**1. Introduction**

Accurate precipitation measurements are essential for the validation of global climate models and for understanding the natural variability of the earth's climate. Moreover, rainfall monitoring can serve as an important element for risk management of severe precipitation events.

Although the importance of quantitative determination of rainfall is well recognized, reliable retrieval of precipitation is often difficult. First, precipitation represents one of the most difficult atmospheric variables to be accurately measured due to its high temporal and spatial variability. Furthermore, the only instruments that guarantee direct measurements of precipitation are rain gauges and disdrometers. Both types of instruments, although, have a quite high temporal resolution, and provide point-like measurements, ensuring a low spatial resolution. On the other hand, ground-based radars provide measurements of rainfall with a relatively high spatial and temporal resolution. Although they represent a valuable source of information, they provide an indirect measurement of precipitation. In addition, radar observations are affected by several uncertainty sources, including miscalibration, ground clutter, beam blocking, attenuation, Wireless Local Area Network (W-LAN) interferences [1–4].

Space-borne monitoring of clouds and precipitation all around the globe has been gaining growing interest from the international scientific community as a primary contribution to the improvement of global precipitation measurement and to the determination and detection of the global climatic changes. Most of the space-borne monitoring systems take advantage of passive instrumentation, (e.g. radiometers), using both infrared (IR) and microwave (MW) emissions to retrieve cloud properties and precipitation estimation. However, it is difficult to establish an exact quantitative relationship between surface rain rate and the cloud physical quantities (e.g., brightness temperatures) measured by the various sensors [5–8]).

IR-based estimates of rainfall exploit the sensitivity of the IR measurements to the uppermost layers of clouds, but the measured cloud-top brightness temperatures do not provide sufficient information to retrieve the actual intensity of surface rainfall with high reliability. However, the relevance of IR estimates lie in the wide coverage of the earth at relatively high spatial and temporal resolution provided by geosynchronous satellites [9–13]), being IR sensors, mainly mounted on geostationary (GEO) satellites (e.g., the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) onboard of Meteosat Second Generation (MSG) and the Geostationary Operational Environmental Satellite (GOES) Imagers).

However, rainfall estimates based on IR and VIS measurements are constantly evolving thanks also to the improved performance of the sensors. In this regard, it should be noted that the IR and VIS based rainfall retrievals have obtained an important improvement by the exploitation of optical and microphysical clouds parameters (e.g., optical thickness, particle radius), thanks to the higher enhanced spectral resolution of the new generation of geostationary sensors (e.g., MSG SEVIRI and GOES Imagers) [14–18]. In addition, the use of optical and microphysical cloud parameters, the use of classification schemes of convective and stratiform precipitation areas has also contributed to improving the accuracy of rainfall estimates [18,19]. Therefore, while the cloud-top temperature is a primary reference to detect deep convection and precipitation, the use of microphysics parameters and of the cloud classification schemes helps to solve the ambiguities in the retrieval and to identify more accurately the rainy area at the ground [20]. It is also worth mentioning that the combined use of both IR and VIS radiation to provide meteorological products supporting nowcasting activities has been widely studied in the EUMETSAT program—Satellite Application Facilities on Support to Nowcasting and Very Short Range Forecasting (NWC SAF) [20–22]. Furthermore, significant progresses are being made in the field of hyperspectral IR detection and substantial impacts are expected on the Numerical Weather Prediction (NWP) [23–25].

On the other hand, MW-based observations have the great advantage of providing a more direct measurement of the precipitation due to the ability of MW radiation to penetrate precipitating clouds and interact with its liquid and ice hydrometeors [26–30]). At the same time, they suffer of the insufficient temporal frequency of Low Earth Orbit (LEO) satellite overpasses (which carry MW instruments), with respect to the high variability of the precipitation in time and space.

To reduce the evidenced limitations and obtain satisfactory precipitation measurements in terms of accuracy, spatial, and temporal resolution, researchers have increasingly moved to using combinations of sensors. The joint use of MW and IR measurements has long been recognized as very effective as it combines the accuracy of the instantaneous MW data and the repetition and coverage characteristics of the IR geostationary measurements [12,31–34]).

The higher number of LEO-GEO satellites orbiting around the globe has made available a significant amount of precipitation estimates. The availability of these estimates are useful to build accurate and reliable multi-satellite datasets. The goal is to provide products with the best short-range estimates, called High Resolution Precipitation Products (HRPP). The Tropical Rainfall Measuring Mission's (TRMM) Multisatellite Precipitation Analysis (TMPA) was produced according to this line, since it combines precipitation estimates from multiple satellites, as well as from rain gauges, where feasible, to generate rainfall data [35,36].

The Climate Prediction Center morphing method (CMORPH) uses motion vectors from dynamic GEO-IR images to fill the temporal gaps between two available Passive Microwave (PMW) rainfall estimates [37]. The Japanese Global Precipitation Measurement (GPM) standard product Global Satellite Mapping of Precipitation (GSMaP) is a PMW–IR precipitation product. The algorithm integrates PMW data with infrared radiometer data to achieve high temporal and spatial resolution global precipitation estimates [38]. The National Oceanic and Atmospheric Administration (NOAA) Self-Calibrating Multivariate Precipitation Retrieval (SCaMPR) algorithm estimates rainfall at a fine temporal resolution using PMW (SSM/I—-Special Sensor Microwave/Imager) and GEO (GOES) satellites. It uses SSM/I data for rain/no-rain pixels classification, and then GOES data to calibrate the relationship between brightness temperature and rain rate via linear regression for the precipitating pixels [39,40]. The PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks) algorithm of the Center for Hydrometeorology and Remote Sensing (CHRS) is an adaptive, multi-platform precipitation estimation algorithm, based on an artificial neural network approach. It merges high quality data from National Aeronautics and Space Administration (NASA), National Oceanic and Atmospheric Administration (NOAA), and Defense Meteorological Satellite Program (DMSP) low-altitude polar-orbit satellites with sampled data from geosynchronous satellites [41–43]. The Integrated Multi-satellitE Retrievals for GPM (IMERG) is a merged precipitation product developed by the US GPM science team. This algorithm is intended to produce fine timeand space-scale estimates for the entire globe using inter-calibrated, merged, and interpolated data from all available PMW satellites, together with microwave-calibrated infrared (IR) satellite estimates, precipitation gauge analyses, and other precipitation estimators [44].

The combination of MW and IR measurements generally follows two main techniques—the so-called "blended" or "microwave-calibrated" and "morphing". The first one is based on a calibration of IR cloud top temperatures measurements using the MW (namely Passive MW-PMW) precipitation estimates, in order to generate local relationships between the IR and PMW observations [31,32,35,43,45–50]). The derived relationships are then applied to the IR data, increasing the spatial and temporal extent of the precipitation estimation with respect to the PMW overpasses. The "morphing" technique is based on the evidence that IR data, locally updated using PMW-based rainfall measurements, can be employed to measure cloud movement, propagating forward in time the rain field, between the consecutive LEO PMW satellite overpasses [37,51–54]. Basically, this technique derives estimates of precipitation from infrared data when passive microwave information is unavailable.

This paper describes an algorithm, named RAINBOW (RAdar INfrared Blending algorithm for Operational Weather monitoring) combining the data collected by SEVIRI and by the Italian ground-based radars network, coordinated by the Italian Department of Civil Protection (IT GR) to provide precipitation estimation over Italy. The main objective of the algorithm is to provide rainfall estimates from SEVIRI observations, by exploiting the portion of IT GR data with the highest quality. The algorithm has been developed by using the "blended" approach taking using the Surface Rainfall Intensity (SRI) composite product obtained by combining the measurements from all the radars of the network. The Italian ground radar network represents a valuable monitoring system for the detection and warning of severe weather and related hydro-geological risks. As a matter of fact, Italy, and more generally the Mediterranean basin, is affected by severe weather events of different nature (e.g., deep convective systems, cyclones, tropical-like cyclones, etc.) hitting coastal as well as inland areas, causing serious damages and casualties [55–62]).

The IT GR is also currently an important part of the ground reference system for the Precipitation Product Validation Group of the EUMETSAT Satellite Application Facility for Support to Operational Hydrology and Water Management project [63]. However, the spatial heterogeneity of the data quality, related to orography and spatial coverage of the IT GR network, imposes the selection of the data to be used for blending.

The RAINBOW algorithm presented in this paper has been developed within the agreement between the Italian Department of Civil Protection and the Institute of Atmospheric Sciences and Climate (ISAC) of the National Research Council of Italy (CNR). The concept is to design an operational product to complement the radar monitoring of relevant precipitation events by covering both sea areas (not covered by IT GR) and areas where the quality of IT GR data is lower due to limited coverage and orographic obstruction. One of the request that has to be satisfied by RAINBOW is the as short as possible running time in order to provide precipitation estimates as soon as the SEVIRI acquisition is available.

This paper is organized as follows. Section 2 presents the instrumentation and methodology used in the design of the algorithm. Section 3 reports the results obtained by the algorithm when it is applied to selected case studies with the relative discussion. The conclusions are then reported in Section 4.

#### **2. Instrumentation and Methods**

Two-and-a-half years of data (from 1 July 2015 to 31 December 2017) collected by the IT GR network, and by the SEVIRI radiometer, have been used to develop the RAINBOW algorithm. The algorithm combines the SEVIRI brightness temperature and the precipitation rate estimated from the ground radars (GRs) to derive a relationship between these two quantities to be applied to each SEVIRI acquisition (i.e., every fifteen minutes). The area of interest is centered on the Italian peninsula, namely between 36–48◦N and 6–20◦E.
