SEMAGI - AN EXPERT-SYSTEM FOR WEED-CONTROL DECISION-MAKING IN SUNFLOWERS

To use herbicides efficiently, decision makers must estimate when weed populations exceed economic treatment thresholds. An interactive micr ocomputer program named SEMAGI has been developed for sunflower (Helia nthus annuus L.) to evaluate the potential yield reduction from multi- species weed infestations and from the parasitic weed broomrape (Oroba nche cernua/O. cumana), and to determine the appropriate selection of herbicides. It combines relational databases on herbicides, weeds and their interaction. Originally, 34 weed species and twenty-six herbicid es were introduced specifying each weed/herbicide efficacy combination . For other agricultural situations, SEMAGI permits the introduction o f new weeds (up to 80), new herbicides (up to 40) and each herbicide-w eed efficacy combination. The expert system processes and selects the herbicide(s) under the constraints of herbicide efficacy data and of a weed-crop competition model. This relates weed-infested crop yield (S YI), potential weed-free yield (SYF), weed density (RD) and weed bioma ss (RBio). The user evaluates the weed infestation by field survey or density counting and the program converts it into equivalent weed biom ass. Weed species are classified in three groups according to their fi nal size. A relationship between weed density, weed size and equivalen t biomass is established for any weed group. In addition, SEMAGI provi des an economic study of any herbicide treatment selected or introduce d by the user, based on herbicide treatment cost, expected yield incre ase from the weed control treatment and sunflower selling price. A com puter capable of running MS-DOS or PC-DOS version 2.0 or greater with a minimum of 2 M bytes of RAM is required. This approach should be app licable to other crops.