1. Introduction
Global atmospheric models remain inadequate to correctly predict boundary layer (BL) cloudiness, a fact that is particularly conspicuous over the eastern borders of ocean basins, frequently covered by large decks of stratocumuli (e.g., Jakob 1999). The cloud cover derived from Advanced Very High Resolution Radiometer (AVHRR) observations and outputs from the Météo- France Numerical Weather Prediction Model (ARPEGE), have been compared in the context of the field experiment SEMAPHORE1 (Mathieu et al. 1999). The comparison showed that high clouds are realistically predicted, whereas low-level clouds are greatly underestimated. From 10 to 17 November 1993, over the North Atlantic region of 19°N to 55°N and 5°W to 41°W, only 11% of BL clouds are correctly predicted by the model, and 75% of the observed low-level clouds are diagnosed as clear sky. In the same region, the ECMWF model was found to suffer from the same deficiency (Bretherton et al. 1995). Martin et al. (2000) also found a similar problem with the United Kingdom Met Office (UKMO) model. The problem we address in this article is the design of a method to evaluate the inefficiencies of global models in representing the cloud-topped boundary layer. Two aspects are involved. First, the process of assimilation of worldwide network data may be inadequate to correctly force the BL subgrid parameterization. Second, BL processes themselves may not be accurately represented in a model.
Weather prediction skill shows that the free troposphere is correctly assimilated from the network. This is not the case for the boundary layer, because the density of measurements cannot resolve the high spatiotemporal variability of the local conditions that determine the state of the boundary layer (Beljaars and Viterbo 1998). In global models, BL processes are represented as a subgrid parameterization, which uses the large-scale profiles and the surface-layer parameters to initiate. Over oceanic surfaces, when SSTs are prescribed, bulk formulas are sufficiently accurate to determine the surface fluxes, whenever extreme conditions are not present. Depending on the water content of each layer, a cloud diagnostic ultimately determines cloudiness. Subsequently, the BL scheme has to represent the vertical transport of energy and water from the surface to the inversion layer, which may then result in cloudiness. Therefore, BL cloudiness may be used for model evaluation, as a signature of the underlying processes involved (i.e., Norris and Weaver 2001; Klein 1997). However, in addition to the difficulty of defining cloud fraction from satellite, another issue can be raised. When a model strongly underestimates cloudiness, the origin of the deficiency is not clear and may be due to the subgrid parameterization or cloudiness as well as the diagnostic. We present an evaluation method independent of the cloud fraction, based on information linked to the boundary layer height (BLH). Minnis et al. (1991) and Betts et al. (1992) already used this parameter to analyze the BL diurnal cycle during the First International Satellite Cloud Climatology Project (ISCCP) Regional Experiment (FIRE) experiment. The authors used the satellite-retrieved cloud-top temperature (CTT) as characterizing the BLH. We however depart from their method in two ways. The first is that they determined a monthly mean diurnal cycle over a small region, whereas we use instantaneous information on a much larger scale. The second is related to their converting CTT to BLH using a statistically based algorithm, whereas we keep the CTT and compare it with an equivalent parameter determined from the model.
Our analysis is performed during an anticyclonic synoptic situation during phase two of the field experiment SEMAPHORE (Mathieu et al. 1999). Two steps are followed: the CTT is used to identify regions for which the data assimilation in ARPEGE realistically forced the BL subgrid scheme. For these selected regions, we evaluate the model and infer the reasons for its underpredicting of BL cloudiness.
As the full method is thoroughly discussed in Mathieu et al. (2003), section 2 only briefly describes how the stratocumulus-topped boundary layer has been characterized from AVHRR observations and how CTT fields have been retrieved. The estimation of a BLH from ARPEGE outputs is described in section 3, along with elements of its validation. Some parameters related to the BLH determination are approximate and we verify in section 3c that this does not significantly impact the spatial variations of the BLH field. The two derived fields are compared to define regions of agreement as presented in section 4. In the last section, 5, the results are discussed. The differences between rejected and selected regions show that the BL characterization is effectively improved in the regions where there is coherence between the two derived fields. The concluding section presents possible reasons for the deficiency of the subgrid parameterization in ARPEGE for predicting BL clouds.
2. ARPEGE reanalysis and satellite retrieval
The current study uses atmospheric analyzes of the SEMAPHORE database. The campaign took place in the North Atlantic during October–November 1993, in a region that is crossed by the Azores oceanic front (Eymard et al. 1996). The atmospheric fields for the experiment were produced by ARPEGE, using version dedicated to the campaign analysis (Giordani and Planton 2000). A stretched coordinate frame is used, where the pole is at the center of the experiment area, using T119 spectral truncation with a stretching factor of 3.5. The resulting size of the grid box is about 30 km near the center. The vertical grid has 24 levels starting at 17 m above surface, and with 12 levels in the lowest 4000 m. Datasets collected between 7 October and 17 November 1993 and operational data from the worldwide network were analyzed by the model. The two types of data are treated separately in ARPEGE analyses; SEMAPHORE observations are assimilated with weights 10 times stronger than the operational data.
During the 10 to 17 November period, meteorological conditions are favorable for BL analyses as the large- scale fields were close to stationary and the region was under the influence of large-scale subsidence. In such a situation, the boundary layer is partially decoupled from the large scale as a strong inversion inhibites coupling with the free troposphere. The low divergence tends also to minimize advection effects from outside regions.
The National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer data were chosen to characterize the boundary layer in terms of the structure of cloud cover, CTT, and sea surface temperature. During the SEMAPHORE experiment, NOAA-11 and -12 data are available. Two images per day are available: one around 0900 and the other around 1700 UTC. The application of a statistical classification method (Sèze and Desbois 1987) provided good identification of low-level clouds and clear areas, as seen in Figs. 1a,b. On the figures, islands and continents are masked out (in white), and cloud-free areas are shown in black. A good dynamic range is obtained on low-level clouds (nine classes), providing a discrimination between homogeneous cloud decks (in green) and broken stratocumulus (in orange).
The cloud classification allows one to identify optically thick stratocumulus (Sc) totally covering the pixels. For these clouds, the CTT is derived from the brightness temperature in channel 4 [centered in the atmospheric window (10.3–11.3 μm)]. Similarly, SSTs are also retrieved following Antoine et al. (1992).
Subsequently, in order to compare satellite data with model output, the spatial resolution of the cloud type, SST, and CTT fields, is matched to the ARPEGE horizontal resolution (Fig. 2a). A CTT error bar is also computed for each grid box (Fig. 2b). For homogeneous Sc decks, the error is within 0.5 K, whereas for other clouds the error ranges from 1 to 2.5 K. This error field is systematically used to weight the point-to-point deviation between model and satellite fields, as described later.
The studied period shows clouds characteristic of the eastern border of an oceanic basin at midlatitude, that is, a region covered with extended decks of Sc. In addition, some days show more-developed cloud structures, revealing the variability of the state of the BL over the ocean. We studied the 13 to 15 November period because extended regions free of high-altitude clouds were present.
3. Synthetic boundary layer height determination from ARPEGE forecast
The inversion height is a key parameter used for BL studies based on observations and also for theoretical studies, similar to the Monin–Obukhov length for the surface layer. It represents a global characteristic of convective BL processes, and we focus our characterization of the boundary layer on this parameter.
a. Method of boundary layer height determination in a GCM and application to the SEMAPHORE analysis
To determine BLH from model output, we adapted a bulk Richardson algorithm, hereafter called thermal algorithm, following Troen and Mahrt (1986) and Holtslag and Boville (1993). The basic idea is to evaluate the altitude reached by a thermal released from the surface layer. The algorithm consists of two steps. First, the thermodynamic properties of the thermal are defined from the state of the surface layer. Second, its equilibrium level is determined.
b. Validation of the thermal algorithm
Because preceding algorithm is adapted to convective boundary layers in the presence of moderate winds, it is appropriate for the selected period of this study.
An example of BLHs diagnosed from an ARPEGE forecast for 1200 UTC 15 November is given in Fig. 3. The map in the figure covers an area approximately 3500 km by 3500 km and the African and Hispanic coasts are reflected in the strong gradients on the right side of the figure; the SEMAPHORE area [Intensive Zone of Observations (IZO)] is symbolized by the black square. Over the region, BLHs are mostly around 1000 m, with maximum around 2500 m. A few grid points have very low values, down to about 30 m, all of these are situated inside the IZO area; we discarded them as explained in section 4. East of the IZO, the general decrease of BLHs from south to north is consistent with the decrease of SST, in particular when approaching the coast of Portugal.
During the period of interest of the SEMAPHORE field experiment, 25 radiosondes were available from ships (every 6 h). Seventeen soundings from airplanes were also recorded, obtained mostly during early afternoons. The measurements are concentrated in the region of the Azores oceanic front. As the model uses an assimilation step of 6 h, outputs in the forecast mode are only valid 3 h later, because the forcing conditions are then sufficiently relaxed, as verified by Giordani et al. (1998). Comparisons between radiosondes and the ARPEGE forecast are done using radiosondes data that have not been assimilated into the model. Experimental profile inversion heights are easily determined from sharp jumps in θ and q profiles (Δθυ ∼ 4–8 K).
Table 1 shows comparison of model thermodynamic values at BLHs with measurements. The bias, a, is derived from the minimization of the mean-square distance running on comparison points, and Q is the residual. Adding 100 m to the BLH suppresses the bias between radiosonde and synthetic BLHs. All following references to the BLH will include this 100 m additional value. From a validation standpoint, the preceding comparison has several drawbacks:
ARPEGE outputs were sometimes delayed in comparison with flight observations.
In the measurement zone, the atmospheric situation was complex and frequently decoupling between the BL and the free atmosphere was observed, enhancing the difficulty of comparison.
The BLH may be very discontinuous on a few ARPEGE grid points. These apparent aberrations in the zone of strong forcing by SEMAPHORE data probably have impact on the whole measurement area.
On another hand, for evaluation of a BLH forecast, the IZO is not the proper region to consider, as it is strongly forced by the assimilation of local measurements in the model. It is also a region influenced by an oceanic front where more complex physics of the boundary layer should be considered (Giordani and Planton 2000). In section 4, the method is applied to a larger scale and region.
c. Sensitivity of the thermal algorithm to arbitrary parameters
In the thermal algorithm, the thermodynamic characteristics of surface-layer fluctuations are key parameters. In addition, some parameters values are chosen arbitrarily, as the critical Ri number and the originating level of the thermal (section 3a). In the following, we analyze the sensitivity of the synthetic BLH to those parameters.
In the algorithm from Holtslag and Boville (1992), the originating level of thermals is taken as the first layer of the model [between 10 and 30 m for National Center for Atmospheric Research (NCAR) model]. A more physically based algorithm should consider surface-layer fluctuations of parcels and determine the level of origination as the level where local-scale convergence results in an updraft. Most of the convective unstable plumes originating from the surface will quickly diffuse in the surface layer. Only the stronger ones will reach a level where surface wind shear can detach them before mixing occurs. The
One important result is that even if absolute BLH values differ when varying Ricr, spatial variations of the model fields taken at the BLH altitude are in general not significantly modified. In the following section, we take advantage of this result to design a selection method for the regions of study.
4. Selection of coherent regions and results
At this stage, we have two independent methods to evaluate the BL height: a synthetic BLH diagnosed from a model and a cloud-top temperature derived from satellite data. The CTT is a signature of the BL height because in our current midlatitude region of study, the strong inversion limits the vertical extension of BL clouds. This has been verified in the campaign IZO by Mathieu et al. (2003). CTT fields are available on a region of about 3500 km by 3500 km. To preserve the most independence between observation and model, the synthetic BLH is converted into the predicted atmospheric temperature at that level (TBLH).
A point-to-point statistical comparison between the two temperature fields was found to give no correlation. This result only tells us that even if coherent signal exists on partial areas, it is undetectable from a background noise. The noise can be filtered out by averaging over long time periods; however, we analyze this “noise” spatially to extract a coherent signal, region by region. Norris and Weaver (2001) promote a similar method. Figure 5 shows examples of consistency between the two temperature fields, along the two transects of Fig. 1. Compared to Fig. 5a, which shows no consistency between synthetic and observed BLH temperatures, the spatial variations in Fig. 5b exhibit good consistency, given the currently available precision limits of GCMs.
How could such consistency be obtained? If we focus on regions outside of the IZO, data assimilation cannot be responsible for it, because the boundary layer is driven by internal air motions which are not resolved by the worldwide network dataset (e.g., Atkinson and Wu Zhang 1996). However for the dynamics and other large-scale fields, the impact of data assimilation is efficient. Hence, we adopt the hypothesis that the regions of consistency between the two BL top temperature fields are the ones for which the boundary layer is forced in a realistic manner by the large scale. This can be justified by the change in dimensionality of the flows; contrary to the large-scale dynamics that are mainly bidimensional, the mesoscale flows in the boundary layer are three-dimensional, and the energy cascades downscale. In other words, the feedbacks of BL processes on the large-scale fields remain moderate in subsidence conditions, so that spatial variation of BLHs mainly indicates varying forcing conditions. Our results of section 3c yield the same: given the variation of the Ricr as a modification of a BL parameterization, the spatial variations of BLH, but also of TBLH which is found to amplify BLH variations, are nearly unaffected.
To emphasize spatial variations over each region, the bias, a, is removed to minimize the residual distance, Q, between the two temperature fields. The calculation follows closely the one of section 3b [Eq. (3)]. Here, O stands for AVHRR-retrieved CTT and σO are the error bars for the temperature fields. The summing index, i, runs on grid points in each subregion. On the satellite side, σO gives both impacts of subgrid fluctuations and the number of participating pixels to the retrieved temperature (cf. Fig. 2b). On the GCM side, the approximate phase is introduced considering the neighboring nine gridpoints variance.
Quality of agreement is characterized first by the residual distance between the two unbiased fields and takes into account the number of grid points participating in the comparison. Hence for each pass of the selection algorithm, a point is eliminated when its weighted squared distance goes over a threshold of 1 K. New bias and residual are computed for the retained grid points, until convergence is attained. The algorithm is applied on approximately 500 km by 500 km regions where CTTs are available. The regions of consistency are selected using (i) the residual distance (threshold of 0.7 K); (ii) the percentage of retained grid points (threshold of 30%); and (iii) a compactness criterion. The compactness criterion is important since it allows the definition of consistency regions to be extended beyond thick cloud regions. In particular, clear sky areas surrounding or separating dense cloud decks are included because a model or a diagnostic has to explain those transitions.
Morning and afternoon observations were compared to ARPEGE forecast for 0600–1000–1200–1800 UTC systematically. Figures 6a,b illustrate results of the selection on 14 and 15 November, respectively. Following all criteria, the area enclosed by (28°N; 35°N) and (10°W; 30°W) in Fig. 6b, is a consistent region as one can verify along the transect in Fig. 5b. However, the transect plotted in Fig. 6a lies in a rejected area (cf. Fig. 5a).
Table 2 gives corresponding values of bias, residual, and percentage of retained grid points, for each image in the analyzed regions. One can see in particular that the very few cases for which CTT could be retrieved in the IZO are systematically rejected from the selection. This is surprising because one could expect that the assimilation of the field-experiment soundings would improve the realism of the state of the boundary layer. We already noted in section 3b the presence of very low values of BLHs in that area. An explanation is given in the next section.
Among the 15 available images, 9 only contain sufficient cloud cover to be used in the selection algorithm (for 10 a.m., and from 13 to 16 November). For those images, from 40% to 60% of cloudy regions are found relevant. This proves that the large-scale forcing fields are not always responsible for the wrong prediction of BL clouds by the model. These results also confirm that it is potentially possible for a GCM to obtain large-scale fields (SST, surface fluxes, and troposphere variables) that correctly force the model BL subgrid parameterization. In the next section, we present examples that demonstrate the interest for using consistent regions in model evaluation.
5. Analyses of the selected regions
In coherent regions, one can assume that the large- scale fields are correctly forcing the BL parameterization. Therefore, these regions represent not only the proper areas to evaluate BL schemes, but also, to appreciate impacts of model development, and more generally get insight into the physics in the boundary layer. Following this perspective, the divergent results obtained on 14 and 15 November for similar situations are examined.
a. Comparison along similar transects on 14 and 15 November
For the two cloud classification results shown in Figs. 1a,b, moderate geostrophic winds were northeasterly to southeasterly. On 14 November, a high was centered at (36°N, 23°W), and on 15 November, it was slightly shifted to (35°N, 26°W). The southeastern decks of BL clouds are quite stable for the period. The two transects A and B plotted in Figs. 1a and 1b, respectively, are now analyzed. Figures 7a and 7b give the BL characterizations from AVHRR and ARPEGE along transects A and B, respectively. The colored squares show the cloud classification result of AVHRR a.m. and p.m. images in terms of cloud types. For both transects, a uniform Sc zone (green color) is evident from west to east, then a zone of broken cloudiness (orange), ending with another Sc deck. In the afternoon, part of the uniform Sc is partially disaggregated, extending the broken cloudiness zone. However A crosses the IZO (21°E; 27°E) and is not part of a consistent region.
Satellite CTTs along A (Fig. 7a) are at a minimum at the western end of the transect (∼6°C) increasing to ∼10°C in the eastern half. Notice that the temperature axis has been reversed to feature more easily BLH variations. ARPEGE TBLH variations are not consistent with CTT variations, at least in the eastern part. The consistency seems to be restored in the western end, outside the IZO. Along B, AVHRR CTTs do not evolve much during the day; the minimum (about 7°C) is east of the Sc zone; CTTs are close to 9°C west of the Sc zone. At 0600 and 1000 UTC, ARPEGE outputs are consistent with AVHRR values, even if the daily evolution of the ARPEGE temperature looks much stronger and the consistency degrades during the day.
One explanation of this unrealistic evolution might be found in the systematic underestimation of low-level cloud cover in the model (Fig. 8), which impacts the diurnal evolution of the boundary layer state. The unrealistic spatial variation observed at 1800 UTC in the western part coincides with the strong underestimation of the model low-level cloudiness in this area.
The synthetic BLHs are mainly negatively correlated with TBLH, varying along B from 800 to 1300 m in the eastern half, and reaching an altitude as low as 600 m in the late afternoon. Coinciding with the BLH variation, we noticed a strong influence of the surface conditions (surface fluxes, temperature and humidity of lowest layers, friction velocity u∗, not shown). Higher BLHs are associated with maximum fluxes, and in general, all spatial variations are strongly dependent on the SST gradient (see Fig. 9b). The only exception is at 1800 UTC, when the surface seems to loose its direct influence on the BLH.
Along A model SSTs (Fig. 9a) have a strong gradient in the east (approximately 2 K per 100 km), which is very different from AVHRR-retrieved SST field. In the thermal algorithm, BLH variation results from the combination of the low-level temperature, u∗, and surface fluxes. We found a minimum of 300 m coinciding with strong local wind variation and a strong dispersion of u∗ in the neighborhood of this minimum (Fig. 10).
We mentioned in the previous analyses that the IZO was systematically rejected from the selection criteria. It is nevertheless difficult to conclude a systematic wrong forecast on the IZO, as very few images give valuable information on that region, because of frequent overcast conditions caused by high clouds. According to Josse (1999), the wrong forecast in the IZO could be due to an assimilation problem of the marisonde buoys. He found that the measured wind speeds were analyzed at 30 m by the model, instead of 5 m. The consequence is an underestimation of low-level winds. We suspect as well a conflict between assimilation of that measurement with the 10-m mast of the ship. This could explain the noisy friction velocity (u∗) found in vicinity of the ship in Fig. 10 ∼(35°N, 26°W).
In conclusion of the comparison between two transects, the method of selecting for consistent regions is effective: transect B shows more consistency between model and satellite-retrieved fields, and the campaign IZO is rejected from the selection, in agreement with assimilation problems. Numerous other transects have been examined that give the same conclusion.
b. Use of selected regions to evaluate BL schemes
The value of this index is given along A and B in Figs. 11a and 11b, respectively. The eastern part of B shows the highest values coincident with the more homogeneous a.m. and p.m. Sc sheets. Going west, the index sharply decreases at the transition to broken clouds, and fluctuates slightly along the disintegrating Sc field. The index of liquid water content seems able to explain observed cloud transitions on a coherent transect but not at all along A. Indeed, except for the eastern part [i.e., far away from the IZO (21°–27°E)], Fig. 11a does not show any possible sensitivity to cloud types.
The conclusion of the exercise is that coherent regions are privileged areas where a BL scheme can be tested against the clouds' characteristics as retrieved from satellites. The scale of comparison is adequate for global model evaluation. Moreover, going to spatial variation on a synoptic scale allows one to evaluate the sensitivity of a diagnosed parameter to local conditions.
6. Conclusions
We have presented a method to evaluate BL parameterization in a global model. The method is designed as to identify regions where the large-scale assimilated fields are correctly forcing the BL subgrid schemes. It is based on a global characterization of the boundary layer by its inversion height. The method has been so far developed for midlatitude characteristics, that is, in regions of large-scale subsidence and dense Sc sheets. Two independent evaluations of the BL top height have been obtained from model outputs and satellite observations: (i) cloud type fields, CTT, and SST maps with associated error bars as retrieved from AVHRR dataset; and (ii) an adaptation of the BLH diagnostic developed by Troen and Mahrt (1986) applied to ARPEGE output fields. Our adaptation of the method leads one to explicitly take into account both latent and sensible surface fluxes to evaluate the virtual potential temperature excess of a standard thermal. The results have been validated against SEMAPHORE measurements. In particular, the synthetic BLH is found to be convincingly realistic.
Despite the strong underestimation of occurrence of BL clouds in ARPEGE, large portions of the analyzed region show a good correspondence between the model and satellite CTT fields. Among the 15 analyzed images, 9 were found with sufficient occurrence of thick BL clouds for which CTT can be retrieved. Consistent regions are found to occur with sufficient frequency to perform useful analyzes (40% to 60% of the retrieved CTT areas). Robustness of the selection method has been verified against the arbitrary choice of some of the thermal algorithm parameters. The selection was successful in rejecting the area of the SEMAPHORE IZO, an area over which an assimilation error had occurred. This is consistent with the very purpose of the method, which is to reject regions where unexplained problems impede pertinent analysis of the physics in the boundary layer. Another application can consist in selecting regions where a mesoscale or large eddy simulation (LES) model is used to resolve the mesoscale flows, because such a model needs to be forced by realistic boundary conditions.
In consistent regions, it can be hypothesized that the analyzed large-scale fields are realistically forcing the boundary layer. Following this hypothesis, the method can be used not only to evaluate an existing BL scheme, but also to evaluate the potential for a model to parameterize the physical processes of the boundary layer responsible for the formation of low-level clouds. We tested this idea with the derivation of a synthetic liquid water content index. It was found that this index indicates more realistic cloudiness and cloud transitions— as observed from satellite—than the cloud diagnostic presently in the model.
The main advantage of the method resides in defining a proper spatial scale to evaluate global models—even if restricted to cloudy regions and their immediate periphery. The method may be looked at as an intermediate between intensive—but short range—field experiments, and large-scale statistical characterization of the boundary layer from satellites. We believe it is offering an alternative to model evaluation, within and possibly without an experimental context.
One may even go further with such a method, introducing the concept of observability—as defined in control theory (e.g., Luenberger 1975). To illustrate the concept, consider that large-scale meteorological fields are recovered from the assimilation of the global terrestrial network. This precisely means that the large- scale atmosphere is observable from the network data (i.e., the state of the atmosphere can be retrieved after a finite number of sequential measurements). Now, focusing on observability of the boundary layer, its spatial variability remains unmeasured by the network. What we have demonstrated here is that some observability of the boundary layer can be extracted from satellites in carefully selected regions.
Acknowledgments
The SEMAPHORE experiment has been supported by CNRS, Météo-France, SHOM, DRET, IFREMER, and ESA, and was initiated by the program PATOM. The authors thank Dr. H. Giordani for providing ARPEGE reanalyses and are grateful to Dr. P. Siebesma and Dr. Van Meijgaard for helpful discussions, and to Dr. M. Haeffelin for numerous corrections to the manuscript.
REFERENCES
Antoine, J. Y., M. Derrien, L. Harang, P. L. Borgne, H. L. Glau, and C. L. Goas, 1992: Errors at large satellite zenith angles on AVHRR derived sea surface temperatures. Int. J. Remote Sens, 13 , 1797–1804.
Atkinson, B. W., and J. Wu Zhang, 1996: Mesoscale shallow convection in the atmosphere. Rev. Geophys, 34 , 403–431.
Beljaars, A. C. M., 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, Eds., Royal Netherlands Academy of Arts and Sciences, 287–304.
Betts, A. K., P. Minnis, W. Ridgway, and D. F. Young, 1992: Integration of satellite and surface data using a radiative–convective oceanic boundary-layer model. J. Appl. Meteor, 31 , 340–350.
Bretherton, C. S., E. Klinker, A. K. Betts, and J. A. Coakley, 1995: Comparison of ceilometer, satellite, and synoptic measurements of boundary-layer cloudiness and the ECMWF diagnostic cloud parameterization scheme during ASTEX. J. Atmos. Sci, 52 , 2736–2751.
Coulman, C. E., 1971: Correlation between velocity, temperature and humidity fluctuations in the air above land and ocean. Bound.- Layer Meteor, 19 , 403–420.
Eymard, L., and Coauthors, 1996: Study of the air–sea interactions at the mesoscale: The SEMAPHORE experiment. Ann. Geophys, 14 , 986–1015.
Giordani, H., and S. Planton, 2000: Modeling and analysis of ageostrophic circulation over the Azores oceanic front during the SEMAPHORE experiment. Mon. Wea. Rev, 128 , 2270–2287.
Giordani, H., S. Planton, B. Benech, and B. Kwon, 1998: Atmospheric boundary layer response to sea surface temperature during the SEMAPHORE experiment. J. Geophys. Res, 103 , 25047–25060.
Holtslag, A. A. M., and B. A. Boville, 1993: Local versus nonlocal boundary-layer diffusion in a global climate model. J. Climate, 6 , 1825–1842.
Jakob, C., 1999: Clouds in the ECMWF reanalysis. J. Climate, 12 , 947–959.
Josse, P., 1999: Modélisation couplée océan–atmosphère à meso- échelle application à la campagne SEMAPHORE. Ph.D. thesis, Université Paul Sabatier, 212 pp.
Klein, S. A., 1997: Synoptic variability of low-cloud properties and meteorological parameters in the subtropical trade wind boundary layer. J. Climate, 10 , 2018–2039.
Lambert, D., 1997: Structure moyenne et turbulente de la couche limite atmosphérique au dessus de l'océan. Ph.D. thesis, Université Paul Sabatier, 212 pp.
Louis, J. F., 1979: A parametric model of vertical eddy fluxes in the atmosphere. Bound.-Layer Meteor, 17 , 187–202.
Luenberger, D., 1975: Introduction to Dynamic Sytems. John Wiley and Sons, 446 pp.
Martin, G. M., M. R. Bush, A. R. Brown, A. P. Lock, and R. N. B. Smith, 2000: A new boundary layer mixing scheme. Part II: Tests in climate and mesoscale models. Mon. Wea. Rev, 128 , 3200–3217.
Mathieu, A., G. Sèze, C. Guerin, H. Dupuis, and A. Weill, 1999: Mesoscale boundary layer clouds structures as observed during the SEMAPHORE campaign. Phys. Chem. Earth, 8B , 933–938.
Mathieu, A., G. Sèze, A. Lahellec, C. Guerin, and A. Weill, 2003: Characterization of the cloud-topped boundary layer at the synoptic scale using AVHRR observations during the SEMAPHORE experiment. J. Appl. Meteor, 42 , 1720–1730.
Minnis, P., P. W. Heck, D. F. Young, C. W. Fairall, and J. B. Snider, 1991: Stratocumulus cloud properties derived from simultaneous satellite and island-based instrumentation during FIRE. J. Appl. Meteor, 31 , 317–339.
Norris, J. A. M., and C. A. Weaver, 2001: Improved techniques for evaluating GCM cloudiness applied to the NCAR CCM3. J. Climate, 14 , 2540–2550.
Sèze, G., and M. Desbois, 1987: Cloud cover analysis from satellite imagery using spatial and temporal characteristics of data. J. Climate Appl. Meteor, 26 , 123–150.
Stull, R. B., 1997: Parameterization of subgrid-scale tracer transport. CAS/JSC Working Group on Numerical Experimentation Rep. 26, Virginia Beach, VA, 148 pp.
Troen, I., and L. Mahrt, 1986: A simple model of the atmospheric boundary layer: Sensitivity to surface evaporation. Bound.-Layer Meteor, 37 , 129–148.
Vogelezang, D. H. P., and A. A. M. Holtslag, 1996: Evaluation and model impacts of alternative boundary-layer height formulations. Bound.-Layer Meteor, 81 , 245–269.
APPENDIX
Evaluating Fluctuations of θυ
The two altitude-dependent scaled variances b1 and b2 were determined from airborne measurements down to 40 m during SEMAPHORE (Lambert 1997). Classical extrapolation formulas give an evaluation from 40 to 10 m of about 70–100 for
As compared with the original algorithm, the current formulation might be preferable for GCM implementation. The first reason is its ability to be scaled with standard parameters of the surface layer, θ∗ and q∗; the second for its giving more flexibility to be adapted to inhomogeneous terrains over land—to take into account specific evaporation laws derived from vegetation models, for instance.
Result of the cloud classification (a) 0900 UTC 14 Nov with white transect (A) from (37°N, 31°W) to (34°N, 14.8°W); (b) 0847 UTC 15 Nov with transect (B) from (31°N, 20°W) to (32°N, 12°W).
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
(a) Boundary layer top temperature retrieved 0847 UTC 15 Nov using AVHRR observations; (b) corresponding error bars.
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
BLHs diagnosed for 1200 UTC 15 Nov
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
Correlation between observed and synthetic BLHs between 10 and 17 Nov; error bars indicate the model's nine neighboring grid points rms errors
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
The BL top temperature retrieved from AVHRR compared with 1200 UTC ARPEGE output: (a) on 14 Nov along the transect (37°N, 31°W)–(34.8°N, 14°W), (AVHRR at 0900 UTC); (b) on 15 Nov along the transect (31°N, 28°W) and (32°N, 12°W) (AVHRR at 0847 UTC)
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
Regions of consistency between AVHRR observations and ARPEGE output: (a) 14 Nov from ARPEGE at 1200 UTC and AVHRR at 0900 UTC; (b) 15 Nov from ARPEGE at 0600 UTC and AVHRR at 0849 UTC
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
The BL top temperatures retrieved along transects, using AVHRR data and ARPEGE output. The three-grid-point-wide colored squares indicate cloud types (from AVHRR). Model results are given at 0600, 1000, 1200, and 1800 UTC; AVHRR at approximately 0900 and 1730 UTC; (a) along transect A on 14 Nov; (b) along transect B on 15 Nov
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
Low-level cloudiness from ARPEGE along transect B
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
ARPEGE SSTs at 1200 UTC compared with AVHRR- retrieved SSTs; (a) along transect A; (b) along transect B
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
Friction velocity u∗ from ARPEGE output along transect A (three meshes wide).
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
Liquid water content index of the boundary layer diagnosed from the ARPEGE output. Model results are given at 0600, 1000, 1200, and 1800 UTC. (a) Along transect A; (b) along B
Citation: Monthly Weather Review 132, 4; 10.1175/1520-0493(2004)132<0915:EOANWF>2.0.CO;2
Bias (model − observation) and residue (Q) evaluated for thermodynamic characteristics at inversions from measured pro files and at BLHs from ARPEGE outputs
Morning mins (a.m.) and afternoon (p.m.) observations are compared to ARPEGE forecast for 0600–1000–1200–1800 UTC systematically. The agreement is estimated using: the percentage of retained points (3rd column), the residual quadratic distances (4th column), the biases (5th column) and compactness criteria (7th and 8th columns)
Structure des Echanges Mer-Atmosphère, Propriétés des Hétérogénéités Océaniques: Recherche Expérimentale.
