4.2. Typology for Mediterranean non-perennial rivers and streams.
In this study, we considered only hydrological indices to categorize Mediterranean NPRS based on statistical criteria. The patterns described by the SOM classification suggest the existence of three groups (clusters I, II, and III) that reflect the interaction between ranges of the five attributes values (Fig.3 and 4, Table2). As expected, cluster I encompasses rivers almost dry during the year with many days and months without flow (usually with an onset of the dry phase in the first months of the year), with low daily flows and high annual rise rate. The cluster II lies between dry rivers (cluster I) and rivers with flows occasionally interrupted by drought (cluster III) with fewer days and months with zero-flow and with combinations of high flows. Our results indicate that the number of days and the percentage of months per year without flow are the most valuable attributes for Mediterranean NPRS classification, since they form well-defined groups. Similarly, CV of Julian date of annual start of zero flow has three compact and differentiated groups, due to that in rivers with no flow during most of the year, the dry periods tend to occur earlier . Conversely, the metrics of mean daily annual flows and annual rise rate (Fig.4) showed less variability, suggesting a lower discrimination capacity, a relationship that has also been shown in other studies .
Similarly to other studies , we included the number of zero-flow days to categorize NPRS. suggests a clear relationship between the duration and frequency of cessation periods for different ecotypes of the Guadiana basin, and other authors classified the NPRS according to the number of days of the dry riverbed as intermittent, ephemeral or episodic. The work by defines intermittent rivers cease seasonally or occasionally the flow (usually for weeks to months), ephemeral streams flow only in response to precipitation or snowmelt events (days to weeks) and episodic streams maintain surface water only during very short periods (hours to days), primarily after heavy rainfall events. In this line, our results are similar to the national legislation of Italy and , where days or months per year without flow are employed to distinguish between permanent, temporal, intermittent, and ephemeral water bodies. For example, in Italy , NPRS are classified according to the number of dry months as intermittent (less than 8 months per year), ephemeral (more than 8 months per year) and episodic (only water circulation after intense precipitation events). In Spain , temporary rivers are defined as those flowing for more than 300 days per year and possible dry periods; intermittent rivers are those flowing between 100 and 300 days per year; and ephemeral rivers are those flowing during less than 100 days per year. Similarly to national legislation, our results indicate three groups (Fig.3 and 4) which we would define as ephemeral rivers (cluster I), intermittent (cluster II), and temporal (cluster III), but we acknowledge notable differences in the range of days and months without flow ranges (Table 2). Apparently, the ranges of our classification are not closed but cluster I (ephemeral) would have more than 181 days or 7 months (59.6%) with zero-flows, while cluster II (intermittent) between 100 and 187 days or between 4 (35.8%) and 7 (56.3%) months without flow, and cluster III (temporal) with less than 46 days and 2 (20.0%) months with zero-flows. As far as we know, this is the first classification in which undisturbed gauging stations to obtain ranges based on statistical analysis of real data were used, providing new insights that improve the understanding of the hydrological dynamics in these water bodies. Their integration into national legislation, which currently classifies NPRS based on expert judgement criteria (e.g., Spain and Italy), would undoubtedly lead to better management strategies aimed at achieving the objectives set by the WFD.
Nevertheless, the applicability of these results might be affected by the limited number of NRPS water bodies controlled by gauging stations . For example, 39% of France’s rivers were NPRS applying a regional predictive model, but only 20% of the total gauging stations in France had zero flow episodes . In addition, some caution is needed when interpreting the results of this study. Our classification is based on data from 69 gauge stations, which represent a small proportion of the total NPRS located in the Mediterranean region. The use of a more extensive dataset would undoubtedly contribute to a wider and more robust classification. For example, we noticed a notable lack of episodic rivers, defined by as rivers with flowing water only during a few hours or days, depending on the rainfall event and its intensity. Since our study has used flow data on a daily scale, it was not possible to detect short-time-scale hydrological changes and none of the gauge sites used in this study was classified as episodic. Given the importance of NPRS in terms of the number of these freshwater ecosystems and considering the limited gauge sites used to measure them, more hydrological data would provide a more robust hydrological classification. For this reason, alternative data acquisition of periods of flow cessation would be necessary. The use of sensor loggers , remote sensing techniques with unmanned aerial vehicles , empirical observation, statistical approaches, process-based hydrological modelling , and multi-temporal satellite data , may solve this problem though notably increasing the costs .