Observations from the NOAA P-3 aircraft during ATOMIC

The Atlantic Tradewind Ocean-Atmosphere Mesoscale Interaction Campaign (ATOMIC) field campaign, part of the larger experiment known as Elucidating the Role of Clouds-Circulation Coupling in Climate (EURECA), was held in the western Atlantic during the period Jan 17 Feb 11 2020. This paper describes observations made during ATOMIC by the US National Oceanic and Atmospheric Administration’s (NOAA) Lockheed WP-3D Orion research aircraft based on the island of Barbados. The aircraft obtained 95 hours of observations over eleven flights, many of which were coordinated with the NOAA 5 research ship R/V Ronald H. Brown and autonomous platforms deployed from the ship. Each flight contained a mixture of sampling strategies including: high-altitude circles with frequent dropsonde deployment to characterize the large-scale environment; slow descents and ascents to measure the distribution of water vapor and its isotopic composition; stacked legs aimed at sampling the microphysical and thermodynamic state of the boundary layer; and offset straight flight legs for observing clouds and the ocean surface with remote sensing instruments and the thermal structure of the ocean with in situ sensors 10 dropped from the plane. The characteristics of the in situ observations, expendable devices, and remote sensing instrumentation are described, as is the processing used in deriving estimates of physical quantities. Data archived at the National Center for Environmental Information include flight-level data such as aircraft navigation and basic thermodynamic information (NOAA Aircraft Operations Center and NOAA Physical Sciences Laboratory, 2020); high-accuracy measurements of water vapor concentration from an isotope analyzer (National Center for Atmospheric Research, 2020); in situ observations of aerosol, 15 cloud, and precipitation size distributions (Leandro and Chuang, 2020); profiles of sea water temperature made with Airborne eXpendable BathyThermographs (AXBTs; NOAA Physical Sciences Laboratory, 2020a); radar reflectivity, Dopper velocity, and spectrum width from a nadir-looking W-band radar (NOAA Physical Sciences Laboratory, 2020b); estimates of cloud

sensing of the ocean surface and clouds. These flight patterns were placed over regions of sea surface temperatures gradients and/or areas being sampled by autonomous ocean vehicles (surface drifters, wave gliders) deployed from the Ron Brown.
Transits between Barbados and the daily operating area offered further opportunities for deploying dropsondes and AXBTs and for remote sensing. The P-3 flew eleven flights during ATOMIC for a total of 95 hours. The first eight flights took place 70 during the day, with nominal take-off times at 13:00 UTC (9:00 local time); the last three took place overnight, with takeoff times between 02:00 -03:30 UTC (local times between 22:00 and 23:30 pm the previous day). Table 1 provides an overview of sampling strategies and other information for each flight. A plan (map) view of the flight tracks is shown in Fig. 1; altitudes are shown as a function of flight time in Fig. 2. Table 1. Flight sampling strategies employed on each flight by the P-3 during ATOMIC. Flight date is UTC and most flights were 8-9 hours long. Numbers in parentheses show the number of AXBTs for which valid data were obtained. "RHB" indicates that the R/V Ronald H. Brown was at the center of a dropsonde circle. Most AXBT patterns deployed 20 instruments. "Cloud" indicates the number of cloud patterns flown; each typically involved sampling at four or five altitudes. See also Table 3 of Quinn et al. (2021). Detailed reports from each flight are available at the EUREC 4 A data portal (https://observations.ipsl.fr/aeris/eurec4a/). 3 Instrumentation and initial data processing 75 Table 2 describes the instrumentation on and deployed from the P-3 during ATOMIC. The instrumentation was similar to that used during hurricane reconnaissance flights and other scientific missions with the exception of the water vapor isotope analyzer provided by the National Center for Atmospheric Research (see Sec. 3.1.2) and the nadir-looking W-band cloud radar provided by NOAA's Physical Science's Lab (Sec. 3.3.1). Many of the basic in situ measurements are combined to provide derived quantities (e.g. wind speed, relative humidity) described in Sec. 4. providing an off-nadir look at the ocean surface useful for calibrating the W-band radar. A change in pilots midway through the experiment led to dropsondes being deployed from circular flight tracks starting on 31 Jan. AXBTs were deployed in lawnmower patterns (parallel offset legs) with small loops sometimes employed to lengthen the time between AXBT deployment to allow time for data acquisition given the device's slow fall speeds. Profiling and especially in situ cloud sampling legs sometimes deviated from straight paths to avoid hazardous weather. The color coding is drawn from a palatte spanning the length of the experiment, so that days that are close in time have similar colors.
for that given instrument, then is marked valid on the QC Checklist included in the Mission Documents (available from https://seb.noaa.gov/pub/acdata/2020/MET/ in directories labeled by flight date and the letter "I" to denote N43RF). In cases 90 where there is more than one reliable sensor, the one sensor is set as the reference, i.e. TDMref. The reference sensor is chosen to minimize data intermittency and maximize both comparisons to independent measurements (e.g. temperature may be compared to dropsondes) and self-consistency among measurements. The intent is to choose a single sensor which best represents the flight overall, even if this sensor might have periods of bad data (e.g. overshooting by the chilled-mirror dew point sensors) during the flight when other sensors might be more reliable. Additional parameters are derived (variable names end in ".d") and 95 corrected (variable names end in ".c") from these data. AOC produces and distributes one netCDF file per flight. Both raw data and the AOC summary file are available for all flights from NOAA's National Center for Environmental Information (see https: //www.ncei.noaa.gov/metadata/geoportal/rest/metadata/item/gov.noaa.ncdc%3AC00581/html where files may be selected by project). Aircraft altitude (km) Figure 2. Flight altitude as a function of time after take-off for the eleven flights by the NOAA P-3 aircraft during ATOMIC, using the same colors as Fig. 1. Sondes were dropped from ∼ 7.5 km, with each circle taking roughly an hour; transits were frequently performed at this level to conserve fuel. Long intervals near 3 km were used to deploy AXBTs and/or characterize the ocean surface with remote sensing.
Stepped legs indicate times devoted to in situ cloud sampling. On most flights the aircraft climbed quickly to roughly 7.5 km, partly to deconflict with other aircraft participating in the experiment. On the three night flights, however, no other aircraft were operating at take-off times and cloud sampling was performed first, nearer Barbados than on other flights.

Water vapor stable isotope analyzer 100
During ATOMIC the P-3 was equipped with a flight-ready Picarro L2130-i water vapor isotopic analyzer which measured the concentration of water vapor and its isotopic composition at 5 Hz frequency. ATOMIC was the first flight campaign for this newly-developed instrument although similar instruments have flown as part of previous airborne research missions (e.g. Sodemann et al., 2017;Herman et al., 2020). The isotope ratio measurements, which are part of a broad suite of such observations made during EUREC4A, are reported elsewhere; here we describe the instrument's fast and accurate measurements of water 105 vapor concentration (i.e. mixing ratio).
While in flight, the isotopic analyzer drew in ambient air through a backwards-facing 0.25-in/6.35 mm copper tube, centered within a National Center for Atmospheric Research HIAPER Modular Inlet (HIMIL). This ensured the selective sampling of water vapor (versus total water). Because mass but not volumetric flow was controlled through the copper tubing, the time delay (τ in seconds) for air entering the HIMIL to reach the isotopic analyzer varied as a function of pressure and temperature.This  Mixing of water vapor within the inlet system, and with molecules that have adsorbed to the copper tubing, also partially 120 smooths high frequency signals (Aemisegger et al., 2012). These effects are, however, fairly small and consistent across flights.
The aircraft hygrometer's time response, in contrast, is quite variable and depends on flight conditions, and the hygrometer is subject to both overshooting (e.g. when the measured signal surpasses the expected value following a rapid rise in environmental water vapor concentration) and ringing (i.e. rapid oscillations around the expected value) during rapid and large changes in water vapor concentration. This is illustrated in Fig. 3 which also highlights the hygrometer's much slower time response (as 125 compared to the isotopic analyzer) in the low humidity conditions found at the highest flight altitudes. Outside of these time periods, the agreement between the hygrometer and isotopic analyzer is quite good (lower panel). Given the more consistently accurate measurements of the isotopic analyzer during ATOMIC we recommend its use in preference to the aircraft hygrometer for characterizing the thermodynamic state of the atmosphere.

Dropsondes
The P-3 released 320 Vaisala Dropsonde RD41s during ATOMIC at the locations shown in Fig. 4. Most were released from 24000 ft/7.5 km though some were released from slightly lower altitudes during transits and others from 9000-10000 ft/2.75-3 145 km during cloud and AXBT flight patterns. The RD41 sensors measures pressure, temperature and humidity as the package falls from the plane, slowed by a parachute (Hock and Franklin, 1999). A GPS package provides location from which wind direction and wind speed are calculated and reported in real time. Measurements are available from the aircraft flight level to the ocean surface. Dropsondes from the P-3 were processed in real time during flight and made available for assimilation over the Global Telecommunications System.

AXBTs
A total of 165 AXBT instruments (Bane and Sessions, 1984;Dinegar Boyd, 1987;Alappattu and Wang, 2015) were deployed from the P-3 over seven flights at locations shown in Figure 5. Most were released at or near 9000 ft /2.75 km. The AXBTs, manufactured by Lockheed Martin Sippican, collect ocean temperature as a function of time after launch. AXBTs normally begin transmitting data when the sensor enters the ocean. One file was produced for each AXBT sensor by removing any 155 extraneous observations obtained before splashdown, then converting time to depth assuming a nominal in-water fall speed of 1.594 m s −1 . Location is determined from the aircraft navigational information at the time the AXBT was released. A median filter was applied to remove most (but not all) spurious outliers in ocean water temperature. Radar data were post-processed following Fairall et al. (2018). Standard processing produces vertical profiles of estimates of reflectivity, Doppler velocity, spectrum width, and cloud-free signal-to-noise ratio, converted to a uniform grid referenced to the sea surface rather than as distance from the aircraft. Reflectivity profiles are corrected for attenuation by atmospheric gases and precipitation. Absorption by water vapor and oxygen is calculated based on temperature, pressure, and relative Attenuation by precipitation is estimated using inversions and relationships from Hitschfeld and Bordan (1954) following Iguchi and Meneghini (1994). Measured Doppler velocity is corrected for the pitch and roll components of aircraft motion.

Remote
The vertical speed of the aircraft is calculated from flight level data (Sec. 3.1.1) and taken into account in the Doppler velocity correction especially during aircraft ascents and descents. An example from an hour of flight (18-19 UTC on Jan 19) is shown 175 in Fig. 6.

Wide Swath Radar Altimeter
The NOAA Wide Swath Radar Altimeter (WSRA, see Walsh et al., 2014;PopStefanija et al., 2020), developed and manufactured by ProSensing, Inc. of Amherst, MA, USA, is a digital beam-forming radar altimeter operating at 16 GHz in the Ku band.
It generates 80 narrow beams spread over ±30°to produce a topographic map of the sea surface waves and their backscattered 180 power. These measurements allow for continuous reporting of directional ocean wave spectra and quantities derived from this including significant wave height, sea surface mean square slope, and the height, wavelength, and direction of propagation of primary and secondary wave fields. Rainfall rate is estimated from path-integrated attenuation (Walsh et al., 2014).

Stepped Frequency Microwave Radiometer
The Stepped Frequency Microwave Radiometer (SFMR) is a nadir-looking microwave radiometer also built by ProSensing.

185
The instrument measures brightness temperatures of the ocean surface and intervening atmosphere at six C-band (4-7 GHz) frequencies. Surface wind speed and average columnar rain rate can be inferred from these brightness temperature values (Uhlhorn et al., 2007). The current implementation (Sapp et al., 2019) has its origins in prototypes developed by the Microwave Remote Sensing Laboratory at University of Massachusetts and NOAA's Hurricane Research Division and deployed in 1980; the hardware and retrieval methods have been improved several times since.

Infrared radiometer
The P-3 deployed three Heitronics KT19.85 passive infrared radiometers which measure radiation in the 9.6 -11.5 µm spectral range. One radiometer points horizontally out the port side of the plane; the others are zenith-nadir-looking. Measured radiation is converted to a brightness temperature. This system has a resolution of 0.1 • C and an accuracy of 0.5 • C plus 0.7% of the difference between the target and instrument housing temperatures. The field of view of 0.5°and the response time 195 approximately 0.5 s.

Post-processed and derived quantities
Data obtained during the experiment have been post-processed and a variety of derived quantities, as described in this section, have been produced. Data files are organized topically as described in Table 3 and explained more fully in this section. One file per day is provided for each of the entries in the table. All data are archived at the US National Center for Environmental

Isotope analyzer
Water vapor measurements from the isotopic analyzer are proportional to the ratio of the moles of water vapor to the moles 210 of moist air (dry air plus water vapor) and are reported as a volume mixing ratio in parts per million volume (ppmv). These are provided alongside estimates of the mass mixing ratio -the ratio of the mass of water vapor to the mass of dry air -at both the analyzer's native time resolution of nominal 5 Hz frequency and at a reduced resolution of 1Hz aligned with the P-3 aircraft data system through boxcar averaging. For convenience, estimates of relative humidity are also provided, obtained by multiplying the 1Hz volume mixing ratios by the aircraft static (ambient) pressure measurements and dividing by the saturation 215 vapor pressure, estimated from the aircraft ambient temperature following Hardy (1998).The files are aligned in time with the flight level data.
The isotopic analyzer's water vapor measurements have been corrected for a low bias of increasing magnitude at concentrations exceeding 10,000 ppmv, identified using a LiCOR 610 dew point generator. The uncertainty associated with this correction spans 26 to 29 ppmv for the humidity range 200 to 30,000 ppmv but reduces to 12 ppmv upon averaging to 1 Hz.

220
The accuracy of volume mixing ratios below 200 ppmv is unverified.

Microphysics
Particle size distributions for microphysical instruments are processed at 1 Hz resolution for each day where data was available.
The CAS and CDP are processed with standard codes available from the manufacturer (although, as noted in Sec. 3.1.3, the CDP did not produce useful data during ATOMIC and is therefore not included in the final archived product). The CIP and 225 PIP provide quick-look data that are qualitatively useful, but accurate quantitative data requires specialized processing of the individual particle images. The System for OAP Data Analysis version 2 (SODA-2, https://github.com/abansemer/soda2) is used to process the images to produce drop size distributions from both instruments. For the PIP, accurate sizing of particles up to 30 mm is possible in post-processing. However, particles larger than approximately 6 mm are typically ice particles; any such particles in the data should be ignored.

230
An aerosol size distribution is inferred from measurements the size range from 0.5 to 2 µm collected by the CAS. (Larger particles are included in the hydrometeor size distribution.) The division is motivated by size distributions observed during ATOMIC, which usually included two modes with the minimum between them typically close to 2 µm. Aerosol concentrations collected in-cloud are higher than expected. This may be a result of cloud drop shattering at the CAS inlet, so these measurements should be treated with caution. We do not anticipate that we can correct this potential issue due to corruption of the CAS 235 particle-by-particle files during the campaign. Despite this small temporal mismatch the thermal structures observed by the radio-and drop-sondes are similar, with a small inversion near 5 km and a larger inversion near 2.5 km, the height of which varies across and around the circle. Large jumps 260 in the moisture field associated with these inversions exhibit similar vertical variability.

AXBTs
Following the processing of dropsonde data for JOANNE we have produced a single file containing all AXBTs profiles obtained during the ATOMIC, interpolated to a standard depth grid at 0.1 m vertical resolution. Figure 9 shows an example from the flight on Jan 19 2020 in which 40 AXBTs were deployed in a lawnmower pattern bracketing five Surface Wave Instrument

265
Floats with Tracking (SWIFT buoys, see Thomson, 2012) deployed from the R/V Ronald H. Brown (see the upper right corner of Fig. 5). Figure 9 shows the ocean temperature as measured by the AXBTs between the surface and 150 m depth, with a near-isothermal mixed layer extending tens of meters and the cooler ocean below 60-80 m. The inset compares the temperature in the first three meters with measurements made by the SWIFT buoys (see Quinn et al., 2021), N RCS m = dBZ e (0) + 10 log 10 (π 5 |K 2 |δR/λ 4 ) − 180 + dBZ attn (1) For the PSL W-band with its 30 m range resolution, the second term on the right-hand side of (1) is 137.9, while the correction factor for attenuation by water vapor and oxygen is roughly dBZ attn = 4 for typical ATOMIC conditions with the aircraft at 3 285 km altitude. During ATOMIC the signal from the surface was strong enough to cause some saturation of the receiver, reducing the sensitivity the nearer the aircraft was to the surface. Values of N RCS m have been further adjusted for this effect as a function of pressure; the correction is small for altitudes higher than 5 km but is as large as 8 dB at 1 km above sea level.
The back-scattered radar return from the ocean surface σ depends on both wind speed and viewing angle θ; this dependence can be exploited to estimate the mean square slope s 2 of surface waves (satellite-borne radar scatterometer wind estimates 290 exploit the same physics). The dependence is usually represented (Walsh et al., 1998;Li et al., 2005) as where Γ 2 is a wavelength-dependent constant with value 0.32 at W-band radar frequencies and the theoretical or calculated normalized radar cross section N RCS c = 10 log 10 (σ). We solve (2) for s 2 , using observations made a nadir viewing angles (θ = 0) and assuming N RCS c = N RCS m . These estimates rely on radar calibration.

295
Following Fairall et al. (2018) precipitation rate is determined from the W-band radar using the vertical gradient of radar reflectivity during light rain and the path-integrated attenuation during the infrequent heavy rain observed during ATOMIC.
Clouds are detectable in both the radar reflectivity profile and the observed infrared brightness temperature. A radar cloud presence index C radar is determined by examining the maximum signal-to-noise ratio (SNR) within each radar column (excluding the surface return). The clear-sky signal-to-noise level of the radar is nominally -20 dB but a value of about -15 is needed to 300 ensure a valid cloud return. We define the index as C radar = max(SNR) + 14 so that values of C radar > 0 indicate clouds. When clouds are detected cloud-top height z ct is estimated as the level closest to the aircraft at which SNR > 14. Cloud-top height diagnosed in this way is shown in the middle panel of Fig. 6. We report the radar reflectivity factor and Doppler velocity at this height as well as the wind speed w(z ct ) and air temperature T (z ct ) as determined from daily-mean in situ aircraft profiles (not dropsondes).

305
An infrared cloud presence index C IR is produced based on the observed nadir-looking brightness temperature T IR (p) made at aircraft operating pressure p. We compare clear-sky measurements of T IR (p) to the near-surface radiometric temperature, determined from the time-mean of infrared radiometer measurements during flight legs at 150 m, to develop a correction term ∆T IR (p − p ) = SST − T IR clear (p − p sfc ) as a quadratic function of p − p . The infrared cloud index C IR is defined as the difference between the observed infrared temperature and the value expected in the absence of clouds, i.e. C IR = T IR (p) − 310 (SST − ∆T IR (p − p sfc )), so that values of C IR < 0 indicate clouds.
The two cloud indexes complement one another. The W-band radar sensitivity is limited and, particularly when the aircraft was transiting or dropping sondes at 7.5 km altitude (about a quarter of the total flight time), many clouds near the surface were beyond the viewing range from the radar are are therefore not detectable in the radar return. For these flight legs the IR cloud index is likely a better indicator of cloudiness. When the aircraft is at or below about 3 km, the radar is very sensitive 315 to clouds and likely detects all clouds with radar reflectivity factor Z e > −35dBZ. Under these circumstances the radar and infrared cloud indexes are quite consistent with one another, as shown for an example hour of observations on 19 January 2020 in Fig. 10.
When cloud top height z ct is available from the radar we use this information to identify cloud-top pressure p ct and, from aircraft soundings, the temperature at that pressure T air ct . This can be compared to IR cloud top temperature corrected for the 320 intervening atmosphere T IR ct = T IR (p)+∆T IR (p−p ct ). Values of T IR ct ≈ T air ct indicate that optically-thick clouds fill the infrared radiometer's field of view. In Fig. 11 we show the difference in cloud-top air temperature and apparent IR cloud-top temperature as a function of W-band reflectivity from January 19. Values of T IR ct − T air ct less than zero indicate the cloud is optically thin or does not completely fill the 0.5 s sample and some radiation from the warm sea surface is adding to the measured IR.

325
ProSensing Inc. processes raw data from the Wide-Swath Radar Altimeter (WSRA, see sec. 3.3.2) to produce information about the wave state of the ocean surface. Most observations, including the power spectrum of surface waves as a function of wavenumber in the north-south and east-west directions; the direction, height, and wavelength of the two most dominant waves; peak spectral variance; and the significant wave height, are reported every 50 seconds. A plan view of observations obtained on 19 Jan 2020 is shown in Fig. 12. Rainfall rate and surface wave s 2 are reported every ten seconds. The latter is 330 computed from the decrease of the intensity of the return with scan angle following (2) so, unlike estimate of s 2 from the W-band radar, estimates from the WSRA do not depend on absolute calibration.
Files with these estimates also contain bookkeeping information (processing parameters and ancillary data) such as aircraft navigation and orientation and other fields that may be useful. In particular, the directional wave spectra calculated from data collected with WSRA inherently contain a 180°ambiguity of the wave propagation which can generally be eliminated using 335 a rough estimate of a predicted dominant ocean wave direction at the location of the observation point. During ATOMIC the prevailing wind direction (typically ENE to WSW) was used as the predicted ocean wave direction for the entire duration of each flight mission. For completeness WSRA files contain directional wave spectra with and without this ambiguity removed.
During ATOMIC the aircraft operated in a number of modes that were unfavorable for collecting WSRA data. Data should not be used if the aircraft altitude is less than 500 m or greater than 4000 m, or when the aircraft's pitch or roll exceeds ±3°. 340 We also recommend using observations only when the peak spectral value is in the range 0.0002 − 0.006 m 2 .

Data availability
Data have been reformatted into netCDF files following CF (Climate and Forecast) conventions (https://cfconventions.org) which provide for units and standard names for variables, a uniform handling of time, and other metadata intended to promote interoperability and interpretability. The files also contain some provenance information following guidance developed for 345 EUREC 4 A. Isotope ratios from the isotope analyzer aboard the P-3 (sec. 3.1.2) will described as part of a paper describing the aircraft-, ship-. and ground-based observations made during the experiments.
Data have been archived at NOAA's National Center for Environmental Information as detailed in Table 4. This represents the version of record. The data are also replicated at the French AERIS data center alongside the wide array of other data from the EUREC 4 A experiment (OpenDAP access via https://observations.ipsl.fr/thredds/catalog/EUREC4A/catalog.html).

350
Code used to generate the figures in this paper is available at http://github.com/RobertPincus/atomic-p3-data-paper/ (doi to be provided on publication). produced the microphysical size distributions. IP obtained and post-processed data from the WSRA and is the point of contact for data 355 from this instrument. GB operated assisted with data processing and conversion to NetCDF. RP coordinated data archiving activities and the production of this manuscript. All authors contributed to the collection and/or processing of the data described.
Competing interests. Author GB is a guest editor for the special issue to which this manuscript is submitted. The authors declare that they have no other conflicts of interest.
Acknowledgements. We are grateful to the pilots, crew, and ground support staff of the WP-3D for their help in obtaining these measurements.

360
The ATOMIC field campaign was supported by the NOAA Climate Program Office under the Climate Variability and Predictability Program; collection of data from the P-3 was supported by award GC19-302a. Additional support for processing and collection of ATOMIC WP-3D data sets was provided by the NOAA Physical Sciences Laboratory. Measurements with the isotope analyzer were supported by the National        Relative humidity (fraction) Figure 8. Profiles of air temperature (left) and relative humidity with respect to liquid (right) as obtained by dropsondes deployed from the P-3 (grey) and a radiosonde launched from the Ron Brown (dark red). Dropsonde data are obtained from JOANNE (see text); radiosonde observations are obtained from the data set described in Stephan et al. (2020). One of the twelve sondes shows much lower humidity from 1.5 -2.5 km than do the others; we are assessing the all the dropsonde circles to understand if this is common or an instrumental artifact. Depth (m) Figure 9. Ocean temperature profiles as measured by AXBTs deployed from the P-3 on Jan 19 2020. Data are shown between the nearsurface and a depth of 150 m though the actual profiles extend to nearly 1000 meter depth. The inset shows ocean temperatures in the first few meters along with measurements from the five SWIFT buoys (see Quinn et al., 2021) surrounded by the AXBT deployments. Three of the forty AXBTs deployed did not provide valid data.  Figure 10. Cloud indexes based on radar (C radar , green) and infrared radiometer (C IR , purple) measurements for the period 18-19 UTC on 19 January (c.f. Fig. 6). Cloud is indicated by values of C radar > 0 and C IR < 0. Absolute values less than one have been removed for clarity.
The two indexes are quite consistent with one another because clouds, when present, are typically opaque enough to be easily detectable in in both infrared and microwave measurements. Air minus IR temperature difference (K) Figure 11. Difference between air temperature at cloud top T air ct and the observed IR temperature T IR ct as a function of W-band radar reflectivity at cloud top for the entire flight made on 19 Jan 2020. The height of the cloud top is determined as the closest position to the observing aircraft at which the radar signal-to-noise exceeds 14 dB; air temperature as a function of height is determined from in situ samples made by the aircraft. Temperature differences near zero indicate that the clouds are optically thick in the infrared. Mean square slope (rad 2 ). Bottom: Significant wave height (m).