Optimization criteria for the representability of numerical models for the estimation of relative reactivity changes, due to localized perturbations in boiling water reactor (BWR) lattices, have been theoretically developed and tested. The validity of the derived theoretical expressions has been assessed for the case of a reactivity perturbation corresponding to the removal of an individual fuel pin from a nominal BWR assembly, thus effectively substituting the pin by water. Such reactivity effects are of importance in the context of evaluating advanced fuel element designs, e.g., those employing part-length rods. Two different geometry models have been implemented for the LWR-PROTEUS critical research facility [full core (FC) and a smaller, reduced geometry (RG)], using the light water reactor assembly code BOXER, and calculations have been performed for the nominal cases (all pins present in the central test assembly) and the perturbed cases (individual pins removed). The FC results have been compared with the results of the RG model with two different boundary conditions (reflective and critical albedo). The comparisons have shown that the results of critical albedo calculations feature superior representability. Differences in relative reactivity effects, with respect to results of the FC calculation, are found to be within the range 1 to 4%.