Model Order Reduction (MOR) techniques have emerged as powerful tools for alleviating the computational burden associated with the repeated solution of large-scale, parameter-dependent partial differential equations (see, e.g., [1], the monograph [2]). However, many classical MOR approaches remain highly sensitive to parameter variations, leading to significant accuracy degradation when parameters deviate from the training set and often requiring costly basis reconstruction or enrichment. This limitation is particularly critical in inverse problems such as Full-Waveform Inversion (FWI), where frequent model updates are intrinsic to the solution process. In this work, we introduce a new reduced-order modeling strategy, termed MOR-TL, designed to enhance robustness with respect to parameter changes while maintaining high computational efficiency [3]. The method is based on the systematic construction of reduced bases using local Taylor polynomial expansions with respect to model parameters. By exploiting Fréchet derivatives of the solution--computed through a multiple right-hand-side strategy--MOR-TL enables the efficient generation of reduced bases with minimal additional computational cost, without the need for extensive retraining or parameter-specific tuning. Numerical experiments on two-dimensional wave propagation problems relevant to seismic applications demonstrate that MOR-TL remains accurate and stable under relatively large parameter perturbations, handling variations of up to 20%. These results highlight the method’s robustness in noisy settings and its effectiveness in optimization procedures such as line-search-based minimization. We further integrate MOR-TL into a Full-Waveform Inversion framework for the reconstruction of two-dimensional subsurface velocity models. The resulting inversion strategy achieves a reduction of approximately 50% in computational cost compared to a high-fidelity spectral element method (SEM), while preserving a comparable level of reconstruction accuracy. These findings demonstrate that MOR-TL provides a robust and efficient alternative for large-scale seismic inversion and offers promising perspectives for subsurface exploration applications, including energy resource characterization, civil engineering, and environmental monitoring such as CO2 storage.