1. Introduction
Marine boundary layer (MBL) clouds are important contributors to Earth’s energy budget, due to their extensive spatial coverage, high albedo, and relatively warm cloud-top temperature (Hartmann et al. 1992). As a result, these clouds have long been subject to observational and modeling scrutiny (Albrecht et al. 1988; Stevens et al. 2005; Wood, 2012). An important climatological feature is the subtropical stratocumulus (Sc) to trade cumulus (Cu) transition (SCT), where shallow Sc-topped MBLs over cool coastal waters advect westward and equatorward in the trade winds, the boundary layer deepens and eventually the Sc cloud breaks up, leaving behind a cumulus MBL with lesser cloud fraction.
The dynamics of the marine cloud-topped boundary layer has a rich history of investigation. Early seminal work was performed by Lilly (1968), who explored cloud-topped MBL mixing, crucially describing the role of cloud -top radiative cooling in driving MBL mixing (in contrast to continental boundary layers predominantly driven by surface fluxes). Randall (1980) examined the role of cloud-top entrainment in generating additional instability and Sc breakup. Nicholls (1984) and Nicholls and Leighton (1986) explored in depth the decoupling of the cloud versus subcloud layers in marine stratocumulus, noting the role of cumulus clouds in recoupling the two layers. From a climatological perspective, Klein and Hartmann (1993) found that lower-tropospheric stability (LTS) correlated strongly with MBL cloudiness and highlighted the importance of a strong inversion for a cloudy MBL. The 1992 ASTEX field campaign provided a great deal of insight into the dynamics of the transition (Bretherton and Pincus 1995; Bretherton et al. 1995a,b), by noting the link between decoupling and breakup in two Lagrangian cases, and suggesting a possible microphysical role for aerosol in the transition in some cases (Albrecht et al. 1995). Krueger et al. (1995) reproduced the transition using numerical simulations forced only by increasing sea surface temperature (SST). Wyant et al. (1997) and Bretherton and Wyant (1997) developed a comprehensive model for the SCT, in which boundary layers advecting over warmer waters and away from subtropical maxima in subsidence deepen and decouple, developing a cumulus layer underneath the overlying stratocumulus. The cumulus further dries out the overlying stratocumulus by entraining free-tropospheric air through a weaker inversion, eventually leading to a cumulus-dominated MBL. This model (hereafter W&B model) also helps explain the LTS-cloudiness correlation found by Klein and Hartmann (1993). Wood and Bretherton (2006) define the estimated inversion strength (EIS), a refinement of LTS applicable over a broader range of temperatures.
Further studies of the relationship between MBL vertical structure and cloud evolution have also added to our understanding of the SCT. Consistent with the W&B model, Wood and Bretherton (2004) showed that satellite estimates of decoupling and boundary layer depth are well correlated. The free-tropospheric relative humidity (FT RH) affects low cloud cover by altering both the net longwave radiative cooling and the entrainment at cloud top (Mauger and Norris 2010; Kawai et al. 2017; Adebiyi and Zuidema 2018), and the magnitude of the effect of FT RH anomalies on cloud cover is sensitive to the vertical distribution of moisture (Eastman and Wood 2018). Large-eddy simulation (LES) studies focusing on cloud-top entrainment (e.g., Lock 2009; Xiao et al. 2011) suggest that Sc breakup is affected by the cross-inversion moisture jump as well as the temperature jump.
Observationally inferred relationships are dependent on the time scale considered for MBL evolution. Eastman et al. (2017) found that on seasonal time scales LTS and MBL depth are uniformly anticorrelated, consistent with W&B, but regions of positive correlation between LTS and depth emerge when the seasonal-scale variability is removed, possibly due to other drivers such as FT RH. That work and others have also highlighted the utility of a Lagrangian perspective for studying drivers of MBL evolution, as there is a lag in the MBL response to changing stability (Klein et al. 1995) and upwind factors must be controlled for (Mauger and Norris 2010). However, Eastman et al. (2016) showed that many MBL cloud properties decorrelate faster (in a day or less) following advecting boundary layer air columns (Lagrangian analysis) than in a fixed Eulerian perspective; it was postulated that this was due to the advection of the MBL across persistent but spatially variable cloud-controlling factors.
MBL aerosol concentration affects clouds, primarily by modulating the cloud droplet number concentration Nd (Twomey 1974). Through microphysical feedbacks, this will also influence cloud macrophysical evolution, sometimes called the “lifetime” effect (Albrecht 1989). The lifetime effect involves many small-scale cloud processes and depends on the cloud regime, so it is a major source of uncertainty in modeling anthropogenic aerosol effects on climate. Process modeling studies based on LES have identified two competing mechanisms, working through entrainment and precipitation, for liquid-phase MBL cloud lifetime effects. Ackerman et al. (2004), Jiang et al. (2006), Xue and Feingold (2006), Bretherton et al. (2007), Wood (2007) and others showed that higher Nd, through its impact on cloud droplet size, increases evaporation–entrainment feedbacks that reduce cloud liquid water content. These studies also noted that lower Nd accelerates the onset of precipitation in thicker clouds, which can decouple a Sc-capped MBL (Stevens et al. 1998) and may help transition Sc to Cu (Sandu and Stevens 2011; Berner et al. 2013; Yamaguchi et al. 2017). Both mechanisms are challenging to represent in global climate model parameterizations because of their dependence on interacting small-scale processes.
Observational constraints are required to evaluate modeling of aerosol–cloud–precipitation interactions, and numerous past studies have used in situ and remote sensing studies to this end. For instance, Bretherton and Pincus (1995) documented two ASTEX Lagrangian case studies. The more polluted air mass retained a persistent Cu under Sc structure despite being decoupled, while the cleaner one transitioned from stratocumulus to cumulus. However, observationally assessing aerosol effects on cloud evolution is challenging because meteorological differences between cases also strongly impact cloud formation. The aim of the Cloud System Evolution over the Trades (CSET) campaign was to provide a comprehensive in situ dataset that could be used to better assess aerosol–cloud interaction and our ability to model it, with Lagrangian sampling for many cases supporting an improved separation of meteorological and aerosol controls on MBL clouds.
The goal of this paper is to use this new set of aircraft observations to better understand and document cloud and boundary layer controlling factors in the SCT and provide some attractive case studies for detailed process modeling. The data were collected during the CSET field campaign (Albrecht et al. 2019), which took place 1 July–15 August 2015 in the northeast Pacific subtropical stratocumulus region. The primary observational platform for CSET was the NSF/NCAR Gulfstream V (GV), flying between Sacramento, California (CA), and Kona, Hawaii (HI). Low-level sampling was carried out throughout the depth of the MBL and lower free troposphere using a suite of in situ and remote sensing instrumentation. CSET employed a Lagrangian approach pairing outbound (CA–HI) flights to return (HI–CA) flights two days later. As we use these terms throughout the paper, note that in the context of Lagrangian evolution, the “outbound” data corresponds to the initial upwind sampling, and the “return” data corresponds to the downwind sampling 2 days later. MBL air masses sampled on the outbound flight were tracked using 500 m constant-height HYSPLIT trajectories, and then resampled again on the return flight. Many of these air masses transitioned from stratocumulus-capped on the outbound flight to primarily shallow cumulus on the resampling portions of the return flight, and a diversity of aerosol conditions were encountered. For an in-depth description of the campaign and instrumentation, see Albrecht et al. (2019).
In this paper, we present a Lagrangian view of the subtropical MBL evolution of CSET sampled/resampled air masses, combining aircraft, geostationary satellite, and reanalysis data. Our work complements an Eulerian analysis of the CSET observations presented in Bretherton et al. (2019), which uses compositing techniques to take advantage of the generally similar sampling during all the flight pairs and develop a statistical representation of the SCT. Bretherton et al. (2019) analyzes the mean state and variability of clouds, aerosols and precipitation along a nominal transect between California and Hawaii. Mean states for these quantities are used to assess the capabilities of reanalysis and reanalysis-nudged global climate model single-column output for this region. Bretherton et al. (2019) also include a comparison between satellite retrieved Nd and in situ sampling.
We first describe how the Lagrangian cases are constructed (section 2a), additional datasets (sections 2b and 2c), and how MBL variables are estimated (section 2d). A statistical analysis of overall MBL Lagrangian evolution from all CSET cases and from the expanded trajectory set are presented in section 3a, while two contrasting case studies recommended for future modeling are presented in section 3b. A discussion of results and conclusions follow in section 4.
2. CSET measurements and data analysis methods
a. Lagrangian case data and methodology
1) CSET flights and data
As described by Albrecht et al. (2019), the CSET flights consisted of a ferry leg (i.e., high-altitude transit flight) to reach an area of interest in the remote MBL, followed by approximately 1500 km of low-level sampling, and then another ferry leg to complete the transit to either Kona (outbound flights) or Sacramento (return flights). The low-level sampling consisted of a series of 3–5 sequences of a fixed set of flight maneuvers per flight (Fig. 1, inset): descent to 150 m above the sea surface for a 10-min “below-cloud” level flight leg, ascent to 10-min “in-cloud” level leg just above cloud base, 10-min “porpoise leg” (i.e., repeating sawtooth) through cloud top and inversion layer, and ascent to 1500 m above cloud top for 10-min “above-cloud” level leg, then descending again to begin the next sequence. Each sequence lasted approximately 50 min and sequences were lettered alphabetically, with A being the first sequence of each flight, B the second, etc. An example of a CSET flight pair is shown in Fig. 1. This figure shows the outbound flight RF08 from California to Hawaii, tracking trajectories initialized on the RF08 flight path, and subsequent RF09 return flight intercepting the trajectories (for CSET, even-numbered flights were outbound CA-HI, and odd-numbered flights were return HI-CA). Distortion of the sampled line due to regions of convergence and divergence resulted in a dog-leg (kink in flight path) in RF09. The flight plans were chosen so that the return flight resampled 500 m constant-level trajectories from the outbound flight, calculated as discussed in section 2b.
Example of one of seven CSET mission pairs, here showing RF08/RF09, flown on 22 Jul and 24 Jul 2015, respectively. Ferry legs are shown in black, while low-level sequences are the blue-and-orange portions of each flight. Tracking trajectories initialized on the outbound flight path are shown in red, with a dot marking every 6 h of advection. (inset) CSET flight sequence, showing progression though below-cloud, in-cloud, porpoise, and above-cloud legs relative to cloud base and cloud top.
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
GV in situ instrumentation (Albrecht et al. 2019) included a standard set of temperature and humidity probes, CO and O3 measurements, an Ultrahigh Sensitivity Aerosol Spectrometer (UHSAS) measuring size-resolved aerosol from 60 to 1000 nm diameter, a Cloud Droplet Probe (CDP, measuring 2–50 μm diameter drops), precipitation probes and broadband radiometers. For remote sensing, the plane was equipped with the HIAPER Cloud Radar (HCR) and a High Spectral Resolution lidar (HSRL), both of which could be aimed up or down, as well as broadband radiometers and dropsondes.
2) Construction of CSET Lagrangian cases
In the context of the CSET campaign, “Lagrangian” refers to following the MBL air masses that were tracked by trajectories and twice sampled by the aircraft. To organize the flight and trajectory data into 18 Lagrangian cases (2–3 per flight pair), the tracking trajectories were used to identify that portions of the outbound flight were resampled by which portions of the return flight. An additional constraint was that each case had to contain at least one full sampling sequence (below-cloud leg through above-cloud leg) on both the outbound/upwind and return/downwind flight. Due to variability in divergence of the sampled air mass, either the outbound or return aircraft data could span a longer distance. The resulting cases represented air masses on the scale of 500 km and spanned a significant amount of mesoscale variability. Cases are numbered in chronological order and are referenced as L01–L18 (for Lagrangian cases 1–18). Collocated satellite measurements and reanalysis data along the flight paths and the connecting trajectories complement the aircraft observations. Two such cases are shown in Fig. 2. Maps for other cases can be found in the supplemental figures in the online supplemental material, along with a table showing key BL variables (cloud fraction, Nd, stability, etc.) for every case. As an example, the case in Fig. 2a, L06, consists of aircraft data from the middle of RF06 (sequences B and C, thick blue line), which was resampled at the beginning of RF07 (sequence A, thick orange line). These two flight portions were connected by 5 tracking trajectories (T1.6–T2.6, thinner colored lines). “For both outbound and return flights, aircraft data are averaged over the appropriate legs (e.g. below-cloud or in-cloud) from all relevant sequences (e.g. RF06 BC, RF07 A).”For instance, the Lagrangian change in droplet number concentration (Nd) for L06 is the difference between the initial aircraft Nd from RF06, sequences B and C, in-cloud legs, and the final aircraft Nd from RF07, sequence, A, in-cloud leg. Trajectory data are averaged over all trajectories in the case except where explicitly noted. For instance, the mean subsidence experienced by L06 is the 2-day time-mean ERA5 subsidence along all 5 trajectories T1.6–T2.6 from RF06/07, then again averaged over trajectories (see section 2d for more detail on subsidence and entrainment).
Lagrangian case studies (a) L06 (from RF06/RF07) and (b) L10 (from RF10/RF11) showing the location of the flight portions (thicker blue/orange lines) and airmass trajectories (thinner lines). Insets show enlarged GOES cloud field (from a 2° × 2° box centered on trajectories 2.3 and 6.0, respectively) at the time of initial sampling, a day later, and two days later at the time of resampling. Hour 49 is used instead of 48 due to partially missing data. Pink circles indicate sequences of flight data.
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
To support future process modeling studies and to highlight the range of MBL cloud evolution sampled in CSET, we present in detail two Lagrangian case studies shown in Fig. 2: L06 in Fig. 2a, and L10 in Fig. 2b, (corresponding to the easternmost air mass from RF10, sequence A, which was resampled on RF11 DE). Both cases start out cloudy and experience a strong decrease in EIS, but while the L06 case experiences a classic SCT, the L10 case remains cloudier. The two cases differ in initial MBL depth, decoupling, microphysics, and large-scale forcings. Note that the agreement between outbound flight number and case number (L06 coming from RF06, L10 from RF10) is coincidental.
b. Satellite and reanalysis data
In addition to the GV measurements, satellite and reanalysis data are used. Where possible, these are compared to aircraft observations for additional validation. The satellite data consists of cloud property fields generated from Geostationary Operational Environmental Satellite-15 (GOES-15) observations using algorithms originally developed for the NASA Clouds and the Earth’s Radiant Energy System (CERES) project (Minnis et al. 2008a, 2011) and adapted for application to other imagers on geostationary satellites (Minnis et al. 2008b). Except where noted, all retrievals are averaged over a 2° × 2° box centered on the coordinates in question (e.g., aircraft or trajectory position) and are at an hourly resolution. The 2° box was empirically chosen to balance the competing interests in reducing noise in box-averaged quantities while avoiding including observations from regions subject to significantly different large-scale forcings; a comparison of the GOES cloud fraction estimate to that derived from a radar-lidar cloud mask can be found in Bretherton et al. (2019).
Supplemental data are drawn from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis, version 5 (ERA5) [Copernicus Climate Change Service (C3S) 2017], for large-scale atmospheric conditions. Trajectories are generated using the Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT; Stein et al. 2015), using the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) forecasts and the Global Data Assimilation System (GDAS) analyses as provided by the National Oceanic and Atmospheric Administration Air Resources Laboratory (NOAA ARL) archive.
c. Expanded trajectory set
During CSET, 53 trajectories within the 18 cases were used in total to track sampled air masses. As the satellite and reanalysis data are not restricted to the context of the CSET flights, we generated a larger dataset of approximately 1800 trajectories, spanning July–August 2015. This is used in section 3a to validate the representativeness of the CSET sampled air masses and provides a large sample size for testing certain hypotheses that do not rely on any aircraft data. The trajectories were initialized at 0000 UTC for every day in the two-month span, on the same grid used for CSET flight planning. The resulting trajectories were filtered for direction to exclude 800 trajectories going toward land or the extratropics. Satellite and reanalysis data were extracted along the remaining 1000 trajectories.
d. Calculation of MBL variables
For aircraft data, some variables are drawn only from certain legs (such as below-cloud) in each sequence and some are only from cloudy observations (defined as RH > 95% and CDP liquid water content ql > 0.01 g m−3) or conversely noncloudy observations. Aircraft cloud droplet number concentration (Nd) is measured by the CDP and averaged only over the cloudy time steps on the in-cloud leg. Accumulation-mode aerosol concentration (all particles with diameters between 100 and 1000 nm) is measured by the UHSAS and averaged over the below-cloud leg. Aircraft SST estimates are averages of the radiometric surface temperature from the below-cloud leg. Carbon monoxide (CO) and ozone (O3) measurements were averaged over the below-cloud leg as well. Note that a particular case may span 300–600 km of sampling, especially on the return flight in which it may encompass multiple sequences. Across this distance, significant mesoscale variability is inevitably encountered, and is only partly smoothed by leg averaging. This adds unavoidable noise to relationships between cloud and environmental conditions.
Cloud fraction is defined as the fraction of GOES pixels in a 2° × 2° box (centered on aircraft location or trajectory) that have successful retrievals of cloud type that are not clear sky. We additionally only focus on warm low liquid cloud fraction by selecting for cloud tops under 4 km with liquid tops. A random overlap assumption is made for pixels masked by deep/high clouds. Cloud droplet number concentration is calculated from GOES retrievals following Painemal and Zuidema (2011) using cloud optical thickness and cloud effective radius. Both CF and Nd are averaged along all outbound and return sequences from that Lagrangian case, and for Nd, the 75th percentile of values for warm low liquid cloud pixels is used to mitigate for possible biases caused by subpixel cloud inhomogeneity. Cloud fraction, as well as cloud-top height, span the full diurnal cycle, whereas Nd and liquid water path retrievals are daytime only.
Inversion base height zi (which typically is very close to the top of any extensive MBL stratocumulus), decoupling, and cross-inversion temperature and moisture jumps are calculated from the aircraft profiles, down from either the above-cloud or ferry leg to the below-cloud leg. As such, the number of profiles in each case (outbound or return) is equal to the number of sequences (e.g., case L10 has one profile from the outbound flight and two from the return flight).The estimation of the inversion base height zi, inversion layer, jumps in q and T across the inversion, and decoupling are all related. The appendix describes in detail how these quantities are measured from aircraft profiles and reanalysis. In brief, the boundary layer inversion middle is identified as the level of greatest lapse rate (after smoothing, forced between 300 and 3000 m). The inversion base height, zi, is the nearest level below the center where the lapse rate is (nearly) moist adiabatic, and the inversion top height is the nearest level above the center where the same is true. This method was found to work well with both aircraft profiles and reanalysis, with only a small bias in the inversion center height (see the appendix for details). For estimates of decoupling (see below), the boundary layer top is the level of the inversion base, and the free troposphere begins at the inversion top. Two variables that have been argued to modulate Sc breakup are the jumps across the inversion in temperature T and specific humidity q (e.g., Randall 1980; Lock 2009; Kawai et al. 2017). The jump in specific humidity Δq is taken to be the difference between the value at the inversion base and top. For the temperature jump ΔT, to account for the fact that inversions are sometimes gradual, we consider the difference in temperature if a parcel at the inversion top is brought down to the inversion base along a moist adiabat.
Decoupling is estimated following Wood and Bretherton (2004), and LTS/EIS following Wood and Bretherton (2006). Both LTS and EIS are very similar metrics. We concentrate on using EIS as it is slightly closer to the physical state variable of interest (the strength of the MBL top inversion), though LTS is used to allow comparison to previous literature. For decoupling, though there are a number of comparable metrics with similar qualitative behavior, here we use the αqt (total water alpha parameter), which varies from 0 (perfectly coupled) to 1 (upper MBL qt is equal to FT qt) and measures the degree of stratification of the q profile (see Wood and Bretherton (2004) for a more in-depth definition). In a few cases where the liquid water content was unavailable due to instrumentation issues, only water vapor is used. This has a very minor effect on the decoupling estimates as the vast majority of MBL atmospheric water is vapor.
Subsidence is calculated from ERA5 as follows: for each trajectory in a case, the reanalysis inversion layer is estimated. A time series of subsidence between outbound and return sampling is extracted as the mean value height-averaged over this layer, from which a 2-day mean is then calculated. Since this is entirely based on reanalysis, potentially poorly constrained, and a key variable for our discussion, we include an estimate of the uncertainty in this mean subsidence value by making use of the ERA5 10-member ensemble, as well as the fact that we potentially have multiple trajectories per case. Therefore, the distribution of 2-day mean subsidence values over the ensembles and trajectories is calculated, and the standard deviation of this is taken as the uncertainty in the mean. The mean entrainment rate for each case is calculated by adding the subsidence estimate that to the mean rate of change of the boundary layer depth calculated using initial and final zi estimates from the aircraft profiles (and therefore the subsidence uncertainty can be taken as a reasonable lower bound on the entrainment rate uncertainty). SST, surface fluxes, and 700 hPa free-tropospheric RH are taken from ERA5 reanalysis and averaged along the entire set of flight sequences making up a case.
Cloud-base rain rates are estimated by combining HCR and HSRL data collected during surface legs based on the technique proposed by O’Connor et al. (2005), as detailed in Schwartz et al. (2019). The precipitation rate for each leg was taken as the mean of column-maximum precipitation from all below-cloud legs, the height of which typically corresponds well with cloud base, typically between 400 and 800 m MSL. Columns with no precipitation were included as zero values (the precipitation rate is not rain conditional). The flight height of 500 ft (150 m MSL) meant that the cloud base was usually above the radar dead zone (<200 m from the plane).
3. Results
a. Characteristic Lagrangian evolution
In this section results are compared across all 18 cases considered together to look for robust features of the Sc–Cu transition. We derive mean changes averaged over all cases to define what a “characteristic” 2-day evolution of the subtropical cloud-topped MBL might look like. These values are shown in the final column of Table 1 as well as highlighted in Figs. 3 and 4 .
Means of key variables for two case studies and over all 18 cases. Shown are outbound/return values and differences, as well as mean values along trajectories (bottom 3 rows). Values from cases are bolded if they fall outside the middle 80% of all values.
Lagrangian coherence of (a) ozone, (b) carbon monoxide, (c) inversion height, (d) decoupling index, (e) GOES cloud fraction, (f) cloud droplet number concentration Nd, (g) cloud-base precipitation rate, and (h) accumulation mode aerosol number concentration Na. The x axis is the outbound flight value, and the y axis is the return flight value. Also plotted is the 1–1 line (dashed gray; this is meant as a quick guide to whether values increase or decrease between samplings); r values for a linear best-fit are in the top left. Red numbers mark the two Lagrangian case studies discussed in section 3b (L6 and L10), and black stars mark mean value across all cases.
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
MBL evolution over two days of (a) cloud fraction vs EIS, (b) decoupling vs depth, (c) cloud fraction vs cloud droplet number concentration, (d) accumulation mode aerosol number concentration vs cloud droplet number concentration, (e) precipitation vs droplet concentration, and (f) cross-inversion moisture jump vs cross-inversion temperature jump. Blue dots show values from outbound flights, and red show values from return flights. Thick black line with larger dots shows mean behavior. Red and blue numbers mark the two Lagrangian case studies discussed in section 3b (L6 and L10).
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
1) Lagrangian coherence of MBL properties
As CSET resampled air masses after two days, we can examine the correlations of various variables sampled on the outbound and return flights to analyze the persistence or Lagrangian coherence of anomalies. Assuming a perfect resampling of an air mass, the correlation of outbound with return values should indicate how well-preserved anomalies are over time. We first look at the Lagrangian coherence of long-lived chemical tracers (ozone and carbon monoxide). Figures 3a and 3b show a strong correlation between outbound and return below-cloud measurements of both tracers for all 18 cases, which lends confidence to the Lagrangian measurement strategy. Deviations from a perfect correlation are expected due to entrainment into the MBL of air with different properties, chemical processing within the MBL, and air–sea fluxes. In particular, ozone in the lower troposphere has a relatively short lifetime of only a few days (Seinfeld and Pandis 2006), and so while it does not at as a perfect tracer, anomalies should persist long enough to correlate after 2 days.
A similar analysis on MBL properties of depth, decoupling, cloud fraction, cloud droplet number concentration, precipitation rate, and aerosol concentration shows relatively low Lagrangian coherence (Figs. 3c–h). This analysis is consistent with the finding of Eastman et al. (2016) that these properties have e-folding time scales of 12–24 h, much less than the two day resampling time. The aerosol number concentration is an apparent exception, but the correlation coefficient drops to 0.52 if one very high-aerosol anomaly is removed. The mean precipitation estimates for the cases were highly sensitive to the how precipitation was defined. A more thorough analysis of the precipitation observations from CSET, including comparison of in situ and remotely sensed values, an evaluation of changes in precipitation frequency as well as mean rain rate, and a discussion of the role of precipitation in the transition, can be found in Sarkar et al. (2019, manuscript submitted to Mon. Wea. Rev.). In particular, they find that in situ observations of precipitation rate are higher than those estimates from the radar-lidar retrievals.
We conclude that while it is simpler to think about MBL evolution following air columns, as it matches the conceptual model of an air mass subjected to changing forcings, that evolution is in practice no more coherent than the spatial gradients of MBL and cloud properties between CA and HI.
2) Transition features
We use two metrics to describe composite changes across our set of Lagrangian cases. The first is the mean difference in a variable between the outbound and return sampling over all cases (e.g., cloud fraction on average decreased 31%, from 84% to 53%), and the second is the number of cases experiencing an increase/decrease in a variable (e.g., 16 out of 18 cases experienced a decrease in cloud fraction. Note that the total number of cases can be affected by missing data). These data are also shown in Fig. 4, which shows the evolution of all 18 cases in terms of various variable pairs. Each panel shows each case as a line between outbound (blue) and return (red) values. This highlights both the mean behavior as well as case-by-case variability.
In general these two metrics consistently show that most air masses undergo a “classic” stratocumulus to trade cumulus transition, including MBL deepening (13/18 cases, mean inversion base [zi] change is +560 m), increased decoupling (14/17, +0.12), decrease in cloud fraction (16/18, −31%), decrease in EIS (18/18, −2.2 K), and decrease in Na (10/17, −14 cm−3) (Figs. 4a–d). On average, the moisture inversion becomes more pronounced ([Δq] decreases by 0.9 g m−3, and 14/18 cases show a decrease), while the temperature inversion weakens ([ΔT] decreases by 0.9 K, 12/18 cases show a decrease) (Fig. 4f).
The downstream changes in precipitation rate and cloud droplet number concentration Nd (Fig. 4e) are much less consistent. For precipitation, 14/18 cases showed a decrease, but the mean change was only −1.5 mm day−1 with a large spread. The decrease in precipitation can be explained with reference to Bretherton et al. (2019), their Fig. 10c; the local minimum at −143°E along the CSET transect in precipitation corresponds with the mean location of resampling. Translated into the longitude–prime coordinate used for compositing aircraft data in Bretherton et al. (2019), the initial samplings were on average at −134°E and the return samplings at −144°E. An in-depth analysis of the Lagrangian evolution of precipitation for a subset of CSET cases can be found in Sarkar et al. (2019, manuscript submitted to Mon. Wea. Rev.), which contains a thorough characterization of precipitation frequency and rate, and also compares in situ and radar-sensed precipitation estimates. For Nd, the mean change was a decrease of −14 cm−3, consistent with the decrease in Na, but this was dominated by a few initially polluted cases with large decreases in Nd, and in fact 9/14 cases showed a slight increase in Nd.
The ubiquity of the classic SCT evolution could partly be due to the sampling strategy in CSET. Flights were targeted toward low-level sampling of initially cloudy MBLs that were forecast to advect down the EIS gradient to the southwest. One way of assessing to what extent the CSET sampling strategy biased the sampled airmass population is by comparing satellite data along the CSET trajectories to those from the expanded trajectory set, as shown in Figs. 5a–e. The time offset of the CSET trajectories is due to the difference in initialization time, as trajectories from the larger set were initialized at 0000 UTC while CSET trajectories were usually sampled around 1800 UTC (local morning) on both the outbound and return flights. CSET outbound flights preferentially sampled trajectories originating at the northern end of the sample area. Although we do not have data to assess the representativeness of the decoupling state or aerosol environment of the boundary layers sampled during CSET, we can compare the evolution of GOES-retrieved cloudiness, cloud-top height and Nd, and ERA5-inferred stability between the CSET and expanded populations of trajectories. Although all fields show high variability, there does not appear to be a strong bias in the CSET sampled trajectory set. In particular, due to the filtering of the expanded trajectory set by direction, the evolution of EIS along both sets of trajectories is very consistent, indicating that the CSET sampled trajectories were not much different from a random set within the observational period, given the constraints on direction and initial location, though we cannot say whether they are more broadly climatologically representative given the strong possibility of significant interannual variability. Compared to the trajectory set used in Sandu et al. (2010) and subsequent work, the trajectories used here are farther to the northwest, with more of an easterly wind. They therefore experienced a weaker gradient in stability, which is consistent with the overall slower evolution in cloudiness in our set.
(a) Expanded trajectory set with evolution over 3 days of (b)–(e) GOES-retrieved variables and (f) ERA5 EIS. Red lines show trajectories sampled during CSET, black lines show all summer 2015 trajectories. Plotted are mean values averaged over all trajectories (thick lines with dark shading) and bounding quartiles (thin lines). Trajectories are shifted in time such that the local solar times at start align. In (a), trajectories go from NE to SW, and a random 50-member subset of the expanded trajectory set is shown [full set used for (b)–(e)].
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
Figure 5 also gives some indication of the magnitude of the diurnal cycle of subtropical MBL cloud-top height and fraction. These values are obtained by first detrending and then fitting a sinusoid to the mean trajectory in Fig. 5, and the stated values are twice the fitted amplitude. Both maximize in the early morning and decrease during the day, cloud-top height by 150 m and cloud fraction by 13%. Nighttime deepening and infilling counteract both daytime trends, though the general behavior is still a net decrease in cloud fraction and deepening of the cloud layer toward Hawaii. The diurnal cycle of both depth and cloud fraction in marine stratocumulus is consistent with previous literature (e.g., Rozendaal et al. 1995; Burleyson et al. 2013; Painemal et al. 2013), and a more in-depth discussion of the diurnal cycle as it relates to the transition in CSET can be found in Sarkar et al. (2019, manuscript submitted to Mon. Wea. Rev.). It is worth noting that the aircraft observations, which all occur within a few hours of local noon, do not sample this diurnal variability. This expanded capability highlights the complementarity of the satellite and aircraft sampling. Care must be taken when examining the diurnal cycles of liquid water path and Nd. Though some of the measured variability is undoubtedly physical, there is also a known solar zenith angle bias (SZA) in these retrievals (Grosvenor and Wood 2014). This is mitigated by filtering for SZA above 70° in Fig. 5.
3) Controls on evolution
Past work has explored the extent to which changes in MBL clouds are consistent among variables and whether such changes are predictable. We can validate some of that work using data presented here. The cloud fraction–stability relationship (as measured by EIS or LTS) is well studied, primarily in an Eulerian framework. Klein and Hartmann (1993) used geographical, seasonal and interannual changes in stratus cloud cover and LTS to estimate a 6% K−1 sensitivity of marine low stratus to LTS; Wood and Bretherton (2004) showed a similar sensitivity for EIS. Here we calculate the EIS/CF relationship using the CSET Lagrangian cases. The slope of the “mean Lagrangian” line (thick line) in Fig. 4a, representing the average behavior or the cases in CF/EIS space, is 14% K−1, steeper than those past climatological sensitivity estimates. Alternatively, we can consider the range of CF/EIS slopes for all 18 cases, which have a median and standard deviation of 15% and 21% K−1. CSET deliberately targeted Sc farther offshore of persistent coastal clearing (low CF) due to offshore flow over cold water (high EIS). Inclusion of such regions would likely lower the mean sensitivity more into line with the climatological sensitivity. Another possible explanation for the slope discrepancy is uncertainty in the CSET slope estimate, since it has a large range across individual cases. However, our derived sensitivity compares well to an estimate from a similar satellite study by Sandu et al. (2010) that considered approximately 8500 trajectories in the NEP. Based on their Figs. 2 and 3, considering trajectory evolution from day 1 to 3 so that the starting cloud fraction is closest to our sample, the mean cloudiness decreases 55% and LTS decreases 3.7 K over two days, implying a sensitivity around 15% K−1. This is despite the weaker LTS gradients in our sample.
The relationship between decoupling and depth is also strong, as shown in Fig. 4b. The slope of the mean trajectory behavior (thick line) is approximately 0.21 km−1 (i.e., the decoupling index increases by 0.21 for every km of deepening). The estimate of this slope as provided in Wood and Bretherton (2004) is 0.28, within estimation error (we used a bootstrap method to generate 1000 samples of 18 cases and estimated the mean slope as 0.22 ± 0.06 km−1).
Although the aircraft resampling only allows for studying of 2-day lagged correlations, the satellite and reanalysis data along trajectories has much higher time resolution, so it can give more insight into the time scales of evolution. Figure 6 shows the lag correlation of GOES CF with EIS using the expanded trajectory set. Each hour represents the correlation among all trajectories between EIS lagged that many hours behind/ahead of the GOES CF 2 days into the trajectory (e.g., the 0-h lag correlation is the instantaneous correlation of EIS and CF 2 days after trajectory initialization). Also shown are lag correlations replacing CTH for CF. CF correlates best with upstream EIS 30 h earlier, strongly suggestive of EIS control on CF and consistent with Klein et al. (1995). No such lag exists in the EIS–cloud-top height correlations, though whether this is due to a more immediate response of the MBL depth to large-scale meteorology is uncertain.
Lag correlations of EIS and GOES cloud fraction/cloud-top height. Peak correlation for CF/EIS was at roughly −30 h lag (i.e., upstream EIS with downstream CF). For CTH, instantaneous correlations were strongest.
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
As both entrainment and precipitation are important processes in MBL evolution, and aerosol can affect both of these through modulation of the cloud droplet number concentration, there is potential for strong aerosol influence on MBL evolution. The CSET data shows a strong correlation (r = 0.72) between below-cloud Na and in-cloud Nd (Fig. 4d). However, there is no evidence of a significant aerosol influence on precipitation or entrainment rates. This is most likely due to the small sample size and the resulting inability to control for additional confounding factors such as MBL depth and large-scale meteorology, and not a general conclusion about the existence of indirect aerosol effects. Although a positive correlation was found between GOES Nd and the GV CDP Nd measurements (see Bretherton et al. 2019), use of the expanded trajectory set and their GOES Nd retrievals to investigate cloud lifetime effects is problematic. One reason for this is a possible cloud fraction-dependent bias related to the validity of the plane-parallel, adiabatic, or lapse rate assumptions applied in the retrieval of Nd for cumulus clouds. Correlations between Nd and GOES cloud fraction disappear when considering aircraft Nd instead of GOES, and so we cannot state conclusively where there is an aerosol effect on cloud evolution present in the CSET Lagrangian cases.
b. Lagrangian case studies
In this section, we take a closer look at two Lagrangian cases observed in CSET, both initially similarly cloudy. L06 is shallow, and relatively clean, while L10 is deeper, more decoupled, and more polluted. The “slowly evolving” L10 case remains deep, decoupled, and cloudy, with little development toward a cumulus-dominated boundary layer, while the “rapidly evolving” L06 case experiences a rapid transition from stratocumulus into cumulus cloud involving deepening and decoupling. We explore the differing evolution in these two cases and document them to encourage their future use for process modeling studies.
In L06, cloud fraction decreases from 95% to 39%, the MBL deepens from 750 to 2370 m, and the decoupling parameter αq increases from 0.13 (well-coupled) to 0.34 (moderately decoupled). In contrast, in L10, cloud fraction decreases from 100% only to 78%, the MBL slightly deepens from 1240 to 1700 m, and αq goes from 0.33 to 0.47 (moderately to heavily decoupled). These cases are marked as “6” and “10” in Figs. 3 and 4 from the previous section. Table 1 highlights key variables from these two cases, as well as the average values over all 18 cases. Values that fall in the top or bottom decile of all 18 cases are bolded to highlight ways in which these cases were unusual.
1) Meteorological environment
Much of the difference between these two cases may be attributable to their meteorological environment and MBL forcings. Though it would be ideal to have two cases with only one crucial difference (e.g., aerosol environment), that is not achievable in a nature, and so we explore the large-scale forcings on these two cases first.
L06 was initially stable and experienced a moderate decrease in EIS. L10, though starting with a higher EIS, experienced a strong decrease and ended up with a similar EIS to L06. This can be seen in Fig. 4a, the L06 line is much steeper than the L10 line. The free-troposphere relative humidity was initially higher in L06 but ended higher in L10, such that the two cases experienced entirely opposite trends in FT RH (see Figs. 7b,e). L10 is unusual within our sample of cases, with its initial and final FT RH being anomalously low and high, respectively.
Soundings and relative humidity cross section from L06 and L10. (a),(d) RH and theta from aircraft initial soundings along with inferred inversion base height (dotted lines; also marked on center panels with arrows). (b),(e) Evolution along trajectories (L06: trajectory 2.3; L10: trajectory 6.0) of RH from ERA5. RH curtain (RH as function of height and time) is only from one trajectory, not averaged over all case trajectories, for clarity. Toward the start and end of the trajectory (left and right ends of each panel), the ERA data are blanked out with a gray box and the aircraft RH data are overplotted, showing generally good agreement with ERA5 with regards to the height of the strongest RH gradient. White line marks GOES cloud-top height (75th percentile from 2° × 2° box along trajectory); line is dotted where CF<50%. (c),(f) As in (a),(d), but for return soundings.
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
An important difference between these two cases is the large-scale vertical motion (Fig. 8d). L10 had a strong mean subsidence rate at the inversion (from ERA5) of approximately 3.5 ± 0.6 mm s−1, while L06 very weak subsidence of only 0.2 ± 0.6 mm s−1, consistent with its rapid deepening and decoupling. This plausible conclusion has two major sources of uncertainty. First, we do not have an independent test of the realism of the large oscillations in the ERA5 vertical motion following the trajectories in both cases, let alone their time average. Second, there was a strong vertical gradient in ERA5 vertical motion during L06, which makes the vertical velocity at the inversion base sensitive to uncertainties in how the inversion base is estimated. In fact, subsidence at 700 hPa averaged along the trajectories was comparable for both cases (between 3 and 3.6 mm s−1).
Evolution of cases along trajectories with aircraft validation: (left) L06 and (right) L10. Each line represents data along a different trajectory for that case, and red dots mark aircraft data with interquartile range. (a) GOES low warm cloud fraction, (b) GOES 75th percentile Nd with aircraft CDP Nd, (c) ERA 700 mb specific humidity, (d) ERA subsidence averaged over inversion layer, (e) ERA 1000 mb wind speed with aircraft below-cloud leg (150 m) wind speed, (f) ERA SST with aircraft radiometric surface temperature from below-cloud leg, and (g) ERA surface latent heat flux. Gray shading indicates night time, and thicker lines (trajectories 2.3 and 6.0) are highlighted for comparison with other plots.
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
The entrainment rate for the two cases is estimated by subtracting the vertical motion at the inversion base from the inversion deepening rate. Using the inversion heights from aircraft profiles (Fig. 7), the mean entrainment rates for L06 and L10 were 9.5 and 5.9 mm s−1 (±0.6, compared to an average across all cases of 5.2 mm s−1). The very high L06 entrainment rate, in addition to weak subsidence, contributed to its rapid inversion deepening. Figures 7a and 7c show that the L06 inversion height change from the aircraft profiles is higher than what GOES cloud-top height or ERA5 RH might indicate (in particular considering the inversion height at the start of L06), and a more conservative estimate based on visual inspection of Fig. 7 would be closer to 7 mm s−1. Regardless, L06 shows stronger entrainment, likely due to the weaker inversion, even when considering the uncertainty in subsidence estimates.
Figures 7b and 7e also compare the GOES cloud-top height (white line), the aircraft inversion (colored triangles) and the ERA5 humidity inversion (transition between purple and orange shading). Note that for clarity, the ERA RH curtain is only drawn from one trajectory per case and not averaged over all trajectories. This is for a more accurate portrayal of the ERA5 vertical structure. When the MBL is deep and cloud fraction is low (marked by dash in GOES cloud-top height), much of the cumulus cloud can be seen through the disappearing deeper stratus and the GOES cloud-top height drops well below the other inversion base estimates (consistent with Karlsson et al. 2010), though a possible bias in GOES cloud-top height estimates in broken scenes due to surface contamination (Zuidema et al. 2009) has not been ruled out. This effect also has a strong diurnal signal from daytime stratocumulus burn-off, which complicates naive use of cloud-top height as a proxy for boundary layer depth.
The differences in surface latent heat fluxes (LHF) shown in Fig. 8 can be explained by considering the SSTs and near-surface wind speeds. The former was higher in L06, consistent with its location slightly farther along the climatological SST gradient. This resulted in an initially higher LHF, though the decrease in wind speed toward the latter half of L06 resulted in the LHF dropping below the L10 values. Averaged over the two days between sampling, the L06 LHF was ultimately above average, and conversely L10 LHF was below average. As latent heat fluxes are a key driver of MBL decoupling (Bretherton and Wyant 1997), the increase in L06 decoupling and relative lack of further decoupling of L10 is consistent with these LHF anomalies.
2) Aerosol environment
The aerosol environment also differs between the two cases. Relative to the ensemble of cases, L06 was unusually clean while L10 was unusually polluted by every measure, including initial Na and Nd (as shown in Fig. 4d), CO, and O3 (as shown in Figs. 3a,b).
There are a number of potential consequences of the more polluted environment of L10 on the cloud evolution. According to LES, an initially higher Nd should increase entrainment while decreasing cloud-base precipitation, with an uncertain net effect on cloud liquid water path (Stevens et al. 1998; Ackerman et al. 2004). Though initially under a free-troposphere dry relative to the case average, the elevated FT RH toward the second half of L10 (Figs. 7e and 8c) preferentially pushes the MBL toward a regime where Sc can thicken and the precipitation effect should begin to dominate (Ackerman et al. 2004; Yamaguchi et al. 2017). However, observed precipitation rates were comparable in L10 to L06, and the entrainment rate was higher in L06 rather than L10. This suggests that confounding meteorological differences between L10 and L06 are overwhelming potential aerosol effects. Sarkar et al. (2019, manuscript submitted to Mon. Wea. Rev.) investigated differences in precipitation in the flight data in more depth and found that the precipitation reaching the surface was greater in the L06 case, consistent with the cleaner boundary layer.
The Na/Nd coevolution in case L10 (Fig. 4d) shows a very strong decrease in Nd with no significant change in Na. This can be explained by considering the flight legs from which these measurements were taken and degree of MBL decoupling. Aerosol measurements are from the 150 m subcloud leg (to avoid contamination by liquid water), while Nd measurements are from the in-cloud leg, which in the return flight from L10 were taken in the upper decoupled cloud layer (at 1550 m). The decrease in Nd is likely explained by the cleaning of this decoupled layer through droplet scavenging by precipitation, together with the decoupling from the lower mixed layer aerosol reservoir (Wood et al. 2011). This was confirmed by comparing the L10 downwind Na observations from the cloud layer (filtered to remove samples containing liquid water) and the subcloud Na observations: cloud-layer Na was 114 cm−3 and subcloud Na was 207 cm−3, still elevated above the Nd of 43 cm−3 but not as extreme a difference. We do not use the cloud-layer Na throughout since the required cloud filtering results in too few data points in many cases, and so the decoupling state must be kept in mind. From the observations we cannot support a cloud lifetime effect of aerosol to explain the slow breakup in L10.
In short, the slowly evolving, polluted L10 case, despite a strong reduction in stability, did not deepen or experience much decrease in CF. This is possibly a result of lower surface fluxes, stronger subsidence, and suppression of an aerosol effect from high free-tropospheric relative humidity. That said, as we have shown, there is a lag in the response of cloud fraction to the effect of decreasing stability, and Fig. 8a shows some evidence that breakup became more prominent in the 18 h after resampling. In contrast, L06 had a weaker inversion and higher average surface fluxes, resulting in greater entrainment, which, together with the significant subsidence, allowed for rapid deepening and decoupling. This deepening, combined with a cleaner aerosol environment, likely resulted in vigorous precipitation and further cleaning of the boundary layer and depletion of the Sc layer.
4. Summary and discussion
In this work we have presented the CSET campaign data organized into 18 Lagrangian cases. These cases present an excellent opportunity for future modeling studies and greatly expand the number of possible aircraft-based modeling cases. We have shown the construction and context of these cases, as well as the limitations in their interpretation, and presented an in-depth look at two case studies.
Our analysis of the CSET Lagrangian sampling provides some useful insight into the nature of MBL cloud evolution in the Sc–Cu transition. Many of the trajectories were initially sampled in Sc, but downwind resampling after two days showed a Lagrangian transition to cumulus clusters associated with higher SST, weaker inversion stability, and deeper, more decoupled MBL structure, in accord with existing theory (Bretherton and Wyant 1997). A diversity of aerosol regimes was sampled, but the effect of aerosol variations on MBL evolution was not easy to separate from other meteorological factors.
Results from analysis of the trace gases CO and O3 validate the CSET HYSPLIT Lagrangian track-and-resample flight strategy, even in the face of complications such as vertical shear within the MBL and inexactness of deriving constant-level trajectories from operational analyses. Despite the good resampling, most key cloud and MBL fields had decorrelated at the 2-day resampling, consistent with the satellite-based Lagrangian analysis of Eastman et al. (2016). Variability in fields that do not follow the boundary layer mean flow is likely responsible. In particular, rapid changes in vertical velocity along boundary layer trajectories will affect cloud thickness and thereby precipitation and aerosol removal, causing quick decorrelation in those fields. The CSET sampling strategy may further lower the measured Lagrangian correlation, as each Lagrangian case spans a large range of mesoscale variability (with only intermittent sampling from either profiles or certain level legs). Fields with higher spatial heterogeneity such as precipitation or Nd will be more strongly affected. This limitation should be considered for future Lagrangian aircraft-based studies.
Geostationary satellite data along the Lagrangian trajectories were analyzed to complement the aircraft sampling and more completely document MBL evolution. Aircraft data provide definitive and complete cloud, aerosol, precipitation and thermodynamic profiling, while satellite data provide good space–time coverage for selected variables subject to potential retrieval caveats. The context of satellites and modern global meteorological reanalyses immensely expands the science that can be done with a single research aircraft used in a survey mode as in CSET, and the use of geostationary satellites for Lagrangian studies in particular is effective due to the high temporal sampling they offer. As with many prior studies we found a significant Lagrangian correlation between the evolution of EIS and cloud fraction, maximizing when cloud fraction is compared with the EIS seen along the same trajectory 30 h prior, consistent with previous literature (Klein et al. 1995). Using the expanded trajectory analysis, we showed that the boundary layers sampled by CSET were representative of the region (though interannual variability limits the extent to which this season can be considered climatologically representative) and quantified the strength of the diurnal cycle of CF and CTH.
A motivation of the CSET campaign was to provide datasets suitable for process modeling cases, for which a Lagrangian framework is advantageous because it reduces uncertainties associated with horizontal advective tendencies. The Lagrangian evolution of the MBL was quite variable from day to day. Influences on MBL evolution are strongly state dependent (e.g., the effect of aerosol on stratocumulus entrainment depends on whether the cloud layer is thin or thick). If a composite case is created for model comparison (e.g., de Roode et al. 2016), much of this complexity is lost and indeed it can be challenging to make quantitative comparisons with observations. By providing a variety of potential Lagrangian modeling cases with detailed aircraft observations, our CSET data should provide modelers with more options for exploring MBL evolution. In this paper, these cases have been placed into context, and two cases have been described in detail, which span some of the diversity of SCT evolution observed during CSET. The cases differ in their initial depth, decoupling, microphysics, and large-scale forcings. We plan to present large-eddy simulations of these cases in a forthcoming paper using CSET aircraft data for initialization and comparison after 2 days, with reanalysis extracted along trajectories providing the large-scale forcings, and we hope future researchers will take an interest in which of these factors is most important to the cloud evolution. Most of our other 18 Lagrangian cases could also be fruitful for comparison with process models, as they sample a diversity of meteorological and aerosol conditions. An important caveat is the inevitable oversimplification in considering a flight sequence spanning hundreds of kilometers as representing a single air mass, since it will invariably sample across mesoscale variability and spatial gradients. For instance, the representativeness of a particular aircraft profile for initializing or comparing with a Lagrangian modeling case must be considered.
Acknowledgments
The authors gratefully acknowledge the entire CSET team, and Peter Blossey and Ryan Eastman for insightful discussions. We are also very grateful to the anonymous reviewers for their feedback, which have been helpful in improving this paper. We also acknowledge ECMWF for provision of ERA5 data, and NOAA ARL for provision of the HYSPLIT model. This work was supported by NSF Grants AGS-1445813, AGS-1445832, and AGS-1660609. V. Ghate was supported by the National Science Foundation (NSF) Grant AGS-1445831 awarded to the University of Chicago and the U.S. Department of Energy’s (DOE) Atmospheric System Research (ASR), an Office of Science, Office of Biological and Environmental Research program, under Contract DE-AC02-06CH11357 awarded to Argonne National Laboratory. McCoy was supported by the NSF Graduate Research Fellowship Program under Grant DGE-1762114. CSET campaign data, including trajectories, may be found at the CSET Data Archive (https://data.eol.ucar.edu/master_list/?project=CSET).
APPENDIX
Estimation of Inversion Properties from Profiles
Conceptually, the inversion capping the marine boundary layer is easy to understand as a layer at the top of the marine boundary layer marked by an anomalous temperature increase and moisture decrease with increasing height. In practice, there is significant variability in the presentation of the inversion layer. The presence of moist layers aloft, differences in background free-tropospheric humidity and temperature, and residual decoupled layers with their own inversion complicates finding a single working definition. Additionally, the vertical extent of the inversion varies significantly from the usually strong and vertically compact inversions in the shallow stratiform MBL regimes closer to CA to deep and gradual inversions in the trade regions downwind. When considering multiple data sources, such as the reanalysis and aircraft profiles considered in this work, differences in data resolution add further complexity. Other measures of the depth of the boundary layer, such as those based on cloud-top heights, have their own complications beyond the scope of this appendix. In this appendix we will define our measure of the inversion layer (with a base, center, and top) and how we estimate ΔT and Δq across the inversion. We will also discuss how aircraft data inversion estimates compare with reanalysis, and how they compare with other methods of the inversion height.
Preprocessing of aircraft profile and dropsonde data are required; we first smooth with a 30-s sliding Hanning window (corresponding to a vertical extent of roughly 220 m for aircraft profiles and 180 m for dropsondes). Our approach is as follows (see Fig. A1 for a visual aid): 1) calculate the “center” of the inversion by identifying the region of strongest anomalous stability (strongest difference between actual lapse rate and moist lapse rate, Γ − Γm), 2) calculate the top of the inversion as the lowest level above the center where Γ > Γm, then 3) calculate the lowest extent of the inversion as the highest level below the center where Γ > Γm. This isolates the region of anomalous stability well, with the center of the inversion separating the upper and lower inversion regions. We define Δq as the difference in specific humidity between the top and bottom of the inversion. For ΔT, we integrate (Γ − Γm)dz from the inversion base to the inversion top. This is equivalent to bringing a parcel from the inversion top down to the level of the temperature minimum along a moist adiabat. A slight modification to the method above was needed to account for the occasional presence of a layer of slightly increased stability above sharp Sc inversions caused by strong radiative cooling. In step 2, the inversion top was calculated instead as where Γ − Γm > max(Γ − Γm)/4 (i.e., where the lapse rate gets 75% of the way back from greatest deviation back to moist adiabatic). The same adjustment was made in step 3 for the inversion base to avoid including the entire cloud layer in more decoupled boundary layers. The exact cutoff is somewhat arbitrary but was found to work well for the boundary layers analyzed and falls between two reasonable definitions of the inversion layer, the first being where the lapse rate is subadiabatic and the second being where the lapse rate is positive.
Example profiles from CSET showing estimation of inversion layer (green lines across all plots, marking base, center, and top of the inversion) and ΔT, Δq (red lines on plots of temperature and moisture). (a),(e) Temperature inversion; note that the inversion top temperature is adjusted to the altitude of the inversion base along a moist adiabat. (b),(f) Moisture profile and (c),(g) relative humidity. Note the presence of an elevated moist layer in (g) well above the strong temperature inversion in (e). (d),(h) (smoothed) Lapse rate and moist adiabatic lapse rate.
Citation: Monthly Weather Review 147, 12; 10.1175/MWR-D-19-0053.1
Figure A1 shows a couple aircraft profiles from CSET, marked with the relevant inversion layers (base, middle and top). It also shows the temperature and moisture values used in calculating ΔT and Δq. In Fig. A1b, note the discrepancy between the temperature inversion and the moisture profile; the presence of moist layers aloft confounds a moisture-based estimate of the inversion. Note also in both profiles the slightly subadiabatic layer extending 500 m above the diagnosed inversion top.
This measure was used primarily because it gives reasonably good agreement between aircraft and reanalysis estimates of the inversion center height (r = 0.74, mean bias of 29 m with higher ERA5 inversions, though the thickness of the inversion layer is consistently overestimated in reanalysis), and because it agrees well with the Heffter method used in prior work (Heffter 1980; see Bretherton et al. 2019 and McGibbon and Bretherton 2017) on aircraft and dropsonde profiles (r = 0.85, mean bias of 89 m in with lower Heffter inversion heights. For ERA5 comparisons, the grid point nearest the middle of the aircraft/dropsonde profile is selected and no smoothing is carried out.
REFERENCES
Ackerman, A. S., M. P. Kirkpatrick, D. E. Stevens, and O. B. Toon, 2004: The impact of humidity above stratiform clouds on indirect aerosol climate forcing. Nature, 432, 1014–1017, https://doi.org/10.1038/nature03174.
Adebiyi, A. A., and P. Zuidema, 2018: Low cloud cover sensitivity to biomass-burning aerosols and meteorology over the southeast Atlantic. J. Climate, 31, 4329–4346, https://doi.org/10.1175/JCLI-D-17-0406.1.
Albrecht, B. A., 1989: Aerosols, cloud microphysics, and fractional cloudiness. Science, 245, 1227–1230, https://doi.org/10.1126/science.245.4923.1227.
Albrecht, B. A., D. A. Randall, and S. Nicholls, 1988: Observations of marine stratocumulus clouds during FIRE. Bull. Amer. Meteor. Soc., 69, 618–626, https://doi.org/10.1175/1520-0477(1988)069<0618:OOMSCD>2.0.CO;2.
Albrecht, B. A., C. S. Bretherton, D. Johnson, W. H. Scubert, and A. Shelby Frisch, 1995: The Atlantic Stratocumulus Transition Experiment—ASTEX. Bull. Amer. Meteor. Soc., 76, 889–904, https://doi.org/10.1175/1520-0477(1995)076<0889:TASTE>2.0.CO;2.
Albrecht, B. A., and Coauthors, 2019: Cloud System Evolution in the Trades (CSET): Following the evolution of boundary layer cloud systems with the NSF–NCAR GV. Bull. Amer. Meteor. Soc., 100, 93–121, https://doi.org/10.1175/BAMS-D-17-0180.1.
Berner, A. H., C. S. Bretherton, R. Wood, and A. Muhlbauer, 2013: Marine boundary layer cloud regimes and POC formation in a CRM coupled to a bulk aerosol scheme. Atmos. Chem. Phys., 13, 12 549–12 572, https://doi.org/10.5194/acp-13-12549-2013.
Bretherton, C. S., and M. C. Wyant, 1997: Moisture transport, lower-tropospheric stability, and decoupling of cloud-topped boundary layers. J. Atmos. Sci., 54, 148–167, https://doi.org/10.1175/1520-0469(1997)054<0148:MTLTSA>2.0.CO;2.
Bretherton, C. S., P. N. Blossey, and J. Uchida, 2007: Cloud droplet sedimentation, entrainment efficiency, and subtropical stratocumulus albedo. Geophys. Res. Lett., 34, L03813, https://doi.org/10.1029/2006GL027648.
Bretherton, C. S., and R. Pincus, 1995: Cloudiness and marine boundary layer dynamics in the ASTEX Lagrangian experiments. Part I: Synoptic setting and vertical structure. J. Atmos. Sci., 52, 2707–2723, https://doi.org/10.1175/1520-0469(1995)052<2707:CAMBLD>2.0.CO;2.
Bretherton, C. S., P. H. Austin, and S. T. Siems, 1995a: Cloudiness and marine boundary layer dynamics in the ASTEX Lagrangian experiments. Part II: Cloudiness, drizzle, surface fluxes, and entrainment. J. Atmos. Sci., 52, 2724–2735, https://doi.org/10.1175/1520-0469(1995)052<2724:CAMBLD>2.0.CO;2.
Bretherton, C. S., E. Klinker, A. K. Betts, and J. A. Coakley, 1995b: Comparison of ceilometer, satellite, and synoptic measurements of boundary-layer cloudiness and the ECMWF diagnostic cloud parameterization scheme during ASTEX. J. Atmos. Sci., 52, 2736–2751, https://doi.org/10.1175/1520-0469(1995)052<2736:COCSAS>2.0.CO;2.
Bretherton, C. S., and Coauthors, 2019: Cloud, aerosol, and boundary layer structure across the northeast Pacific stratocumulus–cumulus transition as observed during CSET. Mon. Wea. Rev., 147, 2083–2103, https://doi.org/10.1175/MWR-D-18-0281.1.
Burleyson, C. D., S. P. de Szoeke, S. E. Yuter, M. Wilbanks, and W. A. Brewer, 2013: Ship-based observations of the diurnal cycle of southeast Pacific marine stratocumulus clouds and precipitation. J. Atmos. Sci., 70, 3876–3894, https://doi.org/10.1175/JAS-D-13-01.1.
Copernicus Climate Change Service (C3S), 2017: ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate. Copernicus Climate Change Service Climate Data Store (CDS), accessed 4 May 2018, https://cds.climate.copernicus.eu/cdsapp#!/home.
de Roode, S. R., and Coauthors, 2016: Large-eddy simulations of EUCLIPSE–GASS Lagrangian stratocumulus-to-cumulus transitions: Mean state, turbulence, and decoupling. J. Atmos. Sci., 73, 2485–2508, https://doi.org/10.1175/JAS-D-15-0215.1.
Eastman, R., and R. Wood, 2018: The Competing effects of stability and humidity on subtropical stratocumulus entrainment and cloud evolution from a Lagrangian perspective. J. Atmos. Sci., 75, 2563–2578, https://doi.org/10.1175/JAS-D-18-0030.1.
Eastman, R., R. Wood, and C. S. Bretherton, 2016: Time scales of clouds and cloud-controlling variables in subtropical stratocumulus from a Lagrangian perspective. J. Atmos. Sci., 73, 3079–3091, https://doi.org/10.1175/JAS-D-16-0050.1.
Eastman, R., R. Wood, and T. O. Kuan, 2017: The subtropical stratocumulus-topped planetary boundary layer: A climatology and the Lagrangian evolution. J. Atmos. Sci., 74, 2633–2656, https://doi.org/10.1175/JAS-D-16-0336.1.
Grosvenor, D. P., and R. Wood, 2014: The effect of solar zenith angle on MODIS cloud optical and microphysical retrievals within marine liquid water clouds. Atmos. Chem. Phys., 14, 7291–7321, https://doi.org/10.5194/acp-14-7291-2014.
Hartmann, D. L., M. E. Ockert-Bell, and M. L. Michelsen, 1992: The effect of cloud type on Earth’s energy balance: Global analysis. J. Climate, 5, 1281–1304, https://doi.org/10.1175/1520-0442(1992)005<1281:TEOCTO>2.0.CO;2.
Heffter, J. L., 1980: Transport layer depth calculations. Bull. Amer. Meteor. Soc., 61, 97–97.
Jiang, H., H. Xue, A. Teller, G. Feingold, and Z. Levin, 2006: Aerosol effects on the lifetime of shallow cumulus. Geophys. Res. Lett., 33, 2–5, https://doi.org/10.1029/2006GL026024.
Karlsson, J., G. Svensson, S. Cardoso, J. Teixeira, and S. Paradise, 2010: Subtropical cloud-regime transitions: boundary layer depth and cloud-top height evolution in models and observations. J. Appl. Meteor. Climatol., 49, 1845–1858, https://doi.org/10.1175/2010JAMC2338.1.
Kawai, H., T. Koshiro, and M. J. Webb, 2017: Interpretation of factors controlling low cloud cover and low cloud feedback using a unified predictive index. J. Climate, 30, 9119–9131, https://doi.org/10.1175/JCLI-D-16-0825.1.
Klein, S. A., and D. L. Hartmann, 1993: The seasonal cycle of low stratiform clouds. J. Climate, 6, 1587–1606, https://doi.org/10.1175/1520-0442(1993)006<1587:TSCOLS>2.0.CO;2.
Klein, S. A., D. L. Hartmann, and J. R. Norris, 1995: On the Relationships among low-cloud structure, sea surface temperature, and atmospheric circulation in the summertime northeast Pacific. J. Climate, 8, 1140–1155, https://doi.org/10.1175/1520-0442(1995)008<1140:OTRALC>2.0.CO;2.
Krueger, S. K., G. T. McLean, and Q. Fu, 1995: Numerical simulation of the stratus-to-cumulus transition in the subtropical marine boundary layer. Part I: Boundary layer structure. J. Atmos. Sci., 52, 2839–2850, https://doi.org/10.1175/1520-0469(1995)052<2839:NSOTST>2.0.CO;2.
Lilly, D. K., 1968: Models of cloud-topped mixed layers under a strong inversion. Quart. J. Roy. Meteor. Soc., 94, 292–309, https://doi.org/10.1002/qj.49709440106.
Lock, A. P., 2009: Factors influencing cloud area at the capping inversion for shallow cumulus clouds. Quart. J. Roy. Meteor. Soc., 135, 941–952, https://doi.org/10.1002/qj.424.
Mauger, G. S., and J. R. Norris, 2010: Assessing the impact of meteorological history on subtropical cloud fraction. J. Climate, 23, 2926–2940, https://doi.org/10.1175/2010JCLI3272.1.
McGibbon, J., and C. S. Bretherton, 2017: Skill of ship-following large-eddy simulations in reproducing MAGIC observations across the northeast Pacific stratocumulus to cumulus transition region. J. Adv. Model. Earth Syst., 9, 810–831, https://doi.org/10.1002/2017MS000924.
Minnis, P., and Coauthors, 2008a: Cloud detection in nonpolar regions for CERES using TRMM VIRS and Terra and Aqua MODIS data. IEEE Trans. Geosci. Remote Sens., 46, 3857–3884, https://doi.org/10.1109/TGRS.2008.2001351.
Minnis, P., and Coauthors, 2008b: Near-real time cloud retrievals from operational and research meteorological satellites. Proc. SPIE, 7107, 710703, https://doi.org/10.1117/12.800344.
Minnis, P., and Coauthors, 2011: CERES edition-2 cloud property retrievals using TRMM VIRS and Terra and Aqua MODIS data—Part I: Algorithms. IEEE Trans. Geosci. Remote Sens., 49, 4374–4400, https://doi.org/10.1109/TGRS.2011.2144601.
Nicholls, S., 1984: The dynamics of stratocumulus: Aircraft observations and comparisons with a mixed layer model. Quart. J. Roy. Meteor. Soc., 110, 783–820, https://doi.org/10.1002/qj.49711046603.
Nicholls, S., and J. Leighton, 1986: An observational study of the structure of stratiform cloud sheets: Part I. Structure. Quart. J. Roy. Meteor. Soc., 112, 431–460, https://doi.org/10.1002/qj.49711247209.
O’Connor, E. J., R. J. Hogan, and A. J. Illingworth, 2005: Retrieving stratocumulus drizzle parameters using Doppler radar and lidar. J. Appl. Meteor., 44, 14–27, https://doi.org/10.1175/JAM-2181.1.
Painemal, D., and P. Zuidema, 2011: Assessment of MODIS cloud effective radius and optical thickness retrievals over the Southeast Pacific with VOCALS-REx in situ measurements. J. Geophys. Res., 116, D24206, https://doi.org/10.1029/2011JD016155.
Painemal, D., P. Minnis, and L. O’Neill, 2013: The diurnal cycle of cloud-top height and cloud cover over the southeastern Pacific as observed by GOES-10. J. Atmos. Sci., 70, 2393–2408, https://doi.org/10.1175/JAS-D-12-0325.1.
Randall, D. A., 1980: Conditional instability of the first kind upside-down. J. Atmos. Sci., 37, 125–130, https://doi.org/10.1175/1520-0469(1980)037<0125:CIOTFK>2.0.CO;2.
Rozendaal, M. A., C. B. Leovy, and S. A. Klein, 1995: An observational study of diurnal variations of marine stratiform cloud. J. Climate, 8, 1795–1809, https://doi.org/10.1175/1520-0442(1995)008<1795:AOSODV>2.0.CO;2.
Sandu, I., and B. Stevens, 2011: On the factors modulating the stratocumulus to cumulus transitions. J. Atmos. Sci., 68, 1865–1881, https://doi.org/10.1175/2011JAS3614.1.
Sandu, I., B. Stevens, and R. Pincus, 2010: On the transitions in marine boundary layer cloudiness. Atmos. Chem. Phys., 10, 2377–2391, https://doi.org/10.5194/acp-10-2377-2010.
Schwartz, M., and Coauthors, 2019: Merged cloud and precipitation dataset from the HIAPER GV for the Cloud System Evolution in the Trades (CSET) campaign. J. Atmos. Oceanic Technol., 36, 921–940, https://doi.org/10.1175/JTECH-D-18-0111.1.
Seinfeld, J. H., and S. N. Pandis, 2006: Atmospheric Chemistry and Physics: From Air Pollution to Climate Change. 2nd ed. Wiley-Interscience, 1232 pp.
Stein, A. F., R. R. Draxler, G. D. Rolph, B. J. B. Stunder, M. D. Cohen, and F. Ngan, 2015: NOAA’s HYSPLIT Atmospheric Transport and Dispersion Modeling System. Bull. Amer. Meteor. Soc., 96, 2059–2077, https://doi.org/10.1175/BAMS-D-14-00110.1.
Stevens, B., W. R Cotton, G. Feingold, and C.-H. Moeng, 1998: Large-eddy simulations of strongly precipitating, shallow, stratocumulus-topped boundary layers. J. Atmos. Sci., 55, 3616–3638, https://doi.org/10.1175/1520-0469(1998)055<3616:LESOSP>2.0.CO;2.
Stevens, B., G. Vali, K. K. Comstock, R. Wood, M. C. van Zanten, P. H. Austin, C. S. Bretherton, and D. H. Lenschow, 2005: Pockets of open cells and drizzle in marine stratocumulus. Bull. Amer. Meteor. Soc., 86, 51–57, https://doi.org/10.1175/BAMS-86-1-51.
Twomey, S., 1974: Pollution and the planetary albedo. Atmos. Environ., 8, 1251–1256, https://doi.org/10.1016/0004-6981(74)90004-3.
Wood, R., 2007: Cancellation of aerosol indirect effects in marine stratocumulus through cloud thinning. J. Atmos. Sci., 64, 2657–2669, https://doi.org/10.1175/JAS3942.1.
Wood, R., 2012: Stratocumulus clouds. Mon. Wea. Rev., 140, 2373–2423, https://doi.org/10.1175/MWR-D-11-00121.1.
Wood, R., and C. S. Bretherton, 2004: Boundary layer depth, entrainment, and decoupling in the cloud-capped subtropical and tropical marine boundary layer. J. Climate, 17, 3576–3588, https://doi.org/10.1175/1520-0442(2004)017<3576:BLDEAD>2.0.CO;2.
Wood, R., and C. S. Bretherton, 2006: On the relationship between stratiform low cloud cover and lower-tropospheric stability. J. Climate, 19, 6425–6432, https://doi.org/10.1175/JCLI3988.1.
Wood, R., C. S. Bretherton, D. Leon, A. D. Clarke, P. Zuidema, G. Allen, and H. Coe, 2011: An aircraft case study of the spatial transition from closed to open mesoscale cellular convection over the Southeast Pacific. Atmos. Chem. Phys., 11, 2341–2370, https://doi.org/10.5194/acp-11-2341-2011.
Wyant, M. C., C. S. Bretherton, H. A. Rand, and D. E. Stevens, 1997: Numerical simulations and a conceptual model of the stratocumulus to trade cumulus transition. J. Atmos. Sci., 54, 168–192, https://doi.org/10.1175/1520-0469(1997)054<0168:NSAACM>2.0.CO;2.
Xiao, H., C.-M. Wu, and C. R. Mechoso, 2011: Buoyancy reversal, decoupling and the transition from stratocumulus to shallow cumulus topped marine boundary layers. Climate Dyn., 37, 971–984, https://doi.org/10.1007/s00382-010-0882-3.
Xue, H., and G. Feingold, 2006: Large-eddy simulations of trade wind cumuli: Investigation of aerosol indirect effects. J. Atmos. Sci., 63, 1605–1622, https://doi.org/10.1175/JAS3706.1.
Yamaguchi, T., G. Feingold, and J. Kazil, 2017: Stratocumulus to cumulus transition by drizzle. J. Adv. Model. Earth Syst., 9, 2333–2349, https://doi.org/10.1002/2017MS001104.
Zuidema, P., D. Painemal, S. de Szoeke, and C. Fairall, 2009: Stratocumulus cloud-top height estimates and their climatic implications. J. Climate, 22, 4652–4666, https://doi.org/10.1175/2009JCLI2708.1.
