7.5.2 Disadvantages of Simulation



The disadvantages of simulation as a technique are also primarily due to the fact that simulation is basically an experimental (or empirical) approach to solving problems. As a result (and this is probably the most important disadvantage) it is difficult to develop cause-and-effect relationships through simulation, especially when the system under consideration requires the specification of many input parameters and involves complex interactions, as is most often the case with urban systems. Despite the fact that one can conduct sensitivity analyses by varying systematically some (or all) of the simulation inputs, the range of values that can be tried is usually limited due to time or cost limitations. In view of this, it is dangerous to extrapolate and make predictions on the basis of whatever functional relationships between inputs and outputs one is able to develop through simulation.

A second disadvantage is that, like all empirical techniques, simulation tends to be a relatively expensive mathematical tool. It is almost impossible to find examples where the resources (in terms of manpower, money, and time) necessary to develop, validate, and run an urban simulation model which is truly useful to the analyst or to the agency for which it was developed were not seriously underestimated at the outset.

Finally, as with all experimental results (and, actually, more so) the statistical analysis of simulation results is difficult. What is the effect of the starting conditions of the simulation on the final results? How many data points should be disregarded as reflecting primarily the starting conditions and not the long-term characteristics of the simulated system? What is the statistical confidence that can be attached to the results ? In the course of a parametric analysis, have we discovered a local or a global optimum set of operating conditions? These are all important questions which, because of space limitations, have not been addressed in this chapter. Good introductory treatments can be found in Naylor et al. [NAYL 69] and in Gross and Harris [GROS 74), and a more advanced treatment in Fishman [FISH 73, 78].