*3.1. Spatial Distribution of Precipitation*

In this section, the differences between the in situ observations and NWP/satellite are shown with a 50 km spatial resolution in Figure 3. In the first flood event in northeastern Iran, the UKMO model outperformed other models and indicated lower precipitation differences with the observations. Moreover, the NCEP underestimated the extreme precipitation over the southeast of the Caspian Sea, while IMERG displayed differences over the southeast and no difference over the southwest part of the Caspian Sea. In the second flood event in western and southwestern Iran, the ECMWF outperformed other products. The NCEP performed better in this flood event than it did in the first flood event. However, all models/satellite, especially UKMO, generally overestimated the precipitation amount. In the third flood event in western Iran, IMERG reasonably captured the spatial precipitation for most parts of the region, except in a few pixels in the west. In addition, IMERG indicated a smoother trend in precipitation differences from the northwest to the eastern part of Iran, while it is evident that other products had larger errors in terms of the difference with the in situ observations.

**Figure 3.** Differences between the accumulated precipitation in situ observations, NWP, and IMERG satellite data.

Table 3 presents the quantitative evaluation of precipitation products. Based on this table, during the first flood event, ECMWF and IMERG were closer to the in situ observation data with respect to the Correlation Coefficient (CC) and maximum precipitation amount. Although the NCEP showed the least correlation in the first event, this model indicated relatively close forecasts to the observations for the second and third events. In general, the ECMWF showed slightly better skill in terms of the correlation, min, and max values of precipitation in comparison with the other products (Table 3).



For a more in-depth evaluation of the spatial pattern, the isohyet contours indicate that for the first flood event, which mostly impacted the Gorganrud Basin in northeast Iran, the UKMO and ECMWF models performed similarly, whereas the UKMO model better forecasted the location of heavy precipitation, followed by the satellite (IMERG). The NCEP model, although recognizing the location of heavy precipitation, underestimated the amount of heavy precipitation (Figure 4).

**Figure 4.** Isohyet contours in the Gorganrud Basin during the first flood event.

According to Figure 5 for the second flood event, the UKMO in the Karkheh Basin yielded better results in comparison with other products in the detection of heavy precipitation. In the Karun Basin, although the UKMO and IMERG products overestimated precipitation, they generally demonstrated better skills than the other two NWP models. The NCEP model in the southeast of the Karun Basin better forecasted heavy precipitation than the other products. With respect to the third flood event in the Karkheh Basin, the UKMO and IMERG showed better performance in the detection of the location of heavy precipitation. In the Karun Basin, although all products showed acceptable performance, the UKMO and ECMWF models outperformed the other products (Figure 6).

**Figure 5.** Isohyet contours in the Karkheh (**top**) and Karun (**bottom**) Basins during the second flood event.

**Figure 6.** Isohyet contours in Karkheh (**top**) and the Karun (**bottom**) Basins during the third flood event.
