Xi'an City (the capital of Shaanxi province) has experienced a mushroom expansion in the past decade. Comprehensive exploitation of land resources has posed a dramatic threat to the urban climate and eco-environment responses, further intensifying the human-land conflict. The purpose of this study is to clarify the spatiotemporal patterns of land use and the driving forces in Xi'an City over the past 10 years, thereby promoting the optimal allocation of land resources and the sustainable development of the urban eco-environment. A classification feature dataset was also established to integrate the Landsat TM/OLI optical and PALSAR radar imagery using a new cloud-computing Google Earth Engine (GEE) platform, including the spectral, terrain, textural, and backscattering features. Among them, the remote sensing technology was utilized to acquire both optical and radar imagery for the multi-source data. The GEE platform combined with Random Forest (RF) was used to deal with the long-term series data, in order to improve the data acquisition, and processing efficiency, but to reduce the data volume, during the classification. Specifically, The RF was used to perform the land use classification in 2010, 2015, and 2019, further to determine the single land use dynamic degree and spatiotemporal patterns. A GeoDetector model was adopted finally to explore the driving factors affecting the spatiotemporal evolution patterns in the study area from natural and social aspects. The results indicate that: 1) A relatively high classification accuracy was achieved using the RF. Specifically, the overall classification accuracies were 92.30%, 86.66%, and 90.78% in the study area in 2010, 2015, and 2019, respectively, and the corresponding Kappa coefficients were 0.89, 0.81, and 0.88, respectively. 2) The main types of land use in the study area were the forest-grass and arable land, accounting for more than 85% of the whole area. The arable land decreased dramatically over the past ten years, particularly on the periphery of the central urban areas. Among them, the arable land decreased about 451.13 km², most of which was transferred to the construction land, leading to the rapid expansion of the construction land from 1 056.9 km² in 2010 to 1 529.01 km² in 2019. The prime expansion areas were in the central and northern parts of the city, where a great amount of arable land was substituted by the construction land. The forest-grass land presented a decreasing to increasing fluctuation, especially in the central south and eastern regions. Besides, the water body and unused land decreased gradually, but with very minor variations. 3) The GeoDetector analysis revealed that the natural factors were the fundamental controlling factors in the land use pattern in the study area, including the terrain, temperature, and precipitation. Furthermore, the economic activities were also the important driving factors with increasing the explanatory power, whereas, the explanatory power of the population increased at the early stage and decreased subsequently. Correspondingly, the overall land-use patterns in the study area were dominated by the interaction of terrain, population, and economic factors. In conclusion, the integrated RF classification and GeoDetector model using the multi-source data can provide an effective way to better understand spatiotemporal land use and the driving forces. These findings can widely be expected to serve as the scientific fundamentals for the decision-making on the planning and management of urban land resources. [ABSTRACT FROM AUTHOR]