Weather radars have deepened our understanding of precipitation microphysics, improved quantitative precipitation estimates, and severe weather forecasts and warnings (Crum and Alberty 1993; Hong and Gourley 2015; Weber et al. 2021; Zhang et al. 2016). In the United States, the well-known Weather Surveillance Radar-1988 Doppler (WSR-88D) network, operated by the National Oceanic and Atmospheric Administration (NOAA), provides a quasi-continental coverage of high-quality observations of weather events (Crum and Alberty 1993). Since 2013, all WSR-88D radars have been upgraded to include polarimetric capabilities, leading to an improved understanding of storm microphysics, the development of polarimetric hydrometeor classification algorithms, and enhanced precipitation retrieval techniques (Ryzhkov and Zrnić 1998; Zhang et al. 2001). However, ground-based radars (denoted as GR herein) suffer from beam blockages, particularly in mountainous regions, that can impede real-time observations of severe weather events (Gabella et al. 2006; Wen et al. 2013). Furthermore, the vertical resolution of operational GRs is degraded by the limited number of elevation angle scans that comprise a radar volume, as well as beam broadening at distant ranges from the radar (Cao et al. 2013; Wen et al. 2013, 2016). In contrast, spaceborne radars (SR), such as the CloudSat and the Global Precipitation Measurement Dual-Frequency Precipitation Radar (GPM DPR), provide datasets that are not affected by the same problems that often impact GR data quality. For example, SRs have a less restricted view of the upper rain column and provide enhanced vertical resolutions when compared to GRs. SRs are also well suited to fill the voids where GR coverage is either missing or poor, such as over the ocean or remote areas on land (Battaglia et al. 2020). On the other hand, the temporal sampling rate provided by a single SR for a given location is quite poor when compared to that provided by a GR. Furthermore, the horizontal resolution provided by an SR is limited by the antenna size while the swath width is restricted by considerations of surface clutter and sampling frequency. Given these different advantages and disadvantages, it is clear that the synergetic use of GRs and SR provides a promising opportunity to probe precipitation microphysics (Liao et al. 2005; Liao and Meneghini 2011; Wen et al. 2013).
After the success of the Tropical Rainfall Measuring Mission (TRMM) in 1997, the GPM Core Observatory (CO)—a collaborative effort by the National Aeronautics and Space Administration (NASA) and Japan Aerospace Exploration Agency (JAXA)—was launched in 2014 with the broad goal of enhancing predictive tools for weather and climate (Hou et al. 2014; Li et al. 2020, 2021, 2022; Foufoula-Georgiou et al. 2020; Wen et al. 2016). The GPM CO carried two advanced instruments: the Microwave Imager (GMI) and the DPR, operating at Ku (13 GHz) and Ka (35 GHz) band, and had a one goal of extending the measurement range of light rain and snowfall to an expanded coverage area from 65°S to 65°N (Skofronick-Jackson et al. 2017).
Since its launch in 2014, the GPM DPR has generated great interest in the community by providing not only the quasi-global observations of the precipitation structure, but also the fine vertical resolution to supplement the poor vertical resolution of GRs. This enhanced vertical resolution is especially important in understanding precipitation microphysics, such as precipitation type, hydrometeor identification, and brightband detection (Cannon et al. 2017; Casella et al. 2017; Iguchi et al. 2018; Le et al. 2016; Toyoshima et al. 2015). For example, Cannon et al. (2017) used the capability of the DPR, vertical pointing GRs, and model simulations to identify the melting layer during atmospheric river events in California. Furthermore, precipitation estimates by DPR also serve as a reference database for the level-3 global seamless gridded precipitation product—Integrated Multi-satellitE Retrievals for GPM (IMERG) (Huffman et al. 2020; Kummerow et al. 2015). Owing to its importance, ground validation (GV) of the GPM DPR have been synchronously initiated around the world (Matsui et al. 2013; Morris and Schwaller 2010; Oki et al. 2020; Schwaller and Morris 2011). The GV program involves a set of instruments and software to retrieve collocated and coincident measurements—such as reflectivity, particle size distribution (PSD), and precipitation estimates—by matching ground and airborne instrumentation to SRs (Houze et al. 2017; Tang et al. 2017). The Olympic Mountains Experiment (OLYMPEX), for instance, deployed over 20 different instruments on multiple aircraft to compare the vertical structure of precipitation in complex terrain (Houze et al. 2017). Precipitation rates estimated by DPR were also compared to the Multi-Radar Multi-Sensor (MRMS) rain data (Wang et al. 2021; Zhang et al. 2016) over the continental United States from 2014 to 2015, where the difference was found to be within 5% for the annual rainfall amount (Oki et al. 2020). However, comparisons of total rainfall amounts do not reveal the capabilities and deficiencies of the GRs and SRs. Numerous studies have focused on comparing the raw radar reflectivities from GRs and SRs and have explored possible reasons for the observed differences, including calibration procedure and methods of attenuation correction (Anagnostou et al. 2001; Biswas and Chandrasekar 2018; Cao et al. 2013; Chandrasekar et al. 2003; Keem et al. 2019; Liao and Meneghini 2022; Masaki et al. 2020; Protat et al. 2022). SRs are used as a valid source to calibrate GRs because of the stability of the DPR calibration over time (Protat et al. 2022; Warren et al. 2018; Masaki et al. 2015, 2020).
Despite the significant collective efforts of the many investigations, few studies have attempted to conduct a large-scale comparison of GPM DPR and WSR 88D three-dimensional radar reflectivity in the United States. A single or even a limited number of sites do not provide a comprehensive understanding of the differences between the two systems or how they might be jointly used to improve our understanding of microphysical processes. Motivated by the need to use the two radar system synergistically, in this study we attempt to map all available DPR data since its launch and all available WSR-88D GRs from 2014 to 2020 to a common grid with the objectives of 1) exploring whether systematic differences exist between the DPR and WSR-88D systems, 2) examining whether those differences are consistent for various microphysical processes, 3) determining the primary factors responsible for any systematic differences, and 4) determining the extent to which the latest DPR (version 7) product improves the GR–SR comparisons. In this paper, the primary comparisons are made with version 6, while version 7 is used to check the degree of improvement in the latest version of the DPR algorithm.
WSR-88D radar
The WSR-88D radar network, operated at the NOAA/National Weather Service (NWS), consists of ∼160 operational radars that guidance for nowcasting and warnings (Ansari et al. 2018). The radars, which operate with frequency between 2.7 and 3 GHz (S band), are routinely calibrated and quality controlled by both online and offline calibrations, with the online calibration being performed automatically while the system is operating and the offline calibration routinely performed as needed by maintenance technicians at preventative maintenance intervals (PMI). The data examined as part of the calibration include system noise, system linearity, transmitter power, klystron delay, antenna gain, and transmitter frequency (Free et al. 2006).
The WSR-88D system generates three-level products for public access. Level 1 consists of raw signal data at each radar site. Level-2 data are processed level-1 data, consisting of range gates that have a gate and azimuthal angle resolution of 250 m and 0.5°, respectively. Examples of level-2 base data include radar base reflectivity (Z), velocity, spectrum width, and polarimetric variables differential reflectivity, specific phase, and correlation coefficient. Level-3 products are generated from level-2 data using advanced algorithms to derive hydrometeor types, precipitation estimates, melting layer location and height. In contrast to level-2 data, level-3 data are averaged into different spatial resolution for each variable. In this study, we retrieved the level-2 base reflectivity and level-3 hydrometeor types archived at the Amazon S3 bucket (https://registry.opendata.aws/noaa-nexrad/) and Google Cloud Storage (https://cloud.google.com/storage/docs/public-datasets/nexrad).
The level-3 hydrometeor-type product used in this study is generated by a fuzzy logic–based hydrometeor classification algorithm (HCA) (www.icams-portal.gov/publications/fmh/FMH11/fmh11partC.pdf) (Park et al. 2009). The HCA has 12 classes, defined according to prescribed thresholds by polarimetric variables. These classes are biological scatters, ground clutter, ice crystals, dry snow, wet snow, light/moderate rain, heavy rain, big drop, graupel, hail mixed with rain, unknown, and no data.
GPM DPR
This study uses the DPR precipitation profile product version 6 (V6 herein) and version 7 (V7 herein), released in 2017 and 2022, respectively (Iguchi and Meneghini 2017, 2021). In this study, the primary assessment is made using the V6 product from 2015 to 2020, while V7 is used to examine the relative improvements from V6 (Iguchi 2020). The DPR consists of the Ku-band precipitation radar (KuPR) and Ka-band precipitation radar (KaPR). KuPR and KaPR in V6 have three scanning modes: normal scan (NS), high-sensitivity scan (HS), and matched scan (MS). The NS is solely applied to KuPR (similar to TRMM PR); hence, it is referred to as KuNS. The HS and MS provide reflectivity estimates based on KaPR; hence, they are referred to as KaHS and KaMS, respectively. The KuNS scans the full swath, while the KaPR (KaMS and KaHS) scans only the inner swath prior to May 2018; after May 2018, the KaHS interleaved scan was shifted to the outer swath so that Ka-band data, matched to the Ku band, became available over the full swath (Liao and Meneghini 2022). The range resolution for KuNS and KaMS is 0.25 km but oversampled at 0.125 km. The KaHS has a range resolution of 0.5 km and a sampling resolution of 0.25 km.
The DPR V7 data, which were released in 2022, had major updates to the data format to reflect the change in the Ka-band scanning. The new data format, called the Full Scan (FS), is used for both frequencies with 176 range-bin arrays and 49 angle-bin arrays. The algorithms include several major improvements, including 1) better sidelobe rejection, 2) application of the dual-frequency technique to the full swath of 49 angle bins, and 3) better rain and brightband (BB) detection (Iguchi 2020; Iguchi et al. 2021). To circumvent the ground clutter problem at off-nadir incidence angles (Kubota et al. 2016), we choose only the 10 cross-track bins centered at nadir. This is similar to the approach by Cannon et al. (2017).
The three-dimensional attenuation-corrected reflectivity (KuNS) in V6/V7 (zFactorCorrected/zFactorFinal) is used in this study for comparison with S-band GR reflectivity. In contrast to the measured reflectivity (zFactorMeasured), the zFactorCorrected/zFactorFinal accounts for the attenuation suffered from clouds, atmospheric gases, and precipitation (Meneghini et al. 2021). In contrast, the S-band WSR-88D radar with its longer wavelength is relatively immune to attenuation from atmospheric signals (Biswas and Chandrasekar 2018). Wen et al. (2011) specifically attributed the bias between KuPR and S-band GR to errors in the KuPR attenuation correction. In addition, the BB height, BB top height, BB bottom height, and precipitation type, documented as “BBheight,” “Bbtop,” “Bbbottom,” “typePrecip,” respectively, are retrieved to break down the differences in microphysical processes. The presence of BB in the classification module is jointly determined by KuPR reflectivity and dual-frequency ratio (DFRm). Readers are referred to Iguchi et al. (2018) and Le et al. (2016) for detailed descriptions. The precipitation type (stratiform, convective, and others) is dependent upon the presence of BB and reflectivity thresholds.
Volume matching
Because of different viewing geometries of the DPR and WSR-88D, a direct comparison of their data are not feasible without resampling one or both datasets. Common approaches to reconcile the two systems differences are 1) mapping GR variables onto 3D cartesian coordinates (Anagnostou et al. 2001; Liao and Meneghini 2009; Schwaller and Morris 2011) and 2) retrieving intersected radar rays called volume matching (Biswas and Chandrasekar 2018; Wen et al. 2011). In this study, we applied the volume-matching method, as illustrated in Fig. 1. One sample bin, highlighted in the red hatched area, is overlaid by both DPR and GR radar beams. To determine the bin height (relative to the surface), the vertical datum of both systems needs to be standardized. We convert the DPR vertical datum—geoid—to the WSR-88D datum [North America Vertical Datum 1988 (NAVD88)]. Practically, the GR gate height is calculated by the radar site elevation plus the relative height determined by the elevation angle and distance to the site. For the DPR, we first calculate the elevation of each bin using the available flight information (ellipsoidal elevation offset, local zenith angle, and bin number) (Iguchi et al. 2018). Next, we convert the height at the elliptical geoid to the height at the NAVD88. For variables inside the common bin, we average numerical values such as the reflectivity (in linear units), and BB height. For categorical data such as hydrometer type, we search for the most dominant type (maximum occurrence) within the volume and assign it in our database. Detailed descriptions of the volume matching method can be found in Morris and Schwaller (2010).
Schematic illustration of 3D volume matching for GPM DPR and ground radar.
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
Event selection
To accelerate data processing and minimize storage issues, we consider only significant precipitation events that appear in the NWS storm database (www.ncdc.noaa.gov/stormevents/) (Li et al. 2022). Only recorded precipitation events that occurred within a range of 25–100 km of the individual radar site are considered; this mitigates the GR blind spot near the radar, the nonuniform beamfilling issue due to beam broadening, and radar wave bending in the stratified atmosphere. In addition, as noted above, an event must be located within the 10 DPR angle bins centered around nadir within 10 min of the satellite overpass. With these restrictions, there are in total over 2 million samples suitable for this study. The number of samples per radar site is shown in Fig. 2. This result, in principle, reflects the precipitation climatology of the continental United States (CONUS), where most of the precipitation falls in the Pacific Northwest and east of the Mississippi. In contrast, arid regions like the Southwest have a limited number of samples. A minimum sample size of 1,000 is required for further analysis.
Number of samples (in thousands) collected at each radar site. Dashed circles represent the detectable radar range (∼300 km).
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
Ku–S-band reflectivity adjustment
Note that all the comparisons in our results were done after adjusting the reflectivity differences.
Results
Systematic view of reflectivity.
The volume matched reflectivities are sampled at each WSR-88D radar site, and the mean differences are illustrated in Fig. 3. Each radar site indicates a positive bias (ZDPR > ZGR) across the CONUS, indicating a consistent and systematic positive difference between the DPR and WSR-88D systems. The average reflectivity difference from the 136 radar sites is 2.4 dB, consistent with the results reported by Biswas and Chandrasekar (2018) in which the differences were found to be between 2 and 3 dB at five GR sites. They attributed the difference to the non-Rayleigh scattering evident in Ku-band DPR wavelength. Keem et al. (2019) reported a 2.27 dB difference at three radar locations in Iowa. They explained that the difference was related to GR wet radome attenuation and beam blockage. In other regions, Warren et al. (2018) found an average 4.0 dB difference between three Australian GRs and DPR, in contrast to 0.1 dB difference for TRMM PR. D’Adderio et al. (2018) recently reported a positive bias ranging from 2.19 to 5.57 dB for the three C-band radars in Italy. Despite the variability in the results, almost all studies indicate a significant positive difference between the DPR V6 and GR.
Spatial map of reflectivity differences (DPR minus GR; in dB) at each WSR-88D site.
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
Figure 4 depicts the grouped reflectivity differences for 136 WSR-88D sites. Overall, the majority of the GRs (77.2%), as depicted by the cyan color in Fig. 4, have absolute differences of less than 3 dB, indicating “good” agreement. Those GRs are mainly located in the central United States, Southeast, and East Coast. Around 19.8% of the GRs fall into the second category (difference of 3–5 dB), scattered over the CONUS. Only four radars have severe differences (>5 dB), all located in the Southern California. The limited sample size for these radars (as shown in Fig. 2) is a possible reason for the large deviations. In general, radars in mountainous regions such as the Rockies disagree with DPR more, partly due to beam blockage that can result in a severe reduction of GR reflectivity values.
Reflectivity differences categorized at three bins: 0–3, 3–5, and 5–9 dB. (a) Scatterplot of DPR reflectivity and GR reflectivity at 140 WSR-88D radar locations and (b) map of WSR-88D locations with color-coded categories. Circles in both panels are color-coded by category.
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
Aggregated by hydrometeor types.
The effective radar reflectivity is determined by the combination of the PSD and the backscattering cross section of the hydrometeors (Chandrasekar et al. 2003; Liao and Meneghini 2009; Wen et al. 2011). We group reflectivity values with respect to eight major hydrometeor types, as shown in Fig. 5. Although the correlation is as high as 0.86 for all types combined, a bias is also obvious (+1.98 dB) with the bulk of the samples above and parallel to the 1:1 line. Of these types, light/moderate precipitation and big drops have high positive biases of 1.94 and 1.86 dB, respectively. Solid precipitation, i.e., dry/wet snow and graupel, have a higher positive bias, when compared to liquid precipitation, ranging from 1.97 to 3.64 dB. This cannot be explained by the difference in the backscattering cross sections at different wavelength nor by signal attenuation (Wen et al. 2011). We suspect the positive bias is related to the different calibration procedures used for the two radar systems. One typical example is in ice crystal or dry snow regions where the DPR signal at the storm top is relatively attenuation-free, implying that calibration is the main sources of the difference (Liao and Meneghini 2009). On the other hand, heavy rain and hail with rain exhibit negative biases (−0.31 and −0.33 dB, respectively), which might be caused by undercorrection of attenuation in heavy rain and hail counteracting the calibration differences.
Reflectivity comparisons by different hydrometeor types at all WSR-88D sites.
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
Comparisons by precipitation types.
Figure 6 depicts the comparisons categorized by different precipitation types as determined by the DPR. Overall, the results grouped by both GR (Fig. 5) and DPR (Fig. 6) depict positive biases of the DPR relative to the GR. It also shows that the reflectivity for the two radars, which is centered around the 1:1 line, aligns reasonably well in convective precipitation. The positive bias is thus reduced to be within 1.5 dB. Nevertheless, the reflectivity in stratiform precipitation starts to deviate, with a positive bias of 1.3 dB. More importantly, the difference becomes greater when the reflectivity values estimated by the GR are taken to be larger than 30 dB, as highlighted in the red box (Fig. 6). To further analyze this example, we separate the stratiform precipitation relative to its location with respect to the BB height as shown in Fig. 7.
Reflectivity comparisons by precipitation types for (a) stratiform precipitation, (b) convective precipitation, and (c) others. The red dashed rectangle in (a) highlights the region where the high-density samples start to deviate from the 1:1 line.
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
As in Fig. 6, but for reflectivity comparisons by height relative to the bright band (BB). (a) Beam height above the BB; (b) beam height within the BB; (c) beam height below the BB.
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
Above the BB, there is a positive bias (1.19 dB) that can be attributed to calibration differences. Within the BB, the bias (1.14 dB) is the smallest among all of the cases, yet the RMSD is the largest. The intrinsic positive bias is possibly compensated by undercorrected attenuation within the BB. Below the BB, the bias increases to 1.6 dB, and samples deviating from the 1:1 line show again the elevated tail highlighted in the red box (Fig. 7), which correlates with the large bias in light/moderate rain in Fig. 5e. Despite a positive bias, the CC (0.89) and RMSD (3.37 dB) below the BB are the best among the metrics in the three regions.
One way to check the accuracy of the DPR attenuation correction is to compare vertical profiles of reflectivity between DPR and GR. If the magnitude of differences between DPR and GR increases and decreases monotonically with the DPR range, it would suggest the DPR undercorrects and overcorrects attenuation, respectively. We regrouped the reflectivity values at bins with height above the ground from 1,500 to 5,700 m in intervals of 200 m. The median value and uncertainty range (25th–75th percentile) are shown from each bin as shown in Fig. 8. Note that since the results integrate and smooth all events that have varying BB heights, a typical stratiform profile is not obvious in this case.
Vertical profile of reflectivity for DPR and GR in (left) stratiform rain and (right) convective rain. The shaded color represents uncertainty range, quantified between the 25th and 75th percentiles.
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
We find a constant shift (∼3 dB) between DPR and GR reflectivity from top to bottom for stratiform precipitation, indicating that attenuation correction of the DPR is reasonable. The reflectivity variability range for DPR begins to narrow at the top (5 dB) yet increases to around 10 dB at the bottom. The range of reflectivity in GR is narrower than that of DPR, despite an increase in the middle, which is possibly related to melting processes. For convective precipitation, however, the bias is around 3 dB at the top and 1 dB at the bottom height. This suggests a slight undercorrection (∼2 dB) of the DPR attenuation for convective precipitation. The reflectivity range, though much greater than stratiform precipitation as expected, follows the same pattern—increasing from the top to the bottom. Again, DPR has a larger reflectivity range than GR, which is probably related to the coarse vertical resolution of GR and signal noise of DPR such as sidelobe, ground clutter, or atmospheric attenuation as signals penetrate deep into the storm.
Comparisons to the NASA Validation Network data.
To confirm our matching methods, we compared our matched results with the NASA Validation Network (VN) data that are publicly accessible (Gatlin et al. 2020; Keem et al. 2019; Schwaller and Morris 2011). We processed volume-matched VN dataset from DPR and GR from 2014 to 2021 at all matched WSR-88D sites (only in the eastern United States). To make a fair comparison, we applied the same reflectivity conversion from Ku band to S band, as mentioned in the “Ku–S-band reflectivity adjustment” section. In contrast to our matching method, VN has stringent data selection protocols. For instance, radar data with nonuniform beamfilling greater than 10% are ignored (Schwaller and Morris 2011). In our case, we select events based on the storm reports as mentioned in the “Event selection” section. In the VN algorithm the radar variables are sampled onto a 3D Cartesian coordinates, while we use a direct volume matching method.
Comparisons of reflectivity from all available sites show positive bias (DPR > GR), with the bias being 1.91 dB for VN data and 2.01 dB for our data (Fig. 9). From Fig. 4b, we see that most of these radar sites have relatively “good” agreement with DPR, with the biases below 3 dB. However, other radar sites in the western United States are not included in the current VN data.
Map of reflectivity bias for (a) NASA network validation data and (b) our method.
Citation: Bulletin of the American Meteorological Society 104, 8; 10.1175/BAMS-D-22-0127.1
Possible reasons for the DPR and GR differences
In this section, we discuss the most probable driving factors that are responsible for the DPR–GR differences. First, the primary factor is believed to be calibration differences as, for example, seen in Fig. 5, where a positive bias is prominent in the ice regions that are attenuation-free. The DPR and WSR-88D radars adopt different calibration protocols. The DPR has implemented the “internal” and “external” calibration to ensure the received signal is correctly converted to reflectivity factor (Masaki et al. 2020; Oki et al. 2020). The internal calibration ensures that power losses from the transmitter to receiver are accounted for, and the external calibration, considering the DPR as a whole system, calibrates signals from ground to the DPR, also called absolute calibration. The internal calibration determines the slope of the transfer function while the external calibration determines the intercept (Masaki et al. 2020). The WSR-88D radars routinely conduct coherent calibration procedures at each radar site, adjusting time-varying system noise, reflectivity correction factors (online calibration), and slowly varying parameters (offline calibration) (Free et al. 2006). Meanwhile, external calibration is conducted using either a self-consistency test (Ryzhkov et al. 2005) or by comparisons to ground-based instruments (e.g., disdrometers) on a regular basis. We found a 1.9 dB shift in the regions above the BB. We suspect that there is a disagreement of the intercept in the transfer function, which is determined by external calibration for both systems.
Second, the attenuation correction to the DPR Ku-band measured reflectivity plays a central role in the reflectivity differences. The latest surface reference technique (SRT) is applied as a primary method in path-integrated attenuation (Meneghini et al. 2021). Although the SRT itself is independent of precipitation phase, the allocation of gate corrected reflectivity entails phase information, which is classified based on DFR. Yet, without attenuation-corrected reflectivity, it is difficult to compute DFR and classify precipitation phase correctly (Iguchi et al. 2018). Although this concern is less severe in stratiform rain where precipitation phase can be inferred given the location of bright band, the issue is more problematic in convective rain as is the NUBF effect that influences all the attenuation estimates. Another method of attenuation correction—the Hitschfield–Bordan method—works well for light rains and small-to-moderate attenuation, yet it still suffers from the phase-state problem in convective rain. In summary, the status quo is that the current algorithm does not handle the attenuation correction issue to the degree that would be desired, and there are ongoing efforts to improve this.
Third, differences in radar scanning and radar characteristics hinder the match between the two. Problems such as NUBF depend on the dimension of the volume chosen to resample radar rays. Particularly for GRs, poor vertical resolution and beam broadening as the distance from the radar increases worsen the NUBF effect, which is further exacerbated in complex terrain (Ryzhkov 2007). Recent studies have been seeking data-driven approaches to mitigate the beam blockage effect in occlusive areas, which have potential to overcome this problem (Yin et al. 2021).
In this study, we provide the first evaluation of the latest version of the DPR algorithm. Similar plots to Figs. 3–8 are shown in the online supplementary Figs. S1–6 (https://doi.org/10.1175/BAMS-D-22-0127.2). We also summarized the improvements of bias in Table 1. A continental map shown in Fig. S1 manifests a good overall improvement showing that the systematic bias between the DPR and GR is reduced in V7. In particular, the average reflectivity difference is reduced from 2.4 dB (V6) to 1.0 dB (V7), where in all cases the DPR exceeds the GR. The number of radar sites that have reflectivity differences within 3 dB has increased from 105 (V6) to 131 (V7), an increase of 24.8% (Fig. S2). The overall bias by hydrometeors has decreased from 1.98 dB (V6) to 1.37 dB (V7), and can be attributed to an improvement in dry snow DPR estimates (RMSD—4.1 versus 3.9 dB; bias—2.23 versus 1.01 dB) and wet snow (RMSD—5.3 versus 3.7 dB; bias—3.64 versus 1.45 dB) (Fig. S3). However, persistent differences in light/moderate rains are present in V7, with 2.01 dB bias comparable to V6 (1.94 dB). Biases with regard to rain type have been decreased in V7 (Fig. S4), and also for heights above BB (Fig. S5). For heights within BB and below BB, there are small increases in biases. The vertical profile shown in Fig. S6 now has a narrower gap (3 dB for V6 versus 0.8 dB for V7) and larger overlapping areas. Overall, the improvements from V6 to V7 are substantial and constitute a closer match between GPM DPR and WSR-88D GR. The improvements are primarily ascribed to 1) advances in path-integrated attenuation (PIA) correction, 2) adoption of an improved sidelobe clutter correction scheme (Kanemaru et al. 2021; Seto et al. 2022), and 3) changes in the DPR solver module including use of a range-adjustable parameter (epsilon).
Summary of bias between GPM DPR version 6 and version 7 (boldface numbers indicate better performance).
The presence of soil moisture can modify the surface backscattering cross section and introduce errors in the estimate of path attenuation. While there is no adjustment in V6, in V7 (Iguchi et al. 2021; Seto et al. 2022), an account of this effect tends to increase the rain-rate estimates over land. Second, the sidelobe clutter correction is improved, especially when a BB is present near the surface. This improved filtering decreases the bias. Third, the updated epsilon parameter affects the drop size distribution and thus the attenuation. Despite overall improvement, there is still a significant difference, when considering that a difference of 1 dB in radar reflectivity can produce a change of 15% in rain rates (Nakamura 2021).
Concluding remarks and recommendations
Ground-based radars in synergy with spaceborne radars can improve our understanding of cloud and precipitation microphysics. Caution should be taken when using the both GPM DPR and WSR-88D radars, however, because systematic differences do exist. In this study, we quantify the reflectivity differences from a systematic overview of 136 WSR-88D radars. We find that, on average, the reflectivities from the DPR are greater than those of the WSR-88D radars by 2.4 dB. At a majority (>77%) of the sites, the difference is within 3 dB. Poorly matched DPR–GRs generally reside either over complex terrain or at the Southern California sites, where the sample size is small. An analysis of the results suggests that the discrepancies originate primarily from three sources: 1) different calibration protocols, 2) errors in the DPR attenuation correction, and 3) imperfect matching methods due to differences in the spaceborne and ground-based radar geometries and beamwidths. NUBF effects further exacerbate the attenuation correction and volume matching problems. In the newer version, V7, the DPR–GR reflectivity bias is reduced from 2.4 dB (V6) to 1.0 dB (V7). This improvement is the result of a better sidelobe rejection method and better attenuation correction procedures implemented in V7. Closer matches between GR and DPR in wet and dry snow are found in V7, yet differences in light/moderate rains persist.
This study aims to highlight the systematic difference between DPR and GR as the interest grows in using a multisensor approach to probe clouds and precipitation microphysics. It does not intend to judge which of the two systems is more accurate because each inherits its own strengths and deficiencies. Any future investigation of this topic would benefit from incorporating more radars from around the world. Moreover, the consistency of the Ku-band SR and S-band GR data over time could be addressed by considering TRMM, GPM, and WSR-88D datasets over the last two decades. Opportunistic sensors such as CloudSat, RainCube, and EarthCARE (to be launched in 2024) will provide more confidence in this task. The triple collation method can be used to examine the similarity and difference among the three independent products without assuming ground truth (Li et al. 2020).
Acknowledgments.
We would like to acknowledge NASA Oklahoma EPSCoR Research Initiation Grant. We would like to thank the support of the GPM Ground Validation Program. The first author is sponsored by the University of Oklahoma Hydrology and Water Security (HWS) program and Graduate College Hoving Fellowship. This work was partially supported by the NOAA/Office of Oceanic and Atmospheric Research under NOAA–University of Oklahoma Cooperative Agreement NA21OAR4320204, U.S. Department of Commerce. We thank NASA Goddard Precipitation Measurement Mission scientist Dr. Toshio Iguchi for his constructive comments. We are also thankful for the support from NASA Marshal Space Flight Center Dr. Patrick Gatlin and Berendes who maintain and provide us VN datasets. Additionally, we thank NOAA/NWS/Radar Operations Center Dr. Dan Berkowitz for his help on WSR-88D radar calibration. We thank three anonymous reviewers for their valuable input to improve the quality of this manuscript.
Data availability statement.
GPM DPR level-2 version 6 and version 7 data can be downloaded from the NASA GES DISC: https://doi.org/10.5067/GPM/DPR/GPM/2A/06 and https://doi.org/10.5067/GPM/DPR/GPM/2A/07. The NEXRAD level-2 base reflectivity and level-3 hydrometeor types archived at the Amazon S3 bucket (https://registry.opendata.aws/noaa-nexrad/) and Google Cloud Storage (https://cloud.google.com/storage/docs/public-datasets/nexrad), respectively.
References
Anagnostou, E. N., C. A. Morales, and T. Dinku, 2001: The use of TRMM Precipitation Radar observations in determining ground radar calibration biases. J. Atmos. Oceanic Technol., 18, 616–628, https://doi.org/10.1175/1520-0426(2001)018<0616:TUOTPR>2.0.CO;2.
Ansari, S. S., and Coauthors, 2018: Unlocking the potential of NEXRAD data through NOAA’s Big Data Partnership. Bull. Amer. Meteor. Soc., 99, 189–204, https://doi.org/10.1175/BAMS-D-16-0021.1.
Battaglia, A., and Coauthors, 2020: Spaceborne cloud and precipitation radars: Status, challenges, and ways forward. Rev. Geophys., 58, e2019RG000686, https://doi.org/10.1029/2019RG000686.
Biswas, S. K., and V. Chandrasekar, 2018: Cross-validation of observations between the GPM Dual-Frequency Precipitation Radar and ground based dual-polarization radars. Remote Sens., 10, 1773, https://doi.org/10.3390/rs10111773.
Cannon, F., F. M. Ralph, A. M. Wilson, and D. P. Lettenmaier, 2017: GPM satellite radar measurements of precipitation and freezing level in atmospheric rivers: Comparison with ground-based radars and reanalyses. J. Geophys. Res. Atmos., 122, 12 747–12 764, https://doi.org/10.1002/2017JD027355.
Cao, Q., Y. Hong, Y. Qi, Y. Wen, J. Zhang, J. J. Gourley, and L. Liao, 2013: Empirical conversion of the vertical profile of reflectivity from Ku-band to S-band frequency. J. Geophys. Res. Atmos., 118, 1814–1825, https://doi.org/10.1002/jgrd.50138.
Casella, D., G. Panegrossi, P. Sano, A. C. Marra, S. Dietrich, B. T. Johnson, and M. S. Kulie, 2017: Evaluation of the GPM-DPR snowfall detection capability: Comparison with CloudSat-CPR. Atmos. Res., 197, 64–75, https://doi.org/10.1016/j.atmosres.2017.06.018.
Chandrasekar, V., H. Fukatsu, and K. Mubarak, 2003: Global mapping of attenuation at Ku- and Ka-band. IEEE Trans. Geosci. Remote Sens., 41, 2166–2176, https://doi.org/10.1109/TGRS.2003.815973.
Crum, T. D., and R. L. Alberty, 1993: The WSR-88D and the WSR-88D Operational Support Facility. Bull. Amer. Meteor. Soc., 74, 1669–1688, https://doi.org/10.1175/1520-0477(1993)074<1669:TWATWO>2.0.CO;2.
D’Adderio, L. P., G. Vulpiani, F. Porcù, A. Tokay, and R. Meneghini, 2018: Comparison of GPM Core Observatory and ground-based radar retrieval of mass-weighted mean raindrop diameter at midlatitude. J. Hydrometeor., 19, 1583–1598, https://doi.org/10.1175/JHM-D-18-0002.1.
Foufoula-Georgiou, E., and Coauthors, 2020: Advancing precipitation estimation, prediction, and impact studies. Bull. Amer. Meteor. Soc., 101, E1584–E1592, https://doi.org/10.1175/BAMS-D-20-0014.1.
Free, A. D., N. K. Patel, and R. W. Macemon, 2006: Open radar data acquisition (ORDA) calibration consistency. 22nd Conf. on Interactive Information Processing Systems for Meteorology, Oceanography, and Hydrology, Atlanta, GA, Amer. Meteor. Soc., 9.3, https://ams.confex.com/ams/Annual2006/webprogram/Paper104423.html.
Gabella, M., J. Joss, G. Perona, and S. Michaelides, 2006: Range adjustment for ground-based radar, derived with the spaceborne TRMM Precipitation Radar. IEEE Trans. Geosci. Remote Sens., 44, 126–133, https://doi.org/10.1109/TGRS.2005.858436.
Gatlin, P. N., W. A. Petersen, J. L. Pippitt, T. A. Berendes, D. B. Wolff, and A. Tokay, 2020: The GPM Validation Network and evaluation of satellite-based retrievals of the rain drop size distribution. Atmosphere, 11, 1010, https://doi.org/10.3390/atmos11091010.
Hong, Y., and J. J. Gourley, 2015: Radar Hydrology: Principles, Models, and Applications. 1st ed. CRC Press, 176 pp., https://doi.org/10.1201/b17921.
Hou, A. Y., and Coauthors, 2014: The Global Precipitation Measurement mission. Bull. Amer. Meteor. Soc., 95, 701–722, https://doi.org/10.1175/BAMS-D-13-00164.1.
Houze, R. A., Jr., and Coauthors, 2017: The Olympic Mountains Experiment (OLYMPEX). Bull. Amer. Meteor. Soc., 98, 2167–2188, https://doi.org/10.1175/BAMS-D-16-0182.1.
Huffman, G. J., and Coauthors, 2020: Integrated Multi-satellitE Retrievals for the Global Precipitation Measurement (GPM) mission (IMERG). Satellite Precipitation Measurement, V. Levizzani et al., Eds., Advances in Global Change Research, Vol. 67, Springer, 343–353.
Iguchi, T., 2020: Dual-Frequency Precipitation Radar (DPR) on the Global Precipitation Measurement (GPM) mission’s Core Observatory. Satellite Precipitation Measurement, V. Levizzani et al., Eds., Advances in Global Change Research, Vol. 67, Springer, 183–192.
Iguchi, T., and R. Meneghini, 2017: GPM DPR precipitation profile L2A 1.5 hours 5 km, version 06. GES DISC, accessed 15 December 2020, https://doi.org/10.5067/GPM/DPR/GPM/2A/06.
Iguchi, T., and R. Meneghini, 2021: GPM DPR precipitation profile L2A 1.5 hours 5 km, version 07. GES DISC, accessed 1 August 2021, https://doi.org/10.5067/GPM/DPR/GPM/2A/07.
Iguchi, T., N. Kawamoto, and R. Oki, 2018: Detection of intense ice precipitation with GPM/DPR. J. Atmos. Oceanic Technol., 35, 491–502, https://doi.org/10.1175/JTECH-D-17-0120.1.
Iguchi, T., and Coauthors, 2021: GPM/DPR level-2 algorithm theoretical basis document. NASA Goddard Space Flight Center Algorithm Theoretical Basis Doc., 238 pp., https://arthurhou.pps.eosdis.nasa.gov/Documents/ATBD_DPR_V07A.pdf.
Kanemaru, K., H. Hanado, and K. Nakagawa, 2021: Improvement of the clutter removal method for the spaceborne precipitation radars. 2021 IEEE Int. Geoscience and Remote Sensing Symp., Brussels, Belgium, IEEE, 840–843, https://doi.org/10.1109/IGARSS47720.2021.9554974.
Keem, M., B. C. Seo, W. F. Krajewski, and K. R. Morris, 2019: Inter-comparison of reflectivity measurements between GPM DPR and NEXRAD radars. Atmos. Res., 226, 49–65, https://doi.org/10.1016/j.atmosres.2019.04.010.
Kubota, T., T. Iguchi, M. Kojima, L. Liao, T. Masaki, H. Hanado, R. Meneghini, and R. Oki, 2016: A statistical method for reducing sidelobe clutter for the Ku-band precipitation radar on board the GPM Core Observatory. J. Atmos. Oceanic Technol., 33, 1413–1428, https://doi.org/10.1175/JTECH-D-15-0202.1.
Kummerow, C. D., D. L. Randel, M. Kulie, N. Wang, R. Ferraro, S. J. Munchak, and V. Petkovic, 2015: The evolution of the Goddard profiling algorithm to a fully parametric scheme. J. Atmos. Oceanic Technol., 32, 2265–2280, https://doi.org/10.1175/JTECH-D-15-0039.1.
Le, M., V. Chandrasekar, and S. Biswas, 2016: Evaluation and validation of GPM dual-frequency classification module after launch. J. Atmos. Oceanic Technol., 33, 2699–2716, https://doi.org/10.1175/JTECH-D-15-0253.1.
Li, Z., M. Chen, S. Gao, Z. Hong, G. Tang, Y. Wen, J. J. Gourley, and Y. Hong, 2020: Cross-examination of similarity, difference and deficiency of gauge, radar and satellite precipitation measuring uncertainties for extreme events using conventional metrics and multiplicative triple collocation. Remote Sens., 12, 1258, https://doi.org/10.3390/rs12081258.
Li, Z., and Coauthors, 2021: Two-decades of GPM IMERG early and final run products intercomparison: Similarity and difference in climatology, rates, and extremes. J. Hydrol., 594, 125975, https://doi.org/10.1016/j.jhydrol.2021.125975.
Li, Z., G. Tang, P. Kirstetter, S. Gao, J. F. Li, Y. Wen, and Y. Hong, 2022: Evaluation of GPM IMERG and its constellations in extreme events over the conterminous United States. J. Hydrol., 606, 127357, https://doi.org/10.1016/j.jhydrol.2021.127357.
Liao, L., and R. Meneghini, 2009: Validation of TRMM Precipitation Radar through comparison of its multiyear measurements with ground-based radar. J. Appl. Meteor. Climatol., 48, 804–817, https://doi.org/10.1175/2008JAMC1974.1.
Liao, L., and R. Meneghini, 2011: A study on the feasibility of dual-wavelength radar for identification of hydrometeor phases. J. Appl. Meteor. Climatol., 50, 449–456, https://doi.org/10.1175/2010JAMC2499.1.
Liao, L., and R. Meneghini, 2022: GPM DPR retrievals: Algorithm, evaluation, and validation. Remote Sens., 14, 843, https://doi.org/10.3390/rs14040843.
Liao, L., R. Meneghini, T. Iguchi, and A. Detwiler, 2005: Use of dual-wavelength radar for snow parameter estimates. J. Atmos. Oceanic Technol., 22, 1494–1506, https://doi.org/10.1175/JTECH1808.1.
Masaki, T., and Coauthors, 2015: Current status of GPM/DPR level 1 algorithm development and DPR calibration. 2015 IEEE Int. Geoscience and Remote Sensing Symp., Milan, Italy, IEEE, 2615–2618, https://doi.org/10.1109/IGARSS.2015.7326348.
Masaki, T., T. Iguchi, K. Kanemaru, K. Furukawa, N. Yoshida, T. Kubota, and R. Oki, 2020: Calibration of the dual-frequency precipitation radar onboard the Global Precipitation Measurement Core Observatory. IEEE Trans. Geosci. Remote Sens., 60, 5100116, https://doi.org/10.1109/TGRS.2020.3039978.
Matsui, T., and Coauthors, 2013: GPM satellite simulator over ground validation sites. Bull. Amer. Meteor. Soc., 94, 1653–1660, https://doi.org/10.1175/BAMS-D-12-00160.1.
Meneghini, R., H. Kim, L. Liao, J. Kwiatkowski, and T. Iguchi, 2021: Path attenuation estimates for the GPM Dual-Frequency Precipitation Radar. J. Meteor. Soc. Japan, 99, 181–200, https://doi.org/10.2151/jmsj.2021-010.
Morris, K. R., and M. R. Schwaller, 2010: Data visualization and analysis tools for the Global Precipitation Measurement (GPM) Validation Network. 2010 IEEE Int. Geoscience and Remote Sensing Symp., Honolulu, HI, IEEE, 847–850, https://doi.org/10.1109/IGARSS.2010.5654248.
Nakamura, K., 2021: Progress from TRMM to GPM. J. Meteor. Soc. Japan, 99, 697–729, https://doi.org/10.2151/jmsj.2021-035.
Oki, R., T. Iguchi, and K. Nakamura, 2020: The GPM DPR validation program. Satellite Precipitation Measurement, V. Levizzani et al., Eds., Advances in Global Change Research, Vol. 69, Springer, 503–514, https://doi.org/10.1007/978-3-030-35798-6_3.
Park, H. S., A. V. Ryzhkov, D. S. Zrnić, and K. Kim, 2009: The hydrometeor classification algorithm for the polarimetric WSR-88D: Description and application to an MCS. Wea. Forecasting, 24, 730–748, https://doi.org/10.1175/2008WAF2222205.1.
Protat, A., V. Louf, J. Soderholm, J. Brook, and W. Ponsonby, 2022: Three-way calibration checks using ground-based, ship-based, and spaceborne radars. Atmos. Meas. Tech., 15, 915–926, https://doi.org/10.5194/amt-15-915-2022.
Ryzhkov, A. V., 2007: The impact of beam broadening on the quality of radar polarimetric data. J. Atmos. Oceanic Technol., 24, 729–744, https://doi.org/10.1175/JTECH2003.1.
Ryzhkov, A. V., and D. S. Zrnić, 1998: Discrimination between rain and snow with a polarimetric radar. J. Appl. Meteor., 37, 1228–1240, https://doi.org/10.1175/1520-0450(1998)037<1228:DBRASW>2.0.CO;2.
Ryzhkov, A. V., S. E. Giangrande, V. M. Melnikov, and T. J. Schuur, 2005: Calibration issues of dual-polarization radar measurements. J. Atmos. Oceanic Technol., 22, 1138–1155, https://doi.org/10.1175/JTECH1772.1.
Schwaller, M. R., and K. R. Morris, 2011: A ground Validation Network for the Global Precipitation Measurement mission. J. Atmos. Oceanic Technol., 28, 301–319, https://doi.org/10.1175/2010JTECHA1403.1.
Seto, S., T. Iguchi, and R. Meneghini, 2022: Correction of path-integrated attenuation estimates considering the soil moisture effect for the GPM Dual-Frequency Precipitation Radar. J. Atmos. Oceanic Technol., 39, 803–821, https://doi.org/10.1175/JTECH-D-21-0111.1.
Skofronick-Jackson, G., and Coauthors, 2017: The Global Precipitation Measurement (GPM) mission for science and society. Bull. Amer. Meteor. Soc., 98, 1679–1695, https://doi.org/10.1175/BAMS-D-15-00306.1.
Tang, G., Y. Wen, J. Gao, D. Long, Y. Ma, W. Wan, and Y. Hong, 2017: Similarities and differences between three coexisting spaceborne radars in global rainfall and snowfall estimation. Water Resour. Res., 53, 3835–3853, https://doi.org/10.1002/2016WR019961.
Toyoshima, K., H. Masunaga, and F. A. Furuzawa, 2015: Early evaluation of Ku- and Ka-band sensitivities for the Global Precipitation Measurement (GPM) Dual-Frequency Precipitation Radar (DPR). SOLA, 11, 14–17, https://doi.org/10.2151/sola.2015-004.
Wang, J., W. A. Petersen, and D. B. Wolff, 2021: Validation of satellite-based precipitation products from TRMM to GPM. Remote Sens., 13, 1745, https://doi.org/10.3390/rs13091745.
Warren, R. A., A. Protat, S. T. Siems, H. A. Ramsay, V. Louf, M. J. Manton, and T. A. Kane, 2018: Calibrating ground-based radars against TRMM and GPM. J. Atmos. Oceanic Technol., 35, 323–346, https://doi.org/10.1175/JTECH-D-17-0128.1.
Weber, M., and Coauthors, 2021: Towards the next generation operational meteorological radar. Bull. Amer. Meteor. Soc., 102, E1357–E1383, https://doi.org/10.1175/BAMS-D-20-0067.1.
Wen, Y., Y. Hong, G. Zhang, T. J. Schuur, J. J. Gourley, Z. Flamig, K. R. Morris, and Q. Cao, 2011: Cross validation of spaceborne radar and ground polarimetric radar aided by polarimetric echo classification of hydrometeor types. J. Appl. Meteor. Climatol., 50, 1389–1402, https://doi.org/10.1175/2011JAMC2622.1.
Wen, Y., Q. Cao, P. E. Kirstetter, Y. Hong, J. J. Gourley, J. Zhang, G. Zhang, and B. Yong, 2013: Incorporating NASA spaceborne radar data into NOAA National Mosaic QPE system for improved precipitation measurement: A physically based VPR identification and enhancement method. J. Hydrometeor., 14, 1293–1307, https://doi.org/10.1175/JHM-D-12-0106.1.
Wen, Y., A. Behrangi, B. Lambrigtsen, and P. E. Kirstetter, 2016: Evaluation and uncertainty estimation of the latest radar and satellite snowfall products using SNOTEL measurements over mountainous regions in western United States. Remote Sens., 8, 904, https://doi.org/10.3390/rs8110904.
Yin, X., Z. Hu, J. Zheng, B. Li, and Y. Zuo, 2021: Study on radar echo-filling in an occlusion area by a deep learning algorithm. Remote Sens., 13, 1779, https://doi.org/10.3390/rs13091779.
Zhang, G., J. Vivekanandan, and E. A. Brandes, 2001: A method for estimating rain rate and drop size distribution from polarimetric radar. IEEE Trans. Geosci. Remote Sens., 39, 830–841, https://doi.org/10.1109/36.917906.
Zhang, J., K. Howard, C. Langston, B. Kaney, Y. Qi, L. Tang, and D. Kitzmiller, 2016: Multi-Radar Multi-Sensor (MRMS) quantitative precipitation estimation: Initial operating capabilities. Bull. Amer. Meteor. Soc., 97, 621–638, https://doi.org/10.1175/BAMS-D-14-00174.1.
