1.1 Sinaloa and its regions

Sinaloa is located in northwest Mexico. Together with the southwestern tip of the state of Chihuahua, the southern edge of Sonora and the eastern slopes of Durango, it is part of the Pacific northwest watershed, as delineated by the Comisión Nacional del Agua (National Water Commission, CNA). Essentially based on elevation, three main physiographic regions make up the state of Sinaloa within the watershed: the Coastal Plains (CP), situated along the Pacific coastline, with increasing width towards the northwest of the state; the low elevations of the Sierra foothills (SF), initially scattered, moving east with increasing altitude; and, finally, the SMO, the easternmost region, which runs almost parallel to the Pacific coasts and where all the state’s rivers originate (Fig. 1).

Fig. 1 Left panel: Sinaloa in northwest Mexico and its physiographic regions. Right panel: location of the climate stations. CP: coastal plains (yellow); SF: Sierra foothills (green); SM: Sierra Madre (blue). 

Given the relevance of the rainfall seasonality patterns for this region and the potential effect of ENSO-driven anomalies, the objectives of this analysis were to: (1) characterize the covariance of the seasonal winter-summer rainfall related variables and geographic factors; (2) quantify the covariance between El Niño/La Niña values in each station, with their respective amounts of winter-summer rainfall; (3) estimate regional differences of the quantified covariance magnitude values of El Niño/La Niña-seasonal rainfall amounts; (4) determine the correlation of the quantified covariance of El Niño/La Niña-seasonal rainfall with geographic elements, and (5) describe the variation of seasonal rainfall-related variables from a multivariate perspective.

^{Caso et al. (2007)} stated that along the Mexican Pacific northwest coast, there is a decrease in mean annual rainfall and summer means decrease, while winter rainfall increases towards the north. Annual and summer rainfall are expected to increase altitudinally by the orographic effect in convective rainfall, as described by ^{Nesbitt et al. (2008)} and ^{Rowe et al. (2008)}. These trends have yet to be quantified for the state of Sinaloa. In addition, we also estimated another related seasonal rainfall variable. Likewise, we incorporated longitude to widen the perspective of spatial analysis.

2.1 Climate data

From the records of the Servicio Meteorológico Nacional (National Weather Service, SMN)-CNA website, based on its entirety, we chose and downloaded daily data from 53 stations (Table I). Data from the selected stations was updated through 2022 by the CNA regional office. From every selected station, we calculated monthly totals for each recorded year. We also considered stations in the states of Chihuahua and Durango that belong to the same regional North Pacific watershed since the rivers that flow into Sinaloa originate from these sites.

We settled the SPP season from May 15 to November, following the guidelines of the CNA that determine the major probabilities of hurricane occurrence on the Mexican Pacific coast for the period. Hence, the WPP season was delimited from December to May 14. We quantified the seasonal rainfall amounts as follows: WPP as the sum of monthly rainfall from December to May 14 and SPP as the sum from May 15 to November. As an example, for any considered station here, the WPP of 1973 was the sum of December 1972 plus the totals of January, February, March, April, and the first 14 days of May of 1973; the 1973 SPP was considered as the total sum from the last 16 days of May to November, therefore, the APP amount of 1973 goes from December 1972 to November 1973.

For the analysis, in addition to APP, SPP, and WPP, we included the percentage of WPP calculated as

%WPP = (WPP/APP) × 100 (1)

We also included the mean number of rainfall events per year (PPE), estimated from daily rainfall data provided by the CNA. Finally, we incorporated the precipitation concentration index (PCI) proposed by ^{Oliver (1980)} for an annual scale proposed, estimated as:

where the numerator is the sum of squared monthly rainfall amounts, and the denominator is the squared total annual rainfall amount. This measure of annual rainfall variation indicates that larger values describe a higher level of rainfall seasonality, e.g., a larger amount of annual rainfall in fewer months.

Monthly Oceanic Niño Index values were obtained from the NWS Climate Prediction Center webpage (^{ONI, n.d.}). These values are estimated as a three-month rolling average temperature difference of sea surface temperature in the tropical eastern Pacific. Values of 0.5 or above point to El Niño conditions; values of -0.5 or below indicate La Niña conditions.

2.2 Data analysis

We explored the covariance of the stations’ APP, SPP, WPP, and %WPP means with latitude, longitude, and altitude by linear regression analysis. We chose these geographic elements due to their accessibility and ease of interpretation.

To elucidate the effect of ENSO variation on seasonal rainfall amounts, we separated monthly ONI data values for El Niño (warm) and La Niña (cold) conditions. Then we performed four non-parametric Spearman correlation (ρ) analyses (P < 0.05) on each selected station: (1) between the ONI monthly data set of El Niño and its corresponding WPP amount; (2) between El Niño and its respective SPP amount; (3) we repeated the process between La Niña and WPP, and (4) La Niña and SPP amounts. We selected Spearman’s ρ correlation since monthly seasonal rainfall data in most of the stations were not normally distributed and/or lacked homogeneity of variance. To assess differences among regions of the quantified magnitude of Spearman’s (ρ) coefficients, we performed a one-way ANOVA (P < 0.05).

To explore the covariance of the quantified Spearman’s ρ products of El Niño/La Niña and seasonal rainfall in terms of geographic variation, a linear regression of their magnitudes was performed with values of latitude, longitude, and altitude. Also, a spatial autocorrelation analysis was carried out using Moran’s I index for the Spearman’s ρ products quantified from the correlations between El Niño/La Niña ONI and WPP-SPP.

Finally, a principal component analysis was performed to achieve a simultaneous perspective of the total variation of the seasonal rainfall variables assessed plus the effect of ENSO on the seasonal rainfall. We used an Oblimin oblique rotation since the considered variables were highly correlated.

3.1 Annual and seasonal rainfall

An initial descriptive analysis of the rainfall-related variables showed that the SM, APP, and SPP means almost double the CP mean. The WPP mean is almost four times larger in the SM as compared to CP (Table II). %WPP is also the largest in the SM, while there were no significant differences between the CP and the SF regions. The number of PPE in the CP is half of the mean value detected in the SM region. The highest seasonality for the rainfall regime was detected for CP.

The regression analysis with the geographic features of the sampled stations showed that APP and SPP had negative correlations with latitude and positive correlations with longitude and altitude. So, their means were smaller towards the north, the west, and higher sites. WPP and %WPP showed higher means in northern and higher sites; in contrast, WPP means were lower in western stations, while %WPP values were higher. PPE correlated negatively with latitude and positively with longitude and altitude, exactly the opposite pattern than PCI (Table III).

3.2 El Niño and La Niña effects on seasonal rainfall

The effect of El Niño conditions on the WPP amounts indicated that positive correlations occurred in all the stations, hence increasing rainfall. Only in Rosa Morada, Huahuapan, and Culiacán-UAS were the results statistically non-significant (Fig. 2a). The magnitude of Spearman’s ρ ranged from 0.463 in Acatitán to 0.150 in Mocorito. Contrastingly, in El Niño conditions with SPP (Fig. 2b), 45 stations produced negative correlations, pointing to a reduction in rainfall. Eight stations had positive correlations (yellow squares in Fig. 2) that were non-significant, most of them in the CP region. Larger intensities were detected in the SF and SM regions.

Fig. 2 Spearman’s ρ correlations analysis between El Niño ONI values and (a) winter rainfall amounts and (b) summer rainfall amounts in 53 stations of the North Pacific watershed of northwestern Mexico. Magnitudes of Spearman’s ρ correlations: blue (■): 0.499 to 0.400; cyan (■): 0.399 to 0.300; green (■): 0.299 to 0.200; olive green (■): 0.199 to 0.100; yellow (■): 0.099 to 0.0; orange (■): -0.001 to -0.100; red (■): -0.101 to -0.200; purple (■): -0.201 to -0.360. 

Spearman&apos;s ρ correlation magnitudes between La Niña ONI values and WPP amounts ranged from 0.200 in Acatitán to -0.115 in Rosa Morada; also, Chinipas, El Carrizo, Mocorito, and Yecorato showed negative signs in such correlations (Fig. 3a). The correlations products between La Niña ONI values and SPP amounts were all positive in the studied stations, hence describing increasing rainfall amounts. The SF and SM regions showed larger magnitudes of these correlation products (Fig. 3b).

Fig. 3 Spearman’s ρ correlations analysis between La Niña ONI values and (a) winter rainfall amounts and (b) summer rainfall amounts in 53 stations of the North Pacific watershed of northwestern Mexico. Magnitudes of Spearman’s ρ correlations: blue (■): 0.499 to 0.400; cyan (■): 0.399 to 0.300; green (■): 0.299 to 0.200; olive green (■): 0.199 to 0.100; yellow (■): 0.099 to 0.0; orange (■): -0.001 to -0.100; red (■): -0.101 to -0.200; purple (■): -0.201 to -0.360. 

3.3 Regional differences of Spearman’s ρ magnitude

We found a normal distribution of Spearman’s ρ magnitude for each analysis, and hence, we performed a one-way ANOVA between regions. Overall, the largest mean value was found in El Niño-WPP, and the minimum mean was detected in La Niña-WPP; however, both winter interactions were not regionally differentiated. On the other hand, summer rainfall was differentiated in the direction of the mean correlations; they were negative in El Niño (reducing rainfall) and positive during La Niña (increasing rainfall). Both summer season rainfall correlations with El Niño/La Niña were different between regions, where the SM magnitudes were always larger (Table IV).

3.4 Geographic and spatial covariation of ENSO rainfall amounts

Latitude correlates significantly with Spearman’s ρ coefficients of El Niño-WPP (Fig. 4a) and La Niña-SPP (Fig. 4b), with smaller coefficients toward the north in both cases.

Fig. 4 Regression analysis of 53 stations within three physiographic regions in northwestern Mexico. (a) latitude and El Niño-winter rainfall (R² = 0.182, P < 0.001, slope = -0.028); (b) latitude and La Niña-winter rainfall (R² = 0.219, P < 0.000, slope = -0.033). Red captions: coastal plains region; green captions: Sierra foothills region; blue captions: Sierra Madre region. 

Longitude generated three significant regressions with Spearman’s ρ values: El Niño-WPP, El Niño-SPP, and La Niña-WPP (Fig. 5).

Fig. 5 Regression analysis of 53 stations within three physiographic regions in northwestern Mexico. (a) longitude and El Niño-winter rainfall (R² = 0.129, P < 0.008, slope = 0.023); (b) longitude and El Niño-summer rainfall (R² = 0.077, P < 0.044, slope = -0.033); (c) longitude and La Niña-winter rainfall (R² = 0.222, P < 0.000, slope = 0.032). Red captions: coastal plains region; green captions: Sierra foothills region; blue captions: Sierra Madre region. 

Altitude produced two significant regressions with Spearman’s ρ values: El Niño-SPP correlated negatively with altitude (Fig. 6a), and La Niña-SPP coefficients correlated positively with altitude (Fig. 6b).

Fig. 6 Regression analysis of 53 stations within three physiographic regions in northwestern Mexico. (a) altitude and El Niño-summer rainfall (R² = 0.220, P < 0.000, slope = 4.03^-5); (b) altitude and La Niña-summer rainfall (R² = 0.244, P < 0.000, slope = 2.59^-5). Red captions: coastal plains region; green captions: Sierra foothills region; blue captions: Sierra Madre region. 

Two spatial Moran’s I analyses were significant: El Niño-SPP: I = 0.227, P < 0.001, showing a clustered aggregation, and La Niña-SPP, which also developed a clustered aggregation: I = 0.147, P < 0.028.

3.5 Multivariate analysis

From a multivariate perspective, a principal component analysis (Table V) showed that the three first components explained over 80% of the total variation. PC1, which depicted 51%, was correlated positively with WPP, the number of rainfall events, the percentage of WPP, APP, and the covariance of La Niña-SPP and SPP. The largest negative loadings of PC1 were the PCI and El Niño-SPP. PC2 illustrated 17% of the total variation; here, the main positive loadings were SPP, the covariance of La Niña-WPP, El Niño-WPP, APP, and a number of rainfall events; the largest negative loading was the PCI. The main positive loadings in PC3 were the covariances of El Niño-SPP and El Niño-WPP plus the PCI; the largest negative loadings were SPP, APP, the number of rainfall events, and the La Niña-SPP relationship.

A graphic analysis of the z-standardized scores for PC1 and PC2 (Fig. 7) calculated from the multivariate analysis, depicted that on the negative sector of PC1, the variability of all CP stations plus Acatitán, Badiraguato, El Comedero, El Fuerte, El Varejonal, Ixpalino, Jaina, Las Tortugas, Mocorito, Rosa Morada, San Joaquín, Sanalona, Santa Cruz, and Tamazula from the SF region, was determined by high values of PCI and the correlation of El Niño-SPP. In the positive sector of PC2, all the stations of the SM region besides Huites, San Dimas, El Real, Choix, Chinipas, Guaténipa, and La Huerta from the SF region were linked to high means of APP, SPP, WPP, %WPP, number of rainfall events, and the correlation of La Niña-SPP.

Fig. 7 Scatter plot of z-scores of principal component analysis for rainfall variables and the effect of El Niño and La Niña conditions on seasonal rainfall of 53 stations from three regions of the North Pacific watershed in northwestern Mexico. Red captions: coastal plains region; green captions: Sierra foothills region; blue captions: Sierra Madre region. 

4.1 Rainfall and geography

In the scenario described by ^{Caso et al. (2007)}, the state of Sinaloa is located at the northern end of a summer-rainfall regime along the transition summer-to-winter rainfall dominance on the Pacific coast of Mexico. Here, regional differences within the Pacific Northwest watershed for all the rainfall-related variables indicated that rainfall seasonal patterns are differentiated. These differences were correlated to geographical variations: the APP mean values decrease heading north and west; APP and SPP clearly decrease into the lower altitudes of the CP, an effect described by ^{Houze et al. (2012)}, and WPP and %WPP increase toward the north and in upper stations. In addition, in the northwest and lowlands, there is a decreasing gradient of PPE and an increasing trend in the values of %WPP and rainfall seasonality (PCI).

4.2 Effects of El Niño and La Niña on seasonal rainfall and regional differences

Regarding the ENSO covariance with seasonal rainfall in Sinaloa during winter, El Niño increased rainfall amounts in each studied station as described by ^{Magaña et al. (2003)}, ^{Caso et al. (2007)}, and ^{Bravo-Cabrera et al. (2010)}. The magnitude of El Niño’s effect in each region is similar, with over 30% of explained covariance. On the contrary, in summer, with El Niño warm-ocean conditions, rainfall in Sinaloa developed a similar pattern as in the southern regions of the Mexican Pacific coasts, with 85% of the studied stations recording decreasing rainfall amount trends. Among regions, the SF and the SM showed the largest negative mean correlations, pointing that in these conditions, El Niño may decrease water availability during the summer, thus disrupting or delaying ecosystemic processes.

In reference to La Niña, our results also coincide with those of ^{Magaña et al. (2003)}, ^{Caso et al. (2007)} and ^{Bravo-Cabrera et al. (2010)} for the WPP and SPP regimes, for which they depicted slightly increasing trends in both seasons. The effect of La Niña and WPP in Sinaloa showed mostly increasing rainfall trends. The magnitude of the overall Spearman’s ρ in each of the analyzed regions was the lowest, in accordance with the analysis of ^{Zermeño-Díaz and Gómez-Mendoza (2023)}, who depicted a weak signal of the La Niña-WPP for northwestern Mexico. As for the covariation of La Niña and the SPP, all the analyzed stations in Sinaloa developed increasing trends. Here, the detected differences among regions showed a larger effect on the SM and SM foothills regions. This scenario could potentially mitigate dryness in the seasonal regions by increasing rainfall and/or advancing the start of the rainy season.

A relevant feature of the spatial distribution of the covariances of ENSO variation on the seasonal rainfall amounts is its homogeneity. In the four performed analyses, we did not detect relevant spatial discrepancies in the directions of correlations in nearby stations. These results suggest that spatial physical heterogeneity does affect local rainfall regimes under ENSO variation, although further research needs to be done in this area.

4.3 Geographic and spatial covariance

Although El Niño and La Niña physical conditions are opposed, their Spearman’s ρ correlation products with WPP amounts correlated in the same manner with latitude. These two results, together with those described by regression between longitude-El Niño-WPP and longitude-La Niña-WPP, indicate that rainfall amounts in the southern and eastern stations of the studied area are prone to increase during winter in both extremes of the ENSO variation. In contrast, as depicted by the regression between longitude and El Niño-SPP, southeastern stations are more sensitive to Pacific warmer water conditions; in this case, although these stations possess larger SPP means in such scenario, rainfall below its average would enhance lack of water availability and so dry conditions during summer due to frequent elevated temperatures in such season.

Regarding altitude, summer rainfall regressions generated two contrasting directions. During El Niño conditions, a few CP stations plus Surutato from the SM region tend to develop a minor increase in rainfall amounts. This condition could slightly mitigate dry conditions in that region, but the SF and the SM stations developed negative trends. Hence, in such regions is evident that El Niño decreases water availability in summer. On the contrary, during La Niña in summer, as altitude increases across regions, the covariation of La Niña and rainfall amounts increases accordingly, hence ameliorating dry conditions due to increased rainfall conditions in all stations.

4.4 Multivariate analysis

From a multivariate perspective, the relevance of ENSO variations in seasonal rainfall amounts seems rather scarce, given the explained variance of the principal components. In PC1, which explains half of the total variance, most of the variation is illustrated by the weights of APP, SPP, WPP, %WPP, number of rainfall events, and PCI. Nevertheless, correlations of El Niño-SPP and La Niña-SPP found in PC1 are important. El Niño-SPP correlation and PCI mainly determine the multivariate responses of CP and 14 stations of the SF, indicating that summer rainfall would increase in this area if El Niño is present. On the other hand, if La Niña is present in summer, rainfall amounts would increase in all SM and eight stations of the SF.

For PC2, El Niño-WPP and La Niña-WPP are relevant in 10 of the 12 stations of the SM region, southern CP stations, and 10 SF stations. This depicts that in winter, regardless of the contrasting El Niño and La Niña conditions, rainfall amounts could increase in the aforementioned stations.