<p>In the design of complex engineering systems, the high computational cost of high-precision simulation technology is mostly reduced by constructing surrogate models. Previous studies have proposed hybrid surrogate models to improve the accuracy and expand the scope of application of a specific single surrogate model. However, it has been observed that the global average hybrid surrogate model is difficult to adapt to the local characteristics of the prediction points, and the point-by-point weighted hybrid surrogate model may have poor global fitting accuracy. In this paper, a novel hybrid surrogate modeling method is proposed by combining global and local error measures, which is called pointwise weighted hybrid surrogate model based on hybrid measures (PWHSMHM). In the PWHSMHM, the benchmark model is selected by leave-one-out (LOO) cross-validation, and the local error measure algorithm of surrogate model (LEMASM) is proposed to estimate the local uncertainty of the sub-proxy model based on the sample density near the prediction point. The relationship between the global error measure and the local error measure is adaptively weighted by the global weight coefficient. The experimental results on 40 benchmark functions confirm that PWHSMHM not only has higher fitting accuracy than single surrogate models but also outperforms existing hybrid surrogate models in this regard. Finally, the proposed method is applied to optimize the automobile front cover to prove that it can be a new choice for the design optimization of complex engineering systems.</p>