For Peer Review
The Detection of Mesoscale Convective Systems by the GPM Ku-band Spaceborne Radar
Journal: Journal of the Meteorological Society of Japan
Manuscript ID JMSJ-2019-0034.R1
Manuscript Type: Articles
Date Submitted by the n/a Author:
Complete List of Authors: wang, jingyu; Pacific Northwest National Laboratory, Atmospheric Sciences & Global Change Houze, Robert ; University of Washington, Department of Atmospheric Sciences; Pacific Northwest National Laboratory, Atmospheric Sciences and Global Change Division Fan, Jiwen; Pacific Northwest National Laboratory, Atmospheric Sciences and Global Change Division Brodzik, Stacy; University of Washington, Department of Atmospheric Sciences Feng, Zhe; Pacific Northwest National Laboratory, Atmospheric Sciences and Global Change Division Hardin, Joseph; Pacific Northwest National Laboratory, Atmospheric Sciences and Global Change Division
GPM evaluation, 3D reflectivity structure, Mesoscale Convective System Keywords: tracking, MCS feature, Global MCS distribution, Intense convection
Page 1 of 100 For Peer Review
1 The Detection of Mesoscale Convective Systems by the
2 GPM Ku-band Spaceborne Radar
3
4 Jingyu Wang
5 Pacific Northwest National Laboratory, 6 Richland, Washington, USA
1 7 Robert. A. Houze, Jr
8 University of Washington, 9 Seattle, Washington, USA 10 Pacific Northwest National Laboratory, 11 Richland, Washington, USA
12 Jiwen Fan
13 Pacific Northwest National Laboratory, 14 Richland, Washington, USA
15 Stacy. R. Brodzik
16 University of Washington, 17 Seattle, Washington, USA
18 Zhe Feng
19 Pacific Northwest National Laboratory, 20 Richland, Washington, USA
21 and
22 Joseph C. Hardin
23 Pacific Northwest National Laboratory, 24 Richland, Washington, USA 25 26 27 March 31, 2019 28 ------29 1) Corresponding author: Robert Houze, University of Washington, Seattle, Washington, 30 USA. Email: [email protected] For Peer Review Page 2 of 100
31 Abstract
32
33 The Global Precipitation Measurement (GPM) core observatory satellite launched in
34 2014 features more extended latitudinal coverage (65S-65N) than its predecessor
35 Tropical Rainfall Measuring Mission (TRMM, 35S-35N). The Ku-band radar onboard of
36 the GPM is known to be capable of characterizing the 3D structure of deep convection
37 globally. In this study, GPM’s capability for detecting mesoscale convective systems
38 (MCSs) is evaluated. Extreme convective echoes seen by GPM are compared against an
39 MCS database that tracks convective entities over the contiguous US. The tracking is
40 based on geostationary satellite and ground-based Next Generation Radar (NEXRAD)
41 network data obtained during the 2014-2016 warm seasons. Results show that more than
42 70% of the GPM-detected Deep-Wide Convective Core (DWC) and Wide Convective Core
43 (WCC) objects are part of NEXRAD identified MCSs, indicating that GPM-classified DWCs
44 and WCCs correlate well with typical MCSs containing large convective features. By
45 applying this method to the rest of the world, a global view of MCS distribution is obtained.
46 This work reveals GPM’s potential in MCS detection at the global scale, particularly over
47 remote regions without dense observation network.
48
49 Keywords GPM evaluation; 3D reflectivity structure; Mesoscale Convective System
50 tracking; MCS features; Intense convection; Global MCS distribution
1 Page 3 of 100 For Peer Review
51 1. Introduction
52 As a collaborative effort between the National Aeronautics and Space Administration
53 (NASA) and the Japan Aerospace Exploration Agency (JAXA), the Tropical Rainfall
54 Measuring Mission (TRMM) satellite was equipped with a Ku-band (13.8 GHz) quantitative
55 precipitation radar, together with a variety of sensors including passive microwave, visible,
56 infrared, and lightning (Kummerow et al., 1998; Schumacher et al. 2004; Zipser et al.
57 2006; Houze et al. 2015). Launched in 2014 as the successor of TRMM, the Global
58 Precipitation Measurement (GPM) core observatory satellite carries the first space-borne
59 Dual-frequency Precipitation Radar (DPR) operating at both Ku (13.6 GHz) and Ka (35.5
60 GHz) bands. Compared to TRMM, the DPR system helps improve the accuracy of
61 precipitation measurement and upgrade the detectability of weak rain as low as 0.5 mm hr-
62 1. Another key advancement of GPM is its extended coverage to higher latitudes (65S to
63 65N compared to 35S to 35N for TRMM), providing a near-global view of 3D cloud and
64 precipitation structure every 2-3 hours.
65 Mesoscale Convective Systems (MCSs) are of great importance because of their large
66 area (at least one hundred kilometers in one direction) and intense, long-lasting (up to 24
67 hours) precipitation (Houze 2004, 2018). In midlatitudes, MCSs strongly impact local
68 climate through their precipitation, severe weather, and redistribution of heat and moisture,
69 which further impact the regional to global hydrological cycle and large-scale circulations
70 (Houze et al. 1990; Feng et al. 2016, 2018; Futyan and Del Genio, 2007). Based on the
2 For Peer Review Page 4 of 100
71 detailed 3D radar reflectivity field observed by spaceborne precipitation radar, Houze et al.
72 (2007, 2015) developed a convective echo classification algorithm based on the horizontal
73 and vertical dimensions of echoes of a given intensity. The most intense of such 3D
74 convective echo objects are assumed to be strongly associated with MCSs (Houze et al.
75 2015, 2019). However, direct comparison to other independent MCS data sets is needed
76 to support this argument. Identification of MCSs has been traditionally carried out by
77 tracking convective elements in geosynchronous satellite or ground-based radar data,
78 which provide temporally continuous spatial data. For example, satellite imagery that
79 provides cloud radiative properties such as brightness temperature at high temporal
80 resolution has been used to track deep convective clouds at regional to global scale
81 (Schmetz et al. 1993; Machado et al. 1998; Morel and Senesi 2002; Héas and Mémin
82 2008; Escrig et al. 2013; Roca et al. 2014). However, the incapacity of visible and infrared
83 satellite data to show the structure of the clouds in the lower troposphere leaves large
84 uncertainties in satellite-based MCS tracking. Ground-based radar measurements provide
85 3D structure of precipitating deep convection that can complement satellite imagery to
86 more accurately detect MCSs. A recently developed tracking algorithm called FLEXible
87 object TRacKeR (FLEXTRKR, Feng et al. 2018) jointly uses the geostationary satellite
88 brightness temperature and NEXRAD 3D radar reflectivity structure to identify and track
89 MCSs. A comprehensive 13-year (2004-2016) MCS tracking database east of the Rocky
90 Mountains has been developed using FLEXTRKR (Feng et al. 2019). This database is
3 Page 5 of 100 For Peer Review
91 considered here as ground-truth to evaluate the MCS objects detected by GPM.
92 To determine GPM’s capability of MCS detection, we organize this study as follows. In
93 Section 2, the data sets of GPM Ku-band reflectivity, NEXRAD observations, as well as
94 their corresponding MCS detecting/tracking algorithms are introduced. In Section 3, the
95 quantitative comparison of 3D radar reflectivity fields is performed between the two data
96 sets. This comparison is essential to determine whether the GPM and NEXRAD radar
97 systems detect the 3D radar reflectivity fields in a consistent way. After the consistency
98 between the two data sets is established, the GPM’s snapshots of intense convective echo
99 objects are compared to the NEXRAD MCS tracking database over the Continental United
100 States (CONUS) to quantify GPM’s MCS detection capability. Then, the regionally
101 validated GPM’s MCS detection algorithm is applied elsewhere around the world,
102 presenting a global view of MCS distribution. Finally, conclusions and discussions are
103 provided in Section 4.
104 2. Data and methodology
105 2.1 GPM Data and convective object classification
106 As a major component of the GPM DPR system, the onboard Ku-band precipitation
107 radar features a swath width of 245 km. The horizontal resolution is 0.05, or
108 approximately 5 km. The vertical resolution is 125 m, provided in 176 levels. In this study,
109 the GPM Ku band radar reflectivity data were downloaded from the University of
110 Washington GPM-Ku Data Set located at http://gpm.atmos.washington.edu, which
4 For Peer Review Page 6 of 100
111 geolocates and interpolates the DPR Level 2A Ku band version 5B data (Iguchi et al.
112 2017) from “radar coordinates” to a Cartesian grid with the aforementioned resolutions.
113 Detailed methodology regarding the geolocation correction and interpolation can be found
114 in Houze et al. (2007).
115 NASA and JAXA divide the GPM Ku-band radar echoes into convective, stratiform, and
116 other categories. In this study, the echo-object classification scheme developed by Houze
117 et al. (2007, 2015) is applied to the convective echoes. The analysis for evaluating GPM
118 3D radar reflectivity and GPM’s MCS detection capability is conducted in the CONUS
119 region for the warm seasons (April-September) of 2014-2016. The classification scheme
120 defines convective echoes as deep, wide, or deep and wide depending on radar reflectivity
121 intensity thresholds, and criteria for convective echo object height and area. Both strong
122 and moderate criteria (as defined by Houze et al. 2015) are used in this study to examine
123 the sensitivity of the results to the criteria. Strong/moderate deep convective core (DCC)
124 objects contain echoes that are greater than or equal to 40 dBZ/30 dBZ (everywhere in the
125 column) in intensity and have a maximum altitude of at least 10 km/8 km. Strong/moderate
126 wide convective core (WCC) objects contain echoes that are greater than or equal to 40
127 dBZ/30 dBZ and have a maximum horizontal extent of at least 1000 km2/800 km2. The
128 category of DCC has no areal coverage criterion, and is believed to be associated with
129 young, vigorous convection. The WCCs have no height criterion, but rather correspond to
130 where the intense convection has organized horizontally upscale into mesoscale areas of
5 Page 7 of 100 For Peer Review
131 active, widespread convection. Although the two categories are classified independently,
132 they do have overlap. Deep and wide convective core (DWC) objects meet the criteria for
133 both DCCs and WCCs. The wide categories, DWC and WCC, are thought to be strongly
134 related to the MCSs, as a typical MCS contains a large intense precipitation area with a
135 major axis longer than ~100 km, which is similar to the coverage criteria for the DWC and
136 WCC echo objects.
137 2.2 NEXRAD data and the MCS database
138 Densely distributed across the entire U.S., the NEXRAD network consists of 159 high
139 resolution S-band Doppler radars (WSR-88D) operated by the National Weather Service
140 (NWS). This study uses the mosaic NEXRAD data set named the Gridded Radar data
141 (GridRad, Bowman and Homeyer 2017) that combines all NEXRAD radar data covering
142 the region 155W – 69W, 25N – 49N. The GridRad data set has 2 km horizontal and 1
143 hour temporal resolutions, and a fixed 1 km vertical resolution (24 levels).
144 The 13-year MCS database (2004-2016) developed by Feng et al. (2019) uses both
145 geostationary satellite infrared brightness temperature (Tb) and GridRad radar data to
146 identify and track MCSs east of the Rocky Mountains (110W – 70W, 25N – 49N). The
147 native GridRad radar data were regridded to 4 km resolution to match the geostationary
148 satellite data. In Feng et al. (2019), an MCS was defined when a convective cloud system
4 149 satisfies both criteria in size (cold cloud system with Tb < 241 K area greater than 6 x 10
150 km2, and the radar-observed precipitation feature major axis length exceeds 100 km, with
6 For Peer Review Page 8 of 100
151 embedded convective core > 45 dBZ) and duration (persistence longer than 6 hours). The
152 identification of a convective core is based on the storm labeling in three dimensions
153 (SL3D) classification (Starzec et al. 2017). In SL3D, convective core is labeled if any one
154 of the following criteria is satisfied: (1) 25 dBZ echo-top height ≥ 10 km, (2) echo
155 peakedness ≥ 50% of the echo column between the surface and 9 km, and (3) 45 dBZ
156 echo-top height is above the melting-layer height. In the MCS database (Feng et al. 2019),
157 the melting-layer height is determined using the 6-hourly ERA-Interim reanalysis (Dee et
158 al. 2011). After an MCS is tracked, its life cycle is further separated into the initiation,
159 genesis, mature, and dissipation stages. Initiation starts at the first hour when the first
160 MCS-related cold cloud system is detected, followed by the genesis stage when the major
161 axis length of the convective core exceeds 100 km. As the convective core maintains its
162 size, the upscale growth of the MCS’s stratiform rain area concludes the genesis
163 stage. Finally, when the convective core length falls below the 100 km threshold, or the
164 stratiform rain area is lower than the mean value throughout the entire MCS life cycle, the
165 system is classified as in the dissipation stage.
166 2.3 Matching the GPM data to the NEXRAD Coordinate
167 The GPM Ku-band radar operates at a wavelength of 2.2 cm, which is capable of
168 detecting large raindrops or snowflakes (Rauber and Nesbitt 2018), while the 3 GHz
169 ground-based NEXRAD radar has a wavelength is 10 cm. The GPM Ku-band radar has a
170 swath width of 245 km, which is equivalent to the range of a single NEXRAD radar
7 Page 9 of 100 For Peer Review
171 (maximum operational range of 230 km, although better coverage and resolution is
172 typically provided within 150 km radius of the radar). The GPM radar scans downward,
173 nearly vertically and it has a vertical resolution ~100 m, while the NEXRADs scan quasi-
174 horizontally with a beamwidth of ~1 degree so that at great horizontal distance the vertical
175 resolution is coarse. More importantly, there is a temporal mismatch between the two data
176 sets, as the GPM provides instantaneous snapshots of 3D radar reflectivity field whereas
177 the GridRad temporally averages the observations from different individual NEXRAD
178 radars within a 9-minute window. The different wavelengths, scanning geometries, and
179 data processing of the GPM radar and NEXRAD require that care be taken in comparing
180 data from the two systems.
181 Evaluation of the GPM Ku-band calibration and its attenuation correction has been
182 performed throughout its pre-launch and post-launch stages in multiple field campaigns,
183 such as the 2011 Mid-Continent Convective Clouds Experiment (MC3E, Wang et al. 2015;
184 Jensen et al. 2016), the 2012 GPM Cold-season Precipitation Experiment (GCPEx,
185 Skofronick et al. 2015), the 2014 Integrated Precipitation and Hydrology Experiment
186 (IPHEX, Grecu et al. 2018), and the 2015-2016 Olympic Mountain Experiment
187 (OLYMPEX, Houze et al. 2017, McMurdie et al. 2018; Zagrodnik et al. 2018). These field
188 experiments have found the GPM Ku-band radar data are consistent with the 10 cm
189 wavelength ground-based radar measurements.
8 For Peer Review Page 10 of 100
190 In comparing radar datasets from GPM and NEXRAD, differences can result from
191 processing the data at different spatial resolution. It is therefore critical for us to
192 quantitatively examine the consistency of the interpolated 3D radar reflectivity databases.
193 The Earth System Modeling Framework (ESMF) package
194 (https://www.earthsystemcog.org/projects/esmpy/) has been used to bi-linearly re-
195 interpolate the GPM 3D reflectivity data (spatial resolution of 0.05o, vertical resolution of
196 125 m) to the GridRad coordinate (4 km horizontal, 1 km vertical resolution). The hourly
197 NEXRAD data are extracted at the GPM overpass locations and the nearest coincident
198 hour. As described in Bowman and Homeyer (2017), the GridRad data are generated
199 using the 4-dimensional binning (averaging) algorithm that merges multiple individual radar
200 volumes (the mean from 4-minute before to 4-minute after the sample time). To make a
201 fair comparison between the averaged GridRad reflectivity to the instantaneous GPM
202 reflectivity measurements, we apply a Gaussian smoothing technique (Reinhard 2006)
203 vertically to the GPM data to mimic the binning procedure used in GridRad. Following a
204 suggestion from C. Homeyer (2018, personal communication), a 1 km running mean along
205 the vertical direction, and a Gaussian smoothing with kernel width of 9 and sigma value of
206 2.5 are applied to the GPM reflectivity field. After the horizontal regridding and vertical
207 smoothing, the two data sets are in the same coordinates. The smoothing procedure plays
208 a significant role in matching the two observation data sets. If not smoothed, the raw
209 GPM’s reflectivity values are systematically higher than the NEXRAD data by 20-25%.
9 Page 11 of 100 For Peer Review
210 Note the minimum detectable reflectivity value of GPM Ku-band radar is between 12
211 dBZ (Hou et al. 2014; Toyoshima et al. 2015; Hamada and Takayabu 2016) and 13 dBZ
212 (Olson et al. 2016). Through the examination of all available Ku-band radar data, it is
213 extremely rare to find reflectivity lower than 13 dBZ (with occurrence frequency of 5 x 10-8).
214 As a result, this study adopts the higher threshold of 13 dBZ, and the corresponding
215 NEXRAD reflectivity is truncated accordingly for fair comparison.
216 As revealed in previous studies (e.g., Krajewski et al. 2006; Cui et al. 2016 and 2017),
217 NEXRAD radar beam blockage is a common issue over mountainous regions, making the
218 radar reflectivity comparison with GPM difficult. However, the MCSs commonly initiate
219 from the lee side of the Rocky Mountains (Feng et al. 2019). Therefore we choose the
220 study domain just east of the Rocky Mountains Front Range (eastern Colorado). In
221 addition, NEXRAD radars deployed along the coast have extended observations over the
222 ocean. Those observations lack overlap with other radars and their sample volume is
223 larger than normal (Rinehart 2001), thus the data quality offshore is questionable.
224 Moreover, from the perspective of MCS tracking, convection farther offshore away from
225 the coastal NEXRADs, are prone to have larger uncertainty because of reduced low-level
226 radar coverage. As a result, through the examination of CONUS elevation data
227 (TerrainBase, 1995), the study domain is constrained to the central U.S. as shown in Fig. 1
228 (110W – 85W, 31N – 48N), which avoids most of the complex terrain over Rocky and
229 Appalachian Mountains and excludes all offshore areas.
10 For Peer Review Page 12 of 100
230 3. Results
231 3.1 The comparison of 3D radar reflectivity between GPM and NEXRAD
232 Vertical echo structure is indicative of the kinematic, dynamic, and thermodynamic
233 features of convective clouds (Hence and Houze 2008, 2011, 2012a, b), as well as cloud
234 microphysical properties (Leary and Houze 1979; Cetrone and Houze 2011; Rowe and
235 Houze 2014; Fan et al. 2015, 2017; Liu et al. 2015; Barnes and Houze 2016; Han et al.
236 2019). As mentioned in Section 2.3, the GPM data are instantaneous snapshots of 3D
237 radar reflectivity field from above whereas the NEXRAD provides the mean composite
238 data from multiple ground radars. This may result in temporal mismatch between the two
239 measurements, since we have binned the GPM overpasses to the nearest hour, i.e., the
240 hourly NEXRAD data are compared to the GPM overpasses that occur within the time
241 window of +/- 30 minutes for each hour. Previous studies (e.g., Feng et al. 2009, Wang et
242 al. 2016 and 2018) have shown that the time mismatch can significantly impact the
243 reflectivity measurements between different radars, as convective features could change
244 substantially in evolution and/or location on a time scale less than an hour. Therefore, we
245 adopt the statistical method of contoured-frequency-by-altitude-diagrams (CFADs) to
246 compare the vertical distribution of radar reflectivity measurements from the two radar
247 platforms.
248 As shown in Fig. 2, CFADs are generated for GPM (a, d, g, j, m, p) and NEXRAD (b, e,
249 h, k, n, q) from all the collocated data for each month (April - September) within the
11 Page 13 of 100 For Peer Review
250 sampling area for the 2014-2016 period. The CFADs display the frequency distribution in a
251 coordinate system of reflectivity bins (x-axis) and the altitude (y-axis), and represent the
252 occurrence frequency of reflectivity spectra normalized by the total number of bins in which
253 reflectivity is recorded. The bin sizes in reflectivity and height are 1 dBZ and 1 km,
254 respectively. The frequency is calculated as the number of non-zero echo values in a bin
255 divided by the total number of bins containing reflectivity (all heights). The integral of the
256 frequency over the entire 2D graph equals 1.
257 From the examination of the CFADs structure, the GPM is in good agreement with
258 NEXRAD for each month regardless of the overall shape or magnitude of the frequency at
259 various altitudes. Both data sets demonstrate consistent seasonal variations: the deepest
260 echo top at various reflectivity thresholds increases from spring (April - May) to summer
261 (June - July - August), then decreases towards the early fall (September). This seasonality
262 is also seen for the altitude change of the frequency contour of 0.6% and above. The peak
263 altitudes of this higher frequency are broadly distributed below 6 km in spring, and the
264 altitudes increase to 9 km in the summer, and finally decrease to lower altitude in early fall.
265 The seasonal changes in echo-top heights and the altitude of the higher frequency
266 reflectivity values implies the seasonal variation of the convective intensity. The convective
267 updraft intensity maximum in summer is consistent with the largest convective available
268 potential energy being observed at that time (Xie et al., 2014). Meanwhile, the seasonality
269 in the altitude of higher frequency values is also indicative of the shift in storm types
12 For Peer Review Page 14 of 100
270 between seasons. Spring and fall storms tend to have more stratiform clouds with bottom-
271 heavy reflectivity profiles (e.g., strong bright-band signature near the melting level),
272 resulting in higher reflectivity at mid-to-lower levels. During the summer, more occurrence
273 of convective clouds shifts higher reflectivity values aloft.
274 The radar reflectivity distributions normalized at each level are also examined using the
275 box-whisker plots shown in Fig. 2c, f, i, l, o, r. General agreements are also found for the
276 median, interquartile and extreme values between the two data sets at different altitudes
277 for every month. The differences are commonly less than 2 dBZ except for the 95th
278 percentiles above 12 km, where the NEXRAD shows larger reflectivity values than the
279 GPM. From the CFADs, the frequency of occurrence for these extreme values are very
280 low, thus the difference there is not of particular concern.
281 3.2 Evaluation of GPM’s capability in MCS detection
282 Having established that the GPM and WSR-88D reflectivity fields are consistent on the
283 interpolated grids, we now assess GPM’s capability in MCS detection. As mentioned in
284 Section 2.1, it has been suggested that the GPM DWC and WCC echo objects correlate
285 with MCSs, which feature convection organized on larger horizontal scale. Here we test
286 this argument by comparing GPM DWC and WCC echo objects (strong criteria), which are
287 observed as snapshots, to the MCS database, which is unambiguously determined from
288 tracking convective features in time. Figure 3 shows one example of a GPM-defined DWC
289 echo object and the coincident MCS at the nearest hour from the NEXRAD data, as well
13 Page 15 of 100 For Peer Review
290 as the timing of GPM’s overpass with the tracked MCS’s life cycle. By comparing Fig. 3a
291 (GPM) and Fig. 3b (NEXRAD), the collocated radar reflectivity fields as observed from the
292 two platforms show that they are in as good agreement as could be expected. The GPM-
293 defined DWC snapshot occurred during the time period that the MCS tracked by the
294 FLEXTRKR algorithm was in the genesis stage (Fig. 3c), meaning it was experiencing
295 upscale growth at the time of the GPM observation.
296 In order to form long-term statistics, all 158 DWC objects and 230 WCC objects
297 detected by GPM (strong criteria) within the three warm seasons are evaluated with the
298 NEXRAD MCS database. By overlaying the mask of GPM-defined DWC or WCC object to
299 the mask of the NEXRAD-tracked MCS cold cloud shield, we can determine if the GPM
300 DWC or WCC objects are part of a NEXRAD-defined MCS. If overlap is found, then this
301 GPM-defined DWC or WCC object is considered as a HIT. Otherwise, it is treated as a
302 false alarm (FAR) because the GPM snapshot of the convective echo objects does not
303 align with any NEXRAD-tracked MCS. Note we do not define a missing category (i.e.,
304 MCSs tracked by the NEXRAD but not identified by the GPM) because of the substantial
305 difference in spatial coverage. As described in Section 2, the GPM has a limited swath
306 width of 245 km, where the NEXRAD network covers the entire CONUS. Therefore, the
307 “missing” detection from the GPM is highly possible simply because there is no GPM
308 overpass for a particular MCS.
14 For Peer Review Page 16 of 100
309 In addition to the objective comparison, all the GPM DWC/WCC and collocated
310 NEXRAD reflectivity images are manually screened for further confirmation. For the GPM-
311 detected echo objects in the FAR category, there are several cases that the GPM
312 detections are closely located (usually within ~5 km) to the NEXRAD-tracked MCS but
313 without overlap. We suspect that this could be caused by the temporal mismatch between
314 the two data sets as aforementioned and should be treated as HIT instead. As a result, by
315 including the NEXRAD MCS masks from 1 hour before to 1 hour after the coincident hour,
316 overlaps are found for those GPM detections (11 cases).
317 The final result shows that 115 of 158 DWC objects and 158 of 230 WCC objects can
318 be verified as MCSs when compared to the NEXRAD data set, leaving 115 cases as FAR.
319 In summary, 70% of DWC and WCC systems classified by the GPM are MCSs identified in
320 the NEXRAD data set. Interestingly, most of the GPM-detected MCS snapshots are during
321 the genesis (41%) and mature stages (32%) of the tracked MCSs. There are only about
322 21% in the dissipation stage and 6% in the initiation stage. These results are consistent
323 with the finding of Feng et al. (2019) that warm season MCS convective features during
324 the upscale growth stage are the largest and deepest. These convective features are most
325 likely to meet the GPM DWC or WCC criteria.
326 After revisiting the 115 cases in the FAR category by examining the NEXRAD images
327 before and after the GPM detection, 19% of them last more than 6 hours but their major
328 axis length of precipitation area couldn’t exceed 100 km to satisfy the MCS criteria defined
15 Page 17 of 100 For Peer Review
329 by FLEXTRKR. For the rest of 81% cases, they all dissipate too soon and therefore fail to
330 satisfy the MCS duration requirement of 6 hours. These false alarms reveal the intrinsic
331 limitation of the criteria used to identify MCSs by tracking NEXRAD echoes and the
332 conditions used to define the GPM echo-object categories are arbitrary.
333 Because the majority of false alarms result from the insufficient duration, a question
334 arises as to whether the HIT rate (defined as HITs / (HITs + FARs)) could increase with
335 lowering the MCS duration threshold. Some previous studies defined MCSs in this region
336 using shorter duration of 4 hours (e.g., Geerts 1998; Haberlie and Ashley 2019). To
337 examine the impact of MCS duration criterion to our results, we performed a sensitivity test
338 by reducing the MCS duration threshold from 6 hours to 4 hours in FLEXTRKR, and the
339 total number of MCSs tracked by NEXRAD in the 3 warm seasons increased from 740 to
340 1193. However, only 12 GPM FAR cases are changed to the HIT category. Upon close
341 examination of the FAR cases, we found several reasons that explain this result. First,
342 90% of the additional 453 NEXRAD-tracked MCS cases occur outside the GPM
343 overpasses, thus they have no impact to the GPM statistics. Secondly, for those short-
344 lived false alarms detected by the GPM, their lifespan are commonly less than 3 hours,
345 which fail to meet even the shortened MCS duration criteria. Lastly, short duration and
346 insufficient coverage are not exclusive, i.e., many GPM-detected objects that do not last
347 longer than 6 hours also do not satisfy the size criteria. In this case, it is complex to
348 differentiate the actual causes of false alarms. The fact that changing the MCS duration
16 For Peer Review Page 18 of 100
349 threshold produces limited impact to GPM’s MCS statistics may further confirm the
350 correspondence between the majority of GPM-detected DWCs/WCCs and the largest,
351 deepest MCSs, as the latter require longer period of time to form. However, there remain
352 exceptions of short-lived DWCs/WCCs, which raise the necessity of the temporal
353 dimension. More extreme sensitivity test of 2 hours duration threshold is performed, now
354 the systems tracked by the NEXRAD increases to 1671 and 49 false alarms change to the
355 category of HIT, making the HIT rate 83%. However, this comparison may not be
356 meaningful because the 2 hours duration is too short for MCS definition, which makes
357 almost no difference to the direct radar echo comparison between the two platforms.
358 Based on these tests, we conclude that 70% accuracy reached in our first comparison
359 remains a good overall estimate of the capability to determine MCS existence from the
360 GPM radar data.
361 By using the original MCS tracking data set as reference, Fig. 4 and Table 1 show the
362 GPM’s detection skills in each month, where the number of HIT, FAR, as well as the
363 number of NEXRAD-tracked MCSs and their averaged precipitation area coverage and
364 duration are compared. The number of NEXRAD-tracked MCSs demonstrates a strong
365 seasonality, which increases from April to June, then diminishes towards early fall. This
366 variation follows the seasonal change in baroclinic instability over the CONUS. During
367 spring, large-scale forcing brought by the mid-latitude trough frequently occurs, providing a
368 favorable environment for organized convection (Maddox et al 1979; Peters and
17 Page 19 of 100 For Peer Review
369 Schumacher 2014). In April, MCSs produced under strong baroclinic forcing usually
370 feature broad stratiform rain region, which results in the largest precipitation area, but the
371 number is relatively low compared to the midsummer, when the CONUS has the weakest
372 baroclinic instability and minimal frontal forcing. However, due to the favorable
373 thermodynamic conditions and possible influence from sub-synoptic disturbances, local
374 convection can nevertheless frequently grow upscale into MCSs (Wang et al. 2011; Song
375 et al. 2019; Feng et al. 2019). The combination of spring-like baroclinic waves and a
376 continuously warming surface possibly favors the peak number of MCSs in late spring and
377 early summer (i.e., June), but the MCS precipitation areas are generally smaller than in
378 spring. Although fall has similar large-scale environments as in spring, MCSs are less
379 frequent and no increase in average coverage is found. One possible explanation is the
380 seasonality in surface temperature gradient, which is weaker in fall than in spring, possibly
381 causing weaker baroclinic waves (Feng et al. 2019). The average MCS duration was also
382 calculated for each month, and no seasonal variation was found.
383 From the perspective of MCS detection by the GPM, several factors contribute to a
384 higher HIT rate: more frequent MCS occurrence, larger precipitation area and longer
385 duration (i.e., the propagating nature results in larger area covered by MCSs). The trend of
386 monthly HIT is consistent with the variation of the MCS numbers, indicating that the
387 probability of GPM detected MCSs increases with more frequent MCS occurrence. Spring
388 corresponds to the lowest False Negative Rate (defined as FARs / (HITs + FARs)), which
18 For Peer Review Page 20 of 100
389 is consistent with the larger MCS coverage favored by the synoptic forcing. Thus higher
390 accuracy in GPM’s MCS detection can be expected relative to the rest of the months. In
391 contrast, for the August-September period, as the result of poleward expansion of the
392 subtropical ridge (Wang et al. 2019), the baroclinic environment for supporting large MCSs
393 no longer exists, resulting in a higher False Negative Rate.
394 3.3 Global MCS distributions
395 After establishing the GPM’s MCS detection capability of about 70% by validating
396 against the NEXRAD-tracked MCS database over CONUS in the previous section, we can
397 apply the DWC and WCC criteria to the GPM data globally to examine the probably global
398 distribution of MCSs. Because ground-based radar networks like NEXRAD do not exist
399 over most of Earth, GPM is the best available resource for determining the global pattern
400 of MCS occurrence. Figure 5 shows the geographical distribution of MCS occurrence
401 frequency determined from GPM during the boreal summer June-July-August (JJA, Fig 5a)
402 and winter December-January-February (DJF, Fig 5b) over the 5-year period (2014-2018)
403 with available GPM observations. The frequency is computed as the number of pixels
404 identified as either DWC or WCC divided by the total number of pixels sampled by the
405 GPM Ku-band radar within a 0.25 x 0.25 gridbox. It is evident that the occurrence of
406 MCSs is more concentrated over land, which is consistent with previous studies showing
407 that convection over vast oceans is generally less intense than over land (e.g., Futyan and
408 Del Genio 2007; Houze et al. 2015). A comparison between Fig. 5a and Fig. 5b shows that
19 Page 21 of 100 For Peer Review
409 MCSs occur more frequently during the boreal summer than winter for both hemispheres.
410 In North America during summer, MCSs are densely distributed over the Great Plains
411 (GP), where they are fed by warm moist air transported from the Gulf of Mexico by the
412 climatological low-level jet on the lee side of the Rocky Mountains. In both summer and
413 winter, MCSs frequently occur offshore of the east coast of North America. These MCSs
414 tend to result from the initiation of gravity waves on the lee side of the Appalachian
415 Mountains (Keighton et al. 2007; Letkewicz and Parker 2010). In winter, MCSs are absent
416 over the GP but occur over the southeast and the offshore of the east coast of the U.S.,
417 consistent with NEXRAD-based MCS frequencies reported by Feng et al. (2019).
418 Over the landmass of tropical South America and central equatorial Africa, large
419 clusters of MCSs occur during JJA, but the hot zones of MCSs shift southward to the
420 subtropics and mid-latitudes in DJF, consistent with previous studies (e.g., Romatschke
421 and Houze 2010; Rasmussen and Houze 2011; Rasmussen et al. 2014). The high-
422 frequency MCS areas are displaced from the areas with the most regional rainfall (Houze
423 2015). Over Asia, three hot zones are identified in JJA, namely the Indian monsoon region,
424 east Asian monsoon region, and the maritime continent consisting Indonesia, Malaysia,
425 and Northern Australia (Ramage 1968). The former two regions are strongly influenced by
426 the summer monsoon flow, hence the MCSs peak in JJA but diminish in DJF. In contrast,
427 the maritime continent MCS occurrence is dominated by diurnal forcing associated with
428 the islands and peninsulas of the region, and in DJF the diurnal convection is enhanced by
20 For Peer Review Page 22 of 100
429 surges of the boreal winter monsoon (Johnson and Houze 1987) and modulated by
430 passages of the Madden-Julian Oscillation (Madden and Julian, 1971,1972, 1994).
431 Note all of the above MCS analyses are based on the strong thresholds applied to GPM
432 detection as described in Section 2.1. A sensitivity test using moderate thresholds is
433 performed with the corresponding global MCS distributions shown in Fig. 6. As revealed in
434 Houze et al. (2015), strong thresholds better represent the behavior of convection over
435 land, while the weak oceanic convective features can be exhibited more easily using the
436 moderate thresholds. By comparing Fig. 6 with Fig. 5, the locations of high MCS
437 occurrence are seen to be well maintained, but the frequency is greatly increased,
438 potentially including many non-MCS objects. The DWC and WCC objects detected using
439 moderate thresholds are also compared to the FLEXTRKR data set over the CONUS.
440 Although the number of GPM-detected DWC and WCC objects increases from 388 to 620,
441 the MCS HIT rate drops to 49%.
442 The global MCS distribution shown in Fig. 5 extends the MCS detection to higher
443 latitudes (beyond 35S-35N covered by TRMM), where notable MCS occurrence
444 frequency is found above 60N in Siberia, Northern Europe, and Canada. For the high-
445 latitude events, Houze et al. (2019) further found that these high-latitude MCSs occur
446 where global warming has been most intense.
447
448
21 Page 23 of 100 For Peer Review
449 4. Conclusions and Discussions
450 In this study, the spaceborne GPM Ku-band radar data sets are quantitatively
451 compared to the ground-based NEXRAD radar data sets during a 3-year period (2014-
452 2016) over the CONUS. Based on the morphology of GPM-detected radar echoes, two
453 types of GPM-detected extreme convective echo objects, DWC and WCC, are compared
454 with an MCS database constructed using feature tracking on synthesized geostationary
455 satellite and NEXRAD radar observations. The major findings of this study are
456 summarized as follows.
457 (1) The GPM radar captures consistent 3D distribution of radar reflectivity with
458 NEXRAD across a wide range precipitating cloud systems, including seasonal variations
459 from spring to fall, when the GPM radar and NEXRAD data are interpolated to the same
460 grid and appropriately smoothed. The comparison demonstrates the two independent
461 radar systems observe the 3D radar reflectivity fields in a consistent manner.
462 (2) The GPM classified DWC and WCC objects are compared to the NEXRAD-tracked
463 MCS database. More than 70% of these GPM-defined extreme convective echo objects
464 are collocated with tracked MCSs in the NEXRAD data set, indicating the GPM-detected
465 DWC and WCC objects are highly correlated with MCSs. The GPM’s capability in MCS
466 detection demonstrates strong seasonality, and a better performance is found in spring
467 and summer than fall. This seasonal variation follows the change of large-scale
468 environments that alter MCS characteristics. For the GPM-detected objects not
22 For Peer Review Page 24 of 100
469 corresponding to NEXRAD-tracked MCSs, the majority of them have shorter lifespan than
470 6 hours. These false alarms highlight the unavoidable uncertainty associated with the
471 arbitrariness of criteria used to identify MCSs in the two datasets.
472 (3) After validating the GPM’s performance in MCS detection over the CONUS, the
473 DWC and WCC classification algorithm is applied to global GPM observations to obtain a
474 global view of MCS distribution. The well-known MCS hot zones, namely the US GP and
475 the offshore of the east coast in North America, subtropical South America, central
476 equatorial Africa, monsoon regions in Asia, and the maritime continent, are further
477 confirmed in this study. Moreover, MCSs in high-latitude regions (above 60N in Siberia,
478 Northern Europe, and Canada) are revealed, and the relative numbers of MCSs over
479 different parts of Earth are now quantitatively measured.
480
481
482
483
484
485
486
487
488
23 Page 25 of 100 For Peer Review
489 Acknowledgements
490 We acknowledge the support of the Climate Model Development and Validation (CMDV)
491 project at PNNL. The effort of J. Wang, J. Fan, Z. Feng, and J. Hardin was supported by
492 CMDV. Robert A. Houze was supported by NASA Award NNX16AD75G and by master
493 agreement 243766 between the University of Washington and PNNL. Stacy R. Brodzik
494 was supported by NASA Award NNX16AD75G and subcontracts from the CMDV and
495 Water Cycle and Climate Extreme Modeling (WACCEM) projects of PNNL. PNNL is
496 operated for the US Department of Energy (DOE) by Battelle Memorial Institute under
497 Contract DE-AC05-76RL01830. This research used resources of the National Energy
498 Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of
499 Science User Facility operated under contract DE‐AC02‐05CH11231. The GPM reflectivity
500 data are download at University of Washington GPM-Ku Data Set at
501 (http://gpm.atmos.washington.edu/), the Global Merged IR dataset is obtained at NASA
502 Goddard Earth Sciences Data and Information Services Center
503 (https://dx.doi.org/10.5067/P4HZB9N27EKU), the GridRad radar dataset is obtained at the
504 Research Data Archive of the National Center for Atmospheric Research (NCAR)
505 (https://rda.ucar.edu/datasets/ds841.0/).
506
507
508
24 For Peer Review Page 26 of 100
509 References
510
511 Baldwin, M. E., and K. E. Mitchell, 1997: The NCEP hourly multisensor U.S. precipitation
512 analysis for operations and GCIP research. Preprints, 13th Conf. on Hydrology, Long
513 Beach, CA, Amer. Meteor. Soc., 54–55.
514 Barnes, H. C., and R. A. Houze, Jr., 2016: Comparison of observed and simulated spatial
515 patterns of ice microphysical processes in tropical oceanic mesoscale convective
516 systems. J. Geophys. Res. Atmos., 121, 8269-8296, doi:10.1002/2016JD025074.
517 Bowman, K. P., and C. R. Homeyer, 2017: GridRad - Three-Dimensional Gridded
518 NEXRAD WSR-88D Radar Data, V3.1. National Center for Atmospheric Research,
519 Computational and Information Systems Laboratory, Boulder, CO, accessed 1
520 November 2017, https://doi.org/10.5065/D6NK3CR7.
521 Box, J. F. Guinness, Gosset, Fisher, and small samples. Stat Sci 1987; 2: 45–52,
522 https://doi.org/10.1214/ss/1177013437.
523 Cetrone, J. and R.A. Houze, 2011: Leading and Trailing Anvil Clouds of West African
524 Squall Lines. J. Atmos. Sci., 68, 1114–1123, https://doi.org/10.1175/2011JAS3580.1
525 Clark, A.J., R.G. Bullock, T.L. Jensen, M. Xue, and F. Kong, 2014: Application of Object-
526 Based Time-Domain Diagnostics for Tracking Precipitation Systems in Convection-
527 Allowing Models. Wea. Forecasting, 29, 517–542, https://doi.org/10.1175/WAF-D-13-
528 00098.1.
25 Page 27 of 100 For Peer Review
529 Cui, W., X. Dong, B. Xi, and A. Kennedy, 2017: Evaluation of Reanalyzed Precipitation
530 Variability and Trends Using the Gridded Gauge-Based Analysis over the CONUS. J.
531 Hydrometeor., 18, 2227–2248, https://doi.org/10.1175/JHM-D-17-0029.1.
532 Cui, W., X. Dong, B. Xi, and R. Stenz, 2016: Comparison of the GPCP 1DD Precipitation
533 Product and NEXRAD Q2 Precipitation Estimates over the Continental United States. J.
534 Hydrometeor., 17, 1837–1853, https://doi.org/10.1175/JHM-D-15-0235.1.
535 Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: configuration and
536 performance of the data assimilation system. Quarterly Journal of the Royal
537 Meteorological Society, 137, 553-597.
538 Doviak, R. J., and D. S. Zrnic, 1993: Doppler Radar and Weather Observations. 2d ed.
539 Academic Press, 562 pp.
540 Fan, J, Y‐C Liu, K‐M Xu, K. North, S. Collis, X. Dong, G. J. Zhang, Q. Chen, P. Kollias,
541 and S. J. Ghan, 2015: Improving representation of convective transport for scale‐aware
542 parameterization: 1. Convection and cloud properties simulated with spectral bin and
543 bulk microphysics. J. Geophys. Res. Atmos., 120, 3485–3509.
544 https://doi.org/10.1002/2014JD022142.
545 Fan, J., et al. 2017: Cloud‐resolving model intercomparison of an MC3E squall line case:
546 Part I—Convective updrafts, J. Geophys. Res. Atmos., 122, 9351–9378,
547 https://doi.org/10.1002/2017JD026622.
548 Feng, Z., X. Dong, and B. Xi, 2009: A Method to Merge WSR-88D Data with ARM SGP
26 For Peer Review Page 28 of 100
549 Millimeter Cloud Radar Data by Studying Deep Convective Systems. J. Atmos. Oceanic
550 Technol., 26, 958–971, https://doi.org/10.1175/2008JTECHA1190.1
551 Feng, Z., L. R. Leung, R. A. Houze, S. Hagos, J. Hardin, Q. Yang, B. Han, and J. Fan,
552 2018: Structure and evolution of mesoscale convective systems: Sensitivity to cloud
553 microphysics in convection-permitting simulations over the United States. J. Adv.
554 Model. Earth Syst., 10, 1470–1494.
555 Feng, Z., R. A. Houze, L. R. Leung, F. Song, J. Hardin, J. Wang, W. Gustafson, and C.
556 Homeyer, 2019: Spatio-temporal Characteristics and Large-scale Environments of
557 Mesoscale Convective Systems East of the Rocky Mountains. J. Clim., In revision.
558 Feng, Z., Leung, L. R., Hagos, S., Houze, R. A., Burleyson, C. D., & Balaguru, K. 2016:
559 More frequent intense and long‐lived storms dominate the springtime trend in central
560 US rainfall. Nature Communications, 7, 13429. https://doi.org/10.1038/ncomms13429.
561 Feng, Z., Leung, L. R., Houze, R. A., Jr., Hagos, S., Hardin, J., Yang, Q., et al. (2018).
562 Structure and evolution of mesoscale convective systems: Sensitivity to cloud
563 microphysics in convection‐permitting simulations over the United States. J. Adv.
564 Model. Earth Syst, 10, 1470–1494. https://doi.org/10.1029/2018MS001305.
565 Feng, Z., X. Dong, B. Xi, S. A. McFarlane, A. Kennedy, B. Lin, and P. Minnis, 2012: Life
566 cycle of midlatitude deep convective systems in a Lagrangian framework, J. Geophys.
567 Res., 117, D23201, doi: 10.1029/2012JD018362.
568 Fritsch, J. M., and G. S. Forbes, 2001: Mesoscale convective systems. Meteorological
27 Page 29 of 100 For Peer Review
569 Monographs, 28(50), 323–358.
570 Futyan, J. M., and A. D. Del Genio (2007), Deep convective system evolution over Africa
571 and the tropical Atlantic, J. Clim., 20, 5041–5060, doi:10.1175/JCLI4297.1.
572 Geerts, B. (1998), Mesoscale convective systems in the southeast United States during
573 1994-95: A survey. Wea. Forecasting, 13, 860-869. Doi 10.1175/1520-
574 0434(1998)013<0860:Mcsits>2.0.Co;2
575 Grecu, M., L. Tian, G.M. Heymsfield, A. Tokay, W.S. Olson, A.J. Heymsfield, and A.
576 Bansemer, 2018: Nonparametric Methodology to Estimate Precipitating Ice from
577 Multiple-Frequency Radar Reflectivity Observations. J. Appl. Meteor. Climatol., 57,
578 2605–2622, https://doi.org/10.1175/JAMC-D-18-0036.1.
579 Haberlie, A. M., and W. S. Ashley, (2019), A Radar-Based Climatology of Mesoscale
580 Convective Systems in the United States. J. Clim., 32, 1591-1606. 10.1175/jcli-d-18-
581 0559.1.
582 Hamada, A. and Y. N. Takayabu, 2016: Improvements in Detection of Light Precipitation
583 with the Global Precipitation Measurement Dual-Frequency Precipitation Radar (GPM
584 DPR). J. Atmos. Oceanic Technol., 33, 653–667, https://doi.org/10.1175/JTECH-D-15-
585 0097.1.
586 Han, B., Fan, J., Varble, A., Morrison, H., Williams, C. R., Chen, B., et al, 2019:
587 Cloud‐Resolving Model Intercomparison of an MC3E Squall Line Case: Part II –
588 Stratiform Precipitation Properties. J. Geophys. Res. Atmos.,124.
28 For Peer Review Page 30 of 100
589 https://doi.org/10.1029/2018JD029596
590 Héas, P., E. Mémin Three-dimensional motion estimation of atmospheric layers from
591 image sequences IEEE Trans. Geosci. Remote Sens., 46 (8) (2008), pp. 2385-2396.
592 Hence, D. A., and R. A. Houze Jr., 2008: Kinematic structure of convective-scale elements
593 in the rainbands of Hurricanes Katrina and Rita (2005). J. Geophys. Res., 113, D15108,
594 doi:10.1029/2007JD009429.
595 Hence, D. A., and R. A. Houze Jr., 2011: Vertical structure of hurricane eyewalls as seen
596 by the TRMM Precipitation Radar. J. Atmos. Sci., 68, 1637–1652,
597 https://doi.org/10.1175/2011JAS3578.1.
598 Hence, D. A., and R. A. Houze Jr., 2012: Vertical structure of tropical cyclones with
599 concentric eyewalls as seen by the TRMM Precipitation Radar. J. Atmos. Sci., 69,
600 1021–1036, https://doi.org/10.1175/JAS-D-11-0119.1.
601 Hence, D.A. and R.A. Houze, 2012: Vertical Structure of Tropical Cyclone Rainbands as
602 Seen by the TRMM Precipitation Radar. J. Atmos. Sci., 69, 2644–2661,
603 https://doi.org/10.1175/JAS-D-11-0323.1, https://doi.org/10.1175/JAS-D-11-0323.1.
604 Homeyer, R. C., K. P. Bowman, 2017: Algorithm Description Document for Version 3.1 of
605 the Three-Dimensional Gridded NEXRAD WSR-88D Radar (GridRad) Dataset, 12 pp.
606 [Available online at http://gridrad.org/pdf/GridRad-v3.1-Algorithm-Description.pdf]
607 Hou, A. Y., and Coauthors, 2014: The Global Precipitation Measurement Mission. Bull.
608 Amer. Meteor. Soc., 95, 701–722, doi:https://doi.org/10.1175/BAMS-D-13-00164.1.
29 Page 31 of 100 For Peer Review
609 Houze, R. A., Jr., D. C. Wilton, and B. F. Smull, 2007: Monsoon convection in the
610 Himalayan region as seen by the TRMM Precipitation Radar. Quart. J. Roy. Meteor.
611 Soc., 133, 1389–1411.
612 Houze, R. A., Jr., 2014: Cloud dynamics. 2d ed. Oxford: Elsevier/Academic Press, 329 pp.
613 Houze, R. A., Jr., 2018: 100 years of research on mesoscale convective systems.
614 Meteorological Monographs, 59, 17.1–17.54,
615 https://doi.org/10.1175/AMSMONOGRAPHS-D-18-0001.1.
616 Houze, R. A., Jr., 2004: Mesoscale convective systems. Rev. Geophys., 42, RG4003,
617 doi:https://doi.org/10.1029/2004RG000150.
618 Houze, R. A., Jr., 2012: Orographic effects on precipitating clouds. Rev. Geophys., 50,
619 RG1001, doi:10.1029/2011RG000365.
620 Houze, R. A., Jr., S. G. Geotis, F. D. Marks, Jr., and A. K. West, 1981: Winter monsoon
621 convection in the vicinity of north Borneo. Part I: Structure and time variation of the
622 clouds and precipitation. Mon. Wea. Rev., 109, 1595-1614.
623 Houze, R. A., Jr., K. L. Rasmussen, M. D. Zuluaga, and S. R. Brodzik, 2015: The variable
624 nature of convection in the tropics and subtropics: A legacy of 16 years of the Tropical
625 Rainfall Measuring Mission (TRMM) satellite. Rev. Geophys., 53, 994–1021,
626 doi:https://doi.org/10.1002/2015RG000488.
627 Houze, R. A., Jr., L. A. McMurdie, W. A. Petersen, M. R. Schwaller, W. Baccus, J. D.
628 Lundquist, C. F. Mass, B. Nijssen, S. A. Rutledge, D. R. Hudak, S. Tanelli, G. G. Mace,
30 For Peer Review Page 32 of 100
629 M. R. Poellot, D. P. Lettenmaier, J. P. Zagrodnik, A. K. Rowe, J. C. DeHart, L. E.
630 Madaus, H. C. Barnes, and V. Chandrasekar, 2017: The Olympic Mountains
631 Experiment (OLYMPEX). Bull. Amer. Meteor. Soc., 98, 2167–2188,
632 https://doi.org/10.1175/BAMS-D-16-0182.1.
633 Houze, R. A., J. Wang, J. Fan, S. Brodzik, and Z. Feng, 2019: Extreme convective storms
634 over high‐latitude continental areas where maximum warming is occurring. Geophys.
635 Res. Lett., 46, 4059– 4065. https://doi.org/10.1029/2019GL082414.
636 Iguchi, T., S. Seto, R. Meneghini, N. Yoshida, J. Awaka, M. Le, V. Chandrasekar, and T.
637 Kubota, 2017: GPM/DPR level-2. Algorithm Theoretical Basis Doc., 68 pp. [Available
638 online at
639 https://pmm.nasa.gov/sites/default/files/document_files/ATBD_DPR_201708_whole_1.p
640 df.]
641 Janowiak, J., B. Joyce, and P. Xie, 2017: NCEP/CPC L3 Half Hourly 4km Global (60S -
642 60N) Merged IR V1, edited by Andrey Savtchenko, Greenbelt, MD, Goddard Earth
643 Sciences Data and Information Services Center (GES DISC), Accessed: [1 February
644 2018], 10.5067/P4HZB9N27EKU.
645 Jensen, M.P., W.A. Petersen, A. Bansemer, N. Bharadwaj, L.D. Carey, D.J. Cecil, S.M.
646 Collis, A.D. Del Genio, B. Dolan, J. Gerlach, S.E. Giangrande, A. Heymsfield, G.
647 Heymsfield, P. Kollias, T.J. Lang, S.W. Nesbitt, A. Neumann, M. Poellot, S.A. Rutledge,
648 M. Schwaller, A. Tokay, C.R. Williams, D.B. Wolff, S. Xie, and E.J. Zipser, 2016: The
31 Page 33 of 100 For Peer Review
649 Midlatitude Continental Convective Clouds Experiment (MC3E). Bull. Amer. Meteor.
650 Soc., 97, 1667–1686, https://doi.org/10.1175/BAMS-D-14-00228.1.
651 Johnson, R. H., and R. A. Houze, Jr., 1987: Precipitating cloud systems of the Asian
652 monsoon. In Monsoon Meteorology (C.-P. Chang and T. N. Krishnamurti, Eds.), 298-
653 353.
654 Keighton, S., J. Jackson, J. Guyer, and J. Peters, 2007: A preliminary analysis of severe
655 quasi-linear mesoscale convective systems crossing the Appalachians. Preprints, 22nd
656 Conf. on Weather Analysis and Forecasting/18th Conf. on Numerical Weather
657 Prediction, Park City, UT, Amer. Meteor. Soc., P2.18. [Available online at
658 http://ams.confex.com/ams/pdfpapers/123614.pdf].
659 Krajewski, W. F., A. Ntelekos, and R. Goska, 2006: A GIS based methodology for the
660 assessment of weather radar beam blockage in mountainous regions: Two examples
661 from the U.S. NEXRAD network. Comput. Geosci., 32, 283–302.
662 Kummerow, C., W. Barnes, T. Kozu, J. Shiue, and J. Simpson 1998: The Tropical Rainfall
663 Measuring Mission (TRMM) sensor package, J. Atmos.Oceanic Technol.,15, 809–817.
664 Leary, C. A., and R. A. Houze, Jr., 1979: Melting and evaporation of hydrometeors in
665 precipitation from the anvil clouds of deep tropical convection. J. Atmos. Sci., 36, 669-
666 679.
667 Letkewicz, C.E. and M.D. Parker, 2010: Forecasting the Maintenance of Mesoscale
668 Convective Systems Crossing the Appalachian Mountains. Wea. Forecasting, 25,
32 For Peer Review Page 34 of 100
669 1179–1195, https://doi.org/10.1175/2010WAF2222379.1.
670 Lin, Y. 2011. GCIP/EOP Surface: Precipitation NCEP/EMC 4KM Gridded Data (GRIB)
671 Stage IV Data. Version 1.0. UCAR/NCAR - Earth Observing Laboratory.
672 https://doi.org/10.5065/D6PG1QDD. Accessed 15 Nov 2018.
673 Lin, Y., and K. E. Mitchell, 2005: The NCEP stage II/IV hourly precipitation analyses:
674 Development and applications. Preprints, 19th Conf. on Hydrology, San Diego, CA,
675 Amer. Meteor. Soc., CD-ROM, 1.2.
676 Liu, Y‐C, J. Fan, G. J. Zhang, K‐M Xu, and S. J. Ghan, 2015: Improving representation of
677 convective transport for scale‐aware parameterization: 2. Analysis of cloud‐resolving
678 model simulations. J. Geophys. Res. Atmos., 120, 3510–3532.
679 https://doi.org/10.1002/2014JD022145.
680 Machado, L.A., W.B. Rossow, R.L. Guedes, and A.W. Walker, 1998: Life Cycle Variations
681 of Mesoscale Convective Systems over the Americas. Mon. Wea. Rev., 126, 1630–
682 1654, https://doi.org/10.1175/1520-0493(1998)126<1630:LCVOMC>2.0.CO;2.
683 Madden, R., and P. Julian, 1972: Description of global scale circulation cells in the tropics
684 with a 40-50 day period. J. Atmos. Sci., 29, 1109–1123.
685 Madden, R. A., and P. R. Julian, 1994: Observations of the 40-50-day tropical oscillation—
686 A review. Mon. Wea. Rev., 122, 814-837.
687 Maddox, R. A., Chappell, C. F., & Hoxit, L. R., 1979: Synoptic and mesoscale aspects of
688 flash flood events. Bull. Amer. Meteor. Soc., 60, 115–123.
33 Page 35 of 100 For Peer Review
689 McMurdie, L. A., A. K. Rowe, R. A. Houze, Jr., S. R. Brodzik, J. P. Zagrodnik, and T. M.
690 Schuldt, 2018: Terrain-enhanced precipitation processes above the melting layer:
691 Results from OLYMPEX. J. Geophys. Res.
692 Atmos., 123, https://doi.org/10.1029/2018JD029161.
693 Morel, C. and Senesi, S., 2002: A climatology of mesoscale convective systems over
694 Europe using satellite infrared imagery. II: Characteristics of European mesoscale
695 convective systems. Q. J. R. Meteorol. Soc., 128: 1973-1995.
696 https://doi.org/10.1256/003590002320603494.
697 Olson, W. S., H. Masunaga, and the GPM Combined Radar-Radiometer Algorithm Team,
698 2016: GPM Combined Radar-Radiometer Precipitation Algorithm Theoretical Basis
699 Document (Version 4), 7 pp. [Available online at
700 https://pmm.nasa.gov/sites/default/files/document_files/Combined_algorithm_ATBD.V0
701 4.rev_.pdf.]
702 Peters, J.M. and R.S. Schumacher, 2014: Objective Categorization of Heavy-Rain-
703 Producing MCS Synoptic Types by Rotated Principal Component Analysis. Mon. Wea.
704 Rev., 142, 1716–1737, https://doi.org/10.1175/MWR-D-13-00295.1.
705 Ramage, C. S., 1968: Role of a tropical “maritime continent” in the atmospheric circulation.
706 Mon. Wea. Rev., 96, 365–370.
707 Rasmussen, K. L., and R. A. Houze, Jr., 2011: Orogenic convection in South America as
708 seen by the TRMM satellite. Mon. Wea. Rev., 139, 2399-2420.
34 For Peer Review Page 36 of 100
709 Rasmussen, K. L., M. D. Zuluaga, and R. A. Houze Jr., 2014: Severe convection and
710 lightning in subtropical South America. Geophys. Res. Lett., 41, 7359–7366,
711 doi:10.1002/2014GL061767.
712 Rauber, R. M., and S. L. Nesbitt, 2018: Radar Meteorology: A first course. Wiley Blackwell,
713 467pp. ISBN-13: 978-1118432624
714 Reinhard, E., 2006: High dynamic range imaging: Acquisition, Display, and Image-Based
715 Lighting. Morgan Kaufmann, pp. 233–234.
716 Rinehart, R. E., 2001: Radar for Meteorologists. 3d ed. Rinehart, 428 pp.
717 Roca, R., Aublanc, J., Chambon, P., Fiolleau, T., & Viltard, N., 2014: Robust Observational
718 Quantification of the Contribution of Mesoscale Convective Systems to Rainfall in the
719 Tropics. J. Clim., 27(13), 4952-4958. http://dx.doi.org/10.1175/JCLI-D-13-00628.1
720 Romatschke, U. and R.A. Houze, 2010: Extreme Summer Convection in South America. J.
721 Clim., 23, 3761–3791, https://doi.org/10.1175/2010JCLI3465.1.
722 Rowe, A. K., and R. A. Houze, Jr., 2014: Microphysical characteristics of MJO convection
723 over the Indian Ocean during DYNAMO. J. Geophys. Res. Atmos., 119, 2543-2554,
724 doi:10.1002/2013JD020799.
725 Savtchenko, A., 2018: README Document for the NOAA NCEP/CPC Half Hourly 4km
726 Global (60S - 60N) Merged IR, 11 pp. [Available online at
727 https://disc2.gesdisc.eosdis.nasa.gov/data/MERGED_IR/GPM_MERGIR.1/doc/READM
728 E.MERGIR.pdf]
35 Page 37 of 100 For Peer Review
729 Schmetz, J., K. Holmlund, J. Hoffman, B. Strauss, B. Mason, V. Gaertner, A. Koch, L. Van
730 De Berg Operational cloud-motion winds from meteosat infrared images J. Appl.
731 Meteorol., 32 (7) (1993), pp. 1206-1225.
732 Schumacher, C., R. A. Houze Jr., and I. Kraucunas (2004), The tropical dynamical
733 response to latent heating estimates derived from the TRMM Precipitation Radar, J.
734 Atmos. Sci., 61, 1341–1358.
735 Schwaller, M.R. and K.R. Morris, 2011: A Ground Validation Network for the Global
736 Precipitation Measurement Mission. J. Atmos. Oceanic Technol., 28, 301–319,
737 https://doi.org/10.1175/2010JTECHA1403.1.
738 Skofronick-Jackson, G., D. Hudak, W. Petersen, S.W. Nesbitt, V. Chandrasekar, S.
739 Durden, K.J. Gleicher, G. Huang, P. Joe, P. Kollias, K.A. Reed, M.R. Schwaller, R.
740 Stewart, S. Tanelli, A. Tokay, J. R. Wang, and M. Wolde, 2015: Global Precipitation
741 Measurement Cold Season Precipitation Experiment (GCPEX): For Measurement’s
742 Sake, Let It Snow. Bull. Amer. Meteor. Soc., 96, 1719–1741,
743 https://doi.org/10.1175/BAMS-D-13-00262.1.
744 Starzec, M., C. R. Homeyer, and G. L. Mullendore, 2017: Storm Labeling in Three
745 Dimensions (SL3D): A Volumetric Radar Echo and Dual-Polarization Updraft
746 Classification Algorithm. Mon. Wea. Rev., 145, 1127-1145. 10.1175/mwr-d-16-0089.1
747 Song, F., Z. Feng, L. R. Leung, R. A. Houze, J. Wang, J. Hardin, and C. R. Homeyer,
748 2019: Contrasting spring and summer large-scale environments associated with
36 For Peer Review Page 38 of 100
749 mesoscale convective systems over the U.S. Great Plains. J. Clim., In revision.
750 TerrainBase, National Geophysical Data Center/NESDIS/NOAA/U.S. Department of
751 Commerce. 1995: Global 5 Arc-minute Ocean Depth and Land Elevation from the US
752 National Geophysical Data Center (NGDC). Research Data Archive at the National
753 Center for Atmospheric Research, Computational and Information Systems Laboratory.
754 https://doi.org/10.5065/E08M-4482. Accessed 28 Mar. 2019.
755 Toyoshima, K., H. Masunaga, and F. A. Furuzawa, 2015: Early evaluation of Ku- and Ka-
756 band sensitivities for the Global Precipitation Measurement (GPM) Dual-Frequency
757 Precipitation Radar (DPR). SOLA, 11, 14–17, doi:https://doi.org/10.2151/sola.2015-004.
758 Wang, J, X. Dong, and B. Xi, 2015: Investigation of ice cloud microphysical properties of
759 DCSs using aircraft in situ measurements during MC3E over the ARM SGP site. J.
760 Geophys. Res. Atmos., 120, 3533–3552. https://doi.org/10.1002/2014JD022795.
761 Wang, J., X. Dong, B. Xi, and A. J. Heymsfield, 2016: Investigation of liquid cloud
762 microphysical properties of deep convective systems: 1. Parameterization of raindrop
763 size distribution and its application for stratiform rain estimation, J. Geophys. Res.
764 Atmos., 121, 10,739–10,760, https://doi.org/10.1002/2016JD024941.
765 Wang, J., Dong, X., and Xi, B., 2018: Investigation of liquid cloud microphysical properties
766 of deep convective systems: 2. Parameterization of raindrop size distribution and its
767 application for convective rain estimation. J. Geophys. Res. Atmos, 123, 11,637–
768 11,651. https://doi.org/10.1029/2018JD028727.
37 Page 39 of 100 For Peer Review
769 Wang, J., X. Dong, A. Kennedy, B. Hagenhoff, and B. Xi, 2019: A Regime Based
770 Evaluation of Southern and Northern Great Plains Warm-Season Precipitation Events in
771 WRF. Wea. Forecasting, 0, https://doi.org/10.1175/WAF-D-19-0025.1
772 Xie, S., Y. Zhang, S. E. Giangrande, M. P. Jensen, R. McCoy, and M. Zhang, 2014:
773 Interactions between cumulus convection and its environment as revealed by the MC3E
774 sounding array, J. Geophys. Res. Atmos., 119, 11,784–11,808,
775 https://doi.org/10.1002/2014JD022011.
776 Yang, Q., R. A. Houze Jr., L. R. Leung, and Z. Feng, 2017: Environments of long-lived
777 mesoscale convective systems over the central United States in convection permitting
778 climate simulations. J. Geophys. Res. Atmos., 122, 13 288–13 307,
779 https://doi.org/10.1002/2017JD027033.
780 Zagrodnik, J. P., L. A. McMurdie, and R. A. Houze, Jr., 2018: Stratiform precipitation
781 processes in cyclones passing over a coastal mountain range. J. Atmos. Sci., 75, 983–
782 1004, doi:10.1175/JAS-D-17-0168.1.
783 Zhang, J., K. Howard, C. Langston, B. Kaney, Y. Qi, L Tang, H. Grams, Y. Wang, S.
784 Cocks, S. Martinaitis, A. Arthur, K. Cooper, J. Brogden, and D. Kitzmiller, 2015: Multi-
785 Radar Multi-Sensor (MRMS) Quantitative Precipitation Estimation: Initial Operating
786 Capabilities, Bull. Amer. Meteor. Soc., 97, 621–638, https://doi.org/10.1175/BAMS-D-
787 14-00174.1.
788 Zhang, J., K. Howard, C. Langston, S. Vasiloff, B. Kaney, A. Arthur, S. V. Cooten, K.
38 For Peer Review Page 40 of 100
789 Kelleher, D. Kitzmiller, F. Ding, D-J Seo, E. Wells, and C. Dempsey, 2011: National
790 Mosaic and Multi-Sensor QPE (NMQ) System: Description, Results, and Future Plans,
791 Bull. Amer. Meteor. Soc., 92, 1321–1338. doi: http://dx.doi.org/10.1175/2011BAMS-D-
792 11-00047.1.
793 Zipser, E. J., D. J. Cecil, C. Liu, S. W. Nesbitt, and D. P. Yorty, 2006: Where are the Most
794 Intense Thunderstorms on Earth? Bull. Amer. Meteor. Soc., 87, 1057–1072,
795 https://doi.org/10.1175/BAMS-87-8-1057
796 Zuluaga, M.D. and R.A. Houze, 2013: Evolution of the Population of Precipitating
797 Convective Systems over the Equatorial Indian Ocean in Active Phases of the Madden–
798 Julian Oscillation. J. Atmos. Sci., 70, 2713–2725, https://doi.org/10.1175/JAS-D-12-
799 0311.1.
800
801
802
803
804
805
806
807
808
39 Page 41 of 100 For Peer Review
809 List of Figures
810
811 Fig. 1 The domain for GPM and NEXRAD comparison is marked with the red box (110W
812 – 85W, 31N – 48N) overlaid on the CONUS elevation map.
813 Fig. 2 The contoured-frequency-by-altitude-diagrams (CFADs) normalized by the total
814 number of samples at all altitude levels for GPM (a, d, g, j, m, p) and NEXRAD (b, e, h,
815 k, n, q) for the months from April to September in 2014-2016 period. The box-whisker
816 plots (c, f, i, l, o, r) for GPM (red) and NEXRAD (blue) are calculated using
817 normalization at each individual level, where the center of the box represents the 50%
818 percentile value, the lower quartile (25%) and upper quartile (75%) form the left and
819 right boundary of the box, and whiskers correspond to the 5% and 95% values.
820 Fig. 3 (a) The Deep Wide Convection (DWC) object defined by GPM on 9 June 2014 at
821 04:26:46 UTC and (b) the coincident NEXRAD observed MCS tracked by FLEXTRKR
822 at 04:00 UTC, as well as (c) the intersection of GPM overpass time (dash-line) with
823 respect to the evolution of the MCS precipitation area during its entire life cycle
824 identified by the NEXRAD.
825 Fig. 4 Monthly total number of MCSs HIT (GPM detection validated by NEXRAD, red
826 bars) and false alarm (FAR, not identified as MCS by NEXRAD, blue bars), overlaid by
827 the monthly total number of MCSs detection by NEXRAD (black line) and their average
828 precipitation area coverage (green line).
40 For Peer Review Page 42 of 100
829 Fig. 5 Geographical distribution of the probability of MCS occurrence frequency detected
830 by GPM during the months of (a) JJA and (b) DJF. The gray shaded areas inside the
831 continental regions represent the 700 m elevation. The probability is on a scale of 0 to
832 100% and is computed as the number of pixels identified as either DWC or WCC
833 divided by the total number of GPM overpasses within a 0.25 × 0.25 gridbox.
834 Fig. 6 Similar to Figure 5 but using the moderate criteria.
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
41 Page 43 of 100 For Peer Review
851
852
853 Fig. 1 The domain for GPM and NEXRAD comparison is marked with the red box (110W
854 – 85W, 31N – 48N) overlaid on the CONUS elevation map.
855
856
857
42 For Peer Review Page 44 of 100
858
859 Fig. 2 The contoured-frequency-by-altitude-diagrams (CFADs) normalized by the total
860 number of samples at all altitude levels for GPM (a, d, g, j, m, p) and NEXRAD (b, e, h,
861 k, n, q) for the months from April to September in 2014-2016 period. The box-whisker
862 plots (c, f, i, l, o, r) for GPM (red) and NEXRAD (blue) are calculated using
863 normalization at each individual level, where the center of the box represents the 50%
864 percentile value, the lower quartile (25%) and upper quartile (75%) form the left and
865 right boundary of the box, and whiskers correspond to the 5% and 95% values.
866
867
868
43 Page 45 of 100 For Peer Review
869
870 Fig. 3 (a) The Deep Wide Convection (DWC) object defined by GPM on 9 June 2014 at
871 04:26:46 UTC and (b) the coincident NEXRAD observed MCS tracked by FLEXTRKR
872 at 04:00 UTC, as well as (c) the intersection of GPM overpass time (dash-line) with
873 respect to the evolution of the MCS precipitation area during its entire life cycle
874 identified by the NEXRAD.
875
876
877
878
879
44 For Peer Review Page 46 of 100
880 881 Fig. 4 Monthly total number of MCSs HIT (GPM detection validated by NEXRAD, red
882 bars) and false alarm (FAR, not identified as MCS by NEXRAD, blue bars), overlaid by
883 the monthly total number of MCSs detection by NEXRAD (black line) and their average
884 precipitation area coverage (green line).
885
886
887
45 Page 47 of 100 For Peer Review
888
889 Fig. 5 Geographical distribution of the probability of MCS occurrence frequency detected
890 by GPM during the months of (a) JJA and (b) DJF. The gray shaded areas inside the
891 continental regions represent the 700 m elevation. The probability is on a scale of 0 to
892 100% and is computed as the number of pixels identified as either DWC or WCC
893 divided by the total number of GPM overpasses within a 0.25 × 0.25 gridbox.
894
895
896
897
46 For Peer Review Page 48 of 100
898
899 Fig. 6 Similar to Figure 5 but using the moderate criteria.
900
901
902
903
904
905
906
907
908
909
910
47 Page 49 of 100 For Peer Review
911 List of Tables
912 Table 1 The statistics of GPM’s MCS detection skill in each month.
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
48 For Peer Review Page 50 of 100
937 Table 1 The statistics of GPM’s MCS detection skill in each month. Average Average GPM False GPM GPM GPM HIT NEXRAD MCS MCS Negative HIT (#) FAR (#) Rate MCS (#) Coverage Duration Rate (km2) (hour) April 43 8 84% 16% 104 20,374 21.5 May 47 9 84% 16% 140 13,742 20.8 June 58 26 69% 31% 160 11,961 20.4 July 61 23 73% 27% 137 10,683 20.9 August 37 31 54% 46% 122 10,885 20.6 September 27 18 60% 40% 77 10,141 19.8
938
939
940
49 Page 51 of 100 For Peer Review
1 The Detection of Mesoscale Convective Systems by the
2 GPM Ku-band Spaceborne Radar
3
4 Jingyu Wang
5 Pacific Northwest National Laboratory, 6 Richland, Washington, USA
1 7 Robert. A. Houze, Jr
8 University of Washington, 9 Seattle, Washington, USA 10 Pacific Northwest National Laboratory, 11 Richland, Washington, USA
12 Jiwen Fan
13 Pacific Northwest National Laboratory, 14 Richland, Washington, USA
15 Stacy. R. Brodzik
16 University of Washington, 17 Seattle, Washington, USA
18 Zhe Feng
19 Pacific Northwest National Laboratory, 20 Richland, Washington, USA
21 and
22 Joseph C. Hardin
23 Pacific Northwest National Laboratory, 24 Richland, Washington, USA 25 26 27 March 31, 2019 28 ------29 1) Corresponding author: Robert Houze, University of Washington, Seattle, Washington, 30 USA. Email: [email protected] For Peer Review Page 52 of 100
31 Abstract
32
33 The Global Precipitation Measurement (GPM) core observatory satellite launched in
34 2014 features more extended latitudinal coverage (65S-65N) than its predecessor
35 Tropical Rainfall Measuring Mission (TRMM, 35S-35N). The Ku-band radar onboard of
36 the GPM is known to be capable of characterizing the 3D structure of deep convection
37 globally. In this study, GPM’s capability for detecting mesoscale convective systems
38 (MCSs) is evaluated. Extreme convective echoes seen by GPM are compared against an
39 MCS database that tracks convective entities over the contiguous US. The tracking is
40 based on geostationary satellite and ground-based Next Generation Radar (NEXRAD)
41 network data obtained during the 2014-2016 warm seasons. Results show that more than
42 70% of the GPM-detected Deep-Wide Convective Core (DWC) and Wide Convective Core
43 (WCC) objects are part of NEXRAD identified MCSs, indicating that GPM-classified DWCs
44 and WCCs correlate well with typical MCSs containing large convective features. By
45 applying this method to the rest of the world, a global view of MCS distribution is obtained.
46 This work reveals GPM’s potential in MCS detection at the global scale, particularly over
47 remote regions without dense observation network.
48
49 Keywords GPM evaluation; 3D reflectivity structure; Mesoscale Convective System
50 tracking; MCS features; Intense convection; Global MCS distribution
1 Page 53 of 100 For Peer Review
51 1. Introduction
52 As a collaborative effort between the National Aeronautics and Space Administration
53 (NASA) and the Japan Aerospace Exploration Agency (JAXA), the Tropical Rainfall
54 Measuring Mission (TRMM) satellite was equipped with a Ku-band (13.8 GHz) quantitative
55 precipitation radar, together with a variety of sensors including passive microwave, visible,
56 infrared, and lightning (Kummerow et al., 1998; Schumacher et al. 2004; Zipser et al.
57 2006; Houze et al. 2015). Launched in 2014 as the successor of TRMM, the Global
58 Precipitation Measurement (GPM) core observatory satellite carries the first space-borne
59 Dual-frequency Precipitation Radar (DPR) operating at both Ku (13.6 GHz) and Ka (35.5
60 GHz) bands. Compared to TRMM, the DPR system helps improve the accuracy of
61 precipitation measurement and upgrade the detectability of weak rain as low as 0.5 mm hr-
62 1. Another key advancement of GPM is its extended coverage to higher latitudes (65S to
63 65N compared to 35S to 35N for TRMM), providing a near-global view of 3D cloud and
64 precipitation structure every 2-3 hours.
65 Mesoscale Convective Systems (MCSs) are of great importance because of their large
66 area (at least one hundred kilometers in one direction) and intense, long-lasting (up to 24
67 hours) precipitation (Houze 2004, 2018). In midlatitudes, MCSs strongly impact local
68 climate through their precipitation, severe weather, and redistribution of heat and moisture,
69 which further impact the regional to global hydrological cycle and large-scale circulations
70 (Houze et al. 1990; Feng et al. 2016, 2018; Futyan and Del Genio, 2007). Based on the
2 For Peer Review Page 54 of 100
71 detailed 3D radar reflectivity field observed by spaceborne precipitation radar, Houze et al.
72 (2007, 2015) developed a convective echo classification algorithm based on the horizontal
73 and vertical dimensions of echoes of a given intensity. The most intense of such 3D
74 convective echo objects are assumed to be strongly associated with MCSs (Houze et al.
75 2015, 2019). However, direct comparison to other independent MCS data sets is needed
76 to support this argument. Identification of MCSs has been traditionally carried out by
77 tracking convective elements in geosynchronous satellite or ground-based radar data,
78 which provide temporally continuous spatial data. For example, satellite imagery that
79 provides cloud radiative properties such as brightness temperature at high temporal
80 resolution has been used to track deep convective clouds at regional to global scale
81 (Schmetz et al. 1993; Machado et al. 1998; Morel and Senesi 2002; Héas and Mémin
82 2008; Escrig et al. 2013; Roca et al. 2014). However, the incapacity of visible and infrared
83 satellite data to show the structure of the clouds in the lower troposphere leaves large
84 uncertainties in satellite-based MCS tracking. Ground-based radar measurements provide
85 3D structure of precipitating deep convection that can complement satellite imagery to
86 more accurately detect MCSs. A recently developed tracking algorithm called FLEXible
87 object TRacKeR (FLEXTRKR, Feng et al. 2018) jointly uses the geostationary satellite
88 brightness temperature and NEXRAD 3D radar reflectivity structure to identify and track
89 MCSs. A comprehensive 13-year (2004-2016) MCS tracking database east of the Rocky
90 Mountains has been developed using FLEXTRKR (Feng et al. 2019). This database is
3 Page 55 of 100 For Peer Review
91 considered here as ground-truth to evaluate the MCS objects detected by GPM.
92 To determine GPM’s capability of MCS detection, we organize this study as follows. In
93 Section 2, the data sets of GPM Ku-band reflectivity, NEXRAD observations, as well as
94 their corresponding MCS detecting/tracking algorithms are introduced. In Section 3, the
95 quantitative comparison of 3D radar reflectivity fields is performed between the two data
96 sets. This comparison is essential to determine whether the GPM and NEXRAD radar
97 systems detect the 3D radar reflectivity fields in a consistent way. After the consistency
98 between the two data sets is established, the GPM’s snapshots of intense convective echo
99 objects are compared to the NEXRAD MCS tracking database over the Continental United
100 States (CONUS) to quantify GPM’s MCS detection capability. Then, the regionally
101 validated GPM’s MCS detection algorithm is applied elsewhere around the world,
102 presenting a global view of MCS distribution. Finally, conclusions and discussions are
103 provided in Section 4.
104 2. Data and methodology
105 2.1 GPM Data and convective object classification
106 As a major component of the GPM DPR system, the onboard Ku-band precipitation
107 radar features a swath width of 245 km. The horizontal resolution is 0.05, or
108 approximately 5 km. The vertical resolution is 125 m, provided in 176 levels. In this study,
109 the GPM Ku band radar reflectivity data were downloaded from the University of
110 Washington GPM-Ku Data Set located at http://gpm.atmos.washington.edu, which
4 For Peer Review Page 56 of 100
111 geolocates and interpolates the DPR Level 2A Ku band version 5B data (Iguchi et al.
112 2017) from “radar coordinates” to a Cartesian grid with the aforementioned resolutions.
113 Detailed methodology regarding the geolocation correction and interpolation can be found
114 in Houze et al. (2007).
115 NASA and JAXA divide the GPM Ku-band radar echoes into convective, stratiform, and
116 other categories. In this study, the echo-object classification scheme developed by Houze
117 et al. (2007, 2015) is applied to the convective echoes. The analysis for evaluating GPM
118 3D radar reflectivity and GPM’s MCS detection capability is conducted in the CONUS
119 region for the warm seasons (April-September) of 2014-2016. The classification scheme
120 defines convective echoes as deep, wide, or deep and wide depending on radar reflectivity
121 intensity thresholds, and criteria for convective echo object height and area. Both strong
122 and moderate criteria (as defined by Houze et al. 2015) are used in this study to examine
123 the sensitivity of the results to the criteria. Strong/moderate deep convective core (DCC)
124 objects contain echoes that are greater than or equal to 40 dBZ/30 dBZ (everywhere in the
125 column) in intensity and have a maximum altitude of at least 10 km/8 km. Strong/moderate
126 wide convective core (WCC) objects contain echoes that are greater than or equal to 40
127 dBZ/30 dBZ and have a maximum horizontal extent of at least 1000 km2/800 km2. The
128 category of DCC has no areal coverage criterion, and is believed to be associated with
129 young, vigorous convection. The WCCs have no height criterion, but rather correspond to
130 where the intense convection has organized horizontally upscale into mesoscale areas of
5 Page 57 of 100 For Peer Review
131 active, widespread convection. Although the two categories are classified independently,
132 they do have overlap. Deep and wide convective core (DWC) objects meet the criteria for
133 both DCCs and WCCs. The wide categories, DWC and WCC, are thought to be strongly
134 related to the MCSs, as a typical MCS contains a large intense precipitation area with a
135 major axis longer than ~100 km, which is similar to the coverage criteria for the DWC and
136 WCC echo objects.
137 2.2 NEXRAD data and the MCS database
138 Densely distributed across the entire U.S., the NEXRAD network consists of 159 high
139 resolution S-band Doppler radars (WSR-88D) operated by the National Weather Service
140 (NWS). This study uses the mosaic NEXRAD data set named the Gridded Radar data
141 (GridRad, Bowman and Homeyer 2017) that combines all NEXRAD radar data covering
142 the region 155W – 69W, 25N – 49N. The GridRad data set has 2 km horizontal and 1
143 hour temporal resolutions, and a fixed 1 km vertical resolution (24 levels).
144 The 13-year MCS database (2004-2016) developed by Feng et al. (2019) uses both
145 geostationary satellite infrared brightness temperature (Tb) and GridRad radar data to
146 identify and track MCSs east of the Rocky Mountains (110W – 70W, 25N – 49N). The
147 native GridRad radar data were regridded to 4 km resolution to match the geostationary
148 satellite data. In Feng et al. (2019), an MCS was defined when a convective cloud system
4 149 satisfies both criteria in size (cold cloud system with Tb < 241 K area greater than 6 x 10
150 km2, and the radar-observed precipitation feature major axis length exceeds 100 km, with
6 For Peer Review Page 58 of 100
151 embedded convective core > 45 dBZ) and duration (persistence longer than 6 hours). The
152 identification of a convective core is based on the storm labeling in three dimensions
153 (SL3D) classification (Starzec et al. 2017). In SL3D, convective core is labeled if any one
154 of the following criteria is satisfied: (1) 25 dBZ echo-top height ≥ 10 km, (2) echo
155 peakedness ≥ 50% of the echo column between the surface and 9 km, and (3) 45 dBZ
156 echo-top height is above the melting-layer height. In the MCS database (Feng et al. 2019),
157 the melting-layer height is determined using the 6-hourly ERA-Interim reanalysis (Dee et
158 al. 2011). After an MCS is tracked, its life cycle is further separated into the initiation,
159 genesis, mature, and dissipation stages. Initiation starts at the first hour when the first
160 MCS-related cold cloud system is detected, followed by the genesis stage when the major
161 axis length of the convective core exceeds 100 km. As the convective core maintains its
162 size, the upscale growth of the MCS’s stratiform rain area concludes the genesis
163 stage. Finally, when the convective core length falls below the 100 km threshold, or the
164 stratiform rain area is lower than the mean value throughout the entire MCS life cycle, the
165 system is classified as in the dissipation stage.
166 2.3 Matching the GPM data to the NEXRAD Coordinate
167 The GPM Ku-band radar operates at a wavelength of 2.2 cm, which is capable of
168 detecting large raindrops or snowflakes (Rauber and Nesbitt 2018), while the 3 GHz
169 ground-based NEXRAD radar has a wavelength is 10 cm. The GPM Ku-band radar has a
170 swath width of 245 km, which is equivalent to the range of a single NEXRAD radar
7 Page 59 of 100 For Peer Review
171 (maximum operational range of 230 km, although better coverage and resolution is
172 typically provided within 150 km radius of the radar). The GPM radar scans downward,
173 nearly vertically and it has a vertical resolution ~100 m, while the NEXRADs scan quasi-
174 horizontally with a beamwidth of ~1 degree so that at great horizontal distance the vertical
175 resolution is coarse. More importantly, there is a temporal mismatch between the two data
176 sets, as the GPM provides instantaneous snapshots of 3D radar reflectivity field whereas
177 the GridRad temporally averages the observations from different individual NEXRAD
178 radars within a 9-minute window. The different wavelengths, scanning geometries, and
179 data processing of the GPM radar and NEXRAD require that care be taken in comparing
180 data from the two systems.
181 Evaluation of the GPM Ku-band calibration and its attenuation correction has been
182 performed throughout its pre-launch and post-launch stages in multiple field campaigns,
183 such as the 2011 Mid-Continent Convective Clouds Experiment (MC3E, Wang et al. 2015;
184 Jensen et al. 2016), the 2012 GPM Cold-season Precipitation Experiment (GCPEx,
185 Skofronick et al. 2015), the 2014 Integrated Precipitation and Hydrology Experiment
186 (IPHEX, Grecu et al. 2018), and the 2015-2016 Olympic Mountain Experiment
187 (OLYMPEX, Houze et al. 2017, McMurdie et al. 2018; Zagrodnik et al. 2018). These field
188 experiments have found the GPM Ku-band radar data are consistent with the 10 cm
189 wavelength ground-based radar measurements.
8 For Peer Review Page 60 of 100
190 In comparing radar datasets from GPM and NEXRAD, differences can result from
191 processing the data at different spatial resolution. It is therefore critical for us to
192 quantitatively examine the consistency of the interpolated 3D radar reflectivity databases.
193 The Earth System Modeling Framework (ESMF) package
194 (https://www.earthsystemcog.org/projects/esmpy/) has been used to bi-linearly re-
195 interpolate the GPM 3D reflectivity data (spatial resolution of 0.05o, vertical resolution of
196 125 m) to the GridRad coordinate (4 km horizontal, 1 km vertical resolution). The hourly
197 NEXRAD data are extracted at the GPM overpass locations and the nearest coincident
198 hour. As described in Bowman and Homeyer (2017), the GridRad data are generated
199 using the 4-dimensional binning (averaging) algorithm that merges multiple individual radar
200 volumes (the mean from 4-minute before to 4-minute after the sample time). To make a
201 fair comparison between the averaged GridRad reflectivity to the instantaneous GPM
202 reflectivity measurements, we apply a Gaussian smoothing technique (Reinhard 2006)
203 vertically to the GPM data to mimic the binning procedure used in GridRad. Following a
204 suggestion from C. Homeyer (2018, personal communication), a 1 km running mean along
205 the vertical direction, and a Gaussian smoothing with kernel width of 9 and sigma value of
206 2.5 are applied to the GPM reflectivity field. After the horizontal regridding and vertical
207 smoothing, the two data sets are in the same coordinates. The smoothing procedure plays
208 a significant role in matching the two observation data sets. If not smoothed, the raw
209 GPM’s reflectivity values are systematically higher than the NEXRAD data by 20-25%.
9 Page 61 of 100 For Peer Review
210 Note the minimum detectable reflectivity value of GPM Ku-band radar is between 12
211 dBZ (Hou et al. 2014; Toyoshima et al. 2015; Hamada and Takayabu 2016) and 13 dBZ
212 (Olson et al. 2016). Through the examination of all available Ku-band radar data, it is
213 extremely rare to find reflectivity lower than 13 dBZ (with occurrence frequency of 5 x 10-8).
214 As a result, this study adopts the higher threshold of 13 dBZ, and the corresponding
215 NEXRAD reflectivity is truncated accordingly for fair comparison.
216 As revealed in previous studies (e.g., Krajewski et al. 2006; Cui et al. 2016 and 2017),
217 NEXRAD radar beam blockage is a common issue over mountainous regions, making the
218 radar reflectivity comparison with GPM difficult. However, the MCSs commonly initiate
219 from the lee side of the Rocky Mountains (Feng et al. 2019). Therefore we choose the
220 study domain just east of the Rocky Mountains Front Range (eastern Colorado). In
221 addition, NEXRAD radars deployed along the coast have extended observations over the
222 ocean. Those observations lack overlap with other radars and their sample volume is
223 larger than normal (Rinehart 2001), thus the data quality offshore is questionable.
224 Moreover, from the perspective of MCS tracking, convection farther offshore away from
225 the coastal NEXRADs, are prone to have larger uncertainty because of reduced low-level
226 radar coverage. As a result, through the examination of CONUS elevation data
227 (TerrainBase, 1995), the study domain is constrained to the central U.S. as shown in Fig. 1
228 (110W – 85W, 31N – 48N), which avoids most of the complex terrain over Rocky and
229 Appalachian Mountains and excludes all offshore areas.
10 For Peer Review Page 62 of 100
230 3. Results
231 3.1 The comparison of 3D radar reflectivity between GPM and NEXRAD
232 Vertical echo structure is indicative of the kinematic, dynamic, and thermodynamic
233 features of convective clouds (Hence and Houze 2008, 2011, 2012a, b), as well as cloud
234 microphysical properties (Leary and Houze 1979; Cetrone and Houze 2011; Rowe and
235 Houze 2014; Fan et al. 2015, 2017; Liu et al. 2015; Barnes and Houze 2016; Han et al.
236 2019). As mentioned in Section 2.3, the GPM data are instantaneous snapshots of 3D
237 radar reflectivity field from above whereas the NEXRAD provides the mean composite
238 data from multiple ground radars. This may result in temporal mismatch between the two
239 measurements, since we have binned the GPM overpasses to the nearest hour, i.e., the
240 hourly NEXRAD data are compared to the GPM overpasses that occur within the time
241 window of +/- 30 minutes for each hour. Previous studies (e.g., Feng et al. 2009, Wang et
242 al. 2016 and 2018) have shown that the time mismatch can significantly impact the
243 reflectivity measurements between different radars, as convective features could change
244 substantially in evolution and/or location on a time scale less than an hour. Therefore, we
245 adopt the statistical method of contoured-frequency-by-altitude-diagrams (CFADs) to
246 compare the vertical distribution of radar reflectivity measurements from the two radar
247 platforms.
248 As shown in Fig. 2, CFADs are generated for GPM (a, d, g, j, m, p) and NEXRAD (b, e,
249 h, k, n, q) from all the collocated data for each month (April - September) within the
11 Page 63 of 100 For Peer Review
250 sampling area for the 2014-2016 period. The CFADs display the frequency distribution in a
251 coordinate system of reflectivity bins (x-axis) and the altitude (y-axis), and represent the
252 occurrence frequency of reflectivity spectra normalized by the total number of bins in which
253 reflectivity is recorded. The bin sizes in reflectivity and height are 1 dBZ and 1 km,
254 respectively. The frequency is calculated as the number of non-zero echo values in a bin
255 divided by the total number of bins containing reflectivity (all heights). The integral of the
256 frequency over the entire 2D graph equals 1.
257 From the examination of the CFADs structure, the GPM is in good agreement with
258 NEXRAD for each month regardless of the overall shape or magnitude of the frequency at
259 various altitudes. Both data sets demonstrate consistent seasonal variations: the deepest
260 echo top at various reflectivity thresholds increases from spring (April - May) to summer
261 (June - July - August), then decreases towards the early fall (September). This seasonality
262 is also seen for the altitude change of the frequency contour of 0.6% and above. The peak
263 altitudes of this higher frequency are broadly distributed below 6 km in spring, and the
264 altitudes increase to 9 km in the summer, and finally decrease to lower altitude in early fall.
265 The seasonal changes in echo-top heights and the altitude of the higher frequency
266 reflectivity values implies the seasonal variation of the convective intensity. The convective
267 updraft intensity maximum in summer is consistent with the largest convective available
268 potential energy being observed at that time (Xie et al., 2014). Meanwhile, the seasonality
269 in the altitude of higher frequency values is also indicative of the shift in storm types
12 For Peer Review Page 64 of 100
270 between seasons. Spring and fall storms tend to have more stratiform clouds with bottom-
271 heavy reflectivity profiles (e.g., strong bright-band signature near the melting level),
272 resulting in higher reflectivity at mid-to-lower levels. During the summer, more occurrence
273 of convective clouds shifts higher reflectivity values aloft.
274 The radar reflectivity distributions normalized at each level are also examined using the
275 box-whisker plots shown in Fig. 2c, f, i, l, o, r. General agreements are also found for the
276 median, interquartile and extreme values between the two data sets at different altitudes
277 for every month. The differences are commonly less than 2 dBZ except for the 95th
278 percentiles above 12 km, where the NEXRAD shows larger reflectivity values than the
279 GPM. From the CFADs, the frequency of occurrence for these extreme values are very
280 low, thus the difference there is not of particular concern.
281 3.2 Evaluation of GPM’s capability in MCS detection
282 Having established that the GPM and WSR-88D reflectivity fields are consistent on the
283 interpolated grids, we now assess GPM’s capability in MCS detection. As mentioned in
284 Section 2.1, it has been suggested that the GPM DWC and WCC echo objects correlate
285 with MCSs, which feature convection organized on larger horizontal scale. Here we test
286 this argument by comparing GPM DWC and WCC echo objects (strong criteria), which are
287 observed as snapshots, to the MCS database, which is unambiguously determined from
288 tracking convective features in time. Figure 3 shows one example of a GPM-defined DWC
289 echo object and the coincident MCS at the nearest hour from the NEXRAD data, as well
13 Page 65 of 100 For Peer Review
290 as the timing of GPM’s overpass with the tracked MCS’s life cycle. By comparing Fig. 3a
291 (GPM) and Fig. 3b (NEXRAD), the collocated radar reflectivity fields as observed from the
292 two platforms show that they are in as good agreement as could be expected. The GPM-
293 defined DWC snapshot occurred during the time period that the MCS tracked by the
294 FLEXTRKR algorithm was in the genesis stage (Fig. 3c), meaning it was experiencing
295 upscale growth at the time of the GPM observation.
296 In order to form long-term statistics, all 158 DWC objects and 230 WCC objects
297 detected by GPM (strong criteria) within the three warm seasons are evaluated with the
298 NEXRAD MCS database. By overlaying the mask of GPM-defined DWC or WCC object to
299 the mask of the NEXRAD-tracked MCS cold cloud shield, we can determine if the GPM
300 DWC or WCC objects are part of a NEXRAD-defined MCS. If overlap is found, then this
301 GPM-defined DWC or WCC object is considered as a HIT. Otherwise, it is treated as a
302 false alarm (FAR) because the GPM snapshot of the convective echo objects does not
303 align with any NEXRAD-tracked MCS. Note we do not define a missing category (i.e.,
304 MCSs tracked by the NEXRAD but not identified by the GPM) because of the substantial
305 difference in spatial coverage. As described in Section 2, the GPM has a limited swath
306 width of 245 km, where the NEXRAD network covers the entire CONUS. Therefore, the
307 “missing” detection from the GPM is highly possible simply because there is no GPM
308 overpass for a particular MCS.
14 For Peer Review Page 66 of 100
309 In addition to the objective comparison, all the GPM DWC/WCC and collocated
310 NEXRAD reflectivity images are manually screened for further confirmation. For the GPM-
311 detected echo objects in the FAR category, there are several cases that the GPM
312 detections are closely located (usually within ~5 km) to the NEXRAD-tracked MCS but
313 without overlap. We suspect that this could be caused by the temporal mismatch between
314 the two data sets as aforementioned and should be treated as HIT instead. As a result, by
315 including the NEXRAD MCS masks from 1 hour before to 1 hour after the coincident hour,
316 overlaps are found for those GPM detections (11 cases).
317 The final result shows that 115 of 158 DWC objects and 158 of 230 WCC objects can
318 be verified as MCSs when compared to the NEXRAD data set, leaving 115 cases as FAR.
319 In summary, 70% of DWC and WCC systems classified by the GPM are MCSs identified in
320 the NEXRAD data set. Interestingly, most of the GPM-detected MCS snapshots are during
321 the genesis (41%) and mature stages (32%) of the tracked MCSs. There are only about
322 21% in the dissipation stage and 6% in the initiation stage. These results are consistent
323 with the finding of Feng et al. (2019) that warm season MCS convective features during
324 the upscale growth stage are the largest and deepest. These convective features are most
325 likely to meet the GPM DWC or WCC criteria.
326 After revisiting the 115 cases in the FAR category by examining the NEXRAD images
327 before and after the GPM detection, 19% of them last more than 6 hours but their major
328 axis length of precipitation area couldn’t exceed 100 km to satisfy the MCS criteria defined
15 Page 67 of 100 For Peer Review
329 by FLEXTRKR. For the rest of 81% cases, they all dissipate too soon and therefore fail to
330 satisfy the MCS duration requirement of 6 hours. These false alarms reveal the intrinsic
331 limitation of the criteria used to identify MCSs by tracking NEXRAD echoes and the
332 conditions used to define the GPM echo-object categories are arbitrary.
333 Because the majority of false alarms result from the insufficient duration, a question
334 arises as to whether the HIT rate (defined as HITs / (HITs + FARs)) could increase with
335 lowering the MCS duration threshold. Some previous studies defined MCSs in this region
336 using shorter duration of 4 hours (e.g., Geerts 1998; Haberlie and Ashley 2019). To
337 examine the impact of MCS duration criterion to our results, we performed a sensitivity test
338 by reducing the MCS duration threshold from 6 hours to 4 hours in FLEXTRKR, and the
339 total number of MCSs tracked by NEXRAD in the 3 warm seasons increased from 740 to
340 1193. However, only 12 GPM FAR cases are changed to the HIT category. Upon close
341 examination of the FAR cases, we found several reasons that explain this result. First,
342 90% of the additional 453 NEXRAD-tracked MCS cases occur outside the GPM
343 overpasses, thus they have no impact to the GPM statistics. Secondly, for those short-
344 lived false alarms detected by the GPM, their lifespan are commonly less than 3 hours,
345 which fail to meet even the shortened MCS duration criteria. Lastly, short duration and
346 insufficient coverage are not exclusive, i.e., many GPM-detected objects that do not last
347 longer than 6 hours also do not satisfy the size criteria. In this case, it is complex to
348 differentiate the actual causes of false alarms. The fact that changing the MCS duration
16 For Peer Review Page 68 of 100
349 threshold produces limited impact to GPM’s MCS statistics may further confirm the
350 correspondence between the majority of GPM-detected DWCs/WCCs and the largest,
351 deepest MCSs, as the latter require longer period of time to form. However, there remain
352 exceptions of short-lived DWCs/WCCs, which raise the necessity of the temporal
353 dimension. More extreme sensitivity test of 2 hours duration threshold is performed, now
354 the systems tracked by the NEXRAD increases to 1671 and 49 false alarms change to the
355 category of HIT, making the HIT rate 83%. However, this comparison may not be
356 meaningful because the 2 hours duration is too short for MCS definition, which makes
357 almost no difference to the direct radar echo comparison between the two platforms.
358 Based on these tests, we conclude that 70% accuracy reached in our first comparison
359 remains a good overall estimate of the capability to determine MCS existence from the
360 GPM radar data.
361 By using the original MCS tracking data set as reference, Fig. 4 and Table 1 show the
362 GPM’s detection skills in each month, where the number of HIT, FAR, as well as the
363 number of NEXRAD-tracked MCSs and their averaged precipitation area coverage and
364 duration are compared. The number of NEXRAD-tracked MCSs demonstrates a strong
365 seasonality, which increases from April to June, then diminishes towards early fall. This
366 variation follows the seasonal change in baroclinic instability over the CONUS. During
367 spring, large-scale forcing brought by the mid-latitude trough frequently occurs, providing a
368 favorable environment for organized convection (Maddox et al 1979; Peters and
17 Page 69 of 100 For Peer Review
369 Schumacher 2014). In April, MCSs produced under strong baroclinic forcing usually
370 feature broad stratiform rain region, which results in the largest precipitation area, but the
371 number is relatively low compared to the midsummer, when the CONUS has the weakest
372 baroclinic instability and minimal frontal forcing. However, due to the favorable
373 thermodynamic conditions and possible influence from sub-synoptic disturbances, local
374 convection can nevertheless frequently grow upscale into MCSs (Wang et al. 2011; Song
375 et al. 2019; Feng et al. 2019). The combination of spring-like baroclinic waves and a
376 continuously warming surface possibly favors the peak number of MCSs in late spring and
377 early summer (i.e., June), but the MCS precipitation areas are generally smaller than in
378 spring. Although fall has similar large-scale environments as in spring, MCSs are less
379 frequent and no increase in average coverage is found. One possible explanation is the
380 seasonality in surface temperature gradient, which is weaker in fall than in spring, possibly
381 causing weaker baroclinic waves (Feng et al. 2019). The average MCS duration was also
382 calculated for each month, and no seasonal variation was found.
383 From the perspective of MCS detection by the GPM, several factors contribute to a
384 higher HIT rate: more frequent MCS occurrence, larger precipitation area and longer
385 duration (i.e., the propagating nature results in larger area covered by MCSs). The trend of
386 monthly HIT is consistent with the variation of the MCS numbers, indicating that the
387 probability of GPM detected MCSs increases with more frequent MCS occurrence. Spring
388 corresponds to the lowest False Negative Rate (defined as FARs / (HITs + FARs)), which
18 For Peer Review Page 70 of 100
389 is consistent with the larger MCS coverage favored by the synoptic forcing. Thus higher
390 accuracy in GPM’s MCS detection can be expected relative to the rest of the months. In
391 contrast, for the August-September period, as the result of poleward expansion of the
392 subtropical ridge (Wang et al. 2019), the baroclinic environment for supporting large MCSs
393 no longer exists, resulting in a higher False Negative Rate.
394 3.3 Global MCS distributions
395 After establishing the GPM’s MCS detection capability of about 70% by validating
396 against the NEXRAD-tracked MCS database over CONUS in the previous section, we can
397 apply the DWC and WCC criteria to the GPM data globally to examine the probably global
398 distribution of MCSs. Because ground-based radar networks like NEXRAD do not exist
399 over most of Earth, GPM is the best available resource for determining the global pattern
400 of MCS occurrence. Figure 5 shows the geographical distribution of MCS occurrence
401 frequency determined from GPM during the boreal summer June-July-August (JJA, Fig 5a)
402 and winter December-January-February (DJF, Fig 5b) over the 5-year period (2014-2018)
403 with available GPM observations. The frequency is computed as the number of pixels
404 identified as either DWC or WCC divided by the total number of pixels sampled by the
405 GPM Ku-band radar within a 0.25 x 0.25 gridbox. It is evident that the occurrence of
406 MCSs is more concentrated over land, which is consistent with previous studies showing
407 that convection over vast oceans is generally less intense than over land (e.g., Futyan and
408 Del Genio 2007; Houze et al. 2015). A comparison between Fig. 5a and Fig. 5b shows that
19 Page 71 of 100 For Peer Review
409 MCSs occur more frequently during the boreal summer than winter for both hemispheres.
410 In North America during summer, MCSs are densely distributed over the Great Plains
411 (GP), where they are fed by warm moist air transported from the Gulf of Mexico by the
412 climatological low-level jet on the lee side of the Rocky Mountains. In both summer and
413 winter, MCSs frequently occur offshore of the east coast of North America. These MCSs
414 tend to result from the initiation of gravity waves on the lee side of the Appalachian
415 Mountains (Keighton et al. 2007; Letkewicz and Parker 2010). In winter, MCSs are absent
416 over the GP but occur over the southeast and the offshore of the east coast of the U.S.,
417 consistent with NEXRAD-based MCS frequencies reported by Feng et al. (2019).
418 Over the landmass of tropical South America and central equatorial Africa, large
419 clusters of MCSs occur during JJA, but the hot zones of MCSs shift southward to the
420 subtropics and mid-latitudes in DJF, consistent with previous studies (e.g., Romatschke
421 and Houze 2010; Rasmussen and Houze 2011; Rasmussen et al. 2014). The high-
422 frequency MCS areas are displaced from the areas with the most regional rainfall (Houze
423 2015). Over Asia, three hot zones are identified in JJA, namely the Indian monsoon region,
424 east Asian monsoon region, and the maritime continent consisting Indonesia, Malaysia,
425 and Northern Australia (Ramage 1968). The former two regions are strongly influenced by
426 the summer monsoon flow, hence the MCSs peak in JJA but diminish in DJF. In contrast,
427 the maritime continent MCS occurrence is dominated by diurnal forcing associated with
428 the islands and peninsulas of the region, and in DJF the diurnal convection is enhanced by
20 For Peer Review Page 72 of 100
429 surges of the boreal winter monsoon (Johnson and Houze 1987) and modulated by
430 passages of the Madden-Julian Oscillation (Madden and Julian, 1971,1972, 1994).
431 Note all of the above MCS analyses are based on the strong thresholds applied to GPM
432 detection as described in Section 2.1. A sensitivity test using moderate thresholds is
433 performed with the corresponding global MCS distributions shown in Fig. 6. As revealed in
434 Houze et al. (2015), strong thresholds better represent the behavior of convection over
435 land, while the weak oceanic convective features can be exhibited more easily using the
436 moderate thresholds. By comparing Fig. 6 with Fig. 5, the locations of high MCS
437 occurrence are seen to be well maintained, but the frequency is greatly increased,
438 potentially including many non-MCS objects. The DWC and WCC objects detected using
439 moderate thresholds are also compared to the FLEXTRKR data set over the CONUS.
440 Although the number of GPM-detected DWC and WCC objects increases from 388 to 620,
441 the MCS HIT rate drops to 49%.
442 The global MCS distribution shown in Fig. 5 extends the MCS detection to higher
443 latitudes (beyond 35S-35N covered by TRMM), where notable MCS occurrence
444 frequency is found above 60N in Siberia, Northern Europe, and Canada. For the high-
445 latitude events, Houze et al. (2019) further found that these high-latitude MCSs occur
446 where global warming has been most intense.
447
448
21 Page 73 of 100 For Peer Review
449 4. Conclusions and Discussions
450 In this study, the spaceborne GPM Ku-band radar data sets are quantitatively
451 compared to the ground-based NEXRAD radar data sets during a 3-year period (2014-
452 2016) over the CONUS. Based on the morphology of GPM-detected radar echoes, two
453 types of GPM-detected extreme convective echo objects, DWC and WCC, are compared
454 with an MCS database constructed using feature tracking on synthesized geostationary
455 satellite and NEXRAD radar observations. The major findings of this study are
456 summarized as follows.
457 (1) The GPM radar captures consistent 3D distribution of radar reflectivity with
458 NEXRAD across a wide range precipitating cloud systems, including seasonal variations
459 from spring to fall, when the GPM radar and NEXRAD data are interpolated to the same
460 grid and appropriately smoothed. The comparison demonstrates the two independent
461 radar systems observe the 3D radar reflectivity fields in a consistent manner.
462 (2) The GPM classified DWC and WCC objects are compared to the NEXRAD-tracked
463 MCS database. More than 70% of these GPM-defined extreme convective echo objects
464 are collocated with tracked MCSs in the NEXRAD data set, indicating the GPM-detected
465 DWC and WCC objects are highly correlated with MCSs. The GPM’s capability in MCS
466 detection demonstrates strong seasonality, and a better performance is found in spring
467 and summer than fall. This seasonal variation follows the change of large-scale
468 environments that alter MCS characteristics. For the GPM-detected objects not
22 For Peer Review Page 74 of 100
469 corresponding to NEXRAD-tracked MCSs, the majority of them have shorter lifespan than
470 6 hours. These false alarms highlight the unavoidable uncertainty associated with the
471 arbitrariness of criteria used to identify MCSs in the two datasets.
472 (3) After validating the GPM’s performance in MCS detection over the CONUS, the
473 DWC and WCC classification algorithm is applied to global GPM observations to obtain a
474 global view of MCS distribution. The well-known MCS hot zones, namely the US GP and
475 the offshore of the east coast in North America, subtropical South America, central
476 equatorial Africa, monsoon regions in Asia, and the maritime continent, are further
477 confirmed in this study. Moreover, MCSs in high-latitude regions (above 60N in Siberia,
478 Northern Europe, and Canada) are revealed, and the relative numbers of MCSs over
479 different parts of Earth are now quantitatively measured.
480
481
482
483
484
485
486
487
488
23 Page 75 of 100 For Peer Review
489 Acknowledgements
490 We acknowledge the support of the Climate Model Development and Validation (CMDV)
491 project at PNNL. The effort of J. Wang, J. Fan, Z. Feng, and J. Hardin was supported by
492 CMDV. Robert A. Houze was supported by NASA Award NNX16AD75G and by master
493 agreement 243766 between the University of Washington and PNNL. Stacy R. Brodzik
494 was supported by NASA Award NNX16AD75G and subcontracts from the CMDV and
495 Water Cycle and Climate Extreme Modeling (WACCEM) projects of PNNL. PNNL is
496 operated for the US Department of Energy (DOE) by Battelle Memorial Institute under
497 Contract DE-AC05-76RL01830. This research used resources of the National Energy
498 Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of
499 Science User Facility operated under contract DE‐AC02‐05CH11231. The GPM reflectivity
500 data are download at University of Washington GPM-Ku Data Set at
501 (http://gpm.atmos.washington.edu/), the Global Merged IR dataset is obtained at NASA
502 Goddard Earth Sciences Data and Information Services Center
503 (https://dx.doi.org/10.5067/P4HZB9N27EKU), the GridRad radar dataset is obtained at the
504 Research Data Archive of the National Center for Atmospheric Research (NCAR)
505 (https://rda.ucar.edu/datasets/ds841.0/).
506
507
508
24 For Peer Review Page 76 of 100
509 References
510
511 Baldwin, M. E., and K. E. Mitchell, 1997: The NCEP hourly multisensor U.S. precipitation
512 analysis for operations and GCIP research. Preprints, 13th Conf. on Hydrology, Long
513 Beach, CA, Amer. Meteor. Soc., 54–55.
514 Barnes, H. C., and R. A. Houze, Jr., 2016: Comparison of observed and simulated spatial
515 patterns of ice microphysical processes in tropical oceanic mesoscale convective
516 systems. J. Geophys. Res. Atmos., 121, 8269-8296, doi:10.1002/2016JD025074.
517 Bowman, K. P., and C. R. Homeyer, 2017: GridRad - Three-Dimensional Gridded
518 NEXRAD WSR-88D Radar Data, V3.1. National Center for Atmospheric Research,
519 Computational and Information Systems Laboratory, Boulder, CO, accessed 1
520 November 2017, https://doi.org/10.5065/D6NK3CR7.
521 Box, J. F. Guinness, Gosset, Fisher, and small samples. Stat Sci 1987; 2: 45–52,
522 https://doi.org/10.1214/ss/1177013437.
523 Cetrone, J. and R.A. Houze, 2011: Leading and Trailing Anvil Clouds of West African
524 Squall Lines. J. Atmos. Sci., 68, 1114–1123, https://doi.org/10.1175/2011JAS3580.1
525 Clark, A.J., R.G. Bullock, T.L. Jensen, M. Xue, and F. Kong, 2014: Application of Object-
526 Based Time-Domain Diagnostics for Tracking Precipitation Systems in Convection-
527 Allowing Models. Wea. Forecasting, 29, 517–542, https://doi.org/10.1175/WAF-D-13-
528 00098.1.
25 Page 77 of 100 For Peer Review
529 Cui, W., X. Dong, B. Xi, and A. Kennedy, 2017: Evaluation of Reanalyzed Precipitation
530 Variability and Trends Using the Gridded Gauge-Based Analysis over the CONUS. J.
531 Hydrometeor., 18, 2227–2248, https://doi.org/10.1175/JHM-D-17-0029.1.
532 Cui, W., X. Dong, B. Xi, and R. Stenz, 2016: Comparison of the GPCP 1DD Precipitation
533 Product and NEXRAD Q2 Precipitation Estimates over the Continental United States. J.
534 Hydrometeor., 17, 1837–1853, https://doi.org/10.1175/JHM-D-15-0235.1.
535 Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: configuration and
536 performance of the data assimilation system. Quarterly Journal of the Royal
537 Meteorological Society, 137, 553-597.
538 Doviak, R. J., and D. S. Zrnic, 1993: Doppler Radar and Weather Observations. 2d ed.
539 Academic Press, 562 pp.
540 Fan, J, Y‐C Liu, K‐M Xu, K. North, S. Collis, X. Dong, G. J. Zhang, Q. Chen, P. Kollias,
541 and S. J. Ghan, 2015: Improving representation of convective transport for scale‐aware
542 parameterization: 1. Convection and cloud properties simulated with spectral bin and
543 bulk microphysics. J. Geophys. Res. Atmos., 120, 3485–3509.
544 https://doi.org/10.1002/2014JD022142.
545 Fan, J., et al. 2017: Cloud‐resolving model intercomparison of an MC3E squall line case:
546 Part I—Convective updrafts, J. Geophys. Res. Atmos., 122, 9351–9378,
547 https://doi.org/10.1002/2017JD026622.
548 Feng, Z., X. Dong, and B. Xi, 2009: A Method to Merge WSR-88D Data with ARM SGP
26 For Peer Review Page 78 of 100
549 Millimeter Cloud Radar Data by Studying Deep Convective Systems. J. Atmos. Oceanic
550 Technol., 26, 958–971, https://doi.org/10.1175/2008JTECHA1190.1
551 Feng, Z., L. R. Leung, R. A. Houze, S. Hagos, J. Hardin, Q. Yang, B. Han, and J. Fan,
552 2018: Structure and evolution of mesoscale convective systems: Sensitivity to cloud
553 microphysics in convection-permitting simulations over the United States. J. Adv.
554 Model. Earth Syst., 10, 1470–1494.
555 Feng, Z., R. A. Houze, L. R. Leung, F. Song, J. Hardin, J. Wang, W. Gustafson, and C.
556 Homeyer, 2019: Spatio-temporal Characteristics and Large-scale Environments of
557 Mesoscale Convective Systems East of the Rocky Mountains. J. Clim., In revision.
558 Feng, Z., Leung, L. R., Hagos, S., Houze, R. A., Burleyson, C. D., & Balaguru, K. 2016:
559 More frequent intense and long‐lived storms dominate the springtime trend in central
560 US rainfall. Nature Communications, 7, 13429. https://doi.org/10.1038/ncomms13429.
561 Feng, Z., Leung, L. R., Houze, R. A., Jr., Hagos, S., Hardin, J., Yang, Q., et al. (2018).
562 Structure and evolution of mesoscale convective systems: Sensitivity to cloud
563 microphysics in convection‐permitting simulations over the United States. J. Adv.
564 Model. Earth Syst, 10, 1470–1494. https://doi.org/10.1029/2018MS001305.
565 Feng, Z., X. Dong, B. Xi, S. A. McFarlane, A. Kennedy, B. Lin, and P. Minnis, 2012: Life
566 cycle of midlatitude deep convective systems in a Lagrangian framework, J. Geophys.
567 Res., 117, D23201, doi: 10.1029/2012JD018362.
568 Fritsch, J. M., and G. S. Forbes, 2001: Mesoscale convective systems. Meteorological
27 Page 79 of 100 For Peer Review
569 Monographs, 28(50), 323–358.
570 Futyan, J. M., and A. D. Del Genio (2007), Deep convective system evolution over Africa
571 and the tropical Atlantic, J. Clim., 20, 5041–5060, doi:10.1175/JCLI4297.1.
572 Geerts, B. (1998), Mesoscale convective systems in the southeast United States during
573 1994-95: A survey. Wea. Forecasting, 13, 860-869. Doi 10.1175/1520-
574 0434(1998)013<0860:Mcsits>2.0.Co;2
575 Grecu, M., L. Tian, G.M. Heymsfield, A. Tokay, W.S. Olson, A.J. Heymsfield, and A.
576 Bansemer, 2018: Nonparametric Methodology to Estimate Precipitating Ice from
577 Multiple-Frequency Radar Reflectivity Observations. J. Appl. Meteor. Climatol., 57,
578 2605–2622, https://doi.org/10.1175/JAMC-D-18-0036.1.
579 Haberlie, A. M., and W. S. Ashley, (2019), A Radar-Based Climatology of Mesoscale
580 Convective Systems in the United States. J. Clim., 32, 1591-1606. 10.1175/jcli-d-18-
581 0559.1.
582 Hamada, A. and Y. N. Takayabu, 2016: Improvements in Detection of Light Precipitation
583 with the Global Precipitation Measurement Dual-Frequency Precipitation Radar (GPM
584 DPR). J. Atmos. Oceanic Technol., 33, 653–667, https://doi.org/10.1175/JTECH-D-15-
585 0097.1.
586 Han, B., Fan, J., Varble, A., Morrison, H., Williams, C. R., Chen, B., et al, 2019:
587 Cloud‐Resolving Model Intercomparison of an MC3E Squall Line Case: Part II –
588 Stratiform Precipitation Properties. J. Geophys. Res. Atmos.,124.
28 For Peer Review Page 80 of 100
589 https://doi.org/10.1029/2018JD029596
590 Héas, P., E. Mémin Three-dimensional motion estimation of atmospheric layers from
591 image sequences IEEE Trans. Geosci. Remote Sens., 46 (8) (2008), pp. 2385-2396.
592 Hence, D. A., and R. A. Houze Jr., 2008: Kinematic structure of convective-scale elements
593 in the rainbands of Hurricanes Katrina and Rita (2005). J. Geophys. Res., 113, D15108,
594 doi:10.1029/2007JD009429.
595 Hence, D. A., and R. A. Houze Jr., 2011: Vertical structure of hurricane eyewalls as seen
596 by the TRMM Precipitation Radar. J. Atmos. Sci., 68, 1637–1652,
597 https://doi.org/10.1175/2011JAS3578.1.
598 Hence, D. A., and R. A. Houze Jr., 2012: Vertical structure of tropical cyclones with
599 concentric eyewalls as seen by the TRMM Precipitation Radar. J. Atmos. Sci., 69,
600 1021–1036, https://doi.org/10.1175/JAS-D-11-0119.1.
601 Hence, D.A. and R.A. Houze, 2012: Vertical Structure of Tropical Cyclone Rainbands as
602 Seen by the TRMM Precipitation Radar. J. Atmos. Sci., 69, 2644–2661,
603 https://doi.org/10.1175/JAS-D-11-0323.1, https://doi.org/10.1175/JAS-D-11-0323.1.
604 Homeyer, R. C., K. P. Bowman, 2017: Algorithm Description Document for Version 3.1 of
605 the Three-Dimensional Gridded NEXRAD WSR-88D Radar (GridRad) Dataset, 12 pp.
606 [Available online at http://gridrad.org/pdf/GridRad-v3.1-Algorithm-Description.pdf]
607 Hou, A. Y., and Coauthors, 2014: The Global Precipitation Measurement Mission. Bull.
608 Amer. Meteor. Soc., 95, 701–722, doi:https://doi.org/10.1175/BAMS-D-13-00164.1.
29 Page 81 of 100 For Peer Review
609 Houze, R. A., Jr., D. C. Wilton, and B. F. Smull, 2007: Monsoon convection in the
610 Himalayan region as seen by the TRMM Precipitation Radar. Quart. J. Roy. Meteor.
611 Soc., 133, 1389–1411.
612 Houze, R. A., Jr., 2014: Cloud dynamics. 2d ed. Oxford: Elsevier/Academic Press, 329 pp.
613 Houze, R. A., Jr., 2018: 100 years of research on mesoscale convective systems.
614 Meteorological Monographs, 59, 17.1–17.54,
615 https://doi.org/10.1175/AMSMONOGRAPHS-D-18-0001.1.
616 Houze, R. A., Jr., 2004: Mesoscale convective systems. Rev. Geophys., 42, RG4003,
617 doi:https://doi.org/10.1029/2004RG000150.
618 Houze, R. A., Jr., 2012: Orographic effects on precipitating clouds. Rev. Geophys., 50,
619 RG1001, doi:10.1029/2011RG000365.
620 Houze, R. A., Jr., S. G. Geotis, F. D. Marks, Jr., and A. K. West, 1981: Winter monsoon
621 convection in the vicinity of north Borneo. Part I: Structure and time variation of the
622 clouds and precipitation. Mon. Wea. Rev., 109, 1595-1614.
623 Houze, R. A., Jr., K. L. Rasmussen, M. D. Zuluaga, and S. R. Brodzik, 2015: The variable
624 nature of convection in the tropics and subtropics: A legacy of 16 years of the Tropical
625 Rainfall Measuring Mission (TRMM) satellite. Rev. Geophys., 53, 994–1021,
626 doi:https://doi.org/10.1002/2015RG000488.
627 Houze, R. A., Jr., L. A. McMurdie, W. A. Petersen, M. R. Schwaller, W. Baccus, J. D.
628 Lundquist, C. F. Mass, B. Nijssen, S. A. Rutledge, D. R. Hudak, S. Tanelli, G. G. Mace,
30 For Peer Review Page 82 of 100
629 M. R. Poellot, D. P. Lettenmaier, J. P. Zagrodnik, A. K. Rowe, J. C. DeHart, L. E.
630 Madaus, H. C. Barnes, and V. Chandrasekar, 2017: The Olympic Mountains
631 Experiment (OLYMPEX). Bull. Amer. Meteor. Soc., 98, 2167–2188,
632 https://doi.org/10.1175/BAMS-D-16-0182.1.
633 Houze, R. A., J. Wang, J. Fan, S. Brodzik, and Z. Feng, 2019: Extreme convective storms
634 over high‐latitude continental areas where maximum warming is occurring. Geophys.
635 Res. Lett., 46, 4059– 4065. https://doi.org/10.1029/2019GL082414.
636 Iguchi, T., S. Seto, R. Meneghini, N. Yoshida, J. Awaka, M. Le, V. Chandrasekar, and T.
637 Kubota, 2017: GPM/DPR level-2. Algorithm Theoretical Basis Doc., 68 pp. [Available
638 online at
639 https://pmm.nasa.gov/sites/default/files/document_files/ATBD_DPR_201708_whole_1.p
640 df.]
641 Janowiak, J., B. Joyce, and P. Xie, 2017: NCEP/CPC L3 Half Hourly 4km Global (60S -
642 60N) Merged IR V1, edited by Andrey Savtchenko, Greenbelt, MD, Goddard Earth
643 Sciences Data and Information Services Center (GES DISC), Accessed: [1 February
644 2018], 10.5067/P4HZB9N27EKU.
645 Jensen, M.P., W.A. Petersen, A. Bansemer, N. Bharadwaj, L.D. Carey, D.J. Cecil, S.M.
646 Collis, A.D. Del Genio, B. Dolan, J. Gerlach, S.E. Giangrande, A. Heymsfield, G.
647 Heymsfield, P. Kollias, T.J. Lang, S.W. Nesbitt, A. Neumann, M. Poellot, S.A. Rutledge,
648 M. Schwaller, A. Tokay, C.R. Williams, D.B. Wolff, S. Xie, and E.J. Zipser, 2016: The
31 Page 83 of 100 For Peer Review
649 Midlatitude Continental Convective Clouds Experiment (MC3E). Bull. Amer. Meteor.
650 Soc., 97, 1667–1686, https://doi.org/10.1175/BAMS-D-14-00228.1.
651 Johnson, R. H., and R. A. Houze, Jr., 1987: Precipitating cloud systems of the Asian
652 monsoon. In Monsoon Meteorology (C.-P. Chang and T. N. Krishnamurti, Eds.), 298-
653 353.
654 Keighton, S., J. Jackson, J. Guyer, and J. Peters, 2007: A preliminary analysis of severe
655 quasi-linear mesoscale convective systems crossing the Appalachians. Preprints, 22nd
656 Conf. on Weather Analysis and Forecasting/18th Conf. on Numerical Weather
657 Prediction, Park City, UT, Amer. Meteor. Soc., P2.18. [Available online at
658 http://ams.confex.com/ams/pdfpapers/123614.pdf].
659 Krajewski, W. F., A. Ntelekos, and R. Goska, 2006: A GIS based methodology for the
660 assessment of weather radar beam blockage in mountainous regions: Two examples
661 from the U.S. NEXRAD network. Comput. Geosci., 32, 283–302.
662 Kummerow, C., W. Barnes, T. Kozu, J. Shiue, and J. Simpson 1998: The Tropical Rainfall
663 Measuring Mission (TRMM) sensor package, J. Atmos.Oceanic Technol.,15, 809–817.
664 Leary, C. A., and R. A. Houze, Jr., 1979: Melting and evaporation of hydrometeors in
665 precipitation from the anvil clouds of deep tropical convection. J. Atmos. Sci., 36, 669-
666 679.
667 Letkewicz, C.E. and M.D. Parker, 2010: Forecasting the Maintenance of Mesoscale
668 Convective Systems Crossing the Appalachian Mountains. Wea. Forecasting, 25,
32 For Peer Review Page 84 of 100
669 1179–1195, https://doi.org/10.1175/2010WAF2222379.1.
670 Lin, Y. 2011. GCIP/EOP Surface: Precipitation NCEP/EMC 4KM Gridded Data (GRIB)
671 Stage IV Data. Version 1.0. UCAR/NCAR - Earth Observing Laboratory.
672 https://doi.org/10.5065/D6PG1QDD. Accessed 15 Nov 2018.
673 Lin, Y., and K. E. Mitchell, 2005: The NCEP stage II/IV hourly precipitation analyses:
674 Development and applications. Preprints, 19th Conf. on Hydrology, San Diego, CA,
675 Amer. Meteor. Soc., CD-ROM, 1.2.
676 Liu, Y‐C, J. Fan, G. J. Zhang, K‐M Xu, and S. J. Ghan, 2015: Improving representation of
677 convective transport for scale‐aware parameterization: 2. Analysis of cloud‐resolving
678 model simulations. J. Geophys. Res. Atmos., 120, 3510–3532.
679 https://doi.org/10.1002/2014JD022145.
680 Machado, L.A., W.B. Rossow, R.L. Guedes, and A.W. Walker, 1998: Life Cycle Variations
681 of Mesoscale Convective Systems over the Americas. Mon. Wea. Rev., 126, 1630–
682 1654, https://doi.org/10.1175/1520-0493(1998)126<1630:LCVOMC>2.0.CO;2.
683 Madden, R., and P. Julian, 1972: Description of global scale circulation cells in the tropics
684 with a 40-50 day period. J. Atmos. Sci., 29, 1109–1123.
685 Madden, R. A., and P. R. Julian, 1994: Observations of the 40-50-day tropical oscillation—
686 A review. Mon. Wea. Rev., 122, 814-837.
687 Maddox, R. A., Chappell, C. F., & Hoxit, L. R., 1979: Synoptic and mesoscale aspects of
688 flash flood events. Bull. Amer. Meteor. Soc., 60, 115–123.
33 Page 85 of 100 For Peer Review
689 McMurdie, L. A., A. K. Rowe, R. A. Houze, Jr., S. R. Brodzik, J. P. Zagrodnik, and T. M.
690 Schuldt, 2018: Terrain-enhanced precipitation processes above the melting layer:
691 Results from OLYMPEX. J. Geophys. Res.
692 Atmos., 123, https://doi.org/10.1029/2018JD029161.
693 Morel, C. and Senesi, S., 2002: A climatology of mesoscale convective systems over
694 Europe using satellite infrared imagery. II: Characteristics of European mesoscale
695 convective systems. Q. J. R. Meteorol. Soc., 128: 1973-1995.
696 https://doi.org/10.1256/003590002320603494.
697 Olson, W. S., H. Masunaga, and the GPM Combined Radar-Radiometer Algorithm Team,
698 2016: GPM Combined Radar-Radiometer Precipitation Algorithm Theoretical Basis
699 Document (Version 4), 7 pp. [Available online at
700 https://pmm.nasa.gov/sites/default/files/document_files/Combined_algorithm_ATBD.V0
701 4.rev_.pdf.]
702 Peters, J.M. and R.S. Schumacher, 2014: Objective Categorization of Heavy-Rain-
703 Producing MCS Synoptic Types by Rotated Principal Component Analysis. Mon. Wea.
704 Rev., 142, 1716–1737, https://doi.org/10.1175/MWR-D-13-00295.1.
705 Ramage, C. S., 1968: Role of a tropical “maritime continent” in the atmospheric circulation.
706 Mon. Wea. Rev., 96, 365–370.
707 Rasmussen, K. L., and R. A. Houze, Jr., 2011: Orogenic convection in South America as
708 seen by the TRMM satellite. Mon. Wea. Rev., 139, 2399-2420.
34 For Peer Review Page 86 of 100
709 Rasmussen, K. L., M. D. Zuluaga, and R. A. Houze Jr., 2014: Severe convection and
710 lightning in subtropical South America. Geophys. Res. Lett., 41, 7359–7366,
711 doi:10.1002/2014GL061767.
712 Rauber, R. M., and S. L. Nesbitt, 2018: Radar Meteorology: A first course. Wiley Blackwell,
713 467pp. ISBN-13: 978-1118432624
714 Reinhard, E., 2006: High dynamic range imaging: Acquisition, Display, and Image-Based
715 Lighting. Morgan Kaufmann, pp. 233–234.
716 Rinehart, R. E., 2001: Radar for Meteorologists. 3d ed. Rinehart, 428 pp.
717 Roca, R., Aublanc, J., Chambon, P., Fiolleau, T., & Viltard, N., 2014: Robust Observational
718 Quantification of the Contribution of Mesoscale Convective Systems to Rainfall in the
719 Tropics. J. Clim., 27(13), 4952-4958. http://dx.doi.org/10.1175/JCLI-D-13-00628.1
720 Romatschke, U. and R.A. Houze, 2010: Extreme Summer Convection in South America. J.
721 Clim., 23, 3761–3791, https://doi.org/10.1175/2010JCLI3465.1.
722 Rowe, A. K., and R. A. Houze, Jr., 2014: Microphysical characteristics of MJO convection
723 over the Indian Ocean during DYNAMO. J. Geophys. Res. Atmos., 119, 2543-2554,
724 doi:10.1002/2013JD020799.
725 Savtchenko, A., 2018: README Document for the NOAA NCEP/CPC Half Hourly 4km
726 Global (60S - 60N) Merged IR, 11 pp. [Available online at
727 https://disc2.gesdisc.eosdis.nasa.gov/data/MERGED_IR/GPM_MERGIR.1/doc/READM
728 E.MERGIR.pdf]
35 Page 87 of 100 For Peer Review
729 Schmetz, J., K. Holmlund, J. Hoffman, B. Strauss, B. Mason, V. Gaertner, A. Koch, L. Van
730 De Berg Operational cloud-motion winds from meteosat infrared images J. Appl.
731 Meteorol., 32 (7) (1993), pp. 1206-1225.
732 Schumacher, C., R. A. Houze Jr., and I. Kraucunas (2004), The tropical dynamical
733 response to latent heating estimates derived from the TRMM Precipitation Radar, J.
734 Atmos. Sci., 61, 1341–1358.
735 Schwaller, M.R. and K.R. Morris, 2011: A Ground Validation Network for the Global
736 Precipitation Measurement Mission. J. Atmos. Oceanic Technol., 28, 301–319,
737 https://doi.org/10.1175/2010JTECHA1403.1.
738 Skofronick-Jackson, G., D. Hudak, W. Petersen, S.W. Nesbitt, V. Chandrasekar, S.
739 Durden, K.J. Gleicher, G. Huang, P. Joe, P. Kollias, K.A. Reed, M.R. Schwaller, R.
740 Stewart, S. Tanelli, A. Tokay, J. R. Wang, and M. Wolde, 2015: Global Precipitation
741 Measurement Cold Season Precipitation Experiment (GCPEX): For Measurement’s
742 Sake, Let It Snow. Bull. Amer. Meteor. Soc., 96, 1719–1741,
743 https://doi.org/10.1175/BAMS-D-13-00262.1.
744 Starzec, M., C. R. Homeyer, and G. L. Mullendore, 2017: Storm Labeling in Three
745 Dimensions (SL3D): A Volumetric Radar Echo and Dual-Polarization Updraft
746 Classification Algorithm. Mon. Wea. Rev., 145, 1127-1145. 10.1175/mwr-d-16-0089.1
747 Song, F., Z. Feng, L. R. Leung, R. A. Houze, J. Wang, J. Hardin, and C. R. Homeyer,
748 2019: Contrasting spring and summer large-scale environments associated with
36 For Peer Review Page 88 of 100
749 mesoscale convective systems over the U.S. Great Plains. J. Clim., In revision.
750 TerrainBase, National Geophysical Data Center/NESDIS/NOAA/U.S. Department of
751 Commerce. 1995: Global 5 Arc-minute Ocean Depth and Land Elevation from the US
752 National Geophysical Data Center (NGDC). Research Data Archive at the National
753 Center for Atmospheric Research, Computational and Information Systems Laboratory.
754 https://doi.org/10.5065/E08M-4482. Accessed 28 Mar. 2019.
755 Toyoshima, K., H. Masunaga, and F. A. Furuzawa, 2015: Early evaluation of Ku- and Ka-
756 band sensitivities for the Global Precipitation Measurement (GPM) Dual-Frequency
757 Precipitation Radar (DPR). SOLA, 11, 14–17, doi:https://doi.org/10.2151/sola.2015-004.
758 Wang, J, X. Dong, and B. Xi, 2015: Investigation of ice cloud microphysical properties of
759 DCSs using aircraft in situ measurements during MC3E over the ARM SGP site. J.
760 Geophys. Res. Atmos., 120, 3533–3552. https://doi.org/10.1002/2014JD022795.
761 Wang, J., X. Dong, B. Xi, and A. J. Heymsfield, 2016: Investigation of liquid cloud
762 microphysical properties of deep convective systems: 1. Parameterization of raindrop
763 size distribution and its application for stratiform rain estimation, J. Geophys. Res.
764 Atmos., 121, 10,739–10,760, https://doi.org/10.1002/2016JD024941.
765 Wang, J., Dong, X., and Xi, B., 2018: Investigation of liquid cloud microphysical properties
766 of deep convective systems: 2. Parameterization of raindrop size distribution and its
767 application for convective rain estimation. J. Geophys. Res. Atmos, 123, 11,637–
768 11,651. https://doi.org/10.1029/2018JD028727.
37 Page 89 of 100 For Peer Review
769 Wang, J., X. Dong, A. Kennedy, B. Hagenhoff, and B. Xi, 2019: A Regime Based
770 Evaluation of Southern and Northern Great Plains Warm-Season Precipitation Events in
771 WRF. Wea. Forecasting, 0, https://doi.org/10.1175/WAF-D-19-0025.1
772 Xie, S., Y. Zhang, S. E. Giangrande, M. P. Jensen, R. McCoy, and M. Zhang, 2014:
773 Interactions between cumulus convection and its environment as revealed by the MC3E
774 sounding array, J. Geophys. Res. Atmos., 119, 11,784–11,808,
775 https://doi.org/10.1002/2014JD022011.
776 Yang, Q., R. A. Houze Jr., L. R. Leung, and Z. Feng, 2017: Environments of long-lived
777 mesoscale convective systems over the central United States in convection permitting
778 climate simulations. J. Geophys. Res. Atmos., 122, 13 288–13 307,
779 https://doi.org/10.1002/2017JD027033.
780 Zagrodnik, J. P., L. A. McMurdie, and R. A. Houze, Jr., 2018: Stratiform precipitation
781 processes in cyclones passing over a coastal mountain range. J. Atmos. Sci., 75, 983–
782 1004, doi:10.1175/JAS-D-17-0168.1.
783 Zhang, J., K. Howard, C. Langston, B. Kaney, Y. Qi, L Tang, H. Grams, Y. Wang, S.
784 Cocks, S. Martinaitis, A. Arthur, K. Cooper, J. Brogden, and D. Kitzmiller, 2015: Multi-
785 Radar Multi-Sensor (MRMS) Quantitative Precipitation Estimation: Initial Operating
786 Capabilities, Bull. Amer. Meteor. Soc., 97, 621–638, https://doi.org/10.1175/BAMS-D-
787 14-00174.1.
788 Zhang, J., K. Howard, C. Langston, S. Vasiloff, B. Kaney, A. Arthur, S. V. Cooten, K.
38 For Peer Review Page 90 of 100
789 Kelleher, D. Kitzmiller, F. Ding, D-J Seo, E. Wells, and C. Dempsey, 2011: National
790 Mosaic and Multi-Sensor QPE (NMQ) System: Description, Results, and Future Plans,
791 Bull. Amer. Meteor. Soc., 92, 1321–1338. doi: http://dx.doi.org/10.1175/2011BAMS-D-
792 11-00047.1.
793 Zipser, E. J., D. J. Cecil, C. Liu, S. W. Nesbitt, and D. P. Yorty, 2006: Where are the Most
794 Intense Thunderstorms on Earth? Bull. Amer. Meteor. Soc., 87, 1057–1072,
795 https://doi.org/10.1175/BAMS-87-8-1057
796 Zuluaga, M.D. and R.A. Houze, 2013: Evolution of the Population of Precipitating
797 Convective Systems over the Equatorial Indian Ocean in Active Phases of the Madden–
798 Julian Oscillation. J. Atmos. Sci., 70, 2713–2725, https://doi.org/10.1175/JAS-D-12-
799 0311.1.
800
801
802
803
804
805
806
807
808
39 Page 91 of 100 For Peer Review
809 List of Figures
810
811 Fig. 1 The domain for GPM and NEXRAD comparison is marked with the red box (110W
812 – 85W, 31N – 48N) overlaid on the CONUS elevation map.
813 Fig. 2 The contoured-frequency-by-altitude-diagrams (CFADs) normalized by the total
814 number of samples at all altitude levels for GPM (a, d, g, j, m, p) and NEXRAD (b, e, h,
815 k, n, q) for the months from April to September in 2014-2016 period. The box-whisker
816 plots (c, f, i, l, o, r) for GPM (red) and NEXRAD (blue) are calculated using
817 normalization at each individual level, where the center of the box represents the 50%
818 percentile value, the lower quartile (25%) and upper quartile (75%) form the left and
819 right boundary of the box, and whiskers correspond to the 5% and 95% values.
820 Fig. 3 (a) The Deep Wide Convection (DWC) object defined by GPM on 9 June 2014 at
821 04:26:46 UTC and (b) the coincident NEXRAD observed MCS tracked by FLEXTRKR
822 at 04:00 UTC, as well as (c) the intersection of GPM overpass time (dash-line) with
823 respect to the evolution of the MCS precipitation area during its entire life cycle
824 identified by the NEXRAD.
825 Fig. 4 Monthly total number of MCSs HIT (GPM detection validated by NEXRAD, red
826 bars) and false alarm (FAR, not identified as MCS by NEXRAD, blue bars), overlaid by
827 the monthly total number of MCSs detection by NEXRAD (black line) and their average
828 precipitation area coverage (green line).
40 For Peer Review Page 92 of 100
829 Fig. 5 Geographical distribution of the probability of MCS occurrence frequency detected
830 by GPM during the months of (a) JJA and (b) DJF. The gray shaded areas inside the
831 continental regions represent the 700 m elevation. The probability is on a scale of 0 to
832 100% and is computed as the number of pixels identified as either DWC or WCC
833 divided by the total number of GPM overpasses within a 0.25 × 0.25 gridbox.
834 Fig. 6 Similar to Figure 5 but using the moderate criteria.
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
41 Page 93 of 100 For Peer Review
851
852
853 Fig. 1 The domain for GPM and NEXRAD comparison is marked with the red box (110W
854 – 85W, 31N – 48N) overlaid on the CONUS elevation map.
855
856
857
42 For Peer Review Page 94 of 100
858
859 Fig. 2 The contoured-frequency-by-altitude-diagrams (CFADs) normalized by the total
860 number of samples at all altitude levels for GPM (a, d, g, j, m, p) and NEXRAD (b, e, h,
861 k, n, q) for the months from April to September in 2014-2016 period. The box-whisker
862 plots (c, f, i, l, o, r) for GPM (red) and NEXRAD (blue) are calculated using
863 normalization at each individual level, where the center of the box represents the 50%
864 percentile value, the lower quartile (25%) and upper quartile (75%) form the left and
865 right boundary of the box, and whiskers correspond to the 5% and 95% values.
866
867
868
43 Page 95 of 100 For Peer Review
869
870 Fig. 3 (a) The Deep Wide Convection (DWC) object defined by GPM on 9 June 2014 at
871 04:26:46 UTC and (b) the coincident NEXRAD observed MCS tracked by FLEXTRKR
872 at 04:00 UTC, as well as (c) the intersection of GPM overpass time (dash-line) with
873 respect to the evolution of the MCS precipitation area during its entire life cycle
874 identified by the NEXRAD.
875
876
877
878
879
44 For Peer Review Page 96 of 100
880 881 Fig. 4 Monthly total number of MCSs HIT (GPM detection validated by NEXRAD, red
882 bars) and false alarm (FAR, not identified as MCS by NEXRAD, blue bars), overlaid by
883 the monthly total number of MCSs detection by NEXRAD (black line) and their average
884 precipitation area coverage (green line).
885
886
887
45 Page 97 of 100 For Peer Review
888
889 Fig. 5 Geographical distribution of the probability of MCS occurrence frequency detected
890 by GPM during the months of (a) JJA and (b) DJF. The gray shaded areas inside the
891 continental regions represent the 700 m elevation. The probability is on a scale of 0 to
892 100% and is computed as the number of pixels identified as either DWC or WCC
893 divided by the total number of GPM overpasses within a 0.25 × 0.25 gridbox.
894
895
896
897
46 For Peer Review Page 98 of 100
898
899 Fig. 6 Similar to Figure 5 but using the moderate criteria.
900
901
902
903
904
905
906
907
908
909
910
47 Page 99 of 100 For Peer Review
911 List of Tables
912 Table 1 The statistics of GPM’s MCS detection skill in each month.
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
48 For Peer Review Page 100 of 100
937 Table 1 The statistics of GPM’s MCS detection skill in each month. Average Average GPM False GPM GPM GPM HIT NEXRAD MCS MCS Negative HIT (#) FAR (#) Rate MCS (#) Coverage Duration Rate (km2) (hour) April 43 8 84% 16% 104 20,374 21.5 May 47 9 84% 16% 140 13,742 20.8 June 58 26 69% 31% 160 11,961 20.4 July 61 23 73% 27% 137 10,683 20.9 August 37 31 54% 46% 122 10,885 20.6 September 27 18 60% 40% 77 10,141 19.8
938
939
940
49