Atmos. Meas. Tech., 13, 4947–4961, 2020 https://doi.org/10.5194/amt-13-4947-2020 © Author(s) 2020. This work is distributed under the Creative Commons Attribution 4.0 License. A new Orbiting Carbon Observatory 2 cloud flagging method and rapid retrieval of marine boundary layer cloud properties Mark Richardson1,2, Matthew D. Lebsock1, James McDuffie1, and Graeme L. Stephens1,2,3 1Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA 2Department of Atmospheric Science, Colorado State University, Fort Collins, CO 90095, USA 3Department of Meteorology, University of Reading, Reading, RG6 7BE, UK Correspondence: Mark Richardson ([email protected]) Received: 8 April 2020 – Discussion started: 18 May 2020 Revised: 16 July 2020 – Accepted: 28 July 2020 – Published: 18 September 2020 Abstract. The Orbiting Carbon Observatory 2 (OCO-2) car- 1 Introduction ries a hyperspectral A-band sensor that can obtain informa- tion about cloud geometric thickness (H). The OCO2CLD- LIDAR-AUX product retrieved H with the aid of collocated Hyperspectral O2 A-band measurements near λ D 0:78 µm, CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder such as those taken by the Orbiting Carbon Observatory-2 Satellite Observation) lidar data to identify suitable clouds (OCO-2), may provide unique new information about bound- H and provide a priori cloud top pressure (Ptop). This colloca- ary layer clouds by retrieving their geometric thickness ( ) tion is no longer possible, since CALIPSO’s coordination fly- or droplet number concentration (Nd), provided coincident ing with OCO-2 has ended, so here we introduce a new cloud information about effective radius (re) from other channels. flagging and a priori assignment using only OCO-2 data, re- They are able to do this because the spectrum responds stricted to ocean footprints where solar zenith angle < 45◦. to the photon path length between the Sun, Earth and the Firstly, a multi-layer perceptron network was trained to iden- sensor. Increased H or decreased Nd with all other cloud tify liquid clouds over the ocean with sufficient optical depth properties held constant leads to increased distance between (τ > 1) for a valid retrieval, and agreement with MODIS– within-cloud scattering events and therefore a longer pho- CALIPSO (Moderate Resolution Imaging Spectroradiome- ton path length and decreased transmittance in wavelengths ter) is 90.0 %. Secondly, we developed a lookup table to si- where O2 absorbs. This leads to spectrally varying changes multaneously retrieve cloud τ, effective radius (re) and Ptop in observed A-band spectra that can allow for joint retrievals τ P from A-band and CO2 band radiances, with the intention of cloud optical depth ( ), cloud top pressure ( top) and that these will act as the a priori state estimate in a future H, provided there is sufficient spectral resolution and low retrieval. Median Ptop difference vs. CALIPSO is 12 hPa enough noise (O’Brien and Mitchell, 1992; Richardson and with an inter-decile range of [−11;87UhPa, substantially Stephens, 2018). better than the MODIS–CALIPSO range of [−83;81UhPa. The basic principle of A-band absorption for cloud height The MODIS–OCO-2 τ difference is 0:8[−3:8;6:9U, and re is well established (Fischer and Grassl, 1991; Rozanov and is −0:3[−2:8;2:1Uµm. The τ difference is due to optically Kokhanovsky, 2004; Yamamoto and Wark, 1961), and nu- thick and horizontally heterogeneous cloud scenes. As well merous spaceborne A-band instruments retrieve cloud prop- as an improved passive Ptop retrieval, this a priori infor- erties (Koelemeijer et al., 2001; Kokhanovsky et al., 2005; mation will allow for a purely OCO-2-based Bayesian re- Lindstrot et al., 2006; Loyola et al., 2018; Preusker et al., trieval of cloud droplet number concentration (Nd). Finally, 2007; Vanbauce et al., 1998), but most lack the spectral reso- our cloud flagging procedure may also be useful for future lution or noise characteristics to obtain H (e.g. Schuessler et partial-column above-cloud CO2 abundance retrievals. al., 2014). Others rely on multi-angle (Ferlay et al., 2010) or combined A- and B-band information (Yang et al., 2013), al- though these tend to contain little information on low-altitude Published by Copernicus Publications on behalf of the European Geosciences Union. 4948 M. Richardson et al.: Rapid OCO-2 cloud flagging and retrievals and relatively thin clouds like marine stratocumulus (Davis et retrievals used a fixed re, and the addition of varying re is al., 2018; Merlin et al., 2016). eased by a new Python RT interface using the ReFRACtor An OCO-2-based retrieval of τ, Ptop and H has (Reusable Framework for Retrieval of Atmospheric Com- been developed (OCO2CLD-LIDAR-AUX, available position) software described in Sect. 2.3. Our new LUT re- at http://www.cloudsat.cira.colostate.edu/data-products/ trieval of a prior re will allow for a more appropriate re to be level-aux/oco2cld-lidar-aux, last access: 1 September assumed in the iterative OE. 2020), which uses lidar-based retrievals from the Cloud- The paper is organised as follows: Sect. 2 describes the Aerosol Lidar and Infrared Pathfinder Satellite Observation relevant OCO-2 details, data selection and radiative-transfer (CALIPSO) satellite to help identify cloudy scenes and calculations before detailing the methodology. Section 3 re- constrain prior Ptop (Richardson et al., 2019). This retrieval ports the performance statistics of the classifier, compares is targeted at single-layer liquid clouds over the ocean LUT retrieved cloud properties vs. MODIS and CALIPSO whose response, both to warming and aerosols, is a major where the instrument footprints overlap, and compares the source of uncertainty in climate simulations (e.g. Bony et final pre-processor throughput against that of OCO2CLD- al., 2005; Bodas-Salcedo et al., 2019; Zelinka et al., 2020). LIDAR-AUX. Section 4 discusses and contextualises the re- Independent information about cloud structure may help to sults and proposes actionable future work that could address address timely questions, where other sensors which rely identified biases, and Sect. 5 concludes. on different retrieval approaches and assumptions can lead to apparently contradictory conclusions (Rosenfeld et al., 2019; Toll et al., 2019). 2 Methods and data With CALIPSO leaving the A-Train constellation in 2018, collocation between OCO-2 and CALIPSO footprints is no 2.1 Instruments and data selection longer possible. Our future retrievals require a new cloud flagging method plus a priori cloud top information for our it- The OCO-2 measurement approach and instrumentation are erative Bayesian optimal-estimation (OE) retrieval (Rodgers, detailed in Bösch et al. (2017); the Level 2 Full Physics 2000). This paper describes a new pre-processor for OCO- (L2FP) RT’s application to clouds is detailed in Richardson 2-based liquid cloud property retrievals that provides the et al. (2017); and the MODIS–CALIPSO–OCO-2 matchup requisite cloud flagging and a priori information. Details of data are as used in Taylor et al. (2016). The datasets used OCO2CLD-LIDAR-AUX are summarised in Table 1, which here are listed in Table 2; in particular, from the OCO-2 Level also lists the main changes introduced in this study. 1b Science (L1bSc) data we obtain calibrated radiances and We do not use the published OCO-2 cloud flag, as it RT inputs such as solar zenith angle (SZA) and instrument was not developed for ocean nadir scenes (Taylor et al., characteristics. 2016), since they were considered too dark for OCO-2’s The OCO-2 satellite flies in the Sun-synchronous A-Train main mission of column CO2 (XCO2) retrievals (Crisp, constellation (L’Ecuyer and Jiang, 2010) and measures dur- 2008; Crisp et al., 2004; Eldering et al., 2016). Therefore ing the daytime ascending node with an Equator crossing we train a multi-layer perceptron network to rapidly identify time near 13:30. Its orbits are committed primarily to either liquid cloud scenes using collocated CALIPSO and Mod- glint or nadir view, and we use nadir-only orbits to provide erate Resolution Imaging Spectroradiometer (MODIS) re- complementary vertical information on clouds that are too trievals. For the prior cloud property retrieval we develop low or thin to be adequately profiled by CloudSat’s nadir- lookup tables (LUTs) that jointly retrieve τ, re and Ptop using view radar. Glint-view footprints would preclude our use of OCO-2 O2 A-band and strong-CO2-band (λ ∼ 2:06µm) radi- the nadir-only CALIPSO lidar data, and atmospheric photon ances. These are similar to the Nakajima–King tables used in path lengths would be longer, thereby reducing the retrieval MODIS cloud retrievals (Nakajima and King, 1990) but add sensitivity. Given our retrieval’s computational expense we an A-band absorption ratio that is sensitive to Ptop. limit to nadir orbits to optimise the likelihood of good re- Our OCO-2 OE retrievals are computationally expensive trievals. due to the complex radiative transfer (RT), so we aim to avoid OCO-2 carries three co-boresighted grating spectrome- footprints which are unlikely to yield good retrievals. The ters centred over the O2 A-band (λ ∼ 0:78µm), weak CO2 cloud flagging and prior LUT retrieval developed here are a band (λ ∼ 1:68µm) and strong CO2 band (λ ∼ 2:06µm). The necessary step in excluding these footprints, and we further satellite operates in a push-broom fashion with a swath of exclude those where solar zenith angle (SZA) > 45◦ based eight footprints whose orientation relative to the track ro- on OCO2CLD-LIDAR-AUX’s retrieval statistics. It is pos- tates through the orbit as the satellite angles to optimise solar sible that a future partial-column (i.e. above-cloud) XCO2 power generation.