*2.1. IT GR Network*

At the time of the work, the Italian ground radar (GR) network includes 20 C-band and 3 X-band radar, managed by 11 administrations. Moreover, 7 C-band and 3 X-band systems (all with dual-polarization capability) are managed by the Department of Civil Protection (DPC), which is also the developer and distributor of the national precipitation product. The spatial distribution of the IT GR network with the associated Quality Index (QI) is depicted on Figure 1. The processing architecture is basically composed of two main steps, where the radar measurements are first locally processed by a unique software system, then all the products are centralized to generate the national level products.

There are different sources that can increase the uncertainty in the radar precipitation estimation [64]. The main errors can be identified by contamination by non-weather returns (clutter), partial beam blocking, beam broadening at increasing distances, vertical variability of precipitation [32,65,66], and rain path attenuation [1,67–69]. Due to the morphology of the Italian territory, the uncertainty can be mainly associated to the orography-related effects, especially in southern Italy where the radar coverage as well as the radar overlapping is poor [3,70]. Another error source is the Radio Local Area Network (RLAN) interferences, which are properly dealt with and filtered out using an effective algorithm based on a multi-parameter fuzzy logic approach that also make use of the Signal Quality Index (SQI).

**Figure 1.** Italian ground radar network (IT GR) spatial coverage and its associated Quality Index (QI).

The processing system aims at identifying most of the uncertainty sources in order to compensate them, whenever it is possible, before estimating precipitation. As described in [71], the data quality index results from the combination of the partial QIs associated to each identified error source. A point-by-point description of the operational radar processing chain can be found in [72]. A sensitivity analysis, previously conducted, compared hourly rain gauge and radar data for increasing QI values. The results evidenced that the error (i.e., the difference between radar and rain gauges estimates) has its minimum value for QI = 0.60. At higher QI values, the error increased because of the presence of outliers together to a marked decrease of the sample size [72]. Following this analysis, only QI values equal or greater than 0.60 are considered reliable and are used within RAINBOW algorithm.

Furthermore, a filtering process is applied to the GR data by comparing them with the data collected by the Italian rain gauges network. The radar data showing marked differences with respect to the rain gauges measurements are discarded. The analysis is based on the ratio between the rain gauges and ground radars hourly cumulated data. Namely, the GR data are discarded if the ratio is less than 0.1, being this value chosen because it is much smaller than the average value that the ratio assumes close to the location of a calibrated polarimetric weather radar [73]. At the end of this operational process chain, the Surface Rainfall Intensity (SRI) product is provided over a <sup>1</sup> <sup>×</sup> 1 km<sup>2</sup> grid with a temporal resolution of 10 min. The SRI is obtained taking into account the orography (and the clutter associated), the technical characteristics of the radar (e.g., the various elevation angles and the scanning time frequency, the correction of the partial beam blocking [74,75]. In particular, the single-site SRI is estimated considered the whole radar volume in polar coordinates, then the national composite is computed in Cartesian coordinates. For a given geographical location, the single site SRI is retrieved combining the radar observations at all elevation scans θ*k*, through a quality-weighted average [71,75,76]. Finally, the national SRI composite is built by combining the single-radar rainfall maps through a quality-weighted approach. In case in a given geographical location two or more radar SRI estimates are available, the one with the highest quality weights more.
