Design and Implementation of Farm Weed Management System Based on GIS

Geographic information system (GIS)-ArcView and spatial variable model were used to design and implement an ArcView-based farm weed management system. The objective of system design was to maximize net returns from herbicide application. ArcView provides the necessary customization and language environment tools in an easy framework for this system design and implementation. A simulation model in the framework was used to determine weed distribution within field, to select suitable herbicide, to compute the related dose and to provide a herbicide application map for spatially selective herbicide spray. The system architecture was determined or described using a layered model that is especially suitable to ArcView framework. The layered model approach was proved to be a practical method of designing and implementing a GIS-based weed management system, and should be widely applicable to the design and implementation of other similar systems.

1. Introduction

This article describes a framework of requirement needed to support GIS-based weed control system implementation and summarizes the most critical areas for improvement in the current computer- based weed control systems.

2. Collection weed distribution data

-Collecting weed distribution data about the spatial variability within field;

-Determining weed plan how to execute spatially selective herbicide spray;

-Implementing the plan within field.

Because the spatial variability of weed distribution within a field may lead to unsuitable application of costly herbicide over some areas, an essential step for effective weed control is accurate identification of weed distribution. In fact, the potential reduction in herbicide use and increment of crop yield from the use are two important motivations for implementing spatially selective herbicide spray according to the weed distribution within a field. Thus, the problem of determining weed distribution becomes a necessary prerequisite to the spray. However, the physical characteristics of weeds make usually detection of distribution of weeds in a growing crop become a very difficult problem. Thompson et al.[3] suggested that suitable real-time weed identification algorithms were unlikely to be available in the near future. The conclusion remains particularly true of the identification of economically important grass weeds in cereal crops. But, Miller and Stafford[6] have proposed a map-based approach in which weed populations are first located on a map which can then be converted to a treatment or application map and used to control the sprayer. The treatment map provides usual information as the followings.

(1)Selection of the dose and mixture of herbicides that will give a level of control close to the optimum; and

(2)Selection of boundary areas around patches to allow a margin of error for both detection and spray control.

Otherwise, Stafford et al. [7] described a technique for generation of weed distribution maps with GPS. In their experiment, a backpack weed distribution map recorder enables a farmer to log information on weed patches on a hand-held computer as he walks a field. The information is automatically tagged with a grid positioning method through a Global Positioning System(GPS) receiver carried in a backpack. On the other hand, in some developing countries that have large farm labor, the information can be manually tagged with an experienced farmer by a hand-held computer.

In our system, an ArcView 2.0-based[8] weed mapping program in a hand-held computer is used to construct a weed distribution map. The program performs the following basic tasks:

-Display a field map,

-Move around a map and display current GPS position,

-Turn various weed-related geographic features on and off ,

-Identify map features including field, crop, weed, herbicide and other weed-related information,

-Label or Record weed features on a map, from GPS and insight,

-Add text or description to a map or weed,

-Plot a weed distribution map.

Weeds are usually distributed in patches with large areas of the fields with weed free or with very low weed density. Some weed patches remain spatially stable over long periods in a conventional cropping method. Thus, the information on a map for describing weed distribution includes weed species and density of each weed patch. Further, the density with each weed patch is set to have three levels: low, medium and high. The patch is defined as four forms: Irregular polygon, Rectangular, Circular and point forms. The GPS information and ArcView function are used to input each patch form and position information into a weed distribution map. The powerful function of ArcView makes the work to generate weed distribution map become very simple. The weed feature or weed distribution map can improve efficiency of herbicide application in order to yield long-term cost savings.

3. Determining weed control plan

In a GIS-based weed management system, dose and weed distribution accuracy may be even more important than in a conventional weed control when determining a weed control plan. Otherwise, environmental factors are extremely important in determining a plan. Thus, soil type, weather and history for treatment success may be more important in determining a plan than weed distribution problem. A GIS-based weed control plan can fail for a number of reasons, including: accuracy of applied herbicide dose, environmental factors, spatial resolution, dose range, dose resolution, simultaneous application of products, dilute herbicide disposal, and range of suitable formulations and active ingredients[9]. However, in developing countries, one effective way is for the farmer to manually operate his spray equipment, so that it responds properly to the spatially variable weed control. Thus, some of these problems are easily solved in our weed management system due to application of manual spray and GPS positioning.

For determining weed control plan, a database of herbicide products is used to store much label information to describe how to apply all registered crop herbicides. It provides much information including application doses, costs, times for these herbicides. Other some important data relevant to the information include also susceptible and resistant weeds, environmental protection limits, application sequences, tank mixes and application notes of these herbicides. Otherwise, the database provides a cross-referencing function that gives all the active ingredients in these herbicides. On the other hand, a reasoning algorithm is used to look into various optimal recommendations of herbicide application with product name, active ingredient and other label information.

Further, the dose can be represented as a function of spatial variation about weed distribution and other factors. A dose applied to part of a field may be obviously less than the manufacture recommendation. Final weed control plan is usually represented as a weed spray control map. A sorting algorithm uses information from manufacture labels and weed distribution maps to recommend herbicides for the weed control. Then, the recommendation is input to a weed control map. In our ArcView-based system, a weed control map is usually related to a project. A project is really a collection of views, tables and other documents. A view is an ArcView document that displays a map and its legend. A view is made up of layers of field information for a particular geographic area. For example, each layer of field information is a collection of field features, such as crops, weeds, herbicides, rivers, roads, boundaries and counties. These layers are called themes. The legend lists the symbols representing features of these themes. A theme is a collection of field features in a view. All the features in one theme are represented by the same type of shape, such as points, lines, or areas. Themes also reference the attributes or characteristics of field and weed features. Usually, a weed control map includes some themes as following.

-Weed distribution theme,

-Crop distribution theme,

-Herbicide application theme,

-Field theme,

-Soil theme.

Weed distribution theme describes weed characteristics and information from weed identification, including weed identity, growth stage, type and geographic distribution. Crop distribution theme represents crop type and growth stage. Herbicide application theme represents positions, times and doses of herbicide application, and other information such as compatibility, cost-effectiveness, weed spectrum, herbicide resistance, crop/cultivator tolerance and environmental restrains information.

In order to reduce field operation cost and improve application timeliness, most farmers apply usually more than one herbicide during one spray operation. Thus, an algorithm is used to select herbicides which are approved for tank mix use, with the help of input information such as crop growth stage, herbicide and label information. In fact, the tank mix spray is also necessary to control simultaneously different weed species in a patch with different herbicides. The sequential spray of different herbicides is another important method of reducing cost and increasing yield. For arriving at the goal, a mechanism is used to select herbicides which may be included in a sequential spray program. The selection is based on information such as product label recommendation, crop growth stage and application timeline.

In general, traditional dose decisions are made on the basis of average whole field weed infestation. However, our method considers the effects of spatial aggregation of weed populations on yield loss, and the accuracy of these dose decisions is obviously improved. Otherwise, in traditional weed control process, disposal of unused herbicide is an expensive and difficult problem. But, our method solved the problem by accurate dose computation and herbicide application map based on practical weed distribution and area. Thus, the success of weed control plan application depends greatly on choosing safe, reliable and cost-effective weed control strategy.

4. System implementation

An architecture of FWMS can be described by a seven-layer architecture model shown in Fig. 1. Simple function of its various layers can be discussed as the followings. The field layer describes the geographic position, shapes and other characteristics of managed fields. The soil layer describes all information about soils of these fields, which includes various key data from a detailed soil survey. The main task of climate layer is to collect and generate a number of basic climate data for supporting weed control decision in its high layer. The environmental layer consists of a number of basic data of environmental restrains to generate various maps of environmental restrains because environmental protection is an essential objective for effective weed control.

In the crop layer, a number of data to represent growth of current crops are used to describe crop status and information about weed control. The weed layer is used to represent current weed distribution position and weed distribution levels within these fields. The herbicide layer is used to generate several different herbicide application decisions for different fields or areas according to practical requirement of these fields and with the help of various basic services provided by its lower layers. In the layered architecture, the lower layer provides always its service for its higher layer, the higher layer uses the service from its lower layer. Our experience shown that to construct the layered architecture model is a very effective method for designing and implementing a high quality ArcView- based application system.

For showing practical application value and steps of the above method, a typical GIS application system for farm weed management was designed for implementing an optimal management of herbicide application within fields or areas. The real project resulted in a prototype called FWMS. The primary consideration of FWMS computer implementation is to allow farms to access and use all available relevant data and decision information provided by FWMS to optimally control their weed through the above model in FWMS. The FWMS tasks include herbicide spray management, determination of weed distribution, herbicide management, and evaluation of relevant weed control result and output of various maps. FWMS was developed on a PC under the Microsoft Windows system using the VISUAL BASIC, VISUAL C++ and Avenue language[10].

The partial source codes of FWMS are written in modular form using the macro programming language Avenue of ArcView 2.1. In FWMS, each unique procedure or activity is written as a separate script. Totally, FWMS includes more than 70 scripts. Table.1 shows a list of partial FWMS' scripts.

/////////////////////////////////////////////////////////////////////////////////////////

FWMS.About.JaasSVW

FWMS.Add.GlobalVariables

FWMS.AddTheme.Choice

FWMS.AddTheme.Crop

FWMS.AddTheme.Weed.Distribution

FWMS.AddTheme.Soil

FWMS.AddTheme.WeatherZones

FWMS.AddTheme. Herbicide

FWMS.Calculate. Herbicide

FWMS.Change.FWMS.Soil.fle

FWMS.Change.Sim.File.ForRecords

FWMS.Change.Sim.File.ForView

FWMS.ChangeTheme.Choice

FWMS.Create.ModelDictionary

FWMS.Map.New

FWMS.Map.Open

FWMS.Theme.Properties

FWMS.View.ConvertToShapefile

FWMS.View.Run.Simulation

FWMS.Delete.Documents

FWMS.Delete.Tablefiles

FWMS.HasVisibleThemes.Update

FWMS.Join.Semi.Automatic

FWMS.Join.Simulation.Results

FWMS.Label.VisibleTheme

FWMS.Layout.Create. Appl.Map

Table.1: A list of partial FWMS' scripts

////////////////////////////////////////////////////////////////////////////////////

ArcView provides the necessary customization and language environment tools in an easy framework. The ArcView-based framework provides some common functions for FWMS as the followings.

(i)To create the graphical user interface according to user requirements,

(ii)To establish some initial properties for graphical controls to support interaction between user and these controls,

(iii)To write Avenue codes that satisfy the requirement of user interface and simulation model,

(iv)To link scripts written in Avenue to events,

(v)To integrate the framework with other language modules.

Avenue is an object-oriented scripting language. The key point in Avenue is on identifying objects and then sending requests them to complete various complex operations. An object can be considered as a package that is composed of tightly-coupled data and functionality. In Avenue, a request can be sent to an object. When an object receives this request, it performs some relevant action or operation. In ArcView, objects are members of a class hierarchy or architecture. Further, these members can be organized into functional categories related to all aspects of the weed control.

Avenue's statements are used to organize and determine when and how requests are made. A request specifies what an instance of a given class will do and a method specifies how it is done. By sending a request to an object, a method appropriate to the class of which the object is an instance is activated. An object in Avenue always responds to a request by returning an object; in some cases, the request creates a new object or returns an existing object.

The prototype system has been used to analyze and process a number of collected data from fields. Weed control of FWMS was compared with traditional none-spatially selective weed control. When the traditional approach was completed with the weed control operation, the control quality of traditional approach is not usually satisfactory due to its none-spatially selective operation. However, the FWMS-based approach may provide an objective method for performing field weed distribution assessments and making optimal weed control recommendations that maximize yield or profit. The profit of FWMS-based weed control approach increased about 9-12% by comparison with the above traditional approach.

These successes show that the shift from non-spatially selective weed control to GIS-based spatially selective one is a profound advance in research of farm weed management system. Further, some major steps of the method application are as the followings: scooping of an application development project, in which the project area, requirement and relevant variables are defined by objective of the application plan, based on the user requirement of spatially selective weed management system and GIS(ArcView) environment; collecting data relevant to the model of spatially selective weed control in the application field, in which these data must be represented as a layered architecture model form relevant to GIS thematic or map form by the method for designing a layered architecture; developing, evaluating and selecting all important logical relationships or functions between variables of these layers, in which these relationships are used to build a basic framework of GIS-based application system or model; integrating these relationships and data relevant to the model, in which final application system or model is built; refining this system and improving its performance. Currently, the system is being revised to incorporate more decision functions. Thus, a new ArcView-based weed management system is being developed that has more powerful GPS function and generate better spatially selective weed control recommendations for strategic planning of scientific weed control.

5. Conclusions

This ArcView-based weed management system for farm scientific weed control was developed to study the principle and application possibility of spatially selective weed control model in developing country. The determination of weed distribution map and the generation of spatially selective herbicide spray map are two keys to design and implement the model. The layered architecture model is an efficient method to design and implement a GIS(ArcView)-based application system. In fact, the layered model is also a theme-oriented processing model. The work to develop more efficient GIS- based model for spatially selective weed control will become a future challenge in the domain. The design and computer implementation of management system of farm weed will change from current 2D plane into future 3D space. ArcView software tools will also provide more powerful supports for this challenge or objective. However, the future works about the computer implementation are to solve some key problems that include how to improve the architecture of spatially selective weed control model, how to construct the operation rules in the architecture, and how to design and implement these rules in GIS environment. In summary, the design, implementation and application of GIS-based spatially selective weed control system will become an important research direction in precision farming domain.

6. Acknowledgments

7. References

[2]Pasqual, G. M., Development of an export system for the identification and control of weeds in wheat, triticale, barley and oat crops. Computers and Electronics in Agriculture 10(1994):117-134.

[3]Thompson, J. F., Stafford, J. V. and Miller, P. C. H., Potential for automatic weed detection and selective herbicide application. Crop Prot., 10(1991): 254-259.

[4]Tim, O. S., Emerging technologies for hydrologic and water quality modeling research. Transactions of the ASAE 39(2)(1996): 465-476.

[5]Yule, I. J., P.J. Cain, E. J. Evans and C. Venus. A spatial inventory approach to farm planning. Computers and Electronics in Agricultural 14(1996): 151-161.

[6]Miller, P.C.H. and Stanfford, J. V., Herbicide application to targeted patches. Proc. of British Crop Protection Conference on Weeds, 18-21 November 1991, Brighton, pp. 1249-1256.

[7]Stafford, J. V., J. M. LeBars and B. Ambler. A hand-held data logger with integral GPS for producing weed maps by field walking. Computers and Electronics in Agriculture 14(1996): 235- 247.

[8]Esri. Introducing ArcView. Environmental Systems Research Institute Ltd., Redlands, CA, 1994, 98 pp.

[9]Paice, M. E. R., P. C. H. Miller and W. Day. Control requirements for spatially selective herbicide sprayers. Computer and Electronics in Agriculture 14(1996): 163-177.

[10]Esri, Customizing ArcView with Avenue. Environmental Systems Research Institute, Inc., 1994, USA.