2.2.2. Meteorological Data

Due to the lack of detailed and evenly distributed meteorological time-series in Spercheios river basin covering the entire simulated period (1960/61–2004/05), the necessary data for the hydrological modeling concerning daily temperature (minimum, average and maximum values) and precipitation were retrieved from the high-resolution gridded data set of daily climate over Europe termed E-OBS from Copernicus Climate Change Service [45,46]. The specific data set covers the period back to 1950 and provides high-resolution gridded fields at a spacing of 0.1◦ × 0.1◦ in regular latitude/longitude coordinates. The ensemble version of E-OBS v.19.0e (the dataset produced from averaging multiple equally probable interpolations of station-based observations, so as to provide the best representation of the spatial and temporal distribution of climate parameters and a measure of uncertainty [47]) is based on the European Climate Assessment and Dataset (ECA&D) initiative that combines collation of daily series of observations at meteorological stations, quality control, analysis of extremes, and dissemination of both the daily data and the analysis results (Figure 1; [48,49]).

The reliability of the E-OBS dataset was evaluated by comparing the time-series of in-situ observations from meteorological stations installed in the wider area by various agencies (Table 2; Figure 1) against the corresponding grid point of the E-OBS dataset. The statistical criteria used to investigate the dataset reliability were the following: mean error *ME*; mean absolute error *MAE*; root mean squared error *RMSE*; standard deviation *STDEV*; and correlation coe fficient *R*, while also the *p*-value was calculated to estimate the significance of the results.


**Table 2.** Meteorological stations used in estimation of E-OBS e fficiency.

MEE: Ministry of Environment and Energy of Greece; HNMS: Hellenic National Meteorological Service; PPC: Public Power Corporation S.A.; P: precipitation; Tmin: minimum air temperature; Tmax: maximum air temperature; Tav: mean air temperature.

The lack of the necessary climatological data (relative humidity, solar radiation and wind speed) precluded the use of the Penman-Monteith equation for the estimation of daily reference evapotranspiration *ET*. Therefore, *ET* was estimated using the Hargreaves empirical approach [50], which is recommended only in cases of lack of other meteorological data and is considered to provide satisfactory results with an error rate of 10–15% or 1 mm/d, whichever is greater [51,52]. In the Hargreaves approach, except for daily average, minimum and maximum temperature, all the other required parameters (solar radiation, latent heat of vaporization) can be estimated using empirical relationships [52].

#### 2.2.3. Land Cover Spatial Distribution

The oldest o fficial and most detailed information concerning land cover distribution in Spercheios river basin was available from National Statistical Service of Greece for the year 1960. These data were part of the preparatory activities taken place prior the Agricultural and Livestock Census of March 19, 1961 [42] and concerned the main land cover types per local community: agricultural land, communal or private pastures for grazing animals, forest, artificial surfaces and water. It should be noted that in the 1960s land cover census, all agricultural activities (annual, crops, vineyards, tree plantations, and fallow land) were grouped together, while areas covered by shrubs, transitional woodland—shrub areas or areas with dense vegetation were characterized as pastures. During this procedure, forests were defined as areas mainly covered by ligneous plants clearly supported by a trunk and branching out to no less than 1 m from the ground. The category artificial surfaces included cities, settlements, roads, mines, and bare rocks. Finally, the category water included lakes, permanent inland and salt marshes, coastal areas and lagoons, estuaries, water courses, river beds, and areas covered by water for the greatest part of the year. Areas temporarily covered by water and areas lying near rivers or lakes dried and usually cultivated in summer were included in arable land (Table 3).


**Table 3.** Land cover nomenclature used in the present study.

The production of the 1960 land cover map (LC1960) was carried out by distributing the land uses per local community, by also taking into consideration the land cover distribution of the CORINE Land Cover (CLC) inventory for the year 1990 [44] and the Census of Agricultural and Livestock Holdings for the year 1961 [53]. As mentioned above, detailed agricultural activities were not distinct in 1960s land cover documented distribution [42]; therefore, the estimation of the different agricultural classes was based on their corresponding distribution per local community for the year 1990 [44]. Natural grasslands, pastures, sclerophyllous vegetation, transitional woodland-shrub, moors and heathland, and sparsely vegetated areas were classified as pastures, while artificial surfaces included continuous and discontinuous urban fabric, airports, industrial or commercial units, mineral extraction sites, construction sites, and road and rail networks (Table 3). The spatial distribution of forest classes (broad-leaved, coniferous and mixed) was also based on the CORINE Land Cover (CLC) inventory for the year 1990 [44].

The land cover maps for the years 1990 (LC1990) and 2018 (LC2018) were retrieved from CORINE Land Cover (CLC) inventory for the corresponding years [44]. It should be noted that based on the methodological approach of CORINE Land Cover (CLC), the density of houses is the main criterion to attribute a land cover class to the discontinuous urban fabric or to the agricultural area, in complex cultivation patterns class. In case of patchwork of small agricultural parcels and scattered houses, the cut-off-point to be applied for discontinuous urban fabric is 30% at least of urban fabric within the

patchwork area [43]. Therefore, documented sparsely populated areas in 1960 and 1990 land cover distributions were in many cases classified as complex cultivation patterns (Table 3; Figure 3).

**Figure 3.** Land cover maps for the year: (**a**) 1960 produced for the present study, (**b**) 1990, and (**c**) 2018 from CORINE Land Cover (CLC).
