**3. Results**

#### *3.1. Long-Term Land Cover Changes*

Based on the results, the land cover of Spercheios river basin has changed considerably over the last five decades. The artificial surfaces have increased during the years reaching from 1% in 1960 and 1990 to 3% of the total river basin area in 2018. This can be partly attributed to the fact that in some cases, small settlements in 1990 were classified as agricultural areas due to the 30% threshold adopted in the methodology by European Environmental Agency in CLC inventory for distinguishing discontinuous urban fabric and complex cultivation patterns [43].

Agricultural land ranges from 28% (470 km2) in 1960, through 32% (531 km2) in 1990, to 30% (498 km2) in 2018. Permanently irrigated land has increased from 2% (40 km2) in 1960, through 4% (71 km2) in 1990, to 8% (135 km2) in 2018. On the contrary, non-irrigated land has decreased from 12% (196 km2) in 1960, through 10% (171 km2) in 1990, to 5% (79 km2) in 2018. Other agricultural activities, the majority of which were also irrigated, range from 14% (235 km2) in 1960, through 17% (289 km2) in 1990, to 17% (284 km2) in 2018. Pastures have decreased over the last decades (from 636 km2—38% in 1960, through 599 km2—36% in 1990, to 538 km2—32% in 2018). Finally, forested land change ranges from 31% (517 km2) in 1960, through 30% (492 km2) in 1990, to 34% (558 km2) in 2018 (Figures 3 and 4).

**Figure 4.** Distribution of land cover type for the three land cover case studies examined.

Figure 5 presents the differences in agricultural land (Figure 5a), pastures (Figure 5b) and forests (Figure 5c) for the three land cover case studies examined. Some of the areas characterized by transitions in land cover among the cases studies were used to investigate the impact of land cover change on annual actual evapotranspiration.

**Figure 5.** Differences in agricultural land (**a**), pastures (**b**) and forests (**c**) for the three land cover case studies examined.

## *3.2. Meteorological Data*

Based on the results of the comparison between in-situ observations from ground meteorological stations and the E-OBS dataset, there was sufficient agreemen<sup>t</sup> regarding precipitation (Table 5). The correlation coefficient *R* ranged between low (0.3) to very high positive (0.9; based on the criteria for correlation interpretation proposed by Hinkle et al. [39]); nevertheless, the *p*-value in all cases was statistically significant at the 0.05 level, except in the case of the meteorological station Ano Mpralos. Overall, the E-OBS dataset systematically underestimated annual precipitation for the entire period of evaluation, except in the case of Zilefto meteorological station for which the *ME* was calculated to be positive. E-OBS dataset was not able to sufficiently estimate the altitude effect on the precipitation rate, leading to a higher value of *ME* in meteorological stations of higher elevation (Table 5; Figure 6). This led to an average 37% underestimation of spatially-averaged annual precipitation of Spercheios river basin.

**Table 5.** Statistical characteristics and efficient criteria of annual observed precipitation measurements and E-OBS dataset.


*N*: number of observations; *AV*: Average; *ME*: mean error (=E-OBS − station observed values); *MAE*: mean absolute error; *RMSE*: root mean square error; *R*: correlation coefficient; AMpr: Ano Mpralos; AYp: Ano Ypati; DVou: Dyo Vouna; GrOx: Grammenni Oxia; Lam: Lamia; Mous: Mousounitsa; Neo: Neochori; Pits: Pitsiota; Pyr: Pyra; Rent: Rentina; Tril: Trilofo; Tymf: Tymphristos; Zil: Zileuto.

**Figure 6.** Precipitation lapse rates of Spercheios river basin based on observational meteorological data and E-OBS dataset.

Concerning air temperature, the E-OBS dataset managed to represent the actual measurements efficiently, except in the case of minimum air temperature, based on the higher *MAE* statistics calculated in minimum temperature in all stations (and maximum temperature of Mousounitsa station). The correlation coefficient *R* ranged between moderate (0.49) to very high (0.98) positive [39]; nevertheless, the *p*-value was not statistically significant at the 0.10 level in the case of minimum temperature at Lamia station and at 0.05 level in the case of maximum temperature at Mousounitsa station. Overall, temperature was underestimated (*ME* negative in all cases), especially in the case of minimum air temperature of Lamia station and of minimum and maximum air temperature of Mousounitsa station (Table 6).

**Table 6.** Statistical characteristics and efficient criteria of annual observed air temperature measurements and E-OBS dataset.


Tmin: minimum temperature; Tmax: maximum temperature; Tav: average temperature; *N*: number of observations; *AV*: Average; *ME*: mean error (=E-OBS − station observed values); *MAE*: mean absolute error; *RMSE*: root mean square error; *R*: correlation coefficient.
