3.3 Observed changes in WWEs
Our new method of assessing WWEs begins with identification of long-duration events (LDEs) in reanalysis fields. We find that the most common patterns (Fig. 1 ) also tend to exhibit a higher occurrence of LDEs per year (Fig. S6 ), which is expected given that a larger number of days in one pattern increases the chances of three or more consecutive days occurring. Significant (>95% confidence) increasing trends in LDEs exist in node #1, while nodes #10 and #12 exhibit declining trends (while the f-test detected significant (>90%) trends in nodes #2 and #6, the number of LDEs is too small to be consequential). These findings are consistent with our earlier work on weather-regime persistence (Francis et al., 2018; Vihma et al., 2019; Francis et al., 2020). Specifically, Francis et al., 2018 found an increased occurrence of LDEs particularly in patterns featuring positive height anomalies in high latitudes of the North American domain, the same as that used in the present study. Increased LDEs in warm-Arctic nodes are consistent with the counterintuitive downward trend in extreme summer heat over much of the upper Midwest during recent decades, and the finding suggests that amplified Arctic warming favors more persistent weather conditions that can lead to extreme weather events.
The matrix of time series shown in Fig. 6a displays the frequency per year with which a WWE originates from an LDE in each node. The infrequency of WWEs (13.2 per year on annual average) and large interannual variability challenge the detection of frequency changes over the observational record, but a few nodes exhibit statistically significant trends. Nodes #1 and #12 initiate substantially more WWEs than do other nodes, and significant trends are evident for node #3 (increasing) and node #12 (decreasing). These trends are attributable mainly to the changes in frequency of these nodes (Fig. 3 ), in that LDEs are more likely to occur when a node becomes more prevalent, and vice versa.
Winter: Changes over time are more evident when assessed as differences in frequency between the two 20-year periods at the beginning and end of the observation record. During winter months (Fig. 6b ), only node #9 exhibits a significant increase in WWEs, while those initiated in node #12 have declined. The pattern in node #9 features a strong meridional ridge/trough pattern across the continent, which according to Fig. 2 , is most likely to shift to node #1 or #12 after a WWE (note node #10 is adjacent, thus does not qualify as a WWE), both of which are more zonal patterns. This transition would suggest that persistent warm, dry weather in the western states and cool conditions across central North America would likely transition to a more progressive synoptic regime in a WWE. The significant decrease in WWEs originating in node #12, in contrast, would suggest fewer abrupt transitions from broad western troughing/eastern ridging to patterns similar to either node #4 or #10, the two most likely nodes following an LDE in node #12. A WWE originating in the strong zonal jet conditions depicted by the pattern in node #12, featuring cold (warm) high (middle) latitudes, would most likely transition along one of two paths: 1) the cold migrates westward over the northeast N. Pacific and Alaska while warmth invades much of North America (node #4), or 2) the cold migrates eastward over northeastern North America, while warmth bulges northward into Alaska (node #10).
Summer: Our results suggest that WWEs originating in node #3 during JAS have increased (Fig 6c ). An LDE in node #3 is most likely to shift to node #1 in a WWE (Fig. S8 ), which would replace anomalous troughing (ridging) in the northwestern (northeastern) parts of the continent with opposite height anomalies. In terms of sensible weather, the increased frequency in WWEs from node #3 to node #1 shift anomalously high temperatures in the Hudson Bay region to Alaska and British Columbia, accompanied by increased negative anomalies throughout most of the western interior. Anomalously wet conditions shift from southeast Alaska and the British Columbia coast to southern California (Fig. 5 ). Note that a transition from node #3 to #1 in summer would be consistent with the reported increased precipitation extremes in western North America (Swain et al., 2018). A decrease in WWEs originating in nodes #4 and #10 has also occurred. An LDE in node #4 is most likely to shift to node #6 or #12 during a WWE. The shift in temperatures and precipitation from node #4 to #6 would reduce the likelihood of hot days over central Canada, reduce the likelihood of abnormally cool days over Alaska, and decrease the chances of heavy precipitation in western Canada while increasing chances in southeastern North America. A WWE that started in node #4 and shifted to #12 would also decrease hot days in central Canada but would also increase cold days over the northwest part of the domain, while heavy precipitation would increase from the Gulf of Mexico to New England A WWE that was initiated in node #12 is most likely to shift to node #10 or # 4. The most notable changes in extremes during a transition to #10 will be an increase in high-temperature extremes in the Gulf of Alaska, more widespread cool spells over the Rocky Mountains and High Plains, and a transition from anomalously wet to dry conditions along the coast of British Columbia. A shift from node #12 to #4 is likely to bring abnormally warm conditions to central Canada, a lower probability of anomalously cold days in the Arctic, and more focused heavy precipitation in the Pacific Northwest. The decreased frequency of WWEs originating in node #12 will mean fewer of these types of abrupt transitions.