*3.2. Freshwater Changes*

Figure 5 shows the regional conditions and changes in the atmospheric variables *T* and *P* (top left in the general approach illustration, Figure 4), in terms of the spatial distribution of their temporal averages over the whole study period 1930–2009 (Figure 5a,b) and the changes in period-average *T* and *P* from 1930–1949 to 1990–2009 (Figure 5c,d). Spatially, long-term average *T* increases from a local range of around 7–11 ◦C in the north to around 15–18 ◦C in the south, while average *P* increases from a local range of around 460–800 mm/year in the east to around 800–1400 mm/year in the west. The corresponding changes also follow a north-south gradient for *T* and an east-west gradient for *P*. The temperature *T* increases in the cooler north and decreases in the warmer south (i.e., cool gets warmer and warm gets cooler), while *P* increases (or decreases the least) in the dry east and decreases (the most) in the wet west (i.e., dry gets wetter—or dries the least and wet gets drier—or dries the most).

Catchment-average results for *T* and *P* (Table 2, for the base case scenario) are consistent with the mapped grid-cell results (Figure 5) in showing that the most north-extending and coolest catchment (Aegean) has warmed the most (increase in average *T* of 0.4 ◦C) while the most southern and warmest catchment (Peloponnese) has cooled (decrease in average *T* of −0.2 ◦C). Moreover, while average *P*

has decreased in all catchments, the most east-extending and driest catchment (Aegean) exhibits the smallest *P* decrease (−35 mm/year) and the most western and wettest catchment (Ionian) exhibits the largest *P* decrease (−94 mm/year).

Overall, the atmospheric hydro-climatic changes in *T* and *P* from 1930–1949 to 1990–2009 have driven most catchments (except the Peloponnese and Ionian) toward somewhat warmer conditions and all catchments toward drier conditions. Furthermore, these changes have decreased the variability range of average *T* and *P* among the nested catchments, from earlier ranges of 3.4 ◦C (from minimum 11.5 to maximum 15 ◦C) and 224 mm/year (from 637 to 861 mm/year) to recent ranges of 2.8 ◦C (from 12 to 14.8 ◦C) and 145 mm/year (from 622 to 767 mm/year), respectively.

With regard to irrigation, this has increased greatly in Greece in terms of all associated variables: irrigated area (Figure 2a) and amount of water used for irrigation in terms of absolute volume (Figure 2b) and per irrigated area (Figure 2c). These changes imply corresponding irrigation increases also in the different study catchments (Table 2). As a consequence of these human-driven irrigation developments, combined with the atmospheric hydro-climatic changes (Figure 5b), the long-term average values of the landscape hydro-climatic variables *ET* and *R* have also changed over Greece and in the different study catchments. Figures 6 and 7 illustrate the changes in *ET* and *R*, respectively, for all catchments. Further result details are listed in Table 4 for the total regional catchment; corresponding results for the other catchments are listed in Supplementary Table S2 for Mainland, Table S3 for Peloponnese, Table S4 for Ionian and Table S5 for Aegean.

**Figure 5.** The spatial distribution of temporal average temperature (**a**) and precipitation (**b**) over the period 1930–2009 and corresponding period-average changes ((**<sup>c</sup>**,**d**), respectively) from the period 1930–1949 to the period 1990–2009. In these maps, the Ionian and Aegean catchments are indicated by black outlines and the three local catchments in mainland Greece with reasonable runoff data time series are indicated by blue outlines.

**Figure 6.** Evapotranspiration changes from 1930–1949 to 1990–2009 in total (Δ*ET*) and their climate (Δ*ETclim*) and irrigation (Δ*ETirr*) components. Main bars show mid-range results and error bars show the range of change estimates for different evaluation scenarios (Table 3 and Supplementary Table S1). Results are shown for the total regional catchment (TOTAL) and the nested Mainland (ML), Peloponnese (PEL), Ionian (IONIAN) and Aegean (AEGEAN) catchments.


**Table 4.** Freshwater changes and main uncertainty estimates for the total regional catchment. The terms *P*, *ET* and *R* represent precipitation, evapotranspiration and runoff, respectively, while Δ stands for change and subscript *clim* and *irr* indicate climate driven and irrigation driven change, respectively. Scenario definitions (Base, Alt.) are as given and explained in Table 3.

Overall, *ET* has increased to some greater or lesser degree, while *R* has decreased considerably over all catchments. For *ET*, the irrigation developments have driven the largest change component Δ*ETirr* (overall increase of around 40 mm/year), while the climate-driven change Δ*ETclim* is in the opposite direction (decrease) and mostly of smaller absolute magnitude (around or less than 10 mm/year for the total, Mainland and Aegean catchments). For the Ionian and Peloponnese catchments, the absolute magnitude of the decrease Δ*ETclim* is relatively close to that of the increase Δ*ETirr*. As a consequence, the total net increase Δ*ET* is relatively small in these two catchments (around 15 mm/year in the Ionian and 4 mm/year in the Peloponnese) and larger in the other catchments (around 25–30 mm/year).

The main reason for the relatively large decreases Δ*ETclim* of around −27 mm/year and −35 mm/year in the Ionian and the Peloponnese catchment, respectively, is that they are subject to the largest precipitation decreases Δ*P* (of −94 mm/year and −80 mm/year). These are also combined with Δ*T* conditions of zero warming (Ionian) or even cooling (of −0.2 ◦C in Peloponnese), which do not drive *ET* increase. A question to investigate in further climate-change research for this region is whether the generally large irrigation-driven *ET* increase (Δ*ETirr* component) and the additional latent heat flux that this implies from the land surface to the atmosphere have contributed to significant local cooling of the land surface, thereby counteracting local warming effects of global climate change. This has been found in other irrigated regions of the world [17,23] and such irrigation-driven surface cooling would be consistent with the relatively small increases of local surface *T* over Greece, including the cooling and zero warming in the Peloponnese and Ionian catchments with the largest and second largest increase components Δ*ETirr*, respectively.

**Figure 7.** Runoff changes from 1930–1949 to 1990–2009 in total (Δ*R*) and their climate (Δ*Rclim*) and irrigation (Δ*Rirr*) components. Main bars show mid-range results and error bars show the range of change estimates for different evaluation scenarios (Table 3 and Supplementary Table S1). Results are shown for the total regional catchment (TOTAL) and the nested Mainland (ML), Peloponnese (PEL), Ionian (IONIAN) and Aegean (AEGEAN) catchments.

In contrast to *ET*, the climate and irrigation change components of Δ*R* are both in the same direction, thus reinforcing each other and leading to considerable decrease in *R* across all catchments (Figure 7). In total, Δ*R* is around: −75 mm/year, −73 mm/year and −65 mm/year in the three largest catchments, total regional, Mainland and Aegean, respectively; and −91.1 mm/year and −119 mm/year in the two smallest catchments, Peloponnese and Ionian, respectively. The irrigation driven change component Δ*Rirr* = −Δ*ETirr* (Equation (3)) has similar magnitude (decrease of around −40 mm/year) across all catchments. In most catchments (total, Mainland, Aegean), the climate-driven change component Δ*Rclim* = Δ*P* − Δ*ETclim* (Equation (3)) is also a decrease of somewhat smaller magnitude (around −35 to −30 mm/year) than Δ*Rirr*. However, in the Ionian and Peloponnese catchments, the decrease Δ*Rclim* is larger (around −75 mm/year and −50 mm/year, respectively) than Δ*Rirr*. Overall, the decrease Δ*Rclim* is mainly due to the precipitation decrease Δ*P* across all catchments.

The results of net total increase in average *ET* and decrease in average *R*, which for *ET* masks and for *R* exacerbates the decrease expected from only the observed atmospheric climate change (in *T* and *P*), are consistent with corresponding results for other irrigated areas of the world. For example, in the Indian Mahanadi River catchment, draining into the Bay of Bengal, with decreased long-term average *P* by −60 mm/year and increased irrigation water use by 81 mm/year (from 1901–1955 to 1956–2000), the average *ET* increased by around 55–70 mm/year while the average *R* decreased by around −130–−115 mm/year [39]. Furthermore, in the case of the Aral Sea catchment (ASC), *P* increased by 11 mm/year and irrigation water use increased by 23 mm/year (from 1901–1950 to 1983–2002), while the average *ET* increased by 15 mm/year and the average *R* decreased by −28 mm/year [39]. In both of these irrigation cases, as also found here for Greece, the irrigation development has led

to *ET* increase in spite of *P* decreasing (or to greater *ET* increase than the increase in *P* for ASC) and consequently to much greater decrease in *R* than the decrease in *P* (or to *R* decrease in spite of *P* increasing for ASC); in both cases, as in Greece, the large *ET* increase is also not explainable by just the observed warming (increase in average *T*) over each catchment. A global study of *ET* changes around the world's land areas has also shown statistically that irrigated areas (and areas with dam and reservoir developments for meeting the irrigation and other increased water demands) have experienced significantly greater *ET* increases than other, more undisturbed land areas with regard to such human developments [8].

The irrigation-driven increases in *ET* imply greater losses of freshwater from the irrigated catchments. For most such catchments, these water losses are not compensated by corresponding increases in observed *P*, since the observed *P* changes (even combined with the observed *T* changes) cannot explain the total *ET* increases [8,39]. From its use for helping crops to survive dry season and drought conditions [16], the irrigation water adding to increased *ET* in an irrigated catchment is thus lost from that catchment and goes to feed other catchments or maybe even the sea. This leaves less freshwater for other uses in the irrigated catchment, such as for households, industry, energy generation and/or ecosystems. Water managers and decision makers need to understand the involved trade-offs and make conscious, informed choices for sustainably balancing freshwater uses among societal sectors and ecosystems. Further research is also needed to support such choices by investigating and revealing the implications of different development scenarios.
