Abstract: To address the large storage overhead, heavy reading burden, and low data-access efficiency of large-scale 3D traveltime tables in microseismic localization, we introduce the error-bounded predictive compression (squeeze, SZ) algorithm to regular-grid traveltime tables generated using the fast marching method, and establish an integrated workflow of "traveltime computation, compressed storage, decompression, and migration localization", which embeds the compressed traveltime tables into a migration stacking localization program with joint source-mechanism inversion. Using a GPU-accelerated migration stacking localization framework, we analyze the influences of compressed traveltime tables on event detection, imaging focusing, and overall localization efficiency at both the data-access and localization-solving layers. Tests on synthetic microseismic data from the 3D overthrust model and field monitoring data show that the SZ algorithm significantly reduces the size of traveltime tables under controlled error conditions. Moreover, the compressed traveltime tables remain highly consistent with the original ones in terms of event-detection curves, imaging-peak positions, and spatial localization results. These results demonstrate that the error-bounded predictive compression method significantly improves the overall processing efficiency of large-scale microseismic migration stacking localization while preserving localization accuracy.