ELSEV IER Crop Profech’on Vol. 16. No. 2, pp. 10’+116, 1997 0 1997 Elsevier Science Ltd PII: SO261-2194(96)0008&I-9 Printed I” Great Brian. All rights reserved 0261-2194/97 $17.00 + 0.00 GESTINF: a decision model for post-emergence weed management in soybean (Glycine znax (L.) Merr .) A. Berti and G. Zanin Centro di Studio sulla Biologia ed il Legnaro (PD), Italy Controllo delle Piante Infestanti, C. N. R., Agripolis, 35020 The selection of the best weed control option can be improved using decision-support systems considering the different factors affecting the efficacy (weed species, growth stage, climatic conditions) and the economics of the treatments. An interactive microcomputer program called GESTINF has been developed to assist in the selection of weed control options in soybean and winter wheat. Using observed weed densities, crop weed-free yield and grain price as input data, the program estimates potential crop damage from multispecies weed complexes and ranks the different weed control options according to expected net returns. The program also gives estimates of yield loss due to weeds surviving the treatment and an environmental index indicating how hazardous the treatment is for the water-table, thus allowing a selection of treatments both on an economic and an environmental basis. The system has been tested for 4 years in different locations of northteastern Italy. The system forecasted the yield losses observed in the field fairly accurately and proved capable of selecting appropriate interventions on the basis of type of flora and weed growth stage. @ 1997 Elsevier Science Ltd. All rights reserved. Keywords: weed interference; decision aids; economic thresholds; soybean, environmental impact The economic threshold concept embodies the notion that some functional relationships exist between weed density, relative time of weed emergence, efficacy of control and crop yield. If these relationships are known or can be estimated, then an economic analysis can be undertaken. To establish an economic threshold it is necessary to predict: (a) the effect the weed will have on the crop; (b) crop yield in the absence of weeds; (c) the efficacy of the weed control methods available (kill rate); (d) the price to be realised for the crop and for assessing the costs of control methods. Excluding herbicide cost, all the other points listed require estimations of phenomena that are subject to a wide variability depending on specific environmental characteristics, crop practices and, for the price of agricultural products, external factors. The economic threshold concept has been subjected to some criticism (Zanin, Berti and Sattin, 1994) because a single threshold does not exist for a given weed species in a given crop. Even without taking into account the competitivity of a species being modified by environmental conditions and time of emergence, which can substantially alter its behaviour (Bauer et al., 1991; Morin, Blanc and Darmency, 1993), as many thresholds exist as there are possible treatments and these values depend on the combination between herbicide efficacy and cost. Furthermore, herbicides are almost always characterised by a wide action spectrum that complicates the implementation of thresholds: the control treatment for one specific weed also affects the other weeds so its economic validity must be evaluated taking the whole infestation into account. The selection of the best weed control option thus requires a decision-support system that must consider the different factors affecting the efficacy (weed species, growth stage, climatic conditions) and the economics of the treatments. Some examples of this type of system have been presented in recent years (Aarts and Visser, 1985; Black and Dyson, 1993; Castro-Tender0 and Garcia-Torres, 1996; Cussans and Rolph, 1990; Forcella, Buhler, Swinton and King, 1993; Gerowitt and Heitefuss, 1990; Streibig, Andreasen and Fredshavn, 1990; Wilkerson, Modena and Coble, 1991). At the CNR Centre for Weed Biology and Control a decision making system for post- emergence weed control in soybean and winter wheat, called GESTINF, has been developed at the end of a research project that began in 1985. To assess the validity of this type of system two types of evaluation are required: (a) on the capability of the system to estimate yield losses and to identify the more suitable weed control method for a given infestation; and (b) on the impact of this type of approach at the farm level, then evaluating the economic consequences of a change in the decision making process. Crop Protection 1997 Volume 16 Number 2 109