Understanding urban state transitions is critical for sustainability, yet a unified framework to characterize the interactive dynamics of key urban components—green vegetation (GV), impervious surface (ISA), and bare soil (BS)—has been lacking at a national scale. To bridge this gap, we developed a novel framework that moves beyond static mapping to directly quantify the dynamic interactions within the V-I-S nexus. Leveraging over 200,000 Landsat images on Google Earth Engine, we generated the first continuous, annual V-I-S fraction maps for China (1990-2019) using a unified spectral endmember space with change trajectory model to characterize and quantify multi directional urban state transitions. Through our pixel-level trajectory analysis, where we integrated the magnitude, timing, and trend of V-I-S changes into a set of conversion rules to systematically identify five transition archetypes: stable, expansion, intensification, green recovery, and soil exposure. This framework demonstrated superior capability in tracking the complex intra-urban landscape components with high accuracy (i.e., smaller mean squared error of -2.07% to 3.44%) and low uncertainties (i.e., fractional biases less than 0.10), as well as achieving highly accurate historical state transitions mapping (87.29%) at fine-scale feature resolution. Our analysis reveals a national-scale GV loss (-31,109 km²) alongside ISA (+28,294 km²) and BS (+2,814 km²) gains, primarily driven by expansion (46%) and intensification (28%). Crucially, we uncover a pivotal shift from expansive growth to intensification and greening as cities develop. This study provides the first spatially explicit, nationally consistent record of intra-urban dynamics, offering a powerful tool for modeling urban social-ecological systems and guiding targeted governance for sustainable urban futures.