California margin temperatures modulate regional circulation and extreme summer precipitation in the desert Southwest

In August 2022, Death Valley, the driest place in North America, experienced record flooding from summertime rainfall associated with the North American monsoon (NAM). Given the socioeconomic cost of these type of events, there is a dire need to understand their drivers and future statistics. Existing theory predicts that increases in the intensity of precipitation is a robust response to anthropogenic warming. Paleoclimatic evidence suggests that northeast Pacific (NEP) sea surface temperature (SST) variability could further intensify summertime NAM rainfall over the desert southwest. Drawing on this paleoclimatic evidence, we use historical observations and reanalyzes to test the hypothesis that warm SSTs on the southern California margin are linked to more frequent extreme precipitation events in the NAM domain. We find that summers with above-average coastal SSTs are more favorable to moist convection in the northern edge of the NAM domain (southern California, Arizona, New Mexico, and the southern Great Basin). This is because warmer SSTs drive circulation changes that increase moisture flux into the desert southwest, driving more frequent precipitation extremes and increases in seasonal rainfall totals. These results, which are robust across observational products, establish a linkage between marine and terrestrial extremes, since summers with anomalously warm SSTs on the California margin have been linked to seasonal or multi-year NEP marine heatwaves. However, current generation earth system models (ESMs) struggle to reproduce the observed relationship between coastal SSTs and NAM precipitation. Across models, there is a strong negative relationship between the magnitude of an ESM’s warm SST bias on the California margin and its skill at reproducing the correlation with desert southwest rainfall. Given persistent NEP SST biases in ESMs, our results suggest that efforts to improve representation of climatological SSTs are crucial for accurately predicting future changes in hydroclimate extremes in the desert southwest.


Introduction
In August 2022, regions of the desert southwest including Death Valley, the driest place in North America, experienced once-in-a thousand year flooding.This was a result of summertime storms that dumped up to 75% of normal annual precipitation amounts in the span of a few hours (Canon 2022).These storms occurred at the northern side of the North American monsoon (NAM), the primary source of summer rainfall in southwestern North America (Adams andComrie 1997, Cook andSeager 2013).These types of events are associated with loss of life and infrastructure damage.In addition, above-average monsoon rainfall has been linked to increases in invasive plant biomass, increasing fire risk (Moloney et al 2019).The profound socioeconomic and ecological impacts of the region's precipitation extremes highlight the need to understand the mechanisms underlying their variability.
An increase in precipitation intensity is expected as a result of global warming, since saturation specific humidity increases with temperature (O'Gorman 2015).In the NAM domain, earth system models (ESMs) and regional models all suggest that warming will intensify individual storms, despite predictions of an overall decrease in summer rainfall by the end of the 21st century (albeit with considerable structural uncertainty-Meyer and Jin 2017, Pascale et al 2017, 2018, Moon and Ha 2020, Almazroui et al 2021).Nevertheless, the entire summer of 2022 featured above-average rainfall, especially on the northern edge of the NAM domain (figure S1(a)).Not only were rainfall rates above average in Nevada, Arizona, and New Mexico, but values of daily outgoing longwave radiation (OLR) show a systematic shift to a longer left tail, suggesting a greater frequency of cooler cloud tops associated with convective rainfall.
We hypothesize that north Pacific sea surface temperature (SST) variability helped drive these higher rainfall rates.The summer of 2022 featured positive SST anomalies in coastal regions of the northeast Pacific (NEP), as well as a La Niña event in the eastern equatorial Pacific (EEP).Observational analyses have shown that La Niña events feature a stronger monsoon ridge, enhancing the strength of the circulation (Castro et al 2001).Previous analyses of intraseasonal variability in the monsoon found that cool conditions in the EEP cold tongue could drive an earlier monsoon onset (Castro et al 2001, Bieda et al 2009).Modeling results, however, are equivocal about the impact of extratropical Pacific SST anomalies on the monsoon.Some studies suggest that warm NEP SST anomalies, similar to the configuration observed in summer 2022, should weaken the monsoon (Castro et al 2001).In contrast, more recent work suggests that these types of events should strengthen the monsoon (Fu et al 2022, Beaudin et al 2023).The latter two studies used idealized atmosphere-only simulations forced by fixed SSTs, and require corroboration by observations.Yet, observations and paleoclimatic reconstructions are equivocal about the relationship between southwest summer precipitation and largescale SST patterns (Griffin et al 2013, Coats et al 2015, Carrillo et al 2016, Demaria et al 2019).Given these contradictory findings, more work is needed to contextualize the role of SST variability in driving extremes similar to summer 2022.Along these lines, paleoclimatic evidence from past warm intervals has identified a link between warm SSTs off the southern California margin and an intensification of monsoon rainfall (Bhattacharya et al 2022, Fu et al 2022).If this relationship holds in the modern, it suggests that warm NEP SSTs, especially on the California Margin, would have played a role in August 2022 flooding.
In this study, we explore the link between SST on the CA margin and precipitation in the northern NAM domain.Because of our interest in events similar to the Death Valley flooding in August 2022, we focus on the northern edge of the NAM domain, or regions of the desert southwest in Arizona, New Mexico, California and Nevada.Using observational data, reanalyzes, and ESM simulations, we analyze the relationship between large-scale SST patterns and both extreme precipitation and seasonal precipitation totals.We show that summers of a greater frequency of days with extreme precipitation in the northern NAM domain tend to co-occur with intervals of warm SSTs on the CA margin.There is also a statistically significant relationship between CA margin SSTs and seasonally-averaged summertime rainfall in the northern NAM domain.Our results therefore provide important context for the processes underlying recent extreme precipitation in the desert southwest, and also identify a mechanism of inter-annual variability in extremes that may continue to influence regional precipitation into the 21st century.However, the ability of Coupled Model Intercomparison Project (CMIP) phase 5 and 6 ESMs to reproduce this relationship is highly variable, and depends in part on the magnitude of a model's warm SST bias on the CA margin.This has implications for our ability to use ESMs to estimate future changes in extreme precipitation, signal-to-noise ratios, and hydroclimate-related risks in the desert southwest.

Composite analyses and conditional probability
Our work focuses on the northern NAM domain, consisting of land regions of the desert southwest between 28 • -37 • N and 115 • -108 • W. We used daily Global Precipitation Climatology Centre's (GPCC) 1.0 • data to define the 95th percentile of daily precipitation rates at each grid point.This threshold, otherwise known as 'p95' , is a widely accepted metric of extreme precipitation (Sillmann et al 2013).We then calculated the number of days between 14 June and the end of September in each year that exceed p95 for the interval between 1982 and 2014 (Schneider et al 2008).While heavy rainfall is relatively rare in this desert setting, events like August 2022 highlight the profound impact of extremes.
To contextualize extreme precipitation, we analyzed composites of daily fields of OLR, as an indicator of cold cloud tops and convective rainfall, from NOAA (Liebmann and Smith 1996); zonal and meridional moisture flux; and daily maximum 3hourly convectively available potential energy (CAPE, a measure of energetic favorability of the atmosphere for deep convection and rainfall) from the North American Regional Reanalysis (NARR -Mesinger et al 2006).We also analyze monthly sea level pressure (SLP) and 850 mb geopotential heights from the NCEP-NCAR reanalysis (Kalnay et al 1996).In order to link changes in the frequency of daily precipitation extremes to SST, we analyzed anomalies of monthly SST (Ishii et al 2005) associated with summers that contain the greatest frequency of extreme precipitation.
We also quantified the probability of seeing a summer with above-average days (e.g.greater than 6 days) of extreme precipitation in the northern NAM domain in years with anomalously warm CA margin temperatures (years where SST anomalies 25 and 32 • N and 125 and 110 • W are >1σ above average).This is the conditional probability that a summer will have greater than 6 days of precipitation exceeding p95 given anomalously warm SSTs on the CA margin.We assessed this probability for warm CA margin SST anomalies at different lead times (e.g. the preceding April-June SSTs), to see whether preceding CA margin SSTs can serve as predictors of northern NAM domain precipitation extremes.The statistical significance of these conditional probabilities was calculated using a 1-sided binomial test to assess whether it was significantly different from random chance (i.e. that the observed conditional probability is not significantly larger than expected from a process with a 50% probability of occurrence (Wilks 2011)).

Correlation analysis and comparison with ESM simulations
We quantified the relationship between CA margin SSTs and monthly mean precipitation over the northern NAM domain by correlating the same index of CA margin SSTs anomalies with monthly mean precipitation from multiple precipitation products to establish the robustness of the relationship (0.25 • GPCC product between 1960 and 2020, the CPC merged analysis of precipitation (CPC), the Global Precipitation Climatology Project, and the NARR (Mesinger et al 2006, Xie et al 2007, Schneider et al 2008, Adler et al 2018)).We also analyze the correlation of our SST index and fields of SLP, 850 mb geopotential height (Kalnay et al 1996) between 1950and 2020, and zonal and meridional moisture flux between 1979and 2022(NARR, Mesinger et al 2006).The focus on monthly mean correlations facilitates comparison to CMIP models, since only monthly mean fields are available for many models.
To determine the extent to which CMIP5 and CMIP6 ESMs reproduce the observed relationship between CA margin SSTs, precipitation, and circulation, we analyzed historical model simulations from 23 ESMs, including several from the HighResMIP project (Eyring et al 2016, Haarsma et al 2016, Chang et al 2020) (table S1).Because the set of simulations we analyze differ in their input forcing datasets (e.g.CMIP5 vs. CMIP6 historical simulations), we focused on analyzing the sensitivity of precipitation to CA margin SSTs instead of making inferences about how trends in CA margin SSTs may influence the future behavior of the NAM.We quantified the linear correlation between southern CA Margin SSTs (e.g. the previously defined SST index, averaged between 25 • and 32 • N and 125 • and 110 • W) and northern NAM domain precipitation (e.g. the area between 28 • -37 • N and 115 • -108 • W).Finally, we compared the correlation between the SST index and large-scale fields (e.g.vertically integrated moisture flux, 850 mb geopotential height, and SLP) between a subset of models that perform well (e.g.exhibit a realistic SSTprecipitation correlation) and a subset of models that perform poorly.This analysis required regridding historical output to a common 1 • by 1 • grid before computing cross-correlations between each models' SST index and large-scale fields (see SI).

California margin temperatures and northern NAM domain precipitation
Between 1979 and 2014, 5 summers featured the greatest number of days with rainfall exceeding p95 (Cavazos et al 2008, Sillmann et al 2013) (figure 1(a)).These years also show a statistically significant shift to larger values of daily maximum 3-hourly CAPE, indicating an increase in the energetic potential of the atmosphere to generate deep convection (figure 1(d)).This shift in CAPE is accompanied by a shift in daily OLR indicative of colder cloud tops associated with deep convection (figure S2).These shifts in the distribution of CAPE and OLR suggest that summers with greater extreme precipitation days featured an atmospheric environment that was more conducive to convective activity.This is consistent with station-based observations associated with the initiation of 9-10 August storms over Las Vegas (figure S11).Atmospheric soundings from Las Vegas indicate a shift to low-level southerly flow overnight, coincident with the development of a deep, moist layer and stronger instability despite relatively low early morning surface temperatures.This results in a near-doubling of CAPE (figure S11).
The composite SST pattern associated with these five summers reveals a 'horseshoe' pattern of warmth in the NEP similar to the warm phase of Pacific decadal variability and the extratropical expression of warm ENSO events (Di Lorenzo et al 2023) (figure 1(d)), with stronger anomalies near Alaska and off the CA margin.The extratropical portion of this SST pattern resembles the SST anomaly pattern from summer 2022 (figure S1(d)).However, unlike 2022, the EEP cold tongue shows only weakly positive SST anomalies, reflecting that these 5 years featured different phases of ENSO.While some years featured La Niña events (e.g.1984,1990,1996), others  (1983, 1984, 1990, 1996, and 2014) had moderate or decaying El Niño events (e.g. 1983, 2014).The lack of a consistent ENSO phase reflects the relatively weak relationship between ENSO and NAM precipitation (Castro et al 2001).It also suggests that NEP SST patterns, especially on the CA margin, may play a more important role in modulating the NAM than the EEP, although we note that tropical and extratropical variability are connected (Di Lorenzo et al 2023).It is notable that at least two of the five years used in this analysis coincide with significant seasonally-persistent extratropical marine heat waves (e.g.1990, 2014) (Capotondi et al 2022), suggesting a link between marine and atmospheric extremes.

Implications for seasonal prediction
The association between CA margin SSTs and northern NAM precipitation extremes raises the possibility that coastal SSTs could aid in efforts to predict inter-seasonal precipitation extremes.This may especially be true given the strong seasonal persistence of temperature anomalies in this region on seasonal timescales, despite some sub-seasonal variability (Wei et al 2021) (figure S3).To test this possibility, we quantified the conditional probability of a summer containing an above-average number of days exceeding p95 given anomalously warm CA margin temperatures (e.g.greater than 1σ above average).When CA margin SSTs are 1σ above average in JAS, there is a 85% of seeing greater than average days with precipitation exceeding p95 (figure 2).Warm CA margin SST anomalies during early summer (JJA, MJJ) and spring (AMJ) are also associated with summers with above-average extreme precipitation days.
Each of these relationships are statistically significant at the 95% level (1-sided binomial test (Wilks 2011)).These results highlight the potential utility of coastal SSTs for predicting summers with greater extreme precipitation days in the desert southwest.

Mechanism of SST-monsoon linkage
Consistent with an energetic environment more conducive to summertime convection, the summers highlighted in figure 1(a) are associated with higher monthly mean rainfall rates over the northern NAM domain.They are also associated with higher geopotential heights over the North American continent and the southern CA margin, with a trough to the northwest (figures 1(e), S11 and S14).This is coupled with stronger southwesterly moisture flux, and a cyclonic circulation anomaly centered over Baja California, similar to the changes observated during the August 2022 floods (figures 1(g) and S11).The anomalous moisture flux results in a large increase in precipitable water over Baja California, the northern NAM domain, and California margin, with an almost 50% increase in humidity in some of these regions (figures 1(g) and S4).Changes in evaporation are modest compared to changes in moisture flux and precipitable water (figure 1(h)).The link between total precipitable water off the southern California coast and extreme NAM precipitation has been noted in studies of individual flood events (Mazon et al 2016, Yang et al 2017), but the link to coastal SSTs and large-scale circulation has not been explicitly studied.
Correlations of NARR precipitation with the CA margin SST index are largely positive, especially over Baja California and the southern Great Basin, northern margin of the present-day NAM domain (figure 2(a)).This correlation pattern is statistically significant and robust across multiple observational products (figures S6 and S7).CA margin SSTs are correlated with stronger southerly moisture flux over Baja and coastal California (figure 2 We contend that warm CA margin SST anomalies drive stronger southwesterly moisture transport because they help weaken the southeast edge of the North Pacific subtropical high pressure system (NPSH).Correlations with 850 mb geopotential heights reveal that warm CA margin SSTs are associated with a weakening of the geopotential gradient on the southern edge of the NPSH (figure 2(c)).Further support for this comes from the strong negative correlation between CA margin SSTs and SLP on the southeast edge of the NPSH (figure 2(d)).Given previous work showing that cool SSTs over eastern ocean margins help maintain the strength of the subtropical highs, we suggest that anomalously warm temperatures on the CA margin weaken the NPSH in this region by reducing local static stability (Seager et al 2003, Bhattacharya 2022).The strong negative correlation between CA margin SSTs and SLP centered west of Baja California and extending along the edge of the NPSH is not present in the correlation field between ENSO and SLP (figure S8).This is likely because variability in CA margin SSTs reflects additional processes beyond ENSO-induced variability (Di Lorenzo et al 2023).There is some similarity between the SLP correlation pattern with CA margin SSTs and the PDO, consistent with recent findings that the PDO modulates SST variability and MHW intensity off Baja California (figure S8) (Ren et al 2023).
Our analyses suggest that CA margin SSTs weaken the southeast edge of the NPSH, favoring stronger moisture flux into the desert southwest.Subsequent increases in humidity would promote higher dewpoints and fuel instability over these normally dry desert regions.From this perspective, warm CA margin SSTs help enhance the positive CAPE generated by a warm summertime Gulf of California (Johnson and Delworth 2023).Given that NPSH strength and underlying SSTs are tightly coupled via air-sea interactions (Seager et al 2003), observations alone are insufficient to establish the direction of causality between SST and NPSH strength.In fact, outflow from the NAM may help amplify SST anomalies on the California margin (Clemesha et al 2023).However, our results agree with several atmosphereonly regional and global simulations showing that CA margin SSTs play a causal role in driving a cyclonic circulation anomaly and stronger meridional moisture flux into the southwest (Fu et al 2022, Beaudin et al 2023).

Model representation of SST-summer rainfall relationship
Figures 1 and 2 suggest that summers with warm CA margin SSTs not only feature more extreme precipitation days, but also see an atmospheric shift conducive to higher seasonal rainfall totals.Accurately representing this SST-monsoon linkage is therefore key for quantifying and predicting future risks related to extreme precipitation (e.g.flooding, infrastructure damage).Since most future projections of the NAM system rely on downscaled (or direct) output from ESMs, we next assess whether ESMs reproduce the CA margin SST-monsoon relationship found in observational products (figure 2).
Only a small subset of ESMs simulate a similar correlation between CA margin SST and northern NAM domain precipitation as compared to observational products.Furthermore, there is a significant negative correlation between a given models' warm SST bias on the CA margin and the strength of the correlation with northern NAM domain precipitation: ESMs that are too warm in the CA margin relative to observations underestimate the correlation between summer precipitation and CA margin SSTs (figure 3).For two ESMs, increasing resolution improves both the SST bias and the strength of the correlation (figure 3).This coheres with previous  the higher resolution configuration of the CNRM model produces a weaker CA Margin SST-NAM summer precipitation correlation than its low resolution counterpart (figure 3).In addition, higher resolution models do not necessarily have a more realistic climatology of NAM precipitation (figure S9).Instead, we suggest that biases in their simulation of the large-scale climate play an important role in models' ability to capture coupling between SST variability, atmospheric circulation, and regional hydroclimate, as captured in figure 3.
To further investigate this hypothesis, we quantify the correlation between SLP, 850 mb geopotential heights, and meridional moisture flux for the ESMs that are best able to (CESM1.3-HR;CanESM5-1; ACCESS-CM2; MPI-ESM1-2-HR) and least able to (MIROC6; FIO-ESM2; NESM3) reproduce the observed SST-monsoon linkage (figure 4(a)).Model precipitation correlations are shown in figure S10.SLP fields reveal that the best performing models produce a NPSH that is extended slightly farther southeast and weaker than the worst performing models (e.g.1012 mb contour that extends south of 20 • N, similar to observations (figure 2(c))).
The best performing models produce a strong negative correlation between SLP and the CA margin SST index, especially on the southeast edge of the NPSH (figure 4(a)).This resembles the pattern seen in observational data (figure 4(b)).In contrast, the worst performing models produce the wrong sign of correlation between the CA margin SST index and SLP.The best performing models also produce positive correlations between CA margin SSTs and 850-mb geopotential heights over the California Margin and a negative correlation to the west, similar albeit slightly different in pattern to observations (figures 4(b) and 2(a)).The worst performing models only produce a localized region of positive correlation, failing to reproduce the east-west dipole in correlation seen in observations.These differences in large-scale correlation patterns directly translate into differences in the correlation pattern across models between CA margin SSTs and moisture flux: the best performing models produce a much stronger correlation between meridional moisture transport and the CA margin SST index than the worst performing models (figures 4(c) and (f)), especially in the regions over Baja California and southern California.
These results suggest that the difference in skill between the best and worst performing models in our analysis relates to the fact that the best performing models exhibit a tighter coupling (and stronger correlation) between the underlying anomalies of SST, atmospheric circulation (e.g. the NPSH), and moisture flux than seen in the worst performing models.This appears to be a direct function of the stronger warm SST bias on the California margin in the worst performing models.We hypothesize that for models with a strong warm bias on the CA margin, a given SST anomaly represents a smaller fractional or percent change in SST and hence a smaller perturbation to the overlying atmosphere.Therefore, in more strongly warm biased models, a given SST anomaly may be less efficient at altering atmospheric circulation (e.g.weakening static stability).ESMs with a stronger warm bias may therefore generate less variability in the southeast edge of the NPSH, as well as a weaker correlation between SST, moisture flux, and NAM precipitation.

Conclusions
In this paper, we used observational data and reanalyzes to demonstrate a linkage between California Margin SSTs and precipitation over the US southwest, Baja California, and western Mexico, which comprise the northern edge of the NAM domain.Warm SSTs on the CA margin result in greater southwesterly moisture flux and increases in precipitable water over the desert, creating a summertime energetic environment that is more favorable for moist convection.Warmer SSTs drive increases in the number of days with extreme precipitation, as well as an overall increase in summertime rainfall rates.Because SST anomalies on the CA margin show strong monthly to inter-seasonal persistence, spring or early summer SST anomalies could be used to predict years with a greater frequency of daily precipitation extremes in the northern NAM domain.Our results suggest that the extreme precipitation observed in August 2022 was at least in part modulated by the large-scale SST pattern, and that other events with similar underlying dynamics have occurred over the observational record.
The link between SSTs and NAM precipitation has been explored in previous observational and modeling studies, but results have been equivocal (Castro et al 2001, Griffin et al 2013, Carrillo et al 2016, Demaria et al 2019, Beaudin et al 2023).This may stem from the fact that the SST pattern associated with greater daily precipitation extremes does not resemble a canonical warm ENSO or positive PDO phase.Moreover, the strongest SST-rainfall correlations occur in the northern NAM domain and are relatively weak in the core monsoon domain in western Mexico.Studies that focus on mode-based indices of the ENSO or the PDO, or analyze precipitation only in the core monsoon domain, may therefore have missed this association between SST anomalies on the CA margin and NAM rainfall.Our observational analyses support previous work emphasizing the importance of extratropical North Pacific SSTs in governing the spatial footprint of NAM precipitation (Bhattacharya et al 2022, Fu et al 2022, Beaudin et al 2023).
Over the recent observational record, interannual SST variability on the CA margin has been linked to persistent multi-year marine heat waves (Meyer andJin 2017, Fewings andBrown 2019).While previous work has explored the linkage between marine heat waves and winter precipitation over western North America (Swain et al 2014), we provide an observational link between extreme events in the marine realm and extreme summertime precipitation on land.Given that observational data and paleoclimate records suggest strong decadal variability of CA margin SSTs, our results also raise the possibility for decadal modulation of precipitation extremes in the desert southwest (O'Mara et al 2019).While there is some evidence for decadal variability in NAM precipitation extremes, more long-term precipitation datasets, including paleoclimate proxy datasets, are needed to explore this possibility (Griffin et al 2013, Demaria et al 2019).
It is possible that the strength of the CA margin SST-northern NAM domain rainfall relationship is modulated by equatorial Pacific SSTs.El Niño events result in a southward shift of the intertropical convergence zone in the EEP, enhancing atmospheric stability over the southwest (Pascale et al 2017).In addition, central Pacific El Niño events may reduce NAM rainfall by inhibiting the development of disturbances that can serve as precursors to strong surges of NAM convection (e.g.Gulf of California surges) (Kim et al 2011).We are unable to disentangle the relative influence of subtropical versus tropical SSTs on northern NAM precipitation in this study because of the limited observational record that offers few realizations of extreme precipitation events, especially since tropical and subtropical SST variability are highly correlated.Long integration of ESM simulations, as well as AMIP-style simulations, would be a useful next step for disentangling the importance of subtropical versus tropical SST variability on desert southwest precipitation.
Finally, we showed that historical simulations of ESMs show varying skill at reproducing the correlation between CA Margin SSTs and northern NAM precipitation.ESMs featuring a strong warm on the CA Margin show less skill at reproducing the observed correlation.Indeed, some ESMs produce significant correlations of the wrong sign.We hypothesize that ESMs with a strong warm bias underestimate the coupling strength of atmospheric circulation and SST on the CA margin.Both CMIP5 and CMIP6 ESMs exhibit systematic warm SST biases in the subtropical NEP, likely stemming from biases in shortwave radiation and ocean heat transport (Wills et al 2022, Zhang et al 2023).Given the results presented herein, many ESMs are likely to systematically misrepresent an important source of interannual variability in desert southwest precipitation.This in turn undermines confidence in studies that use direct or downscaled ESM outputs to quantify future changes in precipitation extremes, estimate signal-to-noise ratios for regional hydroclimate, or analyze future changes in hydroclimaterelated risk over the desert southwest (Marvel et al 2019, AghaKouchak et al 2020).CMIP6 models are known to underestimate the severity and duration of multi-month or multi-year marine heat waves, similar to those that cause warming on the southern CA margin, and may contain persistent biases that influence their ability to reproduce observed SST trends (Seager et al 2019, 2022, Plecha and Soares 2020).Efforts to improve ESM representation of climatological SSTs and SST variability will therefore greatly improve our ability to estimate variability and trends in precipitation extremes, with broad implications for our understanding of future regional hydroclimaterelated risks, especially in arid regions like the desert southwest.
It remains an open question as to whether the relationship between SSTs on the CA margin and northern NAM precipitation will persist into the 21st century.Modeling studies predict a weakening of the NAM with anthropogenic warming, in part from a dynamic response resulting in a warmer, more stable troposphere over the southwest (Pascale et al 2017) and thus a higher threshold for convection (Pascale et al 2018).A given CA margin SST anomaly may become less effective at generating positive anomalies of CAPE over the northern NAM domain, decreasing the correlation to summertime precipitation in this region.We briefly assess this possibility by analyzing the four ESMs that best reproduce the observed CA margin SST-northern NAM precipitation correlation, and find that all produce a strong, statistically significant correlation well into the 21st century, with two ESMs even producing a strengthening of this association (figures S12 and S13).While further analyses, especially using large ensemble approaches or AMIPstyle simulations, are needed to disentangle the relative influence of interannual SST variability and forced changes on future precipitation in the northern NAM domain, our results suggest that the CA margin SST-NAM monsoon linkage could aid the effort to predict monsoon extremes well into the 21st century.

Figure 1 .
Figure 1.Composites associated with summers featuring the greatest number of extreme precipitation days.(a) number of summertime days with precipitation exceeding the 95th percentile (p95) between 1981 and 2020 over the region between 28 • and 37 • N and 115 • and 108 • W (this region is outlined in box in panel (f).Gray bars highlight the five years with the greatest number of extreme precipitation days(1983, 1984, 1990, 1996, and 2014).(b) Daily maximum of 3-hourly CAPE for climatology (tan), each individual summer (light blue) and a composite of all summers (dark blue).(c) Probability of seeing greater than 6 days of precipitation above p95 between July and September in the presence of a warm SST anomaly on the CA margin at different leading seasons.Seasons with * are statistically significant at the 95% level.(d) and (e) SST and 850-mb geopotential height composite anomaly associated with these five summers, with box showing region used for SST index in subsequent figures.(f) Monthly-mean NARR rainfall composite during these summers; (g) Anomalies of vertically integrated moisture flux (vectors) and total precipitable water (shading); (h) evaporation composite anomaly (shading).
Figure 1.Composites associated with summers featuring the greatest number of extreme precipitation days.(a) number of summertime days with precipitation exceeding the 95th percentile (p95) between 1981 and 2020 over the region between 28 • and 37 • N and 115 • and 108 • W (this region is outlined in box in panel (f).Gray bars highlight the five years with the greatest number of extreme precipitation days(1983, 1984, 1990, 1996, and 2014).(b) Daily maximum of 3-hourly CAPE for climatology (tan), each individual summer (light blue) and a composite of all summers (dark blue).(c) Probability of seeing greater than 6 days of precipitation above p95 between July and September in the presence of a warm SST anomaly on the CA margin at different leading seasons.Seasons with * are statistically significant at the 95% level.(d) and (e) SST and 850-mb geopotential height composite anomaly associated with these five summers, with box showing region used for SST index in subsequent figures.(f) Monthly-mean NARR rainfall composite during these summers; (g) Anomalies of vertically integrated moisture flux (vectors) and total precipitable water (shading); (h) evaporation composite anomaly (shading).
(b)), consistent with composites showing southwesterly moisture transport during extreme precipitation summers (figure 1(g)).Over the Gulf of California, this correlation represents an enhancement of the climatological southerly moisture transport (Bordoni and Stevens 2006, Johnson and Delworth 2023).Over the California margin, positive correlations with meridional moisture transport reflect a weakening of northerly and easterly flow that diverges moisture away from coastal southern California (figure 2(b)).

3.
Relationship between SST bias and hydroclimate in observational products compared to CMIP5 and CMIP6 ESMs.(a) Scatterplot of SST bias over 25 • and 32 • N and 125 • and 110 • W on the CA margin (x axis), and the strength of the correlation between SSTs in this region and precipitation in the northern NAM domain (27 • and 37 • N and 115 • and 107 • W-y axis).Subsets of ESMs used in figure 4 outlined in dashed rectangles.(b) Relationship between the strength of precipitation-SST correlation shown in panel (a) (x axis) and the correlation between SLP (20 • and 30 • N and 125 • and 110 • W) and SST over CA margin (y axis); (c) relationship between SST bias on the CA margin (x axis) and the SLP-SST correlation shown in panel (b) (y axis).findings that higher resolution ESMs perform better at simulating the NAM because of their ability to resolve topography and produce realistic statistics of transient disturbances (Pascale et al 2016, Meyer and Jin 2017, Varuolo-Clarke et al 2019).However, higher resolution does not always improve the correlation:

Figure 4 .
Figure 4. Contrasting relationship between large-scale climate fields and SST in ESMs that perform well (a)-(c) and ESMs that perform poorly (d)-(f) at reproducing the SST-precipitation relationship.SST index calculated from the dashed box.(a) and (d) Show climatological SLP (contours) and correlation between SLP and the SST index (shading).(b) and (e) as in panels (a) and (d) but for 850 mb geopotential height.(c) and (f) Show climatological vertically integrated moisture flux (vectors) and correlation between meridional moisture flux and the SST index (shading).