Dryland ecosystems with alternative stable states may undergo a regime shift from a relatively healthy state to a degraded state under the increasing climate change and disturbances from human activities such as land reclamation and grazing, resulting in the decline of ecosystem functions. The identification of early warning signals is a hot topic in the study of ecosystem regime shifts and a key aspect of management practices to prevent ecosystem degradation. Previous studies on early warning signals have focused on generic signals such as autocorrelation, variance and other statistical indicators, however, these indicators may not be applicable to dryland ecosystems with specific mechanisms. The spatial indicators developed based on landscape pattern characteristics provide a uniquely spatial perspective for ecosystem regime shift identification, which has scientific significance and practical value for understanding ecosystem degradation processes and mechanisms in drylands. In this review, the phenomenon of regime shifts in dryland ecosystems and its characteristic mechanisms are first introduced. Then, focusing on the indicators and methods of landscape ecology, the key early warning indicators (vegetation coverage, vegetation patch morphology, vegetation patch size and frequency distribution and hydrological connectivity, etc.) based on the characteristics of landscape pattern in drylands are summarized from a spatial perspective, and the concepts, quantitative methods, identification characteristics and practical applications of these key indicators are analyzed. Finally, based on the advantages and limitations of indicators, the review points out the future direction of the research, including exploring potential landscape indicators, strengthening research on the interaction mechanisms of multiple driving factors of ecosystem change in drylands, carrying out empirical research at multiple spatial and temporal scales, constructing an overall analytical framework for early warning signals of ecosystem regime shifts, and strengthening the quantitative research on indicator thresholds.