*4.4. Influence of Elevation on Satellite Precipitation Products*

To analyze further the influence of elevation and satellite precipitation products, we grouped all the rain gauge stations into three categories according to their elevation (0–300 m, 300–600 m, >600 m), and compared the evaluation metrics across the different elevation ranges. The annual rBias results for the both GPM and TRMM products are presented in Tables 5 and 6, respectively. Minimum and maximum respective values are provided in brackets along with the corresponding number of satellite products cell for each elevation category.

Results as presented in Tables 5 and 6, respectively, portray that, on an annual scale, both of the two SP products overestimate the precipitation below an altitude of 300 m, with IMERG presenting the largest overestimation (mean annual RB values: RBIMERG = 123.8%, RB3B43 = 107.7%). On the contrary, both SP products underestimate precipitation with increasing elevation, with 3B43v7 displaying a more apparent underestimation than IMERG, with RBIMERG = 93.0%, RB3B43 = 78.0% and RBIMERG = 70.3%, RB3B43 = 65.6% for elevation ranges between 300 and 600 m and >600 m, respectively. Regarding the Pearson correlation coefficient, the performance for 3B43 v7 was slightly better than IMERG at all of the categories, with *r* values increasing with elevation.


**Table 5.** Annual rBias and Pearson correlation coefficient performance for GPM data according to elevation (the range of values is given in parentheses).


**Table 6.** Annual rBias and Pearson correlation coefficient performance for TRMM data according to elevation (the range of values is given in parentheses).

More evaluation metrics (Bias, RMSE, MAE) were used to evaluate the performance of GPM and TRMM monthly data according to elevation (Table 7). Overall, the performance of both satellite products metrics (Bias, RMSE, MAE) were worst in higher altitude areas than in lower altitude areas. Regarding Bias, findings established that TRMM performed better in the elevation range 0–300 m, while GPM exhibited lower bias values in higher altitudes. TRMM exhibited lower RMSE and MAE values than those of GPM in the elevation ranges 0–300 m and >600 m and higher values in the elevation range 300–600 m.

**Table 7.** Metrics (Bias, RMSE, MAE) performance for GMP and TRMM monthly data according to elevation (the range of values is given in parentheses).


The monthly and annual spatio-temporal variations of bias for both satellite precipitation products for the study period are presented in Figures 8 and 9, respectively, while the corresponding seasonal spatio-temporal variation is presented in Figure 10. It is clear that both SP products underestimate precipitation in higher elevation areas and overestimate in areas with lower elevation fluctuations. The underestimation is more evident in the winter.

**Figure 8.** Monthly bias for both precipitation products for the study period: (**a**) TRMM and (**b**) GPM.

**Figure 9.** Annual bias for both precipitation products for the study period: (**a**) TRMM and (**b**) GPM.

**Figure 10.** *Cont*.

These findings are in agreement with previous studies reported in the literature [45–50]. The performance of both SP products could be due to the products themselves and to topography. Both satellite precipitation products combine data from both satellite sensors and ground gauges. Since data from only one gauge station are used in mountainous areas in Cyprus, while three are located in rather flat areas, the accuracy of satellite precipitation products may be affect. Chen and Li [51] and Tang et al. [52] also reported that the accuracy of satellite precipitation products in high mountainous areas in west China could be attributed to the sparse gauge network. Moreover, estimated differences could be also attributed to the differences of the rainfall process, which is rather complicated in mountainous areas than in low altitude areas due to the influence of topography.
