Acp 2020 18
Acp 2020 18
Acp 2020 18
5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
1 Assessment of vertical air motion among reanalyses and qualitative comparison with direct
3 Kizhathur Narasimhan Uma1, Siddarth Shankar Das1, Madineni Venkat Ratnam2, and Kuniyil
4 Viswanathan Suneeth1
5
1
6 Space Physics Laboratory, Vikram Sarabhai Space Centre, ISRO, Trivandrum-695022, India
2
7 National Atmospheric Research Laboratory, Dept. of Space, Gadanki-517112, India
8
9 *e-mail : urmi_nmrf@yahoo.co.in
10 Abstract
11 Vertical wind (w) is one of the most important meteorological parameters for
12 understanding different atmospheric phenomena. Only very few direct measurements of w are
13 available and most of the time one must depend on reanalysis products. In the present study,
15 JRA-55) and qualitative comparison of those datasets with direct VHF radar measurements over
16 the convectively active regions Gadanki (13.5oN and 79.2oE) and Kototabang (0oS and 100.2oE)
17 are presented for the first time. The magnitude of w derived from reanalyses is 10-50% less than
18 that from the direct radar observations. Radar measurements of w show downdrafts below 8 to 10
19 km and updrafts above 8-10 km over both locations. Inter-comparison between the reanalyses
20 shows that ERAi is overestimating NCEP-2 and underestimating all the reanalyses. Directional
21 tendency shows that the percentage of updrafts captured is reasonably good, but downdrafts are
22 not well captured by all reanalyses. Thus, caution is advised when using vertical velocities from
23 reanalyses.
24 Key Words: Vertical velocity, MST Radar, Equatorial Atmosphere Radar, Reanalysis
25
1
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
26 1 Introduction
27 Vertical air motion (w) in any region of the Earth’s atmosphere reflects the structure and
28 dynamical features of that region. Importantly, in the lower part of the atmosphere, sudden
29 widespread changes in weather are usually associated with variations in the vertical air motion.
30 The magnitude of w is a factor of ten or more smaller than the horizontal wind; nevertheless, it is
31 the crucial component for the evolution of severe weather (Peterson and Balsley, 1979).
32 Adiabatic cooling associated with upward motion leads to the formation of clouds and
33 precipitation and adiabatic warming associated with downward motion leads to the dissipation of
34 clouds. Extensive studies have been done on the relationships between w and
35 precipitation/convection over the tropics (Back and Bretherton, 2009; Uma and Rao, 2009a; Rao
36 et al., 2009; Uma et al., 2011 and references therein). Thus, w plays a vital role in controlling
37 day-to-day changes in weather. Different scales of variability exist in w like microscale to meso
38 synoptic, and planetary-scales (Uma and Rao, 2009b). It also controls the energy and the mass
39 transport between the upper troposphere and lower stratosphere (Yamamoto et al., 2007, Rao et
40 al., 2008). In a nutshell, knowledge of w is crucial for evaluating virtually all physical processes
41 in the atmosphere. Hence precise measurements of w could serve a guiding factor for studying
43 The small magnitudes of w make it very difficult to measure, as the errors involved in
44 measurements are often larger than the actual values. Direct and indirect methods exist to
45 measure w (e.g. Doppler measurements using radars for profiling and sonic anemometers in the
46 boundary layer) as well as indirect computational methods (e.g., adiabatic, kinematic and quasi-
48 where radars are situated. Global estimates are derived diagnostically from horizontal winds and
49 temperatures. Indirect estimation, gives a general view on the distribution of ascending and
2
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
50 descending motion on the synoptic scale within the quasi-geostrophic framework (Tanaka and
52 Reanalyses evaluate the vertical pressure velocity (omega) using indirect estimation (e.g.,
53 Dee et al., 2011). However, reanalyses combine both observations and model outputs to produce
54 systematic variation in the atmospheric state (e.g., Fujiwara et al., 2017). For example, in the
55 kinematic method, omega is estimated by integrating the mass continuity equation assuming
56 inviscid adiabatic flow. However, this kinematic estimate suffers from errors in the observations
57 as omega is estimated from horizontal divergence (Tanaka and Yatagai, 2000). A 10% error in
58 the wind may lead to a 100% error in the estimated divergence (Holton., 2004). Omega from the
59 thermodynamic energy equation is less sensitive to horizontal winds as it mainly depends on the
60 temperature gradient. However, in this method the local rate of change in temperature must be
61 measured accurately, meaning that observations must be taken at frequent intervals in time to
62 estimate ∂T /∂t accurately (Holton., 2004). This methodology fails in areas of strong diabatic
63 heating, especially where condensation and evaporation are involved. The quasi-geostrophic
64 method for estimating omega neglects ageostrophic effects, friction and diabatic heating
65 (Stepanyuk et al., 2017). It is to be noted from the above discussions that reanalyses are not
66 error-free owing to the many underlying approximations and assimilations involved (Kennedy et
67 al., 2012).
68 There are few indirect methods by which we can derive w from radar measurements in
69 the middle and upper atmosphere, where direct measurement of vertical wind are not possible
70 due to technical constraints. These methods include Doppler weather radar, Medium Frequency
71 (MF) radar and meteor radar. Doppler weather radar uses an indirect method to calculate vertical
72 winds (Liou and Chang, 2009; Matejka, 2002). Meteor radar also cannot determine vertical
3
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
73 velocity directly as the winds are determined from meteor showers using a wide beam width. As
74 a consequence, Laskar et al. (2017) calculated vertical wind from meteor wind radar data based
75 on a “kinematic” method using the continuity equation and hydrostatic balance. Dowdy et al.
76 (2001) have calculated vertical wind using the horizontal momentum and mass continuity
77 equations from the MF radar data. However, indirect methods are only adopted when direct
79 Very-high frequency (VHF) and ultra-high frequency (UHF) vertical pointing radars are
80 the most powerful tools for determining the vertical air motion (velocity) directly with high
81 temporal and vertical resolution. However, the magnitude may still not be directly comparable
82 between reanalyses products and observations as the reanalyses provide the intensity of vertical
83 air motion over wide areas (> 25 km2), whereas the direct radar measurements provide
84 information for the column over a single location. Thus, the best way to assess reanalysis
85 estimates of w is to compare its directional tendencies with those of radar. To the author’s
86 knowledge, no studies yet exist concerning with the assessment of w products derived from
87 different reanalyses and evaluation of these products against radar measurements. The present
88 study, which is therefore first of its kind, focuses on assessment of w among various reanalyses
89 using VHF radar measurements from two tropical stations where convective activity is frequent:
90 Gadanki (13.5oN and 79.2oE) and Kototabang (0.2oS and 100.2oE). Evaluations of this type are
91 critically important as reanalyses estimates of w are widely used by the scientific community to
92 understand and simulate a variety of atmospheric processes. In section 2, the data and
93 methodology are described. Section 3 contains the main results followed by a discussion and
4
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
98 Troposphere Radar (IMSTR) located at Gadanki and the Equatorial Atmosphere Radar (EAR)
99 located at Kototabang. Both the IMSTR and EAR are pulsed coherent radars operating at 53
100 MHz (IMSTR) and 47 MHz (EAR) respectively. These instruments are used to estimate w by
101 measuring the Doppler shift in the vertical beam. The technical details and operational
102 parameters of the IMSTR have been given by Rao et al., (1995) while those for the EAR have
104 In the present study direct measurements of w from VHF radars are used to assess
105 vertical motion between the surface and the lower stratosphere. Data collected from the IMSTR
106 between 17:30 and 18:30 LT (LT=GMT+5:30 hr) from 1995 to 2015 are analyzed using the
107 adaptive method (Anandan et al., 2001). This is the common operational mode of the IMSTR for
108 deriving the winds, and represents the only data available for such a long period of time. In
109 general, 4-8 vertical profiles are averaged to create daily profiles. Averaging is conducted using
110 the arithmetic mean as it represents the central tendency, which is generally used for wind
111 averaging. In a vertically pointing beam, signal-to-noise ratio (SNR) decreases with height
112 except in areas of stable layer (like the tropopause) and in the presence of strong turbulence.
113 Above 25 km, the SNR becomes constant in the absence of atmospheric signals. Data in this
114 region can be therefore treated as noise and used to estimate the threshold SNR (Uma and Rao,
115 2009b). It is found that noise levels lie between -17 dB and -19 dB with a 2 value of 3 dB
116 (where is the standard deviation). Thus data having SNR less than -15 dB are discarded from
117 the present analysis. Data from intense convective days (checked for individual profiles), defined
118 as w being less/greater than ± 1 ms-1 are also discarded as these data severely bias the
5
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
119 climatological mean vertical velocity (e.g. Uma and Rao, 2009b). The EAR provides quality
121 continuously and this study uses every hour data (diurnal data of single day) from 2001 to 2015.
122 The EAR data during convective periods are eliminated following the same criteria as for the
123 IMSTR, a second screening step. Each full diurnal cycle (after removing convective profiles) is
124 averaged and considered as a single daily profile for the EAR. For both radars, vertical velocity
126 (1)
127 where is the radar wavelength (in cm) and fd is the Doppler velocity (Hz).
128 It is known that estimates of w derived from VHF radar measurements are vulnerable to
129 biases due to tilting layers, strong horizontal winds (e.g., jet-stream), complex topography,
130 Kelvin-Helmholtz instabilities and gravity waves (Rao et al., 2008 and references therein). Rao
131 et al., (2008) has discussed in detail the biases that can cause spurious diagnosis of downward
132 wind as proposed by Nastrom & VanZandt (1994). In addition, they have also discussed the
133 potential biases caused by beam pointing errors as mentioned by Hauman and Balsley (1996) and
134 have conducted critical analysis to rule out beam pointing biases from VHF radar data. As
135 proposed by Nastrom & VanZandt (1994) on the bias caused by gravity waves, Rao et al., (2008)
136 have investigated biases caused by gravity waves by calculating the variances and found that
137 downward wind below 10 km are not affected by gravity waves. Their analysis clearly showed
138 that the mean downward motion below 10 km and upward motion above 10 km are real and not
139 caused by measurement biases, and also that the existing biases do not change the direction of
140 the background w when measurements are averaged over longer periods.
6
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
143 Weather Forecasts (ECMWF) Interim reanalysis (ERAi) from 1995 to 2015 (Dee et al., 2011).
144 The nearest grid points are taken for Gadanki (13.68oN, 79.45oE) and Kototabang (0.35oS,
145 100.54oE). Although 37 pressure levels up to 1 hPa resolution are available, we have restricted
149 spatial (30 km) and temporal resolution (hourly) from the surface up to 80 km (137 levels).
150 ERA5 also features much improved representation especially over the tropical regions of the
151 troposphere and better global balance of precipitation and evaporation. Many new data types not
152 assimilated in ERAi are ingested in ERA5 (Hoffmann et al., 2018). The details are available in
153 Copernicus climate change service report (Hersbach and Dee 2016 and
154 https://cds.climate.copernicus.eu/cdsapp#!/home). The nearest grid points are again taken for
155 Gadanki (13.63oN, 79.31oE) and Kototabang (0.14oS, 100.40oE), and the data period is 2002-
156 2015.
159 (MERRA-2) is the latest reanalysis of the modern satellite era produced by the National
160 Aeronautics and Space Administration’s (NASA) Global Modelling and Assimilation Office
161 (GMAO). MERRA-2 data are provided on 42 pressure levels from the surface to 0.01 hPa with a
162 temporal resolution of 3 h and horizontal resolution of 0.5o in latitude by 0.625o in longitude◦.
163 Details have been provided by Gelaro et al. (2017). The nearest grid points are used for Gadanki
164 (13.5oN, 79.37oE) and Kototabang (0.14oS, 100.00oE), with coverage from 1995 to 2015.
7
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
167 Research (NCEP-NCAR) reanalysis is based on the NCEP operational model with a horizontal
168 resolution of 209 km and 28 vertical levels. Its temporal coverage is four times per day. NCEP-2
169 products are improved relative to NCEP-1, having fixed errors and updated parameterizations of
170 physical processes, as evaluated by Kanamitsu et al. (2002). The data for the present study
171 covers 1995 to 2015 and is extracted at the nearest grid points to Gadanki (12.5oN, 77.5oE) and
175 with new data assimilation and prediction systems (Kobayashi et al., 2015). New radiation
176 schemes, higher spatial resolution and 4D-var data assimilation with variational bias correction
177 for satellite radiances have been used to generate the JRA-55 products. This reanalysis includes
178 variation in greenhouse gas concentrations with time, as well as the new representations of land
179 surface parameters, aerosols, ozone and SSTs. The horizontal resolution of the forecast model is
180 ~60 km for JRA-55. The nearest grid points are taken for Gadanki (13.75oN, 78.75oE) and
182
183 For all the reanalyses data, w (in cm s-1) is estimated using the formula:
184 (2)
185 where is the vertical velocity in pressure coordinates (in Pa s-1), T is the absolute temperature
186 (K), p is the atmospheric pressure (hPa) and R (=287 J kg-1 K-1) is the gas constant. To compare
8
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
187 measured vertical wind with the reanalysis products, we take the reanalysis data corresponding to
188 12 GMT for Gadanki and the daily mean for Kototabang.
191 IMSTR (observations) and the ERAi, ERA5, MERRA-2, NCEP-2 and JRA-55 reanlalyses data
192 sets over Gadanki. Although the magnitudes are of the same order between the observations and
193 reanalyses,significant differences are identified in the figures. It is to be noted that convective
194 days are discarded in the radar analysis (observations) as mentioned in the previous section and
195 those days are also eliminated from all the reanalysis data sets. These differences may be
196 attributed to the spatial averaging implicit in the reanalyses products, whereas the radar
197 measurements are for a single point. Thus in the present study, we only discuss the tendency of w
198 as it is used to represent the global variation of w, rather than its magnitude. The IMSTR
199 observations show updrafts between 8 and 20 km, with the largest values in the tropical
200 tropopause layer (TTL, 12-16 km), from December to April. These features are not reproduced
201 by any of the reanalyses, which all show downdrafts from December to April between 1 km and
202 the tropopause level (mean tropopause is ~ 16.5 km). Comparatively, downdrafts are observed in
203 the IMSTR below 6 km in April, which may be attributed to pre-monsoon (March-May)
204 precipitation and evaporation (Uma and Rao, 2009a). Vertical velocity in ERAi differs in both
205 magnitude and direction from other reanalyses, especially in the lower troposphere from March
206 to June. Meanwhile, the magnitude of vertical velocity in ERA-5 is a little larger (than that in the
207 other reanalyses) from May to June. Updrafts are observed in the TTL by the IMSTR during
208 June, when all reanalyses show similar features but located below the TTL. During July and
209 August both the radar observations and the reanalyses show updrafts in the vicinity of the TTL.
9
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
210 Updrafts are observed in the TTL from September to November but the peak in the updrafts is
211 shifted lower than that observed by the IMSTR. Below 8 km, IMSTR shows downdrafts from
212 April to October. It is notable that the reanalyses only produce downdrafts below 2 km and are
213 unable to reproduce the downdrafts above 2 km. Earlier studies using the IMSTR showed similar
215 Uma and Rao (2009b) have reported the diurnal variation of w in different seasons. Their
216 observations have only 1-2 diurnal cycles per month over Gadanki. They found significant
217 variations as large as 6 cm/s over Gadanki using IMSTR. The present observations are limited to
218 16:30 to 17:30 IST, with all reanalysis data over Gadanki taken at 12 GMT (17:30 IST). Thus,
219 time-averaged climatological mean biases can be neglected. We have also analyzed w from the
220 EAR (Kototabang) where the observations are available for the full diurnal cycle (measurements
221 of hourly averages for 24 hrs of observations). All reanalysis data over Kototabang are averaged
222 for the full diurnal cycle. Figure 2 shows the monthly mean climatology of daily mean of w
223 observed by the EAR and five reanalyses over Kototabang. All the reanalyses agree well with
224 each other over Kototabang. Radar measurements of w at this location consistently show updrafts
225 in TTL region and downdrafts below 6 km (e.g. Rao et al., 2008). The updrafts in the TTL are
226 well reproduced by all the reanalyses although the peak magnitude of w and its vertical location
227 remain lower than observed.. However none of the reanalyses reproduces the downdrafts. A
228 distinct bimodal distribution in w from May to September (two peaks between 8-10 km and 14-
229 17 km) with a local minimum between 12 and 13 km is observed in the EAR measurements. The
230 magnitudes of both updrafts and downdrafts are larger than those observed over Gadanki. JRA-
231 55 produces the largest w among the reanalyses. The monthly means show significant differences
232 in the direction of w between the observations and the reanalyses below 6 km.
10
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
233 To establish the robustness of the results obtained from both the observations and
234 reanalyses we have used different averaging procedures to assess the consistency of the
235 variability in w at monthly scales. Monthly mean climatological profiles of w from radar
236 observations and various reanalyses over Gadanki and Kototabang respectively are shown in
237 Figure 3. Downdrafts in the troposphere are not captured by any of the reanalyses over either
238 location. By contrast, updrafts in the TTL are generally reproduced in the monthly mean though
239 they are often overestimated by the reanalyses. ERAi underestimates the magnitude of both
240 updrafts and downdrafts over Gadanki and while NCEP-2 underestimates the magnitude of
242 Monthly means calculated over five-year periods from both radar and ERAi are shown in
243 Figure 4 for Gadanki and Figure 5 for Kototabang. The reanalysis shows a similar behavior to
244 the overall climatology in each five-year average. The overall patterns of updrafts and
245 downdrafts in the radar measurements of vertical velocity are also similar, indicating a consistent
246 performance of the radar over the full 20 year analysis period.
247 To further elucidate potential biases in the results due to averaging, we have taken ERA-5
248 at 12 GMT and compared it to the daily mean (obtained by averaging w at different times of the
249 day) to show that the sampling restrictions at Gadanki do not bias the results obtained. Figures 6
250 and 7 show the mean w obtained at 12 GMT and also the mean obtained by averaging hourly
251 analysis for each day for Gadanki and Kototabang, respectively. ERA5 is chosen for this
252 evaluation as the data are available at one-hour interval. The analysis shows in the magnitude of
253 w, with 12 GMT generally showing larger magnitudes compared to the daily means over
254 Gadanki (although no such systematic differences is observed in Kototabang). The directional
11
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
255 tendencies are also similar in both the profiles at both locations. This analysis shows that the
256 results are not biased by taking data only at 12 UTC over Gadanki.
257 Our analysis to this point shows the level of consistency between the features observed
258 by the radar and the reanalyses. To further understand the relative differences among the
259 reanalyses we perform a monthly mean comparative analysis among the reanalyses, as shown in
260 Figure 8. In this case, we took ERAi as a reference and compare it with w products from other
261 reanalyses. We chose ERAi, because the zonal and meridional winds from this reanalysis have
262 been shown to compare well with radiosonde and rocket sounding observations over the Indian
263 equatorial region (Das et al., 2015). The solid lines in Figure 8 show the differences over
264 Gadanki, while the dashed lines show differences over Kototabang. Over Gadanki, the difference
265 between the ERAi and other reanalyses is less than ±0.5 cm/s during December-January-
266 February (DJF, winter). ERAi underestimates ERA5 compared to other reanalyses, while values
267 based on MERRA-2 are relatively larger than those in other reanalyses. During MAM, strong
268 downdrafts are found below 5 km with comparable magnitudes in all five reanalyses. ERAi
269 underestimates ERA5 and NCEP-2 during March, and all other reanalyses from April to
270 September. Values of w in ERAi are larger than those in NCEP-2 above 8 km. All five
271 reanalyses compare well at all atltiudes above 18 km. As expected, magnitudes are larger during
272 JJA than during other months. From October to November, the magnitude reduces to ±1 cm/s
273 with values from ERAi smaller than those from all other reanalyses except NCEP-2.
274 Over Kototabang, the magnitude of w is relatively larger than over Gadanki. It is
275 interesting to note that, ERAi underestimates MERRA-2 in all months over this location also
276 (MERRA-2 shows larger magnitudes compared to other reanalyses). Similarly values based on
277 EARi are larger than those based on NCEP-2. From December to February ERAi underestimates
12
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
278 MERRA-2 below 10 km and ERA5 between 10 and 15 km while overestimates NCEP-2 and
279 JRA-55. The overall bias pattern remains the same during MAM, except for differences relative
280 to JRA-55. From June-November, ERAi underestimates NCEP-2 and overestimates all the other
282 The direction of w is an essential metric for comparing the observations and reanalyses.
283 We therefore show the directional tendencies from the IMSTR and the EAR measurements with
284 relative to those from the reanalysis data. Figure 9a shows the directional tendencies based on the
285 IMSTR and the reanalyses over Gadanki, while Figure 9b shows the directional tendencies based
286 on the EAR and the reanalyses over Kototabang. The directional tendency is calculated at each
287 height for every month when the radar or reanalysis data exceed 1 cm/s in either direction. The
288 directional tendency for each month is estimated and then aggregated into seasons. These
289 directional tendencies are given in terms of percentage of occurrence with respect to height. The
290 directional tendency is calculated for w only if the magnitudes lie above ± 0.1 cm/s for both radar
291 retrievals and reanalyses. The tendency is calculated separately for updrafts and downdrafts.
292 Over Gadanki during DJF all reanalyses produce updrafts at rates of less than 10 % of
293 updrafts throughout the profile. During MAM these ratios increase to 15 %, with NCEP-2
294 producing updrafts about 25 % of the time. During JJA and SON, the percentage occurrence
295 increases with height from 25 % to a maximum of 50 % between 12 and 14 km. The percentage
296 of updrafts occurrence then decreases from 14 to 20 km. This tendency trend is similar for all the
297 reanalyses except ERA5 for which the percentage occurrence is less than 25 % during all
298 seasons. The maximum ratio of updrafts over Gadanki is located between 12 and 15 km altitude.
299 The percentage occurrence of downdrafts over Gadanki is also less than 50 % at all the
300 levels. During DJF and MAM the reanalyses produce downdrafts 40 to 50 % of the time a much
13
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
301 higher frequency compared to the updrafts (<10 %). However, these ratios decrease above 10
302 km. By contrast, the percentage of downdrafts produced during JJA and SON is less than that of
303 the updrafts, with frequencies less than 25 % in all the levels during these seasons. The
304 performance of ERA5 over Gadanki is very poor as the occurrence frequencies are very small for
306 Over Kototabang the percentage occurrence of updrafts increases with height in all
307 seasons reaching a maximum of 75- 90 % between 10 and 14 km. Above 14 km the percentage
308 decreases to a minimum of 5 % at 19 km. Updrafts are rarely produced by the reanalyses
309 altitudes less than 4 km. It is important to note that none of the reanalyses produce daily mean
310 downdrafts exceeding 1 cm/s between 6 and 16 km. The percentage of downdrafts increases both
312 NCEP-2 and JRA-55 show occurrence frequencies of downdrafts around 65 to 75 % above 18
313 km. The performance of ERA5 appears to be poor compared to the other reanalyses over this
315 4 Summary
316 The present study assesses the vertical motion (w) in reanalyses against direct radar
317 observations from the convectively active regions Gadanki and Kototabang. The assessment is
318 carried out for five different reanalyses, ERA-Interim, ERA-5, MERRA-2, NCEP-2 and JRA-55.
319 Measurements were collected using VHF radar at both locations. We have used 20 years of data
320 from Gadanki and 17 years of data from Kototabang. The following points summarize the results
322 a. The magnitude of w obtained from reanalyses is underestimated by 10-50% relative to the
14
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
324 b. Observations over Gadanki showed updrafts from 8 to 20 km year around. The reanalyses
325 only reproduced this feature during JJA and SON when magnitudes were larger than 0.5 cm/s
326 in the reanalyses. However, the vertical location of the updrafts differs between the
327 observations and the reanalyses. Downdrafts below 8 km are not captured well by reanalyses.
328 c. Over Kototabang, the reanalyses did not consistently produce downdrafts below 8 km in all
329 months. Updrafts in the UTLS are captured well; however, the peak in the vertical
331 d. Inter-comparison among the reanalyses shows that ERAi overestimates NCEP-2 and
332 underestimates the other three reanalyses with respect to the magnitude of w over both
334 e. Assessment of directional tendencies shows that updrafts are reproduced reasonably well in
335 all the five reanalyses but downdrafts are not reproduced at all.
336 Our analysis reveals that downdrafts are not well produced in reanalyses, and also the location of
337 the largest updrafts is shifted lower than in the observations. Hence the reanalyses should be used
338 with care for representing various atmospheric motion calculations (viz. diabatic heating,
339 convection, etc.,) that mainly depend on the direction of w. This study provides the reanalysis
340 community an initial basis to improve the methodology for calculating w in reanalyses, as this is
342
343 Acknowledgements
344 Authors would like to acknowledge all the technical and scientific staffs of National
345 Atmospheric Research Laboratory (NARL) and Research Institute of Sustainable Humanosphere
346 (RISH), who directly or indirectly involved in the radar observations. Thanks to all the reanalysis
15
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
347 data centre for providing the data through the portal of Research data archival (RDA) of
348 NCEP/UCAR. One of the author KVS thank Indian Research Organisation for providing
350
351 Data availability: Analysed data (both radars and reanalyses) used in this study can be obtained
352 on request. Raw time series data are available through open access in the following websites:
358 KNU conceived the idea for validation of vertical velocity among the reanalyses. SSD, MVR,
359 and KVS collected and analysed the MST radar spectrum data. All the authors contribute for
360 generation of figures, interpretation and manuscript preparation. The data used in the present
364
365
366
367
368
369
16
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
370 References
371 Anandan, V.K., Reddy, G.R., Rao, P.B.: Spectral analysis of atmospheric signal using higher
372 orders spectral estimation technique, IEEE Trans. Geosci. Remote Sens., 39, 1890.
374 Back, L. E., and C.S. Bretherton.: Geographic variability in the export of moiststatic energy
375 and vertical motion profiles in the tropical Pacific, Geophys. Res. Lett., 33, L17810,
377 Das, S. S., K. N. Uma, V. N. Bineesha, K. V. Suneeth and G. Ramkumar.: Four decadal
379 data: results from Thumba Equatorial Station, Q. J. Roy. Met. Soc., DOI:10.1002/qj.2632,
380 2015.
381 Dee D. P et al.: The ERA-Interim reanalysis: Configuration and performance of the data
382 assimilation system. Q. J. R. Meteorol. Soc., 137, 553–597, doi: 10.1002/qj.828, 2011.
383 Dowdy, A., R. A. Vincent, K. Igarashi, Y. Murayama and D.J. Murphy.: A comparison of mean
384 winds and gravity wave activity in the northern and southern polar MLT. Geophy.Res.Lett.,
387 Hagio, and Y. Yabugaki.: Equatorial Atmosphere Radar (EAR): System description and first
389 Fujiwara, M., Wright, J. S., Manney, G. L., Gray, L. J., Anstey, J., Birner, T., Davis, S., Gerber,
390 E. P., Harvey, V. L., Hegglin, M. I., Homeyer, C. R., Knox, J. A., Krüger, K., Lambert, A.,
391 Long, C. S., Martineau, P., Molod, A., Monge-Sanz, B. M., Santee, M. L., Tegtmeier, S.,
392 Chabrillat, S., Tan, D. G. H., Jackson, D. R., Polavarapu, S., Compo, G. P., Dragani, R.,
17
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
393 Ebisuzaki, W., Harada, Y., Kobayashi, C., McCarty, W., Onogi, K., Pawson, S., Simmons,
394 A., Wargan, K., Whitaker, J. S., and Zou, C.-Z.: Introduction to the SPARC Reanalysis
395 Intercomparison Project (S-RIP) and overview of the reanalysis systems, Atmos. Chem.
397 Gelaro, et al.: The Modern-Era Retrospective Analysis for Research and Applications, Version 2
399 Hersbach, H. and Dee, D.: ERA5 reanalysis is in production, ECMWF Newsletter, Vol. 147, p.
401 2016.
402 Hoffmann, L., Günther, G., Li, D., Stein, O., Wu, X., Griessbach, S., Heng, Y., Konopka, P.,
403 Müller, R., Vogel, B., and Wright, J. S.: From ERA-Interim to ERA5: the considerable
406 Huaman, M., and B. B. Balsley.: Long-term average vertical motions observed by VHF wind
407 profilers: The effect of slight antenna pointing inaccuracies, J. Atmos. Oceanic Technol., 13,
410 NCEP-DOE AMIP-II reanalysis (R-2). Bull. Am. Meteorol. Soc., 83: 1631–1643, doi:
412 Kennedy, A. D., X. dong and B. Xi.: Comparison of MERRA and NARR Reanalyses with the
414 Kobayashi, S., et al.: The JRA‐55 Reanalysis: General specifications and basic characteristics, J.
18
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
416 Laskar, F. I., et al.: Experimental evidence of arctic summer mesospheric upwelling and its
417 connection to cold summer mesopause, Geophys. Res. Lett., 44, 9151- 9158.
419 Liou, Y.C., and Chang., Y. J.: A Variational Multiple-Doppler Radar Three-Dimensional Wind
420 Synthesis Method and Its Impacts on Thermodynamic Retrieval, Mon. Wea. Rev., 137:11,
422 Matejka, T.: Estimating the most steady frame of reference from Doppler radar data, J. Atmos.
424 Nastrom, G. D., and T. E. VanZandt.: Mean vertical motions seen by radar wind profilers, J.
426 Peterson, V. L., and B. B. Balsley.: Clear air Doppler measurements of the vertical component of
427 wind velocity in the troposphere and stratosphere. Geophys. Res. Lett., 6(12), 1979,
428 Rao, P. B., A. R. Jain, P. Kishore, P. Balamuralidhar, S. H. Damle, and G. Viswanathan.: Indian
429 MST radar 1. System description and sample vector wind measurements using ST mode,
431 Rao, T.N, K. N. Uma, D. Narayana Rao, and S. Fukao.: Understanding the transportation process
432 of tropospheric air entering the stratosphere from direct vertical air motion measurements
433 over Gadanki and Kototabang, Geophys. Res. Lett., 35, L15805,
435 Rao, T. N., Uma, K. N, T. M. Satyanarayana and D. N. Rao.: Differences in Draft Core Statistics
436 from the Wet to Dry Spell over Gadanki, India (13.5oN, 79.2oE), Mon.Wea.Rev, 137, 4293-
19
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
438 Rao, T. N., K. K. Kumar, S. S,. Das, T. N. Rao, T. M. Satyanarayana.: On the Vertical
439 Distribution of Mean Vertical Velocities in the Convective Regions during the Wet and
440 Dry Spells of the Monsoon over Gadanki, Mon.Wea.Rev, 140, 398-410, 2011.
441 Rao, V., D. Rao, M. V. Ratnam, K. Mohan, and S. Rao.: Mean Vertical Velocities Measured by
442 Indian MST Radar and Comparison with Indirectly Computed Values, J. App. Meteo, 42(4),
444 Stepanyuk, O., R. Jouni, V. Sinclair, Heikki, , Järvinen.: Factors affecting atmospheric vertical
445 motions as analyzed with a generalized omega equation and the OpenIFS model. Tellus., 69.
447 Tanaka, H. L., and A. Yatagai.: Comparative study of vertical motions in the global atmosphere
449 Uma, K. N., and Rao, T. N.: Characteristics of Vertical Velocity Cores in Different Convective
450 Systems Observed over Gadanki, India, Mon.Wea.Rev., 137, 954-974, DOI:
452 Uma, K. N., and Rao, T. N.: Diurnal variation in vertical air motion over a tropical station,
453 Gadanki (13.5oN, 79.2oE), and its effect on the estimation of mean vertical air motion, J.
455 Uma, K. N., Kumar, K. K., Das, S.S., Rao, T. N., and Satyanarayana, T. M.: On the Vertical
456 Distribution of Mean Vertical Velocities in the Convective Regions during the Wet and Dry
457 Spells of the Monsoon over Gadanki, Mon. Weather Rev., 140, 398–410,
20
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
459 Yamamoto, M. K., N. Nishi, T Horinouchi, M Niwano, and S Fukao.: Vertical wind observation
460 in the tropical upper troposphere by VHF wind profiler: A case study, Rad, Sci., 42, RS3005,
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
21
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
485 Figure 1. Climatological monthly mean altitude profile of vertical velocity obtained from MST
486 Radar and 5-reanalysis at 12 GMT over Gadanki. Horizontal lines indicate the standard error.
487 Figure 2. Same as Fig.1, but for diurnal mean over Kototabang.
488 Figure 3 : Monthly mean climatology of vertical velocity obtained from (a) radars, (b) ERAi, (c)
489 ERA-5, (d) MERRA-2, (e) NCEP-2, and JRA-55 over Gadanki (left) and Kototabang (right).
490 Gadanki data are at 12 GMT and Kototabang data are diurnal mean.
491 Figure 4. Monthly mean vertical velocity obtained from (a) MST Radar and (b) ERAi for 5
492 years interval (from top to bottom) over Gadanki (12 GMT).
493 Figure 5. Same as Fig.4 but for diurnal mean over Kototabang.
494 Figure 6. Height profile of vertical velocity at 12 GMT and diurnal mean (with 1 hour
495 resolution) over Gadanki extracted from ERA-5 (highest available time resolution).
497 Figure 8. Comparison of relative differences in vertical velocity (w) between the reanalysis for
498 Gadanki (solid line) and Kototabang (dash line). Individual month differences are estimated and
499 then averaged for each month. Over Gadanki, data is taken for 12 GMT and for Kototabang it is
500 diurnal.
501 Figure 9. Comparison of directional tendency simultaneously observed in radar and various
502 reanalysis data sets for (a) Gadanki and (b) Kototabang. Updrafts are shown in top and third
503 panels and downdrafts are shown in middle and bottom panels (for details see text).
22
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
Figure 1. Climatological monthly mean altitude profile of vertical velocity obtained from MST
Radar and five reanalyses over Gadanki at 12 UTC. Horizontal lines indicate the standard error
in each data set.
23
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
Figure 2. Same as Fig.1, but for daily mean profiles over Kototabang.
24
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
Figure 3 : Monthly mean climatologies of vertical velocity obtained from (a) radars, (b) ERAi,
(c) ERA5, (d) MERRA-2, (e) NCEP-2, and JRA-55 over Gadanki (left) and Kototabang (right).
Gadanki data are at 12 GMT and Kototabang data are daily means.
25
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
Figure 4. Monthly mean vertical velocity obtained from (a) MST Radar and (b) ERAi for 5-
years intervals (from top to bottom) over Gadanki (12 GMT).
26
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
27
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
Figure 6. Height profiles of vertical velocity for 12 GMT and from daily mean (with 1 hour
resolution) over Gadanki extracted from ERA5 (highest available time resolution).
28
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
29
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
Figure 8. Comparison of relative differences in vertical velocity (w) between the reanalysis for
Gadanki (solid line) and Kototabang (dash line). Individual month differences are estimated
relative to ERAi and then averaged for each month.
30
https://doi.org/10.5194/acp-2020-18
Preprint. Discussion started: 27 March 2020
c Author(s) 2020. CC BY 4.0 License.
Figure 9. Comparison of directional tendencies between the radars and various reanalysis data
sets for (a) Gadanki and (b) Kototabang. Updrafts are shown in the upper panels and downdrafts
are shown in the lower panels for each site (for details see text).
31