Multi-temporal landslide inventory (2014–2022) associated with the Ms 6.5 Ludian earthquake in Yunnan, China

Large earthquakes in mountainous regions often trigger widespread landslides that can persist over years or decades. Despite this, current post-seismic landslide inventories are often temporally limited and spatially constrained. To address this gap, we present a multi-temporal landslide inventory spanning from 2014 to 2022, associated with the 2014 Ms 6.5 Ludian earthquake in Yunnan, China. This inventory integrates high-resolution satellite imagery with deep learning–based workflow and manual validation to identify landslides over 11 temporal phases, including one pre-seismic, one co-seismic, and nine post-seismic periods. Landslides were mapped as both point and polygon features, with attribute data including area, perimeter and estimated volume. Accuracy assessment based on confusion matrices indicates an overall mapping accuracy of 87.2% and a Kappa coefficient of 0.81. The final dataset comprises ∼11,000 individual landslides and is provided in ESRI Shapefile format with full spatial metadata. This publicly accessible dataset provides a robust foundation for the spatiotemporal evolution of landslides, supporting future research on landslide recovery dynamics, post-seismic hazard modelling, and geomorphic disturbance assessment in strike-slip fault zones.