Weather Patterns Driving Atmospheric Rivers, Santa Ana Winds, Floods, and Wildres During California Winters Provide Evidence for Increasing Fire Risk

Floods caused by atmospheric rivers and wildres fanned by Santa Ana winds are common occurrences in California with devastating societal impacts. Planning for these types of events is critical to protect life and property, and extending the lead-times of predictability for these types of events improves emergency response. A better understanding of linkages between large-scale atmospheric circulation patterns and extreme weather represents an important step towards improving predictability and preparedness on subseasonal-to-seasonal (S2S) timescales and for climate change adaptation. In this work, we show that winter weather variability in California, including the occurrence of extreme and impactful events, is linked to four atmospheric circulation regimes over the North Pacic Ocean previously named and identied as the “NP4 modes”. These modes interact on daily timescales to produce recurring winter weather patterns that are major drivers of atmospheric river landfalls, Santa Ana winds, oods, and wildres. Many recent California natural hazard events resulted from compounding and cascading extremes including frequent atmospheric river landfalls (wet, fuel-producing years) or lack thereof (drought) and followed by Sana Ana-driven wildres that render the landscape susceptible to hydrologic hazards posed by short-duration high-intensity precipitation events. This historical perspective of atmospheric circulation and impacts over 70 years reveals that weather patterns are changing in a way that enhances wildre risk in California, while the frequency of weather patterns linked to historical oods is not diminishing. These changes highlight the rising hazards of compounding weather extremes in California’s present and future.


Introduction
Extreme winter weather variability in California is largely driven by the interplay between atmospheric rivers and downslope winds (Santa Ana winds (SAWs) in Southern California or Diablo winds in Northern California). Atmospheric rivers (ARs) are narrow laments of concentrated atmospheric water vapor that can bring extreme precipitation during fall and winter. While critical for California's water supply, ARs are also the major cause of historical oods (Ralph et al. 2006Corringham et al. 2019). Downslope SAWs and their northern counterpart, the Diablo winds, are characteristically hot, dry, offshore winds that bring re weather to coastal and inland California (Abatzoglou et al. 2021). Typically, the SAW season begins in fall and peaks in December (Guzman-Morales 2016) whereas Diablo winds peak during October . The timing of offshore winds relative to ARs is an important factor for hazardous wild res in California, where the risk of wild re is enhanced if a strong SAW or Diablo occurs before the rst seasonal rain when fuel moisture is climatologically low. ARs and downslope winds are driven by certain recurring weather patterns over and offshore of western North America that favor onshore or offshore ow over California, respectively (e.g. Ralph  Previous studies have demonstrated the importance of certain atmospheric circulation patterns in driving AR landfalls over California and western North America. AR landfalls are often associated with a ridge situated over/near Alaska (~ 160°W) with a trough to the southeast and a strong pressure gradient that channels marine moisture toward the coast (e.g. Weaver 1962; Strobin and Reynolds 1995; Ralph et al. 2004; Neiman et al 2008; Guirguis et al. 2018Guirguis et al. , 2019Guirguis et al. , 2020. The position, strength, and orientation of these circulation features relative to coastal topography modulate the landfall location as well as the amount and spatial distribution of moisture transport and precipitation (Neiman et (Gershunov et al. 2021). These hot, dry, and windy conditions present substantial re risk to California when vegetation is dry. Harmful health impacts associated with SAW-driven heat waves (Schwartz et al. 2020) and wild re smoke (Aguilera et al. 2020(Aguilera et al. , 2021a have also been reported. While all SAWs warm adiabatically, they do not always produce heat waves at the coast. Hot and cold SAWs result from distinctly different synoptic dynamics and Great Basin land surface conditions (Gershunov et al. 2021). All SAWs involve a surface high pressure over the northeastern Great Basin, which develops in association with a blocking high aloft in the case of hot SAWs, and with Rossby wave breaking in the wake of transient cold frontal cyclones in the case of cold SAWs; the latter are also associated with anomalous Great Basin snow cover (Gershunov et al. 2021). Southern California wild res are preferentially spread by hot SAWs (Gershunov et al. 2021).
Typically, the interplay between ARs, SAWs, extreme precipitation, and wild res is thought of as a situation where one extreme (ARs) is the remedy or prevention of the other extreme (wild res). However, when ARs occur following wild res, they can cause deadly debris ows over the burn area if shortduration, high intensity rainfall occurs during cold frontal passage (Oakley et al. 2018), compounding the original impacts. Here, we show that ARs and SAWs are more closely linked than previously discovered through interacting synoptic-scale atmospheric circulation regimes. Speci cally, we demonstrate how four key circulation regimes over the North Paci c Ocean ("North Paci c Modes" or "NP4 modes"), identi ed and described in previous work (Guirguis et al. 2020, hereinafter GGR'20), interact on daily timescales to produce recurring, regionally-important weather patterns responsible for impactful weather in California and along the West Coast. These four circulation regimes (shown in Fig. 1a) are known as the Baja-Paci c (BP), Alaskan-Paci c (AP) Canadian-Paci c (CP), and Offshore-California (OC) modes. The NP4 modes share characteristics with established, well-known teleconnections identi ed in previous work. For example, the Alaskan-Paci c and Canadian-Paci c modes are related to the well-known West Paci c Oscillation (WPO) and Paci c North American (PNA) Pattern, respectively (GGR'20), which were

Atmospheric Variables
Daily 500 mb geopotential height (Z500) and wind elds for the period spanning Nov-Feb 1949-2021 are from NCEP/NCAR 2.5° global reanalysis (Kalnay et al. 1996). Daily anomalies were calculated relative to the annual/semiannual seasonal cycle t using least squares regression.

Atmospheric River Detection and AR Intensity Scale
The SIO-R1 catalog of Gershunov et al. (2017) is used to identify landfalling ARs along the coast of western North America. The detection method uses threshold criteria for integrated water vapor (IWV > 15 mm) and integrated vapor transport (IVT > 250 kg m -1 s -1 ) to identify elongated objects of high moisture content that intersect the coast. The AR scale of Ralph et al. (2019) is used to categorize the magnitude of storms. This 5-scale classi cation is based on IVT magnitude and duration where the scale 1-5 describes ARs as weak, moderate, strong, extreme, and exceptional, respectively. AR scales 1-2 are mostly bene cial providing needed precipitation without damage, whereas AR scales 4-5 are mostly hazardous with peak IVT in the range of 750 to more than 1250 kg m -1 s -1 . Flood damages in the western U.S. increase exponentially with the AR scale (Corringham et al. 2019). The SIO-RI AR catalog is available at https://weclima.ucsd.edu/data-products/.

Santa Ana Winds
Historical SAW events are identi ed using the daily Santa Ana Winds Regional Index (SAWRI) of Guzman-Morales (2016, 2019). This data product is based on 65 years of hourly data from the dynamically downscaled CaRD10 dataset (Kanamitsu and Kanamaru, 2007) and validated against station observations. SAWs are de ned by strong, sustained, northerly to easterly winds (Guzman-Morales et al., 2016). This study uses the updated daily version of the Guzman-Morales and Gershunov (2019) SAW index, regionalized over Southern California (~ 32-35 °N, 116-118 °W) and available at https://weclima.ucsd.edu/data-products/. We de ne moderate (extreme) SAWs as those exceeding the 50th (90th ) percentile.

Historical Temperature and Precipitation
Daily precipitation and maximum temperature (Tmax) data are from Livneh et al. (2013Livneh et al. ( , 2015, which is a gridded data product available at a 1/16° spatial resolution over the period 1950-2015 (available at https://climatedataguide.ucar.edu/climate-data). The source data are from approximately 20,000 stations in the Global Historical Climatology Network (GHCN). Extreme precipitation is de ned as daily accumulation above the 95th percentile. Heat waves (cold extremes) are de ned following Gershunov and Guirguis (2012) as daily temperatures persisting for at least one day above the 95th (below the 5th ) percentile after removing the seasonal cycle.

Sierra Nevada Snowpack
We use monthly snow water equivalent data from a single snow course in the Northern Sierra Nevada.
Speci cally, we selected Donner Summit (39.3 °N, -120.3 °W, elevation 2100 meters, station ID 20K10) from the Natural Resources Conservation Service (https://wcc.sc.egov.usda.gov/nwcc/rgrpt? report=snowcourse&state=CA). This location sits at the intersection of three key headwater regions (Yuba, Truckee, and American Rivers) involved in managing California and Nevada water resources. The Donner Summit snow course has a long (1913-2021) and complete record.

Wild re
Wild re information is from the California Fire Protection's Fire and Resource Assessment Program (FRAP) re perimeters database (https://frap. re.ca.gov/frap-projects/ re-perimeters/). We focus on re start dates for historical wild res with burn sizes exceeding 1000 acres in the Southern California (SoCal) counties of Santa Barbara, Ventura, Los Angeles, Orange, Riverside, San Bernardino, and San Diego. Our sample includes 76 wild res spanning 1949-2018, with most occurring in November (n = 40) or December (n = 21). It is worth noting that most wild res in California occur in October, which is outside of our extended winter focus season. However, the wild re season can and does extend into winter during drought years or when delayed rains extend dry vegetation into the peak of SAW season.

North Paci c Atmospheric Circulation Regimes
We use daily amplitudes of four North Paci c atmospheric circulation regimes (NP4 modes), which were identi ed and shown to be essential drivers of vapor transport and AR activity along the coast of western North America by Guirguis et al. (2018, hereinafter GGR'18) and GGR'20. These four modes are named according to their geographic centers of action: Baja-Paci c (BP), Alaskan-Paci c (AP), Canadian-Paci c (CP), and Offshore-California (OC) modes. The NP4 modes were identi ed using rotated Empirical Orthogonal Function (EOF) analysis applied to Z500 anomaly elds over a large domain spanning the northern Paci c Ocean and western North America. While the NP4 modes are not the most important in terms of eld variance (i.e., they correspond to higher-order rotated principal components), they were found to be the most important modes for coastal vapor transport, AR landfalls, and extreme precipitation in California (GGR'18, GGR '20). The EOF loading patterns (maps) associated with the positive phase of each mode are shown in Fig. 1a. The portion of the domain where these modes are most in uential is shown in Figs. 1b and 1c. Speci cally, Fig. 1b shows the locations where one of the four modes is by itself responsible for more than 50% of the Z500 variance. Figure 1c shows the collective in uence of the four modes, where we see that together these modes explain 25-90% variance in the portion of the domain most relevant for west coast weather. The NP4 dataset used in this study is described in Guirguis et al. (2020b) and is available at https://doi.org/10.6075/J0154FJJ.

Weather regime Classi cation
Daily weather regimes are classi ed based on the joint phase combination of the NP4 modes. As the NP4 modes oscillate over the North Paci c, their daily interactions determine the position and strength of ridges and troughs along the coast of western North America. Unique NP4 phase combinations result in distinct weather patterns that drive extreme weather in California (GGR'18 GGR '20). On a given day, each of the four NP4 modes can be positive or negative, which yields 16 possible phase combinations, and 16 corresponding weather regimes (introduced and described in Sect. 3).
We use the terminology "weather pattern" and "weather regime" somewhat interchangeably. "Weather pattern" is taken to mean the Z500 anomaly eld itself while "weather regime" refers to the classi cation of that pattern. A catalog of historical weather regimes observed over 1949-2021 is developed as an accompaniment to this publication and is available at https://weclima.ucsd.edu/data-products. 3. Winter Weather Patterns Impacting The West Coast Figure 2 shows Z500 anomaly patterns associated with the sixteen different phase combinations of the NP4 modes, which are used to de ne winter weather regimes (WR) impacting the West Coast. We rst provide examples of how the NP4 modes interact to create distinct weather patterns (Sect. 3.1) and then provide an overview of circulation characteristics and impacts associated with each weather regime (Sect. 3.2).

How the NP4 modes interact to produce distinct weather patterns
Examples of how the NP4 modes interact to produce distinct weather patterns are provided in Fig. 3 and  . We highlight three events (December 7-16, January 3-11, and February 1-8) when the NP4 modes exhibited the same joint phase con guration (labeled E1, E2, and E3; Fig. 3). Speci cally, during these three events the BP mode was positive (which favors a ridge over Baja California, c.f. Figure   1a), the AP mode was negative (favoring high pressure over Alaska), and the OC and CP modes were negative (favoring low pressure over western Canada and offshore from California, respectively). This mode con guration results in the atmospheric circulation and IVT patterns shown in Figs. 3c and 3d, respectively. This type of pattern (classi ed as WR10, see Fig. 2) is associated with strong onshore ow, AR landfalls, and wet conditions over Central and Southern California with heavy precipitation over the Sierra Nevada. In this example, the February 1-8 event highlights the Oroville Dam crisis (Vahedifard et al. 2017) that resulted in the evacuation of over 180,000 people due to spillway damage from extreme runoff during intense and prolonged AR precipitation with high snow levels causing rain-on-snow. These three events also produced ooding on the leeside of the Sierra Nevada (Sterle et al. 2019).
The second example is for WY 2021 (Fig. 4). Here we highlight three different events (November 1-4, November 27-December 8, and January 9-17) sharing a common NP4 phase combination. For these events (labeled E4-E6), the BP and AP modes were positive (favoring a ridge over Baja California and a trough over Alaska, respectively, c.f. Figure 1a) and the CP and OC modes were negative (favoring high pressure over western Canada and offshore California, respectively). This combination results in the atmospheric circulation (Fig. 4c) and IVT patterns (Fig. 4d), featuring anomalous onshore ow over British Columbia/Paci c Northwest and offshore ow over Southern California. This pattern of circulation (classi ed as WR5, see Fig. 2) is associated with ARs in the northern latitudes and SAW conditions over Southern California. The two longer events (E5 and E6) prompted Red Flag Warnings in Southern California along with large-scale public safety power shutoffs initiated by electric companies to prevent accidental ignitions by utility equipment (Murphy, 2019; Chediak and Sullivan 2021).
3.2. Impacts and circulation characteristics associated with each weather regime Impacts associated with each weather regime are shown in Fig. 5 for AR landfall probabilities (a), precipitation anomalies (b), probability of extreme precipitation (c), and percent of total historical precipitation (d). Additional impacts are shown in Fig. 6 for SAW probabilities (a), temperature anomalies (b), heat wave probabilities (c), and cold extreme probabilities (d). The speci c NP4 phase combination associated with each weather regime is shown in Figure S1 and is also displayed as highlighted text in the lower left of each map in Fig. 2. For all weather regimes shown in Fig. 2, the salient ridge/trough features are well represented by the vast majority of days within each weather regime class. For example, Figure S2 shows the percent agreement among days within each sample in the sign of the Z500 anomaly, which reaches 94-100% for ridge/trough positions at the point of maximum overlap.
Weather regimes 1-6 feature a strong ridge positioned over/offshore western North America (Fig. 2), creating atmospheric blocking conditions that produce dry weather over much of the western US, especially California ( Fig. 5b-d). Offshore ow occurs over Southern California along the eastern boundary of the ridge, bringing elevated chances of SAW conditions (Fig. 6a). Four of these patterns (WR1, WR4, WR5 and WR6) favor warm temperatures, with an elevated chance of heat wave occurrence, whereas two patterns (WR2&WR3) bring cold conditions (Figs. 6b-d). These warm and cold avors of SAWs are consistent with Gershunov et al. (2021). In the northern latitudes, WR 2-6 result in enhanced onshore ow along the northern/northwestern boundary of the ridge, bringing elevated AR probabilities for British Columbia and the Paci c Northwest (Fig. 5a). Weather regimes 7-12 bring wetter conditions to the western US coast (Fig. 5) due to the position of an offshore trough that enhances transport of oceanic moisture over land (Fig. 2). Weather regimes 7-8 feature a trough positioned near the Paci c Northwest coastline. The location of this trough is favorable to precipitation and AR landfalls over Oregon and Washington and into Northern California. Weather regimes 9-10 feature a trough offshore California, which brings precipitation and AR landfalls over Central and Northern California. For weather regimes 11-12, the trough is further equatorward, favoring a more southerly AR track with wet conditions centered over Southern California and penetrating into the interior Southwest.
Weather regimes 15-16 bring anomalous onshore ow over Baja California (Fig. 2), with wet conditions into the Sonoran Desert region (Fig. 5). Concurrently, these weather patterns bring strong easterly/northeasterly ow to Northern California/Paci c Northwest along the eastern boundary of a ridge posited over Alaska or British Columbia. Previous work has noted the link between ARs over Baja

Atmospheric Rivers, Extreme Precipitation, And California Water Resources
The weather regimes responsible for most of California's total and extreme precipitation are WR9-12 ( Fig.  5c-d). WR9 is characterized by a deep and expansive trough centered over the Gulf of Alaska, with a ridge positioned over Southern California that helps to channel moisture towards Central and Northern California (Fig. 2). This weather pattern elevates AR probability 3-fold along the coast from Central California to the Paci c Northwest (Fig. 5a) and brings a high probability of extreme precipitation (Fig.  5c). In Central and Northern California, by contributing upwards of 30% of annual precipitation, WR9 is the most important source of precipitation (Fig. 5d). WR10 is also vital for California water resources (Fig.  5d) and is an important driver of AR landfalls (Fig. 5a) and extreme precipitation (Fig. 5c). It is similar to WR9 in the position of the high and low pressure centers along the West Coast (Fig. 2). However, WR10 is characterized by a ridge over Alaska (negative phase of the Alaska-Paci c Mode) and a positively tilted trough that extends towards the tropics.
WR12 is similar to WR10, except it lacks the ridge over Southern California, which is an important distinction as its absence allows the trough to expand equatorward (Fig. 2) bringing more AR activity and precipitation to Southern California (Fig. 5). WR11 is also impactful for Southern California, bringing a high probability of extreme precipitation when it occurs (Fig. 5c), but it accounts for a smaller proportion of total SoCal precipitation (Fig. 5d) due to the lower frequency of occurrence (i.e., WR11 is more transitory and is observed on 2% of days compared to 9% for WR12). Figure 7 summarizes just how critical these weather patterns are to California precipitation extremes and water resources. Collectively, while they account for only 25% of November-February days, they account for up to 72% of total precipitation and up to 80% of extreme precipitation for some locations in California (Fig. 7a-b). Using the AR intensity scale of Ralph et al. (2019), we nd that WR9 and WR10 are the primary synoptic weather patterns for the most intense and destructive ARs (AR scale 4&5, Fig. 7c). WR9 and WR10 collectively account for approximately 50% of AR category 4 storms and more than 80% of AR category 5 storms impacting California. Median ood damages associated with AR category 4 and 5 events are on the order of $20 million and $260 million, respectively (Corringham et al. 2019).
Previous work has shown how the presence or absence of one-to-three intense storms can make-or-break a water year (Dettinger 2011) or terminate a drought (Dettinger 2013), or prevent the onset of a megadrought (Hatchett et al. 2016). Our ndings suggest that these make-or-break storms most likely carry synoptic signatures similar to WR9-12. These few but large storms are important in building the Sierra Nevada snowpack that later melts and provides bene cial warm season water resources to meet human and ecological demands. Figure 8 shows the relationship between the seasonal frequency of WR9-12 and interannual snowpack variability in the Northern Sierra Nevada, which demonstrates a very strong relationship between the number of days in a season that WR 9-12 were present and snow water equivalent (SWE) at Donner Summit on March 1. The relationship is strong (p < 0.001, r = 0.61) even though important details including precipitation itself or snowmelt events are not considered.

Historic California Floods
We selected eight ood events to investigate from a weather regime perspective. These historic events were selected because they have achieved notoriety for the amount of destruction they caused. These are the oods of December 1955, December 1964, January 1969, February 1986, January 1995, December/January 1996-97, February 2017, and February 2019.  Figure 10 shows Z500 anomalies on the peak day of each event along with the weather regime classi cation, where we see that all peak days are classi ed as either WR9 or WR10. These patterns favor very strong vapor transport over California (Fig. 11). While the patterns are similar, there are distinctions in the position of the ridge located over/near Alaska and the position/orientation of the trough relative to the coast. The positioning of these synoptic features modulates the location of the AR track as it intersects the coast, which is critical from a local impacts perspective (e.g. note the latitudinal differences in the AR landfall location in Fig. 11). While there are slight variations in the synoptic signatures, the weather patterns represent a recurring type of atmospheric circulation linked to extreme precipitation in California, which is consistent with many studies in the literature investigating these or other extreme AR events (e.g.

Santa Ana Winds And Fire Weather
We next investigate the role of the sixteen weather patterns in driving the opposite regional weather phenomena: Santa Ana winds and wild re. Wild res in California are of increasing concern as recent years have brought extremely destructive and widespread wild res with devastating loss of life and property and long periods of unhealthy air quality. The largest Southern California wild res are typically associated with anomalously hot temperatures, dry vegetation, and persistent Santa Ana winds. Here we discuss the types of weather patterns most conducive to wild re in California during winter.
Weather regimes 1-6 all support offshore ow over Southern California (Fig. 2) and SAW probabilities in the range of 39-65%, or 20%-100% above average (Fig. 6a). These six patterns are also anomalously dry for California generally (Fig. 5), although WR6 can produce AR impacts in Northern California (Fig. 5a) in cases when the offshore trough (Fig. 2) moves closer to shore.
Two of these patterns are associated with cold SAWs (WR2&3), and the rest are hot SAW patterns with elevated chances of a heat wave (Fig. 6). Hot SAWs are associated with a ridge centered over the Great Basin, while cold SAWs feature a trough over the interior Southwest and a positively tilted offshore ridge that brings northeasterly ow to coastal California ( Figure S3). The synoptic signatures associated with hot and cold avors of SAWs are consistent with recent ndings (Gershunov et al. 2021). Figure 12 shows the weather regime distribution for start dates of the largest Southern California wild res. As expected, the weather regimes associated with SAWs (WR1-6) are those responsible for most (78%) large wild re starts, and the hot SAW patterns are responsible for more than twice as many starts (58%) as cold SAWs (24%). Gershunov et al. (2021) found hot SAWs to be responsible for the largest wild re acreages burned in Southern California, partly because hot SAWs tend to last longer, high temperatures support enhanced drying of vegetation and lower relative humidity, but also because cold SAWs tend to be preceded by precipitation. Weather regime 5 is associated with the largest number of wild re ignitions (30%), which results from strong offshore ow (Fig. 2) favoring re weather: hot, dry, SAW conditions (Figs. 5&6). The WR5 pattern also occurs more frequently than other hot, dry, SAW patterns (i.e. WR5 occurs on 10% of days compared to 3% for each WR1 and WR4). Composites of atmospheric circulation associated with Southern California wild re start days classi ed as WR1-6 ( Figure S4) highlight the similarity between synoptic circulation on speci c wild re start days and the de ned weather regimes (Fig. 2).

The California Thomas Fire of 2017
In addition to hot, dry, and windy conditions needed for a wild re to grow, the largest wild res are associated with the persistence of these conditions over a period of time without rain. The Thomas Fire of 2017-2018 burned for over a month, igniting December 4 before being fully contained on January 12. were those associated with hot, dry, SAW conditions. From Fig. 13b, the most common weather regimes during December 4-January 12 were WR1 (11 days), WR4 (6 days), WR5 (6 days), and WR6 (5 days).
Containment of the re was aided by a weak atmospheric river on January 9-10 (classi ed as WR8, Fig.  13a). While this storm nally brought rainfall, it also brought tragedy as it included short-duration, high intensity precipitation associated with cold frontal passage that triggered fatal post-re debris ows in the coastal community of Montecito (Oakley et al. 2018).

Observed Trends In Weather Regime Frequency And Implications For California Water Resources And Fire Risk
In the past decade, California has endured two extreme and persistent droughts, ongoing extreme heat, and eight of the ten largest wild res in over a century of records (Krishnakumar and Kannan 2020). Here, we investigate if and how winter weather patterns are changing across the western US, and how these changes might affect winter weather variability and extremes in California. Figure 14 gives the seasonal frequency of each weather regime over the historical record, along with statistically signi cant (p < 0.05) increasing (red) and decreasing (blue) trends. The weather patterns found to be increasing in frequency are WR1, WR4, WR5, and WR7. While WR7 is not linked to extreme California weather in our analysis, the other three weather regimes (1, 4 &5) are major contributors to historical wild res ( Fig. 12 and Fig. 13b) and are associated with heat waves in California and the western US (Fig. 6c). The combined seasonal frequency of these three weather regimes is provided in Fig. 15b, which shows a strong increase in the frequency of these impactful weather patterns (slope = 0. Along with increasing re risk due to the rising frequency of hot, dry, SAW patterns, we also nd evidence of decreasing frequency of weather patterns associated with cold SAWs (WR3; see also Gershunov et al. 2021) and precipitation in the Southwest (WR11, WR12, and WR14). Figure 15a shows the combined frequency of the patterns associated with southwest precipitation (WR11, WR12, and WR14, slope=-0.18% per year, p < 0.001). These are all associated with a deep trough over and offshore from western North America (Fig. 2) with moisture transport along a southern track bringing precipitation to Southern California and the Desert Southwest (Fig. 5). These weather regimes also tend to be colder (Figs. 6b and   6d), therefore this decreasing trend could mean less precipitation in general, but also less cold precipitation which has implications for snow accumulation and storage. We nd no long-term trends for the weather patterns associated with the strongest ARs and the most destructive California oods (WR9 and WR10). This suggests that atmospheric circulation is changing to favor re weather at the expense of precipitation in Southern California, but the risk of extreme precipitation and ooding caused by the strongest ARs is not diminishing.
The observed trends in the frequency of certain weather patterns is related to the rising frequency of ridging over land along the west coast. Speci cally, we observe a positive trend in the Baja-Paci c mode, which makes ridging over Baja California more likely, and a negative trend in the Canadian-Paci c mode, which corresponds to an increase in ridging over the Paci c Northwest and western Canada ( Figure S5 and Fig. 1a). These geospatial differences in trends of height anomalies over land relative to those over the Paci c Ocean favor atmospheric blocking conditions associated with dry weather in California.

Timescales Of Persistence And Relevance For S2s Predictability
Within a season, the persistence of one or more NP4 modes in one phase or another makes certain weather conditions more or less likely as ridges or troughs persist over and offshore from western North America (c.f. Figure 1a). As the NP4 modes oscillate with different periodicity and come in-and out-ofphase with each other (as seen in Figs. 3a and 4a), the weather regimes transition on relatively short daily timescales (median = 3 days, Figure S6, purple bars). However, the NP4 modes themselves are more persistent, uctuating on extended-range (10-21 days) to S2S (21-90 days) timescales, and this persistence is modulated by the more slowly-varying climate system (GGR '20). The timescale of persistence of the NP4 modes (number of consecutive days in one phase before transitioning) is on the order of a few days to several weeks ( Figure S6, orange lines). Although the median persistence of the NP4 modes over the historical record is 10-13 days, within a given season much longer persistence commonly occurs. For example, Figure S6 (green line) shows the longest-lasting event event identi ed in each season. This shows that during all seasons in the historical record there has been at least one longduration event where one or more modes persisted in one phase for at least 19 days and up to 62 days before transitioning. During 68% of seasons, such phase persistence has lasted for at least 30 days. This type of persistence makes particular extreme weather hazards more or less likely on timescales up to two months. Additionally, this behavior can determine if California reaches, exceeds, or falls short of achieving normal precipitation in a given water year. On seasonal timescales, it is common for one or more NP4 modes to exist predominantly in one phase or another during a season due to modulation by large-scale teleconnections such as ENSO (GGR '20). Thus, the interaction between large-scale climate forcing and the NP4 modes, and the resulting seasonal weather regime distribution, represents important S2S linkages that should be explored in future research.

Discussion
In this study, we demonstrate how four modes of atmospheric circulation over the North Paci c Ocean interact to produce distinct weather patterns that recur throughout the historical record. These recurring weather patterns drive AR landfalls, extreme precipitation, historic oods, Santa Ana winds, wild res, and temperature extremes during California winters. Relationships between the driving NP4 modes and climate-scale teleconnections like ENSO have already been discovered (GGR'20) and the results presented here suggest important linkages between both seasonal-and weather-scale processes that project onto S2S predictability of the various meteorological phenomena explored in this work. The methodology developed here can easily be applied to dynamical weather forecast models to investigate the predictability of extreme weather impacts at S2S lead times. Predicting the phase of one or more NP4 modes, using dynamical, statistical, or hybrid methods, would be highly useful for situational awareness regarding the likelihood of extreme weather events. Dynamical models have been shown capable of producing skillful forecasts of large-scale circulation features at lead times of 2.5 weeks (e.g. Ferranti et al. 2018Ferranti et al. , 2019 to several weeks (Gibson et al. 2020b). Additional lead time and higher con dence in long-lead forecasts can be translated by forecasters into improved impact-based decision support (Uccellini and Ten Hoeve 2019) to partner agencies and used to provide earlier warning to public audiences through social media before extreme events (e.g., Lambrecht et al. 2019Lambrecht et al. , 2021).
An important outcome of this work is the generation of a catalog of daily weather patterns spanning 73years (https://weclima.ucsd.edu/data-products), which can be used for monitoring and predictability studies. The methodology offers a perspective on weather pattern transitions not previously explored. In contrast to more traditional methods that use a categorical perspective (e.g., cluster analysis or selforganizing maps), our approach identi es weather patterns as a function of the continuous wave-like atmospheric oscillations that occur in the region of the globe most relevant for California (c.f. . Our weather regime classi cation has been shown to be skillful at representing important circulation features associated with impactful events, including destructive oods and Southern California wild res. Using this approach, we nd that the seasonal frequency of weather patterns associated with wild re is increasing while other patterns that bring precipitation to southern California and the Desert Southwest are decreasing in frequency. This is an unsettling set of trends, as during 2020 California suffered its worst wild re season on record, causing extensive impacts to ecosystems, property, and public health (e.g. Aguilera et al. 2020Aguilera et al. , 2021a. The continued dry conditions experienced during winter 2020-2021 have further exacerbated already extreme re hazards and as of this writing, the 2021 re season is on track surpass the previous year's record in terms of acreage burned. Importantly, while circulation changes appear to be reducing the likelihood of some types of precipitation events (WR11, WR12, WR14), there is no observed change in the weather patterns responsible for the most intense ARs and destructive oods (WR9 and WR10). With warming due to anthropogenic climate change, these patterns could potentially be even more destructive as the atmosphere can more effectively transport larger quantities of water vapor (Rhoades et al. 2021). This is consistent with recent work focused on precipitation regime changes in California using climate model projections ( There are many research questions that could be explored using this methodology and data catalog. While we provided only a cursory look at the relationships with snowpack, the role of different weather patterns in driving snow accumulation and depletion throughout the western United States, would be valuable knowledge from a water resource perspective, particularly as snowpack declines continue (Siirila-Woodburn et al. 2021). Further investigation on the timing and transition of different weather patterns could provide insight for S2S predictability. While this analysis focused on the California winter season, the methodology could be applied to other seasons to study warm season heat waves and/or re weather, as well as to other regions. Improving our understanding of the linkages between climate change, seasonal-scale teleconnections, and extreme weather will be critical for effective emergency planning, resource management, and climate change adaptation efforts at local, regional, and state levels.

Ethics Declaration
The authors declare no competing interests.

Availability of Data and Material
The weather regimes catalog developed for this study, along with the NP4 dataset, Santa Ana winds regional index, and the SIO-R1 AR catalog are available at https://weclima.ucsd.edu/data-products/. The Livneh temperature and precipitation data are available at https://climatedataguide.ucar.edu/climatedata. Wild re information is from the California Fire Protection's Fire and Resource Assessment Program (FRAP) re perimeters database available at https://frap. re.ca.gov/frap-projects/ re-perimeters/. Snow water equivalent for the Sierra Nevada are from the Natural Resources Conservation Service available at https://wcc.sc.egov.usda.gov/nwcc/rgrpt?report=snowcourse&state=CA.

Code Availability
Code developed for this study will be made available upon request 10. References Figure 1 (a) The four North Paci c atmospheric teleconnection patterns (NP4 modes) de ned and described in GGR'20 shown here in the phase most associated with wet conditions over California (de ned as the positive phase as in GGR '20). The color scale gives the temporal correlation (r) between each teleconnection pattern and the standardized Z500 anomalies at each grid cell (b) locations where more than 50% of Z500 variance is explained by one of the NP4 modes using linear regression with r2>0.5, (c) total Z500 variance explained by the joint NP4 modes using multiple linear regression. Figures (a) and (c) are adapted from GGR'20.    Impacts associated with each weather regime showing (a) AR landfall probability, (b) precipitation anomaly, (c) probability of extreme precipitation, (d) percent of total historical precipitation. Precipitation extremes are daily accumulations above the local 95th percentile. Sample size is important for comparing (c) and (d) where some weather regimes bring high probabilities of extreme precipitation but are less common and thus do not contribute substantially to the total historical precipitation (e.g. WR7 and WR11, see sample sizes in Figure 2)  The relationship (r=0.61) between snow water equivalent on March 1 at Donner Summit in the Northern Sierra and the frequency of weather regimes 9-12 during November-February.  Observed Z500 anomalies for the day with the most intense AR footprint over California for each ood event, de ned by the summed IVT over California. The weather regime classi cation for each day is given as highlighted text in the lower left corner of each map.

Figure 11
Observed IVT for the same days shown in Figure 10. The weather regime distribution of Southern California wild re starts.

Figure 14
Observed frequency of each weather regime during Nov-Feb (black) and statistically signi cant (p<0.05) trends colored red for increasing trends or blue for decreasing trends.