We perform seismic hazard analysis for three urban centers (Cali, Colombia; Quito, Ecuador; and Santiago, Dominican Republic) within the TREQ project, explicitly accounting for the local soil response in each city. The two requirements for this analysis are (1) hazard estimates on reference bedrock at each city, for which we use the models described in D2.2.2 and D2.2.3 and (2) soil response models that quantify the amplification (or deamplification) of ground shaking throughout the three urban centers due to the shallow soil layers. In this study, in collaboration with local experts, we develop the soil response models for each city using available local geotechnical and geophysical data. We simulate the soil response using 1D equivalent linear analysis using pySRA, and a large suite of input motions generated from (and therefore fully compatible with) the underlying hazard model, which accounts for uncertainty in the input motions. The computed soil response models consist of a set of soil amplification factors (AF) and their uncertainty (�lnAF), covering the respective urban centers. The AFs are defined for periods relevant for risk analysis (PGA – 2.0 s), as well as for a wide range of bedrock shaking intensity levels (0.05 – ~4 g), and can therefore be readily used for probabilistic hazard and risk analysis. Finally, hazard curves are computed at the surface by convolving the bedrock hazard with the AFs and �lnAF at a set of sites. The results are compared to those computed using the ergodic site terms within the GMPEs, which depend on the shear wave velocity in the upper 30 m (Vs30), for which we find significant differences at some sites. Specifically, using the local soil response models can either increase (up to a factor ~2) or decrease the hazard results compared to when an inferred Vs30 is used. For soft sites at longer return periods, the hazard can be lower (and sometimes lower than the hazard on rock) due to soil nonlinearity. These differences can have a significant impact on the risk results, which are described in Deliverable D2.3.5. This study demonstrates the importance of incorporating local soil response when the goal is to model hazard at the urban scale with a higher level of detail compared to more standard approaches using Vs30.