BEE AWARE, an expert system for honey bee diseases, parasites, pests and predators

BEE AWARE, an expert system for honey bee diseases, parasites, pests and predators

Computers and Electronics in Agriculture, 9 ( 1993 ) 111-122 Elsevier Science Publishers B.V., Amsterdam 111 BEE AWARE, an expert system for honey b...

598KB Sizes 0 Downloads 90 Views

Computers and Electronics in Agriculture, 9 ( 1993 ) 111-122 Elsevier Science Publishers B.V., Amsterdam

111

BEE AWARE, an expert system for honey bee diseases, parasites, pests and predators Janice E. McClure a, M a r y a n n T o m a s k o b and Clarence H. Collison c aExpert Systems Development Group and bDepartment of Entomology, Penn State University, University Park, Pennsylvania, USA CDepartment of Entomology, Mississippi State University, MS 39762, USA (Accepted 3 March 1993 )

ABSTRACT The BEE AWARE expert system for the diagnosis and management of honey bee diseases, pests, parasites and predators has been developed and tested as a new means of transferring up-to-date information to beekeepers. Funding for government agencies traditionally responsible for the control of honey bee diseases is being reduced in many states, and completely eliminated in others. Extension positions and budgets for apiculture programs are also being cut. With the introduction of two new parasitic mites and incidents of disease on the rise, there is increasing need to make up-to-date information readily available to beekeepers. BEE AWARE consists of educational information, diagnosis, and other information on the control and management of diseases, pests, predators and parasites of honey bees. The program has been tested with users for a year and is available commercially for minimal cost.

INTRODUCTION

In recent years, two new parasites and one disease have been introduced into the United States and have become serious problems for beekeepers. The varroa mite, Varroajacobsoni is considered the most serious pest of honey bees in the world and was first reported in the United States on September 25, 1987. Since that time this mite has spread throughout most of the south and much of the northern U.S. A second parasitic mite, the honey bee tracheal mite, Acarapis woodi (Rennie) was first discovered in the U.S. on July 3, 1984 and is now present in nearly every state. This parasite has caused extremely heavy winter loss of honey bee colonies, especially in the northeast (Eischen, 1987; Otis et al., 1988; Furgala et al., 1989). A survey of Pennsylvania beekeepers found that apiaries not infested with tracheal mite experienced an average winter loss of less than 11% while apiaries infested with the Correspondence to: J.E. McClure, Coordinator, Expert Systems Development Group, 501 ASI Bldg., Penn State University, University Park, PA 16802, USA.

0168-1699/93/$06.00 © 1993 Elsevier Science Publishers B.V. All rights reserved.

112

J.E. McCLURE ET AL.

mite experienced a loss greater than 31.4% (Tomasko, 1990). In addition to the diseases that have been present in this country almost as long as the honey bee itself (American foulbrood, European foulbrood, sacbrood, nosema and paralysis) a new fungal disease, chalkbrood was reported in 1968 (Baker and Torchio, 1968). Traditionally, the apiary inspection service in each state has been charged with the control of American foulbrood (AFB), a severe and contagious brood disease of honey bees. However, in recent years apiary inspection services have been saddled with the additional responsibilities of sampling for and attempting to control the spread of the two newly introduced parasitic mites. In 1980, apiary inspection in Pennsylvania was reduced about 70% and has continued at a reduced level ever since. In 1979, the Pennsylvania apiary inspection service found only 1.1% of the colonies and 1.7% of the apiaries inspected to be infected with AFB. In 1986, levels of AFB had increased to 6.0% of the colonies and 12% of the apiaries inspected (James Steinhauer, Pennsylvania Department of Agriculture, personal communication, July 10, 1992 ). This trend is not unlike what is being experienced in other states. In addition to decreases in funding for apiary inspection, the Cooperative Extension Service is experiencing nation-wide budget cuts at the county and state levels. These funding cuts have resulted in a loss of apicultu.re specialists in some states. While there has been a shift in research emphasis to parasitic mites and Africanized bees, avenues for making this information available to beekeepers have been limited. In particular, chemical and cultural tactics available for controlling mites are constantly being changed and updated, a trend that is expected to continue. This information has been slow to reach beekeepers through traditional methods. The problem faced by many of the agricultural scientists and extension specialists is how to package and deliver these large amounts of technical information in an organized and timely fashion. Expert systems are being explored as a new and innovative method of transferring timely, up-to-date information to farmers, growers and beekeepers. New research information is currently transferred through county extension meetings, demonstrations, extension publications, videos and by one-on-one contact. However, because of the increasing complexity of farming in the United States, additional methods are needed to deliver the increasing amount and variety of information available to assist farm managers. The organizational structure of expert systems and their ability to handle both heuristic and numerically based decision support programs make them ideal for organizing and delivering large amounts of technical information (Plant and Stone, 1991 ). By integrating information from many different subject areas the human expert is able to formulate a specific solution to a particular problem. In general, agricultural expert systems are capable of integrating the knowledge of several disciplines (for example, entomology, microbiology, botany, meteorology, economics) into a

BEE AWARE

1 13

single knowledge based system of decision support. Properly developed expert systems can be a powerful tool for providing growers or beekeepers with the day-to-day decision support needed to grow crops or manage bees. DESIGN AND DEVELOPMENT

Expert system development at The Pennsylvania State University (PSU). Although expert systems have been used in industry for years, their application in agriculture is recent, and the use of these systems in commercial agriculture is rare to date (Travis and Latin, 1991 ). In 1986, USDA Extension Service/ Smith-Lever 3 (d) funds for Integrated Pest Management supported a formal effort at Penn State in developing systems. Researchers and extension personnel in agricultural engineering, entomology, plant pathology, horticulture, agronomy and agricultural economics developed several interdisciplinary programs in the agriculture sciences.

PennShell. PennShell is a collection of C language macros that facilitate development of frame-based expert systems using both backward and forward chaining (Foster et al., 1991 ). PennShell handles the I/O functions, graphics, and menu handling. The programmer creates the rules, frames, and organization of the program along with any special features. PennShell exists on both MS-DOS and Macintosh platforms, allowing simultaneous delivery with only a few special platform modifications. BEE AWARE is written in the C language with the use of the PennShell expert system tool kit.

BEE A WARE development. The development of any expert system includes design, testing, evaluation, delivery and maintenance (Parsaye and Chignell, 1988 ). In the design phase the goals, audience and scope of the program must be decided. Biological and other technical experts are required to perform several tasks including: defining the problem or problem areas, defining the boundaries of the system, determining interactions among the fields of expertise and, within problem areas, reproduce problem solving logic on paper. This logic is then implemented by the programmer on the computer. The ideal expert system problem is one that needs an expert level of knowledge in order to arrive at a solution and yet is not so encompassing that it cannot be reasonably described and documented. The audience and purpose of the expert system needs to be clearly defined since the expert system must be compatible with the needs and management style of the end user. BEE AWARE was initially designed in 1988-1989 with the goal of delivering up to date information to beekeepers in an easy to use format. The scope was limited to diseases and pests of apiaries. Initially, the audience was defined as Pennsylvania beekeepers, but since has been broadened to include

114

J.E. McCLURE ET AL.

North American beekeepers. These items were decided upon before starting the actual logic of the system. As the logic was committed to diagrams and programmed, these design considerations were used as guiding principles. Testing and evaluation are as integral to the development of a practical, useful expert system as the creation of the computer program itself. There are at least three levels in the testing phase: testing the code, testing the interface, and testing the knowledge base (Parsaye and Chignell, 1988 ). The code was tested by the primary programmer with the aid of other programmers experienced with the current development practices. Testing the knowledge base (challenging the program) of the expert system BEE AWARE was accomplished by other apiculturists at the research and production level. Interface testing was accomplished at association and extension meetings, short courses and in sessions with individual beekeepers. Early in the development of BEE AWARE, it was realized that the end user (client) of the system must be involved in the design of the program. For the user to operate the program effectively and for the program to receive the accurate information it needs, the interface had to ask questions and receive input in an strict, but friendly manner. Also, the recommendations from the expert system had to be in the context of the user's situation. Fourteen individuals including producers, extension personnel and researchers used the program over the course of a year, and c o m m e n t e d on its design and features. Modifications in the presentation, wording, methods of information input, and types of recommendations were incorporated from the suggestions of end users as well as disciplinary peers and others. Criticism provided by the test group of users accounted for more substantial changes in the expert system than criticism from other groups. The delivery phase requires support and training. Users of BEE AWARE are encouraged to call the development group with questions or problems. Pennsylvania extension agents are encouraged to attend an in-service to acquire general information on beekeeping and training in the use of the expert system. Maintenance is a continuing process, the code must be updated for new information, and hopefully, new areas added. The shell and the compiler used also are updated by their firms regularly, the program is updated to the newest versions of the shell and compiler as needed to insure easier maintenance and support. The BEE AWARE program is usually updated on an annual basis. RESULTS

BEE A W A R E expert system. Currently, the system gives advice on the identification and management of honey bee diseases, pests, parasites, predators

BEE AWARE

1 15

and some information on pesticides. The program is divided into three modules: Information, Diagnosis and References (Fig. 1 ). The information module provides specific information on individual diseases, pests, parasites and predators. This information is broken down into categories such as cause, symptoms, transmission, life cycle and treatment or recommendations. The diagnostic module is designed to assist the beekeeper in identifying an unknown problem. Through a series of questions, answered by the beekeeper, the computer narrows down the possibilities until it comes up with a suitable diagnosis. If the computer is given conflicting or inconsistent information and cannot make a reliable diagnosis, the beekeeper is advised to consult a local (human) expert. Once a problem is identified, the beekeeper can then obtain

I BeeAware 1 ( Information)

( Diagnosis)

(References1

I'0°"d'--.]~

I 0u-.0- I

I 0,o-.r~ !

IBr-°d,--.,~

[-~,~l

~.',r.n°-I

I ~0... I

l"~"n'"~J//

I Predators

/

Hare,nformatioIn

Parasitesi ~ I Pesticides Fig. 1. Modules of BEE AWARE and how they interlink.

I~"'°'°"°'°'1 Pesticide

I'oxicittabl tj e

l ]6

J.E. McCLURE ET AL,

specific information (including current control recommendations) concerning the disease, pest, parasite, etc. from the information module. Several aspects of BEE AWARE are educationally motivated. The information contained in the individual description sections is designed to educate beekeepers on the causes and prevention of diseases. The reference module is divided into four sections. The first section is a list of the scientific publications used in creating the system. Additional sections include a list of apiary inspectors in the U.S. and Canadian provinces, a glossary of the technical terms used in the system and a list of commonly used pesticides and their relative toxicity to honey bees.

Information. The information module consists of the following categories: adult diseases, brood diseases, pests, predators, parasites and pesticides. Each of these categories is further subdivided into specific types of brood or adult diseases, types of pests, predators or parasites. If the user is aware of the specific problem and simply wants additional information, he/she selects the category under which the problem falls. A list of specific diseases, parasites, etc. appears under the chosen category. The user then chooses the appropriate item; for example, American foulbrood under the category brood disease (Fig. 2 ). Once the disease, pest, parasite, etc., of interest is selected, specific topics (such as cause, symptoms, transmission, treatment, etc. ) can be chosen that lead to text and graphics. These topics vary depending on the disease or problem (Fig. 3 ). Each of these topics can be viewed or printed as needed. For example, a graphic from the topic 'symptoms', is available under American foulbrood (Fig. 4). The user can select other information areas On which disease would you like information? Pick as many as needed. Click OK to obtain the information or when finished.

[ ] American Foulbrood

[ ] Chalkbrood

[ ] European Foulbrood

[ ] Stonebrood

[ ] Sacbrood

[ ] No more information needed

C oK

I

Fig. 2. Information screenfor brood diseases.

SEE AWARE

] 17

Select items for more information on AFB.

[ ] Cause

[ ] Transmission

[-]Symptoms

[ ] Recommendations

[ ] Testing

[ ] No more information needed

IoK

J

[ ] Print selected items

Fig. 3. Information screen for American foulbrood.

Pupa "melted down" with pupal tongue visible

Healthy pupa

Scale with pupal tongue Fig. 4. Illustration from AFB information module showing disease progression.

118

J.E. McCLURE ET AL.

under the disease American foulbrood or move to different problem categories as desired. The system is designed so that the user can move easily around the information module.

Diagnosis. The diagnosis module helps the user to identify an unknown problem and is the core of the expert system. The system uses backward chaining until it narrows the problem down to one or two possible solutions and then forward chains to complete the diagnosis. The user is asked questions about the location and description of the unknown problem. The design of these questions follows logically in the way the beekeeper would most likely explore the problem himself. The questions start with the location of the problem in the hive (bottom board, entrance, honey supers, brood chamber, etc.). The questions proceed logically according to previous answers. A section of the decision tree for apparent damage to frames is shown (Fig. 5 ). The boxes with rounded ends denote questions, the rectangular boxes denote where the system has found an intermediate solution and will then forward chain to that area and find a final solution, and the darkened boxes denote final solutions. BEE AWARE contains 197 rules and 150 frames leading to 23 possible outcomes. When the system arrives at a solution, the user is asked if he needs more information, an explanation or a new diagnosis (if the current diagnosis is unsatisfactory). An option screen appears after a diagnosis is made (Fig. 6). After the system arrives at a probable solution, one of the options available to the user is an explanation, more information, a new diagnosis, or exit. The explanation is written in sentences that not only reflect the answers to the

Choose an item to continue.

New Diagnosis

1

I

Explain

IM°relnf°rmati°n

I

I

Exit

I

Fig. 5. Option screen after diagnosisis made.

BEEAWARE

1 19

questions, but also state some of the logic that the system used to come to this solution. The text is determined from the rules and questions that were used (Fig. 7).

References. The reference module consists of a list of references used in BEE AWARE, a glossary of terms relating to honey bees, a toxicity table, and a list of apiary inspectors and extension apiculturists in the U.S. and Canada. The glossary is arranged alphabetically. The user selects the first letter of the term he wants to define (Fig. 8 ). The system then displays the terms and Is the problem associated with a) wax

Is the problem associated with a) capped honey b) brood ceils

I .....

I

Is wood physically damaged? yes no

Are wax ceils damaged or Is Are frames spotted spotting (bee feces) evident? (bee feces)? a) damaged yes b) spotting = y j no

./

leC~l~l

\

Fig. 6. Decision tree of hive super problem of the BEE AWARE system. The problem is located on the bottom board, with the brood. Dead brood is present on the bottom board. The bees have been actively flying and bringing pollen or nectar Into the hive possibly exposing them to pesticides. There are 4 symptoms of pesticide poisoning. Regurgitated nectar Is an indicator of pesticide poisoning. Climbing and falling bees may indicate pesticide poisoning. Uncoordinated movements may indicate pesticide poisoning. Aggressive behavior when accompanied with other symptoms may indicate pesticide poisoning.

Fig. 7. Explanation for pesticide poisoning diagnosis.

120

J.E. McCLURE ET AL.

definitions that correspond to that letter. The glossary can be accessed during a diagnosis so that i f a technical term is used in a question during a diagnosis, the user can define that term. Once the term has been defined, the system returns to the current question. Glossary Choose any letter Choose Return when finished

I

Return

1

Fig. 8. Glossary access screen.

For a list of the local apiary inspector and expert, choose the appropriate button. To print press ~-P when the address is visible.

Canada

0 Puerto Rico 0 Alberta

O Nova Scotia

O British Columbia

0 Ontario

0 Manitoba

0 Prince Edward Island

0 New Brunswick

0 Quebec 0 Saskatchewan

Done

}

Fig. 9. Apiary inspector access screen.

I U.S. States 1

BEE AWARE

121

The reference list is a bibliography for the program. The user can use this to find more general information on bees and beekeeping. A toxicity table is also included that lists the relative toxicity to honey bees of selected pesticides according to residual times. This table may be used to help beekeepers, growers or pesticide applicators choose chemicals that are less toxic to honey bees or identify chemicals that may have been responsible for a bee kill. The apiary inspector list is divided into two access screens, one for the United States and the other for Puerto Rico and Canada (Fig. 9). The user selects the state or province, and a screen appears listing the inspector and university or extension personnel as appropriate.

Testing and delivery. During 1990, fourteen users tested the BEE AWARE expert system. This group included researchers, beekeepers, and extension personnel. From their direct input, changes were made to the system. Some of these changes included print out features and the addition of a glossary that is available during diagnostic sessions. The information module interface was changed to allow direct access and access after a diagnosis. More direct print out features were added to allow for printing of individual screens or of entire sections of information. Modules were interconnected to create additional paths for getting to an individual diagnosis. BEE AWARE was taken to five regional beekeeping meetings where beekeepers were asked to use the program as they wished and their comments were recorded. The system was also tested during apiculture classes conducted at Penn State in 1990 and 1991. BEE AWARE was made available for sale to individual beekeepers in the spring of 1991. The response so far has been good. Beekeepers from all over the country and from Germany and Spain have purchased the system. Pennsylvania county extension offices have been provided with copies of the expert system in order to extend the knowledge of state beekeeping specialists. Problem reports have been few. CONCLUSIONS

In order to protect honey bee colonies from diseases, parasites, pests and predators it is essential that beekeepers maintain strong, healthy colonies and develop the ability to recognize early, the symptoms of the various maladies affecting honey bees. The diagnostic capabilities of the BEE AWARE expert system can assist beekeepers in identifying problems early, making it possible for them to take appropriate action to correct or treat the problem. In addition, the ability of this system to handle experiential and experimental information in a timely fashion makes it ideal for delivering specific technical information on current control methods and recommendations. An expert system has certain advantages over a conventional reference or text book. First, an expert system can actually assist in diagnosing a particular

122

J.E. McCLURE ET AL.

problem. The knowledge of m a n y h u m a n experts can be incorporated into a system and, unlike a text book, an expert system can be easily and quickly updated. Unlike a h u m a n expert, BEE AWARE is always readily available and provides advice o f consistent content and quality. BEE AWARE can be used by beginning and experienced apiculturists to help with the problems o f increasing information management. However, more information could still be added to the system. Particularly useful would be an apiary m a n a g e m e n t section with information on queen management, seasonal management, honey production and record keeping. We hope to incorporate these additional features in future work.

REFERENCES Baker, G.M. and Torchio, P.F., 1968. New record ofAscosphaera apis from North America. Mycologia,60:189-190. Eischen, F.A., 1987. Overwintering performance of honey bee colonies heavily infested with Acarapis woodi (Rennie). Apidologie, 18: 293-304. Foster, M.A., Saunders, M.C., Mierzejewski,K. and Twardus, D., 1991. Development and validation of GypsEX: a knowledge-based system for aerial application of pesticides against gypsy moth. Comput. Electron. Agric., 5: 327-345. Furgala, B., Duff, S., Aboulfaraj,S., Ragsdale,D. and Hyser, R., 1989. Some effectsof the honey bee tracheal mite (Acarapis woodi Rennie) on wintering non-migratory, honey bee (Apis mellifera L. ) colonies in East Central Minnesota. Am. Bee J., 129: 195-197. Otis, G.W., Bath, J.B., Randall, D.L. and Grant, G.M., 1988. Studies of the honey bee tracheal a mite (Acarapis woodi Rennie) (Acari: Tarsonemidae) during winter. Can. J. Zool., 66: 2122-2127. Parsaye, K. and Chignell,M., 1988. Expert Systemsfor Experts. Wiley,New York, 395 pp. Plant, R.E. and Stone, N.D., 1991.Knowledge-BasedSystemsin Agriculture. McGraw-Hill,New York, 342 pp. Tomasko, M., 1990. Tracheal mite and nosema survey in Pennsylvania. BEE AWARE: Notes and news on bees and beekeeping. Penn State Coop. Ext. 59. Travis, J.W. and Latin, R.X., 1991. Development, implementation, and adoption of expert systems in plant pathology.Annu. Rev. Phytopathol., 29: 343-360.