The PreVOCA experiment: modeling the lower troposphere in the Southeast Pacific

M. C. Wyant; R. Wood; C. S. Bretherton; C. R. Mechoso; J. Bacmeister; M. A. Balmaseda; B. Barrett; F. Codron; P. Earnshaw; J. Fast; C. Hannay; J. W. Kaiser; H. Kitagawa; S. A. Klein; M. Köhler; J. Manganello; H.-L. Pan; F. Sun; S. Wang; Y. Wang

doi:10.5194/acp-10-4757-2010

Wyant, M. C.;Wood, R.;Bretherton, C. S.;Mechoso, C. R.;Bacmeister, J.;Balmaseda, M. A.;Barrett, B.;Codron, F.;Earnshaw, P.;Fast, J.;Hannay, C.;Kaiser, J. W.;Kitagawa, H.;Klein, S. A.;Köhler, M.;Manganello, J.;Pan, H.-L.;Sun, F.;Wang, S.;Wang, Y.;

2010-05-26 00:00:00

Atmos. Chem. Phys., 10, 4757–4774, 2010 Atmospheric www.atmos-chem-phys.net/10/4757/2010/ Chemistry doi:10.5194/acp-10-4757-2010 © Author(s) 2010. CC Attribution 3.0 License. and Physics The PreVOCA experiment: modeling the lower troposphere in the Southeast Paciﬁc 1 1 1 2 3 4 5 M. C. Wyant , R. Wood , C. S. Bretherton , C. R. Mechoso , J. Bacmeister , M. A. Balmaseda , B. Barrett , 6 7 8 9 4 10 11 4 F. Codron , P. Earnshaw , J. Fast , C. Hannay , J. W. Kaiser , H. Kitagawa , S. A. Klein , M. Kohler ¨ , 12 13 2 14 15 J. Manganello , H.-L. Pan , F. Sun , S. Wang , and Y. Wang Department of Atmospheric Sciences, University of Washington, Seattle, WA, USA Department of Atmospheric and Oceanic Sciences, University of California, Los Angeles, Los Angeles, CA, USA Global Modeling and Assimiliation Ofﬁce, NASA Goddard Space Flight Center, Greenbelt, MD, USA Research Department, European Center for Medium-Range Weather Forecasts, Reading, UK Department of Geophysics, University of Chile, Santiago, Chile Universite ´ Pierre et Marie Curie, Laboratoire de Meteorologie Dynamique, Paris, France Met Ofﬁce, Exeter, Devon, UK Atmospheric Sciences and Global Change, Paciﬁc Northwest National Laboratory, Richland, WA, USA Climate and Global Dynamics Division, National Center for Atmospheric Research, Boulder, CO, USA Meteorological College, Japan Meteorological Agency, Tokyo, Japan Program for Climate Model Diagnosis and Intercomparison, Lawrence Livermore National Laboratory, Livermore, CA, USA Center for Ocean-Land-Atmosphere Studies, Calverton, MD, USA Environmental Modeling Center, National Centers for Environmental Prediction, Camp Springs, Maryland, USA Marine Meteorology Division, Naval Research Laboratory, Monterey, CA, USA International Paciﬁc Research Center, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, Honolulu, HI, USA Received: 13 October 2009 – Published in Atmos. Chem. Phys. Discuss.: 11 November 2009 Revised: 22 April 2010 – Accepted: 7 May 2010 – Published: 26 May 2010 Abstract. The Preliminary VOCALS Model Assessment America. Mean-monthly model surface winds agree well (PreVOCA) aims to assess contemporary atmospheric mod- with QuikSCAT observed winds and models agree fairly well eling of the subtropical South East Paciﬁc, with a particular on mean weak large-scale subsidence in the region next to the focus on the clouds and the marine boundary layer (MBL). coast. However they have greatly differing geographic pat- Models results from fourteen modeling centers were col- terns of mean cloud fraction with only a few models agreeing lected including operational forecast models, regional mod- well with MODIS observations. Most models also underes- els, and global climate models for the month of October timate the MBL depth by several hundred meters in the east- 2006. Forecast models and global climate models produced ern part of the study region. The diurnal cycle of liquid wa- ◦ ◦ daily forecasts, while most regional models were run con- ter path is underestimated by most models at the 85 W 20 S tinuously during the study period, initialized and forced at stratus buoy site compared with satellite, consistent with pre- the boundaries with global model analyses. Results are vious modeling studies. The low cloud fraction is also under- compared in the region from 40 S to the equator and from estimated during all parts of the diurnal cycle compared to ◦ ◦ 110 W to 70 W, corresponding to the Paciﬁc coast of South surface-based climatologies. Most models qualitatively cap- ture the MBL deepening around 15 October 2006 at the stra- tus buoy, associated with colder air at 700 hPa. Correspondence to: M. C. Wyant ([email protected]) Published by Copernicus Publications on behalf of the European Geosciences Union. 4758 M. C. Wyant et al.: The PreVOCA experiment 1 Introduction eling during REx. Available for evaluating the models are ship-based observations from National Oceanographic and The atmosphere-ocean system in Southeast Paciﬁc (SEP) is Atmospheric Administration (NOAA) and the Woods Hole interesting for many reasons. The SEP is a region of sig- Oceanographic Institution (WHOI) cruises from the coast to ◦ ◦ niﬁcant coastal upwelling and extensive persistent low-level the stratus buoy at 20 S 85 W, a large suite of satellite mea- clouds (Klein and Hartmann, 1993), whose sensitivity to cli- surements, and operational meteorological analysis and re- mate change is of great interest. The region plays a large analysis. The month of October 2006 was chosen because of role in the El-Nino Southern Oscillation. It also hosts large the availability of diverse satellite and ship observations and contrasts of aerosol concentrations, from extremely clean because it matches the seasonal timing of REx. remote maritime conditions to highly polluted conditions While a major goal of VOCALS is to study the interaction downstream of large coastal point sources such as copper of aerosols and the marine boundary layer, most of the mod- smelters. els participating in PreVOCA have very limited representa- A major ﬁeld campaign, the East Paciﬁc Investigation of tions of such interactions. Although many models do include Climate (EPIC), was conducted in 2001 in the SEP (Ray- the direct radiative effect of aerosol, most do not treat the mond et al., 2004; Bretherton et al., 2004). One of EPIC’s transport and dispersion of aerosols, but instead use clima- goals was to examine the interaction of microphysics, clouds, tological values of aerosol concentration. Most models use and the marine boundary layer (MBL). At the stratus buoy lo- single moment bulk microphysical schemes, which do not ◦ ◦ cation (20 S 85 W) EPIC found a strongly diurnally vary- utilize aerosol information. Because of these limitations, we ing stratocumulus cloud cover with very little cumulus con- proceed under the assumption that aerosol impacts are sec- vection (Bretherton et al., 2004). Wood et al. (2002) also re- ondary to other physics, and instead focus on the modeling ported a strong diurnal variation in liquid water path (LWP) of the MBL and clouds independent of aerosol effects. in the region. Other SEP large-scale cloud-related studies We are motivated then by the following questions: Do the have focused on the seasonal cycle of clouds in the region models simulate the large scale conditions adequately? Do (Klein and Hartmann, 1993), inﬂuence of clouds on the cou- the models agree on the vertical structure of the MBL? Do pled climate system (Ma et al., 1996), the effects of topog- the models capture the basic cloud regimes and the MBL suf- raphy on subtropical stratocumulus clouds (Xu et al., 2004; ﬁciently well? Are the simulations and predictions of such Richter and Mechoso, 2004, 2006), effects of the diurnal a quality that will support the models use in studies of cli- heating over the Andes on the MBL (Garreaud et al., 2001; mate change, aerosol and chemical transport, aerosol indirect Garreaud and Munoz, 2004; Wood et al., 2009), sub-seasonal effects, and aerosol-cloud interactions? MBL variability (Xu et al., 2005), and inter-annual variations The outline of this paper is as follows: Sect. 2 describes of MBL depth (Zuidema et al., 2009). the setup of PreVOCA and brieﬂy describes the participat- Despite advances in observing and understanding the SEP, ing models. Additional model details are provided in Ap- general circulation models (GCMs) typically do not repre- pendix A. Section 3 compares the mean monthly prediction sent this region well. The recent study of Hannay et al. of the models with observations. The diurnal cycle of the (2009) compared forecasts of the Geophysical Fluid Dynam- MBL is examined in Sect. 4, and the response of the MBL ics Laboratory (GFDL) and the National Center for Atmo- region to synoptic variations is discussed in Sect. 5. A con- spheric Research (NCAR) Community Atmospheric Model cluding discussion is presented in Sect. 6. (CAM) GCMs and the ECMWF forecast model to six days of October 2001 EPIC observations at the stratus buoy. All models produced a shallower boundary layer than observed 2 Experiment setup and had amplitude and phase errors in the diurnal cycle of LWP compared to observations, while the ECMWF model Our focus is on the maritime region off the west coast of ◦ ◦ performed better than the GCMs in terms of cloud predic- South America from the equator to 40 S and from 70 W to ◦ ◦ tion. 110 W. This region encompasses the stratus buoy at 85 W The SEP is also the focus of VOCALS (VAMOS Ocean 20 S and the region east of it, which was the primary focus Cloud-Atmosphere-Land Study). In preparation for the VO- of REx. CALS Regional Experiment (REx), which took place from The diverse collection of models participating in Pre- mid-October through mid-November 2008, we organized an VOCA and some of their run parameters are summarized in assessment of current atmospheric modeling capability. Pre- Table 1. We loosely categorize the models into three groups: VOCA (the Preliminary VOCALS model Assessment) com- operational, regional, and climate. Operational models are pares a large and diverse collection of models simulating the short and medium range forecast models that are run globally SEP during the period of October 2006. These include op- and typically involve a data assimilation system. Regional erational forecast models, regional models, and global cli- models are run over a more limited area at higher horizon- mate models. The main goals are to understand current tal resolution. They rely on boundary conditions provided model biases and their implications for forecasting and mod- by other models, and are most frequently used for mesoscale Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/ M. C. Wyant et al.: The PreVOCA experiment 4759 Table 1. Participating models. Name Type Forecast Freq. Forecast Hours Horizontal Vertical Levels Investigators −1 [d ] Analyzed Resolution (σ>0.8) (inner domain) [km] NRL COAMPS Regional 2 6–18 81 (27) 45(24) S. Wang COLA RSM Regional – – 50 28(8) J. Manganello V. Misra IPRC Reg-CM Regional – – 25 28(10) Y. Wang (IRAM) LMDZ Regional – – 250(50) 38(10) F. Codron PNNL Regional – – 45(15) 44(26) J. Fast (WRF-Chem) W. Wang E. Chapman U. Chile (WRF) Regional – – 45 43(19) B. Barrett UCLA (WRF) Regional – – 45(15) 34(8) F. Sun A. Hall X. Qu ECMWF OPER Operational 2 0–12 25 91(16) M. Kohler ¨ J. Kaiser ECMWF 5-DAY Operational 1 48–72 40 91(16) M. Kohler ¨ NASA GMAO Operational 4 0–6 55 72(14) J. Bacmeister GEOS5-DAS JMA Operational 4 24–30 60 60(13) H. Kitagawa T. Komori H. Onoda NCEP GFS Operational 1 12–36 38 64(15) H.-L. Pan R. Sun UKMO Operational 1 12–36 40 50(9) P. Earnshaw S. Milton ECMWF Climate – – 125 62(16) M. Kohler ¨ Coupled Ens. M. Balmaseda NCAR CAM 3.5/CAM 3.6 UW Climate 1 48–72 250 26(4)/30(8) C. Hannay GFDL AM2 Climate 1 48–72 250 24(10) S. Klein M. Zhao research. The climate models are the atmospheric compo- There are two simulation modes among the runs presented nents of coupled atmosphere-ocean climate models. They here: forecast, and continuous. In forecast mode, models have much coarser horizontal resolution than the regional made daily forecasts initialized by operational analysis, with models and are designed to balance global energy budgets. a few models making more frequent forecasts (see Table 1). We choose to place the ECMWF coupled ensemble forecast For these runs, a speciﬁed subset of forecast hours for each model in the “climate” category, while we place the LMDZ daily run was selected, and these were stitched together to climate model in the “regional” category because it has been provide a continuous month of model output for comparison run in a mode and resolution in the SEP more typical of re- with other models. gional models than climate models. Typical horizontal reso- For each forecast, we expect the vertical structure of the lutions are about 50 km for regional and operational models, MBL to drift away from the initial analysis towards a model- and 250 km for climate models. Some models use nested dependent preferred state, while other supporting features of grids over the study region with as high as 15 km resolu- the forecast do not deviate far from analysis, highlighting bi- tion. The number of vertical levels varies from 24 to 91, ases in the MBL. This approach to identifying parameteri- and all models except for CAM 3.5 have 8 or more levels zation biases applied to GCMs is described in Phillips et al. in the boundary layer. Detailed descriptions of the models (2004) and has been used in many recent studies (e.g., Klein including boundary-layer and cloud schemes, microphysics et al., 2006; Williamson and Olson, 2007; Boyle et al., 2008; schemes, and aerosol treatments are provided in Appendix A. Hannay et al., 2009). www.atmos-chem-phys.net/10/4757/2010/ Atmos. Chem. Phys., 10, 4757–4774, 2010 4760 M. C. Wyant et al.: The PreVOCA experiment Fig. 1. October 2006 mean AMSR-E SST(K) and QuikSCAT winds −1 (10 ms scale plotted). In continuous mode, the model ﬁelds are initialized by analysis, and then run for the entire month from initial con- ditions provided by analysis datasets, which also provide time dependent boundary conditions at the edges of the do- main. Because the model ﬁelds are not re-initialized regu- larly, greater model biases are expected than in the forecast- mode runs. The models’ output is provided every 3 simulated hours. All runs were made with speciﬁed SST except for the ECMWF coupled ensemble, which runs coupled to the Ham- burg Ocean Primitive Equation model (HOPE, Wolf et al., −1 1997). The SSTs in the study region typically vary by as Fig. 2. October 2006 mean 10-m wind speed (ms ) and wind vectors for QuikSCAT (upper left) and a selection of models. much as 0.5 K across models with discrepancies as large as 1 K near the coast. colder to warmer water promoting surface sensible and la- tent heat ﬂux into the MBL. Those winds are very steady 3 Monthly averages in the north part of the domain, which is free from strong We ﬁrst examine monthly mean ﬁelds corresponding to Oc- disturbances. In the southern part of the domain, westerlies tober 2006 to illustrate important aspects of the model sim- prevail and eastward propagating disturbances typically pass ulations of the MBL. For many ﬁelds we will show means every few days. In the eastern part of the domain, the lower from a representative subset of models rather than showing tropospheric stability is large (Klein and Hartmann, 1993; all models in order to simplify the presentation. Wood and Hartmann, 2006), typically associated with very strong inversions at the top of the MBL. Surface precipita- Figure 1 shows the monthly mean SST from the Advanced −1 tion, as estimated by AMSR-E, is light (<0.5 mmd ) across Microwave Scanning Radiometer-EOS (AMSR-E) and 10- the study region, except south of 30 S where mid-latitude m wind vectors from Quick Scatterometer (QuikSCAT, used disturbances pass, and the modeled surface precipitation is here as processed in Field and Wood, 2007). There is limited −1 similarly weak (<1 mmd ) for most models. blending of 10-m winds with NCEP (National Centers for Environmental Prediction)/NCAR Reanalysis (Kalnay et al., We compare QuikSCAT 10-m wind direction and speed 1996) for missing data. The study region is dominated by with a representative selection of models in Fig. 2 (Note that the eastern part of the South Paciﬁc subtropical high. Sur- the CAM 3.5 winds are instead from the lowest model grid −1 face southeasterly winds with speeds of 5–9 ms prevail level at about 64m). The models generally show excellent over most the northern half of the region, with southerly agreement with the observed surface winds. The mean po- winds along coast inﬂuenced by the high coastal topogra- sition of the surface anticyclone is well agreed upon by the phy. The winds over most of the study region blow from models. The weakness of the southerlies along the coast in Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/ M. C. Wyant et al.: The PreVOCA experiment 4761 the GCMs is likely due to their coarse horizontal resolution. Note that QuikSCAT observations are assimilated into all of the operational model systems, though the model winds are free to diverge from QuikSCAT during their forecasts. The good agreement of the monthly mean winds with observa- tions and with each other can be attributed to good initial- ization and the to the relative importance of static large-scale features such as the subtropical high and trade wind circula- tions which are relatively easy to forecast, compared to the time-dependent features. Associated with the South Paciﬁc high is broad subsi- dence. The mean October 2006 subsidence at 850 hPa is shown for several models in Fig. 3. Weak subsidence −1 −1 (∼0.03 Pas or ∼25 hPad ) prevails across most of the −1 study region, with stronger subsidence (>0.05 Pas ) near the Chilean coast at about 30 S. The main region of model disagreement is the northwest part of the domain, where, for example, the IPRC model has stronger subsidence and the PNNL has slightly weaker subsidence. Note that the domain- mean subsidence in regional models is determined by the horizontal winds imposed at the side boundaries. The small- scale spatial variability of monthly-mean subsidence in some models is a common behavior of high-resolution models and depends largely on their numerics and handling of topogra- phy. The models also generally agree on the geographic pat- tern of mean lower tropospheric stability (not shown). Since the models’ SSTs are speciﬁed, this is mostly an indicator of agreement of temperature within a few degrees K at 700 hPa. −1 Fig. 3. October 2006 mean subsidence at 850 hPa (Pas ). NCEP Despite the generally close agreement among the horizon- and ECMWF analysis is shown in lieu of observed subsidence. The tal wind, vertical velocity, and static stability ﬁelds, the mod- stratus buoy location is indicated with a triangle. els show a large disagreement in their cloud properties. The mean-monthly cloud fraction from the Moderate Resolution Imaging Spectroradiometer (MODIS) and low-cloud fraction but is still around 0.7 near the stratus buoy, dropping off to for all models are shown in Fig. 4. The MODIS cloud frac- 0.4 in the peripheral parts of the study region. The mod- tion in Fig. 4a (as computed in Field and Wood, 2007) is els show a large disparity in cloud fraction despite the sim- based on liquid-water retrieval and mostly represents low ilarity in forcing discussed above. A common model prob- cloud. It excludes ice cloud which is primarily found south lem is too little low cloud near the coast from 30 S to the of 30 S associated with passing synoptic disturbances. The equator, as exempliﬁed in the IPRC and GFDL runs and timing of MODIS data corresponds to about 10:30 a.m. LT, in the NCEP analysis. The models also vary greatly in the but modeled and observed stratocumulus cloud properties at amount of cloud in the north central and northwest part of this time are typically not too far from their 24-h mean. the study region where trade-cumulus convection is more sig- Model low cloud fractions are computed from the surface niﬁcant, with many models producing too much cloud com- to 800 hPa for many of the models, though for several mod- pared to MODIS (e.g. PNNL, UCLA, and COLA, LMDZ, els (CAM, COLA, GFDL, IPRC, and LMDZ) low cloud is and UKMO), and several models too little cloud in this re- computed from the surface to about 700 hPa. This differ- gion including IPRC, JMA, UCHILE and especially CAM ence has minimal consequence north of 25 S except for the 3.6 UW. The modeled geographic patterns in the subsidence LMDZ model which has substantial cloud between 700 hPa are generally not reﬂected in the modeled low-cloud frac- and 800 hPa that is captured in Fig. 4. Models differ substan- tion, except for the signiﬁcant clearing near the coast south tially in their cloud overlap assumptions, which are utilized of 30 S collocated with very strong subsidence. This local- in computing low-cloud fraction in Fig. 4. ized strong subsidence forces the entrainment of very warm The MODIS cloud fraction is 0.8–0.9 near the coast ex- and dry above-inversion air into the boundary layer, dissipat- ◦ ◦ cept for a relatively clear region between 30 S and 40 S as- ing cloud. Overall the ECMWF family of models and the sociated with strong subsidence there. Mean cloud fraction UKMO do fairly well in matching the mean MODIS cloud decreases moving away from the cloudy part of the coast, fraction in the region. www.atmos-chem-phys.net/10/4757/2010/ Atmos. Chem. Phys., 10, 4757–4774, 2010 4762 M. C. Wyant et al.: The PreVOCA experiment Fig. 4. October 2006 MODIS total cloud fraction (upper left) and modeled monthly-mean low-cloud fraction. The mean liquid water path (LWP) from several satel- sonal cycle (e.g., Colbo and Weller, 2007). Downwelling lites (AMSR-E, TRMM Microwave Instrument (TMI), the shortwave radiation is shown in Fig. 6 for almost all mod- Special Sensor Microwave Imager (SSM/I) F13 and F15, els together with the observed monthly mean value from L. O’Neill, personal communication, 2009) is compared with International Satellite Cloud Climatology Project (ISCCP) the models’ total (clear-sky plus cloudy sky) grid-box LWP FD data (Zhang et al., 2004). The downwelling radiation −2 −2 in Fig. 5. Setting aside the extreme south part of the study varies between about 220 Wm and 300 Wm at the stra- region which is inﬂuenced by mid-latitude synoptic systems, tus buoy location. For each model, geographical biases the observed LWP has a very broad maximum in the north in downward shortwave radiation compare very closely to central part of the study region. This is well west of the near- those in cloud fraction. This connection is further shown coastal low-cloud maximum observed in MODIS, and is re- in Fig. 7, which plots each model’s mean downward short- ◦ ◦ lated to higher SSTs away from the coast. Several models wave ﬂux versus low cloud fraction in a 5 ×5 box centered ◦ ◦ broadly underestimate LWP (e.g. GFDL). Some models un- at 85 W 20 S. While there are clearly many factors inﬂu- derestimate LWP more in the eastern part of the region (e.g. encing the shortwave radiation reaching the surface, cloud PNNL and NCEP (not shown)), while others underestimate fraction plays a major role. In many models, the substantial LWP in the western part (CAM 3.5). A few models obtain under-prediction of clouds from the buoy region eastwards the basic mean pattern that is qualitatively correct (ECMWF to the South American coast results in very large positive bi- −2 models, GMAO, and UKMO, mostly not shown). ases (sometimes larger than 100 Wm ) in downwelling SW compared to ISCCP. Biases such as these would substantially The large model discrepancies in LWP and cloud fraction increase regional SSTs in coupled simulations. have substantial implications for the surface energy budget, especially the downwelling shortwave radiation at the sur- face, a major driver of the SST spatial distribution and sea- Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/ M. C. Wyant et al.: The PreVOCA experiment 4763 Model soundings at the stratus buoy are compared next to further explore inter-model differences. Figure 8 com- pares the mean October 2006 soundings of speciﬁc humidity and potential temperature for the models with an observed climatological sounding. The NOAA/ESRL soundings (see de Szoeke et al., 2009) are the average of data from 169 rawinsondes launched near the stratus buoy during October 2001, 2005, 2006, and 2007 covering 30 total days. The mean sounding indicates a fairly well mixed layer capped by a strong inversion at about 870 hPa (1450 m) and very stable and dry above-inversion conditions at the buoy. Of the three categories, the operational model sound- ings show the least spread and agree fairly well with the ESRL soundings. The MBLs in most models are shal- lower and moister than the climatology. All of the model- mean-inversions are less sharp than observed. Note that this sharpness is inﬂuenced by temporal variability in MBL depth. Above 800 hPa the agreement among operational soundings with the ESRL climatology is generally excel- lent. The MBLs in the regional models are also generally shallower and slightly moister than the climatology, and the mean-inversions are less sharp than observed, but with much larger spread than the operational models. The larger spread in MBL depths appears to be connected to a similar feature in potential temperatures above 700 hPa. The climate model soundings spread is less than that of the regional models, but they also exhibit shallower and moister boundary layers than the climatology. CAM 3.6 UW appears to be an improve- ment over CAM 3.5 in regard to the latter behavior. The variation in vertical structure of the boundary layer Fig. 5. October 2006 mean satellite liquid water path (upper left, is also evident in proﬁles of cloud condensate along 20 S, −2 −2 gm ) and modeled liquid water path (gm ). shown in Fig. 9, for which we have no direct observa- tional comparison. The transect along 20 S does not follow a boundary layer trajectory (the boundary layer winds are southeasterly), but it does illustrate the expected deepening top of the highest model level where the relative humidity of the boundary layer to the west as SST increases. This is exceeds 60%. For almost all models, this level is coincident similar to the pattern seen with increasing SST in the North with a sharp decline in cloud condensate with height, and East Paciﬁc in observations and models (e.g., Siebesma et al., is near the strongest vertical gradient in potential temper- 2004). Like the North East Paciﬁc, we expect solid stratocu- ature. We use three climatologies of boundary-layer depth mulus near shore transitioning to more broken trade cumulus or cloud-top height to compare with. Constellation Observ- convection further west. ing System for Meteorology, Ionosphere, and Climate (COS- The vertical distribution of condensate varies substantially MIC, Anthes et al., 2008) boundary layer depths are esti- between models. CAM 3.5 produces a shallow fog layer near mated from the height of the largest virtual potential temper- the coast and a shallow and thin cloud layer further west ature gradient. Because of the relative sparsity of measure- compared to the other models shown. Other previous ex- ments, the values shown are averaged from 15–25 S. The periments with CAM suggest that this behavior is not simply Cloud-Aerosol Lidar and Infrared Pathﬁnder Satellite Obser- due to its coarser vertical resolution than most other models vation (CALIPSO) cloud-top height climatology (Wu et al., studied here. At the other extreme the IPRC condensate has 2008) is created from the Level 2 lidar-based cloud layer broad vertical extent and cloud condensate extends above the product, and is here averaged from 17–23 S. The MODIS- 700-hPa level west of 100 W. derived cloud-top heights are based on cloud-top temperature The maximum height of condensate is closely tied to the inferred from measured 11-μm radiances (Zuidema et al., cloud-top height and with the boundary-layer depth. We plot 2009) and are averaged here from 19–21 S. We also plot the model-estimated MBL depths along 20 S together with the mean boundary layer depth estimated from NOAA/ESRL a number of observational climatologies in Fig. 10. For each soundings from Fig. 7, using the height where the RH=60%. model, the model boundary layer depth is estimated as the (Ship based radar estimates of boundary-layer depth from www.atmos-chem-phys.net/10/4757/2010/ Atmos. Chem. Phys., 10, 4757–4774, 2010 4764 M. C. Wyant et al.: The PreVOCA experiment −2 Fig. 6. October 2006 mean downwelling SW radiation at the surface from ISSCP and models (Wm ). a subset of the days sampled by the soundings have a mean depth about 50 m less). East of 90 W, the various observa- tional climatologies agree to within 300 m. West of 90 W COSMIC heights are about 300–500 m higher than the other observed estimates, possibly because cloud-top height be- gins to diverge from inversion base height in the trade-wind MBL. Comparison of MODIS and CALIPSO mean cloud- top heights with model boundary-layer depths west of 90 W, especially in the second half of October, is potentially prob- lematic because clouds are frequently absent in that region. The discrepancy between model and observation is largest near the coast, with the typical modeled MBL depth sig- niﬁcantly shallower than satellite estimates. Moving west from the buoy, the models tend to deepen the boundary layer much more rapidly than MODIS and CALIPSO. To the west of 90 W the model consensus MBL depth agrees well with MODIS and CALIPSO, though not with COSMIC. Fig. 7. Mean October 2006 downwelling shortwave radiation at ◦ ◦ The tendency of GCMs to underestimate MBL depth at the the surface versus low cloud fraction for a 5 ×5 box centered at ◦ ◦ stratus buoy has been noted in multiple studies (Bretherton 85 W 20 S. The observed value (red x) comes from ISCCP FD et al., 2004; Hannay et al., 2009). It is unclear why GCMs data and MODIS liquid cloud fraction. and other models consistently underestimate this and why the underestimate is especially strong near the coast, though in Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/ M. C. Wyant et al.: The PreVOCA experiment 4765 Fig. 9. October 2006 mean 20 S cross sections of model liquid −1 water content (gkg ). munication, 2009). The ﬁt uses data from TMI, SSM/I, and AMSR-E, though the diurnal amplitude and phase effectively comes only from TMI. Most models have a weaker diur- −1 Fig. 8. Soundings of speciﬁc humidity (left, kgkg ) and potential nal cycle in LWP than the observed at the buoy. However, ◦ ◦ temperature (right, K) at 20 S 85 W. Model soundings are Oc- when normalized by 24-h mean LWP, the amplitude of the tober 2006 means for regional (top), operational (middle) and cli- models diurnal cycle compares well with the observed am- mate (bottom). NOAA/ESRL soundings (black) are an average of plitude. For several models the discrepancy with observed several-day October periods in multiple years (see text). LWP could be interpreted as an error in geographic place- ment of the maximum mean LWP; the mean observed TMI liquid water path (Fig. 5) is near its maximum at the buoy the latter case the model under-resolution of topographic fea- while many models have their maximum LWP further to the tures may be to blame (R. Garreaud, personal communica- west or northwest (e.g. PNNL). The observed LWP peaks tion, 2009). around 04:00–05:00 a.m. LT, consistent with the well-known diurnal radiative modulation of stratocumulus cloud thick- ness. For most models the phase of the LWP agrees fairly 4 Diurnal cycle well with the observed phase. We next consider the mean October diurnal cycle in the vicin- The diurnal cycle of cloud fraction near the buoy is com- ity of the stratus buoy. Figure 11 shows the composite diurnal pared to observational climatologies in Fig. 11b. The Ghate cycle in LWP (averaged over cloudy and clear atmospheric et al. (2009) climatology of cloud fraction (thick black line) ◦ ◦ columns) and low-cloud fraction of the models for a 1 ×1 is derived from September–November (SON) measurements box centered on the buoy (hour 00:00 LT is 05:00 UTC). To of downward longwave radiation made at the stratus buoy compare with modeled and observed LWP (Fig. 11a), we from 2001–2005. The Extended Edited Synoptic Cloud Re- have plotted a best ﬁt sinusoid (black line) at the buoy over ports Archive (EECRA) cloud fraction (thin black line) is October 2006 from satellite data (L. O’Neill, personal com- based on 1956–1997 SON climatology of ship-based surface www.atmos-chem-phys.net/10/4757/2010/ Atmos. Chem. Phys., 10, 4757–4774, 2010 4766 M. C. Wyant et al.: The PreVOCA experiment Fig. 10. October 2006 model boundary layer depth (m) compared with observations of boundary layer depth and cloud-top height. Fig. 11. Composite October 2006 diurnal cycle of (a) liquid water Model mean (solid black line), 25–75 percentile range (dark gray), −2 ◦ path (gm ) and (b) low cloud fraction at the IMET buoy at 85 W and model range (light gray) are plotted. COSMIC October 2006 20 S. In (a) satellite observations from 2006 (see text) are plotted boundary layer depth sampled over 15–25 S is plotted in green. (thick black line) and in (b) observed climatologies of cloud fraction CALIPSO cloud-top height is plotted in magenta. MODIS cloud- from the buoy (Ghate et al., 2009, thick black line) and EECRA top heights are plotted in red from Zuidema et al. (2009). The (Hahn and Warren, 1999, thin black line) are plotted. mean depth (blue x) is an October climatology estimated from NOAA/ESRL soundings taken near the stratus buoy, with standard deviation plotted. radiation measurements is signiﬁcant but perhaps not entirely surprising because of the different methodologies used. It is observations (Hahn and Warren, 1999) of whole-sky cover of therefore difﬁcult to assess the realism of the models’ diurnal clouds with cloud-base below 3000 ft. The diurnal mean of cycle of low clouds. both of these measures is slightly higher than that of MODIS Also of interest is the diurnal “upsidence” wave modeled (∼0.7, Fig. 4); this is due at least in part to differences in in Garreaud and Munoz (2004). This wave of upward-motion the deﬁnitions of cloud fraction. The model consensus cloud in the lower troposphere is believed to be a response to di- fraction near the buoy is much lower than any of these obser- urnal heating over the Andes. In Fig. 12 we show the di- vational estimates. ◦ urnal composite pressure velocity ω at 850 hPa along 20 S The Ghate et al. (2009) cloud fraction, which has better in various models (hour 00:00 LT is 05:00 GMT). For most resolution in time than the EECRA, shows a pronounced models (including those not shown), a clear westward prop- afternoon and early-evening drop. Many models also have agating upsidence wave is present with a phase speed similar −1 strong afternoon cloud fraction reduction, though most of to the estimated 30 ms reported in Garreaud and Munoz −1 these same models also have a strong morning peak not (2004) and the estimate of 25 ms in Wood et al. (2009) present in the observations. The amplitudes of the models’ from QuikSCAT-retrieved winds. In the ﬁgure, the white diurnal cycle of low-cloud fraction range widely from about line is the phase of the wave maximum, assumed to be par- 0.04 to 0.6. The strong mean diurnal cycle in LWP is there- allel to the Peruvian coast, projected on to 20 S assuming −1 fore primarily an oscillation of cloud fraction in some models the wave travels southwestwards at 30 ms . This distur- and of cloud thickness in others. bance typically starts near the coast around 05:00 p.m. local The signiﬁcant disagreement in the amplitude of the di- time and reaches the stratus buoy around 02:00 a.m. LT. Of urnal cycle of cloudiness between EECRA reports and buoy the models shown in Fig. 12, the wave is not pronounced Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/ M. C. Wyant et al.: The PreVOCA experiment 4767 Fig. 13. Time series of 24-h averaged (a) 10-m zonal velocity −1 −1 (ms ), (b) 10-m meridional velocity (ms ), and (c) liquid wa- −2 ◦ ◦ ◦ ◦ ter path (kgm ) for a 3 ×3 box centered at 20 S 85 W. Model means (light black line), 25th–75th percentile range (dark gray), and model range (light gray) are plotted. Heavy black lines indicate observations; QuikSCAT in (a) and (b) and AMSR-E LWP in (c), respectively. Day 0 on the time axis corresponds to 0 Z, 1 Octo- ber 2006. dynamical regime is maintained albeit with substantial varia- ◦ ◦ Fig. 12. Hovmuller ¨ diagram along 20 S of a 24-h composite (re- tions in cloudiness. Along 20 S, muted effects of the higher −1 peated twice for clarity) of omega at 850 hPa (Pas ) of selected latitude systems remain. −1 models for October 2006. White line is phase speed of 30 ms . Figure 13a and b shows the daily-averaged modeled and QuikSCAT surface winds at the stratus buoy. The daily-mean u-component of the wind is easterly for the entire month of in the NCEP and IPRC simulation, and rather weak in the −1 October, and its magnitude ranges from near 0 to 9 ms . CAM 3.5. While this upsidence wave appears to result in The models generally capture the synoptic variability well, a slight boundary layer deepening (∼50 m) in models where but overestimate the easterly component during the last half it occurs, its effect on modeled cloud fraction and liquid wa- of the month compared with QuikSCAT. The v-component ter path appears to be small based on diurnal composites of of the wind also blows from the same direction (southerly) these quantities at 20 S (not shown). for the entire month but exhibits substantially less synop- tic variability than the u-component, with magnitudes only −1 ranging from 3 to 6 ms . The LWP at the buoy (Fig. 13c) 5 Synoptic variability as measured by AMSR-E shows strong variations on day-to- day timescales that are not matched by the model consensus We next examine the daily variations by the models through- (Fig. 13c) or by individual models (not shown). The mod- out the month of October 2006 associated with synop- els LWP variations do not agree well with each other, and no tic changes. We focus on 20 S, which includes the particular model matches the AMSR-E LWP variations well. NOAA/WHOI stratus cruises and was major focus of the 2008 REx ﬁeld campaign. South of 20 S there is signiﬁcant The ﬁrst row of Fig. 14 shows daily averaged observations synoptic activity, as austral springtime mid-latitude cyclones of MODIS liquid cloud fraction and cloud-top height de- brush by every few days. To the north of 20 S, the inﬂu- rived from MODIS (Zuidema et al., 2009). The white areas ence of these midlatitude systems is weak and a more static in cloud fraction indicate missing retrievals due to clearing www.atmos-chem-phys.net/10/4757/2010/ Atmos. Chem. Phys., 10, 4757–4774, 2010 4768 M. C. Wyant et al.: The PreVOCA experiment Fig. 14. Hovmuller ¨ diagrams of 24-h means along 20 S for October 2006. MODIS cloud fraction and cloud top height (m) are plotted in the top row. The bottom two rows are low cloud fraction, boundary layer depth (m), and potential temperature at 700 hPa (K) for the ECMWF Fig. 14. Hovmuller ¨ diagrams of 24-h means along 20 S for October 2006. MODIS cloud fraction and operational model and CAM 3.5, respectively. The white line indicates the stratus buoy position. All data are horizontally averaged over ◦ ◦ cloud top height a 1 ×1 box. (m) are plotted in the top row. The bottom two rows are low cloud fraction, boundary layer depth (m), and potential temperature at 700 hPa (K) for the ECMWF operational model and CAM and the need for relatively homogeneous clouds to estimate day 25. The values of cloud fraction are more extreme in 3.5, respectively. The white line indicates the stratus buoy position. All data are horizontally averaged MODIS cloud-top height. Plotted below are the low cloud the MODIS data, and the main clearing starts slightly further ◦ ◦ fraction and MBL depth for the ECMWF operational model westward, but the general cloudiness patterns are very well over a 1 ×1 box. and for CAM 3.5. MODIS cloud fraction, dominated by low characterized. The ECMWF model also shows a widespread cloud, shows a dramatic drop in cloudiness to the west of the and persistent boundary layer deepening beginning at day 13, buoy starting at around day 15 and this clearing extending though it has the typical model underestimate of boundary- eastwards of the buoy at day 25 for a period of 2–3 d be- layer depth east of the buoy. The observed negative tem- fore widespread clouds are reestablished. MODIS cloud-top poral correlation between MBL depth and cloud fraction is height retrievals show a broad deepening west of the buoy also seen in the ECWMF operational model. CAM 3.5 also just before the day 15 clearing. At this time the boundary captures the main observed clearing and the boundary layer layer deepens signiﬁcantly at the buoy and to its east. The deepening just before it. It does not match the day 25 clear- extended period with deeper boundary layer appears to be ing near and to the east of the buoy, and has too little cloud associated with the reduction in cloud fraction. in general to the west of the buoy, a problem also seen south of 20 S in the Fig. 4 . The ECMWF operational model succeeds in capturing much of the 20 S cloud change during the comparison pe- The day 13 change to a deeper boundary layer in ECMWF riod, showing both a dramatic reduction in clouds at day 14 and CAM is strongly tied to a ∼5–10 K cooling of the lower to the west of the buoy and the clearing to the east of it at troposphere above the boundary layer, as can be seen in the Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/ M. C. Wyant et al.: The PreVOCA experiment 4769 modeled potential temperature at 700 hPa in Fig. 13 (right- most column). Associated with this cooling is above-MBL cooling to the south and southeast of the buoy (not shown), the primary directions from which the lower troposphere is advected. This cooling promotes stronger entrainment and deepening of the MBL. Some aspects of these changes are captured by most of the models. Figure 15 shows time series of the MBL depth and cloud fraction for all of the models at the stratus buoy, di- urnally averaged, together with MODIS observations (black solid line). (Each point represents a 24-h mean from 00:00– 24:00 UTC). While there is scatter in the mean MBL depth, especially among regional models, all of the models show deepening at around day 13, and a deeper boundary layer tends to persist for a several day period. Interestingly this change is not as clear in the MODIS retrieval. Most op- erational models show some clearing associated with this change, while regional and climate model cloud changes are not consistent. The operational models also are able to better capture the observed clearing event at day 25, an event which does not appear to be connected directly with boundary layer depth changes, but instead is related to strong increases in modeled subsidence (not shown). 6 Discussion and conclusions The PreVOCA model assessment surveyed the ability of a wide range of contemporary atmospheric models to sim- ulate the SEP region near the Chilean coast during Octo- ber 2006. October in the SEP is characterized by extensive marine stratocumulus boundary layers and weak mean low- level subsidence. Operational and climate models performed daily short-term forecasts for the period, while regional mod- els each ran month-long simulations forced continuously by analysis. Overall the models do a good job of simulating Fig. 15. October 2006 time series of 24-h mean boundary layer ◦ ◦ the observed anticyclonic surface winds. They share similar depth in km (a–c) and low cloud fraction (d–f) for a 3 ×3 box ◦ ◦ Fig. 15. October 2006 time series of 24-h mean boundary layer depth in km (a–c) and low cloud fraction mean subsidence patterns, though these are difﬁcult to evalu- centered at 20 S 85 W. Solid black lines are MODIS cloud-top height and MODIS cloud fraction. Model line types as in Fig. 8. ate by observational comparison. M◦ eanwhile, ◦ the cloud and ◦ ◦ (d–f) for a 3 ×3 box centered at 20 S 85 W. Solid black lines are MODIS cloud-top height and boundary layer properties produced by the models are quite diverse especially in cloud fraction, MBL depth, and LWP. MODIS cloud fraction. Model line types as in Fig. 8. The models generally under-predict the amplitude of the Cloud fraction biases are primary contributors to very large diurnal cycle of liquid water path at the stratus buoy, though biases in the downward shortwave ﬂux at the surface. the amplitude relative to the mean LWP agrees fairly well. The models also have widely varying MBL depths. A very The models’ predictions of the diurnal cycle of low-cloud common model problem to the east of the stratus buoy is the fraction are quite varied, with several models predicting under-prediction of the MBL depth, especially near to the a signiﬁcant peak in morning cloud fraction compared with coast. This does not appear to be simply a problem of insuf- more ﬂat observations. Discrepancies in the observed diur- ﬁcient horizontal resolution (e.g., the ECMWF OPER model, nal cycle, especially in the size of the late afternoon/early with relatively ﬁne 25-km horizontal resolution, substantially evening drop in cloud fraction, make the evaluation of the di- underestimates coastal MBL depth at 20 S). This problem urnal model cloud fraction biases difﬁcult. Hopefully these has important implications for modeling of surface ﬂuxes, comparisons can be improved using the more extensive mea- cloud thickness, and cloud fraction, though there is no clear surements available in VOCALS REx. connection between mean MBL depth bias and mean cloud fraction bias among these models. www.atmos-chem-phys.net/10/4757/2010/ Atmos. Chem. Phys., 10, 4757–4774, 2010 4770 M. C. Wyant et al.: The PreVOCA experiment Most models produce a diurnal upsidence wave which array of in-situ aircraft and ship measurements, and will no propagates southwestward with phase velocity similar to that doubt provide further insights to improve modeling of this reported in previous studies. Though the upsidence wave region. produces clear perturbations in MBL height as it passes, its effect on modeled offshore cloud fraction appears to be min- Appendix A imal in these models. Most of the models qualitatively capture the large varia- Here we provide a description of the model physics and ex- tions in MBL height associated with synoptic variability. The periment setup in more detail. Unless otherwise stated, all primary cause of these variations is the changing temperature the models use single moment bulk microphysical schemes. above the boundary layer in the lower troposphere altering Of the models that use aerosols, most use climatological the LTS. The inﬂuence of variations of large-scale subsidence speciﬁed aerosol concentrations which impact the simula- on MBL-height variations appears to be secondary to other tions through radiative effects only. Exceptions will be noted conditions. At the stratus buoy, in observations and in some below. models, deepening of the MBL is associated with reduced COAMPS – The Coupled Ocean/Atmosphere Mesoscale cloudiness. For most models, however, low cloud changes Prediction System of the Naval Research Laboratory (Hodur, do not agree with observed changes or with each other. Two 1997) is run for 24 h periods, twice daily starting at 0:00 Z forecast models in particular, ECMWF and UKMO, show and 12:00 Z with a smaller nested grid covering the study re- skill at cloud prediction, whereas the regional models do not. gion. It continuously assimilates atmospheric and SST data. There is not a clear relationship between model vertical res- It uses the Navy Operational Global Atmospheric Prediction olution and model skill at predicting MBL height or cloud System (NOGAPS) global model to provide lateral boundary properties. conditions. A moist TKE scheme is used in the PBL. A bulk The differences in performance between the operational microphysics scheme based on Rutledge and Hobbs (1984) forecast models and the regional models are large. The oper- is used. ational model forecasts tend to agree well with one another COLA – The submission from the Center for Ocean- in surface winds and MBL depth, more so than the regional Land-Atmosphere Studies uses the Regional Spectral Model models. Some of this difference in performance may be due (RSM) developed at the Experimental Climate Prediction to the short simulation length of the runs made in forecast Center (ECPC) of the Scripps Institution of Oceanogra- mode. For all of the forecast-mode runs (all operational mod- phy described in Kanamaru and Kanamitsu (2007) with els runs plus CAM and GFDL), the whole domain is reinitial- some modiﬁcations, particularly to the treatment of cloud- ized with analysis for each run, reducing errors and drift from water. A month long simulation was performed, continu- analyzed states. In contrast the regional models are initial- ously forced by NCEP/NCAR reanalysis. The model uses ized only once, and the observational analysis that is applied the non-local PBL scheme of Hong and Pan (1996), bulk at the boundaries can take days to inﬂuence the entire study cloud microphysics (Sundqvist et al., 1989), and the prog- region. The operational models also beneﬁt because the anal- nostic cloud water scheme of Zhao and Carr (1997). yses they use are often created by models with identical or ECMWF – Several ECMWF model results were sub- nearly identical physics, so model adjustments to the bound- ary conditions are greatly reduced. Despite these advantages, mitted, three of which we show here. The operational many of the operational forecast models have substantial de- model (ECMWF-OPER) uses ECMWF-IFS CY31R1; the ﬁciencies in predicting cloud properties. ﬁve day forecasts runs (ECMWF-5DAY) are using CY32R3. The simulations in this study do not have the necessary The coupled ensemble forecasts (ECMWF-CPLD) are using horizontal resolution to accurately simulate pockets of open CY32R3 run eyx6. These models have very similar physical cells (POCs), which are a signiﬁcant observed feature of low parameterizations, but the CY32R3 runs include reﬁnements clouds in the region. For regional and global models in the to convection and stratocumulus representation and the in- foreseeable future, parameterizations for POCs will likely be troduction of McICA radiation. The ECMWF ECMWF- necessary to accurately represent cloud cover in this and sim- 5DAY runs were initialized with the ECMWF analysis. The ilar regions. coupled 5-member ensemble runs were initialized on 1 Au- gust 2006 with differing initial perturbations and the output A major focus of VOCALS is the interaction between shown here are ensemble means. The model runs all use aerosols and clouds, and the cloud and boundary-layer mod- a combined eddy diffusivity-mass-ﬂux scheme using moist eling errors demonstrated here pose substantial challenges to conserved variables for the dry and stratocumulus-topped modeling aerosol and gas concentrations and transport, as boundary layer (Kohler ¨ , 2005; Tiedtke, 1993) microphysics. well as aerosol source and sink processes. A follow-on inter- comparison of a similar suite of models during for October– GFDL – The Geophysical Fluid Dynamics Laboratory November 2008 during REx will be performed with a partic- (GFDL) AM2 model (GFDL-GAMDT, 2004) was run with ular focus on aerosol-cloud interactions. This future study, a ﬁnite volume dynamical core on a cubed-sphere grid. Each the VOCALS Assessment or VOCA, will beneﬁt from a large 00:00 Z daily forecast was initialized with ECMWF analysis Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/ M. C. Wyant et al.: The PreVOCA experiment 4771 data (identical to that used for CAM). A Lock et al. (2000) K- NCEP – The National Centers for Environmental Predic- proﬁle boundary layer scheme with calculated entrainment tion Global Forecasting System (GFS) model operational rate was used and the bulk microphysics scheme of Rotstayn runs use the NCEP data initialization system for initial con- (1997) was used. ditions. A non-local surface-forced K-proﬁle scheme is used IPRC – The International Paciﬁc Research Center IPRC- for the PBL. The bulk microphysics scheme of Zhao and Carr RegCM (Wang et al., 2003, 2004) version 1.2 was run con- (1997) used. tinuously throughout the study period with the lateral bound- PNNL – The WRF-Chem model version 2.2 was run in aries forced with NCEP/NCAR reanalysis. The model uses a number of conﬁgurations with both an inner nested domain a prognostic TKE scheme with an additional non-local ﬂux and an outer domain. We present here the runs which in- parameterization and a bulk mixed-phase Lin-type micro- clude a high resolution domain from 10–30 S and from 70– physical scheme (Wang, 2001). An artiﬁcial smoothly- 90 W nested within a coarser outer domain. The output pre- varying cloud-droplet concentration is speciﬁed over the sented here is from the outer domain. The initial and bound- ocean based upon proximity to land. ary conditions are based on GFS analysis. The PBL scheme JMA – The Japan Meteorological Agency model, version is YSU (Hong et al., 2006) and the microphysics used is GSM0711 was run as a series of 4× daily forecasts, each run a Lin scheme (Lin et al., 1983; Chen and Sun, 2002) mod- for 30 h, with the last 6 h analyzed here. Each run was initial- iﬁed to make autoconversion dependent on droplet number ized from JMA operational global analysis. The model uses based on Liu et al. (2005). Three different categories of runs a Mellor-Yamada level-2 PBL scheme and has a bulk mi- are presented, speciﬁed cloud droplet number concentration crophysics scheme based on Sundqvist (1978) and Sundqvist (PNNL-M), prognostic cloud-droplet number concentration et al. (1989). but constant CNN concentration (PNNL-P), and interactive LMDZ – The LMDZ general circulation model from CNN with full chemistry, variable CCN, and speciﬁed emis- the Laboratoire de Meterologie Dynamique (Hourdin et al., sions of aerosols, SO and other gases (PNNL-C, Chapman 2006) has no active microphysics scheme in the runs pre- et al., 2009). sented here. Runs were submitted using both the default UCHILE – The University of Chile runs use the WRF K-proﬁle turbulent scheme and a new boundary Mellor- model (Skamarock et al., 2005), version 2.2 run continuously Yamada type boundary layer scheme with a moist thermal over October 2006 with NCEP/NCAR reanalysis for initial plume scheme. We present here runs with the newer bound- and lateral boundary conditions. It uses a prognostic Mellor- ary layer-scheme only; the other scheme did not produce Yamada-Janjic ´ TKE scheme (Janjic ´, 2002) with a Lin micro- strongly different results. The month was run continuously physical scheme (Lin et al., 1983; Chen and Sun, 2002). with a ﬁne grid over the study region and relaxation to ERA- UCLA – The UCLA runs also use WRF 2.2 initialized and 40 winds outside of the ﬁne grid. No other variables were forced at the lateral boundaries by NCEP/NCAR reanalysis. relaxed towards reanalysis. A ﬁner domain is nested within a coarser outer domain, with NASA GMAO – The Global Modeling and Assimilation 15 km horizontal resolution for the inner domain. Output for Ofﬁce (GMAO) GEOS-5 DAS output comes from a 4× daily the run is presented here from the inner domain. The YSU forecasts with global data assimilation. The model uses a PBL scheme is used with explicitly treatment of entrainment Lock et al. (2000) boundary layer scheme and a Sundqvist- at the PBL top. The WSM bulk microphysics scheme (Hong type bulk microphysics scheme. et al., 2004) is used. NCAR – Two versions of the National Center for Atmo- UKMO – The Met Ofﬁce Uniﬁed Model (MetUM) was spheric Research Community Atmospheric Model (CAM, run in its operational global model cycle G41 conﬁgura- Collins et al., 2004) were used. For each version, three day tion. The dynamics is a non-hydrostatic two-time level semi- simulations were run initialized from daily 00:00 Z ECMWF implicit, semi-Lagrangian formulation (Davies et al., 2005). analysis, with results from the third day presented here. The The boundary layer scheme is a nonlocal surface-forced K- ﬁrst version, CAM 3.5 uses a non-local K-proﬁle bound- proﬁle scheme (Lock et al., 2000; Martin et al., 2000), the ary layer scheme (Holtslag and Boville, 1993) and a single microphysics scheme is that of Wilson and Ballard (1999) moment bulk microphysics scheme (Rasch and Kristjans- and the cloud fraction scheme is that of Smith (1990). For son, 1998). The second version, CAM 3.6 UW is run using more details of the model formulation and recent changes a prognostic TKE scheme (see Bretherton and Park, 2008) see Allan et al. (2007). and the shallow convection scheme of Park and Bretherton Acknowledgements. Thanks to D. Painemal and P. Zuidema for (2009). CAM 3.6 runs use the double-moment bulk micro- providing MODIS retrieved cloud-top heights. Also thanks to physical scheme with prognostic cloud-droplet number con- S. Park who provided his gridded EECRA data. COSMIC data centration of Morrison and Gettelman (2008) which allows was provided by B. Kuo. Many thanks to L. O’Neill at NRL who aerosol concentration to affect cloud droplet activation. Both provided diurnal ﬁts and monthly mean of LWP from satellite. versions have prognostic aerosols and use the MOZART bulk CALIPSO cloud top-height data was provided by D. Wu of aerosol model (Lamarque et al., 2005). the Ocean University of China. Thanks also to Virendra Ghate for providing diurnal cloud fraction data. QuikSCAT data are www.atmos-chem-phys.net/10/4757/2010/ Atmos. Chem. Phys., 10, 4757–4774, 2010 4772 M. C. Wyant et al.: The PreVOCA experiment produced by Remote Sensing Systems and sponsored by the NASA de Szoeke, S., Fairall, C., and Pezoa, S.: Ship observations of the Ocean Vector Winds Science Team. S. deSzoeke’s archive of ship tropical Paciﬁc Ocean along the coast of South America, J. Cli- observations was very helpful to this work. The ISCCP FD data mate, 22, 458–464, 2009. were obtained from the ISCCP web site http://isccp.giss.nasa.gov Field, P. and Wood, R.: Precipitation and cloud structure in midlat- maintained at NASA GISS. S. A. Klein acknowledges M. Zhao itude cyclones, J. Climate, 20, 5208–5210, 2007. (GFDL) for performing GFDL model integrations, J. Boyle Garreaud, R. and Munoz, R.: The diurnal cycle in circulation and (LLNL) for preparing analysis data, and the U. S. Department of cloudiness over the subtropical southeast Paciﬁc: A modeling Energy’s Ofﬁce of Science Climate Change Prediction and Atmo- study, J. Climate, 17, 1699–1710, 2004. spheric Radiation Measurement programs for ﬁnancial support. Garreaud, R., Rutllant, J., Quintana, J., Carrasco, J., and Minnis, P.: The contribution of S. A. Klein to this work is performed under the CIMAR-5: A snapshot of the lower troposphere over the subtrop- auspices of the US Department of Energy by Lawrence Livermore ical southeast Paciﬁc, B. Am. Meteorol. Soc., 82, 2193–2207, National Laboratory under contract DE-AC52-07NA27344. We 2001. acknowledge the support of NASA award No. NX06AB74G GFDL-GAMDT: The new GFDL global atmosphere and land for C. Hannay. This work was also supported by NSF grant model AM2-LM2, J. Climate, 17, 4641–4673, 2004. ATM0745702 and NOAA grant NA070AR4310282. Ghate, V., Albrecht, B., Fairall, C., and Weller, R.: Climatology of surface meterology, surface ﬂuxes, cloud fraction, and radia- Edited by: J. Quaas tive forcing over the south-east Paciﬁc from buoy observations, J. Climate, 22, 5527–5540, 2009. Hahn, C. and Warren, S.: Extended edited synoptic cloud reports References from ships and land stations over the globe, 1952–1996, Tech. rep., Carbon Dioxide Informational Analysis Center, Oak Ridge Allan, R., Slingo, A., Milton, S., and Brooks, M.: Evaluation of National Laboratory, 1999. the Met Ofﬁce global forecast model using geostationary Earth Hannay, C., Williamson, D., Hack, J., Kiehl, J., Olson, J., Klein, S., Radiation Budget (GERB) data, Q. J. Roy. Meteor. Soc., 133, Bretherton, C., and Kohler ¨ , M.: Evaluation of forecasted south- 1993–2010, 2007. east paciﬁc stratocumulus in the NCAR, GFDL, and ECMWF Anthes, R., Bernhardt, P., Chen, Y., Cucurull, L., Dymond, K., Models, J. Climate, 22, 2871–2889, 2009. Ector, D., Healy, S., Ho, S., Hunt, D., Kuo, Y., Liu, H., Man- Hodur, R.: The Naval Research Laboratory’s coupled ning, K., Mccormick, C., Meehan, T., Randel, W., Rocken, ocean/atmosphere mesoscale prediction system (COAMPS), C., Schreiner, W., Sokolovskiy, S., Syndergaard, S., Thomp- Mon. Weather Rev., 125, 1414–1430, 1997. son, D., Trenberth, K., Wee, T., Yen, N., and Zeng, Z.: The Holtslag, A. and Boville, B.: Local versus nonlocal boundary-layer COSMIC/FORMOSAT-3 – Mission early results, B. Am. Me- diffusion in a global climate model, J. Climate, 6, 1825–1842, teorol. Soc., 89, 313–333, 2008. Boyle, J., Klein, S., Zhang, G., Xie, S., and Wei, X.: Climate model Hong, S. and Pan, H.: Nonlocal boundary layer vertical diffusion forecast experiments for TOGA COARE, Mon. Weather Rev., in a Medium-Range Forecast Model, Mon. Weather Rev., 124, 136, 808–832, 2008. 2322–2339, 1996. Bretherton, C. and Park, S.: A new bulk shallow-cumulus model Hong, S., Dudhia, J., and Chen, S.: A revised approach to ice mi- and implications for penetrative entrainment feedback on updraft crophysical processes for the bulk parameterization of clouds and buoyancy, J. Atmos. Sci., 65, 2174–2193, 2008. precipitation, Mon. Weather Rev., 132, 103–120, 2004. Bretherton, C., Uttal, T., Fairall, C., Yuter, S., Weller, R., Baum- Hong, S., Noh, Y., and Dudhia, J.: A new vertical diffusion pack- gardner, D., Comstock, K., Wood, R., and Raga, G.: The EPIC age with an explicit treatment of entrainment processes, Mon. 2001 stratocumulus study, B. Am. Meteorol. Soc., 85, 967–977, Weather Rev., 134, 2318–2341, 2006. Hourdin, F., Musat, I., Bony, S., Braconnot, P., Codron, F., Chapman, E. G., Gustafson Jr., W. I., Easter, R. C., Barnard, J. C., Dufresne, J., Fairhead, L., Filiberti, M., Friedlingstein, P., Grand- Ghan, S. J., Pekour, M. S., and Fast, J. D.: Coupling aerosol- peix, J., Krinner, G., Levan, P., Li, Z., and Lott, F.: The LMDZ4 cloud-radiative processes in the WRF-Chem model: Investigat- general circulation model: climate performance and sensitivity ing the radiative impact of elevated point sources, Atmos. Chem. to parametrized physics with emphasis on tropical convection, Phys., 9, 945–964, doi:10.5194/acp-9-945-2009, 2009. Clim. Dynam., 27, 787–813, 2006. Chen, S. and Sun, W.: A one-dimensional time dependent cloud Janjic, ´ Z.: Nonsingular implementation of the Mellor-Yamada level model, J. Meteorol. Soc. Jpn., 80, 99–118, 2002. 2.5 scheme in the NCEP Meso model., Tech. rep., NCEP, 2002. Colbo, K. and Weller, R.: The variability and heat budget of the Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., upper ocean under the Chile-Peru stratus, J. Mar. Res., 65, 607– Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Y., 637, 2007. Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K., Collins, W., Rasch, P., Boville, B., Hack, J., McCaa, J., Williamson, Ropelewski, C., Wang, J., Leetmaa, A., Reynolds, R., Jenne, R., D., Kiehl, J., Briegleb, B., Bitz, C., Lin, S.-J., Zhang, M., and Joseph, D.: The NCEP/NCAR 40-year reanalysis project, B. and Dai, Y.: Description of the NCAR Community Atmosphere Am. Meteorol. Soc., 77, 437–471, 1996. Model (CAM 3.0), Tech. rep., NCAR, 2004. Kanamaru, H. and Kanamitsu, M.: Scale-selective bias correction Davies, T., Cullen, M., Malcolm, A., Mawson, M., Staniforth, A., in a downscaling of global analysis using a regional model, Mon. White, A., and Wood, N.: A new dynamical core for the Met Weather Rev., 135, 334–350, 2007. Ofﬁce’s global and regional modelling of the atmosphere, Q. J. Klein, S. and Hartmann, D.: The seasonal cycle of low stratiform Roy. Meteor. Soc., 131, 1759–1782, 2005. Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/ M. C. Wyant et al.: The PreVOCA experiment 4773 clouds, J. Climate, 6, 1587–1606, 1993. iform clouds and precipitation in large-scale models. 1. Descrip- Klein, S., Jiang, X., Boyle, J., Malyshev, S., and Xie, S.: Di- tion and evaluation of the microphysical processes, Q. J. Roy. agnosis of the summertime warm and dry bias over the U.S. Meteor. Soc., 123, 1227–1282, 1997. Southern Great Plains in the GFDL climate model using a Rutledge, S. and Hobbs, P.: The mesoscale and microscale struc- weather forecasting approach, Geophys. Res. Lett., 33, L18805, ture and organization of clouds and precipitation in midlatitude doi:10.1029/2006GL027567, 2006. cyclones. XII: A diagnostic modeling study of precipitation de- Kohler ¨ , M.: Improved prediction of boundary layer clouds, velopment in narrow cold-frontal rainbands, J. Atmos. Sci., 41, ECMWF Newsletter, 104, 18–22, 2005. 2949–2972, 1984. Lamarque, J., Kiehl, J., Brasseur, G., Butler, T., Cameron-Smith, Skamarock, W., Klemp, J., Dudhia, J., Gill, D., Barker, D., Wang, P., Collins, W., Collins, W., Granier, C., Hauglustaine, D., W., and Powers, J.: A description of the advanced research WRF Hess, P., Holland, E., Horowitz, L., Lawrence, M., McKenna, version 2, Tech. rep., NCAR, 2005. D., Merilees, P., Prather, M., Rasch, P., Rotman, D., Shin- Siebesma, A. P., Jakob, C., Lenderink, G., Neggers, R. A. J., Teix- dell, D., and Thornton, P.: Assessing future nitrogen deposi- eira, J., Van Meijgaard, E., Calvo, J., Chlond, A., Grenier, H., tion and carbon cycle feedback using a multimodel approach: Jones, C., Kohler, M., Kitagawa, H., Marquet, P., Lock, A. P., Analysis of nitrogen deposition, J. Geophys. Res., 110, D19303, Muller, F., Olmeda, D., and Severijns, C.: Cloud representation doi:10.1029/2005JD005825, 2005. in general-circulation models over the northern Paciﬁc Ocean: A Lin, Y., Farley, R., and Orville, H.: Bulk parameterization of the EUROCS intercomparison study, Q. J. Roy. Meteor. Soc., 130, snow ﬁeld in a cloud model, J. Clim. Appl. Meteorol., 22, 1065– 3245–3267, 2004. 1092, 1983. Smith, R.: A scheme for predicting layer clouds and their water- Liu, X., Penner, J., and Herzog, M.: Global modeling of aerosol dy- content in a general-circulation model, Q. J. Roy. Meteor. Soc., namics: Model description, evaluation, and interactions between 116, 435–460, 1990. sulfate and nonsulfate aerosols, J. Geophys. Res., 110, D18206, Sundqvist, H.: Parameterization scheme for non-convective con- doi:10.1029/2004JD005674, 2005. densation including prediction of cloud water-content, Q. J. Roy. Lock, A., Brown, A., Bush, M., Martin, G., and Smith, R.: A new Meteor. Soc., 104, 677–690, 1978. boundary layer mixing scheme. Part I: Scheme description and Sundqvist, H., Berge, E., and Kristjansson, J.: Condensation single-column model tests, Mon. Weather Rev., 128, 3187–3199, and cloud parameterization studies with a mesoscale numerical 2000. weather prediction model, Mon. Weather Rev., 117, 1641–1657, Ma, C., Mechoso, C., Robertson, A., and Arakawa, A.: Peruvian 1989. stratus clouds and the tropical Paciﬁc circulation: A coupled Tiedtke, M.: Representation of clouds in large-scale models, Mon. ocean-atmosphere GCM study, J. Climate, 9, 1635–1645, 1996. Weather Rev., 121, 3040–3061, 1993. Martin, G., Bush, M., Brown, A., Lock, A., and Smith, R.: A Wang, Y.: An explicit simulation of tropical cyclones with a triply new boundary layer mixing scheme. Part II: tests in climate and nested movable mesh primitive equation model: TCM3. Part I: mesoscale models, Mon. Weather Rev., 128, 3200–3217, 2000. Model description and control experiment, Mon. Weather Rev., Morrison, H. and Gettelman, A.: A new two-moment bulk strati- 129, 1370–1394, 2001. form cloud microphysics scheme in the community atmosphere Wang, Y., Sen, O., and Wang, B.: A highly resolved regional cli- model, version 3 (CAM3). Part I: Description and numerical mate model (IPRC-RegCM) and its simulation of the 1998 severe tests, J. Climate, 21, 3642–3659, 2008. precipitation event over China. Part I: Model description and ver- Park, S. and Bretherton, C.: The University of Washington shallow iﬁcation of simulation, J. Climate, 16, 1721–1738, 2003. convection and moist turbulence schemes and their impact on Wang, Y., Xie, S., Xu, H., and Wang, B.: Regional model simula- climate simulations with the community atmosphere model, J. tions of marine boundary layer clouds over the southeast Paciﬁc Climate, 22, 3449–3469, 2009. off South America. Part I: Control experiment, Mon. Weather Phillips, T., Potter, G., Williamson, D., Cederwall, R., Boyle, J., Rev., 132, 274–296, 2004. Fiorino, M., Hnilo, J., Olson, J., Xie, S., and Yio, J.: Evalu- Williamson, D. and Olson, J.: A comparison of forecast errors in ating parameterizations in general circulation models – climate CAM2 and CAM3 at the ARM Southern Great Plains site, J. simulation meets weather prediction, B. Am. Meteorol. Soc., 85, Climate, 20, 4572–4585, 2007. 1903–1915, 2004. Wolf, J.-O., Maier-Reimer, E., and Legutke, S.: The Hamburg Rasch, P. and Kristjansson, J.: A comparison of the CCM3 model Ocean Primitive Equation Model, Tech. rep., Deutsches Kli- climate using diagnosed and predicted condensate parameteriza- marechenzentrum, 1997. tions, J. Climate, 11, 1587–1614, 1998. Wood, R. and Hartmann, D.: Spatial variability of liquid water path Raymond, D., Esbensen, S., Paulson, C., Gregg, M., Bretherton, C., in marine low cloud: The importance of mesoscale cellular con- Petersen, W., Cifelli, R., Shay, L., Ohlmann, C., and Zuidema, P.: vection, J. Climate, 19, 1748–1764, 2006. EPIC2001 and the coupled ocean-atmosphere system of the trop- Wood, R., Bretherton, C., and Hartmann, D.: Diurnal cycle of liquid ical East Paciﬁc, B. Am. Meteorol. Soc., 85, 1341–1354, 2004. water path over the subtropical and tropical oceans, Geophys. Richter, I. and Mechoso, C.: Orographic inﬂuences on the annual Res. Lett., 29, 2092, doi:10.1029/2002GL015371, 2002. cycle of Namibian stratocumulus clouds, Geophys. Res. Lett., Wood, R., Kohler ¨ , M., Bennartz, R., and O’Dell, C.: The diurnal 31, L24108, doi:10.1029/2004GL020814, 2004. cycle of surface divergence over the global oceans, Q. J. Roy. Richter, I. and Mechoso, C.: Orographic inﬂuences on subtropical Meteor. Soc., 135, 1484–1493, 2009. stratocumulus, J. Atmos. Sci., 63, 2585–2601, 2006. Wu, D., Hu, Y., McCormick, M., Xu, K., Liu, Z., Smith, B., Omar, Rotstayn, L.: A physically based scheme for the treatment of strat- A., and Chang, F.: Deriving marine-boundary-layer lapse rate www.atmos-chem-phys.net/10/4757/2010/ Atmos. Chem. Phys., 10, 4757–4774, 2010 4774 M. C. Wyant et al.: The PreVOCA experiment from collocated CALIPSO, MODIS, and AMSR-E data to study Zhang, Y., Rossow, W., Lacis, A., Oinas, V., and Mishchenko, M.: global low-cloud height statistics, Geosci. Remote Sens. Lett., 5, Calculation of radiative ﬂuxes from the surface to top of atmo- 649–652, 2008. sphere based on ISCCP and other global data sets: Reﬁnements Xu, H., Wang, Y., and Xie, S.: Effects of the Andes on eastern of the radiative transfer model and the input data, J. Geophys. Paciﬁc climate: A regional atmospheric model study, J. Climate, Res., 109, D19105, doi:10.1029/2003JD004457, 2004. 17, 589–602, 2004. Zhao, Q. and Carr, F.: A prognostic cloud scheme for operational Xu, H., Xie, S., and Wang, Y.: Subseasonal variability of the south- NWP models, Mon. Weather Rev., 125, 1931–1953, 1997. east Paciﬁc stratus cloud deck, J. Climate, 18, 131–142, 2005. Zuidema, P., Painemal, D., de Szoeke, S., and Fairall, C.: Stratocu- mulus cloud-top height estimates and their climatic implications, J. Climate, 22, 4652–4666, 2009. Atmos. Chem. Phys., 10, 4757–4774, 2010 www.atmos-chem-phys.net/10/4757/2010/

http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

Atmospheric Chemistry and Physics Unpaywall

http://www.deepdyve.com/lp/unpaywall/the-prevoca-experiment-modeling-the-lower-troposphere-in-the-southeast-l5wmzH0XGy

The PreVOCA experiment: modeling the lower troposphere in the Southeast Pacific

Loading next page...

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher: Unpaywall
ISSN: 1680-7316
DOI: 10.5194/acp-10-4757-2010
Publisher site: See Article on Publisher Site

Abstract

Journal

Atmospheric Chemistry and Physics – Unpaywall

Published: May 26, 2010

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

The PreVOCA experiment: modeling the lower troposphere in the Southeast Pacific

The PreVOCA experiment: modeling the lower troposphere in the Southeast Pacific

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

The PreVOCA experiment: modeling the lower troposphere in the Southeast Pacific

The PreVOCA experiment: modeling the lower troposphere in the Southeast Pacific

References

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies