4.1 Spatiotemporal patterns and seasonality
Figure 1 shows that for a large number of stations, there are only a small number of years with more than 10% zero-flow days. Only a few stations have zero-flow 90% of the years with available data. As shown on the map (Figure 1), the annual percentage of years with zero-flow days can vary strongly even for neighboring stations. This highlights the influence of local characteristics (geology, land cover, water use…) on zero flow occurrences. Snelder et al. (2013) previously observed over France that the high spatial heterogeneity in small-scale processes associated with intermittence partly explains the low spatial synchronization of zero-flows. There is a weak latitudinal gradient in the occurrence of zero-flow days, with the higher mean annual number of zero-flow days in the South (rho= -0.36 with latitude, significant at the 5% level), but with a very strong spatial variability even for neighboring catchments. This implies that most intermittent streams are not necessarily associated with the aridest climate conditions in Southern Europe.
Clustering has been applied using the variables θ , the mean direction of zero-flow and the variability around this date, r , computed for the winter and summer seasons. As shown in Figure 2, three different seasonality patterns were identified. The largest one, Cluster 2, is composed of 376 stations having a mean date of occurrence for zero-flow days between May and November, at the end of summer. The location of the stations composing this cluster are scattered all across Europe, in different climatic zones ranging from Continental to Mediterranean climate types. The second-largest cluster, Cluster 3, contains 47 stations with a mean occurrence of no-flow events between January and March. It includes stations with a nival regime, such as the Pyrenees of Scandinavia, that experience cessation of flow due to freezing. The Cluster 1 (29 stations) mean corresponds to late fall (November to January). As shown in this analysis, for most stations (Clusters 1 and 3), the zero-flow conditions are more frequently observed in summer months or during winter or early spring due to snow and ice cover. Yet, as shown in the map in Figure 2, there are no clear spatial patterns that could be identified from this analysis though Cluster 3 stations are located predominantly in mountainous or northern areas.