Materials and methods Directory UMM :Data Elmu:jurnal:A:Agriculture, Ecosystems and Environment:Vol80.Issue1-2.Aug2000:

144 O. Roussel et al. Agriculture, Ecosystems and Environment 80 2000 143–158 the Study Centre for a More Autonomous Agricultural Development Centre d’Etude pour un Développement Agricole Plus Autonome CEDAPA was founded in central Brittany. This group, consisting currently of about 100 farmers, proposes a pasture-based milk and meat production system Pochon, 1993. The system should allow a good income and, as the result of a re- duced work load, a better quality of life for the farmer. CEDAPA further claims that, because of a reduction of external inputs, farms are more autonomous, have lower costs and cause less environmental damage. CEDAPA farmers follow guidelines concerning pas- ture and crop management practices, animal feeding and landscape maintenance Pochon, 1998. The pro- duction system aims at a maximum amount of forage in the diet of the cows, mainly provided by a ryegrass — white clover sward. This allows a reduction of the amount of concentrated feed needed to complement the diet. Within this context, annual crops are grown to provide additional forage during winter. According to CEDAPA guidelines, the area dedicated to silage maize should not exceed 15 of the total area ded- icated to pasture and annual forage crops. Silage maize is considered to be a potentially polluting crop mainly because it does not cover the soil during win- ter and spring, and thus, may favour erosion, runoff and leaching. More generally, the guidelines limit ni- trogen fertilisation and use of pesticides for all crops, in order to reduce production costs and pollution risks Pochon, 1998. In 1993 a research programme Système Terre et Eau was set up to evaluate the CEDAPA production system. Within this research programme, a study was carried out in 1996 to characterise pesticide use by CEDAPA farmers. The objectives of this survey were to: 1. Evaluate the way farmers plan their crop pest man- agement; 2. Evaluate the environmental effect of pesticide use by CEDAPA farmers. The results of the survey and an evaluation of crop protection strategies have been reported by Cavelier et al. 1997. The evaluation of the environmental effect of pesticides used by CEDAPA farmers is the subject of this paper. This evaluation can be achieved in a variety of ways. For example, it could be based on measurement of relevant variables or on their estimation by means of a mathematical simulation model Stockle et al., 1994. Neither of these approaches was used here on the grounds of cost and non-availability of an adequate simulation model. Bockstaller et al. 1997 proposed a set of ‘agro-ecological indicators’ as an alternative to measurements and the use of simulation models to evaluate the environmental effects of farming systems. The term ‘indicator’ has been defined as a variable which supplies information on other variables which are difficult to access Gras et al., 1989. Indicators synthesise information and can thus help understand- ing of a complex system Girardin et al., 1999. Sev- eral indicator-type approaches have been proposed to assess pesticide effects Shahane and Inman, 1987; Levitan et al., 1995; Van der Werf, 1996. Here the indicator Ipest Van der Werf and Zimmer, 1998 was used, because it takes into account pesticide proper- ties, site-specific conditions and characteristics of the pesticide application. Ipest is suitable because the in- put variables it requires are available for the farms and the hydrogeologic conditions of Brittany.

2. Materials and methods

2.1. Data used A pesticide application is defined as the application of a single pesticide active ingredient. A commercial product for the control of pests may contain more than one active ingredient, and a pesticide spray treatment carried out by a farmer may contain one or more com- mercial products. Thus, with this definition, a single treatment often corresponds to several pesticide appli- cations. Data on pesticide applications were collected during the 1994–1995 and the 1995–1996 cropping seasons on the farms of 23 CEDAPA members Cave- lier et al., 1997. On a total of 31 crops of winter wheat, 16 maize crops and 14 fodder beet crops 163 treatments were made, corresponding to 329 pesticide applications and involving 54 active ingredients. Pes- ticides applied as seed treatments were ignored. The pesticide characteristics used to calculate Ipest values of pesticide applications were taken from sev- eral databases, as presented in Van der Werf and Zim- mer 1998. Ipest values were calculated for pesticide active ingredients only, not for adjuvants. For nine O. Roussel et al. Agriculture, Ecosystems and Environment 80 2000 143–158 145 Table 1 Input variables and indicator modules Presence, Rsur, Rgro and Rair for the Ipest fuzzy expert system a Input variables Indicator modules Presence Rsur Rgro Rair Pesticides properties: Field half-life — DT50 X X Leaching potential — GUS X Volatility — K H X Aquatic toxicity — Aquatox X Human toxicity — ADI X X Site-specific conditions: Drift percentage X Runoff risk X Leaching risk X Application factors: Rate of application X Position of application X X X a Sources for pesticide properties are in Van der Werf and Zimmer, 1998. active ingredients corresponding to 34 pesticide ap- plications, Henry’s law constant was not available. In this case the value 2.65×10 − 5 was used, which cor- responds to the median value of the transition interval see next section. 2.2. The structure of the pesticide indicator This study used Ipest-B B for Brittany, a modi- fied and enhanced version of the indicator Ipest Van der Werf and Zimmer, 1998, to estimate the potential environmental effect of pesticide applications, accord- ing to the authors’ expert perception. Ipest-B has been adapted to the hydrogeologic conditions of Brittany: shallow aquifers dominated by surface or subsurface waterflows. The Ipest fuzzy expert system will be out- lined here, full details being given by Van der Werf and Zimmer 1998. Ipest consists of four modules. The module Pres- ence reflects the rate of application of the pesticide, the modules Risk of surface water contamination Rsur, Risk of groundwater contamination Rgro and Risk of air contamination Rair reflect the risk for three major environmental compartments. The values of the modules depend on a total of 10 input variables Table 1. Three types of input variables are distinguished: a pesticide properties; b site-specific conditions; c characteristics of the pesticide application. The four indicator modules can be considered individually or can be aggregated into an overall indicator reflecting the total potential environmental influence of a pesti- cide application. This modular structure is flexible: the mode of aggregation of the modules can be changed and new modules e.g. terrestrial biota, soil can be added as availability of data and understanding of pes- ticide effects evolve. This flexibility allowed adapta- tion of the system to the hydrogeologic conditions of Brittany. For each module, a value on a dimensionless scale between 0 no risk of environmental effect and 1 maximum risk of environmental effect is calculated. These values are calculated according to decision rules and to the degree of membership of the input variables to ‘fuzzy’ subsets. The mechanism will be explained briefly below. For all input variables given in Table 1 two fuzzy subsets F Favourable, i.e. the set of values which are considered to give rise to acceptable environmen- tal effect and U Unfavourable, i.e. the set of val- ues which are considered to give rise to unacceptable environmental effect were defined. The membership of values of input variables to the fuzzy subsets F and U is defined by a membership function, which can take any value from the interval [0, 1]. The value 0 represents complete non-membership, the value 1 represents complete membership; values in between are used to represent partial membership. Membership functions have been defined such that the value of an input variable either belongs fully to one of the two fuzzy subsets or partially to both, in the latter case the value is within a ‘transition interval’. Calculations are carried out according to a set of rules of the type: IF premise THEN conclusion. This may be illustrated by an example. Assuming that the output variable Environmental effect of the appli- cation of a pesticide depends on two input variables only: Rate of application and pesticide Field half-life. A first decision rule of the expert system might be: “If Rate of application is favourable i.e. low and if Field half-life is favourable i.e. short then Environmental effect is 0 no effect”. This rule is summarised in the first line of Table 2, the rest of the table shows the decision rules for the other three situations. Sugeno’s inference method Sugeno, 1985, which will not be detailed here, allows a weighting of the values of the 146 O. Roussel et al. Agriculture, Ecosystems and Environment 80 2000 143–158 Table 2 Summary of decision rules describing the effect of the input variables Rate of application and DT50 on the hypothetical module Environmental effect a Rate of application DT50 Environmental effect F F 0.0 F U 0.5 U F 0.5 U U 1.0 a F: Favourable; U: Unfavourable. conclusions of the decision rules to obtain a value for the module. Fig. 1 shows the decision rules from Table 2. This mode of presentation is followed in this paper.

3. Results