*4.2. Seasonal Validation*

The results for the seasonal validation are summarized in Table 3. Both TRMM 3B43 and GPM IMERG estimates presented high R values (0.91 and 0.90, respectively) in winter, which suggests a good agreement with ground-based measurements on a seasonal scale. Similar results were found for autumn (0.84 and 0.83, respectively) and spring (0.81 and 0.80, respectively), while the poorest correlation (0.68 and 0.67, respectively) was established in summer.


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

The BIAS for TRMM 3B43 and IMERG ranged from −12.6 to 0.9 and from −8.3 to 2.6, respectively, in the four seasons. The 3B43 overestimated precipitation in spring (0.9) and autumn (0.3), while underestimation is noticed in summer (−1.3) and is rather significant in winter (−12.6). Similar is the pattern for IMERG, with overestimation in spring (0.8) and autumn (2.6), while underestimation is noticed in summer (−1.8) and is rather significant in winter (−8.3).

In terms of rBias, 3B43 presented larger values than that of IMERG in spring (15.0 and 17.5, respectively) and autumn (6.4 and 18.4, respectively), while IMERG showed better performance in winter (−1.1 and −9.9, respectively). Both products displayed their worst values in summer (440.7 for 3B43 and 299.8 for IMERG, respectively).

Precipitation products displayed a similar trend for RMSE and MAE with higher values in winter, spring and autumn and lower values in summer. For winter, spring and autumn, TRMM 3B43 had RMSE values of 33.4, 15.2 and 13.9, which were slightly lower than those of IMERG, which were 38.6, 19.1 and 16.7, respectively. In summer, 3B43 had lower RMSE values, 6.3, compared to those of IMERG, 8.7, respectively. Similar results occurred for MAE values between 3B43 and IMERG, with larger values noticed in winter (33.4 and 38.6, respectively) and lower in summer (3.8 and 4.9, respectively). These results may indicate that both SP products exhibit a similar error level, on a seasonal scale.
