The confluence of factors driving urban growth is highly complex, resulting from a combination of ecological and social determinants that co-evolve over time and space. Identifying these factors and quantifying their impact necessitates models that capture both why urbanization happens as well as where and when it happens. Using a database that links five satellite images spanning 1976-2001 to a suite of socioeconomic, ecological and GIS created explanatory variables, this study develops a spatial-temporal model of the determinants of built-up area across a 25,900 square kilometer swath across central North Carolina. Extensive conversion of forest and agricultural land over the last decades is modeled using the complementary log-log derivation of the proportional hazards model, thereby affording a means for modeling continuous-time landscape change using discrete-time satellite data. To control for unobserved heterogeneity, the model specification includes an error component that is Gamma distributed. Results confirm the hypothesis that the landscape pattern surrounding a pixel has a major influence on the likelihood of its conversion and, moreover, that the omission of external spatial effects can lead to biased inferences regarding the influence of other covariates, such as proximity to road. Cartographic and nonparametric validation exercises illustrate the utility of the model for policy simulation.