RESULTS AND DISCUSSION

4.1 δ18O, δ2H and d-excess results

Weighted mean values of δ2H and δ18O in precipitation and their standard deviation at both sites are compiled in Table 1 and are in agreement with data of the nearby stations Morón de la Frontera and Gibraltar (TEIJEIRO; ARÉVALO; ZABALETA; CASTAÑO CASTAÑO et al. , 2007). High d-excess values of 11.8 ‰ at Plaza de España and 13 ‰ at Doñana point to an important western Mediterranean influence beside the Atlantic vapor source.
[Insert Table 1]
The standard deviations of Table 1 show a very high temporal variability which may be related to the influence of both, Atlantic and Mediterranean vapor sources origins. Nonetheless, these standard variations are not in agreement with (CELLE-JEANTON; TRAVI; BLAVOUX, 2001) who report these high variabilities only for Barcelona, whereas the nearby Gibraltar station showed much less variability which was attributed to a domination of the Atlantic vapor source.
Time series plots of δ2H, δ18O, d-excess, temperature and precipitation for individual samples are shown in Figure 2. A seasonal fluctuation of δ2H, δ18O behavior is evident with more depleted values in the cold and heavy isotope enrichment during the warm season. Large precipitation events are more likely to be seen between October and April and are almost absent in July and August. Isotope ratios of the warm season show the expected increase in comparison with winter values, but due to the lack of precipitation in July and August there is no clear peak. Since precipitation samples are covering at least 2 weeks, one sample may represent several rain events which reduces the variability of isotopic signatures, typically observed between single events, due to their specific conditions defined by temperature, rain intensity or vapor source. d-excess values are anticorrelated with isotopic enrichment and decrease during the summer months, indicating evaporation effects due to higher temperatures and smaller precipitation amounts, which leads to reduced d-excess values (YANG; MU; GUO; BAOet al. , 2019).
[Insert Figure 2]
The average rainfall during the respective sampling periods registered at both sites are slightly below the average rainfall of 500 mm. The La Rinconada station close to Seville registered average annual rainfall of 487 mm during the years of observation from 2014-2018 and the meteorological station in Doñana measured 491 mm/y from 2016-2019.

4.2 Local meteoric water lines (LMWLs)

A bivariate plot of δ 2H versus δ18O , and the relevant meteoric water lines are presented in Figure 3. Most of the analyzed cumulated samples, but especially those of δ18O signatures smaller than -4‰, plot on the Western Mediterranean Meteoric WaterLine (WMMWL) indicating d-excess which can be explained by higher degrees of Rayleigh distillation of moisture sourced from isotopically similar marine sources, typical for western Mediterranean semiarid climate conditions.
[Insert Figure 3]
The corresponding slopes and intercepts of the meteoric water lines shown in Figure 3 and their respective standard errors are compiled in Table 2.
[Insert Table 2]
The more enriched samples from the summer season with higher temperatures plot more closely to or even below the GMWL. This points to secondary evaporation at higher temperatures leading to slightly lower slopes of the LMWL compared to the GMWL. The signatures of both sites are very similar in general. Small differences result between the two LMWL with a slightly higher slope at the Plaza de España site.
Comparison of amount weighted (PWLSR) versus un-weighted (OLSR) regression leads to almost identical slopes which is not in agreement with (HUGHES; CRAWFORD, 2012) who report that particularly in Mediterranean regions OLSR methods often lead to considerably lower LMWL slopes compared to those calculated by amount weighted regressions (PWLSR). This may be explained the fact that summer rain events with low d-excess values collected at our study sites are represented by medium to high cumulative precipitation amounts of more than 35 mm (Fig. 3) which may be related to the more Atlantic influence.
As illustrated above in Figure 2, the observed wide range of stable isotope compositions of rainwater in one hydrological year is attributed to seasonality. In general, most of the rainfall occurs in the cold season between October and April but with high temporal variability regarding the distribution among these months. Precipitation weighted average of isotope signatures for different climatological seasons (Fig. 4) shows that the cold season leads to more depleted values at both sites.
[Insert Figure 4]
Furthermore, the Plaza de España site during the summer season plots below the LMWL indicating secondary evaporation due to thermal effect with higher temperature due to the larger distance to the coast.

4.3 Effects of temperature on δ18O and d-excess in precipitation

Figure 5 shows the relation of δ18O and d-excess values with weighted temperature and sampled volume for the study period at both locations with the respective regression lines in Table 3. Results show that most rainfall events occur at temperatures below 20 C which is the “colder” season regarding local conditions. Figure 5a indicates weak and statistically non-significant (p > 0.05) positive correlations between δ18O and temperature at both locations with a large scattering which may be related to the influence of Atlantic and Mediterranean vapor source origin and is also visible in the standard deviations compiled in Table 1. Due to the scarce rainfall during the summer months, the major amount of rainfall occurs in a comparatively small range of temperatures of approximately 8⁰C at both sites which considerably reduces the importance of temperature compared to other local or regional effects such as vapor source variability. Considering the fact that the sites are located at low-mid latitudes and also close to the ocean which additionally dampens the annual temperature amplitude, the weak relation is in agreement with general findings that positive correlation of the temperature effect were identified mainly in high and mid-high latitude continents (DANSGAARD, 1964). Nonetheless, the pronounced variability is not in agreement with (CELLE-JEANTON; TRAVI; BLAVOUX, 2001) who report this only for Barcelona, whereas the nearby Gibraltar station showed much less variability, which was attributed to a domination of the Atlantic vapor source.
Figure 5c shows a statistically significant decrease of d-excess with high temperatures at the Doñana site which is attributed to secondary evaporation and confirms results discussed above in Figure 2, although the correlation is weak. In contrast, for the data at Plaza de España site, the decrease of d-excess with high temperatures can be observed but less pronounced and without statistical significance and therefore secondary evaporation was not clearly identified at this site. Larger time series may be necessary to draw a sound conclusion here.
Figure 5b plots δ18O versus sample volumes and shows a weak negative correlation of more depleted values for higher sample volumes at both locations, which can be the consequence of two effects. It may be attributed to the rainfall amount effect leading to a decrease of heavy isotopes with the progress of precipitation but it also may be attributed to the temperature effect since more intense rainfall events are observed during the colder seasons. According to (DANSGAARD, 1964), the amount effect of precipitation was observed typically in low-mid latitude areas in proximity of the ocean which concurs with our study sites. Multiple regression analysis described below will discriminate the individual importance of both effects. However, also here only at the Doñana site the effect shows statistical significance, whereas at Plaza de España, the value is slightly above the 0.05. Figure 5d shows a weak positive correlation of sample volume with d-excess values which is attributed to secondary evaporation when rainfall is low and temperatures are higher and which is also in agreement with observations of other authors (KRAJCAR BRONIĆ; BAREŠIĆ; BORKOVIĆ; SIRONIĆ et al. , 2020; YANG; MU; GUO; BAO et al. , 2019). Again, results are statistically significant only for the Doñana site (Table 3).
[Insert Figure 5]
[Insert Table 3]
To discriminate the particular impacts of the individual parameters precipitation amount and temperature on δ18O as a dependent variable, multiple linear regression analysis was performed for the Doñana site. The model showed that only precipitation amount revealed statistical significance (p-value: 0.00347) whereas temperature has a negligible effect (p-value: 0.46). Due to the lack of statistical significance at the Plaza de España site as shown in Table 3, no multiple linear analyses was performed for these data.

4.4 Vapor source origin and d-excess in precipitation

To examine vapor source origin as a possible reason for d-excess, backward trajectory paths were calculated for the Doñana Site with Hysplit for the periods with the maximum and minimum d-excess values compiled in Table 4.
[Insert Table 4]
Table 4 shows similar conditions for both samples except for the precipitation amount and the average rainfall intensity showing higher values for both parameters the sample with maximum d-excess. The maximum d-excess sample is composed mainly by two major rainfall events whereas one event is dominating the volume of the minimum d-excess sample. The sample from June 2018 plots below the GMWL (Fig. 3) indicating secondary evaporation below the cloud which may have contributed to the low d-excess value.
Figure 6 shows that the low d-excess samples reflects one major event whereas two major events represent the high d-excess sample of precipitation.
[Insert Figure 6]
The backward trayectories were calculated using the startpoints and times of these major events compiled in figure 7.
[Insert Figure 7]
Results show that a significant proportion of vapor mass of the high d-excess sample originates from the Mediterranean pointing to the qualitative importance of vapor source origin for the elevated d-excess variability on the site. Beside the different source origins as possible explanation of d-excess variability, the minor rainfall intensity of the 2018 sample may additionally contribute to the low d-excess value.

4.5 Local effects

To identify potential local effects and spatial variation, the temporal evolution of δ18O and d-excess of samples representing identical time periods of precipitation at both sites were compared (Figure 8).
[Insert Figure 8]
In spite of the larger distance between the two sites results show very similar trends indicating their comparability and potential for detection of possible thermal effects with average annual temperatures of 19.2 °C in Seville (Plaza de España) compared to 18 °C in Doñana. 4 of 6 samples of the urban site (Plaza de España) show lower d-excess values which may point to the influence of higher temperatures but more samples of simultaneous rain events are necessary to confirm this hypothesis. Figure 8c also confirms the trend that heavier isotope signatures concur with lower d-excess values as discussed above.