65-year changes of annual streamflow volumes across Europe with a focus on the Mediterranean basin

Abstract. Determining the spatio-temporal variability of annual streamflow volume plays a relevant role in hydrology for improving and implementing sustainable and resilient policies and practices of water resource management. This study investigates annual streamflow volume trends in a newly-assembled, consolidated and validated dataset of daily mean river flow records from more than 3,000 stations, which cover near-natural basins in more than 40 countries across Europe. Although the dataset contains streamflow time-series from 1850 to 2015 in some stations, the statistical analyses were carried out by including observations from 1950 to 2015 in order to have a consistent and reliable dataset over the continent. Trends were detected calculating the slope of Theil-Sen's line over the annual anomalies of streamflow volume. The results show annual streamflow volume trends emerged at European scale, with a marked negative tendency in Mediterranean regions (about −1 × 103 m3/(km2 year)) and a generally positive trend in northern ones (about 0.5 × 103 m3/(km2 year)). The annual streamflow volume trend patterns appear in agreement with the continental-scale climate change observations in response to climate change drivers. In the Mediterranean area, the declining of annual streamflow volumes started in 1965 and since early 80' volumes are consistently lower than the average. The spatio-temporal annual streamflow volume patterns observed in this work can help to contextualize short-term trends and regional studies already available in the scientific literature as well as to provide a valid benchmark for further accurate quantitative analysis on annual streamflow volumes.



Introduction 25
Elucidating continental patterns of annual streamflow volume changes in the Anthropocene epoch to confirm unequivocally the effects of climate change and human impact on water resources has become a challenge in contemporary hydrology (Bloschl et al. 2019). Although the hydrological scientific community undertook a great effort, almost no research robustly demonstrates an ubiquitous and uniform trend in European annual streamflow volumes, especially in Mediterranean areas where drought periods have been increased during the last fifty years (Caloiero et al. 2018). Few regional studies, mainly located in northern Europe, detected potential trends in river flow, relying mainly on data from the beginning of the second middle of the twenty-first century (Piniewski et al. 2018, Renard et al. 2008, Birsan et al. 2005, KLIWA 2003, Schmocker-Fackel and Naef 2010, Demeterova and Skoda 2005, 2009, Fiale 2008, Fiala et al. 2010. These studies showed a clear seasonal change in streamflow without finding a solid correlation with the geographic position of the catchments (Bard et al. 2015, Bormann et al. 2017). In addition, such studies have faced the extreme sensitivity of the river streamflow on the data selection, 35 the method used for the trend detection and the analyzed time window (Stahl et al. 2010). For these reasons, Kundzewicz et al. (2005) advocated particular caution in interpreting streamflow trend signals resulting from a restricted number of stations with a small recording period as the first consequence of potential gaps in the data time series or missing values which could alter the significance of the statistical tests.
Thus, 3,485 stations were selected to assembly a dataset that guarantees the best balance between the necessities to investigate a dataset as large as possible (which covers a large part of the continent), and to detect a historical variability. Location of the different gauged stations is reported in Fig. 1 on physical European map. The selected gauged stations belong to more than 40 European countries and provided time-series data from 1950 to 2015, characterized by about 1.200 points of measure per year, 100 on average (Fig. 2a). Unlike previous studies, more attention was given to the analysis of river flow time-series over the Mediterranean basin for which a dedicated effort was carried out in this study to fill the gap existing in previous studies. About one-third of gauged stations falls in this area, especially in Spain, southern France and Italy. Gauged stations that enclose catchments with an area more than 100,000 km 2 were excluded by the analysis because human disturbance is unavoidable at this scale (Piniewski et al. 2018). Nevertheless, about 90% of stations belongs to catchments with size less than 1,000 km2 as 105 shown in Fig. 2b. Daily hydrographs for all gauges were inspected to identify dubious patterns, all records were screened visually and those with visible inhomogeneity, problems in low flow range, or missing values for a long period of time ( > 2 year) were excluded (Kundzewicz et al. 2005). Despite potential levels of human-induced alterations of river flow regime could be still present in time-series data, a certain degree of disturbance can be tolerated (Murphy et al. 2013). For this reason, high flow conditions were not investigated and we focused the analysis on annual streamflow volumes. 110

Trend detection
Trend magnitude of a hydro-meteorological series of data is usually estimated using the Theil-Sen's estimator (Theil, 1950;Sen, 1968), a non-parametric test usually adopted for indicating monotonic trend and amplitude of change per unit time. It is a robust estimate of the magnitude of a trend in hydrological and climatic time-series as demonstrated in literature (e.g., 120 Kundzewicz and Robson 2004, Stahl et al. 2010, Burn et al. 2012, Hammanfor et al. 2013). In the present study, the slope of Theil-Sen's line, known as Theil-Sen's slope or Sen's slope, was calculated on the annual anomalies in streamflow volumes, an innovative modality with respect to the application on direct streamflow data. The annual anomalies in volumes were detected by comparing them with the baseline obtained by averaging annual streamflow volumes in the entire period of observation for each station. This strategy allows to emphasize trends, minimizing the random errors derived from uncorrected 125 measures or unexpected signals, as already tested by Pandžić and Trninić (1992). A positive anomaly indicates that the observed annual streamflow volume is greater than the baseline, while a negative anomaly indicates the observed annual streamflow volume is lower than the baseline. To homogenize the annual streamflow volume time-series and further to compare the streamflow volume anomalies among all the gauged stations, the value of each anomaly was divided for the catchment area obtaining volume anomalies per unit of area. Moreover, significance of the annual streamflow volume trend 130 was tested by adopting a non-parametric statistical approach based on Mann-Kendall (MK) (Mann, 1945;Kendall, 1975) test.
Such test has already shown its robustness in trend detection, in particular in case of non-normally distributed data such as the meteorological and hydrological series (e.g., Yue and Wang 2002;Yue et al. 2003;Yue and Pilon 2004;Piniewski et al. 2018).
In particular, if the result of the test is returned in H = 1, it indicates a rejection of the null hypothesis (i.e. presence of trend) at the alpha significance level (here assumed equal to 0.05). Conversely, if H = 0, it indicates a failure to reject the null 135 hypothesis at the alpha significance level (i.e. no presence of trend).

Annual streamflow volume trends in Europe
Anomalies in annual streamflow volumes for each gauged station was calculated, and in Fig. 3a   The European spatial pattern of the annual streamflow volume trend appears congruent also with the observed European temperature and rainfall long-period changes as shown in Fig. 5a  authors found that in some regions such as southeast of England, northeast of France, as well as Danish the contribution of the aquifer to streamflow is high especially in summer periods (i.e. when irrigation occurs). Various studies, moreover, have demonstrated that the mechanisms of interactions between groundwater and river flow contribute to moderate the influence of climate change drivers on streamflow, conversely, basins with less productive aquifers show a more direct response to climate drivers (Fleig et al. 2010, Laize et al. 2010. 195

Annual streamflow volume trend in Mediterranean area
Focusing on the main Mediterranean river catchments (according with European Environmental Agency classification), the number of stations with positive and negative Theil-Sen's slope for each catchment was computed, and the results are reported in Fig. 6. In all main river basins in Spain, France and Italy prevails negative trends of annual streamflow volumes. In Garonne and Rhone river basins is found the larger magnitude of negative annual streamflow volume trends, respectively, of about -2.2 200 10 3 m 3 /(km 2 year) and -3 10 3 m 3 /(km 2 year). No basin with marked artefact trends is found as demonstrated by the very close distance of the 25th and 75th percentiles from the median slope value, confirming, thus, trend homogeneities inside each basin.
Lack of information in the Padana Plain is still present today. This is due to the complexity of its river network and its peculiar ecosystem mainly based on a strong interaction between river discharge and groundwater, which makes difficult to carry out reliable flow measurements (Masseroni et al. 2017).
Negative trend over the entire Mediterranean basin is also confirmed by the analysis performed on the mean annual streamflow volume produced in this area. The annual streamflow volumes of each station were standardized by their mean value, and then the standardized annual streamflow volumes of all stations were averaged. The result is reported in Fig. 7, where the standardized annual streamflow volumes smoothened by a simple rolling average with a sliding window of 5 year length is shown along with the 25 th and 75 th percentile trends. When standardized annual streamflow volume is greater than 1 it means 210 that the annual streamflow volume is greater than the average of annual streamflow volumes, vice versa if standardized annual streamflow volume is lower than 1. The former case can be considered as a positive signal of annual streamflow volume exceedance, whereas in the latter an annual streamflow volume deficit. Fig. 7

4 Conclusions
This study closes the gap between regional researches on annual streamflow volume trend and a continental-scale pattern of its spatio-temporal variability. Starting from a dataset constituted by more than 3.000 gauge stations over more than 40 countries across Europe, anomalies in annual streamflow volume were computed and Theil-Sen's line slope was evaluated for each catchment over a recorded period from 1950 to 2015. A clear and undisputed trend pattern in annual streamflow volumes 230 is recognized by the statistical analysis, showing marked negative trends in Mediterranean areas and positive trends in northern regions of Europe. All main Mediterranean river basins reveal negative trends in annual streamflow volume with an expected decreasing in annual streamflow volume of about -1 10 3 m 3 /(km 2 year). On the contrary, in northern regions of Europe, a positive increase of annual streamflow volume is expected to be on average about 0.5 10 3 m 3 /(km 2 year). This trend patterns agree with the increase of temperatures and the decreasing in rainfall volumes detected by long-period observations on 235 European continent. Indeed, these observations confirm an increase in drought situations in the southern regions of Europe, whereas revel an increase of rainfall volumes and runoff production in the northern European countries. In the Mediterranean area, the effect of climate change caused an inversion of the annual streamflow volume availability with respect to the mean of observations, i.e. from positive to negative values, starting from about 1985. In the recent 30-year period , the streamflow volumes are consistently lower than the average availability of the period 1950-2015. 240 The results of this study, therefore, can pave the way for more detailed quantitative analysis of annual streamflow volume variability (especially during different seasons) in order to meet the needs of managing water resources in agricultural, industrial and civil sectors. https://doi.org/10.5194/hess-2020-21 Preprint. Discussion started: 27 January 2020 c Author(s) 2020. CC BY 4.0 License.