Continuous measurements of total ozone (by Dobson spectrophotometers) across the contiguous United States (US) began in the early 1960s. Here, we analyze temporal and spatial variability and trends in total ozone from the five US sites with long-term records. While similar long-term ozone changes are detected at all five sites, we find differences in the patterns of ozone variability on shorter time scales. In addition to standard evaluation techniques, STL-decomposition methods (Seasonal Trend decomposition of time series based on LOcally wEighted Scatterplot Smoothing, LOESS) are used to address temporal variability and trends in the Dobson data. The LOESS-smoothed trend components show a decline of total ozone between the 1970s and 2000s and a "stabilization" at lower levels in recent years, which is also confirmed by linear trend analysis. Methods from statistical extreme value theory (EVT) are used to characterize days with high and low total ozone (termed EHOs and ELOs, respectively) at each station and to analyze temporal changes in the frequency of ozone extremes and their relationship to dynamical features such as the North Atlantic Oscillation and El Niño Southern Oscillation. A comparison of the "fingerprints" detected in the frequency distribution of the extremes with those for standard metrics (i.e., the mean) shows that more "fingerprints" are found for the extremes, particularly for the positive phase of the NAO, at all five US monitoring sites. Results from the STL-decomposition support the findings of the EVT analysis. Finally, we analyze the relative influence of low and high ozone events on seasonal mean column ozone at each station. The results show that the influence of ELOs and EHOs on seasonal mean column ozone can be as much as ±5%, or about twice as large as the overall long-term decadal ozone trends.