*4.3. Annual Validation*

Mean annual values for the study period (see Figure 6) were calculated for both the set of 136 ground stations and the satellite precipitation products (mean values of the available corresponding grid cells within the study area). It is found that, overall, both SP data exhibit an underestimation, although it is lower for IMERG, with the exception of 2015, when a slight overestimation by the IMERG product was noticed. It should be noted, however, that the available precipitation estimates for 2014 were limited to the period April to December, while for 2018 they were limited to January to June.

**Figure 6.** Mean annual values of gauge stations and Satellite Precipitation (SP) data.

Figure 7 displays the variation of overestimation (>100%) or underestimation (<100%) of annual precipitation values at each of the corresponding SP grid cell, between the SP estimated and calculated mean annual gauge station records. We notice that, more or less, the SP estimates have the same behavior regarding the overestimation/ underestimation of annual rainfall. Underestimation is more evident in the central area of Cyprus (greater area of Troodos mountain range), where higher precipitation records are generally noticed, highlighting the known limitations of satellite products regarding heavy rainfall, while the overestimation is noticed in the coastal or rather flat areas, where again the estimation of precipitation still remains a challenge due to the difficulty in distinguishing between rain and non-rain pixels over a complex background [44]. Furthermore, there are no significant differences between the years, since the cells that generally overestimate/underestimate rainfall have a similar performance regardless of the year.

**Figure 7.** *Cont*.

**Figure 7.** Underestimation/Overestimation of the annual satellite product estimates as compared with gauge station records at the corresponding grid cell: (**a**) TRMM—2014 (April–December); (**b**) GPM—2014 (April–December); (**c**) TRMM—2015; (**d**) GPM—2015; (**e**) TRMM—2016; (**f**) GPM—2016; (**g**) TRMM—2017; (**h**) GPM—2017; (**i**) TRMM—2018 (January–June); (**j**) GPM—2018 (January–June).

The performance of the IMERG product was slightly better than 3B43V7 with an R higher value (0.82 and 0.81, respectively), lower Bias values (−17.9 mm/month and −30.5 mm/month, respectively) and rBias values (3.1 and −3.0, respectively), while presenting worst performance in terms of RMSE (108.8 mm/month and 94.7 mm/month, respectively) and MAE (96.1 mm/month and 83.4 mm/month, respectively) values, as presented in Table 4 (minimum and maximum respective values are provided in brackets).


**Table 4.** Performance evaluation metrics of annual data from 3B43V7 and IMERG and (the range of values is given in parentheses).
