PARENT SESSION
21 - Probabilistic Methods in Risk Assessment
8:00 AM to 6:30 PM, Monday, 13 May 2002
Exhibition Area

(21-16) Multi criterion ranking of potential environmental impact in GIS.

Sørensen, Peter^*,1, Lerche, Dorte¹, Fauser, Patrik¹, Thomsen, Marianne¹, ¹ National Environmental Resarch Institute, Fredriksborgsvej 399, PostBox 358, Roskilde, Denmark

ABSTRACT- A spatial ranking method is developed using Partial Order Theory combined with environmental data calculated using GIS. The methodology will be exemplified using data on spatial distributed pesticide usage. The pesticide usage is calculated in GIS (for Denmark) and used to find the spatial distribution of the potential environmental impact in relation to surface water eco-toxicology. The input for the GIS is geo-reference information on crop type combined with information about the active ingredient toxicity and real data records of dosage levels for every crop type. About 100.000 locations are characterised in this system in terms of the three different criteria for environmental impact. The partial order theory using so-called linear extensions appears to be a useful tool for such a multi criterion ranking. The use of linear extensions is a way to transform the partial order of the GIS locations into a total order that comply with all the element relations in the partial order. The probability for a specific total order can be calculated from the system of all possible linear extensions. Unfortunately, it is typically not possible, even for a supercomputer, to find all linear extensions, for partial ordered sets including more than about 20 elements. However, an estimation of the total order using only a randomly chosen fraction (sub set) of all linear extension decreases the calculation time extensively and it is shown how this method can analyse even very large ranking problems like the GIS data result for pesticides.

Key words: partial order, linear extensions, ranking probability