AbstractThe Earth's mantle transition zone has significant control on material flux between upper and lower mantle, thus constraining its properties is imperative to understand dynamic processes and circulation patterns. Global seismic data sets to study the transition zone typically display highly uneven spatial distribution. Therefore, complementary geometries are essential to improve knowledge of physical structures, thermochemistry, and impact on convection. Here, we present a new automated approach utilizing machine learning to analyze large seismic data sets, and derive high‐resolution maps of transition zone discontinuity properties. Seismic measurements from ScSScS precursors are integrated with mineralogical modeling to constrain thermochemistry of the western Pacific subduction zone. Our models map recent subduction patterns through the transition zone, indicating stagnation of slabs and accumulation of basalt at its base, and interaction between stagnant slabs and plumes. These results suggest that the thermochemical properties of upper mantle discontinuities can provide high‐resolution images of mantle circulation patterns.