1. Introduction
The signals from Global Navigation Satellite Systems (GNSS) are affected by the atmosphere. These impacts can be seen in the phase observations measured at the GNSS receiver’s location. These atmospheric effects include 1) lower signal propagation speeds in an atmosphere than in a vacuum, which is expressed as a delay in signal’s phase, 2) a signal propagation path diverges from a straight line, which is measured in the form of the signal’s bending angle, which is then transformed into temperature- and water vapor–dependent refractivity profiles. The first effect is mostly reflected in slant total delay (STD), and in the mapped to zenith direction total delay [zenith total delay (ZTD)], calculated from ground-based GNSS measurements (Bevis et al. 1992). In contrast, the second effect occurs in the GNSS signal received by satellites in low-Earth orbit (LEO), such as in GNSS radio occultation (RO) measurements (Kursinski et al. 1997).
Provided precise orbits and clocks, STD and ZTD can be calculated from ground-based GNSS measurements through the use of precise point positioning (PPP) or double-difference processing methods. Alternatively, ZTD and STD can be determined through ray tracing, where the trajectory of the GNSS signal is reconstructed based on meteorological parameters from different data sources such as radiosondes, radiometers, or numerical weather prediction (NWP) models (Deng et al. 2011; Hobiger et al. 2008; Hofmeister 2016; Feng et al. 2020). Tropospheric delays calculated from both GNSS and ray-tracing methods can be used as a tool to study severe weather events as well as validate the forecasting capabilities of weather models (Deng et al. 2011; Li et al. 2015; Ha et al. 2002). Ha et al. (2002) pioneered these techniques. These authors compared slant wet delays (SWD) from an MM5 forecast model with 9-km horizontal resolution to GPS retrievals acquired during a squall line event and showed the potential of GPS observations in capturing abrupt, small-scale variations in moisture. Another study focused on the validation of NWP models using GNSS was carried out by Douša et al. (2016), who showed that ZTD exhibited an accuracy within 1.2 cm in extreme precipitation conditions. Hordyniec et al. (2018) studied the potential impact of hydrometeors on GNSS delays during heavy rainfall events in Poland using GNSS observations and STD calculated from the mesoscale Weather Research and Forecasting (WRF) Model. Their results revealed that GNSS estimates tend to be larger than modeled delays and that there was a positive correlation between delay residuals and rainfall intensity. In a recent study, Lasota et al. (2019) applied a ray-tracing approach to two weather models, WRF, Global Forecast System (GFS); these were used in conjunction with ERA-Interim data during the passage of a tropical cyclone (TC) in Taiwan. In general, it was found that there was a good agreement between GNSS and the STDs calculated from GFS, WRF, and ERA-Interim data. However, a large discrepancy was observed during the passage of the eye of the TC and at mountain stations.
The state of the atmosphere above the GNSS receiver is expressed in terms of ZTD, which can be transformed into integrated water vapor (IWV) values. Several studies have investigated the spatiotemporal patterns of IWV during heavy rainfall, thunderstorms, and TCs (Sapucci et al. 2019; Li and Deng 2013; Iwabuchi et al. 2000; Song and Grejner-Brzezinska 2009; Zhang et al. 2015; Adams et al. 2013, 2015, 2017). A common trend in these studies is that IWV tends to increase and peaks before precipitation occurs; it then decreases after the weather event (Sapucci et al. 2019; Li and Deng 2013; Barindelli et al. 2018; Song and Grejner-Brzezinska 2009). However, the magnitude of the increase, as well as the time offset between the IWV and precipitation peaks, has been known to vary on a case-specific basis. IWV growth observed in association with precipitation forms the basis of heavy rainfall warning systems that utilize IWV thresholds as well as combined instability indices (Priego et al. 2017; Benevides et al. 2015, 2019; Guerova et al. 2019). Recently, Łoś et al. (2020) trained a random forest classifier (Breiman 2001) to nowcast summer storms based on IWV and wet refractivity profiles from GNSS tomography. The model was able to predict storms with an accuracy of over 87%, while the precision of the prediction was slightly less than 30%.
The high accuracy, precision, and vertical resolution of RO observations, as well as its global coverage, make these data well suited for the detection of heavy rainfall and convective storms (Bonafoni et al. 2019). RO refractivity profiles during heavy rainfall exhibit positive biases compared to ERA-Interim data, the European Centre for Medium-Range Weather Forecasts (ECMWF) high-resolution analysis, as well as GFS operational analysis, which suffers from the misrepresentation of moist dynamics (Padullés et al. 2018). Biondi et al. (2012, 2013) demonstrated the feasibility of using the RO bending angle as well as temperature profiles to study the vertical structure of convective clouds and TCs. Positive bending angle anomalies of more than 10% observed in the upper troposphere/lower stratosphere correlate well with the height of cloud tops observed with Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) with a root-mean-squared error (RMSE) of less than 1 km and 365 m for convective storms and TCs, respectively. These large spikes in bending angles were also reflected in the temperature profiles, which corresponded to extreme cold anomalies caused by the convective cloud layer. More recently, Lasota et al. (2018) confirmed that the impact of dense convective clouds in TCs was significant, with positive mean biases of 0.5% and 1.6% in RO refractivity and bending angle measurements, respectively. Neiman et al. (2008) were among the first to exploit RO observations in the study of atmospheric rivers (ARs), which are narrow (around 300–500 km) and elongated (more than 2000 km) corridors of enhanced moisture; they are generally identified by vertically integrated water vapor transport (IVT) readings that exceed a threshold of 250 kg m−1 s−1, which can trigger heavy precipitation upon landfall (Zhu and Newell 1998; Guan and Waliser 2015; Ralph et al. 2019). The results of Neiman et al. (2008) revealed a relationship between the thermodynamic and moisture structures recorded in RO observations, as well as dropsondes, satellite images, and reanalyses. Zechiel and Chiao (2021) presented the changes in water vapor and temperature caused by AR based on 13 years of observations from the Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC). In addition, incorporating RO refractivity and bending angle as parameters into the WRF Model improved its rainfall predictions in terms of both the amount and the location of the precipitation (Huang et al. 2016; Yang et al. 2014). Santhi et al. (2014) developed RO-based instability indices and compared them with radiosonde observations; they showed that there was a high degree of correlation of 0.65 between these two sets of observations during convective activity.
Several studies have demonstrated the benefits of the complementary use of both ground-based and satellite-based GNSS observations in terms of the detection and verification of severe weather events. A recent literature review by Bonafoni et al. (2019) revealed that the extreme weather events most frequently studied with GNSS techniques were convective storms and heavy rainfall. However, the IWV values during extreme weather events reported in the literature are restricted to isolated cases, which limits their nowcasting potential. For example, an investigation into the relationship between precipitation and IWV in Portugal found that the probability of precipitation increased as the peak value of IWV increased, eventually reaching probabilities of greater than 70% for IWV values higher than 42 kg m−2 (Benevides et al. 2015). A regional study in Argentina by Calori et al. (2016) concluded that positive IWV anomalies tended to be associated with hail; however, this relationship was only valid for that particular region. A study conducted in a region in Japan found that there was a time lag between peak IWV values and cloud-ground lightning (Inoue and Inoue 2007). A study by Łoś et al. (2020) reported that the GNSS wet refractivity profiles (up to 7 km) 1–2 h before the thunderstorm were among the most important features for the nowcasting of severe storms. Hence, additional studies are required to define GNSS-based thresholds, indicators, and/or indices that can be used to predict heavy rainfall and convective storms.
Ground-based GNSS techniques have many advantages: they can be operated at low costs, acquire permanent measurements, and have a relatively dense spatial coverage over the Northern Hemisphere. However, ground-based observations are only available over landmasses, and mostly in the form of vertically integrated tropospheric delays which give the total column, or precipitable, water vapor for the locations equipped with GNSS sensors. GNSS RO measurements provide accurate and precise bending angles and refractivity profiles with high vertical resolution and, in contrast to ground-based measurements, are also available over oceanic regions. However, RO measurements are limited by their low number of observations as well as their random distribution around the globe due to the prerequisite geometries required by LEO and GNSS satellites in order for them to record the event, which decreases the feasibility of their use in near-real-time weather forecasting. Therefore, the integration of complementary satellite- and ground-based GNSS observations could potentially be well suited for the monitoring of severe weather events such as thunderstorms and hailstorms.
Within the framework of the aforementioned criteria, this study aims to examine the potential of using combined RO and ground-based GNSS observations, in conjunction with weather models and ray-traced STD, by studying two hailstorms in Bulgaria. Bulgaria has one of the highest mean annual occurrences of thunderstorms and hail events in Europe; they mostly occur between April to September and tend to peak in July (Taszarek et al. 2019). Thus, Bulgaria provides multiple case studies with which the usefulness of GNSS observations in extreme weather detection can be evaluated (Guerova et al. 2019, 2020). In this work, we first focus on the products of GNSS RO observations before exploiting ground-based GNSS observations. We analyze the bending angle anomalies as well as the temperature and specific humidity profiles obtained using RO and compare them to the WRF Model, ERA5 data, and radiosonde profiles from Bulgaria. We then utilize the ground-based GNSS observations in conjunction with ray-traced STD and ZTD values to verify the quality of the WRF Model and ERA5. This paper is divided into six sections: Section 2 gives a brief overview of the datasets used in the study, while the methodology of the STD/ZTD assessment from GNSS observations and NWP models is described in section 3. The results for the two case studies—hailstorms which occurred in July 2014 and May 2019—are presented in sections 4 and 5, respectively. The conclusions and potential avenues for future research are discussed in section 6.
2. Data
a. GNSS ground-based network
In 2018, 12 ground-based GNSS stations were installed in Bulgaria as part of the Balkan–Mediterranean real-time severe weather service (BeRTISS) project (Guerova et al. 2020). The Bulgarian Hail Suppression Agency (HSA) operates nine GNSS stations that provide coverage across northwestern and central Bulgaria. These stations deliver hourly data streams to the operational server at Sofia University and are processed using the Bernese GNSS Software, version 5.2 (Dach et al. 2015) in near-real-time (NRT) mode by the Sofia University GNSS Analysis Center. In this work, the NRT results from two GNSS stations [Popovo (POPO) and Petrovo (PETR)] in central Bulgaria were selected and used to analyze the 2019 hailstorm case study. The GNSS data for the 2014 hailstorm case study were acquired from the GNSS station in Sofia (SOFI), which is part of the International GNSS Service network. GNSS receiver and antenna data, as well as the location of the stations used in this study, are presented in Table 1.
GNSS station locations and equipment.
b. Radio occultation profiles
Radio occultation is a limb-sounding technique that generates meteorological profiles with high vertical resolutions as well as a high degree of accuracy in all-weather conditions. In this study, we utilized the data that were processed by the Wegener Center (WEGC) and distributed in the Occultation Processing System version 5.6.2 (OPSv5.6.2) GNSS RO record (EOPAC Team 2020; Angerer et al. 2017); these data are available on the Earth Observation Data Centre (EODC) web page (EODC 2019). In the WEGC solution, the RO data were combined and processed using a geometric/wave optics approach based on amplitude, excess phase, and orbit information from the University Corporation of Atmospheric Research/COSMIC Data Analysis and Archive Center. The RO refractivity profiles were reconstructed by applying the Abel transform to the bending angle profiles computed in the previous step. Since the integral in the Abel transform requires values up to infinity, collocated ECMWF short-range forecasts and the Mass Spectrometer and Incoherent Scatter Radar model were used as a priori information for its initialization at high altitudes. Using external data implies statistical optimization, which employs an inverse covariance weighting technique (Gobiet et al. 2007). At this point, temperature, pressure, and specific humidity can be estimated using a one-dimensional variational (1DVar) approach based on optimized refractivity profiles and background information from the collocated ECMWF forecasts. The final WEGC product includes the vertical profiles of a suite of variables such as bending angle, refractivity, temperature, or specific humidity; these profiles have a 100-m vertical resolution and global coverage from Earth’s surface to a height of 40 km. Observations from multiple missions were incorporated into the dataset, included data from the Challenging Minisatellite Payload (CHAMP) from 2001 to 2008, the Gravity Recovery and Climate Experiment (GRACE-A and GRACE-B) from 2007 to 2017, Meteorological Operational A (MetOp-A) from 2007 to 2017, MetOp-B from 2013 to 2017, Communications/Navigation Outage Forecasting System (C/NOFS) from 2010 to 2011, and the COSMIC from 2006 to 2019. The number of available observations varied over time. On average, less than 250 RO events were detected between 2001 and 2006. There was a significant increase in recorded measurements after the launch of COSMIC satellites in 2006, which yielded more than 2500 observations a day. Similarly, fewer RO events were occurred after 2010 due to the gradual reduction in the number of COSMIC satellites.
c. ERA5
The ERA5 atmospheric dataset was introduced in 2016 to replace the former commonly used ERA-Interim. The ERA5 is the fifth-generation global atmospheric reanalysis produced by ECMWF (Hersbach et al. 2020). Its main improvements include better quality and accuracy, as well as enhanced horizontal and time resolutions of 0.25° (around 31 km) and 1 h, respectively. In contrast, the preceding ERA-Interim had resolutions of 0.75° (around 80 km) and 6 h, respectively. The ERA5 data are distributed across a total of 37 pressure levels; a majority of these have resolutions of 1) 25 hPa between 1000 and 750 hPa and from 250 to 100 hPa, and 2) 50 hPa between 750 and 250 hPa, while 10 additional layers are irregularly spaced above 100 hPa with the top at 1 hPa. The ERA5 also benefits from revised model physics and data assimilation (DA), which is based on a hybrid incremental 4D variational analysis (VAR) with 12-h windows. The DA system assimilates approximately 24 million measurements per day, of which less than 1 million are from conventional meteorological data sources—the rest are obtained from over 200 satellites. Traditional observations include in situ surface and upper-air measurements, weather buoys, wind profiles, radiosondes, and aircraft. Satellite soundings are composed of ozone observations, microwave and infrared radiances from multiple instruments such as HIRS, MSU, TMI, IASI, and AIRS, ocean‐wave height from altimeters, ocean vector wind and land soil moisture provided by scatterometers, and RO bending angles. Preliminary rapid ERA5 outputs are available with a 5-day delay; these are eventually superseded 2 months later by observations that have been thoroughly checked for quality.
d. WRF simulations
The WRF Model is a state-of-the-art mesoscale NWP system designed for both atmospheric research and operational forecasting applications. We used WRF version 3.7.1 and UEMS version 15.99.8 without DA. The initial and boundary conditions were retrieved from the GFS. The WRF Model is computed for 3 domains with horizontal resolutions of 18, 6, and 3 km, respectively, across 44 vertical levels. The cumulus parameterization scheme (CPS) was Kain–Fritsch (KF), which uses a specified minimum entrainment rate and formulations to allow variability in the cloud radius and cloud-depth threshold for deep (precipitating) and shallow (nonprecipitating) convective clouds. Downdrafts formed from the air in the layer at 150–200 hPa above the cloud base, and they detrain over a fairly deep layer below the cloud base. Downdraft mass flux is estimated as a function of the relative humidity and stability just above cloud base but is not related to vertical wind shear. In fundamental terms, the KF-CPS rearranges mass in a column using the updraft, downdraft, and environmental mass fluxes until at least 90% of the convective available potential energy (CAPE) is removed. The parameterization of cloud microphysics is a sophisticated 5-class scheme that includes ice, snow, and graupel processes; it is suitable for high-resolution real-data simulations, which includes ice sedimentation and time-split fall terms (Lin et al. 1983; Rutledge and Hobbs 1984; Tao et al. 1989; Chen and Sun 2002). Longwave radiation was provided by the Rapid Radiative Transfer Model (Mlawer et al. 1997), while a Goddard shortwave, a two-stream multiband scheme with ozone from climatological and cloud effects, was used to acquire shortwave radiation (Chou and Suarez 1994). Finally, the Noah land surface model (Chen et al. 1996) and the Yonsei University scheme (Hong et al. 2006) were adopted as the land surface and planetary boundary layer models, respectively.
3. Methods
a. GNSS data processing
GNSS delays in zenith and satellite directions were computed for the hailstorm that occurred on 15–16 May 2019, described in further detail in section 5. The ZTDs and consequent STDs for SOFI were obtained following the methodology presented in Hordyniec et al. (2018) and Lasota et al. (2019), and references therein. We used the Bernese GNSS software (version 5.2) in PPP mode to process the GPS-only observations with a cutoff elevation angle that limited the satellite measurements to at least 3° above the horizon (Dach et al. 2015). The algorithm used to determine the GNSS tropospheric products comprised several steps. First, the dependent linear horizontal gradients [G(az, ea)] for ZTD, azimuth (az), and elevation angle (ea) were calculated for each station. Postfit phase residuals were then estimated to acquire STDs that were free from the systematic effects. The processing is briefly described below. The tropospheric delays and gradients were computed at time resolutions of 15 min and 1 h, respectively.
The ZTDs for PETR and POPO were computed using the Bernese GNSS Software (Dach et al. 2015) in double-difference mode as described in Guerova et al. (2020) such that at the final stage of the computation, the ZTDs and ZTD horizontal gradients were estimated hourly using the GMF (Boehm et al. 2006b) and Chen and Herring (Chen et al. 1996) mapping functions, respectively.
The ZTD can be decomposed into three parts: hydrostatic [zenith hydrostatic delay (ZHD)], wet [zenith wet delay (ZWD)], and zenith hydrometeors delay (ZHmD); the latter is usually small and neglected in GNSS processing. However, external knowledge about the ZHD is necessary to disentangle the contributions of the ZHD and the ZWD to the total delay. ZHD is responsible for most of the total delay and can be accurately determined based on the Saastamoinen model, which uses atmospheric pressure at the receiver position as an input (Saastamoinen 1972). The ZWD can be up to 10%–20% of the total delay and is influenced by the highly variable tropospheric water vapor. The least squares approach applied to the iono-free linear combination of the GNSS L1 and L2 carrier-phase in PPP processing allowed for the estimation of the ZWD in parallel with the total tropospheric linear horizontal gradients, which reflect the first-order inhomogeneity of the troposphere.
In this study, the horizontal gradients were computed every hour; this is in contrast to the 15 min resolution of the delays. This was done to reduce the number of estimated parameters and to prevent the large differences in gradient (Meindl et al. 2004). Furthermore, additional postfit phase residuals (RES) that were estimated using the methodology presented in Kačmařík et al. (2017) were incorporated into the estimation of the tropospheric products, which served to mitigate the influence of all unmodelled error sources (Shoji et al. 2004).
b. ZTD and STD ray tracing
An alternative method of using GNSS processing to determine tropospheric delays in zenith and satellite directions is a technique called ray tracing. This procedure allows for the reliable reproduction of the propagating GNSS signal path based on the approximation of geometrical optics and meteorological parameters from NWP models (Hobiger et al. 2008). In this study, we used the 2D piecewise linear (PWL) ray-tracing approach; a detailed description is available in the works of Hobiger et al. (2008), Hofmeister (2016), and Hordyniec et al. (2018). The PWL scheme is one of the simplest and easiest means of implementing ray-tracing strategies. Despite its simplicity, it was used effectively in the investigation of slant delays through an intense tropical cyclone. Our methodology is thus comparable to that used by Lasota et al. (2019). The simplicity of the PWL approach comes from neglecting horizontal gradients of refractivity, which has a small impact compared to vertical gradients. This allows for the components of the horizontal gradient to be set to zero, and the propagation path of the GNSS signal to be limited to the vertical plane of the fixed azimuth.
4. Case study 1: Hailstorm on 8 July 2014
a. Rossby wave breaking
On 8 July 2014, a hailstorm developed over the Sofia plain with hailstones of up to 10 cm in size. According to the national weather archive, this hailstorm was the strongest recorded in the last 75 years and resulted in widespread damage to over 50 000 vehicles and over EUR 123 million in insured losses (Bocheva et al. 2018). To analyze the upper-level flow of the 8 July 2014 hailstorm, maps of geopotential height, temperature, wind, and vorticity at 200 and 300 hPa were selected; these are presented in Fig. 1. From Fig. 1a, it can be seen that at 0600 UTC, a warm air mass (pink–red colors) was situated over the Atlantic Ocean between Iceland and France. Simultaneously, a cold air mass (blue–green colors) was spread across the Scandinavian Peninsula between Poland and Serbia. After 6 h, this upper-level temperature gradient propagated eastward, with the cold air mass moving toward the Black Sea (Fig. 1b). At 1200 UTC, both strong positive and negative absolute vorticity advection at 300 hPa was observed in the western Balkans (Fig. 1c). This was associated with a splitting of the upper-level flow into southeast and west components, indicating a local cyclonic flow (marked with T in Fig. 1d). Interestingly, the thickness maps at 0600 and 1200 UTC 8 July reveal high values over the Scandinavian and Balkan Peninsulas (orange color in Figs. 1e,f) and low values over the Atlantic Ocean and France (green color). A distinct U-shape form of 560- and 564-dam geopotential height lines at 500 hPa is seen over France and Germany in both Figs. 1e and 1f (thick black lines). This results in Rossby wave breaking over Europe and widespread instability and cyclogenesis ahead of the midlevel trough.
Geopotential height and temperature at 200 hPa at (a) 0600 and (b) 1200 UTC 8 Jul. (c) Geopotential and advection of absolute vorticity at 300 hPa and (d) wind and relative vorticity at 300 hPa at 1200 UTC 8 Jul. Sea level pressure isolines (white contours) and 500-hPa geopotential heights (thick black lines) on (e) 0600 and (f) 1200 UTC 8 Jul 2014.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
b. GNSS RO, WRF, and ERA5 profiles on 8 July 2014
Further analysis of atmospheric structure requires detailed profile information; therefore, we used GNSS RO bending angle anomaly profiles obtained between 1 and 8 July 2014 in the Balkan Peninsula. Thirty-four RO profiles covered the area encompassed by longitudes between 15° and 32°E and latitudes between 36° and 46°N. Figure 2 clearly shows that below 6 km, the bending angle variations were within ±15% and had a lower variance between 6 and 10 km. Below 10 km, the RO temperature profiles were within a range of ±4 K. However, between 10 and 14 km, some negative temperature anomalies were observed. These had an average value of −2.8 K (red line in Fig. 2b).
Vertical profiles of (a) bending angle, (b) temperature, and (c) distribution of selected RO profiles from 1 to 8 Jul 2014. The bold red lines in (a) and (b) indicate their mean values, while the yellow circles in (c) indicate additional RO profiles that were investigated.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
Two specific RO profiles from 8 July are further discussed as their spatiotemporal coverage is linked to the hailstorms in Bulgaria. The first profile was collected at 0907 UTC and was located to the west of Bulgaria. Figure 3 shows that there was a positive bending angle anomaly (green line) between 12 and 14 km. This results in a large negative temperature anomaly between 10 and 14 km. The RO temperature profile was consistent with the radiosonde profile at 1200 UTC at Sofia (red line). In addition, it can be clearly seen that both ERA5 and the WRF Model have matching large temperature anomalies at this altitude. However, it should be noted that ERA5 assimilates the RO bending angle, while the WRF Model does not. In contrast, the temperature profile at 1526 UTC still exhibits this negative anomaly, but it is smaller. It should be noted that the RO profile acquired at 1526 UTC was located southwest of Sofia. This elevated cold air pool between 10 and 14 km is also observed in the temperature profiles of both WRF and ERA5. Back trajectory analysis with the Hysplit model (not shown) was carried out to identify the origin of the air mass at 12 km. The path was found to have 1) entered northern Canada on 2 July, 2) passed south of Greenland on 6 July, 3) reached the United Kingdom on 7 July, and 4) passed over Sardinia at 0000 UTC 8 July before reaching Bulgaria at 1200 UTC 8 July. A large positive specific humidity anomaly was also observed in the RO profiles, with maximums at 4 km at 0907 UTC (Fig. 3e, top panel) and around 4.2 km at 1526 UTC (Fig. 3f, bottom panel). This elevated humidity was also observed in the radiosonde profile; however, it should be noted that the RO profile at 1526 UTC had much higher values. Thus, it can be concluded that the observed negative temperature anomaly between 10 and 14 km, in combination with a large positive anomaly of specific humidity in the midtroposphere (4–6 km), are likely contributing to the severity of the storm. Specific humidity profiles from the WRF Model tended to overestimate its values below 4 km. The ERA5 profiles were similar to the observed values at 0900 UTC and tended to underestimate them at 1500 UTC. It should be noted that a large bending angle anomaly up to 20% was observed at an altitude between 4 and 6 km at 1526 UTC and a smaller one up to 10% is seen at 4.2 km at 0907 UTC. The dZTDfrac profiles from WRF and ERA5 at 0900 UTC show a ± 2% agreement with the RO profiles above 3 km (Fig. 3g). At 1500 UTC, the WRF dZTDfrac profile reported an overestimation of over 5% below 4 km and underestimation of 4% between 4 and 6 km (Fig. 3h). The ERA5 dZTDfrac profile recorded underestimates of over 3% for the same altitudes.
Vertical profiles of (a),(b) bending angle, (c),(d) temperature, (e),(f) specific humidity, and (g),(h) fractional ZTD (dZTDfrac, RO minus WRF/ERA5 divided by RO) at (top) 0907 and (bottom) 1526 UTC 8 Jul 2014. RO (green line), RS (red line), WRF (blue line), and ERA5 (orange line) profiles are presented. The gray dotted lines in each panel represent the standard deviation of the RO measurements of that particular climatological parameter.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
The specific humidity profiles at both 0907 and 1526 UTC (Fig. 3) reveal the presence of a positive anomaly at around 4 km. This anomaly was investigated using the IVT maps shown in Fig. 4. High IVT values were recorded at 0900 UTC over Sicily, south Italy, the Adriatic Sea, and the western Balkans. At 1500 UTC, high IVT values were seen to advance toward the Black Sea. Interestingly, the IVT analysis indicated that the high values over the Mediterranean on 8 July can be traced back to an atmospheric river over the North Atlantic identified on 4 July (Fig. 5a). The river was connected to tropical cyclone Arthur (red color in Figs. 5a,b) located on the east coast of North America. The IVT readings on consecutive dates shows the trajectories of both the tropical cyclone and the atmospheric river (Figs. 5c,f).
ERA5 IVT at (a) 0900 and (b) 1500 UTC 8 Jul 2014.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
ERA5 IVT from (a) 1200 UTC 4 Jul to (f) 0000 UTC 7 Jul. White circles denote collocated profiles along the HYSPLIT back-trajectory.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
c. GNSS STDs and WRF/ERA5 ray tracing 8 July 2014
In the next step of the analysis, GNSS STDs from SOFI were compared to the ray-traced STDs from the ERA5 and WRF models. Figure 6 presents the hourly values of the fractional STD differences between GNSS and ERA/WRF (dSTDfrac). An interesting feature is the wet WRF bias that starts at 1500 UTC and lasts for 5 h to 2000 UTC 8 July. This difference can be explained by the vertical structure of the humidity as seen in the RO specific humidity profile at 1526 UTC (Fig. 3f). The WRF Model has a positive anomaly with a peak of 2.6 g kg−1 at 2.5 km. The observed profile had a similar peak, but at 4.6 km. In other words, the vertical humidity structure of the model was offset by about 2 km. Higher specific humidity values at lower altitudes will logically result in larger STDs in the WRF Model. In this respect, the ERA5 humidity profile is more consistent with the RO peaks; however, it tends to underestimate these values by a factor of 2. At 0900 UTC, both WRF and ERA5 had a dry bias, which is likely due to the bias observed in the temperature profile.
Hourly values of dSTDfrac for SOFI station for the WRF Model (blue) and ERA5 (orange).
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
The fractional STD (dSTDfrac, calculated by taking ERA5 minus WRF divided by ERA5) toward each satellite is presented in Fig. 7 and gives a 3D view of the WRF and ERA5 at SOFI station at 0900 and 1500 UTC. As expected, the dSTDfrac at 0900 UTC has a smaller range, and its differences are mainly in the bottom 2–3 km. Interestingly, the largest negative dSTDfrac is toward the G18 satellite (dark blue color). A small positive dSTDfrac of 1% is seen above 6 km. At 1500 UTC, the negative dSTDfrac dominates the lower atmosphere while a small positive dSTDfrac (2%) is observed toward G16 and G31. The 3D structure of the ERA5 and WRF differences is an additional tool that can be used to cross-evaluate the vertical structure of the WRF Model and ERA5.
dSTDfrac for SOFI station on (a) 0900 and (b) 1500 UTC 8 Jul 2014.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
5. Case study 2: Hailstorm on 15–16 May 2019
Detailed description of the atmospheric conditions leading to the widespread hailstorms on 15–16 May 2019 are given by Guerova et al. (2020). Individual convective cells developed in the central and east regions of Bulgaria on 15 and 16 May 2019. On 16 May, the registered radar reflectivity was over 70 dBZ (not shown). Like the hail event that occurred on 8 July 2014 in Sofia, a cold upper-air pool (200 hPa) with a temperature of −60°C was observed over the Black Sea region (Fig. 8b). It should be noted that, over the same region, high thickness values can be seen in Fig. 8, indicating the presence of a warm air mass below 500 hPa associated with a midlevel atmospheric ridge. This suggests that a well-formed cyclone was approaching Bulgaria and that the area with convective cells was situated between the cyclone and the ridge over the Black Sea. Elevated IVT values are observed at 2300 UTC 15 May (Fig. 8c) between the two synoptic systems (i.e., the cyclone and the atmospheric ridge).
(a) Sea level pressure (white contours), 500-hPa geopotential height (black line), and thickness (color shading) at 0600 UTC 16 May. (b) Geopotential height and temperature at 200 hPa at 0600 UTC 16 May 2019. IVT at (c) 2300 UTC 15 May and (d) 0800 UTC 16 May.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
a. GNSS RO, WRF, and ERA5 profiles 16 May 2019
The vertical RO profile at 0808 UTC 16 May 2019 clearly shows the presence of a small positive temperature anomaly (Fig. 9b) between 2 and 10 km as well as a negative temperature anomaly with a peak of −3.8 K at 11.3 km. The temperature anomalies captured by WRF and ERA5 are also negative at 11 km, but the observed positive anomaly was not reproduced. Figure 9c shows that the observed specific humidity had a positive anomaly of up to 2.2 g kg−1 between 2 and 4.5 km. The RO temperature and specific humidity anomalies confirm the observations reported in section 4b of a cold air pool at 11 km in combination with a wet pool in the midtroposphere. The WRF specific humidity anomaly profile is consistent with the observed profiles below 4 km. In contrast, ERA5 underestimates the peak anomaly magnitude by 30%. It should be noted that ERA5 also underestimated specific humidity in the 8 July 2014 hailstorm in Sofia. This is likely related to the coarser horizontal resolution of ERA5.
Vertical anomaly profiles of (a) bending angle, (b) temperature, (c) specific humidity, and (d) dZTDfrac at 0808 UTC 16 May 2019. Profiles from RO (green line), WRF (blue line), and ERA5 (orange line) are presented.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
b. GNSS and WRF/ERA5 ray-traced ZTD on 16 May 2019
Since 2019, the two GNSS stations POPO and PETR have been operating in the region of the convective cells and were used for ZTD analysis. The dZTDfrac for WRF and ERA5 are presented in Fig. 10. At both stations, the WRF Model has a negative bias of more than 1% between 1900 and 2300 UTC 15 May 2019. This ZTD underestimation can be linked to the elevated IVT passing over the region (see Fig. 8c).
Fractional ZTD differences for (a) POPO and (b) PETR stations for the WRF Model (blue) and ERA5 (orange).
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
3D view of fractional STD differences between ERA5 and WRF at 0800 UTC 16 May 2019 for (a) POPO and (b) PETR stations, with GPS satellites labeled in boldface.
Citation: Journal of Atmospheric and Oceanic Technology 39, 5; 10.1175/JTECH-D-21-0100.1
The fractional ZTD difference between WRF, ERA5, and corresponding to the GNSS RO vertical profile at 0800 UTC 16 May 2019 is shown in Fig. 9d. The PETR station (Fig. 10b) is located closer to the RO profile and a negative dZTDfrac bias in the range of 3%–4% is observed between 2 and 4 km; this is also visible in the specific humidity anomaly (Fig. 9c). Above 5 km, the small positive dSTDfrac of approximately 1% presented in Fig. 11b is likely the result of the higher temperature anomaly in the ERA5 compared to WRF (Fig. 9b). No significant differences between WRF and ERA5 profiles were visible in the profiles from POPO station (Fig. 11a).
6. Conclusions
In this manuscript, two severe weather events resulting in hailstorms in Bulgaria were analyzed. A warm air mass between the ground and the midtroposphere (5.5 km) and a cold air pool located at the tropopause level were observed in both cases. The most severe case, which occurred on 8 July 2014, was associated with Rossby wave breaking, widespread instability, and cyclogenesis over the western Balkans. The GNSS RO profiles reveal a cold temperature anomaly between 10 and 14 km on 8 July 2014. In addition, a bending angle anomaly of up to 20% between 4 and 6 km was observed; this was the result of a positive specific humidity anomaly. An IVT analysis indicated that the high specific humidity values over the Mediterranean on 8 July can be traced back to an atmospheric river over the North Atlantic observed on 4 July. This river was connected to the tropical cyclone Arthur on the east coast of the United States. The WRF Model also reported a positive specific humidity anomaly but with an altitude offset of about 2 km. ERA5 specific humidity profile was more consistent with the RO peak but tended to underestimate specific humidity by a factor of 2.
The nonzero fractional STD differences between WRF and ERA5 profiles were mainly observed in the bottom 2–3 km. In the case of the 15 May 2019 hailstorm, elevated IVT values were recorded at 2300 UTC between a cyclone and an atmospheric ridge situated over the Balkan Peninsula. A negative temperature anomaly was observed in the RO vertical profile at 11.3 km before the storm on 16 May 2019; this anomaly was well reproduced by both WRF and ERA5. The observed specific humidity had a positive anomaly between 2 and 4.5 km, which was well captured by WRF but underestimated by ERA5, even though ERA5 assimilated the RO bending angle. A negative fractional ZTD bias of around 3%–4% between 2 and 4 km was confirmed by the differences in the specific humidity profiles generated by WRF and ERA5.
The case studies analyzed and presented in this work are the first to exploit the added value of ground-based and RO GNSS signals for the monitoring of severe weather events in southeast Europe. We conclude that the combination of atmospheric dynamics, water vapor transport, negative upper-air temperature anomalies, and elevated humidity in the midtroposphere are common features of these two storms. In both cases, the WRF Model and ERA5 dataset simulated the upper-air temperature anomaly well but had difficulties reproducing the structure and magnitude of midtroposphere humidity. Bocheva et al. (2018) conducted a numerical experiment on the 8 July hailstorm and reported that the WRF Model generated convection, but with a lower intensity and a temporal offset compared to the observed development of the storm. This shift was observed in the WRF Model generated by ray-traced STDs simulated in this study. Another interesting feature was the dynamics of IVT. Recent studies of IVT and associated atmospheric rivers in the Mediterranean region were focused mainly on precipitation extremes.
Acknowledgments.
We thank the Wegener Center RO and climate research team for providing the WEGC OPSv5.6 RO data, which are available at https://doi.org/10.25364/WEGC/OPS5.6:2020.1. Elżbieta Lasota was conducting a research visit under the Leading Research Groups support project from the subsidy increased for the period 2020–25 in the amount of 2% of the subsidy referred to Art. 387 (3) of the Law of 20 July 2018 on Higher Education and Science, obtained in 2019.
Data availability statement.
The radio occultation profiles can be obtained from WEGC OPSv5.6, available via https://doi.org/10.25364/WEGC/OPS5.6:2020.1. The ERA5 data used in this study are available in the Copernicus Climate Data Store, https://cds.climate.copernicus.eu/. The radiosonde observations can be retrieved from the NOAA/ESRL radiosonde database, https://ruc.noaa.gov/raobs/. GNSS SOFI station data are available via the International GNSS Service analysis center, https://igs.org/products-access/#atmospheric-parameters. GNSS POPO and PETR station data were acquired from third-party providers and are available on request. The WRF Model code, compilation script, initial and boundary condition files, and namelist settings are available at https://github.com/NCAR/WRFV3 and can be reproduced.
REFERENCES
Adams, D. K., S. I. Gutman, K. L. Holub, and D. S. Pereira, 2013: GNSS observations of deep convective time scales in the Amazon. Geophys. Res. Lett., 40, 2818–2823, https://doi.org/10.1002/grl.50573.
Adams, D. K., and Coauthors, 2015: The Amazon Dense GNSS Meteorological Network: A new approach for examining water vapor and deep convection interactions in the tropics. Bull. Amer. Meteor. Soc., 96, 2151–2165, https://doi.org/10.1175/BAMS-D-13-00171.1.
Adams, D. K., H. M. Barbosa, and K. P. Gaitán De Los Ríos, 2017: A spatiotemporal water vapor–deep convection correlation metric derived from the Amazon Dense GNSS Meteorological Network. Mon. Wea. Rev., 145, 279–288, https://doi.org/10.1175/MWR-D-16-0140.1.
Angerer, B., F. Ladstädter, B. Scherllin-Pirscher, M. Schwärz, A. K. Steiner, U. Foelsche, and G. Kirchengast, 2017: Quality aspects of the Wegener Center multi-satellite GPS radio occultation record OPSv5.6. Atmos. Meas. Tech., 10, 4845–4863, https://doi.org/10.5194/amt-10-4845-2017.
Barindelli, S., E. Realini, G. Venuti, A. Fermi, and A. Gatti, 2018: Detection of water vapor time variations associated with heavy rain in northern Italy by geodetic and low-cost GNSS receivers. Earth Planets Space, 70, 28, https://doi.org/10.1186/s40623-018-0795-7.
Benevides, P., J. Catalao, and P. M. A. Miranda, 2015: On the inclusion of GPS precipitable water vapour in the now casting of rainfall. Nat. Hazards Earth Syst. Sci., 15, 2605–2616, https://doi.org/10.5194/nhess-15-2605-2015.
Benevides, P., J. Catalao, and G. Nico, 2019: Neural network approach to forecast hourly intense rainfall using GNSS precipitable water vapor and meteorological sensors. Remote Sens., 11, 966, https://doi.org/10.3390/rs11080966.
Bevis, M., S. Businger, T. A. Herring, C. Rocken, R. A. Anthes, and R. H. Ware, 1992: GPS meteorology: Remote sensing of atmospheric water vapor using the global positioning system. J. Geophys. Res., 97, 15 787–15 801, https://doi.org/10.1029/92JD01517.
Biondi, R., W. J. Randel, S.-P. Ho, T. Neubert, and S. Syndergaard, 2012: Thermal structure of intense convective clouds derived from GPS radio occultations. Atmos. Chem. Phys., 12, 5309–5318, https://doi.org/10.5194/acp-12-5309-2012.
Biondi, R., S.-P. Ho, W. Randel, S. Syndergaard, and T. Neubert, 2013: Tropical cyclone cloud-top height and vertical temperature structure detection using GPS radio occultation measurements. J. Geophys. Res. Atmos., 118, 5247–5259, https://doi.org/10.1002/jgrd.50448.
Bocheva, L., T. Dimitrova, R. Penchev, I. Gospodinov, and P. Simeonov, 2018: Severe convective supercell outbreak over western Bulgaria on July 8, 2014. Idojaras, 122, 177–202, https://doi.org/10.28974/idojaras.2018.2.5.
Boehm, J., B. Werl, and H. Schuh, 2006a: Troposphere mapping functions for GPS and very long baseline interferometry from European Centre for Medium-Range Weather Forecasts operational analysis data. J. Geophys. Res., 111, B02406, https://doi.org/10.1029/2005JB003629.
Boehm, J., A. Niell, P. Tregoning, and H. Schuh, 2006b: Global Mapping Function (GMF): A new empirical mapping function based on numerical weather model data. Geophys. Res. Lett., 33, L07304, https://doi.org/10.1029/2005GL025546.
Bonafoni, S., R. Biondi, H. Brenot, and R. Anthes, 2019: Radio occultation and ground-based GNSS products for observing, understanding and predicting extreme events: A review. Atmos. Res., 230, 104624, https://doi.org/10.1016/j.atmosres.2019.104624.
Breiman, L., 2001: Random forests. Mach. Learn., 45, 5–32, https://doi.org/10.1023/A:1010933404324.
Brenot, H., and Coauthors, 2013: Preliminary signs of the initiation of deep convection by GNSS. Atmos. Chem. Phys., 13, 5425–5449, https://doi.org/10.5194/acp-13-5425-2013.
Calori, A., J. R. Santos, M. Blanco, H. Pessano, P. Llamedo, P. Alexander, and A. de la Torre, 2016: Ground-based GNSS network and integrated water vapor mapping during the development of severe storms at the Cuyo region (Argentina). Atmos. Res., 176–177, 267–275, https://doi.org/10.1016/j.atmosres.2016.03.002.
Chen, F., and Coauthors, 1996: Modeling of land surface evaporation by four schemes and comparison with FIFE observations. J. Geophys. Res., 101, 7251–7268, https://doi.org/10.1029/95JD02165.
Chen, G., and T. A. Herring, 1997: Effects of atmospheric azimuthal asymmetry on the analysis of space geodetic data. J. Geophys. Res., 102, 20 489–20 502, https://doi.org/10.1029/97JB01739.
Chen, S.-H., and W.-Y. Sun, 2002: A one-dimensional time dependent cloud model. J. Meteor. Soc. Japan, 80, 99–118, https://doi.org/10.2151/jmsj.80.99.
Chou, M.-D., and M. J. Suarez, 1994: An efficient thermal infrared radiation parameterization for use in general circulation models. NASA Tech. Memo. 104606, Vol. 3, 85 pp.
Dach, R., S. Lutz, P. Walser, and P. Fridez, 2015: Bernese GNSS software version 5.2. University of Bern.
Davis, J. L., T. A. Herring, I. I. Shapiro, A. E. E. Rogers, and G. Elgered, 1985: Geodesy by radio interferometry: Effects of atmospheric modeling errors on estimates of baseline length. Radio Sci., 20, 1593–1607, https://doi.org/10.1029/RS020i006p01593.
Deng, Z., and Coauthors, 2011: Validation of tropospheric slant path delays derived from single and dual frequency GPS receivers. Radio Sci., 46, RS6007, https://doi.org/10.1029/2011RS004687.
Douša, J., and Coauthors, 2016: Benchmark campaign and case study episode in central Europe for development and assessment of advanced GNSS tropospheric models and products. Atmos. Meas. Tech., 9, 2989–3008, https://doi.org/10.5194/amt-9-2989-2016.
EODC, 2019: Wegener Center GNSS radio occultation record. Accessed 25 November 2020, https://eodc.eu/data/wegener-center-gnss-radio-occultation-record/.
EOPAC Team, 2020: GNSS radio occultation record (OPS 5.6 2001–2019). University of Graz, accessed 25 November 2020, https://doi.org/10.25364/WEGC/OPS5.6:2020.1.
Feng, P., F. Li, J. Yan, F. Zhang, and J.-P. Barriot, 2020: Assessment of the accuracy of the Saastamoinen model and VMF1/VMF3 mapping functions with respect to ray-tracing from radiosonde data in the framework of GNSS meteorology. Remote Sens., 12, 3337, https://doi.org/10.3390/rs12203337.
Foelsche, U., G. Kirchengast, A. K. Steiner, L. Kornblueh, E. Manzini, and L. Bengtsson, 2008: An observing system simulation experiment for climate monitoring with GNSS radio occultation data: Setup and test bed study. J. Geophys. Res., 113, D11108, https://doi.org/10.1029/2007JD009231.
Gobiet, A., G. Kirchengast, G. Manney, M. Borsche, C. Retscher, and G. Stiller, 2007: Retrieval of temperature profiles from CHAMP for climate monitoring: Intercomparison with Envisat MIPAS and GOMOS and different atmospheric analyses. Atmos. Chem. Phys., 7, 3519–3536, https://doi.org/10.5194/acp-7-3519-2007.
Graffigna, V., M. Hernández-Pajares, M. Gende, F. Azpilicueta, and P. Antico, 2019: Interpretation of the tropospheric gradients estimated with GPS during Hurricane Harvey. Earth Space Sci., 6, 1348–1365, https://doi.org/10.1029/2018EA000527.
Guan, B., and D. E. Waliser, 2015: Detection of atmospheric rivers: Evaluation and application of an algorithm for global studies. J. Geophys. Res. Atmos., 120, 12 514–12 535, https://doi.org/10.1002/2015JD024257.
Guerova, G., T. Dimitrova, and S. Georgiev, 2019: Thunderstorm classification functions based on instability indices and GNSS IWV for the Sofia Plain. Remote Sens., 11, 2988, https://doi.org/10.3390/rs11242988.
Guerova, G., T. Dimitrova, K. Vassileva, M. Slavchev, K. Stoev, and S. Georgiev, 2020: Balkan-Med real time severe weather service: Progress and prospects in Bulgaria. Adv. Space Res., 66, 2844–2853, https://doi.org/10.1016/j.asr.2020.07.005.
Ha, S.-Y., Y.-H. Kuo, Y.-R. Guo, C. Rocken, and T. Van Hove, 2002: Comparison of GPS slant wet delay measurements with model simulations during the passage of a squall line. Geophys. Res. Lett., 29, 2113, https://doi.org/10.1029/2002GL015891.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hobiger, T., R. Ichikawa, Y. Koyama, and T. Kondo, 2008: Fast and accurate ray-tracing algorithms for real-time space geodetic applications using numerical weather models. J. Geophys. Res., 113, D20302, https://doi.org/10.1029/2008JD010503.
Hofmeister, A., 2016: Determination of path delays in the atmosphere for geodetic VLBI by means of raytracing. Ph.D. thesis, Dept. of Geodesy and Geoinformation, Technische Universität Wien, 309 pp.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, https://doi.org/10.1175/MWR3199.1.
Hordyniec, P., J. Kapłon, W. Rohm, and M. Kryza, 2018: Residuals of tropospheric delays from GNSS data and ray-tracing as a potential indicator of rain and clouds. Remote Sens., 10, 1917, https://doi.org/10.3390/rs10121917.
Huang, C.-Y., S.-Y. Chen, S. P. Rao Anisetty, S.-C. Yang, and L.-F. Hsiao, 2016: An impact study of GPS radio occultation observations on frontal rainfall prediction with a local bending angle operator. Wea. Forecasting, 31, 129–150, https://doi.org/10.1175/WAF-D-15-0085.1.
Inoue, H. Y., and T. Inoue, 2007: Characteristics of the water-vapor field over the Kanto district associated with summer thunderstorm activities. SOLA, 3, 101–104, https://doi.org/10.2151/sola.2007-026.
Iwabuchi, T., I. Naito, and N. Mannoji, 2000: A comparison of global positioning system retrieved precipitable water vapor with the numerical weather prediction analysis data over the Japanese Islands. J. Geophys. Res., 105, 4573–4585, https://doi.org/10.1029/1999JD901007.
Kačmařík, M., and Coauthors, 2017: Inter-technique validation of tropospheric slant total delays. Atmos. Meas. Tech., 10, 2183–2208, https://doi.org/10.5194/amt-10-2183-2017.
Kačmařík, M., J. Douša, F. Zus, P. Václavovic, K. Balidakis, G. Dick, and J. Wickert, 2019: Sensitivity of GNSS tropospheric gradients to processing options. Ann. Geophys., 37, 429–446, https://doi.org/10.5194/angeo-37-429-2019.
Kursinski, E., G. Hajj, J. Schofield, R. Linfield, and K. R. Hardy, 1997: Observing Earth’s atmosphere with radio occultation measurements using the global positioning system. J. Geophys. Res., 102, 23 429–23 465, https://doi.org/10.1029/97JD01569.
Lasota, E., W. Rohm, C.-Y. Liu, and P. Hordyniec, 2018: Cloud detection from radio occultation measurements in tropical cyclones. Atmosphere, 9, 418, https://doi.org/10.3390/atmos9110418.
Lasota, E., W. Rohm, G. Guerova, and C.-Y. Liu, 2019: A comparison between ray-traced GFS/WRF/ERA and GNSS slant path delays in tropical cyclone Meranti. IEEE Trans. Geosci. Remote Sens., 58, 421–435, https://doi.org/10.1109/TGRS.2019.2936785.
Li, G., and J. Deng, 2013: Atmospheric water monitoring by using ground-based GPS during heavy rains produced by TPV and SWV. Adv. Meteor., 2013, 793957, https://doi.org/10.1155/2013/793957.
Li, X., F. Zus, C. Lu, G. Dick, T. Ning, M. Ge, J. Wickert, and H. Schuh, 2015: Retrieving of atmospheric parameters from multi-GNSS in real time: Validation with water vapor radiometer and numerical weather model. J. Geophys. Res. Atmos., 120, 7189–7204, https://doi.org/10.1002/2015JD023454.
Lin, Y.-L., R. D. Farley, and H. D. Orville, 1983: Bulk parameterization of the snow field in a cloud model. J. Appl. Meteor., 22, 1065–1092, https://doi.org/10.1175/1520-0450(1983)022<1065:BPOTSF>2.0.CO;2.
Łoś, M., K. Smolak, G. Guerova, and W. Rohm, 2020: GNSS-based machine learning storm nowcasting. Remote Sens., 12, 2536, https://doi.org/10.3390/rs12162536.
Mahlstein, I., J. Bhend, C. Spirig, and O. Martius, 2019: Developing an automated medium-range flood awareness system for Switzerland based on probabilistic forecasts of integrated water vapor fluxes. Wea. Forecasting, 34, 1759–1776, https://doi.org/10.1175/WAF-D-18-0189.1.
Meindl, M., S. Schaer, U. Hugentobler, and G. Beutler, 2004: Tropospheric gradient estimation at CODE: Results from global solutions. J. Meteor. Soc. Japan, 82, 331–338, https://doi.org/10.2151/jmsj.2004.331.
Mlawer, E., S. J. Taubman, P. Brown, M. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, https://doi.org/10.1029/97JD00237.
Nafisi, V., M. Madzak, J. Böhm, A. A. Ardalan, and H. Schuh, 2012: Ray-traced tropospheric delays in VLBI analysis. Radio Sci., 47, RS2020, https://doi.org/10.1029/2011RS004918.
Neiman, P. J., and Coauthors, 2008: Diagnosis of an intense atmospheric river impacting the Pacific Northwest: Storm summary and offshore vertical structure observed with COSMIC satellite retrievals. Mon. Wea. Rev., 136, 4398–4420, https://doi.org/10.1175/2008MWR2550.1.
Padullés, R., E. Cardellach, K.-N. Wang, C. O. Ao, F. J. Turk, and M. Torre-Juárez, 2018: Assessment of Global Navigation Satellite System (GNSS) radio occultation refractivity under heavy precipitation. Atmos. Chem. Phys., 18, 11 697–11 708, https://doi.org/10.5194/acp-18-11697-2018.
Priego, E., J. Jones, M. J. Porres, and A. Seco, 2017: Monitoring water vapour with GNSS during a heavy rainfall event in the Spanish Mediterranean area. Geomatics Nat. Hazards Risk, 8, 282–294, https://doi.org/10.1080/19475705.2016.1201150.
Ralph, F. M., J. J. Rutz, J. M. Cordeira, M. Dettinger, M. Anderson, D. Reynolds, L. J. Schick, and C. Smallcomb, 2019: A scale to characterize the strength and impacts of atmospheric rivers. Bull. Amer. Meteor. Soc., 100, 269–289, https://doi.org/10.1175/BAMS-D-18-0023.1.
Rüeger, J., 2002: Refractive index formulae for electronic distance measurement with radio and millimetre waves. University of New South Wales UNISURV Rep. S-68, 13 pp.
Rutledge, S., and P. Hobbs, 1984: The mesoscale and microscale structure and organization of clouds and precipitation in midlatitude cyclones. XII: A diagnostic modeling study of precipitation development in narrow cold-frontal rainbands. J. Atmos. Sci., 41, 2949–2972, https://doi.org/10.1175/1520-0469(1984)041<2949:TMAMSA>2.0.CO;2.
Saastamoinen, J., 1972: Atmospheric correction for the troposphere and stratosphere in radio ranging satellites. The Use of Artificial Satellites for Geodesy,Geophys. Monogr., Vol. 15, Amer. Geophys. Union, 247–251, https://doi.org/10.1029/GM015p0247.
Santhi, Y. D., M. V. Ratnam, S. K. Dhaka, and S. V. Rao, 2014: Global morphology of convection indices observed using COSMIC GPS RO satellite measurements. Atmos. Res., 137, 205–215, https://doi.org/10.1016/j.atmosres.2013.10.002.
Sapucci, L. F., L. A. Machado, E. M. de Souza, and T. B. Campos, 2019: Global positioning system precipitable water vapour (GPS-PWV) jumps before intense rain events: A potential application to nowcasting. Meteor. Appl., 26, 49–63, https://doi.org/10.1002/met.1735.
Shoji, Y., 2013: Retrieval of water vapor inhomogeneity using the Japanese nationwide GPS array and its potential for prediction of convective precipitation. J. Meteor. Soc. Japan, 91, 43–62, https://doi.org/10.2151/jmsj.2013-103.
Shoji, Y., and Coauthors, 2004: Tsukuba GPS dense net campaign observation: Improvement in GPS analysis of slant path delay by stacking one-way postfit phase residuals. J. Meteor. Soc. Japan, 82, 301–314, https://doi.org/10.2151/jmsj.2004.301.
Solheim, F. S., J. Vivekanandan, R. H. Ware, and C. Rocken, 1999: Propagation delays induced in GPS signals by dry air, water vapor, hydrometeors, and other particulates. J. Geophys. Res., 104, 9663–9670, https://doi.org/10.1029/1999JD900095.
Song, D.-S., and D. A. Grejner-Brzezinska, 2009: Remote sensing of atmospheric water vapor variation from GPS measurements during a severe weather event. Earth Planets Space, 61, 1117–1125, https://doi.org/10.1186/BF03352964.
Tao, W., J. Simpson, and M. Mccumber, 1989: An ice-water saturation adjustment. Mon. Wea. Rev., 117, 231–235, https://doi.org/10.1175/1520-0493(1989)117<0231:AIWSA>2.0.CO;2.
Taszarek, M., and Coauthors, 2019: A climatology of thunderstorms across Europe from a synthesis of multiple data sources. J. Climate, 32, 1813–1837, https://doi.org/10.1175/JCLI-D-18-0372.1.
Yang, S.-C., S.-H. Chen, S.-Y. Chen, C.-Y. Huang, and C.-S. Chen, 2014: Evaluating the impact of the COSMIC RO bending angle data on predicting the heavy precipitation episode on 16 June 2008 during SoWMEX-IOP8. Mon. Wea. Rev., 142, 4139–4163, https://doi.org/10.1175/MWR-D-13-00275.1.
Zechiel, P. R., and S. Chiao, 2021: Climate variability of atmospheric rivers and droughts over the west coast of the United States from 2006 to 2019. Atmosphere, 12, 201, https://doi.org/10.3390/atmos12020201.
Zhang, K., T. Manning, S. Wu, W. Rohm, D. Silcock, and S. Choy, 2015: Capturing the signature of severe weather events in Australia using GPS measurements. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 8, 1839–1847, https://doi.org/10.1109/JSTARS.2015.2406313.
Zhu, Y., and R. E. Newell, 1998: A proposed algorithm for moisture fluxes from atmospheric rivers. Mon. Wea. Rev., 126, 725–735, https://doi.org/10.1175/1520-0493(1998)126<0725:APAFMF>2.0.CO;2.