1. Introduction
Campbell et al. (2016, hereinafter C16) and Lolli et al. (2017, hereinafter L17) describe multiyear ground-based NASA Micro-Pulse Lidar Network (MPLNET, 532 nm; Welton et al. 2001; Campbell et al. 2002) measurements of cirrus clouds and corresponding estimates of their net daytime top-of-the-atmosphere (TOA) cloud radiative forcing (CRF; i.e., the difference in TOA solar and infrared radiation budgets estimated in the presence of cloud versus that of the corresponding clear sky alone) at Greenbelt, Maryland, in 2012 [38.998°N, 76.848°W; 50 m above mean sea level (MSL)] and Singapore in 2010–11 (1.308°N, 103.778°E; 79 m MSL), respectively. C16 estimate an absolute annual net daytime TOA CRF (defined as the sample-averaged net forcing normalized by the numbers of local daylight hours and absolute cloud occurrence rate estimated from satellite) ranging between 0.03 and 0.27 W·m−2 exclusively over land. L17 estimate net TOA CRF ranging between 2.20 and 2.59 W·m−2 over land and −0.46 and 0.42 W·m−2 over ocean using a common methodology. C16 propose a meridional hemispheric gradient in daytime cirrus cloud net TOA CRF based on their midlatitude analysis, which is confirmed after pairing C16 with the tropical dataset in L17. The remaining question is how this gradient evolves poleward of the midlatitudes.
Cirrus clouds are unique within the radiant planetary energy budget. All clouds induce a positive nighttime net TOA CRF term (i.e., warming, whereby clouds induce an enhanced downward infrared flux with no offsetting solar visible input and/or reflection). However, during daytime, cirrus cloud forcing is influenced by, in no order of significance, cloud heights and temperatures, thermal surface contrast, effective ice particle sizes, total cloud water content, surface albedo, and solar zenith angle (e.g., Stephens et al. 1990). Given their uniquely transmissive (Sassen and Cho 1992) optical and relatively cold nature (e.g., Campbell et al. 2015), cirrus are the only atmospheric cloud genus that regularly induces both positive and/or negative net daytime TOA CRF values depending on the balance of these determining factors for any given cloud scene (Stephens and Webster 1981). Cirrus are ubiquitous in the upper troposphere and lower stratosphere during all seasons and over all regions (e.g., Rossow and Schiffer 1999; Mace and Zhang 2014). This combined results of C16 and L17 and the presence of a significant meridional net daytime forcing gradient are a reflection of their consequential role within the climate system. A positive net daytime cirrus cloud TOA CRF component present over any significant portion of the globe is in direct contrast with the net diurnal cooling globally exhibited by clouds overall (near −20 W m−2; Ramanathan et al. 1989; Yi et al. 2017).
Ground-based lidar observations, typically combined with in situ ice microphysical sampling, have long been the backbone of cirrus cloud process studies (i.e., Sassen 1991). While the community continues reaping the benefits of the now nearly 15-yr global archive of NASA Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP; Winker et al. 2010) datasets, ground-based networks like MPLNET have similarly grown more diverse and widespread (e.g., Pappalardo et al. 2014). Satellite-based analyses of global and regional atmospheric processes benefit greatly from ground-based context. Opportunities for enhanced sampling resolutions and measurement sensitivities from ground, critically combined with diurnal observing, remain a practical and inexpensive resource for long-term climate monitoring and means-testing science hypotheses before working with larger satellite datasets. By understanding and characterizing unique atmospheric phenomenon with ground and suborbital measurements, reconciling satellite-based measurements becomes easier (e.g., Swap et al. 2003; Holben et al. 2018).
Here, we complete a trilogy of MPLNET-based cirrus cloud net TOA CRF studies initiated in C16 and L17 by examining forcing characteristics along the final point of a conceptual meridional gradient. We examine two years of cirrus cloud measurements and corresponding net TOA CRF estimates from 2017 and 2018 as collected at Fairbanks, Alaska (64.86°N, 147.85°W, 300 m MSL) in the subarctic. The latter distinction proves particularly relevant in choosing a final location to pair with Greenbelt and Singapore. That is, we are estimating a daytime-absolute net TOA CRF that is critically dependent on available sunlight hours. Since Fairbanks is situated so far north, solar inputs are highly variable annually, and critically limited during winter. The consideration of two years of data for this study was done to accumulate two “summer” periods versus two “winter” periods (respectively defined broadly here as May–October and November–April on the basis of average, but unique, broadband solar surface albedos; shown below), to analyze the approximate equivalent of one year worth of sample cases from those at lower latitudes.
The goals of this paper are to summarize the Fairbanks dataset, to describe important refinements to our processing methodology compared with the prior two studies and to discuss our results. The Fairbanks results themselves represent a stand-alone result, but we conclude by discussing them within the context of the three MPLNET studies, thus completing the series. Like in the case of Singapore, we highlight the interannual variability within the 2-yr sample. However, we introduce monthly averaged satellite-derived broadband solar surface reflectances in order to best resolve the influence of wintertime snow coverage along the Fairbanks surface. Net daytime TOA CRF increases with increasing surface albedo. We highlight the effect of this mechanism with respect to the monthly accumulation and decline of local surface snowpack. The result is a quantitative estimate of cirrus net daytime TOA CRF along the subarctic, as well as a new recognition of the variability in regional weather processes and their impact on the sign and magnitude of cirrus cloud forcing nearing the poles.
2. Data and experimental design
C16 and L17 describe the methodology applied to MPLNET data for estimating cirrus cloud daytime TOA CRF. Briefly, single-layer cirrus clouds in MPLNET, version 3.0, datasets are resolved according to Campbell et al. (2008, 2015) and Lewis et al. (2016). Radiative transfer calculations are performed using the Fu–Liou–Gu model (FLG; Fu and Liou 1992, 1993; Gu et al. 2003, 2011). We retain the broadband solar and infrared parameterization of ice microphysical and optical properties from Yang et al. (2000, 2005). The relationship between lidar-derived 532-nm extinction coefficient, ice crystal radius, and cloud-top temperature is based on Heymsfield et al. (2014). Thermal profiling comes from NASA Goddard Modeling and Assimilation Office (GMAO), version 5.9.1 (GEOS-5), reanalysis datasets, which are used to initialize FLG simulations. Uncertainties in GMAO-derived temperatures from the upper troposphere are believed to be less than 1°C (M. Rienecker 2013, personal communication). The 532-nm MPLNET cloud extinction coefficient (km−1) is solved in unconstrained MPLNET cirrus cloud retrievals (Campbell et al. 2008; Lewis et al. 2016) using a priori settings for the lidar extinction-to-backscatter ratio (the so-called lidar ratio) of 20 and 30 sr to simulate practical bookend system variance (e.g., Chew et al. 2011; Garnier et al. 2015). Clouds are modeled assuming particulate aerosol- and cloud-free skies above and below then so as to uniquely isolate the cirrus TOA CRF term. Experimental system elements are summarized in Table 1.
A summary listing of primary experimental components and data constraints used in the study.
Refinements to the processing protocol were inevitable over the 6-yr time period taken to conduct these studies, though we consider the impact on the results to be minor if at all significant. The first change was reported by L17 in adding a varying ocean surface albedo term designed for accommodating very low solar zenith angles in the tropics and potential sun glint over water (Jin et al. 2004). We apply a monthly averaged broadband solar surface reflectance term derived from the NASA Clouds and the Earth’s Radiant Energy System (CERES; Wielicki et al. 1996; Rutan et al. 2015) Synoptic (SYN1deg), version 4A, product. This value and the monthly total number of daylight hours at Fairbanks are detailed in Table 2. This distribution depicts the influence of snow on the local surface and annual variability in surface reflectivity overall that influences the solar CRF term at the TOA. Our summer/winter definitions introduced above are borne out, respectively, as albedos less than or greater than about 0.20. Daytime is defined through all three studies as those hours when the incoming solar radiance exceeds the combined outgoing solar and thermal component, which constrains whether or not a cloud can induce a positive net TOA CRF term. Finally, cloud-base temperatures were not allowed to exceed −25°C for fear of aerosol contamination in the MPLNET cloud product from erroneous mineral dust layers (Lewis et al. 2020).
Number of daytime hours per month and the monthly average broadband surface albedos from CERES SYN1deg Ed4A in 2017 and 2018 at Fairbanks.
C16 spend considerable time discussing many of the uncertainties present within our experimental design, and their likely impact on any subsequent analyses. While we strive for accuracy, and are as circumspect as we can be about both the strengths and limitations of the ground-based lidar data sample and our radiative transfer experiments, the net daytime TOA CRF values reported here strictly remain estimates given these constraints.
3. Top-of-the-atmosphere cloud forcing analysis at Fairbanks
a. Method and results
Shown in Fig. 1 are climatological cirrus cloud occurrence rates centered regionally over Alaska at 2.5° × 2.5° resolution derived from nearly 10 years of CALIOP dataset observations (June 2006 to May 2017). Consistent with Campbell et al. (2015), the CALIOP, version 4.10, cloud-layer product is subset for all ice-classified layers featuring a cloud-top temperature colder than −37°C. During the summer months, we apply a frequency of 43.7%, as compared with 44.8% during winter (44.2% annual). Subarctic CALIOP-estimated cloud occurrence rates are broadly consistent with those of the upper midlatitudes across the Northern Hemisphere (not shown), but higher than those of the Arctic (not shown). We believe this a result of the lessening influence of convection working poleward. In normalizing our results by the CALIOP-derived cloud occurrence, however, we are assuming diurnal cloud homogeneity, since the climatological CALIOP estimates come from daytime (~1330 local time) and nighttime (~0130 local time) overpasses. Diurnal cirrus cloud variability globally is an open question. When compared with C16 and L17, however, this is believed to be a considerably more stable approximation in the subarctic as compared with lower latitudes.
Cirrus cloud absolute occurrence rate frequencies at 2.5° × 2.5° resolution estimated from NASA CALIOP level 2 cloud-layer products, and thus subcategorized on the basis of definitions outlined in Campbell et al. (2015), from June 2006 to May 2017 centered near Fairbanks (white star) (a) annually, (b) from May to October, and (c) from November to March.
Citation: Journal of Applied Meteorology and Climatology 60, 1; 10.1175/JAMC-D-20-0077.1
The daytime Fairbanks cirrus cloud dataset is depicted for sample-relative frequencies of cloud-top temperature (CTT; °C) at specific COD ranges in Fig. 2, divided into 2017 and 2018 annual samples and their respective summer/winter distributions. The overall sample sizes are smaller compared with C16 and L17, which is a reflection of the fewer numbers of available daytime cases during the winter months. From Table 2, November, December, and January feature no daylight hours meeting our definition. At 14 266 and 12 468 total cases in each respective year (Fig. 2), the combined 2-yr sample is roughly comparable to the more than 21 000 cases in 2012 at Greenbelt and 15 000–18 000 cases in the two years (2010 and 2011) at Singapore.
Sample-relative daytime cirrus cloud occurrence histograms derived from continuous NASA MPLNET observations at Fairbanks during (top) 2017 and (bottom) 2018 as a function of COD (30-sr solutions only; see insets for bulk percentages) vs CTT (a),(d) annually; (b),(e) from May to October; and (c),(f) from November to March.
Citation: Journal of Applied Meteorology and Climatology 60, 1; 10.1175/JAMC-D-20-0077.1
Samples are shown as a function of three COD groupings, including subvisual (COD ≤ 0.03), optically thin (0.03 < COD ≤ 0.30), and opaque (0.30 < COD) clouds [as defined by Sassen and Cho (1992)]. Both years are similar in relative frequencies of COD (29%, 42%, and 28% for 2017 and 32%, 41%, and 25% for 2018 for subvisual, optically thin, and opaque, respectively). This result is consistent with the midlatitude site at Greenbelt in C16, though surprisingly skewed toward thinner clouds compared with Singapore in L17. Given the propensity for convection and tropopause transition-layer cirrus clouds nearer the equator (e.g., Virts and Wallace 2010), optically thin and subvisual clouds are in fact lower in sample frequency at Singapore than in the midlatitudes and subarctic (16% vs 48% and 35% for 2010 and 15% vs 50% and 35% for 2011 from L17). This is presumably a sampling artifact caused by the zenith-viewing MPLNET measurement configuration, as compared with what would otherwise be apparent from satellite-based CALIOP (Campbell et al. 2015).
CTT distributions in Fig. 2 clearly reflect two very different years, with colder (and presumably higher) cirrus clouds dominant in 2017 and warmer (lower) ones more frequent in 2018. In a forthcoming paper (D. R. Ryglicki et al., unpublished manuscript), a 39-yr (1980–2018) trend analysis of the fifth major global reanalysis by the European Centre for Medium-Range Weather Forecasts (ERA5) Arctic datasets (70°–90°N) includes an investigation of domain-averaged 2.0–potential vorticity unit (PVU; 1 PVU ≡ 10−6 K·kg−1·m2·s−1) height anomaly, a proxy for tropopause height (e.g., Martius et al. 2010), and its subsequent interannual variability after data smoothing via a low-pass filter. These data reflect the influence of varying weather processes on regional circulation and cloud formation/sustainment (e.g., Sassen and Campbell 2001; Parker et al. 2014) and are shown in Fig. 3.
Thirty-nine-year (1980–2018) seasonal 2.0-PVU (summer and winter) height anomalies and trend analyses solved over the Arctic domain (70°–90°N) from ERA5 datasets. Interannual values are solved after daily data point smoothing is done with a low-bandpass filter.
Citation: Journal of Applied Meteorology and Climatology 60, 1; 10.1175/JAMC-D-20-0077.1
In comparing the CTT distributions of Fig. 2 with the polar CTT occurrence rates (|60°| ≤ latitude < |80°|) from 2012 CALIOP observations in Campbell et al. (2015; their Fig. 3d), it is seen that the 2017 distribution is highly consistent. From Fig. 3, both 2012 and 2017 coincide with positive relative summer regional 2-PVU height anomalies (both > 100 m). Relative to surrounding years with negative anomalies, this is indicative of generally more stable regional conditions, higher tropopause heights, more anticyclonic-dominant circulation, greater shortwave insolation, and less low-cloud cover. This finding matches well with the colder and higher cirrus observed by MPLNET. The 2018 summer, in contrast, coincided with a negative 2-PVU-height-anomaly summer season (~−75 m), meaning less stability than in 2012 and 2018, lower tropopause heights, greater cyclonic regional circulation, less insolation and increased warmer and lower cloud cover, which again is borne out in the MPLNET sample. The impact of this finding is found in the respective MPLNET cirrus cloud sample sizes from 2017 and 2018 (14 266 vs 12 468; Fig. 2). The lesser number in the latter year indicates greater low cloudiness and more frequent attenuation of the lidar before profiling the upper troposphere. Conversely, both 2017 and 2018 winter anomalies were opposite to their summer counterpart, which is a result that will be discussed further below.
Depicted in Fig. 4 are annual cloud sample frequencies as a function of COD (0.03 resolution), sample-relative average net daytime TOA CRF per 0.03 COD bin and corresponding sample-normalized net daytime TOA CRF per 0.03 COD bin for the 20–30- sr MPLNET samples from the full 2-yr dataset [note that these are merely sample averages after the monthly-based method introduced above and outlined in Eq. (1) has been applied]. Respective 2017 and 2018 yearly estimates range from −0.49 to 1.10 W·m−2 and from −1.67 to 0.47 W·m−2, meaning that the higher and colder clouds observed in 2017 correspond to greater positive forcing than the lower, warmer sample in 2018 (i.e., most likely an enhanced thermal infrared term). Monthly and annual net daytime TOA CRF values are outlined in Table 3. Figures S1 and S2 in the online supplemental material complement Fig. 4 in respectively depicting the 2017 and 2018 years alongside their winter and summer seasonal components.
For (left) the bulk 2-yr sample and for individual years (center) 2017 and (right) 2018, (a)–(c) the MPLNET-derived daytime cirrus cloud sample-relative frequency, (d)–(f) sample-relative average net TOA CRF (W·m−2), and (g)–(i) absolute normalized net TOA CRF solved at Fairbanks.
Citation: Journal of Applied Meteorology and Climatology 60, 1; 10.1175/JAMC-D-20-0077.1
Absolute (normalized for daytime hours from Table 1 and absolute cloud occurrence frequency derived from CALIOP; see text) net TOA CRFs during daytime and nighttime solved using the 20- and 30-sr solutions for 2017 and 2018.
We introduce an additional net TOA CRF estimate in this analysis, in contrast with C16 and L17. Also shown in Table 3 are nighttime absolute forcing values. Here, nighttime MPLNET cloud profiles taken from the Fairbanks sample were applied and analyzed in the same monthly resolved manner as their daytime counterparts. Sample sizes overall were higher during nighttime (27 250 and 30 932, respectively), which is the result of the greater percentage of nighttime hours occurring at Fairbanks versus daytime (69% vs 31% daytime).
Annual net nighttime TOA CRF across the 2-yr sample is estimated to be between 27.62 and 51.65 W·m−2. Respective 2017 and 2018 yearly values range from 27.79 to 52.59 W·m−2 and from 27.45 to 50.70 W·m−2. The annual net diurnal TOA CRF (summing the respective daytime and nighttime values) is estimated to be between 26.54 and 52.43 W·m−2. Respective yearly values are 27.30–53.69 W·m−2 and 25.78–51.17 W·m−2. Relative differences of nearly 1.5–2.5 W·m−2 in the net annual nighttime aggregates are likely attributable to the same differences in regional weather likely responsible for the difference in the daytime result. Enhanced nighttime thermal forcing from a colder/higher cloud sample induces a larger net value in 2017 when compared with the warmer/lower sample in 2018.
Figure 5 features seasonal depictions of sample-relative average net daytime TOA CRF distributions versus COD, similar in nature to the annual composites in Figs. 4e and 4f (as well as supplemental Figs. S1d and S1e and S2d and S2e for the respective 2017 and 2018 summer and winter periods). The goal here is a simpler intercomparison of these distributions to identify dominant and/or contrasting interseasonal attributes. Due to noise compromising the higher end of the COD spectrum, caused by limited sample sizes, the data are only shown to 1.0. Additionally, data are shown on a logarithmic scale to help better distinguish the optically thin portion of the sample (COD ≤ 0.30).
Sample-relative net average TOA CRF distributions vs log10(COD) for the (a) annual 2017 (solid lines) and 2018 (dashed lines) daytime samples as a function of their 20- (blue) and 30-sr (red) estimates and for the 2017 and 2018 (b) summer and (c) winter periods.
Citation: Journal of Applied Meteorology and Climatology 60, 1; 10.1175/JAMC-D-20-0077.1
Within the annual composite (Fig. 5a), we find that the 2017 sample generates slightly but consistently higher net daytime TOA CRF across all depicted cloud spectra versus 2018, which is consistent with the colder and higher cloud sample overall. From Table 3, only during February, August, and October are 2018 monthly values greater than 2017. The seasonal composites depict greater sensitivity, however. Summertime positive average net daytime TOA CRF along the optically thin cloud spectrum (COD < 0.30) is considerably higher in 2017 (Fig. 5b), a feature that is consistent with the generally higher annual forcing values overall that year compared with 2018. In the limited winter months, however (Fig. 5c), 2018 is higher than 2017 through most of the optically thin cloud spectrum. As introduced above, winter 2-PVU anomalies in 2017 and 2018 were opposite to their summer counterparts. The lack of sufficient hours during winter months compared with summer lessens their impact on the overall annual forcing estimate. While we stress interannual variability in regional weather as the primary driver of differences in magnitude and sign of annual net daytime TOA CRF, this result is a reflection of the causal impact of interseasonal variability as well.
Shown in Fig. 6, similar to C16 and L17, are breakdowns of sample frequencies, sample-relative and sample-normalized net daytime TOA CRF versus solar zenith angle (SZA) for the bulk 2-yr sample and each year (complementary seasonal results are again included as Figs. S3 and S4 in the online supplemental material for 2017 and 2018, respectively). The Fairbanks sample is consistent with C16 and L17 in that they extend to just below 80°, meaning that, based on our definition of daytime hours, incoming solar radiation exceeds outgoing total radiance over the course of the year in the subarctic up to roughly the same SZA maximum as the midlatitude and tropical sites. This result is not necessarily intuitive, however, given varying climatological thermal profiles among the three sites (i.e., mostly colder, particularly at lower levels, moving poleward).
As in Fig. 4, but vs SZA (°).
Citation: Journal of Applied Meteorology and Climatology 60, 1; 10.1175/JAMC-D-20-0077.1
Another compelling attribute apparent from these subarctic data compared with C16 and L17 is the crossover point in SZA between positive and negative net TOA CRF. Given the inherently different climatological vertical thermal profiles at each site, a warmer surface and colder upper troposphere at Singapore should coincide with greater numbers of SZA samples inducing positive TOA CRF. L17 find that the crossover value falls between roughly 55° and 65° over land based both on the 20/30-sr sample and interannual sample differences. We would then expect this crossover value to decrease moving poleward, where surface temperatures gradually decrease and thermal contrast, which drives the positive infrared warming component to the net estimate, similarly decreases.
As expected, at Greenbelt C16 report this value at between 50° and 55°. At Fairbanks, however, the value does not continue to decrease, but instead increases to between roughly 60° and 65°, again based on the interannual and 20/30-sr samples. We speculate that this is caused again by daytime sampling differences, and the fewer numbers of cases at the locally higher SZA that would otherwise be available to the sample during an equally robust winter period. Comparing Fig. 6a with Fig. 4b of C16, for instance, suggests that this interpretation is plausible, given that the Fairbanks sample sizes are generally flatter with increasing SZA compared with the midlatitude dataset.
The final analysis point comes in Fig. 7, which is a continuation of Figs. 4 and 6, but now comparing frequencies, sample-relative and sample-normalized net daytime TOA CRF versus CTT (complementary seasonal plots for each year are featured in Figs. S5 and S6 in the online supplemental material). The same distributions from Fig. 2 are reflected here in Figs. 7a–c, with 2018 exhibiting greater numbers of warmer clouds. Crossover points between negative and positive net TOA CRF range between −47° and −50°C, which agrees well with the Greenbelt result but is a few degrees colder than the Singapore overland result (overocean is warmer). Again, from climatological temperatures and thermal contrast with the surface, we expect the crossover temperature to decrease moving poleward. The relative consistency with the midlatitude result, however, is compelling, suggesting that the meridional variance, at least over land, is small enough to fall within our estimated result uncertainties derived between the three sites.
As in Fig. 4, but vs CTT (°C).
Citation: Journal of Applied Meteorology and Climatology 60, 1; 10.1175/JAMC-D-20-0077.1
b. Discussion
Annual forcing estimates from Fig. 4 and Table 3 are in contrast with the original hypothesis of C16. In hypothesizing the existence of an annual meridional forcing gradient, C16 presume that the net TOA CRF gradually turns from neutral in the midlatitudes to negative moving poleward over land (C16 fail to recognize the corresponding net daytime TOA CRF impact of lower albedo over waters). Herein, the results from Fairbanks suggest that the net forcing can turn negative in the subarctic, as a function of variability in regional/seasonal weather patterns. However, C16 should also be reconsidered with additional years of study to better understand the interannual variability possible there.
The influence of SZA in the subarctic is critical to quantifying the net daytime TOA CRF term. At Fairbanks, polar summer means many daylight hours and lower SZA compared with insufficient insolation during winter and significantly greater relative numbers of nighttime hours. In other words, daytime in the subarctic summer corresponds with the ideal time of the year to induce positive net TOA CRF, which then dominates the net annual daytime estimate without sufficient compensation from daylight cases under higher SZA, and thus lesser (likely negative) net TOA CRF, during winter. This latter process can be visually inferred from Figs. S1 and S2 in the online supplemental material. Therefore, the Fairbanks annual result does not necessarily turn negative, depending on the inherent weather processes that are most dominant during summer in a given year.
A noteworthy ramification of these results comes in the context of the recent study of Yi et al. (2017). They estimate daytime “ice cloud” net TOA CRF using the 2012 Collection 6 Aqua MODIS cloud mask as the input dataset. It has become abundantly clear, however, through a host of studies (e.g., Campbell et al. 2015; Holz et al. 2016; Marquis et al. 2017) that the MODIS cloud mask lacks fundamental sensitivity to optically thin and subvisual cirrus clouds (COD < 0.30). Cirrus clouds are exponentially more common at the lower COD values (e.g., Fig. 4a), with roughly half of all cirrus clouds exhibiting CODs less than 0.30 (Campbell et al. 2015), and they likely to go undetected by MODIS cloud algorithms (e.g., Marquis et al. 2017).
Yi et al. (2017) report a slightly negative zonal-mean net TOA CRF for ice clouds in their analysis. Consistent with the C16 hypothesis, their depiction of a poleward gradient in zonal-mean net daytime TOA CRF (their Fig. 2c) turns negative in the midlatitudes extending through the subarctic. Again, the 2012 Arctic cirrus CTT distribution from CALIOP (Campbell et al. 2015) compares well to the 2017 Fairbanks MPLNET distribution. Both years coincided with positive 2-PVU height anomalies and thus more stable polar-regional weather conditions. Although not conclusive, this evidence points to likely higher relative net daytime TOA CRF compared with surrounding years, and we would expect something similar to the result derived during 2017 (as opposed to 2018) from MPLNET at Fairbanks. This is not what we find in Yi et al. (2017), however. Granted, the absolute differences are relatively small (on the order of 1.0 W·m−2), but the sign is distinctly opposite. We acknowledge that Yi et al. (2017) use a different ice microphysical parameterization and two-stream solution to simulate ice optical properties within their radiative transfer model, which can be responsible as responsible for some significant measure of these differences.
Shown in Table 4 are the results of a complementary analysis to Table 3. However, we have limited the MPLNET sample here to only clouds where COD ≥ 0.30, which is intended to better simulate what a MODIS-like sensor is more likely to resolve. Normalization by the regional CALIOP cloud occurrence frequency was cut in half (21.8% in summer or 22.4% in winter) as an approximation to the balance between optically thin and optically thick clouds overall. Average annual net daytime TOA CRF over the 2-yr sample ranges between −3.35 and −0.27 W·m−2. Respective yearly daytime values vary from −1.94 to 0.36 W·m−2 and from −4.76 to −0.89 W·m−2. When limiting the MPLNET sample to clouds exhibiting COD ≥ 0.30, our results are more consistent with Yi et al. (2017). This consistency implies that, despite differences in optical model and various physical assumptions applied, the impact of including thin cirrus significantly influences the final net TOA CRF estimate. Nighttime values over the full 2-yr sample range between 32.95 and 54.59 W·m−2. By year, they range from 33.63 to 55.69 W·m−2 and from 32.38 to 53.49 W·m−2, respectively. These results are justifiably higher, given the removal of thin clouds from the sample. Future TOA cirrus cloud closure studies involving CERES should make every attempt to represent the contribution of optically thin cirrus.
As in Table 3, although now only for a sample featuring CODs greater than or equal to 0.30, absolute (normalized for daytime hours from Table 1 and half of the absolute regional cloud occurrence frequency derived from CALIOP; see text) net TOA CRFs during daytime and nighttime solved using the 20- and 30-sr solutions for 2017 and 2018.
We offer one final thought with regard to these bulk estimates with respect to regional weather. The MODIS era began in 2000 with the launch of the Terra satellite (e.g., Platnick et al. 2003). CALIOP profiling began in 2006. This study and the Yi et al. (2017) closure studies are mere snapshots of the net daytime TOA CRF estimates, which otherwise exhibit appreciable interannual variability. We hypothesize that a direct correlation between 2-PVU height anomalies and net daytime TOA CRF. Figure 3 shows that 2-PVU height anomalies over the 39-yr ERA5 reanalysis period have trended toward positive values, with a flip in sign occurring just before 2000. It is very likely, then, that cirrus cloud net daytime TOA CRF values along the subarctic have been trending upward over this time period, and that negative net daytime values were far more common, if not predominant, in past years. Considering the rapid acceleration of regional polar surface temperatures, as well as sea ice loss and its impact on regional surface albedos (e.g., Screen and Simmonds 2010), a positive feedback loop between cirrus cloud net daytime TOA CRF and background surface/thermal processes is likely occurring within the regional system. We will evaluate this hypothesis in a forthcoming study.
4. Conclusions and perspective
This paper describes the third and final component in a trilogy of ground-based Micro-Pulse Lidar Network (532 nm; MPLNET) studies focused on global daytime cirrus cloud physical and top-of-the-atmosphere radiative forcing properties. C16 report on a 1-yr analysis that is based on observations collected over land during 2012 at Greenbelt in the midlatitudes, and L17 describe two years (2010–11) of measurements characterized both over land and water that were collected at Singapore. Here, a 2-yr (2017–18) overland analysis is presented for datasets gathered at Fairbanks in the subarctic. C16 motivate this cumulative effort by hypothesizing a global meridional TOA cloud radiative forcing gradient, whereby net daytime TOA CRF is positive near the equator and gradually turns negative approaching the poles. Cirrus clouds are the only phenomenological cloud type in the atmosphere that can induce either a positive and/or negative net daytime TOA CRF depending on cloud temperatures, vertical thermodynamic profile, cloud thermal contrast with the surface, solar zenith angle, surface reflectance, and ice crystal habits and size. Hence, evaluation of this hypothesis has resulted in a three-part study evaluating daytime cloud forcing properties meridionally along three distinct zonal bands in the Northern Hemisphere.
At Fairbanks, the estimated absolute net daytime TOA CRF across the 2-yr sample ranges between −1.08 and 0.78 W·m−2. Annual estimates from 2017 range between −0.49 and 1.10 W·m−2, and in 2018 they range between −1.67 to 0.47 W·m−2. These estimates were reached after normalizing the ground-based sample by a roughly 44% regional cirrus cloud occurrence rate, derived from the satellite-based Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) instrument, and percentages of daylight hours resolved monthly. The months of November, December, and January contributed no data points to the sample given that our definition of daylight requires the incoming TOA solar radiance to exceed the outgoing total (solar and infrared) radiance, which is necessary to induce a negative net TOA CRF term for any given cloud sample.
These estimates are not fully consistent with the hypothesis of C16, whereby higher relative solar zenith angles (increasing relative cloud albedos) and lesser overall thermal contrast between the clouds and the colder ground surface (suppressing thermal infrared warming) poleward of the relatively neutral CRF in the midlatitudes would force the term to net negative overall. We believe this result to be the primary impact of two factors. First, regional weather variability in the subarctic is increasingly more influential than that in the midlatitudes. We quantify this statement by presenting a 39-yr trend analysis of 2.0-PVU height anomalies over the Arctic to estimate interannual and interseasonal (i.e., summer vs winter) differences in the net TOA CRF. While some of this variability has been noted in unusual wintertime cirrus cloud observations at Fairbanks (Campbell et al. 2018), we present additional evidence representing how such variability is manifested overall. Colder, higher cirrus clouds were more frequent in 2017, and warmer, lower cirrus clouds were more frequent in 2018; the former correlates well with positive 2-PVU anomalies, and the latter with negative anomalies. The 2017 cloud-top temperature frequency distribution also matched that derived in the Arctic from 2012 using CALIOP, which was also another year in which positive 2-PVU height anomalies were observed. Second, despite the increase in surface albedo that drives net TOA CRF higher relative to a barren land surface below, the lack of compensating numbers of daylight cases during winter months means that May–July cases dominate the net TOA CRF overall. These tend to be more positive than not.
These findings lead us to two additional hypotheses that we will test and evaluate in future efforts. First, global annual cirrus cloud net daytime TOA CRF will prove most sensitive to variability induced by regional weather processes along the subarctic zonal belt. Singapore results in L17 were generally consistent across the 2-yr sample considered. It is believed that the predominance of convection and ubiquitous presence of cirrus in the tropics are likely to render a relatively stable net TOA CRF effect. Whereas it would be prescient to examine tropical interseasonal cirrus cloud physical and forcing variability with respect to dominant regional-scale circulatory phenomena (e.g., the Madden–Julian oscillation and/or the North Atlantic Oscillation; Hurrell et al. 2001; Zhang 2005), we believe it likelier that conditions are more stable approaching the equator overall. By the time one reaches the subarctic, oscillating variability in the intrusion of midlatitude weather systems versus the extent of polar-dominant dynamics are plausibly likelier to influence cirrus cloud properties in a manner that more significantly impacts the globally aggregated annual-mean daytime cirrus net TOA CRF.
Second, and perhaps more noteworthy, is the causal correlation of our annual and seasonal net daytime TOA CRF with polar-regional tropopause height anomalies. Increasing mean 2-PVU heights over roughly the last 40 years are an indication that cirrus net daytime TOA CRF has been likely increasing regionally over that span. It is very plausible that net TOA CRF was negative at the beginning of that study period, consistent with the original C16 hypothesis. Given the amplification of Arctic warming in recent years (Thorne 2008), there has very likely been a positive cirrus cloud forcing feedback effect driven by colder/higher clouds and greater thermal contrast with warming surface temperatures. However, we only look at this effect over land. Over ice and open waters, the impact of receding sea ice regionally will counteract this effect since darker surfaces decrease relative forcing significantly relative to ice-covered ones. The increase in open waters would thus lessen forcing. Again, we are working to answer this question in a future study.
Although the C16 hypothesis does not play out directly in these results, the presence of a meridional gradient, with both positive and negative components, is verified in this three-part MPLNET analysis, and similarly confirmed in part with the recent work of Yi et al. (2017). The positive net daytime TOA CRF component of daytime diurnal forcing nearing the equator stands in distinct contrast with a wholly negative daytime forcing component found approaching the poles. Total estimated cloud TOA forcing amounts to roughly −20 W m−2, and all cloud phenomenological types induce a negative net daytime TOA CRF except cirrus. Positive net daytime cirrus TOA CRF forcing nearing the equator is thus a compensating component to the net daytime cooling effect from all other clouds. Additional work is necessary to extrapolate the work laid out in these studies to satellite lidar datasets and thus estimates of global annual, interannual, and interseasonal numbers (e.g., Hong et al. 2016). In an evolving climate, we must continue to monitor the distribution and frequency of positive cirrus daytime TOA CRF and its offset to the negative all-cloud daytime CRF over time.
One regrettable lesson of this effort has been in recognizing the inconsistency of ice physical models used for fundamental radiative transfer analysis. The community needs to close the gap between ice optical models, which govern single-scattering, asymmetric, and absorptive properties, and physical ones that relate ice crystal size and water content. The latter are fundamental relationships for solving radiative parameters that associate lidar measurements of optical extinction, in this case, or numerical weather prediction models of condensate as another example. An optical model based on a single or mixed set of habits and a given size distribution implies a corresponding relationship between water content, effective size and spectral extinction that does not yet exist. For these MPLNET papers we have consistently used a set of optical parameterizations for infrared and solar broadband characteristics that are reasonably out of date (Yang et al. 2000, 2005). We further use the relationship of Heymsfield et al. (2014) to relate temperature and effective ice crystal size based on field observations, which may or may not accurately relate to the corresponding optical model, and rely on the original parameterization of Fu and Liou (1993) relating ice water content and particle size to spectral broadband extinction.
The uncertainty reported in the forcing estimates reported here is based only on the input MPLNET lidar observations. No uncertainty is understood either from the use of older and potentially incompatible parameterizations. In a sense, radiative transfer models have become “black boxes” to community users, with no complementary error models to properly characterize results. Incompatibility between optical and physical input parameterizations is at least one inconsistency that can be removed from the process, given the breadth of new and complementary retrievals of particle size from passive radiometric sensors tied to a single, binding ice optical model. In a pending study, we will further do just that.
Thus, in closing, we turn to reinforcing the compulsory value of ground versus satellite remote sensing as an enduring and dependable context to better understanding the increasing number of satellite remote sensing available to climate researchers. Here, the study of Yi et al. (2017) seeking MODIS and CERES cloud closure is better reconciled, and in fact fundamentally corrected by considering thin cirrus clouds that the passive radiometer struggles to detect. Balmes and Fu (2018) make a similar point by conceptualizing the potential undersampling of thin cirrus clouds by satellite sensors. Dolinar et al. (2020) make another case for this same process and its radiative impact. The innovative and cohesive MPLNET has compiled decades of cloud and aerosol observations over many sites and continues expanding to this day. The ground-based datasets thus serve as a means for better testing a specific hypothesis that can then be more thoroughly investigated and characterized from satellite. Ultimately, the context available from ground observations to better understanding datasets, like that of NASA Earth Observing System and prominently including MODIS and CALIOP, is invaluable.
Acknowledgments
The NASA Micro-Pulse Lidar Network is supported by the Earth Observing System project (S. Platnick) and the NASA Radiation Sciences Program (H. Maring). The Radiative Effects of Thin Cirrus Clouds (REThinC) project is supported by the Naval Research Laboratory Base Program (BE033-03-45-T008-17; Drs. S. Chang, J. Hansen, and R. Preller). Author D. R. Ryglicki has participated here through support of the Office of Naval Research Arctic Cyclone Department Research Initiative, Program Element 0601153N (R. Ferek, D. Eleuterio, and J. Doyle). We gratefully acknowledge assistance from the administrative staff of the Geophysical Institute at the University of Alaska Fairbanks in maintaining the instrument and operating laboratory since lidar operations commenced in autumn 2016.
REFERENCES
Balmes, K., and Q. Fu, 2018: An investigation of optically very thin ice clouds from ground-based ARM Raman lidars. Atmosphere, 9, 445, https://doi.org/10.3390/atmos9110445.
Berry, E., and G. G. Mace, 2014: Cloud properties and radiative effects of the Asian summer monsoon derived from A-Train data. J. Geophys. Res. Atmos., 119, 9492–9508, https://doi.org/10.1002/2014JD021458.
Campbell, J. R., D. L. Hlavka, E. J. Welton, C. J. Flynn, D. D. Turner, J. D. Spinhirne, V. S. Scott, and I. H. Hwang, 2002: Fulltime, eye-safe cloud and aerosol lidar observation at Atmosphere Radiation Measurement program sites: Instrument and data processing. J. Atmos. Oceanic Technol., 19, 431–442, https://doi.org/10.1175/1520-0426(2002)019<0431:FTESCA>2.0.CO;2.
Campbell, J. R., K. Sassen, and E. J. Welton, 2008: Elevated cloud and aerosol layer retrievals from micropulse lidar signal profiles. J. Atmos. Oceanic Technol., 25, 685–700, https://doi.org/10.1175/2007JTECHA1034.1.
Campbell, J. R., M. A. Vaughan, M. Oo, R. E. Holz, J. R. Lewis, and E. J. Welton, 2015: Distinguishing cirrus cloud presence in autonomous lidar measurements. Atmos. Meas. Tech., 8, 435–449, https://doi.org/10.5194/amt-8-435-2015.
Campbell, J. R., S. Lolli, J. R. Lewis, Y. Gu, and E. J. Welton, 2016: Daytime cirrus cloud top-of-atmosphere radiative forcing properties at a midlatitude site and their global consequence. J. Appl. Meteor. Climatol., 55, 1667–1679, https://doi.org/10.1175/JAMC-D-15-0217.1.
Campbell, J. R., and Coauthors, 2018: Unusually deep wintertime cirrus clouds observed over the Alaskan subarctic. Bull. Amer. Meteor. Soc., 99, 27–32, https://doi.org/10.1175/BAMS-D-17-0084.1.
Chew, B. N., J. R. Campbell, J. S. Reid, D. M. Giles, E. J. Welton, S. V. Salinas, and S. C. Liew, 2011: Tropical cirrus cloud contamination in sun photometer data. Atmos. Environ., 45, 6724–6731, https://doi.org/10.1016/j.atmosenv.2011.08.017.
Dolinar, E. K., J. R. Campbell, S. Lolli, S. C. Ozog, J. E. Yorks, C. Camacho, Y. Gu, A. Bucholtz, and M. J. McGill, 2020: Sensitivities in satellite lidar-derived estimates of top-of-the-atmosphere optically-thin cirrus cloud radiative forcing: A case study. Geophys. Res. Lett., 47, e2020GL088871, https://doi.org/10.1029/2020GL088871.
Fu, Q., and K. N. Liou, 1992: On the correlated k-distribution method for radiative transfer in nonhomogeneous atmospheres. J. Atmos. Sci., 49, 2139–2156, https://doi.org/10.1175/1520-0469(1992)049<2139:OTCDMF>2.0.CO;2.
Fu, Q., and K. N. Liou, 1993: Parameterization of the radiative properties of cirrus clouds. J. Atmos. Sci., 50, 2008–2025, https://doi.org/10.1175/1520-0469(1993)050<2008:POTRPO>2.0.CO;2.
Garnier, A., J. Pelon, M. A. Vaughan, D. M. Winker, C. R. Trepte, and P. Dubuisson, 2015: Lidar multiple scattering factors inferred from CALIPSO lidar and IIR retrievals of semitransparent cirrus cloud optical depths over oceans. Atmos. Meas. Tech., 8, 2759–2774, https://doi.org/10.5194/amt-8-2759-2015.
Gu, Y., J. Farrara, K. N. Liou, and C. R. Mechoso, 2003: Parameterization of cloud–radiation processes in the UCLA general circulation model. J. Climate, 16, 3357–3370, https://doi.org/10.1175/1520-0442(2003)016<3357:POCPIT>2.0.CO;2.
Gu, Y., K. N. Liou, S. C. Ou, and R. Fovell, 2011: Cirrus cloud simulations using WRF with improved radiation parameterization and increased vertical resolution. J. Geophys. Res., 116, D06119, https://doi.org/10.1029/2010JD014574.
Heymsfield, A., D. Winker, M. Avery, M. Vaughan, G. Diskin, M. Deng, V. Mitev, and R. Matthey, 2014: Relationships between ice water content and volume extinction coefficient from in situ observations for temperatures from 0° to −86°C: Implications for spaceborne lidar retrievals. J. Appl. Meteor. Climatol., 53, 479–505, https://doi.org/10.1175/JAMC-D-13-087.1.
Holben, B. N., and Coauthors, 2018: An overview of mesoscale aerosol processes, comparisons, and validation studies from DRAGON networks. Atmos. Chem. Phys., 18, 655–671, https://doi.org/10.5194/acp-18-655-2018.
Holz, R. E., and Coauthors, 2016: Resolving ice cloud optical thickness biases between CALIOP and MODIS using infrared retrievals. Atmos. Chem. Phys., 16, 5075–5090, https://doi.org/10.5194/acp-16-5075-2016.
Hong, Y., G. Liu, and J. F. Li, 2016: Assessing the radiative effects of global ice clouds based on CloudSat and CALIPSO measurements. J. Climate, 29, 7651–7674, https://doi.org/10.1175/JCLI-D-15-0799.1.
Hurrell, J. W., T. Kushnir, and M. Visbeck, 2001: The North Atlantic Oscillation. Science, 291, 603–605, https://doi.org/10.1126/science.1058761.
Jin, M., and S. Liang, 2006: An improved land surface emissivity parameter for land surface models using global remote sensing observations. J. Climate, 19, 2867–2881, https://doi.org/10.1175/JCLI3720.1.
Jin, Z., T. P. Charlock, W. L. Smith Jr., and K. Rutledge, 2004: A parameterization of ocean surface albedo. Geophys. Res. Lett., 31, L22301, https://doi.org/10.1029/2004GL021180.
Lewis, J. R., J. R. Campbell, E. J. Welton, S. A. Stewart, and P. C. Haftings, 2016: Overview of MPLNET, version 3, cloud detection. J. Atmos. Oceanic Technol., 33, 2113–2134, https://doi.org/10.1175/JTECH-D-15-0190.1.
Lewis, J. R., J. R. Campbell, S. Lolli, S. A. Stewart I. Tan, and E. J. Welton, 2020: Determining cloud thermodynamic phase from the polarized Micro Pulse Lidar. Atmos. Meas. Tech., 13, 6901–6913, https://doi.org/10.5194/amt-13-6901-2020.
Lolli, S., and Coauthors, 2017: Daytime top-of-the-atmosphere cirrus cloud radiative forcing properties at Singapore. J. Appl. Meteor. Climatol., 56, 1249–1257, https://doi.org/10.1175/JAMC-D-16-0262.1.
Mace, G. G., and Q. Zhang, 2014: The CloudSat radar-lidar geometrical profile product (RL-GeoProf): Updates, improvements, and selected results. J. Geophys. Res. Atmos., 119, 9441–9462, https://doi.org/10.1002/2013JD021374.
Marquis, J. W., A. S. Bogdanoff, J. R. Campbell, D. L. Westphal, J. A. Cummings, N. J. Smith, and J. Zhang, 2017: Estimating satellite sea surface temperature retrieval cold biases in the tropics due to unscreened optically thin cirrus clouds. J. Atmos. Oceanic Technol., 34, 355–373, https://doi.org/10.1175/JTECH-D-15-0226.1.
Martius, O., C. Schwierz, and H. C. Davies, 2010: Tropopause-level waveguides. J. Atmos. Sci., 67, 866–879, https://doi.org/10.1175/2009JAS2995.1.
Pappalardo, G., and Coauthors, 2014: EARLINET: Towards an advanced sustainable European aerosol lidar network. Atmos. Meas. Tech., 7, 2389–2409, https://doi.org/10.5194/amt-7-2389-2014.
Parker, T. J., G. J. Berry, M. J. Reeder, and N. Nicholls, 2014: Modes of climate variability and heat waves in Victoria, southeastern Australia. Geophys. Res. Lett., 41, 6926–6934, https://doi.org/10.1002/2014GL061736.
Platnick, S., M. D. King, S. A. Ackerman, W. P. Menzel, B. A. Baum, J. C. Reidi, and R. A. Frey, 2003: The MODIS cloud products: Algorithms and examples from Terra. IEEE Trans. Geosci., 41, 459–473, https://doi.org/10.1109/TGRS.2002.808301.
Ramanathan, V., R. D. Cess, E. F. Harrison, P. Miins, B. R. Barkstrom, E. Ahmad, and D. Hartmann, 1989: Cloud-radiative forcing and climate: Results from the Earth Radiation Budget Experiment. Science, 243, 57–63, https://doi.org/10.1126/science.243.4887.57.
Rossow, W. B., and R. A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80, 2261–2288, https://doi.org/10.1175/1520-0477(1999)080<2261:AIUCFI>2.0.CO;2.
Rutan, D. A., S. Kato, D. R. Doelling, F. G. Rose, L. T. Nguyen, T. E. Caldwell, and N. G. Loeb, 2015: CERES synoptic product: Methodology and validation of surface radiant flux. J. Atmos. Oceanic Technol., 32, 1121–1143, https://doi.org/10.1175/JTECH-D-14-00165.1.
Sassen, K., 1991: The polarization lidar technique for cloud research: A review and current assessment. Bull. Amer. Meteor. Soc., 72, 1848–1866, https://doi.org/10.1175/1520-0477(1991)072<1848:TPLTFC>2.0.CO;2.
Sassen, K., and B. S. Cho, 1992: Subvisual-thin cirrus lidar dataset for satellite verification and climatological research. J. Appl. Meteor., 31, 1275–1285, https://doi.org/10.1175/1520-0450(1992)031<1275:STCLDF>2.0.CO;2.
Sassen, K., and J. R. Campbell, 2001: A midlatitude cirrus cloud climatology from the Facility for Atmospheric Remote Sensing. Part I: Macrophysical and synoptic properties. J. Atmos. Sci., 58, 481–496, https://doi.org/10.1175/1520-0469(2001)058<0481:AMCCCF>2.0.CO;2.
Screen, J. A., and I. Simmonds, 2010: The central role of diminishing sea ice in recent Arctic temperature amplification. Nature, 464, 1334–1337, https://doi.org/10.1038/nature09051.
Stephens, G. L., and P. J. Webster, 1981: Clouds and climate: Sensitivity of simple systems. J. Atmos. Sci., 38, 235–247, https://doi.org/10.1175/1520-0469(1981)038<0235:CACSOS>2.0.CO;2.
Stephens, G. L., S.-C. Tsay, P. W. Stackhouse Jr., and P. J. Flatau, 1990: The relevance of the microphysical and radiative properties of cirrus clouds to climate and climatic feedback. J. Atmos. Sci., 47, 1742–1753, https://doi.org/10.1175/1520-0469(1990)047<1742:TROTMA>2.0.CO;2.
Swap, R. J., H. J. Annegarn, J. T. Suttles, M. D. King, S. Platnick, J. L. Privette, and R. J. Scholes, 2003: Africa burning: A thematic analysis of the Southern African Regional Science Initiative (SAFARI 2000). J. Geophys. Res., 108, 8465, https://doi.org/10.1029/2003JD003747.
Thorne, P. W., 2008: Arctic tropospheric warming amplification? Nature, 455, E1–E2, https://doi.org/10.1038/nature07256.
Virts, K. S., and J. M. Wallace, 2010: Annual, interannual, and intraseasonal variability of tropical tropopause transition layer cirrus. J. Atmos. Sci., 67, 3097–3112, https://doi.org/10.1175/2010JAS3413.1.
Welton, E. J., J. R. Campbell, J. D. Spinhirne, and V. S. Scott, 2001: Global monitoring of clouds and aerosols using a network of micro-pulse lidar systems. Proc. SPIE, 4153, 151–158, https://doi.org/10.1117/12.417040.
Wielicki, B. A., B. R. Barkstrom, E. F. Harrison, R. B. Lee III, G. L. Smith, and J. E. Cooper, 1996: Clouds and the Earth’s Radiant Energy System (CERES): An Earth Observing System experiment. Bull. Amer. Meteor. Soc., 77, 853–868, https://doi.org/10.1175/1520-0477(1996)077<0853:CATERE>2.0.CO;2.
Winker, D. M., and Coauthors, 2010: The CALIPSO mission: A global 3D view of aerosols and clouds. Bull. Amer. Meteor. Soc., 91, 1211–1230, https://doi.org/10.1175/2010BAMS3009.1.
Yang, P., K. N. Liou, K. Wyser, and D. Mitchell, 2000: Parameterization of the scattering and absorption properties of individual ice crystals. J. Geophys. Res., 105, 4699–4718, https://doi.org/10.1029/1999JD900755.
Yang, P., H. Wei, H.-L. Huang, B. A. Baum, Y. X. Hu, G. W. Kattawar, M. I. Mischenko, and Q. Fu, 2005: Scattering and absorption property database for non-spherical ice particles in the near through far-infrared spectral region. Appl. Opt., 44, 5512–5523, https://doi.org/10.1364/AO.44.005512.
Yi, B., A. D. Rapp, P. Yang, B. A. Baum, and M. D. King, 2017: A comparison of Aqua MODIS ice and liquid water cloud physical and optical properties between collection 6 and collection 5.1: Cloud radiative effects. J. Geophys. Res. Atmos., 122, 4550–4564, https://doi.org/10.1002/2016JD025654.
Zhang, C., 2005: Madden–Julian Oscillation. Rev. Geophys., 43, RG2003, https://doi.org/10.1029/2004RG000158.