1. Introduction
The Mediterranean Sea is recognized as a climatic hotspot [
1], often affected by severe weather events that are becoming more and more frequent in the last decades. Among these events, increasing attention has been recently devoted to the so-called Mediterranean hurricanes (Medicanes) or tropical-like cyclones (TLCs). During the TLC phase, these storms share some features with the well-known tropical cyclones, even if smaller in size. TLCs have a typical diameter of 100–300 km, while the associated surface wind speed can occasionally reach 22–28 ms
−1. They are characterized by the presence of a quasi-cloud-free calm eye, strong winds close to the vortex center, spiral-like cloud bands elongated from the center, vertical alignment of pressure minima, weak vertical wind shear, and a warm anomaly [
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12]. These cyclones may last for several days, although the presence of tropical characteristics is generally limited to a few hours.
In the last years many studies tried to investigate the origin of these events and to link their development to specific meteorological conditions, by using both observations and modeling [
6,
7,
13,
14]. Several studies have discussed the crucial role of the latent heating and the sea surface fluxes of heat and moisture, in conjunction with other factors such as the low-level baroclinicity, the orography, jet stream, and an upper tropospheric potential vorticity (PV) anomaly, for the occurrence, intensity, and movement of Medicanes [
2,
6,
8,
9,
10,
11,
13,
14,
15,
16,
17]. The accepted conceptual model for the intensification and maintenance of Medicanes is similar to that of tropical cyclones, i.e., governed by surface energy fluxes within pre-existing organized cyclonic environments. An upper-level cold trough is required to contribute to cool and moisten the low and mid-tropospheric environment, thus increasing the air–sea gradient of saturation moist static energy [
18].
Therefore, Medicanes can be described as nonfrontal synoptic-scale cyclones, originating in baroclinic environments, associated with a cold trough (or a cut-off low) extending in the Mediterranean and developing, in the mature stage, a warm core due to the release of latent heat associated with convection developing around the pressure minimum. Carrio et al. [
19] analyzing tropicalization process of the 7 November 2014 Mediterranean cyclone, show the importance of the upper-level dynamics to generate a baroclinic environment prone to surface cyclogenesis and in supporting the posterior tropicalization of the system. Some baroclinic cyclones evolve into symmetrical structures during their occlusion phase, with intense latent heat release around their core. The intensity of sea surface heat fluxes, enhanced by the contrast between cold air and a relatively warmer sea surface, is very important for the intensification and maintenance of TLCs. Therefore, differently from tropical latitude hurricanes, cold air aloft is required in order to enhance thermodynamic disequilibrium and trigger the formation of Mediterranean TLCs [
20]. Lower and upper level potential vorticity anomalies associated with TLCs have been identified and characterized by Miglietta et al. [
11] as dry PV, associated with the intrusion of dry stratospheric air, and wet PV generated by latent heat release due to condensation. Such latent heat release at low levels would extend to upper levels only in case of long lasting Medicanes. Therefore, previous studies (e.g. [
2,
21,
22]) have shown that the vertical extension of the warm core varies from the lower to the upper troposphere and is linked to duration of the TLC phase (as also evidenced by Emanuel [
18]).
Medicanes are generally considered as quite rare events. Recently, for the first time, Cavicchia et al. [
23], exploiting high-resolution dynamical downscaling of atmospheric fields obtained from the NCEP/NCAR reanalysis, found systematic and homogeneous statistics of Medicanes, over a six-decades period. Their study estimated an average number of about 1.5 Medicanes per year. The advancements in satellite imagery, along with a comprehensive exploitation of all the available observations, have favored the clear identification of such peculiar events.As highlighted by Cioni et al. [
10], since extratropical and tropical-like cyclone characteristics can alternate or even coexist in the same cyclonic system, satellite data and numerical modeling can be very useful to identify and characterize the development and mature phases of these systems.
In this study we benefit from the unprecedented availability of microwave (MW) satellite observations in the NASA/JAXA Global Precipitation Measurement (GPM) mission era [
24,
25] to characterize and monitor the different development phases of a meteorological event, named Numa by the Free University of Berlin’s Meteorological Institute, that occurred in November 2017 in Southern Italy. The GPM international satellite mission was specifically conceived to unify and advance precipitation measurements from research and operational MW sensors for delivering next-generation global precipitation data products. The GPM constellation consists of a variable and continuously updating number of Low Earth Orbit (LEO) satellites equipped with conical and cross-track scanning radiometers. The main satellite of the constellation, the GPM Core Observatory (GPM-CO), a joint effort by NASA and JAXA, carries an advanced combined active/passive sensor package—the GPM Microwave Imager (GMI), acting also as a calibrator for the other MW constellation radiometers, and the Dual-frequency Precipitation Radar (DPR). The GPM-CO is currently considered the reference satellite-borne platform for the quantitative estimation of precipitation and precipitation microphysics characterization from space. The combined active/passive MW sensor measurements are used to derive consistent precipitation estimates from a constellation of satellites carrying MW radiometers provided by a consortium of international partners. The GPM constellation of satellites currently ensures 1-hourly coverage (on average) over the Mediterranean area.
Panegrossi et al. [
26] illustrated the potential of the GPM constellation for monitoring heavy precipitation events in the Mediterranean region. In their work, the authors also showed the improved capabilities of the GMI with respect to other radiometers to depict the rainband structure of Medicane Qendresa, which occurred on 7–8 November 2014. The study focused on the use of GPM MW radiometers for surface precipitation monitoring, but no DPR overpasses were available for that case. In another study, Marra et al. [
27] carried out an observational analysis of an extremely severe hailstorm, developed over the Mediterranean Sea, and they showed the great capabilities of both GMI and DPR to depict the microphysical structure of the storm and to evidence its unique and rare features compared to global GPM observations. More recently, a preliminary study of GPM capabilities in monitoring the development of Medicane Numa and associated surface precipitation has been carried out by Panegrossi et al. [
28].
The goal of the present paper is to illustrate the potential of the GPM-CO and of the GPM constellation to provide unprecedented insight of Medicane Numa, especially during its offshore development, where no ground-based observations are available. In this respect, since Numa is the first Medicane observed by the DPR, to the best of our knowledge, both passive and active MW measurements are used to infer the structure of the precipitation and to identify the main dynamical and microphysical processes responsible for its changes during the transition from the development to the mature stage. Moreover, the present study illustrates the potential of using GPM precipitation products as a tool to verify and improve Numerical Weather Prediction (NWP) model forecast, particularly useful over the sea.
In the first part of this study we describe the evolution of Numa, from its formation to its dissipation, by using a “traditional” approach suitable for Near Real Time (NRT) applications, i.e., based on visible (VIS) observations, environmental conditions provided by the European Centre for Medium-Range Weather Forecasts (ECMWF), and lightning activity associated with the storm, as measured by the LIghtning NETwork (LINET) [
29].
In the second part of the paper we analyze in detail the added value of the GPM in characterizing the precipitation structure associated with the storm, also thanks to an overpass of the GPM-CO precipitation radar available during its development phase. MW observations and LINET strokes are used to identify specific features during the different stages of the storm evolution. We also compare the satellite-based precipitation rates and pattern to the ones provided by a ground-based radar and NWP simulations (specifically RAMS-ISAC [
30]) for this case. The NWP is also used to support what derived from observations and to highlight the TLC characteristics of Numa. Finally, a sensitivity analysis of the RAMS-ISAC short-term forecast of Numa on the assimilation of DPR observations is also shown.
The paper is organized as follows:
Section 2 describes the sensors and products used in this study, while
Section 3 briefly reviews the RAMS-ISAC model setup. In
Section 4, ECMWF forecast and lightning observations are used to analyze the evolution of Numa.
Section 5 is dedicated to the analysis of GPM observations, while
Section 6 and
Section 7 investigate the performances of RAMS-ISAC model and the impact of DPR assimilation.
Section 8 discusses the main results of the paper. Finally some conclusions are drawn in the last Section.
2. Observational Dataset Description
The observational analysis of Numa has been carried out by using both spaceborne and ground-based data.
Two overpasses by the GPM-CO available for Numa are analyzed in this study. The GPM-CO carries the most advanced spaceborne precipitation MW radiometer (GMI) and the first spaceborne dual-frequency precipitation radar (DPR). The GMI is equipped with 13 different channels: 10 GHz, 19 GHz, 37 GHz, 89 GHz, and 166 GHz (in dual polarization V and H), and 23 GHz, 183.31±3 GHz, and 183.31±7 GHz (V polarization only). The DPR allows for a 3D insight of precipitation structure. Simultaneous observations by the DPR and GMI are only available for the overpass during Numa development phase, because of the narrower radar swath size (245 km for Ku-band, and 125 km for Ka-band) compared to GMI (885 km). Measurements by the other GPM constellation radiometers are also considered. At the time of NUMA, ten satellites, equipped with five different radiometers, were available, providing overpasses at 1-h temporal resolution over the Mediterranean area (on average). In the present paper, besides the GMI, the two other conically scanning radiometers in the constellation are considered: the Advanced Microwave Scanning Radiometer 2 (AMSR2) on board JAXA’s GCOM W1 satellite, providing measurements at spatial resolution comparable to GMI (roughly three times higher than the other radiometers) and the Special Sensor Microwave Imager and Sounder (SSMIS) on board the U.S. DMSP satellites F16, F17, F18. The official NASA/JAXA GPM products [
31] are used in the study, specifically 1C-R and 1C intercalibrated brightness temperatures (TBs), for GMI and the other radiometers respectively, the 2A-CLIM GPROF product for GMI, AMSR2, and SSMIS [
32], and the 2A-DPR product for DPR [
33]. Moreover MODIS Corrected Reflectance (true color) images [
34] are used to describe the upper-level cloud structure during Numa TLC phase. Optimized Coastal Ocean wind product [
35], retrieved from ASCAT scatterometer measurements, is used to analyze the RAMS-ISAC surface wind forecast.
Besides space-based observations, ground-based data are also used in the analysis. In particular, the operational dual-polarization weather radar (hereinafter GR) located in Pettinascura (39.37°N, 16.62°E, ~1700 m asl), Calabria region, Italy, and managed by the Civil Protection Department of Italy, provided measurements during Numa’s mature phase, when it was approaching land. In addition, lightning data from the ground-based low-frequency (VLF/LF, between 1 and 200 KHz) LIghtning NETwork (LINET) are used in this study. The LINET network counts over 550 sensors in several countries around the world, with very good coverage over a wide area in central Europe and central and western Mediterranean (from 10° W to 35° E in longitude and from 30° N to 65° N in latitude). The system measures the time (temporal resolution is about 512 ms), the horizontal and vertical location of VLF sources, their amplitude, and it is also able to discriminate between cloud-to-ground (CG) and intracloud (IC) strokes, and their polarity [
29,
36]. The 3D discrimination method, based on time of arrival (TOA) difference triangulation technique, is reliable (with a position accuracy of 125 m) when the sensor baseline does not exceed ~ 200 km. Moreover LINET allows for estimation of IC emission height, although at least four sensors are needed for a reliable determination of the IC stroke height [
36,
37].
4. Numa Track and Evolution
Numa has been classified in slightly different ways by many international meteorological organizations. According to NOAA/NESDIS it is a hybrid storm, with both tropical and subtropical characteristics. In the analysis of CIMMS, that traces its origin back to the remnants of Tropical Storm Rina, cyclone Numa acquired subtropical characteristics, making it a relatively rare Medicane. In the EUMETSAT webpage dedicated to NUMA ([
48]) the storm is clearly classified as a Medicane.
The storm starts developing on 15 November over the Strait of Sicily and on 16 November, as it moves toward East, it shows hybrid characteristics. Then, during its transition over the Ionian Sea, it reaches its mature phase, and, on 17 November at 12:00 UTC, it hits Apulia Region in Southern Italy, where it persists for 24 h, until 18 November at 12:00 UTC. Here the storm shows typical TLC features, with a quasi-cloud-free calm eye, a warm core, and strong winds close to the vortex center. Then, it moves eastward and, on 19 November, it completely dissipates over Greece.
Figure 1 shows two MODIS Corrected Reflectance (true color) images capturing Numa as it hits Apulia region and evidences how its upper-level cloud structure changes as subtropical and tropical-like cyclone characteristics alternate. On 17 November at 12:00 UTC MODIS Aqua (
Figure 1a) shows a spiral-like cloud band elongated from the center (39.1°N, 18.1°E), with the main rainband in the north hitting the coast of Apulia region, without evidencing an axisymmetric structure or a well identifiable cloud-free eye. On 18 November 2017 at 10:30 UTC, while the storm persists over Southern Apulia region, MODIS Terra (
Figure 1b) reveals a well-defined center (39.4°N, 18.7°E) with a cloud-free eye structure, surrounded by a whirl of clouds that appear deeper in the northeastern sector. At both times the storm is characterized by a closed surface wind circulation and a warm core.
The synoptic environment of Numa both at the surface and 500 hPa is shown by ECMWF operational analyses (
Figure 2). On 16 November at 12:00 UTC (
Figure 2a,b), a wide low pressure area (minimum 1004 hPa) is present over the Strait of Sicily and over the Ionian Sea, with two distinct minima. The mean sea level pressure (mslp) minimum over the Ionian Sea evolves in Numa, while the minimum over the Strait of Sicily is filled in the following hours. The wind at 900 hPa (red arrows) shows a cyclonic circulation over the Strait of Sicily and Ionian Sea, which is not completely closed around the Numa mslp minimum (
Figure 2a). At 500 hPa, the synoptic situation shows a low of geopotential height over Southern Italy with two separate cut-offs. One of them is over the two mslp minima at the surface. Interestingly, the cyclone over the Ionian Sea, evolving in Numa, does not show a warm core surrounded by a close circulation at this stage of development (
Figure 3a).
The well-defined mslp minimum, a warm core, as well as the closed cyclonic circulation at 850 hPa, start becoming evident between 16 November 18:00 UTC and 17 November 00:00 UTC (
Figure 3b).
The mslp on 18 November at 00:00 UTC is shown in
Figure 4a. The cyclone center is located over the Sea, just offshore the southernmost tip of Apulia region. At 500 hPa (
Figure 4b), the geopotential height shows a cut-off and a well-defined cyclonic circulation around its center, which is over the mslp minimum. At this stage of development, Numa clearly shows a warm core at several vertical levels (up to 300 hPa). In particular, at 700 hPa (not shown), the temperature gradient around the Numa center is ~4 °C/120 km.
During the following 12–18 h, the mslp and 500 hPa geopotential height minima were almost stationary in both position and value, causing rainfall and intense winds over Southern Apulia. Thereafter, the storm moved towards Greece while the mslp started to increase.
Figure 5 shows the Numa track constructed from the mslp field provided by ECMWF operational analysis/forecast at 0.125°. It illustrates the storm movement traced using the position of the mslp minima every three hours throughout 15 November to 19 November. During its evolution, Numa shows mslp minimum values ranging from 1006 hPa to 1003 hPa, higher than other Medicanes.
Figure 5 shows also the lightning activity associated with the storm, as monitored by the LINET network. The total (both CG and IC) strokes detected within 200 km radial distance and within 1 h around the 3-hourly mslp minimum are shown. The persistence of the storm over Southern Apulia region is highlighted by the fact that the same position of mslp minima is found three times for two consecutive 3-hour time intervals between 17 November at 12:00 UTC and 18 November at 12:00 UTC.
The temporal evolution of the storm in terms of lightning activity and mslp can be seen more in detail in
Figure 6, showing the temporal evolution of mslp and the corresponding number of total strokes detected within 200 km radial distance and within 1 h around the location and time of the mslp minima. The figure evidences a more intense lightning activity at the initial stage of the storm when strokes are distributed around the cyclone center. This activity reaches its peak on 17 November at 12:00 UTC after the transition of the storm to the TLC phase. It is worth noting, however, as shown in the next section, that the peak is due to strokes registered in the outer regions of the cyclone, while the lightning activity is absent in the main rainbands in proximity of the cyclone center. As Numa persists over Southern Apulia regions, maintaining its TLC structure, and with the minimum mslp persisting for 24 h over the same area, the lightning activity strongly decreases (
Figure 6) because it is confined to the outer regions of the cyclone (at distances > 200 km from the cyclone center). It is worthwhile to note that the LINET detection efficiency is high over the whole domain crossed by the storm throughout its evolution and thus the analysis of lightning activity carried out in this study is reliable.
6. Analysis of RAMS-ISAC Simulations
The forecast of Medicane Numa was characterized by a significant change of the prediction of its position depending on the model initialization time. In an unstable flow regime, such as Numa, the sensitivity of the predicted storm features, for example the position of the cyclone eye, to the initial and boundary conditions is well known [
5,
13]. Nevertheless, in this section we highlight important aspects related to an operational NWP model setup [
13]. Here we analyze the operational RAMS-ISAC forecasts issued on 16 and on 17 November. The first operational forecast uses the Global Forecast System (GFS) analysis/forecast cycle issued at 12:00 UTC on 16 November as initial and boundary conditions, while the second forecast uses the GFS analysis/forecast cycle issued 24 h later at 12:00 UTC on 17 November. GFS initial and dynamic boundary conditions are downloaded at 0.25° horizontal resolution. The boundary conditions are updated every 6 h and the finalization date is +60 h after the initialization time. The sea surface temperature (SST) is interpolated onto the RAMS-ISAC grid from GFS analysis at 0.083° horizontal resolution. SST is constant during the simulation and it is that of the initial time of the simulation (12:00 UTC on 16 and 17 November, depending on the simulation).
It is important to note that we do not consider the impact of rapidly generated sea waves on the sea surface roughness, which plays an important role for the exchange of energy and momentum between the sea surface and the atmosphere. The sea surface roughness is estimated by the Charnock relation [
54], which increases with the square of the friction velocity. So, the sea surface roughness is higher for more intense surface winds. We used rain gauges to independently verify the accuracy of the forecasts. Based on rain gauges, the most abundant precipitation registered over land during the event was in the southern part of the Apulia region (Salento), between 12:00 UTC on 17 November and 12:00 UTC on 18 November. The Italian rain gauge network (
Figure 13a,b) recorded 24 h precipitation varying from 64 mm/24h of Corigliano d’Otranto (40.1°N, 18.3°E) to 172 mm/24h of Presicce (39.9°N, 18.3°E), with the second largest precipitation (147 mm/24h) in Ruffano (40.0°N, 18.3°E). It is worth noting that the mean annual (monthly) precipitation in Salento is ~810 mm/year (~125 mm in November) and that the 17–18 November 2017 event was the most intense in terms of 24h cumulated precipitation since 1970 [
55].
Figure 13c shows the 24 h precipitation predicted between on 17 November at 12:00 UTC and on 18 November at 12:00 UTC for the simulation starting on 16 November at 12:00 UTC. The amount of 100 mm/24h is predicted over the Ionian Sea, with the maximum found 50 km south of the Apulia coast. Precipitation is predicted also over Calabria (39.7°N, 16.8°E) with maximum amounts of 90 mm/24h. While the forecast shows that Numa could be a potentially dangerous meteorological system because, in addition to the high winds, it brings considerable precipitation, the forecast on 16 November shows small precipitation amounts over Apulia, and northeastern Calabria. The situation is different for the forecast issued on 17 November (
Figure 13d); in this forecast, the impact of Numa in Southern Apulia is evident, because RAMS-ISAC shows a maximum of 140 mm/h and 120 mm/24 h in Presicce and Ruffano, respectively, in good agreement with the observations.
Interestingly, the simulation starting at 12:00 on 16 November was able to forecast the TLC features of Numa, similarly to that on 17 November (see Figures 15 and 16) and, considering the characteristics of the storm (low vertical wind shear, quasi-neutral moist environment, sea level pressure minima, amount of precipitation at the surface, etc.), the two forecasts issued on 16 and 17 November are similar. Nevertheless, an important difference arises comparing the storm tracks of the two simulations (
Figure 14). The storm track of the forecast issued on 16 November (R16) is shifted to the south compared with that issued on 17 November (R17). Also, while the R17 storm track is almost stationary for the first 12 h, the R16 center of the storm moves more rapidly towards Greece. The R17 track is in better agreement with that of
Figure 5 which, being derived from ECMWF analysis and short-term forecast, can be considered as the reference. The mslp minima have similar values between R16 and R17 and vary from 1003 hPa to 1006 hPa, in good agreement with values in
Figure 5.
From the above analysis, it follows that the differences between R16 and R17 are mainly in the tracks followed by the TLCs of the two simulations, showing the important role of initial and boundary conditions for the forecast of this kind of events, similarly to other studies ([
5,
13] among others). Therefore, the consequences on the precipitation field forecast over the land are significant.
Since the simulation starting on 17 November at 12:00 UTC provides a good rainfall field forecast over Apulia region, in the following the output of this simulation is used to present some characteristics of Numa TLC phase. An important feature of the Medicanes is the surface wind, which can be intense, causing damages and being a threat for the populations. Satellite observations are an important source of data in this context because they can be used to verify the wind field forecast over the sea, where conventional observations are sparse.
Figure 15 shows the comparison between RAMS-ISAC forecast on 17 November at 19:00 UTC, and the ASCAT observations at 19:09 UTC on the same day. The model is able to predict several features of the storm, such as the eye position (shifted few kilometers to the southwest of the real position) and the strong wind, up to 20 m/s, on the western side of the eye. The main areas of moderate-intense wind over the Adriatic and Ionian Seas are also well predicted even if the model underestimates observed winds. There are also local scale features that are not forecast by the model, as the local wind patterns south of 37°N and east of 18 °E.
As already stated, Numa shows TLC features at this time of development. To highlight these features,
Figure 16a shows the cross-section of relative humidity and temperature anomaly fields along 39.2° N latitude on 17 November at 19:00 UTC. The temperature anomaly is computed with respect to the mean value at each vertical level. The warm core caused by the release of the latent heat of the evolving TLC extends from 1 km to 10 km height (with 1.5 K anomaly between 2 and 5 km height), showing a deep vertical extension of Numa. Relative humidity is lower in the warm core region. It is higher below 2 km height and in the outer regions of the TLC, indicating (a) low-level moisture convergence and (b) moisture redistribution by the convection.
Figure 16b shows the vertical cross-section, at the same latitude of
Figure 16a of the equivalent potential temperature and of the meridional wind speed. Typical features of TLC are apparent as the state of nearly-moist neutrality of ascending air parcels, the limited horizontal extension of the TLC, the axisymmetric shape, the windless region in proximity of cyclone center (the TLC calm eye), and the low vertical wind shear up 6 km height [
5,
7,
13].
The results of
Figure 16a,b indicate that the so-called Wind Induced Surface Heat Exchange (WHISHE [
56,
57]) may play an important role in Numa TLC phase maintenance.
7. The Implementation of DPR (Dual Frequency Precipitation Radar) Reflectivity Observation in RAMS-3Dvar
In this section we show the implementation of DPR 3D reflectivity field as an observation in the RAMS-ISAC data assimilation scheme. The RAMS-3DVar assimilation scheme [
39] modified to assimilate reflectivity factor observed by ground radars [
58] is used for this purpose. The background is given by a short-term forecast of the RAMS-ISAC. More specifically, since the DPR observation is available at 13:49 UTC on 16 November, we use the background at 14:00 UTC neglecting the time difference between the DPR observation and the RAMS-ISAC background. The DPR reflectivity is sampled at seven vertical levels from 2000 m to 8000 m to match the model vertical bins, and is assimilated following Caumont et al. [
59]. Hereafter, the simulation with no DPR assimilation is referred to as CTRL, while the simulation with DPR assimilation is referred to as ANL.
Figure 17a shows the difference between ANL and CTRL forecast of the relative humidity (RH) field at 14:00 UTC, at 3172 m in the terrain following coordinate system used by RAMS-ISAC. There are several areas of the domain where the assimilation increases the relative humidity of the model (RH difference up to 30%). For comparison,
Figure 17b shows the DPR reflectivity at 3000 m at 13:49 UTC, quantifying the impact of the DPR reflectivity assimilation on the RH field. The largest impact on the forecast is introduced south of 36°N, in correspondence of the main “comma”-shaped rain band observed by DPR (see
Figure 10b). The RH perturbation is significant and, once assimilated into the model, it may determine saturation and precipitation where/when introduced. Other perturbations to the RH field are given in Northeastern Sicily and in Central-Eastern Italy.
As stated above, the DPR data are assimilated at 14:00 UTC on 16 November. As shown in the previous section, the forecast starting on 16 November at 12:00 UTC missed the precipitation over Southern Apulia 24–48 h later (between 17 November at 12:00 UTC and 18 November at 12:00 UTC). The DPR reflectivity assimilation is not able to correct the precipitation forecast in Southern Apulia, likely because the (only) DPR overpass is available 24 h before rainfall hits Southern Apulia. This result is similar to other studies with radar data assimilation (see Hu et al. [
60] and Jones et al. [
61]), showing that the resilience of the impact of the assimilation is about 6 h.
Nevertheless, weak effects of assimilating DPR can be seen in the comparison of the CTRL and ANL precipitation rate at 16:20 UTC (
Figure 18), i.e., 2h and 20 minutes after the assimilation time, when an SSMIS overpass is available and captures the storm during its transition over the Ionian Sea. The precipitation rate field is compared to that obtained from the NASA GPROF-CLIM-SSMIS V05 product. The comparison shows important differences between SSMIS precipitation rates and both ANL and CTRL. This difference is caused by both forecast errors, and the low spatial resolution of the GPROF precipitation rate (SSMIS has spatial resolution three times lower than GMI and AMSR2). Nevertheless, the GPROF-CLIM-SSMIS rainfall pattern shows precipitation south of 36°N and west of 18°E, which is not simulated by CTRL forecast.
ANL precipitation, however, extends more to the south compared to CTRL and, in particular, it extends to 35°N latitude and east of 18°E, in better agreement with GPROF-CLIM-SSMIS precipitation pattern. The southward shift of the ANL precipitation is confirmed by the comparison of the CRTL and ANL accumulated precipitation between 14:00 UTC and 18:00 UTC (
Figure 18d). In this case differences up to 7 mm/4h are predicted south of 37°N. The southward extension of the precipitation forecast of ANL compared to CTRL is a consequence of DPR data assimilation, as shown in
Figure 17a.
As shown above, the assimilation of the DPR reflectivity did not have a significant impact on the simulation of TLC Numa. However, the above results show the possibility to implement DPR reflectivity as an observation in data assimilation systems, leaving the question of its impact on the forecast open.
8. Discussion
The analysis carried out for Medicane Numa highlights how powerful MW radiometers are in identifying the details of its precipitation structure, not discernible in conventional (VIS or IR) satellite imagery, as it evolves over the Mediterranean Sea. This is particularly evident for AMSR2 and GMI, thanks to their high spatial resolution compared to the other radiometers. The behavior of the measured TBs (at 37 and 89 GHz for example) is a clear indication that the main precipitation formation mechanism in the storm during its TLC stage, and in particular in the rainbands around the eye, is likely slantwise convection characterized by a weak updraft. Ice hydrometeors show different scattering properties from those found in the high convective cores (high TB minimum at 89 GHz, absence of scattering at 37 GHz, and no electrical activity). As the eye becomes more defined and the cyclone strengthens, the main rainband shows signatures of weak convection at 89 GHz (and at higher frequencies), due to scattering by precipitation-size frozen hydrometeors. However, even in this case, the absence of strong vertical motions inhibits the electrical activity. The analysis of GMI TBs and surface rainfall rate at the development phase evidences stronger convection activity. The minimum Polarization Corrected Temperature (PCT) at 89 GHz, computed according to [
62], is 160.65 K and 195.35 K, in the development phase and in the mature phase of Numa, respectively. The 30 K difference of the 89 GHz PCT minimum values confirms that the convective activity is much weaker in the mature phase than in the development phase, in agreement with what found in previous studies (e.g. [
3]). Comparing these values with TMI climatology of convective systems (e.g., Cecil et al. [
51] and Liu and Zipser [
52]), it is evident that the convection identified during the mature phase can be characterized as “weak”, whereas during its development phase Numa shows features typical of “strong” convection. The lack of scattering signal at higher frequency (> 150 GHz) in correspondence of the MODIS VIS cloud image during the TLC phase, and the depolarization at 89 GHz, indicate the presence of supercooled cloud droplets. The GMI and AMSR2 TB behavior finds correspondence to what found in other studies about in situ observations in tropical cyclones [
63,
64,
65]. They confirm that tropical oceanic convection is usually not known for the strength of its updraft (exceptions occasionally occur), which is required to loft substantial quantities of rain drops and dense graupel to high altitude. In situ observations evidence, instead, the presence of supercooled cloud droplets at high levels and that the ice, mostly rimed particles and low density graupel, forms in the weak updraft regions and is redistributed throughout the storm by the upper and midlevel outflows.
In the development phase, we have analyzed for the first time the 3D structure of this kind of storms, as observed by a spaceborne radar. DPR is particularly valuable because it can provide information on the storm features, from both the microphysical and meteorological perspective, thus supporting what already evidenced by the GMI TB analysis. On the other hand, during the TLC phase, GMI allows highlighting precipitation underestimation by the ground radar of Pettinascura, mainly due to the blind zone caused by its viewing geometry. In fact, the area of Numa, although frequently hit by Medicanes or intense precipitation, is not yet adequately covered by operational weather radars. These results confirm that spaceborne MW observations can provide useful insights about rainfall structure, intensity, and pattern, especially over the sea where no ground-based data are available, and when GR rainfall rate estimates might be affected by large uncertainties (as shown also in Panegrossi et al. [
26]).
The investigation of numerical modeling performance has shown that the RAMS-ISAC forecast for this event is sensitive to the initial and boundary conditions. In particular the forecast starting on 16 November at 12:00 UTC was unable to correctly predict the landfall over Apulia, while the forecast starting on 17 November at 12:00 UTC correctly predicted the rainfall amount and landfall over Apulia. The storm track simulated on 16 November is too far south from the real path, and rainfall was forecast mainly over the sea. The simulation starting on 17 November shows a good prediction of the precipitation field and was used to gain insight into the TLC phase of Numa. Wind speeds up to 20 m/s were simulated on 17 November at 19:00 UTC, in good agreement with ASCAT observations. The warm core extended up to 10 km showing a deep vertical extension of Numa, favored by the long-lasting TLC phase [
18]. Other TLC characteristics as the small horizontal extension of Numa, the low vertical wind shear, the axisymmetric shape, and the calm eye of the Medicane, were well captured by the model. These results, in agreement with previous model-based studies on Medicanes (e.g. Lagouvardos et al. [
2], Miglietta and Rotunno [
5], Davolio et al. [
13], Picornell et al. [
21]), support the passive MW observation analysis, showing that during the TLC phase warm rain processes occur in the area surrounding the eye, while weak convective activity is observed in small portions of the eyewall. We have analyzed how DPR measurements, available over the sea, can be also used in NWP data assimilation. However, because only one overpass of the DPR is available, the impact of the data assimilation on the Numa forecast was limited. Nevertheless, DPR reflectivity data assimilation caused an extension of the rainfall forecast toward the South, in better agreement with NASA GPROF-CLIM-SSMIS V05 product.
9. Conclusions
In this study a Medicane event, named Numa, which occurred in November 2017, has been investigated by using both observations (mainly satellite-based) and numerical modeling. We have stressed how MODIS VIS images, ECMWF forecast and LINET lightning data are very useful for a NRT monitoring of the track of a Medicane like Numa. The main focus of the study, however, is on the added value by the GPM constellation of MW radiometers, that, together with 3D DPR measurements, provides a valuable tool for monitoring and characterizing precipitation features of TLCs, especially during their offshore development, when ground-based observations (rain gauges and radars) cannot be used.
The ability of the different MW frequencies to penetrate the cloud at different heights, and their sensitivity to the horizontal and vertical distribution of liquid and frozen hydrometeors within the cloud, allow not only to observe rainbands and eyewall structure, but also to infer the nature of rainfall formation processes. From the comparison of different MW radiometer overpasses it is possible to identify trends in the TLC and eye development (strengthening or weakening phases), to depict the evolution of precipitation structure and intensity from its development throughout its mature phase, and to localize and characterize the convective activity (as confirmed by ground-based lightning network data). For these reasons, MW radiometers are also a unique tool to verify the ability of cloud resolving models to reproduce the observed structure of the storm. This is particularly effective over the sea where the low frequency channels allow clearly depicting the structure and intensity of the precipitation.
RAMS-ISAC high-resolution simulations support what inferred from the observations, evidencing Numa TLC characteristics (closed circulation around a warm core, low vertical wind shear, intense surface winds, heavy precipitation), persisting for more than 24 h. The first attempt to assimilate the DPR observation in RAMS-ISAC has shown an impact, although weak, on the simulated rainfall rates and amounts. This is because the availability of just one DPR overpass throughout the lifetime of the storm does not allow long-lasting impact on the forecast outcomes.
Standing the actual scarce availability of space radar observation, a future development of this study will include the assimilation of precipitation retrieved by the GPM constellation of MW radiometers.