Wayne T. Hamilton
Department of Rangeland Ecology and Management
Texas A&M University, College Station, Texas, U.S.
Abstract
Veterinarians and other professionals working with range livestock producers often encounter noxious plants that present concurrent problems to animal health. In such cases the problem is acute, requiring immediate removal of animals and destruction of the problem plants. In other instances, noxious plants may influence range vegetation components negatively by reducing desirable forage species and diminishing range nutritional value. In the latter case, the need for weed control may be less critical in the short term, but relate significantly to the overall health and productivity potential of livestock and wildlife species. Producers quite commonly ask for advice on weed control methods, or it becomes obvious that brush and weed control should be suggested to them in order to overcome noxious plant problems.
The diversity of problem plant species on grazing lands and the large number of potential treatment alternatives makes selecting the appropriate brush and weed control technology difficult, especially in the case of herbicides. Combinations of chemicals, rates of application, additives, application dates and other recommendations vary with plant species, soils, climatic conditions and geographic region. Regulation governing use of herbicides and differential responses from the various technologies contribute to the complexities of selection.
The objective for developing EXSEL was to assist in selection of appropriate control technology for a target species based on predicted efficacy of the treatment. Users receive instructions for treatment application and guidance on posttreatment responses from target species and associated vegetation. The system requires a minimum of data input by users to perform treatment selection accurately. EXSEL is updated annually to add species, technologies, regulations, responses, or other information as it becomes available.
EXSEL can be made available to veterinarians and others assisting landowners with noxious plant problems to determine the correct treatment alternatives for effective control. The program will ensure that treatment recommendations are within the label requirements for herbicides and that landowners are aware of proper application instructions and regulations regarding the use of chemicals.
Introduction
Texas has 10 major vegetation regions (Hatch et al., 1990) that include approximately 95 million acre of rangeland. This great diversity of vegetation includes many different weeds and woody plants and a total of 42 million acres that are significantly invaded by noxious species, making Texas the nation leader in the need for brush and weed control technology (USDA NRCS, 1991). Chemical, mechanical, pyric and biological methods are all used with the state for brush and weed control. Matching the most appropriate control technology with the brush or weed problem is often difficult, particularly in the case of chemicals. Combinations of herbicides, rates of application, additives, application dates and other recommendations vary with species, soils, climatic conditions and geographic region. Moreover, regulation concerning use of herbicides and the differential responses that can be expected from application of various technologies all contribute to the complexities of choosing the best weed control practices by landowners and those assisting them (Welch, et al., 1991; Hamilton, 1993)
The primary objective for developing EXSEL was to support decisions for selecting appropriate control technology for a target plant species based on predicted efficacy of the treatment (percent plant mortality). The program is primarily designed for general brush and weed control on rangeland, not specifically for poisonous plants. However, plant database does contain several important poisonous species and the recommended control methods and it is realtively easy to add additional species and technologies.
EXSEL needed to be specific to the conditions identified by users in the field and yet not be input intensive, that is, not require great amounts of measurements or technical information. An additional objective was to provide users with current treatment application instructions as well as guidance on what response to expect following the treatment from both the problem plant and associated desirable vegetation. The system was also designed to allow easy updating to include added species to the 160 initially included in the species database, new technologies, regulations, responses or other information as it became available.
Methods
Development of EXSEL began with identification of the essential factors that must be considered when making brush and weed management decisions. The factors identified were target species, soil texture, soil depth, target species density (plants per acre), average mainstem diameter of target species, average height of target species, soil moisture condition, topography, amount of desirable vegetation present (to determine if seeding is necessary) and restrictions on use of aerial application of herbicides. With these factors in mind, information on various brush and weed control methods that could be used in Texas was obtained from the literature (Scifres, 1980; Scifres et al.. 1989; Welch, 1990) and from expert opinion. This information included effectiveness for species and characteristic limitations for the method with respect to plant density, plant size (height and stem diameter), soil texture and soil depth. Rules were then developed to utilize this knowledge along with user input on site characteristics and plant species, density and size to select technically feasible alternatives for the brush or weed management problem. Additional rules were developed to analyze user input on fine fuel characteristics (quantity, continuity and distribution) to determine if prescribed burning is feasible (Fig. 1)
Statements to provide information on expected response of forage target species to broadcast treatments were developed from research and demonstration results and from information obtained from brush and weed management experts. Treatment efficacy based on percent target plant mortality was obtained from research, demonstration and commercial applications and assigned to each brush and weed control method and plant in the species database. Caution or information statements were developed to alert the user of important environmental factors, such as season, soil moisture, soil temperature or unique characteristics of treatments, including limitations that are critical to proper application. Also, regulations that govern the use of brush and weed management practices, such as restrictions on herbicide applications, were included as caution or information statements within the program when the practices are considered.
Selection of mechanical technology in EXSEL is programmed via a forward chaining expert system using rules of inference to deduce conclusions. Chemical technology selection and burn feasibility sections use decision tree logic to determine specific recommendations. EXSEL is written in Microsoft C language 6.00a for the IBM PC. Some of the functions were written in Microsoft assembler to allow for responsive video output. The program was designed to operate with all popular IBM PC video adapters. The software is organized modularly into four areas: 1) the user interface, 2) the mechanical brush and weed management treatments, 3) the chemical treatments and 4) prescribed burn feasibility. NASA’s CLIPS system was used in preliminary development, however, an in-house technique was devised to duplicate the functionality we sought in CLIPS but in a smaller, more easily maintained and faster system.
All user interface-collected information is stored in a text file to completely isolate the user interface from the rest of the product. The user interface was selected on the basis of compatibility with other products developed by the Ranching Systems Group. It was developed in-house and is specific to the IBM PC environment.
The chemical treatments code is composed of a simple decision tree, as there was little advantage in deducing treatments via an expert system. Each plant species has certain possible treatments. Each set of treatments is organized by categories and restrictions so that the process of selection is a matter of printing the treatments that qualify. The prescribed burning feasibility code is also a decision tree which determines if the appropriate conditions are met for a prescribed burn to be possible.
The preparation of the EXSEL documentation was the result of planning meetings that involved all members of the software design team which included Texas Agricultural Experiment Station and Texas Agricultural Extension Service personnel, an Extension graduate assistant, a systems analyst and a technical editor. Program objectives, methods and structure were discussed in detail so that all members of the team were able to contribute to all aspects of the software program.
Results and Discussion
Isolation of the user interface from the rest of the product has two advantages: 1) if the user interface must change, as is likely because of expected changes in user interface technology, it would minimally affect the rest of the application, and 2) other interface modules could be developed so that the software might be more easily ported to other platforms. This decision was made with the expectation that computer technology, both hardware and software, would change over time.
Involvement of expert opinion is essential and serves not only to distill the essence of the expert knowledge, but also to assist the subject matter experts in clarifying their knowledge and identifying those areas where their opinions differ so that the knowledge may be made consistent and homogeneous for both computer and human alike. Developing a rule base is not simply a computer programming task. In testing, there were many cases in which the expert system would arrive at a conclusion with which the human expert would disagree. A rudimentary explanation feature was added to facilitate understanding the machine reasoning involved as an aid to debugging the deduced logic of the expert system.
EXSEL prompts the user to provide information on the essential factors that must be considered when making brush and weed management decisions. Once the required information is entered, EXSEL can determine the most technically feasible control method to overcome the problem. However, EXSEL also provides for personal preferences with respect to general methods of treatments, such as mechanical versus chemical or individual plant treatment versus broadcast treatment, by allowing personal preferences to override system selections.
Figure 2 provides an example of the logic flow using EXSEL for selection of technically feasible mechanical treatments. The initial data entry screen prompts the user for identification of the problem (target) plant species which can be obtained by searching the plant database (Fig. 1). In the case of mechanical practices, the county and soil type (textural class) do not function in the deduction process to determine possible treatments. However, these entries and others not needed to reach conclusions are normally entered to complete a record of the input variables to form a report for a client. They may also be required entries for chemical practice selection if a chemical alternative is subsequently considered. For example, the county entry is necessary to determine if there are regulation limiting the use of herbicides. The entry for soil depth does influence mechanical practice potential. If the entry for soil depth had been "shallow", the practice of rootplowing would not have been selected by deduction. The user input of "moderately deep" does not preclude the use of rootplowing, but a caution statement is invoked so that users know there may be limitation associated with soil depth.
Plants per acre key EXSEL to the practicality of individual plant treatment, such as the "Brush Busters" technology developed here in the San Angelo area by TAES and TAEX personnel. If plant density exceeds 500 per ha, the system invokes a broadcast treatment unless the user specifies individual treatment. Stem diameter determines if practices such as shredding are feasible. For example, 9cm stem diameter would indicate that a shredder could be used to remove woody plant top growth, however, the selection of "seeding required" later in the entries precludes the use of shredding since it does not contribute to appropriate seedbed preparation. The plant height entry would be highly significant to the use of ground chemical application equipment, however, is not limiting in the selection of mechanical practices. There are 4 entries for soil moisture ranging from adequate throughout the soil profile to "low overall." The low soil moisture entry does not constrain the selection of mechanical practices, but it does raise a caution statement that poor soil moisture can negatively affect the practice. The control method to be used can be specified by the user, or it can be deduced by the expert system from data inputs. In the example, the system was allowed to select between a broadcast or individual plant method and selected broadcast on the basis of plant density as discussed above. The user input that there is no restrictive topography keeps this consideration from being a factor in the determination of mechanical treatment feasibility. If topography were restricting (steep slopes, etc.), this entry would inject a caution statement to the user concerning the need for extreme care in proceeding with mechanical measures.
Treatment type refers to mechanical or chemical practices. In Fig. 2, the user is assumed to select mechanical, which forces the consideration of only these practices. When seeding is specified, the system will exhibit to the user potential problems with seedbed preparation in the form of caution statements. For example, there is a statement that indicates that rootplowing without raking prepares a poor seedbed. Since the user selected mechanical treatment type, the input field for possible restrictions to aerially applied herbicides does not contribute to the deductive process. However, this entry could be needed if the user wants to return to the same problem plant and switch to chemical rather than mechanical practices. If a prescribed burn is selected by the user, there are additional input fields for fine fuel load, continuity and distribution: however, in the example it is assumed that an interpretation by the system on the potential for burning is not desired. Those elements identified as pertinent to the conclusion for possible mechanical treatments interact with the rule base and the treatments are deduced and reported.
EXSEL generates reports that can be sent to the screen, a file, or the printer. The report includes a disclaimer statement and information entered by the user so that it can be reviewed. It also contains the specific treatment recommendations, both chemical and mechanical, for the target species. In the case of chemicals, the report includes the recommended rates of application for each compound as well as additives and specific application instructions. The report also provides the expected results from the treatment expressed as the level of efficacy: very high(76-100% mortality), high (56-75% mortality), moderate(36-55% mortality) and low (less than 35% mortality). When individual plant chemical treatments are recommended, the report will provide guidance to herbicide formulation for total volume of spray mix. Tables provide the amount of formulated herbicide to use for the concentration shown in the rate recommendations for individual plant and spot treatments and for a range of total spray volumes desired.
EXSEL also provides expected projections of vegetation responses following application of selected technologies; for example, information is given when maximum production increase should be expected after treatment, how long this level of production will likely last and when it will return to pre-treatment levels. A combination of predicted treatment efficacy and expected response provides information to decision makers for use in economic analysis comparing alternative treatments using net present value of the investment (Conner et al., 1990).
EXSEL makes users aware of environmental factors, such as season, soil moisture and/or temperature or unique characteristics of treatments, including limitations that are critical to proper application. The report includes "comment" statements that call user’s attention to uniqueness about the treatment or provides information that may be helpful in considering the treatment as opposed to alternatives. The report also contains "caution" statements that flag regulations that govern the use of brush and weed management practices, such as Chapter 75 of the Texas Agricultural Code that regulates sale, use and transportation of herbicides in the user’s county, referred to earlier in the Fig. 2 example.
If the user is interested in prescribed fire, EXSEL prompts the input of information on fine fuel quantity, continuity and distribution. EXSEL then determines the potential for use of prescribed burning as a brush and weed control method. A detailed checklist of factors required for the safe implementation of prescribed burning is also provided. A worksheet can also be printed that prompts the user for necessary input information and provides a place to record entries that match the screen input requirements of the software. The worksheet is particularly useful when taken to the field to record data.
The F1, F2, and F3 function keys allow alphabetical searching of the plant database to find the target species, move to the next species of the previous species, respectively. The search feature eliminates the need for the user to spell plant names. Another useful function is the F8 key that allows restoration of all previously entered data after escaping from a report so modifications can be made. This restoration feature is particularly helpful in developing "what if" scenarios by changing input for the same target species.
EXSEL also contains a "search" function which allows the user to search forward to a specific word or phrase in the report as well as to repeat a search. There are also "field-sensitive" help screens in EXSEL that provide more information about the inputs that are required. Another feature is a "readme" file that can be viewed on the screen or printed to provide information for use of the program in addition to those included in the user’s guide. This feature is useful for bringing items to the user’s attention that may have been developed after the user’s guide was last revised.
Conclusions
Field tests indicate EXSEL successfully matches a specific weed or brush problem with the most technically feasible treatment alternatives. EXSEL is user friendly. County Agricultural Extension Agents evaluating the software have indicated that the user’s guide provided sufficient instruction to preclude formal training sessions. Following the field testing phase, the addition of several new herbicide use recommendations demonstrated the ease in updating the program. EXSEL is serving as a powerful tool for service agency personnel, consultants and rangeland producers for not only organizing the complexities of control measures, but also providing optimum control technology for a diversity of brush and weed control problems. The software would be beneficial to veterinarians working with landowners to eliminate or reduce infestations of noxious plants to minimize negative consequences of ingestion by livestock. Moreover, information provided in EXSEL on responses to expect from target species and associated vegetation following treatment has proven useful in the development of response curves that plot benefits for economic analysis.
In addition to being available through all county extension agents in Texas, EXSEL has been purchased by the Natural Resources Conservation Service for use in some of the Texas field offices. The current version of EXSEL is 1.07 and the program is updated on April 1 of each year as necessary. EXSEL is available to the public through the Texas Agricultural Extension Service software catalog and is priced at $50.00 for in-state and $60.00 for out-of-state purchasers. The purchase price includes the user’s guide, installation diskette and notification of updating of the program when changes are incorporated. Updated versions are available to previous purchasers for $10.00. The program works well on lap top computers that are carried on ranches in vehicles by veterinarians, extension personnel, NRCS technicians or others working with landowners.
EXSEL databases do not currently include many of the plant species that veterinarians would identify as toxic to livestock or wildlife. However, the plant database to list these species and the treatment recommendations for effective control can easily be added to the program. We would welcome the cooperation of veterinarians concerned with toxic plant control to identify and include additional problem plants in EXSEL.
References
Conner, J. R., Hamilton, W. T., Stuth, J. W. and Riegel, D. A. (1990) ECON: an investment analysis procedure for range improvement practices. MP-1717, Tex. Agr. Exp. Sta. Tex. A&M Univ., College Station. 46 p.
Giarratano, J. C. (1989) CLIPS user’s guide: version 4.3 of CLIPS. Artificial Intelligence Section, Lyndon B. Johnson Space Center. 459 p.
Hamilton, W. T., Welch, T. G., Myrick, B. R., Lyons, B. G., South, J. W. and Conner, J. R. (1993). EXSEL: expert system for brush and weed control technology selection. In Proc. Conf. on Application of Advanced Information Technologies: Effective Management of Natural Resources. ASAE, Information and Electrical Tech. Div. 500 p.
Hatch, S. L., Kancheepuram, G. N. And Brown, L. E. (1990) Checklist of the vascular plants of Texas. MP-1655, Tex. Agr. Exp. Sta., Tex. A&M Univ., College Station. 158 p.
Scifres, C. J. (1980) Brush Management: principles and practices for Texas and the southwest. Tex. A&M Univ. Press. 360 p.
Scifres, C. J., Koerth, B. H., Crane, R. A., Flinn, R. C., Hamilton, W. T., Welch, T. G., Ueckert, D. N., Hanselka, C. W. And White, L. D. (1989) Management of south Texas mixed brush with herbicides. Bull. 1623, Tex. Agr. Exp. Sta., Tex. A&M Univ., College Station, 66 p.
USDA Natural Resources Conservation Service (1991) Grazingland Resources in Texas. 34 p.
Welch, T. G. (1990) Chemical weed and brush control suggestions fro rangeland. Bull. 1466, Tex. Agr. Ext. Ser., Texas A&M Univ., College Station, 17 p.
Welch, T. G., Hamilton, W. T., Deiss, S. W., Myrick, B. R. and Lyons, B. G. (1991) EXSEL: Expert system for brush and weed control technology selection. In Proc. Intl. Conf. On Decision Support Systems for Resource Manage., Texas A&M Univ., College Station, 158 p.