The influence of ENSO during spring over northwestern Mexico

This study focuses on better understanding how El Niño‐Southern Oscillation (ENSO) influences the occurrence of particularly dry conditions (warm temperature and dry precipitation anomalies) in northwestern Mexico during boreal spring (March–May), a season that is characterized by the driest regional conditions, and by a high socioeconomic sensitivity to the interannual variations in them. For that, surface observations (CRU TS4.05 and ERSSTv5) and low‐tropospheric circulation reanalysis (ERA5) were analysed using linear regressions, contingency tables, and composites over two periods, 1901–2020 and 1959–2020. The results confirm that the positive phase of ENSO (El Niño) tends to reduce the probabilities of a drier than normal spring and that its negative phase (La Niña) increases them, but suggest that such signal is relatively weak. In particular, the drying effects associated to La Niña are generally small, as they explain less than 10% of the interannual variability in precipitation, and most of the driest springs occurred during neutral years. The results also shed light into the causes of the driest springs: internal atmospheric variability. On average, during the driest springs, sea surface temperature anomalies are statistically non existent over the eastern Pacific, but these events are characterized by an expansion of the southeastern (SE) side of North Pacific Subtropical High (NPSH). It is found that during spring, a strengthened SE side of the NPSH weakens the moisture flux from the Pacific Ocean into northwestern Mexico. These results suggest that the seasonal predictability of spring for northwestern Mexico might not only depend on the strength and phase of ENSO but also on the processes controlling the variability of ridges over the extratropical eastern Pacific.

Northwestern Mexico has an arid and semiarid climate (García, 2004) that is prone to droughts (Méndez & Magaña, 2010;Seager et al., 2009).In general, the socioeconomic system of this region is sensitive to variations in water availability (Aguilar-Barajas et al., 2016 and references therein), as they can result in depleted underground water levels, increased incidence of gastrointestinal diseases, and disputes for water between sectors (Galv an & Magaña, 2020;Magaña et al., 2018;Magaña & Conde, 2000).Regional water deficits might disrupt the industrial productivity, increase the social unrest, reduce the general well-being of the population, and might ultimately increase migration from Mexico to the United States (Del-Toro-Guerrero & Kretzschmar, 2020;Feng et al., 2010;Fishman & Li, 2022;Hunter et al., 2013;Nawrotzki et al., 2013).From these threats, it emerges a worrisome prospect for northwestern Mexico, given its history of slow implementation of drought vulnerability-reducing strategies in the region (Aguilar-Barajas et al., 2016), and its pronounced warming trends (Cavazos et al., 2020;Gutiérrez-Ruacho et al., 2010;IPCC, 2021;Magaña et al., 2012;Maloney et al., 2014;Seager et al., 2013), as even in the absence of changes in precipitation, warmer temperatures would increase dryness by increasing evapotranspiration rates.The northwestern domain defined in this study includes different climatic subregimes (Figure 1a), particularly along the Peninsula of Baja California (Figure 1b,  c).While the precipitation regime in the non-Peninsular portion is dominated during summer by the influence of the North American Monsoon (Figure 1d, e) (e.g., Adams & Comrie, 1997), the precipitation regime in the Peninsula of Baja California is dominated during summer and fall by the landfall of recurving eastern Pacific tropical cyclones (particularly in the southern tip of the Peninsula), and during winter and early spring by the rainfall from cold fronts and atmospheric rivers (Inda-Díaz & O'Brien, 2023;Magaña et al., 2003).Despite the different phenomena that produce rain in the whole northwestern domain, its climatic variability coincides in that their driest months tend to be around boreal spring (Figure 1; March-April-May).
From a water-deficit perspective, it might not come as a surprise that boreal spring is likely to be the most important season in northwestern Mexico (Figures 1, 2a).Spring marks the end of the monzonic dry spell with rapidly increasing temperatures; the scarce amounts of humidity during this season help in preventing soil erosion, reduce the risk of wildfires, and nurture incipient crops.If crops during spring fail, less vulnerable farmers (those with irrigation infraestructure) might lose part of their investment, but the most vulnerable ones-rainfed farmers-might lose their entire livelihoods or end up in debt (Appendini & Liverman, 1994;Bocco et al., 2021;Dobler-Morales & Bocco, 2021;Eakin et al., 2018).
Despite the high risk that characterize the dryness during spring season, relatively few studies have investigated exclusively the climate variability of this season, most have focused on winter and/or summer, or have grouped winter and spring into the 'dry' season (Cavazos & Hastenrath, 1990;Ropelewski and Halpert 1996;Pavia et al., 2006;Bhattacharya & Chiang, 2014;Bravo-Cabrera et al., 2017;Fuentes-Franco et al., 2018).Another issue that adds complexity to the understanding of the climate variability of this season is its low forecasting skill, a characteristic known as the spring predictability barrier (Webster, 1995).Combined, these issues result in a real challenge for seasonal forecasters to address, with a relatively high degree of confidence, the question of 'how dry next spring might be?'But there have been encouraging advances suggesting that an adequate spring forecast is possible since forecasts of El Niño-Southern Oscillation (ENSO) for this season have become more reliable (Larson & Kirtman, 2017;Webster & Hoyos, 2010); ENSO is the main source of seasonal predictability (e.g., Jin et al., 2009).
The positive phase of ENSO, El Niño, tends to increase precipitation and reduce temperature during spring in northern Mexico, but its negative phase, La Niña, has the opposite effects, it reduces precipitation and increases temperature (Bhattacharya & Chiang, 2014;Bravo-Cabrera et al., 2017;Cavazos & Hastenrath, 1990;Englehart & Douglas, 2002;Magaña & O., 1999;Pavia et al., 2006;Zolotokrylin et al., 2016).This relation has the potential to be very important for seasonal forecasters, as a potentially robust indication of how dry next spring might be.In this study, we aim for a better understanding of this relation.As explained next, such objective is motivated by three main issues: This relation has been evaluated mostly from a deterministic perspective; it is unclear actually how useful might be to anticipate particularly dry spring seasons; and it is unclear how it might be affected by the prevailing state of the Pacific Ocean.
The importance of the latter issues can be exemplified with a particular case, that of 1998.Spring of 1998 was characterized by El Niño conditions (e.g., Changnon, 2000), but in northwestern Mexico, instead of the expected relatively wet conditions associated with El Niño, the region experienced a particularly dry spring (Figure 2b).This exception might not be surprising for two main reasons.Only a fraction of the interannual variability in the hydroclimate variability of the region might be attributable to ENSO; how large is that fraction for northwestern Mexico during spring?
The answer is unclear.The relation between ENSO and its impacts should be more properly understood not from a deterministic approach but from an approach based on probabilities (e.g., Mason & Goddard, 2001), and at a regional scale.
That northwestern Mexico experienced the opposite conditions to those attributable to El Niño in spring of 1998 might be more of a surprise considering that this season was also characterized by the positive phase of the Pacific Decadal Oscillation (PDO; https://www.ncdc.noaa.gov/teleconnections/pdo); the combination of both positive phases tend to result in a magnification of the expected ENSO patterns (e.g., Wang, Huang, et al., 2014).Although there is a consensus that both of these modes are the modes that most affect the climate of the subtropical, north American sector (Lopez & Kirtman, 2019; see also Zhao et al., 2017), such consensus does not extend to all aspects of the ENSO-PDO relations (McCabe and Dettinger 1999;Gershunov & Barnett, 1998;Englehart & Douglas, 2002;Hu et al., 2011;Kam et al., 2014;Wang, Huang, et al., 2014; see also Pavia et al., 2006Pavia et al., , 2016)).More particularly, for the domain of northwestern Mexico, which is mostly in the subtropics (Figure 1), it is unclear if the combination of the same phase in both ENSO and PDO modes would result in a magnification of the expected ENSO impacts (e.g., Wang, Huang, et al., 2014), or maybe not since tropical surface temperature (SST) anomalies play a less significant role than internal atmospheric variability over subtropical and extratropical regions, and since an extreme SST forcing in the tropics may not result in magnified impacts outside of the latter regions (Kumar et al., 2013).Consistently, previous studies have also suggested that not only SST anomalies but also internal atmospheric variability is a large contributor to the causes associated to prolonged periods of dryness in the southwestern US and Mexico sector (Lin et al., 2017;Linkin & Nigam, 2008;Seager et al., 2014;Seager & Hoerling, 2014;Swain et al., 2016;Wang, Hipps, et al., 2014;Wang & Schubert, 2014;Zhao et al., 2017).In fact, atmospheric variability seems to be the main driver of extreme, episodic droughts, as those of 2011-2012 and 2013-2014 in southwestern US and parts of northern Mexico (Seager et al., 2014;Seager et al., 2015;Swain et al., 2014).These episodes are caused by ridges off the coast of California that are strengthened during La Niña, but they might also appear in the absence of a clear SST forcing (Seager et al., 2015; see also Lin et al., 2017).If these high-pressure systems are also behind particularly dry spring seasons over northwestern Mexico, such as that of 1998, is unclear.
To better understand the influence of ENSO in northwestern Mexico during spring, we applied an observational approach based on diagnostics of composites, linear regressions, and contingency table maps.More details about the used datasets and applied methods are provided in Sections 2 and 3 respectively.The results are presented and discussed in Section 4; after that, Section 5 provides final considerations.

| METHODS
Prior to all diagnostics, linear trends were removed from all datasets.Then, we applied a three-pronged methodological approach.First, we estimated the influence of ENSO on the seasonal precipitation and temperature data from a classic linear approach (Wilks, 2011), by regressing these data onto ONI 3.4 .Second, we estimated the influence of ENSO from a probabilistic approach, following the contingency table analysis of Mason andGoddard (2001, see also Sardeshmukh et al., 2000).With this analysis, we evaluated the occurrence probabilities of upper and lower quartile values in precipitation and temperature at the same time that ONI 3.4 was larger (smaller) than 0.5 C (−0.5 C).We applied this analysis to all data (1901-2020 period) and to two subgroups separately: one group, with 50 years of data, was characterized by the positive phase of the PDO; the other, with 70 years, was characterized by the negative phase of the PDO.
Third, aiming not to include information about the state of ENSO or the PDO in the analysis from the start, we sought SST and atmospheric circulation (Z, u, and v) features associated to the driest and wettest springs using composites.The driest and the wettest springs were identified from the quartiles of a time series of potential water deficit (i.e., the difference between precipitation and potential evapotranspiration), a time series that resulted from the spatial average over the north and northwest domain (Figure 1b).In the lower (upper) quartile of this time series, we identified the years with the driest (coolest) springs.The driest (wettest) years are those in which potential evapotransporation is larger (smaller) than precipitation.As shown in Figure 3, the driest springs (upper quartile composites) received close to 20 mm months −1 less precipitation than the relatively wet springs, an amount that is close to the average of that season (Figure 3a, b).The driest springs are characterized by below-average precipitation over the northern and central Mexico, and by above-normal temperatures over the central northwestern Mexico (Figure 3c, d).Instead of using a water-deficit time series, we also used time series of the related variables, temperature, precipitation and potential evaporation; the results were similar, but those based on the water-deficit time series were more robust.
Our analyzes were constrained by the temporal coverage of the data.We used the 1901-2020 period (that covered by CRU, ERSST, and the ENSO and PDO indexes) in all diagnostics except those involving low-tropospheric circulation, in which we used the 1959-2020 period (that covered by ERA5).Statistical significance in the regression and composite analyzes was tested using the 95% confidence level, based on a Student's t-distribution test.
In the contingency-table analysis, the same confidence level was applied but we used a hypergeometric distribution (Mason & Goddard, 2001).

| Explained variance
During spring, cold (warm) temperature anomalies and wet (dry) precipitation anomalies are characteristic of the influence of El Niño (La Niña) in northern Mexico (Figure 4a, b) (Section 1).The area-mean attributable variance of this relation in the interannual variability of precipitation is 9.86%, and that of temperature is 15.65% (Figure 4c, d).The influence of ENSO during spring is also characterized by negative precipitation anomalies in southcentral Mexico (Figure 4a), a signal that has an attributable variance of the interannual variability that ranges between 5% and 10%.From these relatively small amounts of attributable variance, it might not come as a surprise that even strong El Niño events have had unexpected outcomes during spring, as in 1998.Also, these results confirm the need of approaching the problem not from a deterministic perspective, but from a probabilistic one.

| Probabilities
Figure 5 presents the probabilities associated with ENSO during spring.Given the occurrence of a La Niña event, northwestern Mexico has a probability higher than 60% to experience an spring season that is in the upper quartile of temperature and in the bottom quartile of precipitation; during El Niño, northern (south-central) Mexico has probabilities higher than 60% of having spring conditions in the upper (lower) quartile in precipitation.As shown in Figure 5, both phases of ENSO were associated with the occurrence of opposite but symmetric precipitation anomalies: there are relatively high probabilities of wet (dry) precipitation anomalies in northwestern Mexico during El Niño (La Niña), and high probabilities of cool (warm) temperature anomalies over the central sector of northern Mexico (over the state of Chihuahua).But Figure 5 also shows asymmetric features.For example, in precipitation, El Niño is associated with high probabilities of relatively wet anomalies in northern Mexico and dry anomalies in south-central Mexico, but La Niña is associated with high probabilities of dry anomalies but only in northwestern Mexico.In temperature, El Niño is associated with high probabilities of relatively cool anomalies in the central and eastern sectors of northern Mexico, but La Niña is associated with high probabilities of relatively cool anomalies in south Mexico, and particularly warm anomalies in northwestern Mexico.

| The influence of the Pacific Decadal Oscillation
The influence of the PDO in the precipitation patterns associated with ENSO is larger over northern Mexico than over central Mexico (Figure 6).Consistent with the previous analysis (Figure 5), also Figures 5 and 6 shows that in temperature, neither El Niño nor La Niña resulted in large, significant probabilities over central Mexico.On the other hand, over northern Mexico, the drying influence associated to La Niña are strengthened during the negative phase of PDO (cf., Figures 5b, 6d), but the moistening influence associated to El Niño is weakened (cf., Figures 5a, 6c).Contrastingly, during the positive phase of the PDO, the influence of ENSO tends to be diminished: the moistening influence associated to El Niño is slightly weakened (cf., Figures 5a, 6a), and the drying influence associated to La Niña vanishes (cf., Figures 5b,  6d) (see also Zhao et al., 2017).A particular example of this result might be 1998, in which the large El Niño of this year occurred under a positive PDO phase, but instead of the expected positive precipitation anomalies, negative anomalies occurred during spring (Figure 2b; Section 1).While this result tends to contrast with the idea that equal phases of ENSO and PDO would result in amplified canonical patterns, regardless of their latitude (e.g., Wang, Huang, et al., 2014), it is more consistent with the idea that an extreme SST forcing would not result in an extreme response in extratropical regions (Kumar et al., 2013).
Although the influence of ENSO in temperature tends to be reduced during the negative phase of the PDO, Figure 6 suggests that temperature tends to be less affected than precipitation by the PDO.It is unclear why precipitation in northern Mexico might be more susceptible than temperature to the influence of the PDO, but as explained later, one possible cause is that precipitation during spring is more likely to occur in intense pulse-like events, such as atmospheric rivers (Inda-Díaz & O'Brien, 2023), pulses that might not have a large effect on the seasonal-mean temperature, but that do have a large effect on the typically small amounts of accumulated rainfall during this season.Also, it is worth noting that Figure 6 suggests that the precipitation over the Yucatan peninsula is sensitive to the PDO.In particular, over this region, the influence of El Niño can be associated with opposite spring precipitation anomalies depending on the phases of the PDO: positive anomalies during the negative phase of the PDO, but negative anomalies during the positive phase of the PDO (Figure 6a, c).Although we ignore the causes behind the signals over the latter region, might be associated with the effects of ENSO and PDO on the formation and the trajectories of the Caribbean easterly waves.Clearly, this issue, the influence of ENSO and PDO over southeastern Mexico during spring, needs to be addressed in future work.

| Are the driest springs La Niña springs?
Most of the driest springs in northwestern Mexico have occurred during neutral ENSO conditions (Figure 7).Although La Niña is associated with relatively dry conditions in northwestern Mexico, Figure 7 shows that nearly two-thirds (63.3%) of the driest springs in this region occurred under neutral conditions, and only 26.7% of these springs occurred during La Niña conditions; and, although nearly 40% of the wettest springs occurred during El Niño, most of the wettest springs occurred also under neutral conditions, suggesting that the occurrence of particularly dry or wet spring events in northwestern Mexico might be largely influenced by factors other than ENSO.Seeking for clues of such factors, we analysed composites of SST and lowtropospheric circulation during these events in the following subsections.

| Sea surface temperature
The SST composites are not symmetric during the driest and the wettest springs.On one hand, during the driest springs, the SST anomaly pattern resembles that of La Niña, but there are no statistically significant cold anomalies in the central and eastern Pacific, and around northwestern Mexico (Figure 8a).It is a pattern in which the only characteristic SST anomalies are the warm SST  From a regional perspective, wet events can be associated to warm SST anomalies in the vicinity of northwestern Mexico: the regional warm SST anomalies in the Pacific would increase evaporation and moisture flux from the east Pacific Ocean into northwestern Mexico, conditions that would facilitate cloud formation and reduced incoming solar radiation.Even more, this combination of anomalies might be enhanced by a relatively weak anticyclonic circulation of the North Pacific Subtropical High (NPSH) (Figure 8b): warm SST anomalies are correlated with an increased amount of stratocumulus clouds in the northeastern Pacific sector, and with a weakening in the circulation of the NPSH (Norris et al., 1998;Yun et al., 2015).This anomaly in the regional atmospheric circulation might also reduce advection of cold water, further leading to enhanced regional warming (Amaya et al., 2020).However, dry precipitation anomalies in northwestern Mexico cannot be associated to cold SST anomalies using symmetrically opposite arguments to those associated to warm SST anomalies since the cold SST anomalies around northwestern Mexico are too weak (Figure 8a).As shown next, during the dry events, atmospheric processes might be more important.

| Low-tropospheric circulation
As shown in Figure 9, the only statistically significant atmospheric feature that characterizes the dry events is an expansion of the the southeastward side of the NPSH.The southeastward expansion of the NPSH promotes a low-tropospheric oceanward circulation by means of its anticyclonic circulation, weakening the humidity flux from the Pacific into northwestern Mexico (Figures 10a,  c).However, this drying associated to the NPSH expansion, might not be similar to the drying that the North Atlantic Subtropical High (NASH) cause during summer over southern Mexico and Central America.(During summer, the NASH expands westward into the Caribbean, bringing dry air from the western Atlantic, and discouraging convection over South Mexico and Central America; e.g., Kelly & Mapes, 2011).During spring, the drying associated to the NPSH expansion is associated with the northward deflection of the Pacific-North American storm track (Seager et al., 2014).This deflection might not only reduce the already scarce amounts of Consistent with previous studies (e.g., Kam et al., 2014), Figure 8-10 show no evident influence from the Atlantic ocean sector, only from the Pacific, suggesting that the Pacific Ocean sector hosts the drivers of the driest or wettest springs in northwestern Mexico.We found consistent results by analysing the circulation composites at other low-tropospheric pressure levels (not shown).
Finally, we investigated if a strengthened SE NPSH was an attributable cause of the anomalously dry spring during El Niño of 1998 over northwestern Mexico.As shown in Figure 11, the results are partially consistent with this idea: The SE expansion of the NPSH was found at levels close to the surface, but the horizontal wind pattern at these levels shows that a more likely cause was an enhanced southward flow; this southward flow weakens the moisture input towards northwestern Mexico.While the latter wind pattern is not a typical feature of dry events (Figure 10a), it is a common feature of El Niño events: it is part of their triggered Gill type of response (e.g., Fang & Yu, 2020).This result is consistent with the idea that some El Niño events tend to cause dry springs in northwestern Mexico (Figure 7), despite it being typically associated with wet conditions (Figures 4a, b).But it raises the question of why not the majority of El Niño events tend to cause dry conditions during spring by the southward flow it promotes (Figure 11b)?After all, dry conditions are the expected outcome associated with El Niño during summer over most parts of Mexico (Bhattacharya & Chiang, 2014;Turrent & Cavazos, 2009).Besides the different influence of ENSO on the phenomena that produce rainfall during spring and summer, the answer to this apparent discrepancy during spring might lie in the relative role of the SE expansion of the NPSH against that of SST anomalies.Apparently, in the 1998 case, the role of the warm tropical El Niño SST anomalies was the dominant one, but in other cases, the opposite might occur (subsection 4.5).It would be desired that future work on this problem focus on identifying the root causes that might determine their relative contributions.

| FINAL CONSIDERATIONS
In northwestern Mexico, boreal spring is a particularly important season.Knowing if its going to be extremely dry, or simply average dry, can make a difference for the society and the biodiversity of the region.In order to improve our planning strategies, it is needed a more detailed understanding of ENSO, or other sources of predictability, on the temperature and precipitation at a regional scale.Particularly this study focuses on better understanding the role of ENSO.A question to summarize the latter need is the following: is an adequate seasonal forecast of ENSO enough to anticipate if next spring is going to be drier than normal?Given the relatively small amount of explained variance from a linear perspective, this question has to be addressed from a probabilistic one.Our results confirm that El Niño tend to reduce the probability of having a dry spring, and La Niña increases them, although most of the driest springs have occurred under neutral conditions.However, during the cold phase of the PDO, La Niña is associated with slightly larger probabilities of dry precipitation anomalies.In future work, it might be interesting to separate the regional spring probabilities associated with central Pacific and eastern Pacific ENSO events (e.g., Kirtman, 2019).
An interesting result is that particularly dry springs might be more unpredictable than 'wet' springs, since the dry ones occurred under a weak La Niña SST pattern, with no significant SST in the northeastern Pacific sector.On average, the key difference between these two events seems to be in the expansion (or contraction) of the southeastern edge of the NPSH.Such a feature, however, cannot explain all dry spring events; an example is that of 1998, in which the warm SST anomalies might have had a larger role than that of a rather weak expansion of the SE side of the NPSH.We sought other seasons in which the NPSH expansion could have been associated with a particular pattern in the temperature or precipitation in Mexico but found none.From these results is not surprising that the NPSH has been commonly overlooked as a possible driver of the Mexican hydroclimate variability (see also Grotjahn & Osman, 2007).
The results of this study highlight the need to better understand the relative roles of oceanic and internal atmospheric processes in driving enhanced dryness over the North American sector (Lin et al., 2017;Lopez & Kirtman, 2019;Seager et al., 2014;Seager et al., 2015;Wang, Hipps, et al., 2014), a need that might be particularly important under global warming since changes in the strength and evolution of the NPSH are expected (He et al., 2017;Schmidt & Grise, 2019;Song et al., 2018aSong et al., , 2018b)).At a regional scale, future work might also focus on better understanding the role of atmospheric modes that have been identified to modulate the climate variability in northwestern Mexico and the southwestern US, such as the Pacific North American pattern (PNA) and North Pacific Oscillation (NPO) (Lin et al., 2017;Magaña et al., 2003;Seager et al., 2015).For that, it would be interesting to apply statistical methods more sophisticated than those applied in this study, perhaps methods those to identify and quantify the contribution of different modes on drought severity and occurrence (e.g., Costa et al., 2021;Lyra et al., 2017;Oliveira-Júnior et al., 2018).

F
I G U R E 1 Monthly mean evolution of precipitation (bars) and temperature (solid line) from CRU (1901-2020) over (a) the northwestern domain, as defined in this study, and over four subdomains, focusing on (b) Baja California Norte, (c) the southern tip of Baja California Sur, (d) Sonora, and (e) over the eastern part of the northwestern domain.The area included in these regions is shown over (f) an elevation map.[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 2 Average of spring precipitation from CRU (a) for the1901-2020 period, and (b) for 1998.The dashed box encloses the northwestern domain used in this study.[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 3 Composites of (a), (b) total precipitation, and composites of (c), (d) precipitation and (e), (f) temperature anomalies; they are based in the (a), (c), and (e) lower and (b), (d), and (f) upper quartiles of the time series (1901-2020) northwestern-domain water deficit.Hatched regions mark statistical significance in total values; in anomalies, only significant ones are shown (95% confidence level).[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 4 Spring (a) and (c) precipitation and (b) and (d) temperature regressed onto ONI 3.4 (1901-2020), and their (c), (d) explained variance.The units of the regression maps are Zscores.Only statistically significant results are shown (95% confidence level).[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 5 Probabilities of spring (a), (b) precipitation and (c), (d) temperature to reach upper or bottom quartile values during El Niño and La Niña events from 1901 to 2020 data.Black squares enclose statistically significant probabilities (95% confidence level).[Colour figure can be viewed at wileyonlinelibrary.com]anomalies over the northwestern Pacific.On the other hand, the SST anomaly associated with the wet spring events is that of a weak El Niño (Figure8b).It is an SST pattern with warm SST anomalies over the central and eastern Pacific, and around northwestern Mexico.As shown in Figure8, these patterns are present during the analysed period of surface observations, and during the shorter period in which we analyse the lowtropospheric circulation fields from ERA5 (1959-2020; see Sections 2 and 3).

F
I G U R E 6 Probabilities of spring (a) and (d) precipitation and (e) and (h) temperature to reach upper or bottom quartile values during El Niño and La Niña events, during prevailing (a), (b), (e), and (f) positive and (b), (d), (g), and (h) negative PDO conditions from 1901 to 2020 data.Black squares enclose statistically significant probabilities (95% confidence level).[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 7 Distribution into El Niño, La Niña, and Neutral categories of the cases forming the lower (driest springs) and upper (wettest springs) quartiles in the time series (1901-2020) of northwestern-domain water deficit.[Colour figure can be viewed at wileyonlinelibrary.com]F I G U R E 8 Spring composites of SST anomalies based on the (a) and (c) lower and (b) and (d) upper quartiles of the northwestern-domain waterdeficit time series for two periods (a), (b) 1901-2020 using ERSST and (c), (d) 1959-2020 using ERA5 data.Hatched regions mark statistical significance (95% confidence level).[Colour figure can be viewed at wileyonlinelibrary.com] rainfall produced by the invasion of mid-latitude fronts during that season but more importantly, it might reduce the rainfall produced by atmospheric rivers (Inda-Díaz & O'Brien, 2023).The opposite occurs during the wettest events: the southeastward side of the NPSH is contracted, resulting in a stronger moisture flux input from the Pacific Ocean into northwestern Mexico (Figure 10b, d).

F
I G U R E 9 Spring composites of 850-hPa geopotential height (shaded and contours; gpm) and horizontal wind (vectors; ms −1 ) based on the (a) lower and (b) upper quartiles of the time series (1959-2020) of northwestern-domain water deficit.Hatched regions mark statistical significance in geopotential height (95% confidence level).[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 1 0 Spring composites of 850-hPa horizontal wind (vectors), (a), (b) geopotential height (read and blue colours), and (c), (d) precipitation anomalies based on the (a) lower and (b) upper quartiles of the time series (1959-2020) of northwestern-domain water deficit.All variables are from ERA5, and their units are Z-scores.Only statistically significant (95% confidence level) anomalies in geopotential height and precipitation are shown; statistically significant anomalies in wind are shown as dark vectors.[Colour figure can be viewed at wileyonlinelibrary.com]F I G U R E 1 1 Anomalies in 925-hPa horizontal wind (vectors) on top of anomalies in (a) geopotential height at that pressure level, and (b) SST during spring of 1998.Only values over oceanic regions are shown.Both variables are from ERA5, and their units are Z-scores.[Colour figure can be viewed at wileyonlinelibrary.com]