Joint Analysis of Convective Structure from the APR-2 Precipitation Radar and the DAWN Doppler Wind Lidar During the 2017 Convective Processes Experiment (CPEX)

The mechanisms linking convection and cloud dynamical processes is a major factor in much of the uncertainty in both weather and climate prediction. Further constraining the uncertainty in convective cloud processes linking 3-D air motion and cloud structure through models and observations is vital for improvements in weather forecasting, and understanding limits on atmospheric predictability. To date, there have been relatively few airborne observations specifically targeted for linking the 3-D air motion surrounding developing clouds to the subsequent development (or non15 development) of convective precipitation. During the May-June 2017 Convective Processes Experiment (CPEX), NASA DC-8-based airborne observations were collected from the JPL Ku/Ka-band Airborne Precipitation Radar (APR-2) and the 2-um Doppler Aerosol Wind (DAWN) lidar during approximately 100 flight hours. For CPEX, the APR-2 provided vertical air motion and structure of the cloud systems in nearby precipitating regions where DAWN is unable to sense. Conversely, DAWN sampled vertical wind profiles in aerosol-rich regions surrounding the convection, but is unable to sense the wind 20 field structure within most clouds. In this manuscript, the complementary nature of these data are presented from the June 10-11 flight dates, including the APR-2 precipitation structure and Doppler wind fields, and adjacent wind profiles from the DAWN data.

Abstract. The mechanisms linking convection and cloud dynamical processes is a major factor in much of the uncertainty in both weather and climate prediction. Further constraining the uncertainty in convective cloud processes linking 3-D air motion and cloud structure through models and observations is vital for improvements in weather forecasting, and understanding limits on atmospheric predictability.
To air motion and structure of the cloud systems in nearby precipitating regions where DAWN is unable to sense. Conversely, DAWN sampled vertical wind profiles in aerosol-rich regions surrounding the convection, but is unable to sense the wind 20 field structure within most clouds. In this manuscript, the complementary nature of these data are presented from the June 10-11 flight dates, including the APR-2 precipitation structure and Doppler wind fields, and adjacent wind profiles from the DAWN data.
The mechanisms linking convection and cloud dynamical processes is a major factor in much of the uncertainty in both 25 weather and climate prediction. The associated mesoscale convective systems (MCS) produce much of the Earth's rainfall and are responsible for the bulk of the heat and moisture transport from the Earth's surface into the upper troposphere. The cold pool dynamics are thought to be an important mechanism to facilitate the development of MCSs in the tropical atmosphere (Chen et. al., 2015;Zuidema et. al., 2017), as well as interactions between individual isolated convective storms (Raymond et. al., 2015). These atmospheric boundaries can have significant impact on deep convection, affecting its 30 initiation, updraft strength and longevity. The intensity and size of the cold pools is strongly dependent upon the vertical distribution of the temperature and humidity and the vertical shear of the horizontal wind. While the overall processes responsible for these interactions have been identified for some time, their precise nature and interactions remains underconstrained by observations, due to the difficulty in obtaining accurate, vertically resolved pressure, temperature, wind and water vapor in the proximity of developing convective clouds. Moreover, increasing evidence points to control of 35 convection by the relatively smaller and more variable amount of moisture above the boundary layer, in the free troposphere (Schiro and Neelin, 2019). Further constraining the uncertainty in convective cloud processes linking 3-D air motion and cloud structure through models and observations is vital for improvements in weather forecasting and understanding limits on atmospheric predictability.

40
The resolution of the precipitation radar onboard the Tropical Rainfall Measuring Mission (TRMM;1997-2014 and the subsequent Global Precipitation Measurement (GPM; 2014-current) missions (4-km horizontal resolution; 250-m vertical) have enabled numerous observational-based studies of MCS convective structure and features (Jiang et. al., 2011).
However, the dynamical (air motion) wind field associated with MCS features at this scale not well-represented by current space-based wind profile observing capabilities. The majority of available atmospheric wind observations are primarily 45 water vapor and cloud-tracked atmospheric motion wind vectors (AMV) derived from operational geostationary satellites (Velden et. al, 2005), which can be refreshed as quickly as 15-minutes, but are mainly indicative of large-scale mid-to-upper level air motion patterns. Observations of wind vectors in the periphery of smaller-scale cloud systems, especially in the 2km nearest the Earth (the approximate delineation of the boundary layer) are much less abundant. Outside of ground-based profiling networks, very few over-ocean wind profile observations at a similar GPM-like horizontal resolution are available. 50 A space-based Doppler wind lidar (DWL) capability has been envisioned as one means to overcome this observational shortcoming (Baker et. al., 2014). Over the past decade, airborne DWL field campaigns have been conducted (Lux et. al., 2018), recently in preparation for the deployment (August 2018) of the first-ever spaceborne DWL, the Atmospheric Dynamics Mission (ADM-Aeolus) of the European Space Agency (ESA) (Stoffelen et. al., 2005). Aeolus provides vertical 55 profiles of the horizontal line-of-sight (LOS) winds at an »100-km horizontal resolution and 200-km separation between profiles, with a main application to numerical weather prediction data assimilation (Horányi et. al., 2015). Observations from campaigns with a DWL such as the THORPEX Pacific Asian Regional Campaign (TPARC) were largely focused towards improvement of tropical cyclone forecasts (Pu et. al., 2010). These airborne campaigns have validated the capabilities of a DWL to provide wind profiles in the boundary layer (Bucci et. al., 2018;Zhang et. al., 2018). There has 60 been relatively less focus in collection and analysis of airborne DWL observations in relation to the convective processes linking air motion and transport of water vapor near clouds, and the subsequent development (or non-development) of convection.
One main reason is that previous campaigns often lacked nadir scanning Doppler precipitation radar capabilities on the same aircraft to enable matched radar-DWL observations. A scanning precipitation radar provides the vertical structure (Rowe and Houze, 2014;Rowe et. al., 2012). These data provide one means to validate the forecasted model precipitation structure (e.g., presence/absence of convection, timing, location), that results when the DWL wind vectors are assimilated into cloud resolving models.
In this manuscript, airborne DWL and Doppler precipitation radar observations are presented from the NASA-sponsored 70 Convective Processes Experiment (CPEX), which took place between 25 May and 24 June 2017, based out of Fort Lauderdale, FL. The goals of CPEX were to improve the understanding of convective processes during initiation, growth, and dissipation, using a combination of observations and cloud-resolving models. In particular, to measure what combinations of environmental structure and observed convective properties such as vertical velocity and reflectivity profiles, result in rapid upscale growth of a convective system into a large organized mesoscale convective system (MCS), or 75 alternatively, result in failure to grow or rapid decay. This manuscript will describe and present only the airborne precipitation radar and DAWN observations; a separate manuscript will present the associated mesoscale model simulations and DAWN data assimilation experiment results (Zhang et. al., 2019).

CPEX Overview.
During CPEX, sixteen NASA DC-8 airborne missions were flown into the Gulf of Mexico, Caribbean Sea and the Atlantic 80 Ocean. Each date is summarized in Table 1 The dropsondes system used during CPEX was the High Definition Sounding System (HDSS) dropsonde delivery system 85 developed by Yankee Environmental Services (Black et. al., 2017).
The dropsonde data are not presented in this manuscript.

90
APR-2 is a 2-frequency Doppler radar, originally developed as an airborne prototype for the second-generation GPM/DPR precipitation radar (Sadowy et al., 2003). The APR-2 has flown in numerous airborne field campaigns outside of CPEX, most recently the ORACLES (2016-2018) and CAMP2Ex (2019) campaigns. APR-2 acquires simultaneous measurements of multiple parameters at both Ku-and Ka-band (14 and 35 GHz, respectively), including co-and cross-polarized radar 95 backscatter, and LOS Doppler velocities of hydrometeors, with a maximum unambiguous velocity of ±27.5 (Ku-band) and ±10.4 (Ka-band) m s -1 . From a nominal 10-km flight altitude, the horizontal resolution at the surface is ~800-m, with a vertical range resolution and sampling of 50-and 30-m (slightly oversampled). Based upon analysis of radar surface backscatter measurements from CPEX, the reflectivity calibration is accurate to within 1-2 dB. From these basic measurements, APR-2 can depict the cloud macroscopic structure (extent, vertical air motion) and estimate the 100 microphysical structure (water content, precipitation intensity, hydrometeor size distribution) of the associated precipitation (Durden et. al., 2012). These resolutions are adequate to capture cloud features down to the resolution typical of highresolution cloud models, and appropriate for comparison with DAWN wind profiles in the vicinity near isolated, scattered, and organized deep convection. DAWN is NASA's airborne DWL with a 2-micron laser that pulses at 10 Hz (Kavaya et. al., 2014). It has previously 105 participated in the NASA Genesis and Rapid Intensification Processes (GRIP) (2010)  only two azimuth angles, -45 o and 45 o . Since these LOS wind profiles view the local wind field from multiple azimuth angles, multiple LOS profiles are analyzed to estimate the vertical profile of the horizontal wind components (u, v) at different pressure levels using the Adaptive Signal Integration Algorithm (ASIA) processing (Kavaya et. al., 2014). DAWN data are available in both the native LOS format, and processed wind vector (u, v) profile format. In this manuscript, the wind vector data are used to evaluate the wind field near clouds captured by the APR-2. The individual LOS data are 115 projected (along the viewing direction) through the APR-2 radar scan to illustrate the ability of DAWN to sense in and near cloud structures. The intent of this section is to assess the DAWN sampling density near the cloud systems captured by the APR-2, relative to 125 the cloud evolution. The 10 June 2017 case is highlighted in this section. This case is used since it is a fairly isolated cloud growth case, not greatly affected by large-scale forcing at early stages, and was covered by several repeat DC-8 passes from various directions. On 10 June 2017, the DC-8 took off from Fort Lauderdale near 1800 UTC and headed east towards the area of interest (AOI) with building clouds, located in the box bounded between 24.2N-26.2N latitude and 74W-72W longitude. Figure 2 shows  A series of convective box patterns were executed, to sample the evolution of the air movement surrounding the convection 140 from multiple flight bearings. The intent was to be on-station in order to capture developing cumulus clouds before they had developed any significant glaciation, before they reached a stage of vertical development where the DC-8 was unable to overfly from its nominal 10-km flight altitude. A photograph taken from the DC-8 near 2200 UTC (Figure 3) on this date illustrates an example of a cloud at a desired stage of evolution, where the clouds are captured at an early enough stage such that the DC-8 can safely overfly multiple times during subsequent evolution. 145

150
APR-2 data was collected in tandem with DAWN between 1835-2230 UTC. To explain the DAWN observations relative to the development of the precipitation, the analysis is broken into four one-hour segments, separated by the DC-8 flight track segments during each hour. The APR-2 data will be shown in context to give a sense of when and where (proximity and cloud penetration depth) DAWN can provide valid wind data. Furthermore, the analysis will focus on the wind shear within each quadrant (NE=northeast, SE=southeast, SW=southwest, NW=northwest), relative to the approximate center (25.2N 155 73W) of the AOI flight box in Figure 2. These segments also correspond to the data assimilation interval used in the investigation of these data by Zhang et. al. (2019).

Flight Segment 1 (1830-1930 UTC).
This first DC-8 flight segment flew along a 120-degree bearing approaching the NW and SW quadrants of the AOI, whereas the next three flight segments discussed below take place inside of the main AOI. Figure 4 shows the plan view at 2-km 160 170 For the 2-km level, the maximum APR-2 Ku-band reflectivity between 1-3 km is plotted underneath the DAWN winds; for the 8-km level the maximum Ka-band reflectivity between 7-9 km is shown instead (the rationale being that since there is less path attenuation through rain at Ku-band than at Ka-band, the Ku-band data provide a better depiction of the cloud structure for the deeper 2-km level; the APR-2 is more sensitive to clouds at Ka-band than at Ku-band, so the Ka-band 175 reflectivity was used for the higher 8-km level cloud structure). Peak APR-2 Ku-band reflectivity at 2-km exceeded 30 dBZ.
In Figure 4, the associated cloud and aerosol conditions were such that the processing of these DAWN LOS data produced a  showing 5 m s -1 southerly winds at 2-km, becoming more westerly at 8-km height. In these DAWN data, there is a tendency for increased directional shear between these two vertical levels as the DC-8 approaches the AOI. To enhance this feature, the left panel of Figure 7 displays each DAWN profile in Figure 4 in a twolevel hodograph form, where each vector points from the DAWN (u,v) wind at 2-km to the (u,v) at 8-km, thereby 210 representing the shear between these two levels. When the vector is aligned along a radial direction, that indicates no directional shear, only speed shear. When the vector is aligned away from the radial direction, that indicates directional shear and possible speed shear. The shear vectors are colored according to which quadrant (NE, SE, SW, NW) they are located in, relative to the approximate center (25.2N 73W) of the AOI flight box in Figure 2. During this time there is sustained directional wind shear in the SW and NW quadrants, oriented from west to east. A similar analysis for the shear 215 between 2-km and 6-km (right panel of Figure 7) shows the shear oriented more south to north.

225
To provide a depiction of the DAWN vertical sampling capability, a cross-section of the DAWN vertical profile sampling locations superimposed upon the APR-2 nadir reflectivity is shown in Figure 8. The black points represent locations of valid DAWN (u, v) wind vectors during this time. Several notable features are evident. Depending upon the APR-2 transmit pulse length, there is a blind zone (~ 1.8 km) below the aircraft where the radar processor does not receive any returned 230 signals. This is noted in a short period where the cloud tops were within the APR-2 blind zone (near scan 750), but the cloud top was identifiable in the DAWN profiles (labeled the "upper cloud area" in green shading in Figure 8). Similarly, near the surface where the APR-2 backscatter is affected by ground clutter in the lowest 500-m, DAWN was able to provide wind observations to the surface. In general, DAWN winds are abundant above 6-km (where the SNR is highest), and below 3-km (where the aerosol content is higher), with considerable upper level sampling right up to the edges of the tall developed 235 clouds (near scan 1000). There are several DAWN profiles that bump up close to the small convective cell near scan 1800 (denoted with a red ellipse in Figure 8), which are associated with the clouds shown in Figure 6 (Box 2) top panel, where the Ku-band reflectivity exceed 30 dB. To show this area in more detail, Figure 9 zooms in to the Box 2 area (1924-1930, where three small growing clouds are shown in the middle of this figure. DAWN wind profiles are produced to the surface next to growing convection near scans 100 and 120, but not for the cell near scan 75. This highlights that convective clouds 240 are not continuous "impenetrable" cloud structures, but in nature have gaps or "holes" in them where the DAWN LOS view can penetrate through to lower levels.  DAWN (u, v) wind vectors produced from the DAWN processing of the LOS data. The green shaded "upper cloud" area shows an area where there are clouds in the 1.8-km blind zone (where APR-2 does not process data), but whose cloud top is noted in the DAWN profiles above this shaded area.

Flight Segment 2 (1930-2030 UTC).
From 1930-2030 UTC, the DC-8 conducted a series of flight legs in a counter-clockwise pattern within the AOI, with densest sampling in the NW and SE quadrants, before departing along a 270-degree bearing. Figure 10 illustrates the APR-2 and DAWN data in the same format as used in Figures 5 and 6. Maximum Ka-band reflectivity in the 7-9 km level are near 20-25 dB in the middle of the segment. On the north side of the AOI, the winds were mainly southwesterly near 10 m s -1 , 260 with 2-km level winds more southerly with weaker 5 m s -1 speeds.

270
In the NW quadrant of the AOI, there is a large shear magnitude between the 2-and 8-km levels ( Figure 11), but it is less directional (vectors more aligned in the radial) compared to Figure 7. In Figure 11, the shear between 2-and 6-km in the NW quadrant (green arrows) is similar to Figure 7, but the shear between 2-and 8-km is pointing more towards the east.
The shear between 2-and 8-km in the SW and SE quadrants (red and blue arrows, respectively) points mostly towards the east-southeast directions, but this same signature is not well noted between the 2-and 6-km levels, owing to the reduced 275 DAWN sampling at the 6-km level.

280
The vertical cross section of the DAWN wind profiles sampling locations alongside the APR-2 nadir reflectivity profile is shown in Figure 12 (areas where the DC-8 was making a banking turn are omitted). Similar to flight segment 1, the two main "no-cloud" regions between APR-2 scans 600-900 and 1300-2000 are well sampled at the upper and lower heights levels. Near scan 850, DAWN data stops near 8-km in areas where APR-2 does not show any cloud, and several profiles 285 near scan 900 sense deeper (to nearly 4-km), both of which may be from lidar backscatter off of clouds not sensed by APR-2 (i.e., below the minimum Ka-band detectability). The lowest-most level retrieved by DAWN near scan 300 and again near scan 1200 appear to be the cloud top, which occurred in the 1.8-km blind zone (~ 8.2-10 km height) area where APR-2 does not provide any data. Near scan 400, there are numerous DAWN profiles provided in cloud gaps as the DC-8 passed through some higher-level clouds. 290

Flight Segment 3 (2030-2130 UTC).
Flight segment 3 begins with the DC-8 heading in a northerly direction. The flight revisited some of the area sampled during the previous segment by executing a box pattern in clockwise direction, before exiting to the east along a 90-degree bearing ( Figure 13). Towards the end of this flight segment, the DC-8 dropped to a 9-km flight level. At 8-km height, 25 DAWN wind vectors were estimated from the LOS data processing. 300 The 90-degree directional shear on the SW quadrant of the AOI is still present, measuring about 5 m s -1 in magnitude ( Figure  305 14), but insufficient 8-km winds were obtained in the other quadrants for comparison (at the 2-km level, only 18 DAWN wind vectors were estimated, nearly all concentrated on the south side of the AOI). Figure 15 shows the DAWN vertical sampling density during this flight segment relative to the APR-2 Ka-band reflectivity structure. On the east side of the AOI the DC-8 passed above a region of thin clouds (as shown in the IR background in Figure 11), but were below the sensitivity of APR-2. This could be one reason for the reduced DAWN sampling between APR-2 scans 700-1000 in the 2-8 km height level, but the E-W leg (scans 1200-1400) provided DAWN profiling to the 315 surface in many locations.

Flight Segment 4 (2130-2230 UTC).
Flight segment 4 begins with the DC-8 heading in an easterly direction and then banking to a 225-degree bearing. The DC-8 partially completed a figure-eight pattern, before exiting to the west along a 270-degree bearing and returning to Florida, as 325 shown in Figure 16. The total DAWN profile sampling numbers are higher than segment 3, with 49 and 63 DAWN vectors provided at 2-and 8-km heights, respectively. The cloud system near 25.5N 73.5W has matured considerably relative to its structure in previous flight segments, represented with a fairly well-defined melting level shown near scans 1450-1550.  Figure 17 shows the shear in the NW quadrant between 2-and 8-km (and between 2-km and 6-km), pointing towards the 335 northeast along a near-radial direction (little directional shear). This period also gathers sufficient DAWN data in the SE quadrant (blue vectors) that was not well sampled in the earlier time segments. This shows evidence of shear between 2-km and 6-km pointing to the north, but shear between 2-km and 8-km pointing towards the east.
Narrow growing clouds were first overflown during scans 200-400 ( Figure 18). DAWN vertical sampling density during this time is fairly dense, with more winds provided in the 2-6 km height level than during flight segment 3, notably in the middle and end of this flight segment. When the DC-8 moved to a lower 9-km flight level, the pulse width was changed resulting in the APR-2 blind zone being shorted by one-half (to 0.9 km), which is evident for the tallest clouds near scans 345 400 and 1400. DAWN also provided overall better sampling in the mid-levels from this lower flight altitude, with almost complete top-bottom profiles towards the end of the flight segment. APR-2 also provided vertical air motion and structure of the cloud systems in the cloud-detected regions where the DAWN profiling capability was degraded. The purpose of this section is to examine a method to couple the two wind estimates near 355 clouds. By viewing clouds from multiple viewing directions near nadir, airborne Doppler radars sample a mixture of the vertical and horizontal winds associated with the movement of the hydrometeors being sensed (Heymsfield et. al., 1996). As the DC-8 moves forward and the APR-2 scans across-track, the measured Doppler velocity represents a combination of the vertical and across-track components of the hydrometeor motion within each APR-2 range bin . These data can provide some complementary wind direction information to complement DAWN, and under the right conditions (no 360 significant horizontal shear across the APR-2 scan swath) provide some continuity in the wind measurements between the cloud and no-cloud areas. The received Doppler velocity represents contributions from the motion of the hydrometeors owing to air motion, and the contribution owing to the (reflectivity-weighted) hydrometeor fall speed.
Define as the viewing angle from nadir (e.g., zero represents straight downward, and negative and positive denote the left and right sides of the APR-2 swath, respectively), and ! and " as the vertical and across-track wind components. Then the Doppler wind 365 at corresponding left and right sides of the swath is given by: where the subscripts left and right refer to the corresponding APR-2 beam positions at − (left side of swath) and + (right side of swath), respectively. The vertical (z) and across-track (y) wind components are easily solved for, Note that in this formulation, the effects owing to the hydrometeor fall speeds are still included, so the estimate of vz in (3) is not the same as the vertical (w component) wind due to air motion only. To account for the fall speed, the fall speed-380 reflectively relation developed by Black et al. (1996) is applied and only the 8-km level winds (where there has not yet been significant attenuation) are assessed. After this correction, vz is assumed equal to the w wind owing to air motion.
However, in general more rigorous radar inversion methods that account for the radar attenuation and the hydrometeor Doppler fall speed are required before this formulation can be applied to lower cloud levels (Guimond et. al., 2014) 385 This principle is examined on the APR-2 data gathered between 1800-2100 on 11 June 2017. Figure 19 shows the plan view, where there are abundant DAWN wind vectors at 8-km, including many that are close to clouds. There are six flight legs along a predominant 90-degree (W-E) or 270-degree (E-W) (+u and -u wind component direction, respectively) flight bearings, beginning near 1800, 1815, 1838, 1900, 1920 and 1955 UTC (with some slight deviations along these directions to avoid deep clouds near flight level). The first and last three of these flight legs occurred in predominantly cloud-free and 390 cloud-covered conditions, respectively. The top panel of Figure 20 shows the time intervals corresponding to these 90-and 270-degree bearings. In these flight directions, the APR-2 across-track wind component vy (4)

400
The bottom panel of Figure 20 shows the APR-2 vertical (vz) and across-track (vy) winds estimated from (3) and (4)

415
A second coincidence occurs between the APR-2 data near 1910 and 1925 UTC, where the APR-2 vy component flips sign between similar wind speed values. However, the area at 1925 UTC is so cloud-filled that there are no nearby DAWN wind profile data to compare to. It also represents an area with stronger vertical winds, where the assumption of no significant horizontal shear across the APR-2 scan swath is likely not valid. While this is not a rigorous comparison of DAWN and Doppler precipitation radar horizontal winds, the principle could be applied to any these data from any close time pair of 420 DC-8 flight bearing segments that are separated by 180-degrees. In this example, the flight bearings were fortuitously along easterly or westerly directions. For any arbitrary flight bearing, the cross-track winds estimated by (4) are more generally a combination of (u, v), and the DAWN (u, v) winds could be transformed to these same directions for comparison. This complement of Doppler radar and DWL observations could provide a means to link horizontal wind data outside of clouds and inside clouds (away from strong vertical motion, from APR-2), an important transition region. Space-based Doppler 425 radar measurement methods to estimate the horizontal LOS (HLOS) wind in-cloud have been proposed (Illingworth et. al., 2018), as one means to complement the HLOS winds from Aeolus. However, further investigation from CPEX and other APR-2 airborne data are needed to assess the quality of the radar wind components before they can be used for science or model data assimilation purposes.

5
Conclusions. 430 This manuscript has presented joint observations from the DAWN Doppler wind lidar and the APR-2 (Ku/Ka-band) Doppler precipitation radar, collected during the CPEX campaign in 2017. Data from NASA DC-8 flight segments from two flight dates were examined to assess the ability of DAWN to sense air motion nearby to developing convection.
The flight patterns on June 10-11 were selected for this purpose. For the June 10 flight date, the DC-8 arrived on-station to the area of interest, with sufficient time to capture the evolution of isolated, small-scale (< 10-km horizontal extent, many not yet 435 glaciated) clouds from numerous DC-8 repeat passes for about a 3-hour period. The environment surrounding the clouds on this date exhibited directional shear between the 2-and 8-km levels in the quadrant SW of the developing convection. A number of growing convective clouds with APR-2 echo tops below 5-km were sampled by the APR-2, away from the more developed convection. The capability of DAWN to collect LOS profiles near convection was highlighted for several passes where profile retrievals were possible up to the edges of many APR-2 detected cloud systems. On June 11, the DC-8 440 sampling pattern consisted of successive repeat passes on E-W and W-E flight bearings, where the cross-track winds from APR-2 were examined for consistency with nearby DAWN winds, in the proximity of cloud edges.
As stated in the introduction, this manuscript provides the observational context for a separate mesoscale model data assimilation study, which is aimed at quantifying the impact of the DAWN measurements on the analyzed atmospheric state 445 variables and on the forecasted precipitation when the DAWN wind profile observations were assimilated into the model (Zhang et. al., 2019). While only limited examples are shown, these particular findings highlight the importance of when and where the wind observations are taken, and provide guidance for assessing observational strategies and requirements needed for future airborne field campaigns with similar instrumentation.

Data Availability 450
The DAWN LOS and profile data (ASCII text format) and APR-2 data (HDF5 format) are available from the authors upon request.