The Mediterranean refers to the sea and bordering region located in the middle of the European, African and Asiatic continents. With Köppen's classification the definition designated henceforth a moderate climate and extended geographically beyond the limits of the Mediterranean Sea (Köppen, 1936). However, the Mediterranean could be defined with several boundaries based on the field of practice; The hydrological boundary defined by the set of catchments draining towards the Mediterranean Sea was adopted for this study (Milano et al., 2013).

Since the geographic extent of the study is very wide, the delimitation of catchments was imported from international references. European catchments and some adjacent countries were imported from the Joint Research Centre (JRC) (De Jager and Vogt, 2010), catchments in the Middle East and Northern Africa were imported from HydroSHEDS database of World Wildlife Fund's project (Lehner and Grill, 2013).

The selection of catchments and their hydroclimatic data was very challenging as no single database gathers all the available stations in the region. The selected catchments cover a wide geographic domain spread across 15 countries in different physiographic settings. Despite the extensive work performed on data collection, some minor exceptions were still to be found. Some timeseries presented discontinuities and therefore only complete hydrological years were considered.

With multiple studies raising serious concerns of climate shifts and aridity expansion in the region, Allam has established a new high resolution classification for hydrology purposes based on Mediterranean specific climate indices (Allam et al., 2020b). The proposed approach includes the use of classic climatic indices and the definition of new climatic indices mainly precipitation seasonality index Is or evapotranspiration threshold SPET both in line with river flow regimes. The classification was set and validated by WorldClim-2 at 1 km high resolution gridded data for the 1970–2000 baseline period. Climatic classes coincided with a geographical distribution in the Mediterranean ranging from the most seasonal and dry class 1 in the south to the least seasonal and most humid class 5 in the North, showing up the climatic continuity from one place to another and enhancing the visibility of change trends. The K-means classification shown in Fig. 1 is distributed into five classes.

The functional approach for annual water balance estimation took birth from the empirical model of L'vovich and the process-based approach of Horton; L'vovich advanced an empirical approach for the two-stage distribution of annual precipitation. The first stage separates the precipitation into surface runoff or quickflow S and wetting W with P=S+W and the second stage partitioning separated the wetting W into Slow flow U and Vaporisation V with W=U+V (L'vovich, 1979) while Horton introduced the ratio of the vaporisation V to the wetting W and named as Horton index, H=V/W in a trial to search for catchments capacity to store the infiltrated water, depending on the soil type and the evaporation controlled by the vegetation type (Horton, 1933).

Ponce and Shetty have developed a simple conceptual model from the water balance equations to separate precipitation into different flows (Ponce and Shetty, 1995a, b). This model has been proposed as an alternative to empirical models, which have a limited applicability. Ponce et Shetty has adopted the same 2 partitioning stages as L'vovich and calibrated the initial abstraction coefficients surface runoff λs, baseflow λu, wetting potential Wp, and vaporisation potential Vp through the application of their equations on the same database and tented to classify the climatic regions according to the range of values they obtained. The mountainous conifer forests of North Africa close to the Mediterranean region has shown high Wp with low λs and a high Vp with high λu.

In the empirical approach, the annual water balance is estimated based on a systematic analysis of long-term rainfall and runoff series observed in different climatic regions and an empirical analysis of the effect of physiographic and climatic descriptors on the water balance (Sivapalan et al., 2011). The value of the empirical approaches lies in the simplicity of the relations which govern the annual assessment on the long term, and which appeared from the coevolution and the self-organization of climates, soils, topography and vegetations of the different regions.

The water balance partitioning was carried out and reproduced accurately the empirical relations for most of the catchments. The charts have shown the well-marked interannual variability for all water balance components, with the same observation for the inter-catchment variability with different ranges between catchments. We set as an example Nahr Ibrahim catchment in Lebanon which is highly watered with a mean annual precipitation MAP of 1450 mm, yielding a mean annual runoff MAQ of 1052 mm; a surface flow S ranging between 100 and 550 mm, a baseflow U showing remarkable high values ranging between 200 and 1350 mm, and vaporisation V showing an upper limit, reached every year due to water availability in contrast with other catchments. The water balance partitioning charts are presented in Fig. 2.

The Ponce et Shetty models were fitted over L'vovich partitioning charts. Some catchments did not yield very good results, this inaccuracy returns to either interrupted flows by dams and lakes, catchments interconnection or simply poor quality of data. The sensitivity analysis conducted on the US MOPEX dataset, showed that Ponce and Shetty curves fitting was insensitive to extremes and that the distribution returned a linear fit (Sivapalan et al., 2011). Those extremes were observed in the Mediterranean wherever the flow regime was dominated by the groundwater runoff, case of Nahr el Assi in Lebanon; or wherever the precipitation was very low, case of Andarax in Spain, or Alcantara in Italy. The runoff and baseflow threshold of all the catchments according to Ponce and Shetty (1995a) are summarized in Table 1 and shown in red on the same charts for Ibrahim in Fig. 2.

The spatial distribution of Ponce and Shetty water balance parameters reflected a wide geographical variability across the Mediterranean without any clear pattern or regional clustering despite some adjacent catchments yielding neighbouring values. The French catchments yielded the lowest values for all parameters with both λsWp and λuVp below 100 mm except for Hérault river (λsWp=122 mm; λuVp=1 mm). In Cyprus, λsWp show close values ranging between 175 and 350 mm. These values are among the highest in the Mediterranean due to the high soil permeability on the island; this is confirmed with the high wetting potential Wp values ranging between 5000 and 9000 mm. The Lebanese coast have a wide variability of λsWp ranging from 7 to 270 mm between adjacent catchments as well as for Wp ranging from 5000 to 27 000 mm (except for Nahr el Assi in Lebanon); same observations could be described for λuVp and Vp variability. We noticed across the Mediterranean, some matching catchments which share remarkable water balance similarities despite their geographical, climatic, or landscape difference. These catchments could be considered twin catchments like Lez in Franc and Erzenit in Albania (CC4); Fium-alto in France and Fluminimaggiore in Italy (CC4); Mirna in Croatia and Alento in Italy (CC5). The water balance similarity between these catchments could be caused by an underlying physiographic similarity like snow or karst components which make these catchments interesting to examine closely.

The diversity of both microclimates and physiography between catchments stands behind the water balance variability as hydrological response on the same climatic forcing differs according to catchments' landform, landcover, soil types and other physiographic features, and despite some close catchments between Spain, France and Italy, no geographical pattern was detected.

We verified the nondimensional formulation as suggested by Sivapalan et al. (2011) to check if any water balance pattern could be extracted from Mediterranean site-specific water balance results that express their common hydrological behaviour. The different non dimensional water balance components W*, S*, U* and V* were estimated from the values of W, S, U and V. A summary is shown here representing the annual average of all the catchments and coloured according to their climatic classes in Fig. 3 in a trial to discover any pattern behind, the average values of each climatic class were also added. The charts show the nature of competition between different water balance components of Mediterranean catchments as suggested by Sivapalan, where, at a first stage, quick flow S* is competing with the storage W* and at a second stage slow flow U* is competing with vaporisation V*. The climatic illustration of W* and S* vs P̃ shows that this competition is gradually evolving from climatic CC2 in the South, where low precipitation and low quick flows dominate, to CC5 in the North where high precipitation and high quick flows dominate.

The maps in Fig. 4 show an interesting spatial distribution where KH and KV gradually decrease from South to North while KB and KR gradually decrease from North to South, nonetheless some exceptions were noticed where some Lebanese catchments showed an inverted distribution. The geographical distribution coincided with the Mediterranean climatic classification, therefore we searched for which climatic index is mostly affecting this distribution and found that the water balance metrics of all catchments are slightly correlated with the mean annual aridity index Ep/P (R=0.55) which is also correlated to the non-dimensional aridity index φ with R=0.63 (r2=0.4), hence supporting Sivapalan finding on US MOPEX data where r2=0.46. It is therefore possible to say, to certain extent, that annual water balance metrics KH and KV increase while KB and KR decrease when seasonality increases, which suggests that intra-annual variability is responsible for the inter-catchment variability. We also noticed that some CC3 and CC4 catchments yielded lower aridity φ than CC5 which is unusual. And looking to their physiography, we noticed that those were mountainous karstic catchments under snow influence, case of Nahr Ibrahim in North Lebanon which yielded the highest slow flow peaking at 0.71 and total runoff at 0.94, due to high snowmelt contribution and high karstification with a similar observation for Spercheios, Nahr el Kalb and other.

This finding completes the one advanced by Sivapalan et al. (2011) on baseflow metric variability stating that the wettest catchments have yielded the highest baseflow contribution and that soil types and topographic features govern the subsurface flow as part of the self-organisation of the vegetation with the climate as proposed by Horton (1933).