- Email: [email protected]
Expert system for weed management of wheat fields in iran
Copyright © 2004 IF AC Fifth International Workshop on Artificial Intelligence in Agriculture, Cairo, Egypt
EXPERT SYSTEM FOR WEED MANAGEMENT OF WHEAT FIELDS IN IRAN A. Atri, E. Zand, M. Baghestanl and J. Khalghanl Plant Pests and Diseases Research Institute. P.o.Box 1454. Tehran 19395. Iran. [email protected] Wheat is the most important strategic crop in Iran. Weeds are one of the most important factors of yield loss in wheat. Computerized decision support systems offer a n ideal means of achieving economical, environmentally safe and sustainable weed m anagement. A practical expert system is necessary tom anage 0 fw eed control in wheat fields of Iran. This expert system has a step-by- step problem-solving procedure based on important factors relating to weed management. Expert system for weed management provides advice on the base of interactions of weeds and wheat including chemical and mechanical weeds control treatments and selects the best herbicide with 0 ptimum dosage and method of application. Economical recommendation can be made by estimating yield loss due to the weeds, wheat price, expected yield and herbicide cost. Recommendations a re base on wheat variety, date of planting; growth stage of wheat, weed species, weed and crop density, history of herbicide use and climatic factors. Keywords: Expert System, Weed management
Introduction Wheat is the most important crop in Iran, which covers 3.9 to 4.2 million hectares of agricultural lands (Roustaei et al.. 2002). 'Estimation of yield loss due to weeds of wheat fields is reported more than 25% in Iran (Zand et al.• 2002). Over the past decade, there has been considerable emphasis on using herbicides. Environmental safety concerns, increasing herbicide resistance in weeds and the need to reduce input costs have caused a growing awareness that intensive use of chemical weed control does not fit well in sustainable weed management system. This means that much more work is needed to improve weed management.
programs categorized typically in decision support tools. Weed management is an information -intensive task requiring the integration ef different aspects. These range from biological traits such as crop characteristic, weed flora, qualitative or quantitative composition, time 0 f emergence, relative herbicide efficacy, and weed competitiveness before and after control to economic aspects such as grain and herbicide costs. economic damage caused by the weeds and environmental considerations such as effect of different types of weed control (Berti et al., 2003).
Today, agriculture has been greatly influenced by Information Technology (IT). The different ITs like expert system in decision support system, remote sensing etc have brought caused evolution in many fields of agriculture. Even cooperation like ITC, MSSL, HLL are now looking forward to extract huge benefits out of this collaboration of IT with agriculture (Ghatak, 2002).
The first computerized systems concerning weeds were developed for integrated control in winter wheat (A arts et al., 1985). Recently, several others have been presented for weed management ( Black and Oyson, 1993; Castro-Tendero and GarciaTorres, 1996; Cussans and Rolph. 1990; Lybecker et al., 1991; Mortensen and coble. 1991; Murali et al., 1999; Renner and Black, 1991; Stigliani and Resina, 1993; Streibig et al., 1990; Swinton and King, 1994; Wiles et al., 1996; Wilkerson et al., 1991; Berti et al., 2003). These decision aids have generally fallen in to one of two categories (Montensen and coble. 1991), those that make recommendation based on herbicide efficacy ( Linker et al., 1990; Renner and Black, 1991; Stigliana and Resina, 1993; Thomon and WitIiamson, 1992) and those that consider weed seed bank or weed seedling density and make a
Expert systems as intelligent computer programs could have considerable effects on solving problems that are sufficiently difficult to require human expertise for their solution. These software programs emulate an expert by involving a client in a problem-solving situation, often providing a recommendation in response to a client's request for help in making decision. These software
116
The database will be easy Updating for each kind of data. The logical relations of all data necessary for running program is determined (Figure 2).
recommendation based on economic benefit (Berti and Zanin, 1997; Krishnan et al., 2001; Lybecker et al., 1991; Mortensen et al., 1999; Pannell, 1990; Wiles et al., 1996 Wilkerson et al., 1991)
The inference engine will process production rules and determine the correct path. The rules state the lines of reasoning based on scientific and heuristically knowledge from human experts, textbook and professional experience. Production rules will be processed in a deductive way (Stigliani and Resina, 1993).
In order to solve problems related weed management in wheat fields, weed research . department of plant pests and diseases research institute in collaboration with Ministry of agriculture decided to develop a fully functional version of an expert system for weed management. The o~iective of this paper is to explain the structure and steps of reasoning process of an expert system to help extension agents and farmers by providing advises on the most suitable economic solutions and recommends timely of herbicide application.
Steps in the reasoning process Recommendations will be present on the base of input data including wheat varieties in Iran (from the list), region (from the list), growth stage of wheat and weed (from the list), weed density, history of herbicide use and environmental factors specially precipitation. Final recommendations will be mechanical or chemical treatment with choosing best herbicide in optimum dosage and method of application under best economical situations.
Program Structure This expert system will be consisted of a knowledge base, inference engine and user interface (Figure 1). The knowledge base will be contained to all the information related to the problems. That will be consisted of name of weeds (scientific and local name); weed type (broad or narrow leaf), varieties of wheat, spectrum of herbicide efficacy depend on weed species, growth stage of weeds, weed density, growth stage of wheat, and environmental factors like precipitation. Some necessary economic information like price of wheat and herbicides registered in Iran and application cost will be added. All of this information will be listed in a prolog knowledge base, process by a pro log-based inference engine and will be embedded directly with C'" and Visual Basic languages.
,..... I"--...
This expert system will use expected crop selling price, herbicide prices, application cost and estimated yield loss and expected weed-free yield to calculate expected net return. Therefore, the reasoning process will be divided into three following main steps: • Input required data for determining conditions • Interaction between weed and wheat • Suggestion on mechanical or chemical control under optimum economics circumstances.
-..... __ ]
Weeds
User Input
Ic=:>
Inference Engine
Herbicides Yield loss
c=:> @
Environment
'-~---Knowledge Base Figurei-. System structure
117
Output
Figure2. Data required
Evaluation of competitive ability of weed species is the most important factor for . beginning control strategy. Stigiani and Resinia (1993) used three rules for determination of competitive ability of weeds based on economic threshold, density and life cycle of weeds. Some others programs used competitive indices based on economic threshold (Berti et al., 2003 and wilkerson et al., 1991).
Climatic factors especially precipitation and chilling days can affect on herbicides efficacy. So different forms of a herbicide might be advised in different climates. . In the case, which i~ needed to apply micro nutrition, treatment recommendation will be changed. This is due to sedimentation of some micro nutrition with herbicides.
n,is system will use zero economic thresholds. It is advised to adopt NST (no seed threshold) means that weeds should not be permitted to set seed .. It is based on the rationale that the seed rain sustains the seed bank of our important weed species (Norris, 20(0). Therefore. thi~ ~ystelll will always present control strategy under economics and environmental situations based on density and growth stage of weed.
Since there has been a rapid increase in the incidence of herbicide resistance worldwide and in order to minimize and manage the threat of herbicide resistance, history of herbicide use during the past 3 years will be asked. So, the system will advise suitable herblclde.wlth different mode or action. There will be also considerable emphasize to prevent using ALS (acetolactate synthase) inhibitor herbicides more than 3 years. HADSS· as a desktop program, designed 10 provide u full spectrum of weed control considered the ways to prevent increasing herbicide resistance too.
Weed control in wheat depends upon critical period of weed infestation. There are some limitations for selection and application of herbicides in wheat. For exumple. 2.4.D is un efTcctive herbicide for control of broad leaves which, should be applied before or early of elongation stage in wheat. Mechanical control will be advised on the base of herbicide and application cost, labor cost and yield loss due to weeds and in cases that there is some limitation to use herbicide like after elongation stage of wheat. (figure3)
• North Carolina State University, Raleigh, NC 27695-7620.
118
l
Growth Stage of Wheat
No
•
J
± After elongation growth stage
Yes
•
Estimated yield loss
List of Herbicides
-.t ....
--.J
..... ....
--110 ~
Economical Calculations Yield loss cost> Labor cost
Micro nutrition application
~
Precipitation/chilling days Forecasting
_.G
Mechanical Control
..... ....
--110 r
r
J
History of herbicide use During the Pllst 3 year
--'
Economical Calculations
~
Chemical Control
Figure3. Mechanical and chemical control
Refercnccs A arts, 1-I.F.M. and e.L.M. Visser (1985). A management information system for weed control ill winter. Proceeding of the British Crop Protectiol/ COl/l/cif. pp 679-686
Castro-Tcndero, A. J. and L. Garcia-Torres (1996). SEMAG1- an expert system for weed control decision making in sunflowers. Crop Prot, 14, 543-548
Berti, A. and G. Zanin (1997). GESTINF:a decision model for post-emergence weed management in soybcan (Glycine max (L.) Merr.). Crop Prot, 16, 109-116.
Cussans, G. W. and J. Rolph (1990). HERBMAST- a herbicide selection system for winter wheat. Proceeding of European Weed
Berti, A., B. Francesco and Z., Giuseppe (2003). Application of decision-support software for postemenrgence weed control. Weed Sci. 51, 618-627 Black, I. D. and C. B. Dayson (1993). An economic threshold model for spraying herbicides in cereals. Weed Res, 33, 279-290.
Research Society Symposium. Integrated Weed Management in Ceraels. Helsinki. Finland: European Weed Research Society. pp 451-457 Ghatak, malobika (2003). Use of information technology in agriculture. .[on line] www.indiainfoline.com/bisclitin.pdf [accessed November 10,2003].
119
Krishnan, G., D. A. Mortensen, A. R. Martin, L. B. Bills, A. Dieleman, and C. Nesser.(2001). WeedSOFT: a state of the art weed management decision support system. Weed Sci. Soc. Am. Abslr, 41,41-42. Linkcr, H. M., A. C. York and D. R. Wilhite Jr (1990). WEEDS-a system for developing a computer-based herbicide recommendation program. Weed Technol, 4, 380-385.
Wilkerson, G. G., S. A. Modena and H.D. Coble (1991). HERB: decision model for postemergence weed control in soybean. Agron. J, 83,413-417. Zand, E., MA Baghestani and P. Shimi (2002). The weeds of wheat fields in Iran and methods of control. In: From wheal Grains 10 loaves of bread Wheat congress press. pp450
Lybecker, D. W ., E . E . S chweizer and R. P. King (1991). Weed management decision based on bioeconomic modeling.Weed Sci, 39,124-129. Mortensen, D. A and H. D. Coble.(1991). Two approaches to weed control decision-aid software. Weed Technol, 5, 445-452. Mortensen, D. A., A. R. Martin, F. W. Roeth, et al (1999). WeedSOFT. Version 4.0 User's Manual. Lincoln, NE: Department of Agronomy, University of Nebraska. Norris, R.F (2000). My view. Weed Sci, 48, pp 273 Pannell, D. J (1990). An economic response model of herbicide application for weed control. Allst. J. Agric. Econ, 34, 223-241. Renner, K. A and 1. R. Black (1991). SOYHERB-a computer program for soybean herbicide decision making. Agroll. J, 83, 921-925. Roustaii, M., H. Ketata, K. H oseeiny, T. H osseinPour and M.H Hosni (2002). Weed improvement program in Dry areas of Iran. Proceedillg of First 11111. Wheat Congress. Irall-Tehran 7-IODec Stigliana, L. and C. Resina (1993). SELOMA: expert system for weed management in herbicide-intensive crops. Weed Techllol, 7, 550-559. Thomson, A. 1. and D. R. Williamson (1992). Formation and use of intermediate inference in advisory systems: a herbicide example. Al Applic, 6, 29-39 Wiles, L. 1., R. P. King, E. E. Schweizer, D. W. Lybecker, and S. M. Swinton (1999). GWM : general weed management model. Agric. Sysl, 50,355-376
120
EXPERT SYSTEM FOR WEED MANAGEMENT OF WHEAT FIELDS IN IRAN A. Atri, E. Zand, M. Baghestanl and J. Khalghanl Plant Pests and Diseases Research Institute. P.o.Box 1454. Tehran 19395. Iran. [email protected] Wheat is the most important strategic crop in Iran. Weeds are one of the most important factors of yield loss in wheat. Computerized decision support systems offer a n ideal means of achieving economical, environmentally safe and sustainable weed m anagement. A practical expert system is necessary tom anage 0 fw eed control in wheat fields of Iran. This expert system has a step-by- step problem-solving procedure based on important factors relating to weed management. Expert system for weed management provides advice on the base of interactions of weeds and wheat including chemical and mechanical weeds control treatments and selects the best herbicide with 0 ptimum dosage and method of application. Economical recommendation can be made by estimating yield loss due to the weeds, wheat price, expected yield and herbicide cost. Recommendations a re base on wheat variety, date of planting; growth stage of wheat, weed species, weed and crop density, history of herbicide use and climatic factors. Keywords: Expert System, Weed management
Introduction Wheat is the most important crop in Iran, which covers 3.9 to 4.2 million hectares of agricultural lands (Roustaei et al.. 2002). 'Estimation of yield loss due to weeds of wheat fields is reported more than 25% in Iran (Zand et al.• 2002). Over the past decade, there has been considerable emphasis on using herbicides. Environmental safety concerns, increasing herbicide resistance in weeds and the need to reduce input costs have caused a growing awareness that intensive use of chemical weed control does not fit well in sustainable weed management system. This means that much more work is needed to improve weed management.
programs categorized typically in decision support tools. Weed management is an information -intensive task requiring the integration ef different aspects. These range from biological traits such as crop characteristic, weed flora, qualitative or quantitative composition, time 0 f emergence, relative herbicide efficacy, and weed competitiveness before and after control to economic aspects such as grain and herbicide costs. economic damage caused by the weeds and environmental considerations such as effect of different types of weed control (Berti et al., 2003).
Today, agriculture has been greatly influenced by Information Technology (IT). The different ITs like expert system in decision support system, remote sensing etc have brought caused evolution in many fields of agriculture. Even cooperation like ITC, MSSL, HLL are now looking forward to extract huge benefits out of this collaboration of IT with agriculture (Ghatak, 2002).
The first computerized systems concerning weeds were developed for integrated control in winter wheat (A arts et al., 1985). Recently, several others have been presented for weed management ( Black and Oyson, 1993; Castro-Tendero and GarciaTorres, 1996; Cussans and Rolph. 1990; Lybecker et al., 1991; Mortensen and coble. 1991; Murali et al., 1999; Renner and Black, 1991; Stigliani and Resina, 1993; Streibig et al., 1990; Swinton and King, 1994; Wiles et al., 1996; Wilkerson et al., 1991; Berti et al., 2003). These decision aids have generally fallen in to one of two categories (Montensen and coble. 1991), those that make recommendation based on herbicide efficacy ( Linker et al., 1990; Renner and Black, 1991; Stigliana and Resina, 1993; Thomon and WitIiamson, 1992) and those that consider weed seed bank or weed seedling density and make a
Expert systems as intelligent computer programs could have considerable effects on solving problems that are sufficiently difficult to require human expertise for their solution. These software programs emulate an expert by involving a client in a problem-solving situation, often providing a recommendation in response to a client's request for help in making decision. These software
116
The database will be easy Updating for each kind of data. The logical relations of all data necessary for running program is determined (Figure 2).
recommendation based on economic benefit (Berti and Zanin, 1997; Krishnan et al., 2001; Lybecker et al., 1991; Mortensen et al., 1999; Pannell, 1990; Wiles et al., 1996 Wilkerson et al., 1991)
The inference engine will process production rules and determine the correct path. The rules state the lines of reasoning based on scientific and heuristically knowledge from human experts, textbook and professional experience. Production rules will be processed in a deductive way (Stigliani and Resina, 1993).
In order to solve problems related weed management in wheat fields, weed research . department of plant pests and diseases research institute in collaboration with Ministry of agriculture decided to develop a fully functional version of an expert system for weed management. The o~iective of this paper is to explain the structure and steps of reasoning process of an expert system to help extension agents and farmers by providing advises on the most suitable economic solutions and recommends timely of herbicide application.
Steps in the reasoning process Recommendations will be present on the base of input data including wheat varieties in Iran (from the list), region (from the list), growth stage of wheat and weed (from the list), weed density, history of herbicide use and environmental factors specially precipitation. Final recommendations will be mechanical or chemical treatment with choosing best herbicide in optimum dosage and method of application under best economical situations.
Program Structure This expert system will be consisted of a knowledge base, inference engine and user interface (Figure 1). The knowledge base will be contained to all the information related to the problems. That will be consisted of name of weeds (scientific and local name); weed type (broad or narrow leaf), varieties of wheat, spectrum of herbicide efficacy depend on weed species, growth stage of weeds, weed density, growth stage of wheat, and environmental factors like precipitation. Some necessary economic information like price of wheat and herbicides registered in Iran and application cost will be added. All of this information will be listed in a prolog knowledge base, process by a pro log-based inference engine and will be embedded directly with C'" and Visual Basic languages.
,..... I"--...
This expert system will use expected crop selling price, herbicide prices, application cost and estimated yield loss and expected weed-free yield to calculate expected net return. Therefore, the reasoning process will be divided into three following main steps: • Input required data for determining conditions • Interaction between weed and wheat • Suggestion on mechanical or chemical control under optimum economics circumstances.
-..... __ ]
Weeds
User Input
Ic=:>
Inference Engine
Herbicides Yield loss
c=:> @
Environment
'-~---Knowledge Base Figurei-. System structure
117
Output
Figure2. Data required
Evaluation of competitive ability of weed species is the most important factor for . beginning control strategy. Stigiani and Resinia (1993) used three rules for determination of competitive ability of weeds based on economic threshold, density and life cycle of weeds. Some others programs used competitive indices based on economic threshold (Berti et al., 2003 and wilkerson et al., 1991).
Climatic factors especially precipitation and chilling days can affect on herbicides efficacy. So different forms of a herbicide might be advised in different climates. . In the case, which i~ needed to apply micro nutrition, treatment recommendation will be changed. This is due to sedimentation of some micro nutrition with herbicides.
n,is system will use zero economic thresholds. It is advised to adopt NST (no seed threshold) means that weeds should not be permitted to set seed .. It is based on the rationale that the seed rain sustains the seed bank of our important weed species (Norris, 20(0). Therefore. thi~ ~ystelll will always present control strategy under economics and environmental situations based on density and growth stage of weed.
Since there has been a rapid increase in the incidence of herbicide resistance worldwide and in order to minimize and manage the threat of herbicide resistance, history of herbicide use during the past 3 years will be asked. So, the system will advise suitable herblclde.wlth different mode or action. There will be also considerable emphasize to prevent using ALS (acetolactate synthase) inhibitor herbicides more than 3 years. HADSS· as a desktop program, designed 10 provide u full spectrum of weed control considered the ways to prevent increasing herbicide resistance too.
Weed control in wheat depends upon critical period of weed infestation. There are some limitations for selection and application of herbicides in wheat. For exumple. 2.4.D is un efTcctive herbicide for control of broad leaves which, should be applied before or early of elongation stage in wheat. Mechanical control will be advised on the base of herbicide and application cost, labor cost and yield loss due to weeds and in cases that there is some limitation to use herbicide like after elongation stage of wheat. (figure3)
• North Carolina State University, Raleigh, NC 27695-7620.
118
l
Growth Stage of Wheat
No
•
J
± After elongation growth stage
Yes
•
Estimated yield loss
List of Herbicides
-.t ....
--.J
..... ....
--110 ~
Economical Calculations Yield loss cost> Labor cost
Micro nutrition application
~
Precipitation/chilling days Forecasting
_.G
Mechanical Control
..... ....
--110 r
r
J
History of herbicide use During the Pllst 3 year
--'
Economical Calculations
~
Chemical Control
Figure3. Mechanical and chemical control
Refercnccs A arts, 1-I.F.M. and e.L.M. Visser (1985). A management information system for weed control ill winter. Proceeding of the British Crop Protectiol/ COl/l/cif. pp 679-686
Castro-Tcndero, A. J. and L. Garcia-Torres (1996). SEMAG1- an expert system for weed control decision making in sunflowers. Crop Prot, 14, 543-548
Berti, A. and G. Zanin (1997). GESTINF:a decision model for post-emergence weed management in soybcan (Glycine max (L.) Merr.). Crop Prot, 16, 109-116.
Cussans, G. W. and J. Rolph (1990). HERBMAST- a herbicide selection system for winter wheat. Proceeding of European Weed
Berti, A., B. Francesco and Z., Giuseppe (2003). Application of decision-support software for postemenrgence weed control. Weed Sci. 51, 618-627 Black, I. D. and C. B. Dayson (1993). An economic threshold model for spraying herbicides in cereals. Weed Res, 33, 279-290.
Research Society Symposium. Integrated Weed Management in Ceraels. Helsinki. Finland: European Weed Research Society. pp 451-457 Ghatak, malobika (2003). Use of information technology in agriculture. .[on line] www.indiainfoline.com/bisclitin.pdf [accessed November 10,2003].
119
Krishnan, G., D. A. Mortensen, A. R. Martin, L. B. Bills, A. Dieleman, and C. Nesser.(2001). WeedSOFT: a state of the art weed management decision support system. Weed Sci. Soc. Am. Abslr, 41,41-42. Linkcr, H. M., A. C. York and D. R. Wilhite Jr (1990). WEEDS-a system for developing a computer-based herbicide recommendation program. Weed Technol, 4, 380-385.
Wilkerson, G. G., S. A. Modena and H.D. Coble (1991). HERB: decision model for postemergence weed control in soybean. Agron. J, 83,413-417. Zand, E., MA Baghestani and P. Shimi (2002). The weeds of wheat fields in Iran and methods of control. In: From wheal Grains 10 loaves of bread Wheat congress press. pp450
Lybecker, D. W ., E . E . S chweizer and R. P. King (1991). Weed management decision based on bioeconomic modeling.Weed Sci, 39,124-129. Mortensen, D. A and H. D. Coble.(1991). Two approaches to weed control decision-aid software. Weed Technol, 5, 445-452. Mortensen, D. A., A. R. Martin, F. W. Roeth, et al (1999). WeedSOFT. Version 4.0 User's Manual. Lincoln, NE: Department of Agronomy, University of Nebraska. Norris, R.F (2000). My view. Weed Sci, 48, pp 273 Pannell, D. J (1990). An economic response model of herbicide application for weed control. Allst. J. Agric. Econ, 34, 223-241. Renner, K. A and 1. R. Black (1991). SOYHERB-a computer program for soybean herbicide decision making. Agroll. J, 83, 921-925. Roustaii, M., H. Ketata, K. H oseeiny, T. H osseinPour and M.H Hosni (2002). Weed improvement program in Dry areas of Iran. Proceedillg of First 11111. Wheat Congress. Irall-Tehran 7-IODec Stigliana, L. and C. Resina (1993). SELOMA: expert system for weed management in herbicide-intensive crops. Weed Techllol, 7, 550-559. Thomson, A. 1. and D. R. Williamson (1992). Formation and use of intermediate inference in advisory systems: a herbicide example. Al Applic, 6, 29-39 Wiles, L. 1., R. P. King, E. E. Schweizer, D. W. Lybecker, and S. M. Swinton (1999). GWM : general weed management model. Agric. Sysl, 50,355-376
120