Landscape ecological risk assessment (LERA) is an effective tool developed in recent years to support landscape ecological construction and management. The ecological risk caused by landscape change or land-use and land-cover change has received increased attention by policy-makers. High-intensity human activities disturbed the pattern and function of ecosystem and the hotspot of landscape change in urban areas, which contributed to the uncertainty of ecological risk. Therefore, it is urgent to improve the LERA method to provide quantitative analysis for landscape management and comprehensive risk prevention. There are several methods used for LERA such as landscape metrics and landscape source-sink risk. However, the classic model of traditional ecological risk assessment followed the mode of risk source identification-receptor analysis-exposure and hazard assessment. In order to improve the LERA method in urban areas, this research followed the landscape metrics method, but improved the algorithm of vulnerability for landscape services. A series of landscape services indicators were reselected to denote landscape vulnerability, including carbon storage, water provision and purification, soil conservation, and habitat quality. This study used multi-source data to analyze the spatio-temporal changes and characteristics of landscape ecological risk based on 1 km×1 km spatial resolution in Guangzhou, a typical area experiencing rapid urbanization, for the period 1990-2015. The results were as follows: (1) LERA based on landscape pattern and landscape service can effectively evaluate the spatial distribution of urban landscape ecological risk. The low landscape ecological risk type is typical and these areas were regularly distributed over the 25 years. In the study stage, the year of 2000 was the transition period. (2) The landscape disturbance degree of Guangzhou increased first and then decreased over the 1990-2015 period; 2000 was the year when the degree of landscape disturbance changed. Landscape vulnerability is increasing. Together, medium vulnerability and high vulnerability areas accounted for more than 60% of the area; these areas were more common in the south than the north, and more common in the west than in the east. The landscape ecological risk in Guangzhou was increasing before 2000, and showed a tendency to decrease after 2000. (3)The landscape ecological risks transition showed that landscape changes and ecological risk have a good spatial consistency in Guangzhou during the 25-year study period. The area with a low landscape risk was 40.74%, and this area was mostly distributed in the northern mountainous areas. The area with a high landscape risk was 6.67%, the transition from low risk to high risk was 32.28%, and high risk change to low risk was 20.31%. This shows that landscape changes in urban area has a high risk after rapid urbanization. However, the uncertainty of LERA still plagued the application of LERA in policy-making. Further research can pay more attention to the verification of assessment results or use scenario analysis to quantify spatiotemporal characteristics of risk to improve the scientific basis for the early warning and prediction of landscape ecological risk. Then, the LERA applications in urban areas can be acceptable for landscape management or policy-making.