Streams draining urban heat islands tend to be hotter than rural and forested streams at baseflow because of warmer urban air and ground temperatures, paved surfaces, and decreased riparian canopy. Urban infrastructure efficiently routes runoff over hot impervious surfaces and through storm drains directly into streams and can lead to rapid, dramatic increases in temperature. Thermal regimes affect habitat quality and biogeochemical processes, and changes can be lethal if temperatures exceed upper tolerance limits of aquatic fauna. In summer 2009, we collected continuous (10-min interval) temperature data in 60 streams spanning a range of development intensity in the Piedmont of North Carolina, USA. The 5 most urbanized streams averaged 21.1°C at baseflow, compared to 19.5°C in the 5 most forested streams. Temperatures in urban streams rose as much as 4°C during a small regional storm, whereas the same storm led to extremely small to no changes in temperature in forested streams. Over a kilometer of stream length, baseflow temperature varied by as much as 10°C in an urban stream and as little as 2°C in a forested stream. We used structural equation modeling to explore how reach- and catchment-scale attributes interact to explain maximum temperatures and magnitudes of storm-flow temperature surges. The best predictive model of baseflow temperatures (R²  =  0.461) included moderately strong pathways directly (extent of development and road density) and indirectly, as mediated by reach-scale factors (canopy closure and stream width), from catchment-scale factors. The strongest influence on storm-flow temperature surges appeared to be % development in the catchment. Reach-scale factors, such as the extent of riparian forest and stream width, had little mitigating influence (R²  =  0.448). Stream temperature is an essential, but overlooked, aspect of the urban stream syndrome and is affected by reach-scale habitat variables, catchment-scale urbanization, and stream thermal regimes.