Classifying the flow regimes of Mediterranean streams using multivariate analysis Ons Oueslati, 1 Anna Maria De Girolamo, 1 * Aziz Abouabdillah, 2 Thomas R. Kjeldsen 3 and Antonio Lo Porto 1 1 Water Research Institute, National Research Council, Bari, Italy 2 Department of Agronomy, National School of Agriculture of Meknès, Meknes, Morocco 3 Department of Architecture and Civil Engineering, University of Bath, Bath, UK Abstract: Rivers in the Mediterranean region often exhibit an intermittent character. An understanding and classification of the flow regimes of these rivers is needed, as flow patterns control both physicochemical and biological processes. This paper reports an attempt to classify flow regimes in Mediterranean rivers based on hydrological variables extracted from discharge time series. Long-term discharge records from 60 rivers within the Mediterranean region were analysed in order to classify the streams into different flow regime groups. Hydrological indices (HIs) were derived for each stream and principal component analysis (PCA) and then applied to these indices to identify subsets of HIs describing the major sources of variations, while simultaneously minimizing redundancy. PCA was performed for two groups of streams (perennial and temporary) and for all streams combined. The results show that whereas perennial streams are mainly described by high-flow indices, temporary streams are described by duration, variability and predictability indices. Agglomerative cluster analysis based on HIs identified six groups of rivers classified according to differences in intermittency and variability. A methodology allowing such a classification for ungauged catchments was also tested. Broad-scale catchment characteristics based on digital elevation, climate, soil and land use data were derived for each long-term station where these data were available. By using stepwise multiple regression analysis, statistically significant relationships were fitted, linking the three selected hydrological variables (mean annual number of zero-flow days, predictability and flashiness) to the catchment characteristics. The method provides a means of simplifying the complexity of river systems and is thus useful for river basin management. Copyright © 2015 John Wiley & Sons, Ltd. KEY WORDS flow regime; Mediterranean rivers; hydrological variables; principal component analysis; cluster analysis; regression modelling Received 13 June 2012; Accepted 30 April 2015 INTRODUCTION Mediterranean rivers are characterized by irregularity of flow and harsh hydrological fluctuations (Arab et al., 2004). Extreme seasonal variation in streamflow often causes a marked pattern of zero or low flow and the lowering of the water surface into isolated pools along the river when flow ceases (Kirkby et al., 2011; Gallart et al., 2012; De Girolamo et al., 2014). This leads to intermittent flow in many rivers experiencing a shift between lotic and lentic conditions during the year (Morais et al., 2004), a phenomenon that affects the invertebrate metrics typically used in ecological quality assessment (Buffagni et al., 2009). Prat et al. (2014) have pointed out that methods developed to measure ecological and chemical quality in perennial streams may only be applicable under certain circumstances or indeed may not be applicable at all if the stream’s hydrological conditions are far from those of a perennial stream. Thus, accurate river classification based on hydrological regime is needed before determining surface water quality status (Munné and Prat, 2004). On the other hand, environmental legislation such as the European Union Water Framework Directive (WFD) (EC, 2000), which calls for the protection and improvement of the ecological status of water resources, requires the character- ization of surface water body types. The main objective of WFD river differentiation is to identify sets of streams that are comparable in order to define the reference conditions. † In addition, Nikolaidis et al. (2013) have highlighted that effective river basin management requires the use of *Correspondence to: Anna Maria De Girolamo, Water Research Institute, National Research Council, Bari, Italy. E-mail: annamaria.degirolamo@ba.irsa.cnr.it † ‘Reference condition’ is defined as the best status achievable – the benchmark. This is also known as ‘high status’, which is attributed to a river exhibiting biological, chemical and morphological conditions associated with no or very low human pressure. Reference conditions are type specific and so are different for diverse types of rivers. HYDROLOGICAL PROCESSES Hydrol. Process. (2015) Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/hyp.10530 Copyright © 2015 John Wiley & Sons, Ltd.