1. Introduction
Observations from spaceborne lidar provide a novel perspective on clouds. The lidar is able to directly measure the height of multiple cloud layers, provided the upper layers are not optically thick. A variety of studies have examined overlap statistics (Wang and Dessler 2006), distributions of cloud-top and -base heights (Hart et al. 2005; Dessler et al. 2006b), and the occurrence and backscatter properties of optically thin clouds (Dessler et al. 2006a). There are also studies that show the value of these observations for model evaluation. Miller et al. (1999) use a threat score analysis to test how well the model’s cloud fields match observations in time and space. In Palm et al. (2005), a comparison with lidar observations shows an overestimation of high cloud amount in the European Centre for Medium-Range Weather Forecasts (ECMWF) model. Wilkinson et al. (2008) also evaluate the ECMWF model by comparing the zonally averaged frequency of occurrence and amount of clouds. A lidar forward model is used in this case to account for signal attenuation.
These previous model evaluations have examined overall cloudiness of the model, averaging in time and space to obtain meaningful statistics. In this article, a different approach is followed. Using observations from the Geoscience Laser Altimeter System (GLAS), we focus on one cloud type, marine stratocumulus (MSc), and compare the frequency of occurrence, typical cloud fraction, and cloud-top height for this cloud type. Marine stratocumulus is a good cloud-type candidate for several reasons: MSc exist in areas of subsidence with little upper cloudiness to fully attenuate the lidar signal. The MSc clouds themselves have a very strong backscatter signal, which is unlike the weaker signal of near-surface aerosol layers. The large signal-to-noise ratio of the clouds’ backscatter gives greater confidence in the retrieved cloud-top height. A major drawback is the accompanying rapid signal attenuation, usually before the cloud base can be detected. Thus, the cloud-base height is not considered in this study. The large horizontal extent and persistence of MSc clouds means they are frequently observed by the lidar and individual observations are likely to be similar to the time and space average of the observations. Last but not least, MSc clouds play an important role in the earth’s radiative balance and many models, including the ECMWF model, do not generate the right amount of MSc with the observed properties (Bretherton et al. 2004; Duynkerke and Teixeira 2001; Zeng et al. 2004). Thus, the choice of cloud type capitalizes on the strengths of the lidar observing system, manages to avoid some of its limitations and addresses an area of interest in cloud modeling.
By focusing on one cloud type, instead of global cloud occurrence, observed characteristics of the cloud type can be compared to the equivalent clouds in the model. Systematic errors highlight specific deficiencies and point the way for targeted improvement of parameterizations.
Section 2 will briefly introduce the GLAS data products and the examined model versions. In section 3, the methods for identification of MSc samples are explained. Results of the evaluation are discussed in section 4 and section 5 concludes this article.
2. Observations and model
a. Observational data
The lidar observations used here are from the GLAS on board the Ice, Cloud and Land Elevation Satellite (ICESat). The cloud layer product (GLA09) at 5 Hz from laser period 2A (26 September–18 November 2003) is used (Zwally et al. 2003). Local overpasses occur twice daily around 0700 and 1900 local time near the equator. Both day and nighttime retrievals are used. The lidar’s nominal vertical resolution is approximately 76.8 m and the layer product is based on observations from the green channel (532 nm). The pulse frequency is 40 Hz. Eight backscatter profiles are averaged before processing in the 5-Hz products, which corresponds to an along-track resolution of about 1400 m. The averaged profile is searched from the top down for consecutive vertical bins with backscatter exceeding a threshold. Up to 10 cloud-top and -base height pairs are saved in the GLA09 product (Brenner et al. 2003). Figure 1a shows the attenuated backscatter of a daytime GLAS track in the southeast Pacific Ocean (SEP). Photons from sunlight scattered into the beam lead to a noisy background. Nonetheless, the stratocumulus clouds have a very clear signal about an order of magnitude larger than the background noise. The lidar beam is quickly attenuated in most shots. Figure 1b shows the cloud boundaries from the level 2 product. Retrieved cloud tops and bases are connected by a black line, whereas the gray lines indicate an attenuated profile below the apparent cloud base. A profile is considered fully attenuated when no ground return signal can be found.
b. Model
Two model cycles, CY28R3 (operational from 28 September to 18 October 2004) and CY29R1 (operational from 5 April to 28 June 2005), of the ECMWF’s Integrated Forecasting System (IFS) are evaluated. The IFS is initialized at noon every other day from the operational analysis and run at T511L60 resolution (approximately 40-km grid spacing) for 3 days. Three-hourly output from forecast hours 12–57 is remapped onto a 1° × 1° regular latitude–longitude grid using nearest-neighbor sampling.
While none of the forecast cycles are identical to the analysis cycle used to initialize the forecasts, the boundary layer structure in the regions considered stabilizes within the first 12 h of the forecast. In Fig. 2, profiles of specific humidity and potential temperature are shown at various forecast hours. To remove day-to-day variability, profiles from a 10° × 10° area in the southeast Pacific are averaged together, and all 28 three-day forecasts are composited. The humidity profile of the analysis (thick gray curve in Figs. 2a,b) shows that the boundary layer is less well mixed in the analysis than in any of the subsequent forecast hours shown. Plotted as thin lines are forecast hours 12–72 in 12-h increments. Forecasts verifying at 0000 UTC are shown as thin black lines in Figs. 2a,c, those verifying at 1200 UTC in Figs. 2b,d. While there is a difference between day and night time, composite profiles from subsequent forecast days verifying at the same time of day fall almost on top of each other, thus showing no further drift of the boundary layer structure after forecast hour 12. For reference, the one standard deviation range at forecast hour 12 is shown as the light gray bar in the background. There is some evidence that the free atmosphere above the inversion warms and dries with forecast length, but this appears to have little influence on the moisture and temperature profiles below. Figures are only shown for CY29R1 here, but all other cycles spin up equally quickly.
The model cycles differ primarily in their treatment of the clear and cloudy boundary layer: in CY28R3, the boundary layer is parameterized based on a K-diffusion model using dry variables (Beljaars and Viterbo 1998). All boundary layer clouds are generated either directly by the large-scale cloud scheme or indirectly by the shallow convection parameterization, which provides a source term for the cloud scheme (Tiedtke 1989, 1993). In CY29R1, the eddy diffusivity mass flux (EDMF) framework, adapted to the dry boundary layer and stratocumulus, is introduced. It consists of a diffusive and a mass flux component and uses moist-conserved variables. This enables the scheme to represent local mixing as well as nonlocal transport due to large eddies. The formulation in moist-conserved variables also allows mixing through the cloud base in stratocumulus situations. The EDMF parameterization is described in detail in Tompkins et al. (2004). Aspects of the scheme relevant for this study are described in more detail below.
3. Method
The goal of the method described in this section is to identify one value each for cloud fraction and cloud-top height representative of the boundary layer (BL) clouds in each model column. Since marine stratocumulus clouds are the subject of investigation, we assume that only one layer of BL clouds exists in the lowest 2.5 km of the atmosphere. However, multiple cloud layers can exist within the column. The method outlined below may seem unnecessarily complicated, but it guarantees that the cloud-top height value is truly representative of the BL clouds only. A simple average of the lowest detected cloud layer could, for example, average together nonoverlapping BL and midlevel clouds and thus produce an average height that is representative of neither cloud feature. The following process aims to include all detected cloud tops associated with the BL cloud feature in the gridbox area, but exclude any cloud tops in the column that belong to cloud layers above the BL.
The lidar tracks are collocated in space and time with the nearest 3-hourly gridded model data from the IFS. Hence, the observations are never more than 1.5 h removed from the model data. The collocation procedure is loosely based on the method introduced by Miller et al. (1999). In the horizontal, each lidar shot is associated with the grid point whose 1° × 1° area the shot falls into.
To calculate a cloud fraction on vertical levels from the lidar observations, we must decide for each shot whether it is clear or cloudy on a particular model level. In the left column of Fig. 3, this decision process is shown schematically. Figure 3a shows the model grid, dashed lines marking layer interfaces, overlaid on top of the lidar shots. Each black solid line connects a detected lidar cloud-top and -base pair. Since the lidar cloud tops and bases do not necessarily line up with the layer interfaces, the vertical extent of the clouds is rounded up or down. In this example, shot 2 fills layer 2 completely in the vertical, but layer 1 only partially. Since the shot extends more than halfway into layer 1, the shot is considered to be cloudy in this layer, as indicated by the gray shading in Fig. 3b. In contrast, shot 7 also partially extends into layer 1, but less than halfway through the depth of the layer. In this case the shot is considered to be clear in layer 1 (no gray shading). For very thin clouds, such as shots 6 and 14, the layer in which the cloud occupies the most space is considered to be cloudy—layer 2 in both examples. The gray shading in Fig. 3b illustrates the lidar-derived cloud fraction on each of the model levels using this method.
The average cloud-top height (CTH) for all clouds associated with an individual model layer is defined as the top-height average of all shots considered to be cloudy in the layer. As shown in Fig. 3c, only the shots that contribute to the cloud fraction in layer 2, marked by gray shading, contribute to the CTH in that layer, calculated as the average of the tops marked with black dots. The average CTH for layer 2 is indicated in the figure. A corresponding average CTH is calculated for all model layers. Each layer’s cloud fraction and CTH are checked against the critical values of 80% and 2.5 km to determine whether the column contains stratocumulus clouds. If several layers meet both criteria, the layer with the greatest cloud fraction is chosen to be most representative of the BL cloud feature, as the greatest number of lidar shots has contributed to the average CTH.
The process for deriving a corresponding cloud fraction and CTH from the model is illustrated in the right column of Fig. 3. In Fig. 3d, the model’s cloud fraction is shown. The cloud fraction, indicated by the width of the gray bars, is chosen to be similar to the example in the left column. The same overlap assumption as used in the radiation scheme of the IFS (Räisänen et al. 2004) is used here to generate clear or cloudy subcolumns (Fig. 3e). The boundaries, or tops and bases, of these subcolumns will always fall onto the layer interfaces, so no rounding up or down is necessary. The tops of all subcolumns that are cloudy in layer 2 (black dots) contribute to the average CTH of layer 2.
A lidar simulator based on Chiriaco et al. (2006) is used to determine the level of full signal attenuation in the model subcolumns before the steps above are carried out. This additional check ensures that BL clouds obscured by optically thick layers above are excluded from the comparison, and the cloud base is adjusted to the level of full signal attenuation within optically thick clouds. The actual simulated backscatter is not used beyond this step. For the given circumstances, the signal from the model’s BL clouds is always significantly larger than the simulated molecular backscatter such that the information put into the simulator (clear and cloudy subcolumns) corresponds exactly to the output (simulated backscatter above the background threshold). No additional information about the cloud boundaries is gained by using the backscatter instead of the cloud mask.
This method for deriving the cloud-top height and fraction of the boundary layer cloud feature in each model column allows an equitable comparison between observed and modeled clouds. In Fig. 4, an example of cloud fractions for model (Fig. 4a) and observations (Fig. 4b) are shown for the same track as in Fig. 1.
Two stratocumulus areas are investigated here: the SEP (30°S–0°, 70°–150°W) and the northeast Pacific (NEP; 15°–35°N, 110°–160°W). The southern area has substantial stratocumulus cover in all seasons, and will be the main focus of this study. The Californian stratocumulus region has a maximum cloud cover around July (Klein and Hartmann 1993). The laser 2A period in the fall is not ideal to study the northeast Pacific region, but it is included to illustrate the model’s regionally dependent response to the sensitivity tests discussed further in the next section.
The comparison of the lidar’s along-track cloud fraction to the model’s gridbox area cloud fraction presents a source of error. A satisfactory way of quantifying this error from the lidar observations alone could not be established. Astin et al. (2001) propose a statistical method to calculate error estimates for along-track cloud fraction measurements. However, their method fails in the limit of full cloud cover, as it requires a minimum number of clear and cloudy sections along the track, within the region considered. Applied to the case here, with 1° × 1° grid boxes, an average of 90 lidar shots per grid box and many observed cases of 100% cloud fraction, the method proves to be unsatisfactory. Thus, we present the along-track cloud fraction as is. Qualitatively, the results presented in the following section prove to be robust for changes of the critical cloud fraction threshold.
4. Results
a. Southeast Pacific
The total number of ocean grid points within the SEP region sampled during the laser 2A period is 12 440, for model and observations both. Out of these samples, 3657 (29.4%) GLAS samples are classified as MSc. These samples occur most frequently in the eastern part of the ocean basin (Fig. 5a). Figure 6 shows histograms of along-track cloud fraction (Fig. 6a) and layer-average cloud-top height (Fig. 6b) for all GLAS samples classified as MSc. Both cloud fraction (in the case of lidar observations) and cloud-top height (both lidar and model) have preferred values due to a finite number of lidar shots, vertical bins in the lidar profile, and model layers in the IFS. Bins in the cloud fraction histograms are purposely chosen large (5%), in part to average out the preferred values of the distribution, but also to reflect the potentially large uncertainty associated with comparing along-track cloud fraction to area cloud fractions. For the cloud-top height distributions, a small bin size (50 m, solid black curve) preserves more detail. For an easier comparison with observations, a boxcar running mean over three bins is also shown (thick gray curve). The cloud-top height distribution from the lidar observations has a broad maximum between 1250 and 1500 m. This range is in good agreement with observations of the trade inversion height during the East Pacific Investigation of Climate (EPIC) made two years prior during the same season (Bretherton et al. 2004).
CY28R3 produces only 1391 samples (11.2%) fulfilling the MSc criteria, primarily located in the eastern half of the region (Fig. 5b). Most samples fail the cloud fraction criteria in this cycle. Figure 6c shows in particular the lack of samples with cloud fractions above 95%. The BL clouds in this model version can be generated through the large-scale cloud scheme or shallow convection, which acts as a source term to the large-scale clouds. By keeping track of the shallow scheme’s activity and monitoring cloud fraction from time step to time step, we confirm that the BL clouds in this region are generated primarily by the almost constantly active shallow convection parameterization. The parameterization is optimized for shallow cumulus clouds typically having low cloud fraction. It is not an ideal representation of the processes occurring within the well-mixed, stratocumulus-topped marine boundary layer. The cloud-top height distribution of the MSc is much lower than observed (Fig. 6d) with a broad peak around 750 m.
The EDMF boundary layer scheme in CY29R1 is capable of producing stratocumulus clouds as part of the boundary layer when conditions are appropriate. The number of samples classified as MSc is 2303 (18.5%) for this cycle and the frequency of occurrence map of these samples is similar to the GLAS map (Fig. 5c) a significant improvement over the previous cycle. The additional samples tend to have cloud fractions above 95%, as can be seen in Fig. 6e). However, there are still more MSc observed by GLAS. The additional samples are primarily located farther west in the ocean basin. Unfortunately, the cloud-top height distribution shows that the majority of model samples still have significantly lower cloud tops than observed (Fig. 5f), though an improvement over the previous cycle is evident. About half of this discrepancy in cloud-top height between observations and model is consistent with a lower-than-observed trade wind inversion in the IFS. A study by Hannay et al. (2009) comparing CY29R1, as well as several other models, with in situ observations in the stratocumulus areas off the Chilean coast indicate an underestimation of the trade inversion height by roughly 200 m in the IFS, and an even greater underestimation for the other models examined. These low inversions are not specific to the IFS cycle or boundary layer scheme, but a longer-standing problem in the IFS; they also exist in CY28R3.
The introduction of the EDMF parameterization clearly improves the amount and location of MSc clouds in the model. Two sensitivity runs of CY29R1 are used to address the remaining issues of the excessively low cloud-top heights and the confinement of MSc samples to the near-coastal areas.
Since the depth of the convective boundary layer is determined through test parcel ascent in the EDMF, the effect of a more aggressive test parcel on the stratocumulus cloud-top height is examined. Additional motivation derives from the observation that the test parcel in the parameterization frequently fails to reach the lifting condensation level. In the parameterization, the test parcel is given a temperature and moisture excess at the surface, which is diluted through lateral entrainment of environmental air as the parcel is lifted. This entrainment term takes the form of ε = 1/τw + cε/z, based on large eddy simulation studies (Siebesma and Teixeira 2000; Siebesma et al. 2007). Here, τ is a time scale of 500 s, w is the vertical velocity, z is the height above ground, and cε = 0.55 is a constant factor. The vertical velocity w approaches zero at the top and bottom of the modeled eddy (i.e., surface and BL top) such that ε becomes infinite in the limit. The second term assures this limit at the surface, as the vertical velocity can be nonzero in the lowest model layer. In the sensitivity test, the second term in the expression is dropped, such that ε = 1/τw. This leads to a slighly reduced lateral entrainment near the surface and hence a more energetic parcel.
MSc samples become more numerous (2878, or 23.1%) and appear more frequently in areas farther west in the region (Fig. 7a). The more aggressive parcel ascent leads to a shift of the cloud-top height distribution to higher values (Fig. 8a). The peak of the cloud-top height distribution is now in better agreement.
The second sensitivity test consists of relaxing the lower-level stability criterion used to distinguish between stratocumulus and trade cumulus situations in the EDMF parameterization. This critical value limits the areas where stratocumulus clouds can be modeled by the EDMF parameterization to the near-coastal areas with high stability. An empirical relationship between seasonally averaged stratocumulus amount and lower-level atmospheric stability was established by Klein and Hartmann (1993). Higher stratocumulus amount is generally associated with higher low-level stability, with a range of 14–22 K for various regions and seasons (Klein and Hartmann 1993, see their Fig. 13). In the EDMF parameterization, a critical value of θ700hPa − θsfc = 20 K is used. For the sensitivity test, this value is reduced to 16 K. However, the sample frequency of occurrence [i.e., total samples 2479 (19.9%)] and the cloud-top height distribution (Figs. 7b and 8b) in the SEP region show little change relative to CY29R1.
b. Northeast Pacific
MSc samples are less frequently observed by the lidar in the NEP region (Fig. 9), as indicated by the lighter gray shading in Fig. 9a. While the stratocumulus extent peaks in the SEP during fall, it tends to be at a seasonal low in the NEP (Klein and Hartmann 1993). Out of 5668 grid columns searched, GLAS finds samples meeting the MSc criteria in 1295 (22.8%) cases. The IFS with the EDMF parameterization (CY29R1) has a similar frequency-of-occurrence pattern (Fig. 9b), but finds only 823 (14.5%) MSc samples. As in the SEP region, the modeled cloud tops are lower than observed (Figs. 10a,b). In contrast to the NEP region, the number of samples (1018 or 18.0%) and cloud-top height prove to be more sensitive to the relaxation of the lower-level stability criterion in CY29R1-S (Figs. 9c and 10c,d). The peak of the CTH distribution shifts upward by a model level (approximately 200 m). It appears that the relaxed stability criterion allows the EDMF parameterization to generate clouds at additional grid points.
We conclude that generation of MSc clouds in the SEP is limited by the early termination of the parcel ascent. By decreasing lateral entrainment and thus preserving more of the test parcel’s energy and humidity excess, both the CTH distribution and frequency of occurrence improve. In the NEP, the model proves to be more sensitive to the stability criterion. This is consistent with a generally weaker low-level stability found in the NEP region of the model compared to the SEP region during the observational period.
5. Conclusions
A method for model evaluation using spaceborne lidar observations has been presented in this article. We focus on defining a cloud type and comparing the frequency and location of occurrence as well as characteristic properties of the cloud type between model and observations. In the older version of the ECMWF model (CY28R3), this comparison reveals a lack of marine stratocumulus clouds over the southeast Pacific Ocean. The introduction of the EDMF parameterization focused on stratocumulus clouds greatly improves the location and frequency of occurrence of MSc clouds, but the generated clouds are generally too low by approximately 400 m. This is consistent with, but cannot be completely accounted for by, a lower-than-observed trade inversion in this region. Samples are also more closely confined to the near-coastal areas than the lidar observations indicate. The sensitivity of the cloud-top height and near-coastal confinement of MSc samples is tested by modifying two aspects of the EDMF scheme: the strength of parcel dilution through entrainment of environmental air and the decoupling criterion based on lower-atmospheric stability. In the southeast Pacific, a more aggressive test parcel leads to better (higher) cloud tops and a more realistic frequency of occurrence. In the northeast Pacific, where lower-level stability is weaker during this time of year, the model is more sensitive to the relaxed stability criterion, allowing the EDMF to generate clouds more often and leading to improvements in MSc sample frequency. However, it must be noted that cloud fraction and cloud-top height are only two aspects of the cloudy boundary layer, and further evaluation is required to determine whether the thermodynamic structure of the boundary layer also improves with the implemented changes.
The period during which GLAS provided good-quality cloud observations was unfortunately short. The instrument on the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) is very similar though, and comparable cloud layer products are available for several years now. It is believed that the described analysis method is also applicable to CALIPSO data. Evaluation of trade wind cumulus based on CALIPSO observations will be the subject of future work.
Acknowledgments
This work had been supported through a Fellowship for Graduate Students as part of the Center for Earth Atmosphere Studies (CEAS), Contract NNG06GB41G from NASA, and by the National Science Foundation Science and Technology Center for MultiScale Modeling of Atmospheric Processes, managed by Colorado State University under Cooperative Agreement ATM-0425247.
REFERENCES
Astin, I., L. Di Girolamo, and H. van de Poll, 2001: Bayesian confidence intervals for true fractional coverage from finite transect measurements: Implications for cloud studies from space. J. Geophys. Res., 106 , 17303–17310.
Beljaars, A., and P. Viterbo, 1998: Role of the boundary layer in a numerical weather prediction model. Clear and Cloudy Boundary Layers, A. A. M. Holtslag and P. G. Duynkerke, Ed., Royal Netherland Academy of Arts and Sciences, 287–304.
Brenner, A., and Coauthors, 2003: Geoscience Laser Altimeter System (GLAS) algorithm theoretical basis document, version 4.1, 137 pp.
Bretherton, C., and Coauthors, 2004: The EPIC 2001 stratocumulus study. Bull. Amer. Meteor. Soc., 85 , 967–977.
Chiriaco, M., R. Vautard, H. Chepfer, M. Haeffelin, J. Dudhia, Y. Wanherdrick, Y. Morille, and A. Protat, 2006: The ability of MM5 to simulate ice clouds: Systematic comparison between simulated and measured fluxes and lidar/radar profiles at the SIRTA atmospheric observatory. Mon. Wea. Rev., 134 , 897–918.
Dessler, A., S. Palm, W. Hart, and J. Spinhirne, 2006a: Tropopause-level thin cirrus coverage revealed by ICESat/Geoscience Laser Altimeter System. J. Geophys. Res., 111 , D08203. doi:10.1029/2005JD006586.
Dessler, A., S. Palm, and J. Spinhirne, 2006b: Tropical cloud-top height distributions revealed by the Ice, Cloud, and Land Elevation Satellite (ICESat)/Geoscience Laser Altimeter System (GLAS). J. Geophys. Res., 111 , D12215. doi:10.1029/2005JD006705.
Duynkerke, P., and J. Teixeira, 2001: Comparison of the ECMWF reanalysis with FIRE I observations: Diurnal variation of marine stratocumulus. J. Climate, 14 , 1466–1478.
Hannay, C., D. Williamson, J. Hack, J. Kiehl, J. Olson, S. Klein, C. Bretherton, and M. Köhler, 2009: Evaluation of forecasted Southeast Pacific Stratocumulus in the NCAR, GFDL, and ECMWF models. J. Climate, 22 , 2871–2889.
Hart, W., J. Spinhirne, S. Palm, and D. Hlavka, 2005: Height distribution between cloud and aerosol layers from the GLAS spaceborne lidar in the Indian Ocean region. Geophys. Res. Lett., 32 , L22S06. doi:10.1029/2005GL023671.
Klein, S., and D. Hartmann, 1993: The seasonal cycle of low stratiform clouds. J. Climate, 6 , 1587–1606.
Miller, S., G. Stephens, and A. Beljaars, 1999: Validation survey of the ECMWF prognostic cloud scheme using LITE. Geophys. Res. Lett., 26 , 1417–1420.
Palm, S., A. Benedetti, and J. Spinhirne, 2005: Validation of ECMWF global forecast model parameters using GLAS atmospheric channel measurements. Geophys. Res. Lett., 32 , L22S09. doi:10.1029/2005GL023535.
Räisänen, P., H. Barker, M. Khairoutdinov, J. Li, and D. Randall, 2004: Stochastic generation of subgrid-scale cloudy columns for large-scale models. Quart. J. Roy. Meteor. Soc., 130 , 2047–2067.
Siebesma, A., and J. Teixeira, 2000: An advection–diffusion scheme for the convective boundary layer: Description and 1D results. Preprints, 14th Symp. on Boundary Layers and Turbulence, Aspen, CO, Amer. Meteor. Soc., 133–136.
Siebesma, A., P. Soares, and J. Teixeira, 2007: A combined eddy-diffusivity mass-flux approach for the convective boundary layer. J. Atmos. Sci., 64 , 1230–1248.
Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Mon. Wea. Rev., 117 , 1779–1800.
Tiedtke, M., 1993: Representation of clouds in large-scale models. Mon. Wea. Rev., 121 , 3040–3061.
Tompkins, A., and Coauthors, 2004: Moist physical processes in the IFS: Progress and plans. European Centre for Medium-Range Weather Forecasts Tech. Memo. 452, ECMWF, 91 pp.
Wang, L., and A. Dessler, 2006: Instantaneous cloud overlap statistics in the tropical area revealed by ICESat/GLAS data. Geophys. Res. Lett., 33 , L15804. doi:10.1029/2005GL024350.
Wilkinson, J., R. Hogan, A. Illingworth, and A. Benedetti, 2008: Use of a lidar forward model for global comparisons of cloud fraction between the ICESat lidar and the ECMWF model. Mon. Wea. Rev., 136 , 3742–3759.
Zeng, X., M. Brunke, M. Zhou, C. Fairall, N. Bond, and D. Lenschow, 2004: Marine atmospheric boundary layer height over the eastern Pacific: Data analysis and model evaluation. J. Climate, 17 , 4159–4170.
Zwally, H., R. Schutz, S. Palm, W. Hart, S. Hlavka, J. Spinhirne, and E. Welton, 2003: GLAS/ICESat L2 global cloud heights for multilayer clouds V028. National Snow and Ice Data Center. [Available online at http://nsidc.org/data/gla09.html].
Section of daytime GLAS track from 26 Sep 2003 between 10°N and 15°S in the SEP: (a) attenuated backscatter and (b) the equivalent level 2 cloud layer product. Each detected cloud top and base pair are connected by a black line. In most cases the cloud “base” is in fact the level of full signal attenuation. Gray lines mark the attenuated shots below the level of full attenuation.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Section of daytime GLAS track from 26 Sep 2003 between 10°N and 15°S in the SEP: (a) attenuated backscatter and (b) the equivalent level 2 cloud layer product. Each detected cloud top and base pair are connected by a black line. In most cases the cloud “base” is in fact the level of full signal attenuation. Gray lines mark the attenuated shots below the level of full attenuation.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Section of daytime GLAS track from 26 Sep 2003 between 10°N and 15°S in the SEP: (a) attenuated backscatter and (b) the equivalent level 2 cloud layer product. Each detected cloud top and base pair are connected by a black line. In most cases the cloud “base” is in fact the level of full signal attenuation. Gray lines mark the attenuated shots below the level of full attenuation.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Specific humidity and potential temperature profiles from CY29R1 composited over the area 15°–25°S, 80°–90°W in the SEP, and for all 28 forecasts initialized every 48 h starting at 1200 UTC 25 Sep 2003. Shown as a dark gray solid line in each of the panels is the composite profile of the analysis. Plotted as thin black lines are the composite profiles at forecast hours 12–72 in 12-h steps. (a),(c), Only the forecast hours verifying at 0000 UTC are shown, whereas (b),(d) forecast hours verifying at 1200 UTC are plotted. Profiles differ between night and daytime (cf. left to right), but profiles verifying at the same UTC time are almost identical in the boundary layer. That is, the boundary layer structure does not adjust significantly after forecast hour 12. There is evidence of adjustment during the first 12 h of the forecast, as well as of a slight drift above the inversion. For comparison, the one standard deviation range at forecast hour 12 is shown as wide gray bar.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Specific humidity and potential temperature profiles from CY29R1 composited over the area 15°–25°S, 80°–90°W in the SEP, and for all 28 forecasts initialized every 48 h starting at 1200 UTC 25 Sep 2003. Shown as a dark gray solid line in each of the panels is the composite profile of the analysis. Plotted as thin black lines are the composite profiles at forecast hours 12–72 in 12-h steps. (a),(c), Only the forecast hours verifying at 0000 UTC are shown, whereas (b),(d) forecast hours verifying at 1200 UTC are plotted. Profiles differ between night and daytime (cf. left to right), but profiles verifying at the same UTC time are almost identical in the boundary layer. That is, the boundary layer structure does not adjust significantly after forecast hour 12. There is evidence of adjustment during the first 12 h of the forecast, as well as of a slight drift above the inversion. For comparison, the one standard deviation range at forecast hour 12 is shown as wide gray bar.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Specific humidity and potential temperature profiles from CY29R1 composited over the area 15°–25°S, 80°–90°W in the SEP, and for all 28 forecasts initialized every 48 h starting at 1200 UTC 25 Sep 2003. Shown as a dark gray solid line in each of the panels is the composite profile of the analysis. Plotted as thin black lines are the composite profiles at forecast hours 12–72 in 12-h steps. (a),(c), Only the forecast hours verifying at 0000 UTC are shown, whereas (b),(d) forecast hours verifying at 1200 UTC are plotted. Profiles differ between night and daytime (cf. left to right), but profiles verifying at the same UTC time are almost identical in the boundary layer. That is, the boundary layer structure does not adjust significantly after forecast hour 12. There is evidence of adjustment during the first 12 h of the forecast, as well as of a slight drift above the inversion. For comparison, the one standard deviation range at forecast hour 12 is shown as wide gray bar.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Schematic illustrating strategy for determining a cloud-top height and cloud fraction representative of the boundary layer cloud feature contained within one model grid column. The method is applied to (left) the lidar observations and (right) the model.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Schematic illustrating strategy for determining a cloud-top height and cloud fraction representative of the boundary layer cloud feature contained within one model grid column. The method is applied to (left) the lidar observations and (right) the model.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Schematic illustrating strategy for determining a cloud-top height and cloud fraction representative of the boundary layer cloud feature contained within one model grid column. The method is applied to (left) the lidar observations and (right) the model.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
(a) The model cloud fraction along the same section of track as shown in 1. (b) The corresponding lidar-derived cloud fraction.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
(a) The model cloud fraction along the same section of track as shown in 1. (b) The corresponding lidar-derived cloud fraction.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
(a) The model cloud fraction along the same section of track as shown in 1. (b) The corresponding lidar-derived cloud fraction.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Frequency of occurrence of MSc samples over the laser 2A period in the SEP. The results (a) from the lidar observations, (b) from the IFS with K-diffusion boundary layer parameterization (CY28R3), and (c) from the IFS with EDMF parameterization (CY29R1).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Frequency of occurrence of MSc samples over the laser 2A period in the SEP. The results (a) from the lidar observations, (b) from the IFS with K-diffusion boundary layer parameterization (CY28R3), and (c) from the IFS with EDMF parameterization (CY29R1).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Frequency of occurrence of MSc samples over the laser 2A period in the SEP. The results (a) from the lidar observations, (b) from the IFS with K-diffusion boundary layer parameterization (CY28R3), and (c) from the IFS with EDMF parameterization (CY29R1).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
(left) Histograms of cloud fraction for all samples identified as MSc. (right) Distributions of cloud-top height. The results (a),(b) from the lidar observations; (c),(d) from IFS CY28R3; and (e),(f) from IFS CY29R1. The thin black curve shows the cloud-top height distributions for a 50-m bin size. The thick gray curve is a boxcar running mean over three data points. The dashed gray curve in (d),(f) is the smoothed distribution of lidar cloud tops.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
(left) Histograms of cloud fraction for all samples identified as MSc. (right) Distributions of cloud-top height. The results (a),(b) from the lidar observations; (c),(d) from IFS CY28R3; and (e),(f) from IFS CY29R1. The thin black curve shows the cloud-top height distributions for a 50-m bin size. The thick gray curve is a boxcar running mean over three data points. The dashed gray curve in (d),(f) is the smoothed distribution of lidar cloud tops.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
(left) Histograms of cloud fraction for all samples identified as MSc. (right) Distributions of cloud-top height. The results (a),(b) from the lidar observations; (c),(d) from IFS CY28R3; and (e),(f) from IFS CY29R1. The thin black curve shows the cloud-top height distributions for a 50-m bin size. The thick gray curve is a boxcar running mean over three data points. The dashed gray curve in (d),(f) is the smoothed distribution of lidar cloud tops.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
As in Fig. 5, but with results from (a) the sensitivity test with a more aggressive parcel ascent (CY29R1-E) and (b) the sensitivity test with less restrictive stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
As in Fig. 5, but with results from (a) the sensitivity test with a more aggressive parcel ascent (CY29R1-E) and (b) the sensitivity test with less restrictive stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
As in Fig. 5, but with results from (a) the sensitivity test with a more aggressive parcel ascent (CY29R1-E) and (b) the sensitivity test with less restrictive stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
As in Fig. 6, but (a) results from sensitivity test with more aggressive parcel ascent (CY29R1-E) and (b) results from the second sensitivity test with less restrictive stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
As in Fig. 6, but (a) results from sensitivity test with more aggressive parcel ascent (CY29R1-E) and (b) results from the second sensitivity test with less restrictive stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
As in Fig. 6, but (a) results from sensitivity test with more aggressive parcel ascent (CY29R1-E) and (b) results from the second sensitivity test with less restrictive stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Frequency of occurrence of MSc samples in the NEP region. The results (a) from the observations, (b) from the IFS with EDMF parameterization (CY29R1), and (c) from the sensitivity run with relaxed lower-level stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Frequency of occurrence of MSc samples in the NEP region. The results (a) from the observations, (b) from the IFS with EDMF parameterization (CY29R1), and (c) from the sensitivity run with relaxed lower-level stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Frequency of occurrence of MSc samples in the NEP region. The results (a) from the observations, (b) from the IFS with EDMF parameterization (CY29R1), and (c) from the sensitivity run with relaxed lower-level stability criterion (CY29R1-S).
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Cloud-top height distributions in the NEP region for (a) GLAS observations, (b) the IFS with EDMF parameterization, and the IFS with (c) modified parcel entrainment and (d) stability criterion.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Cloud-top height distributions in the NEP region for (a) GLAS observations, (b) the IFS with EDMF parameterization, and the IFS with (c) modified parcel entrainment and (d) stability criterion.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
Cloud-top height distributions in the NEP region for (a) GLAS observations, (b) the IFS with EDMF parameterization, and the IFS with (c) modified parcel entrainment and (d) stability criterion.
Citation: Monthly Weather Review 137, 12; 10.1175/2009MWR2937.1
