AirHARP is a wide field-of-view imaging polarimeter designed for characterization of cloud and aerosol optical properties. AirHARP is an amplitude-splitting polarimeter: light entering the front lens is decomposed into three orthogonal linear polarization states (0, 45, and 90∘) by a modified three-way Phillips prism. Each polarization state is imaged by a unique detector array, so the first three parameters of the Stokes vector (I, Q, and U) are retrieved at pixel level by combining co-located information from all three detectors. Stripe filters on the detectors define 120 distinct along-track viewing angles across the four HARP wavelengths (spectral widths): 440 (14), 550 (12), 670 (18), and 870 (38) nm, across a total swath of ±57∘ (±47∘) along track (cross track). A total of 60 of these angles are at 670 nm, specifically designed for studying the structure of the polarized cloudbow. The high angular resolution (roughly 2∘) enables AirHARP to measure the cloudbow structure at all individual pixels. The remaining three wavelengths have 20 angles each for characterization of aerosols. Figure 4a shows an illustration of the sampling scheme applied by the HARP polarimeters, where each along-track viewing angle of the instrument produces a full push-broom image. The figure also shows an example of data collected during the ACEPOL campaign, its second deployment. The instrument's maiden field campaign was the Lake Michigan Ozone Study (LMOS) in 2017 (McBride et al., 2020). For ACEPOL, AirHARP collected data over defined targets, as the processing speed of the onboard data acquisition system precluded continuous data collection.

AirHARP is an aircraft demonstration for the HARP CubeSat instrument, a stand-alone satellite that was successfully launched on 19 February 2020 from the International Space Station (ISS) conducting Earth observations for a year-long mission. A third member of the HARP family is the HARP2 sensor, which will provide global coverage in 2 d, as part of the NASA PACE mission due to launch no earlier than 2022 (Werdell et al., 2019).

AirHARP Level 0 ACEPOL data are corrected, reconstructed, geolocated, and calibrated in the Hyper-Angular Image Processing Pipeline (HIPP) Level 1B algorithm. These data products are gridded to a horizontal resolution of 2000 pixels per latitude degree and packaged into HDF5 files for distribution in the ACEPOL data archive. A Level 2 aerosol retrieval algorithm has been implemented using the Generalized Retrieval for Aerosol and Surface Properties (GRASP) scheme (Dubovik et al., 2011; Puthukkudy et al., 2020). Figure 4d shows the AOD map retrieved using AirHARP measurements and the GRASP inversion algorithm for a smoke scene in Fig. 4c. The magnitude and spatial variability of retrieved AirHARP AOD are in good agreement with collocated lidar (i.e., HSRL-2) observations (Puthukkudy et al., 2020) and are also consistent with the AOD retrieved from AirSPEX and RSP instruments (Fu et al., 2020). Cloud droplet size distribution retrievals are implemented using a traditional parametric fit to Mie scattering curves for liquid water droplets (McBride et al., 2020). Retrievals of ice cloud characteristics, aerosols above clouds, ocean and surface reflectance properties, and atmospheric correction are also important potential uses of HARP data.

The Airborne Multiangle SpectroPolarimetric Imager (AirMSPI; Diner et al., 2013b) is push-broom imaging camera used for the characterization of atmospheric aerosols and clouds. In addition to radiometric channels, AirMSPI employs photoelastic modulators (PEMs) to enable accurate measurements of the degree and angle of linear polarization (Diner et al., 2010; van Harten et al., 2018). These polarimetric data help discriminate between different aerosol particle types, which is crucial to improving our understanding on climate and air quality. The instrument flies aboard NASA's high-altitude ER-2 aircraft and acquires Earth imagery with ∼10 m spatial resolution across a 9 km wide swath. Radiance data are obtained at 355, 380, 445, 470, 555, 660, 865, and 935 nm. The polarimetric channels at 470, 660, and 865 nm report both radiances and the linear polarization Stokes components Q and U. AirMSPI is a precursor to the future Multi-Angle Imager for Aerosols (MAIA) satellite instrument (see Sect. 5.3), which will be used to improve our understanding of the health risks associated with airborne particulate matter.

The AirMSPI camera is mounted on a motorized single-axis gimbal to enable multiple views of a science target from different along-track view angles. The number of views and specific set of angles for each observing sequence are programmable prior to flight. While approaching the target in a straight and level flight line, the pilot presses one of three buttons to start the corresponding multi-angle acquisition sequence. Figure 5 shows example imagery of the sequences used during the ACEPOL campaign. All three modes begin and end with views of the onboard dark target and polarization validator light source.

AirMSPI observed a total of 24 targets in the nine-angle step-and-pseudostare mode, 32 targets in the 15-angle mode, and 209 continuous-sweep images. AirMSPI L1B2 geolocated radiometric and polarimetric data and quick looks are available as noted in the Data availability section. For a further analysis and comparison with other MAPs and lidars during ACEPOL, see Fu et al. (2020).

A pair of RSP instruments (denoted RSP1 and RSP2; the latter was used in ACEPOL) have been deployed on more than 25 field missions in the last 20 years. The RSP is an airborne multi-angle and multispectral polarimeter that continuously scans in the aircraft along-track direction. During a complete scanner rotation, a total of 205 samples, each with a 0.8∘ (14 mrad) field of view, are collected. The 205 samples include 152 views between 60∘ forward and aft of the normal to its baseplate, 10 samples viewing an internal dark reference, and 43 samples through an Earth-viewing polarization scrambler. Samples are obtained by a rotating polarization-compensated mirror assembly with six bore-sighted telescopes, to simultaneously obtain nine Stokes parameter spectral bands. Each telescope uses a Wollaston prism to split the incoming intensity into two spatially separated orthogonally polarized components, which are then further split and passed through dichroic filters, defined by the spectral bandpass: (full width at half maximum bandwidths in parentheses) 410.3 (30), 469.1 (20), 555.0 (20), 670.0 (20), 863.5 (20), 960.0 (20), 1593.5 (60), 1880.0 (90), and 2263.5 (120) nm. On the ER-2, the un-vignetted viewing angle range is from 65∘ aft to 45∘ forward, for a total of 120 samples. The single nadir ground pixel size from an altitude of 20 km is 280 m, and successive pixels partially overlap.

RSP2 was calibrated in the Airborne Sensor Facility at NASA Ames Research Center before and after ACEPOL. Radiometric calibration was stable to within roughly 1 % for all bands, except for 410.3 nm, where a 4 % decrease in radiometric throughput was observed (polarimetric calibration was stable to within ∼0.1 %). Data processing to level 1b (calibrated and geolocated at-sensor measurements) consists of dark subtraction for each scan and application of calibration coefficients to generate Stokes parameters, together with the geolocation of each viewing angle sample. Processing from level 1b to level 1c consists of collocating the different viewing angles about a defined altitude, either at the ground or at cloud-top if a cloud is present. The latter requires cloud identification and a cloud top height estimate (Sinclair et al., 2017). Level 2 processing of cloudy scenes is split between water and ice clouds. Water cloud retrievals include cloud optical depth and standard bi-spectral droplet size estimates (Platnick et al., 2016), as well as both parametric (Alexandrov et al., 2012b) and nonparametric (Alexandrov et al., 2012a) droplet size distribution estimates that use polarization (Alexandrov et al., 2018). Most clouds observed during ACEPOL were low-level water clouds, but when ice clouds were detected, cloud optical depth, particle size, and particle shape/roughness retrievals (Van Diedenhoven et al., 2012, 2013; Van Diedenhoven, 2018) were retrieved. Level 2 processing for aerosol retrievals uses the Microphysical Aerosol Properties from Polarimetry (MAPP) algorithm (Stamnes et al., 2018). The MAPP land surface model consists of a Fresnel reflectance with shadowing for the polarized reflectance (Waquet et al., 2009) together with a RossThick (Ross, 1981; Roujean et al., 1992) LiSparse (Li and Strahler, 1992) kernel model (Wanner et al., 1995), similar to that used for operational processing of MODIS land surface products (Schaaf et al., 2002). RSP was operational for all ACEPOL flights, and data are available at the ACEPOL data archive (see Data availability section). Example RSP observations of a liquid phase cloud are shown in Fig. 6.

The SPEX airborne instrument (Smit et al., 2019) employs the spectral modulation technique (Snik et al., 2009), in which the degree and angle of linear polarization are encoded in a modulation of the radiance spectrum as a function of wavelength. This modulation is achieved by placing a set of dedicated optical components (quarter wave retarder, multiple order retarder) in front of the telescope. The resulting two light beams contain a modulation pattern as a function of wavelength (out of phase with each other) that enter a spectrometer and are focused on a detector. Radiance measurements are obtained between 400 and 800 nm at the spectral resolution of the spectrometer (∼2 nm), while for DoLP the spectral resolution is determined by the modulation period, which is conservatively estimated at 10 nm at 400 nm and 40 nm at 800 nm.

An example of a SPEX airborne radiance and DoLP measurement obtained during ACEPOL is shown in Fig. 7.

SPEX airborne has nine view ports that are projected on one detector, at angles (±57, ±42, ±28, ±14, and 0∘). The SPEX airborne cross-track swath is 6∘. The data processing from level 0 (detector counts) to level 1b (calibrated and geolocated radiance and DoLP values) consists of the following steps: dark image subtraction, spectral extraction and line-of-sight annotation, wavelength annotation, spectral alignment of the two modulated spectra, radiometric correction, demodulation, and geolocation. Next, data of the different view ports are all aggregated on the same spatial grid, yielding the level 1C data product, currently computed at 1 km × 1 km spatial resolution.

For level 2 processing (Fu et al., 2020), the SPEX airborne team has focused on aerosol retrievals, building further on the SRON aerosol retrieval algorithm previously used for POLDER-3 processing (Hasekamp et al., 2011; Lacagnina et al., 2015, 2017) and groundSPEX (van Harten et al., 2014; di Noia et al., 2015). The algorithm has been extended from a bimodal retrieval scheme to multimode retrieval scheme for an arbitrary number of modes (Fu and Hasekamp, 2018). For processing ACEPOL data, a setup with five modes has been used, where for each mode the aerosol column number is retrieved corresponding to a set of fine- and coarse-mode spectrally varying refractive indices, the fraction of spherical particles (Dubovik et al., 2006), and the aerosol layer height. SPEX airborne data have been processed using 16 discrete wavelengths between 450 and 750 nm. Wavelengths less than 450 nm and greater than 750 nm are excluded because of lower data quality. SPEX airborne was operational for all ACEPOL flights, and data are available at the ACEPOL data archive.

The Cloud Physics Lidar (CPL) is a nadir-pointing, multi-wavelength (355, 532, 1064 nm) elastic backscatter lidar that measures vertical profiles of cloud and aerosol properties (McGill et al., 2002). CPL has participated in numerous field campaigns since its first deployment in 2000 on the ER-2, often in tandem with the other remote sensing instruments involved in ACEPOL. Raw CPL data are calibrated by normalizing the 355, 532, and 1064 nm signals to the Rayleigh (molecular) backscatter between 15 and 18 km, creating vertical profiles of the total attenuated backscatter (McGill et al., 2007). CPL data products retrieved from the calibrated backscatter include vertical profiles of depolarization ratio, particulate backscatter coefficient, and particulate extinction coefficients, as well as layer properties such as top and base altitudes, aerosol type, and optical depth (Yorks et al., 2011; Hlavka et al., 2012).

CPL data products, with vertical resolution of 30 m and 1 s frequencies (∼200 m horizontal resolution), enable a wide range of applications including the analysis of cloud properties (McGill et al., 2004; Bucholtz et al., 2010; Yorks et al., 2011; Alexandrov et al., 2015), as well as aerosol and dust properties (McGill et al., 2003; Nowottnick et al., 2011; Wu et al., 2016). CPL data from ACEPOL are fully processed and data products are available at the ACEPOL data repository. However, an electronics controller failed in the CPL laser during the second flight of the campaign. The cause of this failure was not immediately apparent, and there was no time to take the instrument out of service during the ACEPOL campaign. Thus, the ACEPOL CPL data products after the first flight are nonstandard, as they required more averaging, sometimes as much as 10 s (2 km horizontal), and removal of some data at specific wavelengths due to poor quality. Nevertheless, data are useful for validation of CALIPSO (McGill et al., 2007; Yorks et al., 2011; Hlavka et al., 2012) and CATS data (Yorks et al., 2016; Pauly et al., 2019.)

HSRL-2 is the second-generation airborne High Spectral Resolution Lidar developed at NASA Langley Research Center. Like the first-generation HSRL-1 (Hair et al., 2008), HSRL-2 independently measures aerosol backscatter and extinction (532 nm) using the HSRL technique, aerosol backscatter (1064 nm) using the standard backscatter technique (Müller et al., 2014; Sawamura et al., 2017), and aerosol depolarization at both wavelengths. HSRL-2 adds measurements of aerosol backscatter and extinction at 355 nm using the HSRL technique as well as measurements of aerosol depolarization at 355 nm. These HSRL-2 measurements are used to compute AOD as well as the aerosol extinction / backscatter ratio (“lidar ratio”) at 355 and 532 nm. Rogers et al. (2009) evaluated the HSRL extinction coefficient profiles and found that the HSRL extinction profiles are within the systematic errors of airborne in situ measurements at visible wavelengths. Derived products include, first, estimates of planetary boundary layer heights which use the vertically resolved profile measurements of aerosol backscatter to derive aerosol mixed-layer heights during the daytime (Scarino et al., 2014). Second, aerosol classification uses an algorithm to interpret the information about aerosol physical properties indicated by the HSRL-2 aerosol intensive parameters to qualitatively infer aerosol type (Burton et al., 2012). HSRL-2 measurements have also been used to retrieve height-resolved parameters such as aerosol effective radius and concentrations (number, surface, volume) (Müller et al., 2014; Sawamura et al., 2017).

HSRL-2 has been deployed on various NASA aircraft for both NASA and DOE field missions (Berg et al., 2016; Sawamura et al., 2017; Burton et al., 2018). During ACEPOL, HSRL-2 operated autonomously from the NASA ER-2 aircraft. When operated from the ER-2, the nominal HSRL-2 aerosol backscatter profiles are reported at a vertical resolution of 15 m and a horizontal (temporal) resolution of 1–2 km (10 s). Aerosol depolarization ratios at all three wavelengths are reported at the same resolutions. For ACEPOL, aerosol extinction, AOD, and lidar ratio from the HSRL methodology are not available in some cases, particularly when the aerosol loading was small. In these cases, the aerosol extinction at 355 and 532 nm is derived using the aerosol backscatter and an assumed lidar ratio of 40 sr and reported at the backscatter resolution. In other cases where the HSRL method is available for extinction products, they are reported at 150 m vertical resolution and at temporal resolution of 60 s generally and 10 s within smoke plumes. Calibrated aerosol backscatter derived using the HSRL method is available in all cases. Problems with the ER-2 coolenol pump and a circuit breaker caused HSRL-2 data gaps on the 27 October flight and loss of HSRL-2 data on the 1 November flight.

The Aerosol Robotic Network (AERONET; Holben et al., 1998) is a system of sun photometers used to produce global measurements of AOD, intensive properties (e.g., refractive index, size distribution), and precipitable water. This NASA product provides data recorded every 15 min utilizing many spectral bands, and the data are available at the AERONET archive (https://aeronet.gsfc.nasa.gov/, last access: 3 September 2020). During ACEPOL, the Modesto, Fresno-2, CalTech, and Bakersfield sites in California were operational and were routinely targeted. Additional AERONET sites that were targeted during the campaign are the USGS Flagstaff site in Arizona and the USC SeaPRISM site. The latter is part of a sub-network designated AERONET-Ocean Color (AERONET-OC). This network comprises enhanced instruments that have the ability to determine water-leaving radiance in addition to aerosol optical properties (Zibordi et al., 2009).

Air particulates are measured by the California Air Resources Board (CARB) throughout the state. A majority of CARB sites report PM2.5 (particulate matter with an aerodynamic diameter of less than 2.5 µm) and PM10 (particulate matter with a diameter of less than 10 µm), as determined by a variety of measurements, including particle counting and mass spectrometry. Gaseous criteria for air pollutants that are regulated in California and at the national level are also measured and reported to ensure compliance with U.S. Environmental Protection Agency (U.S. EPA) air quality standards and are stored at the CARB archive (https://www.arb.ca.gov/adam/index.html, last access: 3 September 2020).

In addition to CARB total PM monitors, several speciated EPA PM monitors are located in California's San Joaquin Valley (SJV) including Modesto, Fresno, Visalia, and Bakersfield. The speciated PM monitors in Modesto, Fresno, and Bakersfield are collocated with AERONET instruments making these sites very useful for determining connections between near-surface PM loading and atmospheric aerosols. Modesto, Fresno, and Bakersfield sites were regularly targeted throughout the campaign.

Vicarious calibration (VicCal) is the process of calibrating an on-orbit or aircraft sensor by observation of an Earth target which is characterized for its optical properties. A radiative transfer program is then used to propagate the surface measurements into a radiance incident on the flight sensor. Ground measurements made in support of ACEPOL vicarious calibration were performed on 25 October 2017, at Rosamond Dry Lake (34.858704∘ N, 118.07638∘ W). This included measurements of surface reflectance and total column optical depth at the time of overflight, roughly 18:00 UTC. Data were processed at JPL and are hosted in the ACEPOL archive. These data were used by the AirMSPI team to derive a multiplier needed to adjust the laboratory-derived radiometric gain coefficient. For the Rosamond VicCal, the ratio of radiances using the laboratory calibration to the VicCal-measured radiances was found to be low for the UV bands (0.85) but within 2 % of unity for the remaining visible bands. The discrepancy in the UV lands is under investigation, but one error candidate is the low light level for these bands during the laboratory calibration.

The VicCal campaign at Rosamond included measurements of surface reflectance and total column optical depth at the time of overflight, roughly 18:00 UTC. Results of these measurements are shown in Figs. 8 and 9. AOD was measured by two Microtops and a Reagan sun photometer. The Microtops are handheld instruments, with spectral bands at 440, 675, 870, 938, and 1020 nm. The Reagan sun photometer (Bruegge et al., 1992; Shaw et al., 1973), built at the University of Arizona, is an auto-tracking sun photometer, with channels at 370, 400, 440, 520, 620, 670, 780, 870, 940, and 1030 nm. The instrument was calibrated via the Langley method by taking early morning data the day before the ER-2 overflight. A discrepancy is noted in the Microtops data, perhaps due to an error in its surface pressure. As the Reagan instrument was calibrated in situ, its output was considered higher in accuracy and used in the AirMSPI VicCal. Also shown in Fig. 8 is the Junge model fit to the Microtops data, which assumes a linear relationship on a log-log scale of aerosol optical depth versus wavelength. Surface reflectance was measured by the Analytical Spectral Devices (ASD) FieldSpec 4 instrument. This measures from 350 to 2500 nm with 1 nm samples with resolution at 3 nm (VIS, NIR) and 10 nm (SWIR). Data were taken over a 500 m× 500 m area, along with spectra of a Spectralon 100 % diffuse reflectance standard. To report radiances, the Spectralon data are interpolated at the target sample times and corrected for Spectralon bidirectional reflectance. The ratio of the target to interpolated Spectralon data provides a per-sample reflectance, which is then averaged to provide a site average.

Operating alongside the vicarious calibration instrumentation was the Ground-based Multiangle SpectroPolarimetric Imager (GroundMSPI) instrument (see Fig. 8). The specifications of the GroundMSPI camera are similar to AirMSPI (see Sect. 3.1.2), including the eight spectral bands within 355–935 nm, with 470, 660, and 865 nm being polarimetric. The camera is mounted ∼1 m from the ground on a motorized altazimuth tripod. GroundMSPI performed continuous elevation scans at different azimuths to image the surface as well as the clear sky. The surface observations provide a direct measurement of the polarized bidirectional reflectance distribution function (p-BRDF), including potential spectral variance, whereas the sky radiance and polarization data serve as input for aerosol retrievals, either independently or combined with the airborne sensors. On 7 November 2017, GroundMSPI performed sky scans at the Fresno AERONET station during ER2 and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) overpasses. GroundMSPI L1B2 rectified and co-registered radiometric and polarimetric data, and quick looks are available as noted in the Data availability section.