5. DISCUSSION AND CONCLUSIONS
This study provides the first European-scale assessment of the trends in intermittence over the last decades. The most striking results are (i) the strong spatial variability of the detected trends, and (ii) the seasonality and relationships with climate drivers. Overall, there is a considerable spatial variability of flow intermittence, with possibly strong dependencies of the different indicators with basin characteristics. Most of detected trends are towards an increasing number of zero-flow days, tending to occur earlier in the season, in particular around the Mediterranean basin. For most basins, there is a strong association of zero-flow days with SPEI negative anomalies, but neighboring basins may exhibit different relationships showing again the strong spatial variability. This implies that regional predictions or generalizations for flow intermittence patterns should be interpreted with caution. Any mapping or extrapolation from such regional results may be prone to considerable errors if not considering basin characteristics that are likely to play a strong role in determining flow intermittence properties. In this study, the individual catchment characteristics (ie. topography, geology, land use, soil types) have not been analyzed, which would require a major work at such a continental-wide scale. In particular, it would be particularly interesting to distinguish basins with strong surface-groundwater interactions that could explain some of the patterns described in the present study. However, as noted by Snelder et al. (2013), flow intermittence is also controlled by processes operating at scales smaller than catchments, thus capturing these processes would require a much more detailed investigation than classical regional approaches to take into account the local physiographic characteristics (Tramblay et al., 2010).
A major uncertainty of the present work and, to a greater extent, applicable to all ecohydrological research on intermittent rivers is the definition of the zero-flow days and the lack of regional representativeness of the monitoring networks. With regard to the first aspect, there is a wide variety of measurements procedures in different European countries, leading to different accuracy of the measured discharge values, in particular the minimal values. For example, in the UK or in France the data is provided in m3 s-1 with an accuracy of three decimals, but for many other countries, the minimum reported discharge is sometimes much greater than 1 L s-1 due to different measuring methods . In addition, many rating curves at open channel stations have uncertainty at low flows caused by instability of the riverbed. In the present work, we considered a strict criterion to identify zero-flow days, but a more adequate selection could be made possible if good quality metadata information were available for most rivers, which is not currently the case for several countries. With regard to the second aspect of regional representativeness, several studies have highlighted the lack of measurements for intermittent rivers (Skoulikidis et al., 2017, Costigan et al., 2017). The number of monitored head water streams which are likely to be intermittent is indeed much smaller than the number of perennial and large streams in national and international databases, although their contribution to the water resources is probably high. This questions the rationale behind national measurement strategies for intermittent streams, in particular, the most ephemeral ones, since they present a possibly reduced interest in water resources management by comparison to perennial streams. Depending on the national monitoring strategies of the river network, it is possible that the selection of intermittent rivers to be monitored may be biased towards a specific type of rivers within a given geographic location of geological properties.
Another important aspect to take into consideration is the regulation status of the monitored rivers, which may evolve over time and not be available in the stations’ metadata. In the present work, the selected basins are those described as natural or weakly altered based on their metadata information. Yet, how to quantitatively define “weakly altered” between different national networks and monitoring protocols? This definition may differ from one country to another without a common objective criterion to define the percentage of natural discharge being diverted or used for water supply. Besides river alterations, often assumed to be an expert judgement, the quantitative evaluation of the water uptake would require extensive work to monitor and collect water consumption data over time. Even for a river considered to be unaltered, diffuse groundwater pumping may occur and therefore have impacts on the groundwater-surface interactions, which could in turn strongly impact intermittence occurrence. Two examples of the influence of river status as described in the metadata on the trends results may be found in the UK. The Coal Burn River is classified as broadly natural, but is an experimental catchment set up to assess the impact of afforestation and the trend analysis indicates an increase in zero-flow days, showing that the influence of land-use change might indeed be significant on this aspect of the flow regime. The limestone Slea River that conversely experienced a decrease in zero-flow days was also classified as natural, but further scrutiny revealed a discharge augmentation scheme installed in 1995. Taking into account all these local specificities in addition to the available metadata would require tremendous effort and the information may not always be as readily available. The main findings of the present study are an incentive to implement process-based studies on the intermittency characteristics for different climatic and physiographic environments.