1JULY 2004AMERICAN METEOROLOGICAL SOCIETY |

S tratocumulus cloud layers are the vast “climaterefrigerators” of the Tropics and subtropics,which they cool by reflecting sunlight back tospace that could otherwise warm the ocean surface.They favor the eastern subtropical oceans, where up-welling and cold currents keep the surface waters coolcompared to the warm, dry subsiding air aloft (Kleinand Hartmann 1993). These low clouds play a keysupporting role in the seasonal cycle of the east PacificOcean and global climate, and may feed back onEl Niño–Southern Oscillation. However, they haveproven particularly difficult for climate models tosimulate because they are only a few hundred meters

thick, typically lie under a sharp inversion, and aremaintained by a blend of physical processes that arecomplex to parameterize. Overlaying the cool south-east Pacific Ocean is the most persistent subtropicalstratocumulus deck in the world.

The East Pacific Investigation of Climate (EPIC)is a National Science Foundation (NSF)- and NationalOceanic and Atmospheric Administration (NOAA)-supported program to study physical processes cen-tral to coupled atmosphere–ocean interaction in theeast Pacific. Its centerpiece was a field experiment inSeptember and October of 2001 along the easternmostline of Tropical Atmosphere Ocean (TAO) buoys at95∞W. Three of its goals were to examine atmosphericdeep convection in the ITCZ, ocean mixing and re-sponse to episodic deep convection, and cross-equatorial flow in the atmospheric boundary layer(ABL). These aspects of EPIC 2001 are described in acompanion paper by Raymond et al. (2004).

In this article, we describe the final component ofEPIC 2001, a stratocumulus study that took a pioneer-ing look at cloud and ABL processes in the southeastPacific region. Comprehensive ship-based remotesensing and surface measurements were taken dur-ing a 2-week cruise through this sparsely traveled re-gion during the month (October) when the stratocu-mulus clouds in this region are most extensive.Scientific objectives included measuring the vertical

THE EPIC 2001STRATOCUMULUS STUDY

BY CHRISTOPHER S. BRETHERTON, TANEIL UTTAL, CHRISTOPHER W. FAIRALL, SANDRA E. YUTER, ROBERT A.WELLER, DARREL BAUMGARDNER, KIMBERLY COMSTOCK, ROBERT WOOD, AND GRACIELA B. RAGA

A cruise to the southeast Pacific Ocean examines how turbulence, drizzle, aerosols,

and the Andes combine to regulate the albedo of the most persistent

stratocumulus cloud regime in the Tropics.

AFFILIATIONS: BRETHERTON, YUTER, COMSTOCK, AND WOOD—Department of Atmospheric Sciences, University of Washington,Seattle, Washington; UTTAL AND FAIRALL—NOAA/EnvironmentalTechnology Laboratory, Boulder, Colorado; WELLER—Woods HoleOceanographic Institution, Woods Hole, Massachusetts;BAUMGARDNER AND RAGA—Universidad Nacional Autonoma deMexico, Mexico City, MexicoCORRESPONDING AUTHOR: Dr. Christopher S. Bretherton,Department of Atmospheric Sciences, University of Washington,Box 351640, Seattle, WA 98195-1640E-mail: breth@atmos.washington.eduDOI: 10.1175/BAMS-85-7-startpage #

In final form 15 December 2003©2004 American Meteorological Society

Au: ABL previ-ously defined

2 JULY 2004|

structure of the ABL, understandingwhat physical processes are deter-mining the stratocumulus cloud al-bedo (how much of the sunlight in-cident on the clouds is reflected backto space), and understanding thefluxes of heat and water that couplethe atmosphere and the ocean in thisregion. Our measurements wereparticularly suitable for examiningfeedbacks between cloud albedo,ABL turbulent dynamics, anddrizzle over the entire diurnal cycle.They also allow the stratocumulus inthis region to be compared to thebetter-studied stratocumulus of thenortheast Pacific, and to thosesampled in the CIMAR-5 cruise inaustral autumn in the much moresynoptically disturbed stratocumu-lus region off of central Chile(Garreaud et al. 2001).

Stratocumulus albedo is strongly dependent oncloud thickness and horizontal continuity. Over 30 yrago, Lilly (1968) first elucidated the tight feedbacksbetween the clouds, radiation, turbulence, and en-trainment that regulate subtropical stratocumuluscloud thickness (Fig. 1). Intense longwave radiativecooling in the cloud top drives turbulent eddies in theABL. These eddies pick up moisture (which maintainsthe cloud layer) from the sea surface, but they alsoentrain filaments of warm, dry air from above thecapping inversion. Entrainment affects both the cloudbase and top. Entrainment constantly lifts the cloudtop, maintaining it against large-scale subsidence. Italso dries out the ABL. ABL turbulence is sensitive tothe cloud base and cloud top, creating a feedback ofthe cloud geometry on the entrainment rate. Thisfeedback helps to establish an ever-changing balancebetween cloud thickness, entrainment, and turbu-lence modulated by external forcings, such as chang-ing sea surface temperatures, subsidence rate, and thedaily cycle of absorption of sunlight.

Subtropical stratocumulus clouds may drizzle ap-preciably, even when they are as little as 250 m thick(Albrecht 1989). Even relatively light drizzle mightdeplete enough liquid water from a stratocumuluscloud to reduce its albedo. In some climate models,drizzle is the primary regulator of modeled subtropi-cal marine stratocumulus cloud thickness and extent.In others, that role falls to entrainment. The two pro-cesses interact, and modeling studies suggest thatdrizzle tends to reduce entrainment (e.g., Pincus and

Baker 1994; Stevens et al. 1998). Very few observa-tions are available to decide which models are correct.

Three months before EPIC 2001, the Dynamicsand Chemistry of Marine Stratocumulus Phase II:Entrainment Studies (DYCOMS-II) experiment alsoset out to improve our understanding of stratocumu-lus entrainment and drizzle processes (Stevens et al.2003). DYCOMS-II used nocturnal flights into north-east Pacific stratocumulus clouds to obtain aremarkable dataset quantifying entrainment, turbu-lence, and drizzle at the time of day when theseprocesses are most vigorous. The EPIC 2001 stratocu-mulus cruise complemented the DYCOMS-II mea-surements by providing comprehensive documenta-tion of the strong diurnal cycle of stratocumulusthickness, a more complete view of the time–spacestatistics of drizzle, and the perspective of a differentregion with a deeper boundary layer and generallylower aerosol concentrations. On the other hand, theaircraft data taken during DYCOMS-II and earlier ex-periments provide crucial context for interpreting ourmeasurements. We are working with DYCOMS-II in-vestigators to best exploit this synergy.

VERTICAL STRUCTURE AND DIURNALCYCLE OF THE SOUTHEAST PACIFIC ABL.The NOAA research vessel Ronald H. Brown (here-after called the Brown) played a key role in all phasesof EPIC 2001, finishing with the stratocumulus cruiseshown on Fig. 2. From the Galapagos Islands, theBrown steamed west on 9 October to 95°W, then

FIG. 1. The interplay of physical processes associated with stratocu-mulus cloud layers.

Au: is therean expansionavailable forCIMAR-5?

3JULY 2004AMERICAN METEOROLOGICAL SOCIETY |

south along the remainder of the TAO buoy line intothe southeast Pacific stratocumulus-capped boundarylayer. It stopped for 6 days to maintain and comparemeasurements with a buoy maintained by R. Wellerof the Woods Hole Oceanographic Institution(WHO) at 20°S, 85°W. This buoy plays an importantpart of EPIC longer-term monitoring (Cronin et al.2002; see http://uop.whoi.edu/stratus/ online for morebuoy information and data). On 25 October, theBrown reached the port of Arica in northern Chile.The entire cruise track was characterized by cool SSTand southeasterly trade winds. Figure 3 shows a Geo-stationary Operational Environmental Satellite(GOES)-East visible channel image for 1445 UTC on19 October, which is typical for the period. Extensivestratocumulus, usually organized into mesoscale cel-lular structures 10–40 km across (often called “me-

Yuter and Bretherton University of Washington Sondes, 5-cm scanning radar, meth blue

Raga and Baumgardner UNAM (Mexico) Aerosol concentration, sulfate fraction

Fairall and Uttal NOAA/ETL Cloud remote sensing, surface fluxes

Weller WHOI IMET buoy, CTDs

TABLE 1. Southeast Pacific stratocumulus cruise research groups and key measurements.

PI Institution Key measurements

FIG. 2. The EPIC 2001 stratocumulus cruise track.Numbers indicate the position of the Brown at the be-ginning of selected days of Oct (UTC). The shadingshows SST from the Tropical Rainfall Measuring Mis-sion (TRMM) Microwave Imager, and the arrows show10-m wind vectors from Quikscat averaged over the9–25 Oct 2001 cruise period. TAO buoys and theWHOI buoy are shown with black squares and a reddot, respectively.

FIG. 3. The southeast Pacific stratocumulus region asviewed by GOES-East. The prominent dark “patchesof open cells” (POCs) of broken cloud in the center,surrounded by almost unbroken cloud cover, are acommonly seen but still a mysterious feature of sub-tropical stratocumulus, possibly associated with regionsof enhanced drizzle. The inset zooms in on the mesos-cale cellularity in the cloud field on 10–40-km scales, andincludes part of a POC.

soscale cellular convection,” e.g., Agee 1982), blan-keted the region throughout most of the cruise. Thebuoy and shipboard sampling also measured signifi-cant small-scale spatiotemporal variability in the up-per ocean in this region in the form of temperature/salinity fronts, eddies, and a diurnal cycle of SST thatwas small while the Brown was nearby, but at othertimes occasionally became much larger during epi-sodes of weak winds and clear skies.

Table 1 lists the participating research groups andsome of their key measurements. Three-hourlysoundings complemented a suite of NOAA/Environ-mental Technology Laboratory (ETL) verticallypointing remote sensing measurements, including aceilometer for measuring cloud base, 8-mm wave-

Au: please defineIMET, CTD—notused in text also“meth blue”

4 JULY 2004|

length Doppler radar for sensing of clouds anddrizzle, and a microwave radiometer for measuringvertically integrated cloud liquid water. The 5-cm-wavelength scanning radar mounted on the Brownwas used to survey the morphology and mean ampli-tude of drizzle within 30 km of the ship. Surface me-teorology, turbulent and radiative flux measurements,occasional samples of drizzle drop-size distribution,and limited measurements of aerosol concentrationand composition provided a comprehensive near-surface view of the PBL. Individual soundings, cor-responding photographs of the cloud field, and radarimages from the EPIC 2001 stratocumulus cruise canbe browsed online (www.atmos.washington.edu/epic/).

Figure 4 shows the temperature and moisturestructure of the lower atmosphere from rawinsondeslaunched 8 times per day. A strong inversion invari-ably capped the ABL, with a 10–15-K potential tem-perature increase with height. The inversion deepenedfrom 900 to 1300 m as the ship headed southwardaway from the equator, then became somewhat shal-lower near the coast. Near the equator it was remark-ably moist above the ABL, but in the core of the stra-tocumulus region it was very dry. Above the ABL,moist and dry layers can be seen descending with timefollowing the ship track; we believe this is a manifes-

tation of the persistent large-scale subsidence over thisregion. The stratocumulus clouds were usually lessthan 500 m thick.

Two features of Fig. 4 were somewhat unexpected.First, vertical gradients of potential temperature andhumidity in the ABL were quite weak throughout thecruise, approximating a “mixed layer” structure. Sucha structure suggests turbulent eddies efficiently mix-ing air through the entire boundary layer. This wassurprising, because at similar distances (1000–2000 km) offshore in the northeast Pacific and Atlan-tic Oceans, the humidity decreases considerably withheight within the ABL, associated with a processcalled “cumulus coupling” (a layer of cumulus cloudsrising into the stratocumulus; Albrecht et al. 1995;Klein et al. 1995; Wyant et al. 1997). Because the areafraction covered by cumulus clouds is usually less than10%, much of the air in such a cumulus layer isnonturbulent, allowing vertical humidity gradients tobuild up. This structure is what we were also antici-pating in the southeast Pacific, but we saw almost nocumulus clouds. One difference between the south-east Pacific and these other regions is that the cap-ping inversion remains very strong at distances over1000 km offshore, which appears to inhibit the devel-opment of cumulus coupling.

The second unexpectedfeature was the strength andregularity of the diurnal cyclein the inversion height. This isseen particularly clearly in thehumidity plot of Fig. 4 duringthe 6-day “buoy period” at20∞S, 85∞W. The inversion(which also marks the stra-tocumulus cloud top) risesroughly 200 m each nightfrom an early afternoon mini-mum value. In contrast, thecloud base shows little diurnalcycle, so the clouds are thin-nest during the early after-noon. During most of thecruise (and particularly in thebuoy period) there was nearly100% cloud cover at night, butthe clouds often became bro-ken in the afternoon. Becausethe diurnal cycle was so regu-lar during the buoy period, wefound it illuminating to con-struct a 6-day buoy-periodmean diurnal cycle of some

FIG. 4. Time–height cross section of (a) potential temperature and (b) wa-ter vapor mixing ratio (bottom) from 8-times daily radiosonde launchesduring the EPIC 2001 stratocumulus cruise. The black curve in (b) is thehourly averaged ceilometer-derived cloud base; cloud top closely coincideswith the ABL capping inversion base. Vertical lines bound the buoy periodduring which the Brown was stationed near the WHOI buoy at 20°S, 85°W.Local time is 6 h behind UTC.

5JULY 2004AMERICAN METEOROLOGICAL SOCIETY |

salient boundary layer and cloud measurements,shown in Figs. 5 and 6.

Figure 5 shows the buoy-period mean diurnalcycle of the potential temperature and humidity pro-files in and above the ABL. At each of the eight dailysounding times, a composite sounding that preservesinversion sharpness was constructed. To do this, wefirst found the inversion base height (local tempera-ture minimum) in each of the six buoy-period sound-ings for that time of day. Then, we vertically stretchedeach of these soundings to have this average inver-sion base height, and averaged them to get thecomposite soundings. The mean millimeter-radar-derived cloud top (not shown) closely matches themean inversion height at each time of day. Figure 5illustrates not only the diurnal cycle of inversionheight, but also the sharpness of the inversion and thewell-mixedness of the ABL. At each time, the poten-tial temperature and humidity are nearly uniformbelow the mean ceilometer-derived cloud base, andhave nearly moist adiabatic gradients above. Slightsubcloud potential temperature and humidity gradi-ents can be seen below cloud base in the early morn-ing (1100 UTC) through late afternoon (2300 UTC).We attribute these gradients to drizzle formation andin-cloud solar absorption, both of which tend to heatthe upper part of the boundary layer, counteractingthe cloud-top longwave cooling and weakening theABL turbulence. During the early part of the night,when there is no solar absorption and the cloud hasnot yet thickened enough to drizzle significantly, thecomposite soundings are particularly well mixed.

Figure 6 shows the buoy-period mean diurnal cycleof the liquid water path measured from a shipboardmicrowave radiometer, and compares it with the“adiabatic liquid water path.” This comparison is ameasure of how efficiently turbulence homogenizesthe thermodynamic properties of air within theclouds. In particular, were we to assume that theclouds were vertically well-mixed or “adiabatic,” theirliquid water content would rise at a predictable ratewith height above cloud base, resulting in an adiabaticliquid water path that depends on the cloud thickness.For Fig. 6, the cloud thickness was deduced as thedifference between the millimeter-radar-derivedcloud top and ceilometer-derived cloud base. Figure 6shows remarkable agreement between the observedand adiabatic liquid water path, suggesting that theclouds are in fact usually close to adiabatic.

Starting with the First International Satellite CloudClimatology Project (ISCCP) Regional Experiment(FIRE-I) in 1987 just off the California coast, manystratocumulus studies have seen a diurnal cycle in

cloud thickness qualitatively similar to that seen inFigs. 5 and 6, with an early afternoon minimum andearly morning maximum in mean cloud thickness andcloud fractional coverage (Albrecht et al. 1990; Minniset al. 1992). This has important consequences for stra-tocumulus feedback on climate, because the cloudsare thinnest and least reflective when the insolationis largest. These studies led to the following explana-

FIG. 5. Buoy-period 6-day composite diurnal cycle of(top) humidity and (bottom) potential temperaturesoundings. Horizontal lines indicate mean ceilometer-derived cloud base. Offsets of 6 g kg-----1 and 10 K are ap-plied to successive 3-hourly profiles.

FIG. 6. The buoy-period mean diurnal cycle of (top)ECMWF-predicted vertical velocity from hourly sam-pling of 12–36-h operational forecasts, (middle) ceilo-meter-derived cloud fraction, and (bottom) liquid wa-ter path derived from the shipboard microwave radi-ometer, and adiabatic liquid water path derived fromcloud thickness. Vertical bars show the standard devia-tion of hourly average values on individual days fromthe 6-day hourly mean.

6 JULY 2004|

tion of this diurnal cycle as a consequence of local ABLprocesses. During the day, solar radiation is absorbedin the clouds, largely cancelling the longwave radia-tive cooling from their tops and greatly inhibitingturbulence. This reduces the efficiency of moisturetransport from the sea surface to the cloud layer, dry-ing it out and lifting the cloud base. Meanwhile, lessturbulence causes less entrainment, allowing meansubsidence to advect the cloud top down. Both effectsthin the cloud during the afternoon.

However, there was a surprising aspect of the stra-tocumulus diurnal cycle in the southeast Pacific.Based on prior observations and modeling studies(e.g., Bougeault 1985), we had expected most of thecloud thickness variations to come from a varyingcloud base, while in fact Fig. 5 shows that they comefrom height variations in the inversion that defines thecloud top. The inversion variations during the buoyperiod were much too large to be plausibly explainedthrough variations in entrainment rate alone. Instead,it became clear that there must also be a strong diur-nal cycle in subsidence as well, which we looked forin output from global operational forecast models.Because long-term surface flux measurements at theWHOI buoy are being used by the European Centrefor Medium-Range Weather Forecasts (ECMWF) formodel validation, ECMWF happened to be saving anextensive suite of hourly model output at this loca-tion from their daily operational 12–35-h forecasts.This output for October 2001 was kindly provided tous by Martin Köhler of ECMWF. H.-L. Pan of theNational Centers for Environmental Prediction(NCEP) also provided us with a similar columndataset using their Medium Range Forecast (MRF)model. Figure 6 shows the buoy-period mean diur-nal cycle in subsidence from the ECMWF model at850 hPa, roughly the height of the inversion.Subsidence was near zero near local midnight, thenrapidly increased to maximum values in the latemorning. Even after averaging over the entire monthof October, the same diurnal cycle could be seen atslightly reduced amplitude. The NCEP–NationalCenter for Atmospheric Research (NCAR) analysisgave similar results. The diurnal cycle of subsidenceamplifies the diurnal cloud thickness variations thatwe would get from internal ABL dynamics, probablyleading to thinner, less reflective stratocumulus dur-ing the day. Through mechanisms such as this, thediurnal cycle may have an impact on the mean climateeven over the remote oceans. But what causes the di-urnal cycle of subsidence in the first place?

Nontidal diurnal cycles of subsidence of diverseorigin affect the ABL over many parts of the tropical

oceans, even well away from continents (Ciesielskiet al. 2001; Dai and Deser 1999; Foltz and Gray 1979).Using 4-times daily ECMWF and NCEP–NCAR re-analysis data, we have done a regional analysis of thediurnal cycle of subsidence over the southeast Pacific.Between the Peruvian coast and the WHOI buoy,there is a 6-h phase delay (and a large reduction inamplitude) in the maximum subsidence, correspond-ing to a phase speed of 50 m s-1. The subsidence wavehas maximum amplitude in the midtroposphere. Bothof these are characteristics of an internal inertia–gravity wave forced by the diurnal cycle of heatingover South America. The regularity of the signal sug-gests that the heating is mainly due to sensible heatfluxes in the ABL, which can reach high into the tro-posphere over the Andes, but latent heating in deepconvective cloud systems also probably contributes tothis. The diurnal cycles of the vertical velocity fieldsin the two reanalyses are qualitatively consistent, buta better model-independent dataset is required to fullytest the subsidence wave hypothesis. A 15-day re-gional model simulation by Garreaud and Muñoz(2004) nicely reproduces this diurnal subsidencewave.

DRIZZLE AND AEROSOL. The southeast Pa-cific stratocumulus regime is among the world’sdriest oceanic regions. No precipitation was mea-sured from October 2001 to October 2002 by a raingauge mounted on the WHOI buoy (though in suc-ceeding years the rain gauge has measured a verysmall amount of rain). However, conventional raingauges are not sensitive to light drizzle, which mayevaporate from the gauge before accumulating.Petty’s (1995) study of the Comprehensive Ocean–Atmosphere Data Set (COADS) climatology of rou-tine ship-observer present weather reports suggeststhat the instantaneous frequency of precipitation inthis region is roughly 1%, so we were optimistic thatwe might find some drizzling stratocumulus at somepoint on the cruise. In fact, our radars observed con-siderable drizzle, particularly at the WHOI buoylocation (even though no rain was accumulated by therain gauges aboard the ship). Figure 7 shows an ex-ample of a horizontal cross section of radar reflectivityat 1 km above the ocean surface from the scanning5-cm-wavelength radar. It shows patches ofreflectivities up to 30 dBZ (light rain showers), orga-nized in cells 1–3 km across, with a hint of polygonalorganization that probably corresponds to the mesos-cale cellular convection visible in the satellite imageof Fig. 3. Also shown is a short time–height sectionof reflectivity from the NOAA 8-mm-wavelength ver-

7JULY 2004AMERICAN METEOROLOGICAL SOCIETY |

tically pointing radar at the same time. The cells areadvecting to the northwest. The high reflectivities ex-tend down to the surface, indicating that some sur-face drizzle must be occurring. The electronics of the8-mm radar saturate at a range-dependent maximumreflectivity, leading to reflectivities noticeably lowerthan the maximum reflectivities seen by the 5-cm ra-dar (Comstock et al. 2004, manuscript submitted toQuart. J. Roy. Meteor. Soc., hereafter COM). Patchydrizzle was common at night and in the early morn-ing during most of the cruise.

Quantifying the precipitation rate from radar datais challenging and involves large uncertainties. Fordrops of diameter D, radar reflectivity depends on D6,while precipitation flux is approximately proportionalto D4. Hence, reliable estimation of precipitation ratefrom radar reflectivity requiresaccurate assumptions about theraindrop size spectrum. Becausethe stratocumulus clouds are sothin, the spectrum of raindropscannot be assumed to be anythinglike that in mature cumulonim-bus clouds. Small calibration er-rors in either radar can bias pre-cipitation retrievals. On selectedoccasions when ship observers feltdrizzle, they exposed filter papertreated with methylene blue tocapture images of the fallingdroplets (Rinehart 1997), whichwere counted and sized to pro-duce estimates of droplet size dis-tribution, instantaneous localrainfall rate, and radar reflectivity.We also used the time-dependentvertical profile of reflectivity fromthe millimeter-wave radar to in-fer the evaporation rate of rain-drops below cloud base and thetypical raindrop size. We com-bined these with the 5-cm radarreflectivity maps to produce esti-mates of area-averaged precipita-tion over a 30-km-radius circlearound the ship. COM docu-ments our precipitation retrievalmethodology and its uncertain-ties; we regard our estimates ofprecipitation to have uncertain-ties of a factor of 2–3.

With these caveats, Fig. 8 showsthe inferred area-averaged pre-

cipitation rate at cloud base and the surface. About4 mm day-1 of water evaporates from the sea surfacein this region, so precipitation rates of 1 mm day-1 ormore are highly significant to the water budget of theclouds. At cloud base, drizzle rates this high are com-mon, but owing to the high cloud base in this region,85% of the drizzle evaporates before reaching the sur-face and the surface precipitation rate seems to be onlya minor sink of water from the ABL. Hence, the im-pact of drizzle is indirect. Water is condensed in theclouds, releasing latent heat there, then falls as drizzleinto the subcloud layer, which it cools by evaporation.This heating-above-cooling couplet is a stratifyinginfluence on the ABL that inhibits turbulence and en-trainment, increases vertical moisture gradients, andmakes the cloud layer more horizontally inhomoge-

FIG. 7. (top) Horizontal cross section of 5-cm radar reflectivity at 1-kmelevation during a heavy drizzle period (ship location at black dot, ar-row shows ABL wind direction) and (bottom) time–height plot of 8-mmradar reflectivity above the ship (black line shows time of upper image).A drizzle cell is advecting over the ship.

8 JULY 2004|

neous (Stevens et al. 1998). Models suggest that thenet effect of these processes is to reduce the liquidwater path (Pincus and Baker 1994; Stevens et al.1998). However, these models have only simulatedshallow stratocumulus-capped boundary layers withlow cloud bases. For such ABLs, drizzle is a more ef-fective ABL water sink than in EPIC 2001, because amuch smaller fraction of precipitation evaporatesbefore reaching the surface. Hence, we cannot be tooconfident of the applicability of these model resultsto the EPIC 2001 ABL, where the subcloud layer isdeep. In any case, the observed drizzle flux throughcloud base seems large enough that the impact ofdrizzle production on ABL turbulence and entrain-ment should at times be quite marked.

In addition, Fig. 8 shows the liquid water path de-rived from the microwave radiometer, and an esti-mate of the cloud droplet concentration, derived asfollows from daytime surface observations of thedownwelling shortwave radiation measured by apyranometer. The cloud transmissivity was calculatedas the ratio of the downwelling shortwave radiationto its expected clear-sky value. The transmissivity wasused to infer the cloud optical depth followingStephens et al. (1984). The cloud optical depth wascombined with the measured liquid water path to in-fer a mean cloud droplet effective radius. The drop-let concentration was then deduced from the cloudthickness, liquid water path, and effective radius fol-lowing a formula of Dong and Mace (2003) modifiedto treat a cloud whose liquid water content increaseslinearly with height, as observed in turbulent stratocu-mulus cloud layers.

The diurnal cycle of cloud thickness and the liq-uid water path strongly modulate the drizzle fallingfrom cloud base. Variations in cloud droplet concen-

tration seem to be playing a similarly important role.The “cleanest” clouds with lowest droplet concentra-tions of 60–80 cm-3 occurred on 18–19 October.During these days, cloud-base drizzle was substantialeven during the daytime. On 15–16 October, whendroplet concentrations rose to as much as 200 cm-3,much less drizzle was observed despite liquid waterpaths that were almost as large. Near the coast, veryhigh cloud droplet concentrations (“dirty” clouds)were observed; satellite retrievals of cloud opticalproperties in this region also tend to show markedlylower effective radii near the coast. Because the ABLwinds usually blow northward up the coast, this isprobably due to dispersal of anthropogenic aerosolproduced by industrial activity along the Chileancoast. Natural salt and dimethyl sulfide aerosol pro-duction may also have substantial spatiotemporalvariability in this region.

Universidad Nacional Autónoma de Mexico(UNAM) aerosol concentration measurements,though limited in scope, also produced a surprise thatwe do not yet fully understand. Figure 9 shows mea-surements of condensation nucleus (CN) concentra-tion made from a 10-m tower on the front of the shipusing air preheated to 343 K then passed through aTSI 3010 CN counter (0.01-mm-size threshold). Atnight, measured CN concentrations were stable andquite long (10–100 cm-3), with some tendency forlower nocturnal CN concentrations, correspondingto the periods of more cloud-base drizzle and asmaller concentration of cloud droplets, as one mightexpect. However, the diurnal signal in CN was muchmore striking. On each day a rapid, roughly two-foldincrease of CN was measured just after dawn. The CN

FIG. 8. Retrievals of hourly averaged drizzle rate, liquidwater path (LWP), and cloud droplet concentration Ndfrom a combination of ship-based remote sensors.Cloud-base drizzle rate showed a marked diurnal cycle,increasing at night in phase with the LWP. Most drizzleevaporated before reaching the ocean surface. Drizzlewas enhanced in cleaner clouds.

FIG. 9. CN concentration and shortwave radiation(SWR) measured on the Brown, showing a strong, un-explained, diurnal cycle in CN apparently driven byphotochemical processes, superposed on backgroundnocturnal CN values that correlate with cloud drop-let concentration.

Au: per AMS stylecannot begin sen-tence w/ acronymshorter than 3characters.

9JULY 2004AMERICAN METEOROLOGICAL SOCIETY |

concentrations during theday showed considerablevariability before suddenlyreturning to nocturnal val-ues 3–5 h after sunset. Wedo not believe that instru-mental artifacts can ac-count for this peculiar be-havior. In fact, Parungoet al. (1978) reported a di-urnal cycle of similar phasein CN measurements takenalong 5∞S between 110∞and 140∞W, and Hoppeland Frick (1990) reportedcomparable CN levels evenfurther to the west. Both studies reported large, un-explained increases in CN concentration for short pe-riods. The measured diurnal cycle of CN suggests thatphotochemical processes may be driving new particleproduction during the daytime (Fitzgerald 1991, andreferences therein), possibly associated with phy-toplankton-produced dimethyl sulfide (Lawrence1993) or organic acids (e.g., Jenkin and Clemitshaw2000). Because the new particles are small they do notimmediately affect the clouds by acting as cloud con-densation nucleii, but over the long run processes likethese may be helping to regulate cloud microphysicsover large parts of the remote oceans, especially inthe Southern Hemisphere.

Given the complex suite of physical processes inter-acting to produce the southeast Pacific cloud-toppedboundary layer, how well do global numerical modelssimulate this important climate regime? Figure 10 com-pares 6-day mean buoy-period profiles of potential tem-perature, water vapor, and liquid water content (thelatter inferred from the time series of ceilometer cloudbase, millimeter-radar cloud top, and microwave liquidwater path, assuming that liquid water increases linearlywith height above the cloud base) with the correspond-ing 6-day mean from ECMWF and NCEP–NCAR op-erational analyses, and with October climatology fromrecent versions of two atmospheric general circulationmodels (GCMs), NCAR’s Community AtmosphereModel version 2.0 (www.ccsm.ucar.edu/models/atm-cam/) and the Geophysical Fluid Dynamics Laboratory(GFDL)’s Atmospheric Model version 2.10 (Andersonet al. 2004, manuscript submitted to J. Climate). Theoperational analyses did not assimilate our radiosondeobservations, and the GCMs used climatological SSTs.

A comparison with our limited measurements canonly be suggestive, especially against October GCMclimatologies, but can expose gross model biases. We

believe that our 6-day observation period was fairlyrepresentative of typical October conditions in thisregion, but with slightly thicker and more extensivestratocumulus cloud. The cloud-top height was simi-lar to the satellite-derived October mean (Wood andBretherton 2004), and the mean ship-observed liquidwater path was roughly 30% larger than the Octobermean inferred from satellite measurements (Woodet al. 2002).

Figure 10 suggests that the ECMWF and NCEP–NCAR models qualitatively represent the main fea-tures of the profiles adequately, but slightly underes-timate the inversion height and cloud liquid waterpath. The GCMs both substantially underestimate thePBL depth and cloud liquid water path, and the cloudis very close to the sea surface, suggesting that theseGCMs underestimate entrainment and instead regu-late stratocumulus thickness by a combination ofdrizzle formation and numerical diffusion in this re-gion. Not all GCMs share these biases, but the EPICresults should prove a useful baseline for comparisonwith GCMs in this otherwise poorly sampled region.

CONCLUSIONS AND FUTURE DIREC-TIONS. The EPIC 2001 stratocumulus cruise was aresounding success. An unexpectedly well-mixedstratocumulus-capped boundary layer capped by astrong inversion was encountered throughout. Astrong diurnal cycle was observed, with thicker cloudsand substantial drizzle (mainly evaporating above thesea surface) during the late night and early morning.This was driven in part by local diabatic processes butwas reinforced by a surprisingly pronounced diurnalcycle of vertical motion. The vertical motion appearsto be an inertia–gravity wave driven by daytime heat-ing over South America that propagates over 1000 kmoffshore. Based on our observations, we surmise that

FIG. 10. Comparison of 6-day buoy-period mean cloud and thermodynamicprofiles—liquid water content (LWC), water vapor (qv ), and potential tem-perature (qqqqq )—from EPIC with ECMWF and NCEP–NCAR operational analy-ses for the same time and location, and with Oct climatological profiles forrecent versions of two leading GCMs.

Au: Parungo etal. 1978 to matchreference cor-rect? If not pleaseprovide correctreference.

10 JULY 2004|

entrainment of dry, warm air is the primary regula-tor of cloud thickness in this region, but that drizzlemay be having an impact by inhibiting turbulence andpromoting mesoscale cloud cellularity.

Resolving the real feedback of drizzle on subtropi-cal stratocumulus clouds has important consequencesfor predicting the “indirect effect” of aerosols on cli-mate through their effect on cloud radiative proper-ties. Increased aerosol concentration tends to producehigher cloud droplet concentration, and smaller meancloud droplet size for a cloud of given liquid watercontent. One consequence, the first indirect effect, isto increase the droplet surface area, making the cloudmore reflective (Twomey 1977). A second and moreuncertain indirect effect is that the smaller mean clouddroplet size inhibits the development of precipitation,thereby increasing the thickness or lifetime of clouds(Albrecht 1989). Our EPIC cruise did not obtain insitu cloud microphysical data or a complete charac-terization of the ABL aerosol, but we did find evidencefor one piece of the second aerosol indirect effect—we measured less drizzle falling from a cloud of agiven thickness during periods of higher cloud drop-let concentration, though our 6-day sample is toosmall to be conclusive.

Modeling studies have also suggested that drizzlesuppression by increased aerosol increases marinestratocumulus thickness and albedo. However, inthese studies, the stratocumulus cloud base is suffi-ciently low that most of the drizzle falling out of thestratocumulus reaches the surface, a major differencefrom the regime observed in EPIC 2001. Sophisticateddynamical–microphysical models simulating the fullspectrum of observed cloud-topped marine bound-ary layers, as well as further observational studies, willbe necessary to understand whether the second indi-rect effect is in fact of a comparable magnitude to thefirst indirect effect, or whether it is in fact muchsmaller.

To this end, we are using various types of ABLmodels, ranging from mixed-layer models to single-column models that mimic what GCMs might simu-late to large-eddy simulation models, to get a betterunderstanding of the drizzle–turbulence–cloudthickness feedbacks hinted at by the EPIC observa-tions. The more comprehensive nighttime snapshotsof turbulence–drizzle–aerosol interactions fromDYCOMS-II should also be invaluable comparisons.The Global Energy and Water Cycle Experiment(GEWEX) Cloud System Study (GCSS) BoundaryLayer Cloud Working Group hopes to conduct aseries of large-eddy simulations (LESs) and single-column model intercomparison studies based on

both of these field experiments to improve ourunderstanding of what we think are large model-to-model differences in the simulation of drizzle micro-physics in stratocumulus and its feedbacks on stra-tocumulus cloud thickness and albedo.

The stratocumulus clouds help keep the underly-ing southeast Pacific Ocean cool by blocking sunlightfrom reaching the sea surface. Climate models mustfaithfully simulate this and other atmosphere–oceancoupling processes in this region to accurately predictthe sea surface temperature and atmospheric circu-lation over the entire Pacific Ocean for such applica-tions as ENSO and climate change prediction. Amajor goal of the ongoing measurements at theWHOI buoy is to better document atmosphere–oceancoupling in this region. For instance, the heat andsalinity evolution of the upper southeast Pacific Oceanare being studied by comparison of WHOI buoyobservations with a one-dimensional ocean model,revealing that the coolness of the SST is not just main-tained by the persistent cloudiness, but also by cool-ing associated with offshore advection modulated byoceanic Rossby waves generated near the SouthAmerican coast. Under the auspices of the interna-tional Climate Variations (CLIVAR) program’sVariability of the American Monsoons (VAMOS)initiative, the VAMOS Ocean Cloud AtmosphereLand Study (www.atmos.washington.edu/~breth/VOCALS/VEPIC_Science_Plan.pdf) is being devel-oped to coordinate continued enhanced measure-ments, and diagnostic and modeling studies of thecoupled atmospheric, oceanic, and land-drivenprocesses that govern the climate of this region.

ACKNOWLEDGEMENTS. The stratocumulus cruiseof EPIC 2001 was a cooperative effort among many sci-entists, students, staff, and the crew and officers of theBrown. Jay Fein of NSF and Mike Patterson of NOAA wereinstrumental in putting together the resources that madeit possible, in particular the last-minute addition ofUNAM aerosol measurements. The authors acknowledgesupport from NOAA grants, NSF Grant ATM-0082384,and an NDSEG Fellowship to K. Comstock.

REFERENCESAgee, E. M., 1982: An introduction to shallow convec-

tive systems. Cloud Dynamics, E. M. Agee andT. Asai, Eds., Reidel, 3–30.

Albrecht, B. A., 1989: Aerosols, cloud microphysics andfractional cloudiness. Science, 245, 1227–1230.

——, C. W. Fairall, D. W. Thomson, A. B. White, andJ. B. Snider, 1990: Surface-based remote sensing of

11JULY 2004AMERICAN METEOROLOGICAL SOCIETY |

the observed and the adiabiatic liquid water contentof stratocumulus clouds. Geophys. Res. Lett., 17, 89–92.

——, M. P. Jensen, and W. J. Syrett, 1995: Marine bound-ary layer structure and fractional cloudiness. J.Geophys. Res., 100, 14 209–14 222.

Bougeault, P., 1985: The diurnal cycle of the marine stra-tocumulus layer: A higher-order model study. J.Atmos. Sci., 42, 2826–2843.

Ciesielski, P. E., W. H. Schubert, and R. H. Johnson,2001: Diurnal variability of the marine boundarylayer durcing ASTEX. J. Atmos. Sci., 58, 2355–2376.

Cronin, M. F., N. Bond, C. Fairall, J. Hare, M. J. McPhaden,and R. A. Weller, 2002: Enhanced oceanic and atmo-spheric monitoring underway in eastern Pacific. Eos,Trans. Amer. Geophys. Union, 83, 210–211.

Dai, A., and C. Deser, 1999: Diurnal and semidiurnalvariations in global surface wind and divergencefields. J. Geophys. Res., 104, 31 109–31 125.

Dong, X., and G. G. Mace, 2003: Arctic stratus cloudproperties and radiative forcing derived fromground-based data collected at Barrow, Alaska. J. Cli-mate, 16, 445–460.

Fitzgerald, J. W., 1991: Marine aerosols: A review.Atmos. Enviorn., 25A, 533–545.

Foltz, G. S., and W. M. Gray, 1979: Diurnal variation inthe troposphere’s energy balance. J. Atmos. Sci., 36,1450–1466.

Garreaud, R. D., and R. Muñoz, 2004: The diurnal cyclein circulation and cloudiness over the subtropicalsoutheast Pacific: A modeling study. J. Climate, 17,1699–1710.

——, J. Rutllant, J. Quintana, J. Carrasco, and P. Minnis,2001: CIMAR-5: A snapshot of the lower troposphereover the subtropical southeast Pacific. Bull. Amer.Meteor. Soc., 82, 2193–2207.

Hoppel, W. A., and J. W. Frick, 1990: Submicron aero-sol size distributions measured over the tropical andsouth Pacific. Atmos. Environ., 24A, 645–659.

Jenkin, M., and K. Clemitshaw, 2000: Ozone and othersecondary photochemical pollutants: Chemical pro-cesses govering their formation in the planetaryboundary layer. Atmos. Environ., 34, 2499–2527.

Klein, S. A., and D. L. Hartmann, 1993: The seasonal cycleof low stratiform clouds. J. Climate, 6, 1587–1606.

——, ——, and J. R. Norris, 1995: On the relationshipsamong low-cloud structure, sea surface temperature,and atmospheric circulation in the summertimenortheast Pacific. J. Climate, 8, 1140–1155.

Lawrence, M. G., 1993: An empirical analysis of thestrength of the phytoplankton dimethylsulfide cloud

climate feedback cycle. J. Geophys. Res., 98, 20 633–20 673.

Lilly, D. K., 1968: Models of cloud-topped mixed layersunder a strong inversion. Quart. J. Roy. Meteor. Soc.,94, 292–309.

Minnis, P., P. W. Heck, D. F. Young, C. W. Fairall, andJ. B. Snider, 1992: Stratocumulus cloud propertiesderived from simultaneous satellite and island-basedinstrumentation during FIRE. J. Appl. Meteor., 31,317–339.

Parungo, F. P., C. T. Nagamoto, R. Madel, J. Rosinski,and P. L. Haagenson, 1978: Marine aerosols in ma-rine upwelling regions. J. Aerosol. Sci., 18, 277–290.

Petty, G. W., 1995: Frequencies and characteristics ofglobal oceanic precipitation form shipboard present-weather reports. Bull. Amer. Meteor. Soc., 76, 1593–1616.

Pincus, R., and M. B. Baker, 1994: Effect of precipita-tion on the albedo susceptibility of marine boundarylayer clouds. Nature, 372, 250–252.

Raymond, D. J., S. K. Esbensen, M. Gregg, and C. S.Bretherton, 2004: EPIC 2001 and the coupled ocean–atmosphere system of the tropical east Pacific. Bull.Amer. Meteor. Soc., in press.

Rinehart, R. E., 1997: Radar for Meteorologists. 3d ed.Rinehardt Publications, 428 pp.

Stephens, G. L., S. Ackerman, and E. A. Smith, 1984: Ashortwave parameterization revised to improvecloud absorption. J. Atmos. Sci., 41, 687–690.

Stevens, B., W. R. Cotton, G. Feingold, and C.-H.Moeng, 1998: Large-eddy simulations of stronglyprecipitating, shallow, stratocumulus-topped bound-ary layers. J. Atmos. Sci., 55, 3616–3638.

——, and Coauthors, 2003: Dynamics and chemistry ofmarine stratocumulus—DYCOMS-II. Bull. Amer.Meteor. Soc., 84, 579–593.

Twomey, S., 1977: Atmospheric Aerosols. Elsevier, 289 pp.Wood, R., and C. S. Bretherton, 2004: Boundary layer

depth, entrainment, and decoupling in the cloud-capped subtropical and tropical marine boundarylayer. J. Climate, in press.

——, ——, and D. L. Hartmann, 2002: Diurnal cycle ofliquid water path over the subtropical and tropicaloceans. Geophys. Res. Lett., 29, 2092, doi:10.1029/2002GL015371.

Wyant, M. C., C. S. Bretherton, H. A. Rand, and D. E.Stevens, 1997: Numerical simulations and a concep-tual model of the subtropical marine stratocumulusto trade cumulus transition. J. Atmos. Sci., 54, 168–192.

Au: pleaseverify 1994 iscorrect year

Au: please pro-vide title of ap-pendix and pagerange if you spe-cifically wouldlike to refer tothat in Rinehartreference