Introduction

The Baja California Peninsula in Mexico is a physiographic region of scientific interest for its floristic richness and its geographical position, in which the marine and aquatic terrestrial ecosystems are influenced by two bodies ofwater, the Pacific Ocean and the Gulf of California. The National Commission for the Knowledge and Use of Biodiversity in Mexico (CONABIO) described it as a narrow strip of land that runs parallel to the Pacific coast along more than 1 400 km, with an average width of 75 km. The east coast of the peninsula is bordered by the warm waters of the Gulf of California and the west coast by the cold waters of the California Current (^{Rodríguez-Moreno and Bullock (2014)}.

The California Current is an ocean flow that originates in the North Pacific and runs along the western United States Coast, moving south to the southern part of the Baja California Peninsula. When reaching this point, turns west towards the Pacific Ocean. Because ofits geographical origin, the current is very cold, relatively shallow and provides a significant load ofcold water promoting upwellings on its way through the waterfront; an upwelling is the vertical movement of water masses of deep levels to the surface. Bograd et al. (2009) reported a significant inter-annual variation in the characteristics of upwelling in the period 1967-2007, with extended periods of high (1970's, 1998-2004) and low (1980-1995) and a trend towards a late and short term in the CCLME (Marine Ecosystem of the California Current- California Current Large Marine Ecosystem). Also reporting that El Niño years are characterized by a delay and weakening of upwelling in this ecosystem.

El Niño (ENSO-The Niño Southern Oscillation) is a disruption of the ocean-atmosphere system the happens in the tropical Pacific and contributes to significant changes in the conditions of weather and climate on the planet. This phenomenon has two components: 1) oceanic, where the Tropical Pacific temperatures surface are contrasted and create El Niño, which occurs when having warm anomalies and La Niña, which occurs when having cold anomalies; and 2) Weather (Walker condition), quantitatively measured by the Southern Oscillation Index which is a reflection of the relative change between the mean monthly atmospheric pressure at sea level between the western sectors (around Darwin, northern Australia) and the central-eastern Pacific Ocean (around Tahiti Island) in French Polynesia. The duration of El Niño is variable. According to ^{McPhaden (1999)} occurs every three to seven years and lasts 12 to 18 months. The NOAA reports that the condition tends to occur every 3-5 years and usually lasts between 9-12 months. Contrary to the condition of El Niño, La Niña can last 1-3 years.

Regarding the ENSO, it is documented that affects the frequency, intensity and spatial distribution of rainfall. The "NationalAeronautics and SpaceAdministration" (NASA) mentioned that its influence is mainly in the coastal areas of the Americas, the Northern Hemisphere from Alaska to central Chile in the Southern Hemisphere. In Mexico, according to ^{Magaña et al. (1998)} winter rainfall in El Niño years increase in intensity, as occurred during the winters of 1982-1983, 1986-1987 and 1991-1992 and become weaker in the summer.

During the Niño winters, the jet stream of midlatitudes moves towards the south causing an increased incidence of cold fronts and rain in the northern and central Mexico (^{Magaña et al, 1998}). The opposite occurs approximately in Niña years. The influence on winter rainfall is not always the same, especially when changes are analysed regionally. The regions of Baja California and Sinaloa have experienced some of the worst floods for years Niño.

Characterizing the climate at the regional level, the most limiting feature is the null or limited availability of historical climate data. In this sense, several authors have documented about the dominant role ofsatellites to observe the dynamics and changes in the Earth system. The most significant advance is the coverage and characterization of the planet on a global scale through increased spatial and temporal resolutions of satellite products, which would be impossible to obtain without this support (^{Dowell et al., 2013}), besides, obtaining regional specific response models improve the representativeness of the data to characterize ecosystems.

In this manuscript, the influence of ENSO is explored in the number and intensity ofrainfall events in the physiographic region of the Baja California Peninsula. The time series of rainfall data was obtained from the records of the satellite "TRMM" (Tropical Rainfall Measurement Mission), estimating daily rainfall globally in pixels of ~ 27 km. Splitting the database at intervals of 5 mm identified the extreme rainfall events as >30 mm. The relationship between extreme events and ENSO episodes has not been studied extensively in the Peninsula although some case studies have been performed.

Materials and methods

Description of the study área

The spatial boundaries of the physiographic province were plotted according to the methodology of the National Research Institute of Statistics, Geography and Informatics (INEGI), which considers the geological origin, landscapes, rock types and geo-forms. In this province three physiographic discontinuities are located: the Vizcaino Desert, with wide plains and dunes that are disrupted in a westerly direction; Magdalena plains, a region with low and depressions that are flooded during the rainy season and whose sandy coast is interrupted for some lagoons (several of them connected to the sea) and the Cape Region with mountains ofgranite (Figure 1). The mountain system runs along the peninsula throughout the east coast, the steepest slopes descend directly to the Gulf ofCalifornia and softer on the lesser slopes into the Pacific Ocean. The northern part of the province is dominated by intrusive rocks while in the western plain patches of alluvial sediments are located.

Figure 1 Spatial distribution of the pixel "TRMM" (left) and physiographic boundaries of the province Baja California (right). Lambert's conformal conic projection. 

Results and discussion

In total, 28 714 rain events were observed distributed throughout the physiographic province, an average of 1914 events per year (Figure 2).

Figure 2 Rain events observed in the historical data set. Data were stratif ied according to the amount of the estimated rain. 

Among the rain events, just under 63% (18, 059) were in the range of between 1 and 5 mm ofrain; 16.7% between 5 and 10 mm (4 788); 7.3% (2 118) between 10 and 15 mm; 4.8% (1 382) were extreme events over 30 mm; for other categories the number of events was 1 166 (15 to 20 mm -4%), 720 (20 to 25 mm -2.5%) and 481 (25 to 30 mm -1.67%). With respect to the extreme events, the results were contrasting and also unexpected. The highest frequency (471 events) was recorded in 2002 (ENSO-NM) and the lowest number (25) in 1999, in ENSO-NiF. However, in other years of ENSO, the observed results were contrasting. In 2009 (ENSO-NM), 134 extreme events were registered,just under 30% ofthose registered in 1999. In another NiF year (2010), the number of extreme events recorded was 37. These contrasts do not give a clear idea of the relationship between ENSO and extreme events, but rather opens up a range of possibilities to be addressed in future researches.

From Table 1, the numbers in bold correspond to the maximum value of the data set; figures in bold and inclined correspond to the minimum value of events in the time series data; underlined numbers correspond to the maximum and minimum of all the historical data series.

Table 1 Distribution of rainfall events by ENSO. 

These results show that in the region of study, the number of rainfall events increases during Weak La Niña, most of the rain events of lower intensity than 5 mm. Additionally we infer that this indicator is opposite (asymmetric) with ENSO-Weak El Niño. An observation of interest to follow up on the statistical analysis of the data is almost parallel relationship to the frequency of rain events between the conditions NiD and N. With respect to extreme events, in condition N was recorded the maximum historical series (372 events) and in the NIF the minimum (62).

The inter-seasonal variation was performed by analysing per calendar-monthly (Table 2).

Table 2 Distribution of rainfall events stratif ied by calendar-month. 

From Table 2, the numbers highlighted in bold correspond to the maximum of the series; those highlighted in bold and inclined are the minimum series and underlined are the maximum and minimum of the entire data set. The Table 2 shows that it happens in September when the highest number of rainfall events appear, while the lowest number of rain events observed in different months and different strata. The <5 mm in June; events between 5 and 20 mm, in May; the >20 mm / event in April -includes extreme events. For marking that in the months of May and June there were no rain events higher than 15 mm.

The largest number of extreme events was observed for the months ofSeptember andAugust. This situation is consistent with those reported by several authors regarding the presence of storms and tropical cyclones affecting western Mexico. According to ^{Dorst (2010)} , the hurricane season is from July to November, with peak activity in late August and early September. Severe storms occurred in September (^{Visher, 1992)} ; Hurricanes greatly impact on the peninsula of Baja California Sur occurred in August and September (^{Vadillo-Romero, 2003}); according to the records of the National Hurricane Center (NHC), in the report of 2006, more than 150 tropical storms have hit Baja California Sur.

These results suggest extending the study, adding new variables to the database such as differentiating the Pacific coastline with the Gulf of California, virtually separate the peninsula in the north and in the south due to the latitudinal extension, considering that the southern part ofthe Peninsula suffers from the increased presence of hurricanes, including data of land-use and incorporate data of most of the extreme weather events such as hurricanes. One aspect ofsignificance and complementary to the rainfall data is the incorporation of direct data from historical series ofmeteorological stations in the study area in order to detect deviations from the estimated probable data TRMM.

Statistical analysis

Analysis of rainfall events

In the general analysis model was declared the number of events as the dependent variable and three independent categorical variables: the ENSO, the calendar month and the layer of rainfall. The results were significant (p< 0.05) for ENSO, Month and Stratum condition. No co-linearity was observed between the variables of the model (0.28> r 2 <0.70). The results of the Bonferroni test are shown in the Table 3.

Table 3 Bonferroni test (a= 0.05) for homogeneous groups. All data series. 

The Bonferroni test for homogeneous groups confirmed the results in the Table 1, with respect that the ENSO N and ENSO NiD are homogeneous for the frequency of rain events. Additionally, they were under the same condition of homogeneity conditions of ENSO NiF and ENSO NM. The ENSO-ND and ENSO-NiM resulted in different groups; respectively, higher and lower average frequency of rainfall events.

Regarding the calendar-month, surprisingly, the highest number of rain events was not observed in September, but continuously during the winter rain season (January, February and March) and in the middle summer (August). The lowest number of rain events on average was observed for the hot months of May and June.

Regarding the strata, the differences between groups were observed mainly for <5 mm, observing the highest average events (2.35), followed by 5 to 10 mm (1.28) and extreme events (>30 mm, -1.19); no significant differences among the groups ofat least 10 mm 30 mm were observed.Although, no history of interval analysis was found to track rainfall patterns in Mexico, ^{Grimm and Tedeschi (2008)} suggested using a threshold of 10 mm day^-1 daily for regions with wet season. The climate in the physiographic province of Baja California Peninsula is predominantly arid and semi-arid, therefore we consider that the proposed analytical strata are considered appropriate.

Analysis of estimated rain

Regarding the volume ofrain, the MANOVAanalysis included the estimated dependent variable of rain and the categorical variables as independent, ENSO and calendar-month. The analysis resulted in significant differences (p< 0.05) for the ENSO and calendar-month [F (5, 19 887)= 91 241, p<0.01; and F (11, 19 877)= 215.03, p< 0.01, respectively] (Figure 3).

Figure 3 Estimated monthly rain by ENSO (A) per calendar month and (B). The value corresponds to the monthly average and the vertical bars to the confidence interval (0.95). 

The highest rainfall is observed in year N and ND (Figure 3a) while the lowest in the NiM and NIF condition, where the estimated average rainfall was the lowest. In the Figure 3b we observe a bimodal behaviour in the distribution ofrainfall per calendar-month, with the months of August and September being the wettest months and the season from April to June is the driest. The Bonferroni test for homogeneity of this groups is shown in the Table 4.

Table 4 Bonferroni test (a= 0.05) for homogeneous groups. All data series. 

According to the Bonferroni test, the six ENSO can be grouped into three groups. The ENSO-NiM and ENSO-NiF condition with the lowest estimated rainfall; ENSO-N and ENSO-ND with the highest average rainfall; and ENSO-NM and ENSO-NiD with average intermediate rainfall. Regarding the calendar-month, the inter-seasonal variation corroborates the hottest months with the lowest average estimated rainfall (April, May and June) and the months of August, September and October as the highest average precipitation.

These results confirm that described by NOAAin the sense that, the condition Niño influences the development of hurricanes in the Eastern Pacific. In addition, we mention that the results suggest that ENSO-N could also cause the same effect. However, the nature of our data allows us to infer the above only because it would be necessary to conduct a specific analysis approach for future researches including the analysis dates and paths of hurricane events and document their influence on the landscape of the ecosystem.

In this context, the results seem to indicate that, the extreme rainfall events that occur during hurricane season, influence by increasing the estimated average monthly rainfall, especially during the late summer and autumn, while during the winter season (December, January and February) is when the rainfall events are more constant and less intense. According to ^{Grimm and Tedeschi (2008)} in addition to the increased frequency of extreme rainfall events, the ENSO episodes can also affect its intensity, which is useful information to refine models of climate prediction and prevention schemes of hydrological disasters.

Discussion

The results document the influence ofENSO on the number of rainfall events and their intensity in the physiographic province of Baja California Peninsula.According to ^{Grimm and Tedeschi (2008)} , the impact ofextreme events are noted for their persistence and defined based on 3-day averages. The average rainfall of 3-days is calculated and the values refer to the central days. The gamma distributions are fixed to these means (considering only values above 0.1 mm), a distribution for each day of the year. Extreme events are those with an average of 3-days above the 90th percentile of stratification in rainfall events at intervals of 5 mm and in particular that relating to extreme events (<30mm), is an option that does not conform strictly as proposed by ^{Grimm and Tedeschi (2008)}.

However, given the volume of data processed and the absence of direct data to locally calibrate TRMM data, and taking account of the exploratory and relatively innovative nature ofthis proposal, the results are sufficient to strengthen the theory of influence of ENSO in the spatial frequency and distribution of rainfall in the study region. An important need relative to the model of analysis of rainfall data is that it most be enriched by the incorporation of other ecosystem response variables, such as indices ofgreenness and moisture in the vegetation and soil, cloudiness and sunshine, among others. Additionally, we might consider studies designed to explore the effect of coastline on the response of marine and terrestrial ecosystems.

An important effect that was not considered in this proposal is the spatial distribution ofinsolation. According to Valdés-Barrón et al. (2013), the distribution of the insolation is predominantly latitudinal. In the field of earth sciences, mapping of solar radiation is important to discuss, because it is a climatic factor that might be used in modelling and mapping of weather elements such as air temperature, evapotranspiration ^{Ninyerola et al. (2000)} ; ^{Fu and Rich (2002)} ; ^{Cristóbal et al. (2005)} and because it is a variable that directly influences the biological processes ofthe plants.

According to ^{Pons and Ninyerola (2008)} , mapping of continuous surfaces is very interesting to integrate and analyse spatial information from irregular distributed points (weather stations) when the goal is to obtain values of solar radiation by topography of a regional scale. In functional logic, the interrelationships in a large geographic area are as variable as the rate of sunlight that reaches the surface is causal ofbiodiversity and endemism. These considerations should be incorporated into future research proposals.

In general, the ENSO is related to changes in the intensity and frequency of extreme precipitation events (^{Grimm and Tedeschi, 2008}) and in periodic floods and droughts (^{Whelan et al., 2013}).According to ^{McBride and Nicholls (1983)} , the region of Australia is particularly sensitive to the ENSO; weather conditions are dry during the phases of El Niño, and with precipitation above the means during La Niña phases. The results in this manuscript are evidence that confirms some asymmetry between declining rainfall events during Weak El Niño and Moderate La Niña and increased frequency and intesity of rainfall events during the conditions ENSO-neutral and ENSO-week La Niña.

Conclusions

According to the analysis ofthe historical database TRMM, in the physiographic province in Baja California, a certain symmetry was observed between the conditions ENSO-neutral and ENSO-Weak La Niña and ENSO-Weak El Niño with ENSO-Moderate La Niña. We found that, the conditions Neutral and Weak La Niña cause an increase in the frequency of rainfall events and, in contrast, a decrease when the conditions of Weak El Niño and Moderate La Niña prevail.

Regarding the intensity of precipitation, asymmetrical relations were observed between the conditions ENSO-neutral with ENSO- Weak El Niño and between ENSO-Moderate La Niña with ENSO-Strong La Niña. During the conditions Neutral and Weak El Niño, the rainfall was more intense while decreasing during the conditions Moderate La Niña and Strong La Niña.

The relations of symmetry and asymmetry between rain events and ENSO, are just the starting point for future research proposals that address the theme, considering the incorporation to the analysis ofnew variables that elucidate more clearly the information gathered in this manuscript. Especially the information related to extreme events affecting most of the southern region of the physiographic province, the spatial distribution of rainfall events and how it affects the development and function of ecosystems, and the role of rain by defining spatial boundaries of land-use.