst example, the problem of machinery selection involved in the strategic planning level is considered. The intended optimisation tool is supporting the planning process choosing fertilising technology information and machinery and returning the output selected fertilizing technology (Fig. 6). A non-linear programming model like the one developed by Sgaard and Srensen (2004) may be invoked and comprise the capacity of the machine/s and the number of the tractors and their power as decision variables. The input factors included eld area, the xed annual cost, operating speed, working rate, repair and maintenance cost, and labour cost, the fuels costs, and expected crop price, and the timeliness cost. The second example regards the operational planning level (Fig. 8). Bochtis et al. (2009) presented a mission planer for an autonomous tractor covering also operations with capacity constraints such as fertilising. The software could support the process formulating execution plan in the operational planning level providing as output the tailored fertilising plans. The specic output is provided in an XML (extendible markup language) le determining several actions related to the execution level including the sequence of points the tractor has to follow, the type of motion between successive points (e.g., straight motion or manoeuvring), the type of predened turning routine used in manoeuvring, and the actions that should be taken once the tractor reaches the desired point (e.g., turning on or turning off the power take-off). The input for the mission planning optimisation problem involve

the eld geometry, facility unit location, eld tracks, etc. This information has been provided by the eld inspection service by the actor external service, i.e. a GIS: geographic information systemdatabase, or in the case that such a base does not exists, by the actor decision maker as a result of the process eld inspection, i.e. recording the eld boundary using an on-board GPS positioning receiver. The information machinery information provides the machinery related input, i.e. the minimum turning radius, operating width, recommended operating speed, and tank capacity. In the same way, the process of the generation of the updated information in real-time systems refers to the execution level (Fig. 9), where this information provided by actors included in the machinery units (e.g., internal sensors, implement ECU, and external sensors). Optimisation algorithms that use information generated at this level are involved in cases such as the real-time planning of a fertilising unit in a sensor-based variable rate precision spraying with some a priory information, i.e. by a satellite image (vehicle routing problem with stochastic demands) (Bochtis and Srensen, 2009), and the real-time planning for a relling unit in the absence of any a priory information (dynamic vehicle routing problem with time windows) (Bochtis and Srensen, 2010). The described information models specify the key activities for the targeted agricultural operations and capture the management cycle of the farm (planning, implementation and evaluation). The design structure facilitates the building of a dedicated information management system as well as provides the foundation for developing targeted decision support systems. Other important

54

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

Fig. 10. Evaluation part where documented raw execution data is processed further and aggregated for the comparative analyses.

enhancement includes recording and storing of implement status/work documentation into farm database, increased usage of numerical/formative data, the Farm database has a central role in both human and machine decision-making, active use of external services increases, increased use of automation, and smart assisting system features to support work are common, used information management technology shifts towards knowledge management technology. 6. Conclusions A user-centric approach to model the information ows for targeted eld operations has been presented. The information models were centred around the farmer as the principal decision maker and involved external entities as well as mobile unit entities as the main information producers. By specifying in detail, the information provided and the information required for the information handling processes, the design and functionalities of the individual information system elements can be derived. This has involved explicitly specifying tacit knowledge of the farmer as a way to extend the FMIS design into automated decision-making. The (IFDs) describe the system interface features of such a FMIS which gives support to farmers core task. In this study, a detailed approach to information modelling that will enable the generation of a Farm Management Information System in crop production, which is the main focus of the FutureFarm project, has been introduced.

The user-centric information ow models propose the implementation of effective managerial functions to the FMIS, but at the same time, they expect the farmers to be ready to adopt new working habits and perhaps also undergo further training. According to the modelling, farmers can utilise different services more efciently and they are able to outsource some of the tasks they had previously performed themselves. Also, farmers would be able to gain increased insight into their production processes and would able to evaluate the performance of the chosen technology. This would lead to better process control as well as an improved capability of documenting the quality of farming e.g. traceability, to markets and administration. Finally, the proposed system would allow the farmers to access and utilise better scientic research and technological developments by providing real process data and the ability to update the systems according to the latest knowledge. The enhancements from this approach would enable recording and storing of implement status or work documentation into farm database. This would also mean increased usage of numerical or formative data. Farm database would have a central role in both human and machine decision-making. Active use of external services as well as use of automation and smart assisting system features to support work would increase. This would be especially important in the future when information management technology shifts towards knowledge management technology.

C.G. Sorensen et al. / Computers and Electronics in Agriculture 73 (2010) 4455

55

Acknowledgements This project was part of the collaborative research project FutureFarm. The research leading to these results has received funding from the European Communitys Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 212117. References
Achten, J., 1997. Models, Methods and Databases for Labour and Machinery in Agriculture. Report No. 1, CIGR-Working Group 17, 45 pp. Akao, Y., Mazur, G.H., 2003. The leading edge in QFD: past, present and future. The International Journal of Quality and Reliability Management 20 (1), 2035. Anderson, J.R., Dillon, J.L., Hardaker, B., 1980. Agricultural Decision Analysis. Iowa State University Press, Ames, Iowa, USA, pp. 344. ASAE, 1974. ASAE Standards 1974. Standards, Engineering Practices and Data developed and adopted by the American Society of Agricultural Engineers. ASAE, MI, USA, pp. 293294. Auernhammer, H., 2001. Precision farmingthe environmental. Computers and Electronics in Agriculture 30 (13), 3143. Bochtis, D.D., Srensen, C.G., 2009. The vehicle routing problem in eld logistics part I. Biosystems Engineering 104 (4), 447457. Bochtis, D.D., Srensen, C.G., 2010. The Vehicle routing problem in eld logistics part II. Biosystems Engineering 105 (2), 180188. Bochtis, D.D., Vougioukas, S.G., Griepentrog, H.G., 2009. A mission planner for an autonomous tractor. Transactions of the ASABE 52 (5), 14291440. Chary, S.N., 2006. Production and Operations Management, 3rd ed. The McGraw-Hill companies. Chatzinikos, T., Fountas, S., Srensen, S.G., Gemtos, T., Vatsanidou, A., Pedersen, S.M., Nash, E., Blackmore, S., 2009. Management strategies and practices. EU Futurearm Project, Report No. 2.1.2, www.futurefarm.eu. Nurkka, P., Norros, L., Pesonen, L., 2007. Improving usability of and user acceptance of ICT systems in farming, Paper presented at the EFITA/WCCA Joint Conngress in IT in Agriculture. Glasgow. Fountas, S., Wulfsohn, D., Blackmore, S., Jacobsen, H.L., Pedersen, S.M., 2006. A model of decision making and information ows for information-intensive agriculture. Agricultural Systems 87, 192210. Gladwin, H., 1989. Ethnographic Decision Tree Modelling. Sage Publications Ltd., London, UK. Goense, D., Hofstee, J.W., 1994. An open system architecture for application of information technology in crop production. In: XII World Congress on Agricultural Engineering, Milan, Italy, August 29September 1, pp. 13561363. Hunt, 1983. Farm Power and Machinery Management. Iowa State University, Ames, Iowa, USA, 320 pp. ISO 13407, 1999. Human-centred Design Process for Interactive Systems. ISO, Geneve, 26. Jensen, A.L., Boll, P.S., Thysen, I., Pathak, B.K., 2000. Pl@nteInfo a web-based system for personalised decision support in crop management. Computers and Electronics in Agriculture 25, 271293.

Kay, D.R., Edwards, M.W., 1999. Farm Manager, fourth ed. The McGraw-Hill, Inc., Boston, MA, USA. McBratney, A., Whelan, B., Ancev, T., 2005. Future directions of precision agriculture. Precision Agriculture 6, 723. Nash, E., Dreger, F., Schwarz, J., Bill, R., Werner, A., 2009. Development of a model of data-ows for precision agriculture based on a collaborative research project. Computers and Electronics in Agriculture 66 (1), 2537. Nielsen, V., Srensen, C.G., 1994. Green FieldsOperational Analysis and Model Simulation. Bulletin no. 59. National Institute of Agricultural Engineering, Dept. for System Engineering, 155 pp. Norros, L., 2004. Acting under Uncertainty. The Core-Task Analysis in Ecological Study of Work. Espoo, VTT, Finland, http://www.vtt./inf/pdf/. Norros, L., Savioja, P., 2006. Integrated validation of complex human-technology systemsdevelopment of a new method. In: Rty, H., Puska, E.K. (Eds.), SAFIR The Finnish Research Programme on Nuclear Power Plant Safety 20032006 Integrated validation of complex human-technology systemsdevelopment of a new method. Espoo, VTT. VTT Research Notes 2363, 275285. Norros, L., Pesonen, L., Suomi, P., Srensen, C., 2009. Implementing systems usability evaluation in the design process of Active Farm Management Information System. In: JIAC Conference, Wageningen, July 58, www.eta.net. Olsson, J., Rydberg, A., 2008. User requirements on ICT for precision crop production. In: Proceedings 9th International Conference on Precision Agriculture, Denver, Colorado, USA, July 2023. Parker, C.G., 2005. Technology Acceptance and the Uptake of Agricultural DSS. EFITA/WCCA Joint Congress in IT in Agriculture. Pesonen, L., Rydberg, A., Koskinen, H. (Eds.), 2008. InfoXTUser-centric Mobile Information Management in Automated Plant Production. Nordic Innovation Centre report series, p. 108. Rosskopf, K., Wagner, P., 2003. Requirements for agricultural software and reasons for adoption constraintsresults of empirical studies. In: EFITA 2003 Conference. Runge, C.F., 2006. Agricultural Economics: A Brief Intellectual History. University of Minnesota Working Paper WP06-1. Skidmore, S., 1997. Introducing Systems Analysis. MacMillan Press Ltd., Oxford, UK. Sgaard, H.T., Srensen, C.G., 2004. A model for optimal selection of machinery sizes within the farm machinery system. Biosystems Engineering 89 (1), 13 28. Srensen, C.G., 2000. A Bayesian Network Based Decision Support System for the Management of Field Operations. Case: Harvesting Operations. Ph.D.-Thesis, Technical University of Denmark, 193 pp. Srensen, C.G., Nielsen, V., 2005. Operational analyses and model comparison of machinery systems for reduced tillage. Biosystems Engineering 92 (2), 143 155. US North-Central Regional Research in Farm Information Systems, 2000. Farm Information Systems: Their Development and use in Decision-making. Research Bulletin 601, Publication No. 345, Iowa State University, Ames, IA, USA. Van Elderen, E., 1977. Heuristic Strategy for Scheduling Farm Operations. Centre for Agricultural Publishing and Documentation, Wageningen, The Netherlands, 217 pp. Van Elderen, E., Kroeze, G.H., 1994. Operational decision making models for arable and grassland farms. IMAG-DLO, Report 94-3.