Large-scale quantification of ecosystem stability in multiple dimensions is crucial to understanding underlying ecological processes and informing ecological management decision-making. However, historically this field has been limited to spatiotemporal scales based on the use of discrete ground-based measurements. The objective of this study was to quantify ecosystem stability in multiple dimensions in a karst peak cluster depression region in southwest China, using dense Landsat time series from 1988 to 2018. Three components of ecosystem stability, namely resistance, resilience, and variability, were derived by applying a time decomposition algorithm, and then the correlations of each component were analyzed to explore the dimensionality of ecosystem stability. Our results revealed following: (1) of the entire karst area, 87.97% of the pixels were disturbed in the past 31 years, the majority of the maximum disturbance events occurred during 2004 in the northeast and mid-west of Guangxi. (2) Only 0.03% of the pixels showed high resistance, whereas a wide part of the study area showed low resilience, with 72.95 % of the pixels had low recovery rate and 39.27% of the pixels being able to be restored to their original state after the disturbance. The south area showed lower variance compared to other areas in karst regions. (3) Correlations between different ecosystem stability indicators obtained from dense remote sensing time series were only weakly correlated or uncorrelated, which provided satellite‐scale evidence that it is necessary to conduct a multi-dimensional evaluation of ecosystem stability, and the effective dimensionalities of ecosystem stability were significantly influenced by different disturbance intensities. These findings expand the understanding of the internal self-maintenance and self-recovery of the ecosystem in response to disturbances, and provide a theoretical basis for ecological engineering construction and regional environmental governance assessment.