Owing to increasing population densities and impervious surface areas, heat island effects increasingly dominate urban environments and hinder sustainable development. The urban spatial form plays an important role in mitigating urban heat islands. Taking Ganjingzi District, Dalian, as an example, this study considered urban spatial form at the community scale using spatial autocorrelation and spatial regression methods to explore 2003–2018 spatial and temporal differentiation characteristics and driving factors of Land Surface Temperature (LST). The LST of each community showed a gradually increasing trend; high values (>30°C) were concentrated in central and eastern areas; low values were (<25°C) was concentrated in the south and west. LSTs were influenced by spatial variables (e.g., land use); however, building form was only weakly related to LST. The global autocorrelation Moran’s I value for LST exceeded 0.7, indicating strong positive correlation in terms of spatial distribution. H-H and L-L LISA values were distributed in central and southern areas, respectively. The spatial error model (SEM) was a better fit than the spatial lag (SLM) or ordinary least squares models (OLS) and was used to explore these relationships. This study focuses on community surface temperature and hopes to provide a valuable reference for community planning, resource allocation and sustainable development.