"... In PAGE 47: ... Value diversity On the basis of on Vesely and Rasmuson apos;s (1984) classification for sources of uncertainty and Funtowicz and Ravetz apos; (1990) classification for types of uncertainty, Van der Sluijs (1995) has proposed a two-dimensional classification scheme defining areas to be addressed in uncertainty management in IAMs. This scheme is presented in Table5 . Inexactness refers to error-bars, probability distribution... In PAGE 49: ... Further, in order to obtain insight into error propagation in the entire integrated model one needs to do more than simply apply Monte Carlo modeling solely to isolated sub models. The IMAGE example shows that even in the area of uncertainty management where adequate tools are currently available (namely the upper left corner of Table5 ), the mapping of uncertainties is only partly done because of lack of resources and competing priorities (Van der Sluijs, 1995). 6.... In PAGE 77: ... vii When IAMs are made stochastic, it should be possible to use them for goal-searching in order to find solutions that are robust against the specified uncertainty in the model. viii We identified a mismatch between the types and sources of uncertainty that should be addressed on the one hand ( Table5 ) and the current practice of uncertainty management in IAMs and the tools available, on the other hand. From our analysis it follows that techniques currently available for uncertainty analysis and uncertainty treatment in IAMs have three major shortcomings: 1.... ..."

"... In PAGE 47: ... Value diversity On the basis of on Vesely and Rasmuson apos;s (1984) classification for sources of uncertainty and Funtowicz and Ravetz apos; (1990) classification for types of uncertainty, Van der Sluijs (1995) has proposed a two-dimensional classification scheme defining areas to be addressed in uncertainty management in IAMs. This scheme is presented in Table5 . Inexactness refers to error-bars, probability distribution... In PAGE 49: ... Further, in order to obtain insight into error propagation in the entire integrated model one needs to do more than simply apply Monte Carlo modeling solely to isolated sub models. The IMAGE example shows that even in the area of uncertainty management where adequate tools are currently available (namely the upper left corner of Table5 ), the mapping of uncertainties is only partly done because of lack of resources and competing priorities (Van der Sluijs, 1995). 6.... In PAGE 77: ... vii When IAMs are made stochastic, it should be possible to use them for goal-searching in order to find solutions that are robust against the specified uncertainty in the model. viii We identified a mismatch between the types and sources of uncertainty that should be addressed on the one hand ( Table5 ) and the current practice of uncertainty management in IAMs and the tools available, on the other hand. From our analysis it follows that techniques currently available for uncertainty analysis and uncertainty treatment in IAMs have three major shortcomings: 1.... ..."

"... In PAGE 9: ... We note that this choice of is not necessarily the best choice in terms of accuracy because of the trade-o between accuracy and resolution in space. But, because of stability, it does allow integration in time with a timestep which is signi cantly larger than that allowed by the Chebychev method, in fact for N = 128 a timestep some 16 times larger can be employed, see Table3 . In cases where physically the time evolution on the small scale is not required this can lead to an enormous reduction in computational e ort.... ..."

"... In PAGE 1: ... Each method uses an optimized mutation rate for that method. The results of testing the algorithm on the two-dimensional landscapes are given in Table1 . The Tukey-Kramer test was... ..."

"... In PAGE 7: ... The fIrSt is based on point scatter graphs, and the second simulates a terrain representation. Point Visualization The two-dimensional coordinates ( Table4 ) were input to Arc View in the fonD of event files (ESRI terminology) which made them readily available for visualization. The coordinate files were linked to ~e original keyword file (refer back to Table 1) through common identifiers, in straightforward GIS manner.... ..."

"... In PAGE 8: ... The top level separator for problems \airfoil 1 quot; and \venkat 2 quot; are shown in Figures 1 and 2; the vertices in the separator are marked by the symbol \ . quot; The results for two dimensional problems are summarized in Table3 . For = 1=3, the sizes of the subgraphs are balanced to within a factor of 2.... ..."

"... In PAGE 8: ... The top level separator for problems \airfoil 1 quot; and \venkat 2 quot; are shown in Figures 1 and 2; the vertices in the separator are marked by the symbol \ . quot; The results for two dimensional problems are summarized in Table3 . For = 1=3, the sizes of the subgraphs are balanced to within a factor of 2.... ..."

"... In PAGE 8: ... The top level separator for problems #5Cairfoil 1 quot; and #5Cvenkat 2 quot; are shown in Figures 1 and 2; the vertices in the separator are marked by the symbol #5C#0F. quot; The results for two dimensional problems are summarized in Table3 . For #0B =1=3, the sizes of the subgraphs are balanced to within a factor of 2.... ..."

"... In PAGE 9: ... In our study we have two variables, real risk and predicted risk, that can assume only two discrete values, low and high, in a nominal scale. Thus the data can be represented by a two-dimensional contingency table, shown in Table3 , with one row for each level of the variable real risk and one column for each level of the variable predicted risk. The intersections of rows... ..."

"... In PAGE 9: ... In our study we have two variables, real risk and predicted risk, that can assume only two discrete values, low and high, in a nominal scale. Thus the data can be represented by a two-dimensional contingency table, shown in Table3 , with one row for each level of the variable real risk and one column for each level of the variable predicted risk. The intersections of rows... ..."