4.3 Links with SPEI anomalies
In this section, the findings of the relationships between the occurrences of zero-flow days and SPEI anomalies are presented. In the first step, a trend analysis has been performed on SPEI over the different basins for different aggregation periods with 6, 12, 18 and 24 months. The results show a clear pattern with positive trends at stations located north of 45°N and negative trends in the south, such findings are fully consistent with Spinoni et al. (2017). To check if the SPEI anomalies could be explanatory covariates for the inter-annual variability of zero-flow day occurrences, the SPEI with different aggregation time has been correlated with the number of zero-flow days and the maximum dry spell lengths. Since it has been shown previously that these variables and SPEI may exhibit trends, the time series have been de-trended in the cases where significant trends were found. Results show a strong association of zero-flow days with SPEI anomalies in particular at the annual and summer timescales (Figure 5), with a lower number of significant correlations during winter. The annual or seasonal sums of zero-flow days are more strongly associated with SPEI anomalies than the maximum length of dry spells, a metric that is probably strongly affected by the hydrogeology and the memory of the basins. Figure 6 shows the map of correlation coefficients obtained with the SPEI18. The correlations are negative for all SPEI timescales, indicating that negative SPEI anomalies, i.e. pronounced precipitation deficits, are linked with a larger number of zero-flow days. For about one-third of stations (36%) there is no significant correlation these stations are mostly located in Spain, southern France, North Africa, Cyprus, but also Belgium. A strong spatial variability of the pattern is once again evident basins indicating local influences. The strength of the correlations is higher for basins with a larger annual average number of zero-flow days for almost all indicators. This demonstrates that strongly intermittent basins are more influenced by climatic variations to determine the annual number of zero-flow days. This is probably due to the fact that these catchments are often close to the wet/dry threshold and consequently immediately impacted by a variation of precipitation availability. There is also statistically significant, but moderate (rho=0.2), the correlation between the strength of the correlations with the SPEI12 and basin size, indicating that the influence of SPEI12 might be stronger for smaller basins since larger basins can be more influenced by human activities. Yet, this relationship is not significant for the SPEI18 and SPEI24. Smaller basins have a reduced buffer capability in the context of climate variability so this behaviour is expected.