Arrays of large numbers of sensors and accompanying complex system identification (SI) methods have been recently used in engineering applications on structures ranging from complex real space trusses to simple experimental beams. However, practical application to strong motion data recorded on large civil engineering systems is limited. In this study, state-space identification methods are used for modal identification from earthquake records with further investigation into the effectiveness of the methods from the viewpoint of sensor layout configuration. The study presented in this paper adopts a deterministic approach, which is complemented with statistical evaluation in another paper. The used SI methods include eigen realization, system realization with information matrix and subspace methods. The application of the methods on instrumented bridge systems in California is included and the performance of these methods is compared in terms of success and feasibility. Subsequently, viability of the methods on arrays of different numbers of sensors on the bridge systems is investigated. This is motivated by the fact that sensor arrays with wireless communication may be conveniently installed on civil engineering structures in the future. Computational costs and limits of applicability are examined using simulated ground motion analysis with detailed finite element models for a bridge system after validation using recorded data from real ground shaking. Moreover, the effect of the configuration of the sensor arrays is considered, accounting for different noise levels in the data to reflect more realistic situations. Copyright 0 (c) 2005 John Wiley & Sons, Ltd.