Much of our understanding on temporary headwater streams is from arid and sub-humid environments. We know less about no-flow periods in humid headwater catchments that experience seasonal snow cover. Our study aimed to 1) characterize the patterns of no-flow periods for forested headwater streams in a snow-dominated landscape, and 2) compare the ability of statistical and process-based models to predict the occurrence of no-flows for these streams under shifting climatic conditions. We used 40 years of streamflow data from 13 headwater catchments within the Turkey Lakes Watershed located in Ontario, Canada, near the eastern shores of Lake Superior. These headwater catchments differ substantially in their number of no-flow days (0 to 166 days per year) despite being clustered in a small geographical area with uniformly wet climatic conditions (mean annual precipitation of 1210 mm/year). We found that between-catchment variability in no-flow occurrences was associated with differences in wetland cover and water travel times. Inter-annual variability in the number of no-flow days was primarily associated with summer and fall precipitation, as well as evapotranspiration rates. Despite the large seasonal snowpacks that form in this region, snow conditions did not appear to influence the timing or extent of no-flow periods. We found that a process-based model provided better predictions of no-flow periods compared to predictions made using statistical models. The advantage of using process-based models was particularly noticeable when making predictions outside the climatic conditions used to calibrate the model (e.g., calibration during wet and cold years and prediction during dry and hot years). Our study provides insights on no-flow periods for headwater streams in northern forested landscapes and how they may respond to changes in climate.