Despite decades of research into air and stream temperature dynamics, paired air-water annual temperature signals have been underutilized to characterize watershed processes. Annual stream temperature dynamics are useful in classifying fundamental thermal regimes and can enhance process-based interpretation of stream temperature controls, including deep and shallow groundwater discharge, when paired with air signals. In this study, we investigated multi-scale variability in annual paired air-water temperature patterns using sine-wave linear regressions of multi-year daily temperature data from streams of various sizes. A total of 311 sites from two spatially intensive regional datasets (Shenandoah National Park and Olympic Experimental State Forest) and a spatially extensive national dataset spanning the contiguous United States (U.S. Geological Survey gages) were evaluated. We calculated three annual air-water thermal metrics (mean ratio, phase lag, and amplitude ratio) to deduce the influence of groundwater and other watershed processes on stream thermal regimes at multiple spatial scales. Site-specific values of the three annual air-water thermal metrics ranged from 0.69 to 5.29 (mean ratio), −9 to 40 days (phase lag), and 0.29 to 1.12 (amplitude ratio). Regional patterns in the annual thermal metrics revealed persistent yet spatially variable influences of shallow groundwater discharge and high levels of thermal variability within watersheds, indicating the importance of local hydrogeological controls on stream temperature. Furthermore, annual thermal metric patterns from the regional datasets were generally concordant with the national dataset suggesting the utility of these annual thermal metrics for analysis at multiple scales. Analysis of the national dataset showed that previously defined thermal regimes based on water temperature alone could be further refined using air-water metrics and these metrics were related to physiographic watershed characteristics such as contributing area, elevation, and slope. This research demonstrates the importance of spatial scale and heterogeneity for inferring hydrological process in streams and provides guidance for the interpretation of annual air-water temperature metrics that can be efficiently applied to the growing database of multi-year temperature records. Results from this research can aid in the prediction of future thermal habitat suitability for coldwater-adapted species at ecologically and management-relevant spatial scales with readily available data.