2. DATABASE OF INTERMITTENT RIVERS
The database of discharge time series of ephemeral and intermittent streams have been collected in the framework of the SMIRES COST action (Datry et al., 2017) in the different European countries in addition to individual contributions and stations from the GRDC (https://www.bafg.de/GRDC) database including countries outside of Europe such as Morocco, Tunisia and Israel. The selected rivers are characterized by natural or moderately influenced regime, and catchment area smaller than 2000km². The absence of dams or reservoirs upstream of the station gauge has been verified from a GIS analysis using the GRanD database (http://globaldamwatch.org/grand/). The metadata originating from different countries should be analyzed with care and can be misleading since the definition of “natural”, and the distinction between “little influenced” and “heavily” influenced rivers may vary strongly between different countries. Also, since this study is focusing on zero-flow days, it is possible that zero values are put in place of missing data; this is the reason why the data had to be checked carefully in the absence of metadata for many rivers.
Instead of using zeros, a threshold of 10-4m3 s-1 (0.1 L s-1) is considered to identify zero-discharge days to account for measurements errors of very small discharge values. However other thresholds, such as 5 L s-1 recommended by Gustard et al. (1992) or Delso et al. (2017) have also been tested, yielding similar results. In addition, the individual time series have been checked to verify if the smallest reported non-zero flows were below 10 L s-1l/s. If there were no non-zero flows below 10L/s in a series, the zero flows of that series were interpreted as wrongly reported missing values. The analyses performed in the present work are focusing on the annual and seasonal timescales, a hydrological year starting April 1 through March 31, and considering summer from April to September and winter from October to March, since this is common practice in low-flow analysis. The definition of the hydrological years was governed by a preliminary analysis on the seasonality of zero-flow events in Europe (mainly in summer and autumn but also in winter). This reduces the chance of observing a flow event in two consecutive hydrological years. The database includes 452 stations with at least two years with five consecutive zero-flow days. This criterion has been chosen to avoid including in the database some missing data in place of actual river intermittence, since it is unlikely that river flow will cease only one day in one year. Indeed, if for an annual time series only a single day with zero-flow is recorded, it could be missing data not properly reported in the metadata. For all stations, all years with more than 5% missing data have been removed. Across most stations (452), there is a common period for analysis between 1970 and 2010 when data are available. Two annual and seasonal metrics of duration, - the duration of the longest no-flow event (maximum length of zero-flow days) and the total duration of no-flow days (sum of zero-flow days) - and the mean date of no-flow days were studied too.