Fuding white tea, particularly premium silver needle, is a high-value Geographical Indication product increasingly vulnerable to cross-regional mislabeling, substitution, and blending. This study developed a dual-modal ultraviolet-visible diffuse reflectance and near-infrared spectroscopic strategy combined with chemometrics for hierarchical authentication. At the macro-geographical level, optimized UV-Vis and NIR PLS-DA models achieved accuracies of 100% and 97.54%, respectively, and correctly classified external Fujian and Yunnan samples. At the town scale within Fuding city, UV-Vis outperformed NIR in predictive ability (Q2 = 0.70 vs. 0.37), indicating greater sensitivity to subtle local origin variation. Earlier-year Panxi samples were also correctly assigned, suggesting preliminary temporal transferability for town-scale authentication. Colorimetric analysis showed limited visual separability between authentic Fuding and non-Fuding silver needle samples, whereas spectroscopic models correctly identified all blended samples. Variable importance in projection analysis revealed complementary contributions from UV-Vis chromophore-related features and NIR matrix-related information, while low-level UV-Vis-NIR fusion further improved blending detection.