2.2.2. Numerical Experiments

Figure 2 illustrates the flow diagrams of the numerical experiments implemented in the current study. Two simulations were performed for each day of the examined events. The first one was carried out without assimilation and refers to the control (CTL) experiment. The second simulation refers to the GNSS ZTD observations assimilation (ZTD) experiment, in which the 3D-var/WRF system under cycling mode was employed. The criterion for selecting the time of ZTD data assimilation for the dry period events was mainly based on the approximate time of the actual start of the convective precipitation and it was set 6 h prior to this time. Thus, data assimilation was applied at 0600 UTC for the May and July cases (Figure 2b), and at 0000 UTC for the June and August cases. For the wet season episodes, the 0000 UTC was chosen as the assimilation hour because the surface low-pressure systems were well developed at that time (Figure 1a).

**Figure 2.** Flow diagrams illustrating the performed WRF simulations without and with ZTD data assimilation for (**a**) 24 h and (**b**) 18 h forecast and verification periods.
