1. Introduction
Decks of stratiform low cloud cover large parts of the subtropical maritime regions, often located off the western coasts of subtropical continental masses. They have a strong radiative cooling effect because of their persistent overcast of large areas at low top heights (Stephens and Greenwald 1991; Hartmann et al. 1992). Marine stratocumulus clouds are considered significant contributors to cloud–climate feedbacks (Bony and Dufresne 2005; Soden and Vecchi 2011) and a large source of uncertainty in climate model simulations (Bony et al. 2006; Wyant et al. 2006; Dufresne and Bony 2008). The radiative importance of marine stratocumulus clouds led to their study through various field campaigns, especially in the last two decades. Most of these campaigns took place in the “ideal” stratocumulus regions of the Pacific Ocean—southeast (Bretherton et al. 2004; Mechoso et al. 2014) or Californian (Stevens et al. 2003)—focusing on various aspects of these long-lasting stratocumulus decks (e.g., their diurnal cycle, their vertical structure and entrainment rate, and the characteristics and life cycle of cells inside). This approach minimizes outside perturbations (e.g., influence of midlatitude storm systems), which are an important factor to consider in other regions where such stratiform cover happens more intermittently as a result of such influences. In fact, stratiform low clouds often form behind midlatitude baroclinic systems, moving along with them in their cold-air wake (Lau and Crane 1997; Norris et al. 1998; Norris and Klein 2000; Field and Wood 2007; Mechem et al. 2010). Consequently, stratiform low clouds have a strong presence in the equatorward sectors of the major midlatitude storm tracks (Tselioudis et al. 2013).
The most recent campaign, the clouds, aerosol, and precipitation in the marine boundary layer (CAP-MBL) field campaign (Wood et al. 2015), occurred over a period of 19 months in the Azores, a subtropical–midlatitude transition region in the middle of the Atlantic Ocean. The stratocumulus deck over the Azores has a more transient nature as the region sits near the northern edge of the Canarian stratocumulus-prone area, and baroclinic systems often influence the region as their main track lies just to its north (Hoskins and Hodges 2002). Using ship-based cloud observations, Klein and Hartmann (1993) determined that the yearly average stratus coverage (including stratocumulus among other low clouds) over the northern Atlantic maximizes north of the Azores region and extends south toward the Canary Islands. More recently, results from the CAP-MBL field campaign show that the selected site experiences boundary layer clouds nearly 50% of the time, all year round, from a variety of sources (Rémillard et al. 2012). This makes it an ideal location to examine the full scale of atmospheric conditions and processes responsible for the formation and support of marine boundary layer clouds, including stratocumulus and shallow cumulus regimes. Preliminary campaign results were interesting enough to justify the installation of a new Atmospheric Radiation Measurement Program (ARM) permanent site in the region (http://www.arm.gov/sites/ena/). The long-term observations of the CAP-MBL campaign and the new data stream from the permanent ARM site provide the opportunity to study cloud changes over the Azores at a variety of time scales and to examine the atmospheric processes responsible for the observed cloud regime variability.
Cloud regime analysis is a valuable tool allowing in situ and field campaign observations from a particular region to be put in the context of the global cloud field structures. This type of analysis also makes it possible to examine cloud interactions with atmospheric processes at a wide range of time and space scales. In the recent past, a number of different methods have been devised for and applied to the study of cloud regimes and the atmospheric processes that produce them. Cloud regime separation methods can be constructed either through the application of data analysis techniques directly on properties of the cloud field itself or through the derivation of regimes from analysis of dynamic and/or thermodynamic atmospheric parameters and the subsequent compositing of the corresponding cloud properties for each regime. An example of the first method is the application of clustering techniques on satellite-derived cloud-top pressure–cloud optical depth distributions to derive cloud-based weather regimes (Jakob and Tselioudis 2003; Rossow et al. 2005; Tselioudis et al. 2013), an approach that has been extensively used to study cloud processes and evaluate model cloud simulations (e.g., Williams and Tselioudis 2007; Jakob and Schumacher 2008). An example of the second method, specifically focused on midlatitude baroclinic storm processes, is the derivation of midlatitude storm areas of influence using an analysis of the sea level pressure field from reanalysis data to study storm effects on cloud field variability (Rudeva and Gulev 2011; Bauer et al. 2015, manuscript submitted to J. Appl. Meteor. Climatol.).
In the present work, the global satellite-based cloud regimes derived in Tselioudis et al. (2013) will be used in combination with the ground observations of the CAP-MBL campaign to put the in situ observations at the Azores site into the context of the global cloud field and to provide the vertical structure detail lacking in passive satellite retrievals. Furthermore, the storm area of influence technique of Bauer et al. (2015, manuscript submitted to J. Appl. Meteor. Climatol.) will be applied to study dynamic influences on the cloud field structure and properties. The objectives are to map cloud regime variability over the Azores in relation to the global cloud variability and to the variability of atmospheric dynamics and to create a prototype for the use of regional observations in global model cloud evaluation. The analysis methods are described in section 2 of the paper. In section 3, the clustering method is used to map cloud regimes in the Azores site and relate them to other stratocumulus regions as well as to the global cloud field. Subsequently, the dynamic influences on those cloud regimes and their properties over the Azores are examined using various storm indicators. Data from the CAP-MBL field campaign are analyzed in order to further characterize the vertical structure of the cloud field and provide the vertical cloud distribution of the satellite-derived cloud regimes. The clustering analysis is finally applied to climate model output to reveal deficiencies in model cloud regime simulations and to examine the relevance of the Azores observations in global model cloud evaluation.
2. Datasets and methods
a. Cloud regime variability—The weather states method
The International Satellite Cloud Climatology Project (ISCCP) provides a cloud-top pressure–cloud optical thickness (PC-TAU) histogram for each 2.5° grid cell on a global scale every 3 h during daytime periods since 1983. This type of histogram contains the frequency of occurrence of clouds in each of 42 cloud-top pressure and cloud optical depth categories. Jakob and Tselioudis (2003) developed a technique to analyze the PC-TAU histograms, looking for distinctive patterns in them. A k-means clustering algorithm (Anderberg 1973) is applied to the 3-hourly histograms for each 2.5° region, including completely clear regions. This allows the identification of patterns describing cloud regime variability.
Applying this technique to the globe as a whole, the optimum number of cloud cluster centroids is determined to be 11 [dataset produced by Tselioudis (2012); see Tselioudis et al. (2013) for more details; the PC-TAU histograms for the centroids are reproduced in Fig. 1 for reference]. Each 3-hourly PC-TAU histogram from each grid cell is assigned to one of these clusters, unless the grid is fully clear, in which case it is assigned to cluster 12. The clusters are referred to as “weather states” (WSs) since it has been shown that the cloud property patterns thus detected are linked to distinct states of the atmosphere (Jakob and Tselioudis 2003; Jakob and Schumacher 2008; Tselioudis et al. 2013). In combination with their spatial distribution (shown in Fig. 3 of Tselioudis et al. 2013), the global WSs are assigned cloud-type names as summarized in Table 1. It should be noted that, as determined by Tselioudis et al. (2013), the WSs are arranged starting from the most convective ones (WS1–2) and moving toward the ones where subsidence dominates (WS10–11). Weakly forced and fair-weather situations (WS6–7) are located in the middle of the list. It is important to realize that the associations in Table 1 represent overall occurrences, and deviations are possible. For instance, WS4 is primarily observed as a polar cloud cluster but can also be found less frequently in subtropical latitudes (e.g., just off the coast of Peru and Chile). Similarly, WS1 represents the deepest and thickest convective regime typically found in the tropics, but the ISCCP cluster analysis places a fair share of WS1 along the track of synoptic systems, mainly equatorward from the main locations of the midlatitude storm WS2.
PC-TAU histograms of the 11 clusters and clear sky (adapted from Tselioudis et al. 2013). The cluster number is indicated at the top of each panel, along with its relative frequency of occurrence.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
PC-TAU histograms of the 11 clusters and clear sky (adapted from Tselioudis et al. 2013). The cluster number is indicated at the top of each panel, along with its relative frequency of occurrence.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
PC-TAU histograms of the 11 clusters and clear sky (adapted from Tselioudis et al. 2013). The cluster number is indicated at the top of each panel, along with its relative frequency of occurrence.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Association between clusters and atmospheric conditions.
b. Dynamic variability—The MCMS method
In the Azores region, located at the southern fringe of the Atlantic storm track (Hoskins and Hodges 2002), it is important to characterize the state of the atmosphere by determining whether the area is under the influence of a baroclinic storm. Here, we determine storm influence using NASA’s Earth science program for modeling, analysis, and prediction (MAP) climatology of midlatitude storminess dataset (MCMS; Bauer et al. 2015, manuscript submitted to J. Appl. Meteor. Climatol.). In this product, storms are located through the application of a tracking technique that identifies and tracks sea level pressure minima obtained from reanalysis data on a 6-hourly basis. The MCMS then defines an area of influence for each synoptic system as the area enclosed by the farthest closed isoline of pressure around the low-pressure center. For every time step, the location of each storm’s center is recorded in the dataset, as well as the list of grid cells influenced by that particular storm center. This MCMS dataset is available for every year since 1979 (Bauer et al. 2011).
In this work, we consider that the ground site is influenced by a storm when the grid cell containing Graciosa Island is flagged by the MCMS, while the Azores region (as defined in Table 2) is considered under a storm influence when any one of its cells is flagged by the MCMS. On the other hand, the conditions are considered nonstormy when the whole Azores region is outside storms’ influences, for both the ground site and Azores region. Moreover, any isolated events (i.e., occurring for a single time step) are excluded from the analysis. These criteria are applied to obtain the best separation between stormy and nonstormy conditions. Throughout this work, it is assumed that the MCMS-defined storms passing near the Azores region are midlatitude synoptic systems. Although tropical depressions have been known to affect the region on occasion, these occurrences are assumed to be rare enough to omit distinguishing them.
Location of various regions prone to experience stratocumulus clouds.
c. Ground-based observations—The CAP-MBL campaign
The Azores is a region composed of nine islands scattered in the middle of the Atlantic Ocean, about 1500 km west of Lisbon, Portugal. It straddles the boundary between the subtropical Azores high pressure system and the midlatitude storm track (see Table 2), thus rendering the region prone to the influence of those two types of atmospheric regime. This dual influence was a main factor behind the choice of this location for the CAP-MBL field campaign (Wood et al. 2015), allowing the investigation of the transition from one regime to the other. From June 2009 to December 2010, a broad variety of instruments were deployed on Graciosa Island, near its north shore to minimize the effects of the other islands. The campaign overlaps with the other datasets presented above, although only for seven months with the ISCCP weather states dataset.
The core of the instrumentation consists of the mobile facility provided by ARM (AMF). This facility includes two active remote sensors of interest here: a W-band cloud radar and a Vaisala ceilometer. They both continuously profile the atmosphere above them, with the radar providing a curtain view of the clouds and their precipitation (with time and vertical dimensions) and the ceilometer detecting the location of the first liquid cloud base with good accuracy. Note that heavy precipitation can affect measurements of both sensors, lowering the retrieved cloud-top and cloud-base heights. Sounding balloons were also launched every 6 h for the duration of the campaign, providing profiles of the thermodynamics structure of the atmosphere under various conditions.
3. Results
a. Climatology of weather states over the Azores
The ISCCP-based clustering results of Tselioudis et al. (2013) are analyzed to extract the distribution of the 12 WSs over the Azores. Figure 2 illustrates the relative occurrences of the WS over the Azores as well as over the globe as a whole (solid line) and over other regions prone to experience stratocumulus clouds (see Table 2 for their location). On the global scale, fair weather (WS7) is, by far, the most frequent state, occurring almost a third of the time in the whole dataset. On the other hand, fully clear skies (WS12) are rather rare, owing to the large size of a grid cell (2.5°). Other relatively frequent states are the mostly storm-related midlevel clouds (WS5) and the shallow cumulus fields (WS8). In the Azores region, every major state is readily observed, from deep convective clouds (WS1–2) to stratiform clouds (WS9–11) and fairer skies (WS7–8). Their frequencies of observation are relatively close to the global-mean ones, except for WS7 and WS8. These latter two states are more equally observed in the Azores, at a frequency halfway between their respective global frequencies. It is important to remember here that each cloud regime includes a range of cloud optical thickness and cloud-top pressure values (Fig. 1). Therefore, clouds of a certain regime over the Azores may have small TAU and PC differences from clouds of the same regime in other regions. For instance, Ghate et al. (2015) recently highlighted differences in the stratocumulus-topped boundary layers of the Azores and the southeastern Pacific. In addition, the deep convective regime (WS1) over the Azores represents mostly convection embedded in frontal systems rather than mesoscale convective towers, as will be shown later in the paper.
Distribution of occurrences of the 11 WSs and clear sky over the whole globe (line) and the main stratocumulus regions (symbols; see Table 2 for the regions’ locations).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Distribution of occurrences of the 11 WSs and clear sky over the whole globe (line) and the main stratocumulus regions (symbols; see Table 2 for the regions’ locations).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Distribution of occurrences of the 11 WSs and clear sky over the whole globe (line) and the main stratocumulus regions (symbols; see Table 2 for the regions’ locations).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
The representativeness of the WS distribution makes the Azores an ideal site to study almost the full range of cloud regimes, to study the interactions of the various weather states with the atmospheric dynamics, and to evaluate model performance in simulating a wide range of cloud systems. On the contrary, stratocumulus-prone regions (e.g., Peruvian, Californian, and Namibian decks) have a disproportionally strong presence of stratiform low-cloud states (WS9–11), albeit with different relative occurrences and very few middle- and high-cloud occurrences (WS1–3 and WS5–6). Regions characterized by midlatitude storms (e.g., North Pacific and Atlantic; results not shown) have a similar state distribution to the Azores but with the deeper systems (WS1–2) better represented at the expense of the fair-weather state (WS7).
The WS distribution over the Azores shows the existence of a suite of storm-related weather states together with the more typical stratocumulus and shallow cumulus regimes. This indicates that two mechanisms capable of creating subsidence and supporting the presence of stratiform low clouds could be operating in that region: a semipermanent subtropical high pressure system (the Azores high in this case) and midlatitude storm-related cold-air outbreaks, created by traveling high pressure systems formed by the subsidence found in the postcold-frontal region of baroclinic storms (Bauer and Del Genio 2006; Mechem et al. 2010). Since the Azores is located near the boundary between these two types of regimes, the region can experience stratiform low clouds coming from both sources. This makes it a target location to study a wider range of boundary layer cloud formation mechanisms than those found in the more typical stratocumulus locations.
The WS distributions presented so far include occurrences from every season. However, it is common knowledge that the location of the main storm track changes with season, moving south during the winter (e.g., Hasanean 2004). To find out how this track movement impacts cloud regimes, the WS monthly occurrences were averaged over the 26.5 years of ISCCP observations for the Azores region. From the results shown in Fig. 3 (black lines), an annual cycle is visible for those WSs whose existence is typically linked to the presence or absence of a storm. For instance, the deep systems (WS1–2) have a stronger presence during wintertime, a time when synoptic systems should be tracking closer to the Azores region. This is also visible for the other storm-related cloud regimes such as WS3 and WS5. On the other hand, fair weather (WS7) is more abundant in the summer months. However, for the low-level clouds (WS8–11), only WS9 (the lowest-top stratocumulus) shows a similar cycle, perhaps as a result of its lower-top character, which would require a stronger subsidence regime (such as the one provided by the Azores high pressure system) than the other low-level cloud regimes. Thin cirrus (WS6) are the only clouds that appear to occur preferentially during a transition season (fall in this case), although WS11 and WS4 hint toward a peak of occurrences in spring.
Average annual cycle of WS occurrences over the Azores region (black line) and over its north and south edges (blue and red lines, respectively). Vertical lines represent one standard deviation in the overall occurrences.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Average annual cycle of WS occurrences over the Azores region (black line) and over its north and south edges (blue and red lines, respectively). Vertical lines represent one standard deviation in the overall occurrences.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Average annual cycle of WS occurrences over the Azores region (black line) and over its north and south edges (blue and red lines, respectively). Vertical lines represent one standard deviation in the overall occurrences.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Figure 3 also contrasts the latitudinal dependence of the WSs, showing their monthly averaged occurrences as obtained using only the four 2.5° northernmost and southernmost cells of the Azores region (blue and red lines, respectively). The results highlight the fact that this region is located near the boundary between the high pressure system to the south (subtropical weather) and the storm track to the north (midlatitude weather). This contrast is most clearly visible for WS7 and WS9 but also for the WS strongly related to the storms (i.e., WS2 and WS5). Note that these results relate to average annual cycles, with every year likely to diverge from them to various degrees. To provide an idea of the variability observed over the 26.5 years on record, the vertical bars in Fig. 3 illustrate the spread from one standard deviation for the overall average cycles.
Combining the measurements from the radar and lidar deployed in the Azores for CAP-MBL, the cloud vertical structure (CVS) of the different WSs can be derived and investigated. Based on Rémillard et al. (2012), the cloudiness is assessed in three layers of the troposphere: high (H), middle (M), and low (L), with heights of 7 and 3 km separating, respectively, the high and middle layers and the middle and low layers. Then, based on Tselioudis et al. (2013; and references therein), each profile is classified as single-layer clouds (1H, 1M, or 1L), continuous double-layer clouds (HxM or MxL where ‘x’ denotes the extension between layers), separated double-layer clouds (HM, HL, or ML), mixed double-layer clouds (HxML or HMxL), separated triple-layer clouds (HML), continuous triple-layer clouds (HxMxL), or clear. Figure 4 depicts the frequency of occurrence of each of these 13 CVSs for each WS, using two hours of ground-based observations surrounding each time stamp from ISCCP. The results show that the different WSs over the Azores present distinct vertical structures that vary from cloud covering the whole atmospheric column in the storm-related states (WS1–2), to clouds with tops at middle levels or below in WS5, to purely low-cloud decks in WS8–11. The results are very similar to the global results presented in Tselioudis et al. (2013) using space-borne active remote sensing instruments (onboard CloudSat and CALIPSO) to derive the vertical structure of the WSs. The main difference is the lack of high clouds associated with WS6 when seen from the ground, which is likely caused by the difficulty for the ground-based instruments to detect thin cirrus.
Cloud vertical structure of each WS, as detected by the radar and lidar from the ground site during CAP-MBL. The width of each CVS bar indicates the frequency of occurrence of this CVS in the particular WS. The white bars (spaces) indicate clear sky, and the light gray bars represent the sum of all CVSs occurring less than 5% of the time. For the separated low-level clouds, the opposing striped patterns illustrate the presence of cumulus or stratocumulus only (right- and left-leaning, respectively). The overlapping stripes indicate the combined observations of both types, and the portion left without stripes marks the presence of undefined low-level clouds.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Cloud vertical structure of each WS, as detected by the radar and lidar from the ground site during CAP-MBL. The width of each CVS bar indicates the frequency of occurrence of this CVS in the particular WS. The white bars (spaces) indicate clear sky, and the light gray bars represent the sum of all CVSs occurring less than 5% of the time. For the separated low-level clouds, the opposing striped patterns illustrate the presence of cumulus or stratocumulus only (right- and left-leaning, respectively). The overlapping stripes indicate the combined observations of both types, and the portion left without stripes marks the presence of undefined low-level clouds.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Cloud vertical structure of each WS, as detected by the radar and lidar from the ground site during CAP-MBL. The width of each CVS bar indicates the frequency of occurrence of this CVS in the particular WS. The white bars (spaces) indicate clear sky, and the light gray bars represent the sum of all CVSs occurring less than 5% of the time. For the separated low-level clouds, the opposing striped patterns illustrate the presence of cumulus or stratocumulus only (right- and left-leaning, respectively). The overlapping stripes indicate the combined observations of both types, and the portion left without stripes marks the presence of undefined low-level clouds.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
It must be noted that the plots include a light gray bar that represents the sum frequency of all CVSs occurring less than 5% of the time. Also, there are discrepancies between the cloud regimes associated with ISCCP WSs and the radar–lidar CVS. Multilayer clouds can explain many of these discrepancies; satellites analyze them as single-layer clouds, giving them an intermediate cloud-top pressure that likens the situation to midlevel clouds. Low-level clouds can also be misconstrued by satellites as midlevel clouds when a strong inversion is topping them. In that sense, these results are similar to those obtained by studies made over the Southern Ocean (e.g., Bodas-Salcedo et al. 2014; Mason et al. 2014). It is important also to realize that ISCCP WSs apply to a wide region (about 280 km), while the radar–lidar analysis is limited to a soda-straw view of what is advected over the instruments.
An attempt was made to further classify boundary layer cloud from radar retrievals, distinguishing cumulus and stratocumulus clouds from other types of low clouds using the classification of Rémillard et al. (2012) based on a cloud’s duration and its top variability. In a nutshell, cumulus clouds are defined as short-lived clouds (i.e., less than 20 min), while stratiform clouds have a stable top (i.e., hourly standard deviation radar-derived cloud-top height smaller than 100 m). Any other boundary layer cloud that does not satisfy both of these criteria remains undefined. Here, we consider four categories for the low clouds: pure cumulus, pure stratiform, a mixture of both during the hour, and presence of an undefined layer. Note that only the third category applies to a whole hour. The others are done on a radar profile basis. In Fig. 4, these categories are illustrated in each vertical structure that has separated low-level clouds using striping patterns (one each for cumulus and stratocumulus coverage, superposing them for their mixed coverage, and none for the rest). The separation illustrates that the low-level WSs that are more optically opaque (WS9–11) have a more continuous cloud cover (>80% per hour) dominated by a stratocumulus sheet (~45%). On the other hand, WS7 and WS8 experience more clear skies and a relatively high fraction of cumulus coverage (about half of the low-cloud coverage). The compositing of the ground-based remote sensing retrievals from the ARM mobile facility further verifies the uniqueness of the satellite-derived WS structures.
b. Dynamics influences
The state of the atmosphere and the mechanisms of cloud formation over the Azores are influenced by the passage of midlatitude storm systems. Using the MCMS dataset, time periods when such systems affect the Azores region are identified. Cumulating these time periods over each month of the dataset creates monthly occurrences of storm influence, which are then averaged over the 33 available years to provide a view of their average annual cycle over the region of interest (see the black solid line in Fig. 5). From its position near the main storm track, the Azores region encounters storm conditions about 40% of the time per year on average. The summer months, June–August (JJA), are the least stormy on average, with the minimum influence usually felt in July. The other months show a storminess nearly constant around 45%, with a slight maximum in spring. Nevertheless, these numbers display a strong interannual variability, as shown by the large standard deviations (see the shaded area in Fig. 5). For instance, 2009 started rather calm, with few storms influencing the ground site before June, while 2010 saw a strong peak of storminess in February. These two years illustrate that storms can occur in any month, especially around wintertime but also in the middle of summer. Storm influence on the Azores site itself lasts typically less than 3 days. Note that no distinction between the storms is attempted here, and a sustained influence could actually come from a succession of storms. Fairer conditions, on the other hand, can prevail much longer, with events often lasting over a week.
Average annual cycle of storm occurrences over the Azores region. The solid line represents the averaged overall occurrences, with the shaded area illustrating the range from one standard deviation. The broken lines separate the occurrences by the location of the storm center relative to the ground site: northwest quadrant (NW; black dashed line), northeast quadrant (NE; black dotted line), southwest quadrant (SW; gray dashed line), and southeast quadrant (SE; gray dotted line).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Average annual cycle of storm occurrences over the Azores region. The solid line represents the averaged overall occurrences, with the shaded area illustrating the range from one standard deviation. The broken lines separate the occurrences by the location of the storm center relative to the ground site: northwest quadrant (NW; black dashed line), northeast quadrant (NE; black dotted line), southwest quadrant (SW; gray dashed line), and southeast quadrant (SE; gray dotted line).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Average annual cycle of storm occurrences over the Azores region. The solid line represents the averaged overall occurrences, with the shaded area illustrating the range from one standard deviation. The broken lines separate the occurrences by the location of the storm center relative to the ground site: northwest quadrant (NW; black dashed line), northeast quadrant (NE; black dotted line), southwest quadrant (SW; gray dashed line), and southeast quadrant (SE; gray dotted line).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
The storm influences can be further separated according to the location of the storm center, as indicated in the MCMS dataset. Using Graciosa Island as a reference, each center’s location is classified into one of the four typical Cartesian quadrants: northwest, northeast, southeast, and southwest (see the broken lines in Fig. 5). The majority of the storms pass to the north of the Azores, usually affecting the region while their center is located to the northwest and secondarily to the northeast of the ground site. This probably relates to their track taking them farther north as they move eastward. Similarly, since the typical storm track is located well north of the ground site, only a few of the storms affecting the region have a center to the south. This separation indicates that the Azores region experiences primarily cold-frontal cloud structures, which extend south of the storm center, with fewer possible warm-frontal clouds, which generally extend to the east of the storm center.
We now investigate the overall impact of storms on the cloud systems of the Azores region. First, Fig. 6 compares the distributions of WS occurring in the grid cell containing the ground site in two situations: 1) as obtained when MCMS reports a storm influencing that 2.5° grid cell, regardless of its location relative to the site, and 2) as obtained when no storms are affecting the whole 10° × 10° Azores region. As expected, it shows that storms are accompanied by more of the deeper and higher cloud systems (WS1–3 and somewhat WS5), while clouds confined in the low levels (WS7–11) tend to characterize conditions not influenced by a storm center. However, WS8 (shallow cumulus) occurs with similar frequencies when within a storm or when well outside its influence. It is important to note that the deep convection WS1 occurs almost exclusively under the storm influence, indicating that it represents storm-embedded convection.
Comparison of the distribution of WS occurrences over the ground site when it is under storm influence (closed symbols) and when the whole Azores region is free of storm (open symbols). The solid line represents the overall distribution of WS occurrences over the ground site.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Comparison of the distribution of WS occurrences over the ground site when it is under storm influence (closed symbols) and when the whole Azores region is free of storm (open symbols). The solid line represents the overall distribution of WS occurrences over the ground site.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Comparison of the distribution of WS occurrences over the ground site when it is under storm influence (closed symbols) and when the whole Azores region is free of storm (open symbols). The solid line represents the overall distribution of WS occurrences over the ground site.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
To look deeper into cloud variability during the life cycle of storm events, a timeline of a storm influence over Graciosa Island was built using the storminess dataset. First, separate storm events were identified, with two or more nonstormy MCMS time steps between events needed to consider them as separate. For each event, the initial and final MCMS time steps were identified, and two 6-h periods (corresponding to three ISCCP time steps) were selected around them. Note that the two transition ISSCP time steps (corresponding to the one just before the first known influence and the one just after the last known influence) are excluded since the storm influence remains undetermined. Then, combining all the events, the frequency of occurrence of WS over Graciosa Island was calculated at these four characteristic periods: before (6 h before the initial step), at the beginning of (6 h following the initial step), at the ending of (6 h preceding the final step), and after (6 h after the final step) a storm passage. It is important to note that these four distributions are obtained from a smaller subset of the data than the one selected for Fig. 6. From the four resulting WS distributions (see Fig. 7), it appears that, at the beginning and ending stages of a storm passage, the Azores region is marked by the presence of the typical storm-related cloud structures (WS1–3 and WS5). The widespread observation of deep convective regimes within the storm links back to the idea put forward above that a front is the main impacting storm feature. Occurrences of these storm-related WSs decrease as the storm passes, leaving room for low-level clouds and fair-weather conditions (WS7–8). As for the stratocumulus states (WS9–11), they are rarely observed under storm influence, suggesting a preference to occur well outside of storms (open symbols in Fig. 7). The WS timeline in Fig. 7 establishes that WS8 is a postfrontal cloud structure as it occurs preferentially at the ending or after the passage of a storm. This hints toward a cold-air outbreak environment favoring the formation of the shallow cumulus fields of WS8. At the same time, the midlevel WS5 and the thick cirrus WS3 states are established as prefrontal cloud structures, as they happen mainly before or around the arrival of a storm center.
Distribution of WS occurrences around the beginning (6-h periods before and after the initial time step; square and x-shaped symbols, respectively) and ending (6-h periods before and after the final time step; plus and circle symbols, respectively) of a storm event as observed above Graciosa Island.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Distribution of WS occurrences around the beginning (6-h periods before and after the initial time step; square and x-shaped symbols, respectively) and ending (6-h periods before and after the final time step; plus and circle symbols, respectively) of a storm event as observed above Graciosa Island.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Distribution of WS occurrences around the beginning (6-h periods before and after the initial time step; square and x-shaped symbols, respectively) and ending (6-h periods before and after the final time step; plus and circle symbols, respectively) of a storm event as observed above Graciosa Island.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
To further explore the relationships between WS occurrence and dynamic regime, an additional regime analysis is performed that uses the direction and speed of the lower-tropospheric winds to identify the prevailing dynamic system. Figure 8a shows the distributions obtained with the 850-hPa winds from ERA-Interim over the Azores site covering the ISCCP analysis period. Similar results were obtained using different levels, or winds retrieved by the radiosondes launched as part of the field campaign. In the overall picture (Fig. 8a), winds have a strong westerly component, with a fairly equal separation between the north and south directions. An interesting feature to note in that distribution is the minimum occurrence in the southeastern quadrant. This feature suggests that the island effect is minimal on the ARM site on Graciosa Island; since that island has a northwest–southeast main axis, the minimum in southeastern winds indicates a strong lack of air masses passing above the island itself, especially its main topographical features, prior to their arrival above the site near the northern shore.
Wind roses of the 850-hPa winds obtained by ERA-Interim above the Azores ground site: (a) compositing all years together, (b),(c) separating those associated with a storm influence or not, (d)–(i) and separating the periods for WS7–12. The number of cases considered for each panel is indicated in parentheses.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Wind roses of the 850-hPa winds obtained by ERA-Interim above the Azores ground site: (a) compositing all years together, (b),(c) separating those associated with a storm influence or not, (d)–(i) and separating the periods for WS7–12. The number of cases considered for each panel is indicated in parentheses.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Wind roses of the 850-hPa winds obtained by ERA-Interim above the Azores ground site: (a) compositing all years together, (b),(c) separating those associated with a storm influence or not, (d)–(i) and separating the periods for WS7–12. The number of cases considered for each panel is indicated in parentheses.
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Separating the wind directions according to the presence or absence of a storm in the vicinity of the Azores results in the two wind roses shown in Figs. 8b,c. The separation emphasizes two wind regimes differentiated by both speed and direction. On one hand, a storm influence has a very focused distribution, which is strongly made up of fast, southwesterly winds. This distribution is reminiscent of cold front environment while the storms pass north of the region (Simmonds et al. 2012), as noted above. On the other hand, the distribution outside of storm influences is much broader and includes weaker winds indicative of a calmer, more stable atmospheric state associated with high pressure regimes. Postfrontal cold-air outbreaks are also likely present in that distribution, as marked by its peak around the northwestern direction. A few trade wind situations appear to have been sampled, too, as visible in the tail of northeastern winds. The wind roses for the low-cloud-dominated states (WS7–11) and the clear-sky WS12 are plotted in Figs. 8d–i. The distribution of winds aloft during WS8 (see Fig. 8e) confirms the idea that the WS8 shallow cumulus fields form in cold-air outbreaks, showing a distribution containing mostly strong northwesterly winds. On the other hand, the other low-level clusters are characterized by fairly different distributions with generally weaker winds and showing a preference to veer toward northeasterly trade winds as the stratocumulus decks get optically thicker.
c. Model evaluation
The results of the satellite and radar analysis presented so far indicate that the Azores is a location where a large fraction of global cloud structures is well represented and that the WS analysis captures well those separate cloud regimes and can be used to put the local cloud structures into the global cloud field context. Consequently, we can use the Azores site results to evaluate the cloud field properties of climate models and use local site evaluation to understand global cloud deficiencies. Therefore, an evaluation of climate model cloud simulations over the Azores is performed next. Figure 9 shows the WS distribution over the Azores for a suite of 10 climate models that participated in phase 5 of the Coupled Model Intercomparison Project (CMIP5; Taylor et al. 2012) and provided ISCCP simulator output for their present-day climate simulations, which makes it possible to calculate their respective WS distribution. The plot also shows the ISCCP Azores WS distribution as shown in Fig. 2. There is a spread in the overall differences of the model and ISCCP WS amounts, but the most striking feature is the underestimation in all models of the shallow cumulus WS8, which was established here to occur in postfrontal cold-air outbreaks. This result is in agreement with Southern Ocean model evaluation studies (e.g., Williams et al. 2013; Bodas-Salcedo et al. 2014; Naud et al. 2014), which found cloud underprediction in the cold-air sector of Southern Hemisphere fronts, and extends their result to the North Atlantic storm track. All models also underpredict the storm related middle-cloud WS5, whereas most models underpredict the primary stratocumulus weather state (WS10) while predicting correctly the amounts of the other two stratocumulus weather states (WS9 and WS11). Also, all models overpredict the fair weather WS7 and the clear sky WS12. The same model deficiencies observed over the Azores are also found when the global model cloud fields are compared to the ISCCP observations (results not shown). This implies that analysis results from climate model evaluation over the Azores can be used to understand the mechanisms behind cloud simulation deficiencies occurring in the global domain.
Comparison of the distribution of WS occurrences over the Azores region for 10 climate models (symbols) with the observations (solid line).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Comparison of the distribution of WS occurrences over the Azores region for 10 climate models (symbols) with the observations (solid line).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Comparison of the distribution of WS occurrences over the Azores region for 10 climate models (symbols) with the observations (solid line).
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
4. Discussion
The Azores region is characterized by a wide variety of cloud scenes, as evidenced by a weather state analysis performed on PC-TAU histograms obtained from ISCCP data. The vertical structure of these scenes was observed from the ground during a 19-month period in 2009/10, as a wide variety of instruments were deployed on Graciosa Island for the CAP-MBL field campaign. Although low-level cloud structures predominantly feature in that region, storm-related scenes are also quite frequent, more so than in the main stratocumulus regions. This is caused by the Azores region being located close enough to the midlatitude storm track, enabling storms to regularly influence the site, especially during winter months. The dynamic regime analyses performed here established clear relationships between WS occurrence and the prevailing dynamic systems of the Azores region. It showed that the deep convective WS1 constitutes storm-embedded convection, that the shallow cumulus WS8 occurs mostly in cold-air outbreaks following cold front passages, and that the stratocumulus decks (WS9–11) occur outside storm influence mostly under trade wind conditions.
Measurements and retrievals from field campaigns are typically used to evaluate simulated clouds from cloud-resolving models and large-eddy simulation (LES) models. However, the results of the present study indicate that the Azores site is also a good location to evaluate cloud simulations of global models, as it experiences most cloud regimes with frequencies similar to the ones found for the global domain. A comparison of the WS distribution from 10 climate models to the observed distribution shows that the models are capturing a large part of the variability of cloud regimes at that site, despite some obvious spread between the models. The main model biases appear to be related to the underprediction of WS5 and WS10 and the relative distributions of WS7 and WS8; all analyzed models exhibit a strong lack of WS8 (shallow cumulus fields), while all but one produced too much of WS7 (fair-weather cloud fields). A possible explanation for this discrepancy could be that the models do not create enough shallow cumulus clouds to obtain the radiative signature of WS8, relegating these scenes to the WS7 category. Another scenario would be that the models create shallow cumulus structures but cannot sustain them long enough, treating them as a more transient state than it really is. In this case, evaporating cumulus clouds would again relegate these scenes to WS7. The analysis of frontal system passage and winds aloft indicate that the presence of WS8 is for the most part associated with cold-air outbreaks produced behind cold fronts, which can last long after the storm has passed. This result suggests that the models’ bias originates from a poor representation of cold-air outbreak conditions, either by not creating the right cloud regimes or by not sustaining them long enough. Note that previous studies have attributed climate model shortwave radiation biases over the Southern Ocean to the underprediction of clouds in the cold sector of baroclinic storms (e.g., Bodas-Salcedo et al. 2012, 2014; Williams et al. 2013).
Once input from the WS analysis establishes which cloud regimes are more difficult to reproduce in models, as described above, then selecting case studies when the identified WSs dominate the region makes it possible to employ cloud process models and field campaign data to study in detail cloud formation and dissipation processes. Two examples are presented in Fig. 10, where visible imagery acquired by the MODIS onboard the Aqua satellite while flying over the Azores region are shown with the WS numbers from the ISCCP analysis superimposed on each image. These images were selected at a time instant when WS8 dominates the region (13 November 2009; Figs. 10a,c,e) and one when WS10 is the dominant regime (22 November 2009; Figs. 10b,d,f). The figure also shows a few hours of the ARM mobile facility (AMF) W-band radar observed reflectivity with the lidar-derived cloud base, as well as the MCMS analysis of sea level pressure and storm area of influence around the time of the satellite images. It can be seen that the WS8-dominated time period is characterized by the presence of an extended broken-cloud field consisting of open cell clouds that are commonly characterized as shallow cumulus clouds in the literature. The radar image also illustrates the intermittency of the cloud field and shows that individual cells have high water contents and are, for the most part, precipitating with the rain reaching the ground. The presence of much weaker cells can also be noted at times. Further, the MCMS analysis reveals the presence of a cold-air outbreak in the aftermath of a strong storm located to the north of the site. On the other hand, for the WS10-dominated time period, the satellite image shows a more consistent cloud deck that is characteristic of marine stratocumulus decks. The radar image confirms that the cloud deck is persistent, with moderate water content amounts, and indicates some drizzle below the cloud base that, for the most part, does not reach the ground. For this case, the MCMS analysis exposes the influence on the site of a well-defined Azores high pressure system. The two cases presented here are therefore representative of cold-air outbreak shallow cumulus and high-pressure-dominated stratocumulus conditions and can be used in process studies utilizing cloud-resolving or LES models to better understand the prevailing cloud processes.
Examples of open and closed cell formation of low clouds, as seen from (a),(b) satellite imagery [Aqua MODIS amalgam provided by NASA/GSFC rapid response team, centered on Graciosa Island (blue dot) with the ISCCP-based WS indicated for the 9 closest cells], (c),(d) ground-based instruments (radar reflectivity in colors, with lidar cloud base as black dots), and (e),(f) model reanalysis [field of sea level pressure (colors) with the storm center and influence from MCMS (symbols and shaded regions, respectively) and a white rectangle indicating the Azores region].
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Examples of open and closed cell formation of low clouds, as seen from (a),(b) satellite imagery [Aqua MODIS amalgam provided by NASA/GSFC rapid response team, centered on Graciosa Island (blue dot) with the ISCCP-based WS indicated for the 9 closest cells], (c),(d) ground-based instruments (radar reflectivity in colors, with lidar cloud base as black dots), and (e),(f) model reanalysis [field of sea level pressure (colors) with the storm center and influence from MCMS (symbols and shaded regions, respectively) and a white rectangle indicating the Azores region].
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
Examples of open and closed cell formation of low clouds, as seen from (a),(b) satellite imagery [Aqua MODIS amalgam provided by NASA/GSFC rapid response team, centered on Graciosa Island (blue dot) with the ISCCP-based WS indicated for the 9 closest cells], (c),(d) ground-based instruments (radar reflectivity in colors, with lidar cloud base as black dots), and (e),(f) model reanalysis [field of sea level pressure (colors) with the storm center and influence from MCMS (symbols and shaded regions, respectively) and a white rectangle indicating the Azores region].
Citation: Journal of Climate 28, 24; 10.1175/JCLI-D-15-0066.1
The combined inspection of satellite and radar images shows that the WS cluster analysis captures well the low-cloud regimes and allows us to generalize the conclusions of case study analyses over the Azores site. This study also demonstrates that regime-based methods applied to in situ and satellite observations can be used to study cloud processes and evaluate models ranging from process-resolving to global climate models. Future studies will benefit from the upcoming expansion of both the satellite and the in situ datasets since the ISCCP analysis is being extended to years beyond 2009, providing a full overlap of the 19-month field campaign, and a permanent site is under development near the field campaign site that will include similar instruments. The presence of a permanent site will provide a wealth of data to study a wide range of cloud fields and their environment. Together, these will enhance the work presented here, allowing for a better characterization of climatological values (by reducing the impact of interannual variability) and storm influences. The present study demonstrates that all the tools are now in place to perform process-resolving LES or single-column model (SCM) simulations of CAP-MBL individual cases and, using the cloud regime analysis, to generalize the case study results and attempt to explain whether major GCM cloud deficiencies relate to the poor representation of atmospheric dynamics mechanisms or to issues related to the parameterization of cloud microphysical processes.
Acknowledgments
The authors acknowledge the support of the DOE Atmospheric Systems Research (ASR) Program, under Grant DE-SC0006712.
REFERENCES
Anderberg, N. R., 1973: Cluster Analysis for Applications. Elsevier, 359 pp.
Bauer, M., and A. D. Del Genio, 2006: Composite analysis of winter cyclones in a GCM: Influence on climatological humidity. J. Climate, 19, 1652–1672, doi:10.1175/JCLI3690.1.
Bauer, M., G. Tselioudis, and W. B. Rossow, 2011: MAP climatology of mid-latitude storminess, version using NCEP–DOE Reanalysis 2. NASA/GISS, accessed 1 February 2013. [Available online at http://gcss-dime.giss.nasa.gov/mcms/].
Bodas-Salcedo, A., K. D. Williams, P. R. Field, and A. P. Lock, 2012: The surface downwelling solar radiation surplus over the Southern Ocean in the Met Office model: The role of midlatitude cyclone clouds. J. Climate, 25, 7467–7486, doi:10.1175/JCLI-D-11-00702.1.
Bodas-Salcedo, A., and Coauthors, 2014: Origins of the solar radiation biases over the Southern Ocean in CFMIP2 models. J. Climate, 27, 41–56, doi:10.1175/JCLI-D-13-00169.1.
Bony, S., and J.-L. Dufresne, 2005: Marine boundary layer clouds at the heart of tropical cloud feedback uncertainties in climate models. Geophys. Res. Lett., 32, L20806, doi:10.1029/2005GL023851.
Bony, S., and Coauthors, 2006: How well do we understand and evaluate climate change feedback processes? J. Climate, 19, 3445–3482, doi:10.1175/JCLI3819.1.
Bretherton, C. S., and Coauthors, 2004: The EPIC 2001 stratocumulus study. Bull. Amer. Meteor. Soc., 85, 967–977, doi:10.1175/BAMS-85-7-967.
Dufresne, J.-L., and S. Bony, 2008: An assessment of the primary sources of spread of global warming estimates from coupled atmosphere–ocean models. J. Climate, 21, 5135–5144, doi:10.1175/2008JCLI2239.1.
Field, P. R., and R. Wood, 2007: Precipitation and cloud structure in midlatitude cyclones. J. Climate, 20, 233–254, doi:10.1175/JCLI3998.1.
Ghate, V. P., M. A. Miller, B. A. Albrecht, and C. W. Fairall, 2015: Thermodynamic and radiative structure of stratocumulus-topped boundary layers. J. Atmos. Sci., 72, 430–451, doi:10.1175/JAS-D-13-0313.1.
Hartmann, D. L., M. E. Ockert-Bell, and M. L. Michelsen, 1992: The effect of cloud type on Earth’s energy balance: Global analysis. J. Climate, 5, 1281–1304, doi:10.1175/1520-0442(1992)005<1281:TEOCTO>2.0.CO;2.
Hasanean, H. M., 2004: Variability of the North Atlantic subtropical high and associations with tropical sea-surface temperature. Int. J. Climatol., 24, 945–957, doi:10.1002/joc.1042.
Hoskins, B. J., and K. I. Hodges, 2002: New perspectives on the Northern Hemisphere winter storm tracks. J. Atmos. Sci., 59, 1041–1061, doi:10.1175/1520-0469(2002)059<1041:NPOTNH>2.0.CO;2.
Jakob, C., and G. Tselioudis, 2003: Objective identification of cloud regimes in the tropical western Pacific. Geophys. Res. Lett., 30, 2082, doi:10.1029/2003GL018367.
Jakob, C., and C. Schumacher, 2008: Precipitation and latent heating characteristics of the major tropical western Pacific cloud regimes. J. Climate, 21, 4348–4364, doi:10.1175/2008JCLI2122.1.
Klein, S. A., and D. L. Hartmann, 1993: The seasonal cycle of low stratiform clouds. J. Climate, 6, 1587–1606, doi:10.1175/1520-0442(1993)006<1587:TSCOLS>2.0.CO;2.
Lau, N.-C., and M. W. Crane, 1997: Comparing satellite and surface observations of cloud patterns in synoptic-scale circulation systems. Mon. Wea. Rev., 125, 3172–3189, doi:10.1175/1520-0493(1997)125<3172:CSASOO>2.0.CO;2.
Mason, S., C. Jakob, A. Protat, and J. Delanoë, 2014: Characterizing observed midtopped cloud regimes associated with Southern Ocean shortwave radiation biases. J. Climate, 27, 6189–6203, doi:10.1175/JCLI-D-14-00139.1.
Mechem, D., Y. L. Kogan, and D. M. Schultz, 2010: Large-eddy observation of post-cold-frontal continental stratocumulus. J. Atmos. Sci., 67, 3368–3383, doi:10.1175/2010JAS3389.1.
Mechoso, C. R., and Coauthors, 2014: Ocean–cloud–atmosphere–land interactions in the southeastern Pacific: The VOCALS program. Bull. Amer. Meteor. Soc., 95, 357–375, doi:10.1175/BAMS-D-11-00246.1.
Naud, C. M., J. F. Booth, and A. D. Del Genio, 2014: Evaluation of ERA-Interim and MERRA cloudiness in the Southern Ocean. J. Climate, 27, 2109–2124, doi:10.1175/JCLI-D-13-00432.1.
Norris, J. R., and S. A. Klein, 2000: Low cloud type over the ocean from surface observations. Part III: Relationship to vertical motion and the regional surface synoptic environment. J. Climate, 13, 245–256, doi:10.1175/1520-0442(2000)013<0245:LCTOTO>2.0.CO;2.
Norris, J. R., Y. Zhang, and J. M. Wallace, 1998: Role of low clouds in summertime atmosphere–ocean interactions over the North Pacific. J. Climate, 11, 2482–2490, doi:10.1175/1520-0442(1998)011<2482:ROLCIS>2.0.CO;2.
Rémillard, J., P. Kollias, E. Luke, and R. Wood, 2012: Marine boundary layer cloud observations in the Azores. J. Climate, 25, 7381–7398, doi:10.1175/JCLI-D-11-00610.1.
Rossow, W. B., G. Tselioudis, A. Polak, and C. Jakob, 2005: Tropical climate described as a distribution of weather states indicated by distinct mesoscale cloud property mixtures. Geophys. Res. Lett., 32, L21812, doi:10.1029/2005GL024584.
Rudeva, I., and S. K. Gulev, 2011: Composite analysis of North Atlantic extratropical cyclones in NCEP–NCAR reanalysis data. Mon. Wea. Rev., 139, 1419–1446, doi:10.1175/2010MWR3294.1.
Simmonds, I., K. Keay, and J. A. T. Bye, 2012: Identification and climatology of Southern Hemisphere mobile fronts in a modern reanalysis. J. Climate, 25, 1945–1962, doi:10.1175/JCLI-D-11-00100.1.
Soden, B. J., and G. A. Vecchi, 2011: The vertical distribution of cloud feedback in coupled ocean-atmosphere models. Geophys. Res. Lett., 38, L12704, doi:10.1029/2011GL047632.
Stephens, G. L., and T. J. Greenwald, 1991: The Earth’s radiation budget and its relation to atmospheric hydrology: 2. Observations of cloud effects. J. Geophys. Res., 96, 15 325–15 340, doi:10.1029/91JD00972.
Stevens, B., and Coauthors, 2003: Dynamics and chemistry of maritime stratocumulus—DYCOMS-II. Bull. Amer. Meteor. Soc., 84, 579–593, doi:10.1175/BAMS-84-5-579.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498, doi:10.1175/BAMS-D-11-00094.1.
Tselioudis, G., 2012: Global weather states from the ISCCP D1 cloud dataset. NASA/GISS, accessed 11 January 2013. [Available online at http://isccp.giss.nasa.gov/gcluster.html.]
Tselioudis, G., W. Rossow, Y. Zhang, and D. Konsta, 2013: Global weather states and their properties from passive and active satellite cloud retrievals. J. Climate, 26, 7734–7746, doi:10.1175/JCLI-D-13-00024.1.
Williams, K. D., and G. Tselioudis, 2007: GCM intercomparison of global cloud regimes: Present-day evaluation and climate change response. Climate Dyn., 29, 231–250, doi:10.1007/s00382-007-0232-2.
Williams, K. D., and Coauthors, 2013: The Transpose-AMIP II Experiment and its application to the understanding of Southern Ocean cloud biases in climate models. J. Climate, 26, 3258–3274, doi:10.1175/JCLI-D-12-00429.1.
Wood, R., and Coauthors, 2015: Clouds, aerosol, and precipitation in the marine boundary layer: An ARM mobile facility deployment. Bull. Amer. Meteor. Soc., 96, 419–440, doi:10.1175/BAMS-D-13-00180.1.
Wyant, M. C., C. S. Bretherton, J. T. Bacmeister, J. T. Kiehl, I. M. Held, M. Zhao, S. A. Klein, and B. J. Soden, 2006: A comparison of low-latitude cloud properties and their response to climate change in three AGCMs sorted into regimes using mid-tropospheric vertical velocity. Climate Dyn., 27, 261–279, doi:10.1007/s00382-006-0138-4.
