METHODS

Samples of rainwater for stable isotopes (δ2H and δ18O) analysis at the two sites were collected in a 5-L polyethylene bottle between January 2014 and April 2019 in the Doñana National Park and in Seville city (Plaza de España). In total, 77 rainwater samples were collected with 40 rainwater samples from Plaza de España (January 2014 to June 2018) and 37 rainwater samples in Doñana (February 2016 to April 2019, see supplemental material). Sampling was performed within intervals of at least 14 days basis if rain occurred but frequently intervals were longer according to the rainfall incidence especially during the summer season. To avoid contamination of the Laser Spectrometer with Paraffin, only samples with a minimum of 1 L sampling volume were analysed. In case of high intensity rainfall events leading to complete filling of the rainwater collector, sampling was performed after the respective event to avoid losing precipitation of subsequent events. The amounts of rainfall which exceeded the collector volume and therefore have not been collected are 114 mm of a total rainfall of 2135 mm between 24.01.2014 and 12.06.2018 at Plaza de España and 212 mm of a total rainfall of 1373 mm between 5.02.2016 and 26.04.2019 at Doñana. This corresponds to a loss of 5.3% and 15.5% at Plaza de España and Doñana respectively whereby 67% of the not analyzed precipitation volume at Doñana corresponds to only one meteorological event of 205 mm in early December 2016, when 142 mm exceeded the rain collector volume. Therefore, the individual samples are considered as representative averaged precipitation weighted results of cumulated rainfall for the respective sampling periods. Parallel sampling at the same dates was performed for 6 intervalls to allow direct comparison between the two sites.
The samples were collected by attaching a funnel to a high-density polyethylene bottle (5L) containing paraffin oil to prevent evaporation and the respective accumulated sample volume was registered for each sampling event. Samples were collected without air bubbles in polyethylene bottles and conserved immediately at 4°C in the refrigerator until isotopic analysis in the laboratory. At both sampling sites, some rainwater samples were taken twice on the same day for double checking. All double checked samples concurred within the measurement uncertainty and we used the average of both samples. Meteorological data were taken from nearby official meteorological stations, la Rinconada for the Plaza de España site whereas precipitation and temperature data of the Doñana site were recorded on site by a Meteorological station including a tipping bucket pluviometer (Meter Group, Inc., USA).
The analyses of the stable isotopes of the rain (δ18O and δ2H) were carried out at the laboratory of the Center of Hydrogeology of the University of Malaga by using a PICARRO™ L2120‐i Cavity Ring-Down Laser Spectrometer. The analytical method consisted of a continuous sequence of three IAEA referred internal standards and seven samples to process the instrumental drifting during the measurement cycles. From each sample, a six injections running-mode was performed and the first three injections were discarded to avoid memory effects. Blanks for the isotope analyses were analyzed sporadically to confirm there was no memory effect during combustion. The corrected isotope data for δ18O and δ2H were reported as ”delta” (d) values in parts per thousand (denoted as ‰) enrichments or depletions relative to the Vienna-Standard Mean Ocean Water (V-SMOW) (ROZANSKI; ARAGUÁS-ARAGUÁS; GONFIANTINI, 1993). The accuracy of isotope measurements was ± 0.1‰ for δ18O and ± 1‰ for δ2H.
Statistical parameters of the regression lines were calculated by the lm function of the R software (RCORETEAM, 2019) using ordinary least square regression (OLSR) and precipitation weighted least square regression (PWLSR). Calculation of local meteoric water lines was based on data covering entire years ranging from February 2014-February 2018 for Plaza de España and February 2016- February 2019 for Doñana. For the analysis of the amount effect only samples without loss of precipitation with a sample volume < 5L were considered. Data of precipitation and temperature of the Doñana site was recorded on site by a Meteorological station including a tipping bucket pluviometer (Meter Group, Inc., USA). Data of the official meteorological station La Rinconada en Seville were used for the Plaza de España site. Average temperatures assigned to samples representing a certain time interval were calculated based on daily temperatures at the respective site weighted by precipitation amounts of the corresponding days.
Back trajectories for the Doñana site were computed with the Hysplit Single Particle Lagrangian Integrated Trajectory (HYSPLIT) Model fromhttp://ready.arl.noaa.gov/HYSPLIT. Considering an average residence time of 10 days, back trajectories of 240 h were produced for predominant rain events of samples with maximum and minimum d-excess values starting at 500 m height which was considered as representative cloud altitude. Start times were defined in UTC times at the end of the respective major rain events. For each simulation 24 back trajectories were started at 3 h intervals representing a 3 day time period of the respective rain event.