The Sikkim–Bhutan seismic gap has witnessed a long earthquake quiescence since the 1714 M7.5~8.5 earthquake. The state of stress accumulation beneath the Sikkim–Bhutan Himalaya and its spatial correlation with seismicity remains unclear due to the lack of geodetic measurements and the low levels of seismic activity. We compile Global Positioning System (GPS) measurements in southern Tibet with the available velocities in the Sikkim–Bhutan Himalaya to reveal the characteristics of strain buildup on the Main Himalayan Thrust (MHT). We correct non-tectonic hydrological loading effects in a GPS time series to accurately determine the Three-Dimensional (3D) velocities of each continuous station. Extensive GPS measurements yield convergence rates of 16.2~18.5 mm/y across the Sikkim–Bhutan Himalaya, which is quite consistent with that observed elsewhere in the Himalaya. Based on a double-ramp structure of the MHT, a refined 3D coupling image is inverted using a dense network of GPS velocities. The result indicates significant along-strike variations of fault coupling beneath the Sikkim–Bhutan Himalaya. The locking width (coupling > 0.5) of western Bhutan reaches ~100 km, which is 30~40% wider than Sikkim and eastern Bhutan. An obvious embayment of decoupling zone near the border between Sikkim and western Bhutan is recognized, and coincides spatially with the rupture terminates of the 1934 Mw8.2 and the 1714 M7.5~8.5 earthquakes, indicating that the large megathrust earthquakes along the Sikkim–Bhutan Himalaya are largely segmented by the spatial variation of frictional properties on the MHT. Using a new compilation of seismic records in the Sikkim–Bhutan Himalaya, we analyze the spatial correlation between fault coupling and seismic activity. The result suggests that the seismicity in the Bhutan Himalaya is broadly distributed, instead of restricted in the lower edge of the interseismic locking zone. This implies that the seismic activity in the Bhutan Himalaya is not uniquely controlled by the stress accumulation at the downdip end of the locked portion of the MHT.

In recent years, geographic knowledge graphs (GeoKGs) have shown great promise in representing spatio-temporal and event-driven knowledge. However, existing knowledge graph embedding approaches mainly focus on structural patterns and often overlook the dynamic evolution of entities in both time and space, which limits their effectiveness in downstream reasoning tasks. To address this, we propose a spatio-temporal evolutionary knowledge embedding approach (ST-EKA) that enhances entity representations by modeling their evolution through type-aware encoding, temporal and spatial decay mechanisms, and context aggregation. ST-EKA integrates four core components, including an entity encoder constrained by relational type consistency, a temporal encoder capable of handling both time points and intervals through unified sampling and feedforward encoding, a multi-scale spatial encoder that combines geometric coordinates with semantic attributes, and an evolutionary knowledge encoder that employs attention-based spatio-temporal weighting to capture contextual dynamics. We evaluate ST-EKA on three representative GeoKG datasets—GDELT, ICEWS, and HAD. The results demonstrate that ST-EKA achieves an average improvement of 6.5774% in AUC and 5.0992% in APR on representation learning tasks. In question answering tasks, it yields a maximum average increase of 1.7907% in AUC and 0.5843% in APR. Notably, it exhibits superior performance in chain queries and complex spatio-temporal reasoning, validating its strong robustness, good interpretability, and practical application value.