Clouds are both produced by and interact with the mesoscale and synoptic-scale structure of extratropical cyclones (ETCs) in ways that are still not well understood. Cloud-scale radiative and latent heating modifies the thermal environment, leading to a response in the dynamics that can in turn feed back on cloud distribution and microphysical properties. Key to the structure of ETCs is the warm conveyor belt (WCB); the poleward-ascending airstream that produces the bulk of the clouds and precipitation. This paper examines a long-lived WCB that persisted over the western North Atlantic Ocean in nearly the same location for several days. During this time, the storm was sampled multiple times by NASA’s A-Train satellite constellation, and a clear transition from stratiform to convective clouds was observed. Examination of coincident temperature and water vapor data reveals destabilization of the thermodynamic profile after the cyclone reached maturity. CloudSat radar reflectivity from two sequential overpasses of the warm front depicts a change from stratiform to convective cloud structure, and high-frequency microwave data reveal an increase in the amount of ice hydrometeors. The presence of convection may serve to strengthen the warm frontal trough while slowing the movement of the primary low pressure center. The stratiform–convective transition cannot be detected from passive measurements of cloud-top pressure. The results demonstrate the effectiveness of multivariate satellite observations for examining the outcome of dynamic processes in ETCs, and highlight the need for more rapid temporal profiling in future remote sensing observing systems.
Extratropical cyclones play a vital role in regulating Earth’s energy balance through poleward meridional transport of energy and through the effects of the clouds they produce within the climate system. Extratropical cyclones (ETCs) are an important source of freshwater in the midlatitudes, as they can produce most of the precipitation received in the planet’s temperate zones (Heideman and Fritsch 1988; Hawcroft et al. 2012), with up to 90% of the rainfall in storm-track regions being produced by fronts associated with ETCs (Catto et al. 2012). The precipitation produced along the frontal regions originates in the planetary boundary layer (PBL) to the south and east of the cyclone center, and is transported by the poleward-ascending warm conveyor belt airstream (Harrold 1973; Browning et al. 1973; Carlson 1980; Browning 1986; Wernli 1997; Eckhardt et al. 2004; Madonna et al. 2014).
Warm conveyor belts (WCBs) efficiently precipitate nearly all the moisture they transport, and make a substantial contribution to the total precipitation in the extratropics. They produce approximately half of the wintertime precipitation (Eckhardt et al. 2004; Field and Wood 2007), and are associated with more than 70% of extreme precipitation events in the southeastern United States, eastern China, Japan, and South America (Pfahl et al. 2014). WCBs are more frequent in the winter seasons compared to summer in both hemispheres, have a stronger seasonal cycle in the Northern Hemisphere, and are particularly frequent in regions of intense baroclinicity and high low-level moisture content (Stohl 2001; Eckhardt et al. 2004; Madonna et al. 2014). Given that WCBs mainly originate in the moist subtropical marine PBL between 20° and 47° in the Northern Hemisphere (Wernli and Davies 1997; Madonna et al. 2014), they can transport large quantities of sensible and latent heat upward and poleward (Browning 1990). WCBs are crucial for the formation of mixed-phased clouds and precipitation within ETCs, along with cirrus clouds in their outflow (Browning 1990; Madonna et al. 2014). Moisture sources for WCBs are typically located over the ocean and close to their respective starting points, with their uptake of moisture often being associated with anomalies of latent heat flux due to a reduction of near-surface relative humidity and enhanced wind velocity (Neiman and Shapiro 1993; Pfahl et al. 2014). As air in the WCB progresses rapidly poleward during its ascent, the specific humidity decreases significantly as the water vapor condenses (Madonna et al. 2014), leading to the production of clouds and precipitation, which are strongly correlated to the amount of moisture transport by the WCB (Field and Wood 2007). While recent studies have been careful to define WCBs in terms of their origin and depth of ascent using parcel trajectories (e.g., Madonna et al. 2014), we characterize the WCB more generally as a poleward-moving airstream within an ETC featuring large amounts of water vapor observed through precipitable water and cloud cover.
Numerous case studies have examined the dynamics of the flow through the WCB, as well as the effect of condensational heating in the WCB on cyclone strength and frontal structure. Several have examined the effect of latent heating on storm strength and structure (e.g., Reed et al. 1988; Kuo et al. 1991; Davis 1992; Whitaker and Davis 1994; Stoelinga 1996; Businger et al. 2005), while others have focused more specifically on fronts (Posselt and Martin 2004; Reeves and Lackmann 2004; Igel and van den Heever 2014) and the WCB itself (Boutle et al. 2011; Joos and Wernli 2012; Flaounas et al. 2016). While significant progress has been made in recent years, there is much yet to examine regarding the interaction between cloud processes and the thermodynamic and dynamic structure of frontal zones, and how changes in cyclone circulation and environment are related to variations in the distribution and intensity of clouds and precipitation. To this end, this paper examines a long-lived WCB associated with a marine ETC that occurred in late November 2006, and the dynamics of its warm frontal zone. This cyclone formed east of the Florida coast as an upper-level trough moved into the southeastern portion of the United States. For the next five days, the cyclone evolved similarly to the marine cyclone model proposed by Shapiro and Keyser (1990) and slowly traveled northeastward parallel to the United States, exhibiting a significant WCB. Once the cyclone matured, the cyclone center remained in relatively the same location for over 24 h while the WCB advanced poleward ahead of the cyclone. As the WCB remained in approximately the same longitudinal location, it was sampled several times by NASA’s Afternoon constellation of Earth Observing System (EOS) satellites (A-Train).
The research presented here leverages a unique opportunity to conduct a multivariate analysis of the time evolution of the cloud structure and precipitation within a long-lived WCB, as most other studies have used composites of satellite observations to examine the connection between extratropical cyclone dynamics and cloud features (Evans et al. 1994; Lau and Crane 1995, 1997; Klein and Jakob 1999; Tselioudis et al. 2000; Tselioudis and Rossow 2006; Naud et al. 2006, 2010, 2012, 2014, 2015; Field and Wood 2007; Field et al. 2008; Posselt et al. 2008; Berry et al. 2011; Govekar et al. 2011, 2014; Bender et al. 2012). The instruments on board the A-Train reveal a clear transition from stratiform to convective cloud structures during the evolution of the WCB, and examination of the coincident thermodynamic structure reveals a discernible transition from stable to unstable thermodynamic structure within the cyclone and its warm conveyor belt.
The remainder of this paper is organized as follows: an overview of the A-Train observations and analysis data used in this study is presented in section 2. A synoptic overview of the case is presented in section 3, while results from A-Train data are detailed in section 4. A discussion of the implications of the results for the design of future satellite missions is offered in section 5. A summary and conclusions are contained in section 6.
2. Data and methods
This paper utilizes data from NASA’s A-Train satellite constellation, specifically from the Moderate Resolution Imaging Spectroradiometer (MODIS; Barnes et al. 1998), the Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E; Kawanishi et al. 2003), and CloudSat (Stephens et al. 2002). AMSR-E and MODIS are located on board the Aqua satellite, whose aim is to explore the global hydrological cycle and its role in the Earth’s climate system (Parkinson 2003). The AMSR-E instrument is a conically scanning passive microwave radiometer with 6 horizontally and vertically polarized frequencies (6.925, 10.65, 18.7, 23.8, 36.5, and 89.0 GHz), observing water-related geophysical variables. Retrieval algorithms combine the brightness temperatures observed from the multiple dual-polarized channels to retrieve estimates of precipitation, integrated water vapor, and integrated cloud liquid water, along with other atmospheric and surface variables (Kawanishi et al. 2003).
MODIS has 36 channels that span the visible and infrared spectral bands, with wavelengths ranging from 0.4 to 14.5 μm. MODIS radiances can be used to retrieve cloud-top properties (i.e., pressure and temperature), precipitable water, and temperature and water vapor profiles (Parkinson 2003). It can also retrieve aerosol optical depth and size distribution in the daytime, but these are limited in the presence of cloud cover and sun glint (King et al. 2003). Cloud optical thickness and effective radius is derived globally using six visible and near-infrared bands at 1-km spatial resolution (King et al. 2003; Parkinson 2003). Of all of the instruments on board the Aqua satellite, MODIS has the finest spatial resolution, ranging from 250 m in the visible to 500 m in the shortwave infrared and 1 km in the thermal infrared (Parkinson 2003; Kawanishi et al. 2003).
The CloudSat mission is designed to produce a global survey of clouds, and to quantitatively evaluate their representation in global atmospheric circulation models. CloudSat provides a rich source of information for evaluating cloud properties derived from other satellite data, such as those observed by the instruments on board Aqua. CloudSat is by nature limited in its ability to observe the horizontal extent of clouds since it is a nonscanning nadir-pointing instrument. Nevertheless, as the first 94-GHz space-borne cloud profiling radar (CPR), CloudSat is designed to measure vertical structure of clouds and precipitation from space, stimulate important research on clouds and precipitation, and provide a demonstration of 94-GHz space-borne technology (Stephens et al. 2002). The relatively high frequency was chosen to optimize cloud detection, at the expense of attenuation in moderate to heavy precipitation. Radar measurements along CloudSat’s track are averaged at 0.32-s time intervals, which provides cloud and precipitation information over a 1.4-km elliptical footprint with 500-m vertical range gates that are oversampled to give an effective vertical resolution of 250 m between the surface and 30 km in altitude (Stephens et al. 2002, 2008). Thermodynamic (temperature and water vapor) fields along the CloudSat track are obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) operational analysis, which were interpolated in space and time to the CloudSat track at a 0.5° resolution (termed the ECMWF-AUX dataset in the set of CloudSat data products; Posselt et al. 2008).
Three-dimensional wind, temperature, and water vapor fields are obtained from the National Oceanic and Atmospheric Administration (NOAA) National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) analysis on a 1.0° × 1.0° grid. Data were obtained from the NOAA National Operational Model Archive and Distribution System (NOMADS; http://nomads.ncdc.noaa.gov/). We identify the location of the WCB using vertically integrated water vapor content (precipitable water vapor) and winds in the lower free troposphere. The WCB in our case is defined to be an airstream that is rooted in the tropics and proceeds poleward, curving cyclonically and anticyclonically as it ascends east of the cyclone center. While all clouds observed in the MODIS, AMSR-E, and CloudSat data ultimately result from rising motion produced by the cyclone, we note that a portion of the cloud shield in frontal zones and around the cyclone center was likely not produced by trajectories that satisfy the Lagrangian criterion used to define WCBs (e.g., as defined via detailed trajectory analysis by Madonna et al. 2014).
3. Synoptic overview
During 20–21 November 2006, a deep upper-tropospheric trough moved into the southeastern United States, and by 1800 UTC 21 November, a weak region of relatively low surface pressure (1008 hPa) associated with the upper-level trough had formed over the western North Atlantic Ocean east of South Carolina (Figs. 1a and 1g). East of the newly formed cyclone, southerly winds began to advect water vapor northward from the tropics, as can be seen in the observed precipitable water overlaid with the 700-hPa wind field (Fig. 1d). As the WCB formed on 21 November, two outflow branches were observed, curving cyclonically and anticyclonically away from northward edge of the WCB (Figs. 1d,e) (Martinez-Alvarado et al. 2014). Strong water vapor flux convergence is indicated in the region that stretched from Cape Hatteras, North Carolina, to the northeast. The northern boundary of the WCB is approximately marked by the 40-mm precipitable water contour, which extended along the low-level warm front at this time. Over the subsequent 24 h, the surface low pressure center remained nearly stationary (Figs. 1g–i) while the northernmost extent of the WCB migrated slowly poleward (Figs. 1d–f). The low pressure center at upper levels also moved poleward, and by 23 November the upper- and lower-tropospheric low pressure centers were collocated (Figs. 1a–c and 2a,d,g). Precipitable water values in the WCB south of the warm front remained consistently high, exceeding 55 mm to the east of the surface low pressure center.
As the WCB progressed poleward during 22–24 November, the southerly winds increased in magnitude, exceeding 40 kt (1 kt = 0.5144 m s−1) within the WCB at the 700-hPa level (Fig. 2d). Clear evidence of a decrease in the cyclonically curving branch of the WCB is evident in Figs. 2d–f. Spatial separation between the cyclone center and the anticyclonically curving branch of the WCB led to seclusion of a region of relatively high water vapor near the cyclone center. In addition to the anticyclonic curvature in the WCB, the low-level anticyclonic circulation east of the parent low caused the WCB to narrow as it approached the cyclone (Figs. 2e,f). Recent work has highlighted the role of latent heat release in the WCB in amplifying the downstream ridge (Pomroy and Thorpe 2000; Massacand et al. 2001; Grams et al. 2011; Schemm et al. 2013; Chagnon et al. 2013). In this case, the upper-level flow above the northern edge of the WCB developed a shortwave ridge within the amplified downstream jet starting on 23 November, which was also visible at the 700-hPa level (Figs. 2a–f). After 0000 UTC 24 November, as the WCB narrowed, the upper-level low acquired a positive tilt and began to merge with the northern branch of the jet stream, and the surface low rapidly propagated northeastward (Fig. 2c). By 25 November, the cyclone center had moved over the western North Atlantic Ocean, and the WCB had weakened and thinned, no longer advecting large amounts of water vapor northward from the tropics (not pictured).
4. Analysis of A-Train products
a. 2D overview
At 1800 UTC 21 November, the surface low pressure system was still developing (Fig. 1), and MODIS cloud-top pressures depicted upper-level clouds that were primarily linear and aligned from the southwest to the northeast along the lower-tropospheric baroclinic zone (Fig. 3a). The nascent cyclonically turning WCB airstream can be seen in the high clouds observed by MODIS over the eastern Carolinas at this time as well. Examination of the surface precipitation reports over the United States, along with retrieved rainfall from AMSR-E, indicated that moderate precipitation of around 5 mm h−1 was being produced in the northwest quadrant of the storm.
Cyclonic and anticyclonic curvature in the WCB can be seen in the retrieved precipitable water at 0600 UTC 22 November (Fig. 4a); however, the bulk of the integrated cloud water and precipitation were located in the cyclonically curving portion of the flow (Figs. 4b and 4c). Cloud-top pressure from MODIS reflects the transport of water primarily northward and westward at this time, as clouds along the eastern side of the warm front were not as geographically extensive (Fig. 3b). Along the leading edge of the WCB near the approximate location of the warm front, cloud liquid water values exceeded 1.5 mm (Fig. 4b) with over 70 mm of water vapor present in the same region (Fig. 4a), and rain rates were estimated to be over 8 mm h−1 (Fig. 4c). Between 0600 and 1800 UTC 22 November, the primary cloud shield moved to the north and east of the cyclone center (Fig. 3c), consistent with the development of the low-level trough in the region at this time (Figs. 2c,f,i). There was a significant decrease in water vapor, cloud liquid water, and rain rate directly north of the cyclone center (Figs. 4e,f), as the bulk of the water vapor transport had shifted to the east and into the anticyclonically curving portion of the warm conveyor belt (Fig. 2f).
Throughout 23 November, MODIS imagery showed the primary cloud shield to be located east of the surface low (Figs. 2g,h) at the northern edge of the WCB, and it remained remarkably stationary from 1800 UTC 22 November to 1800 UTC 23 November (Figs. 3c–e). At the same time, a break in the cloud cover was observed directly north of the cyclone center (Figs. 3d,e) as the cyclonic branch of the WCB became separated from the anticyclonic branch, as discussed in the synoptic overview in section 3 (Fig. 2d). By comparing Figs. 2e and 3e, we notice that most of the cloud cover coincidentally remained concentrated over the location of the warm conveyor belt, with less cloud cover remaining over the East Coast of the United States, northwest of the cyclone center (Fig. 3e). The orbit of the A-Train, combined with the more limited spatial extent of the AMSR-E swath, led to a gap in retrievals of cloud and precipitation over the cyclone center during 23 November. The next overpass of the storm center occurred at approximately 0600 UTC 24 November, and revealed a significant change in the character of the water vapor, cloud, and rain distribution (Figs. 4g–i). The distribution of water vapor became more spatially nonuniform compared to the observations on 22 November, and the AMSR-E retrievals indicated pockets of high cloud liquid water and rain rate (Figs. 4h,i). This is an indication of a transition in the nature of the cloud structure and precipitation processes within the warm front.
b. Vertical structure
Examination of the cloud horizontal extent, integrated cloud liquid water, and rain rate provide some indication of the hydrological features within the storm. However, details of the microphysical influence on the thermodynamic structure and the vertical condensate distribution cannot be obtained using 2D and vertically integrated observations. CloudSat is capable of observing the internal structure of clouds (Stephens et al. 2002), and in this case, it intersected the WCB and the warm front multiple times as the WCB remained in nearly the same longitudinal location for a few days. The observations closest to the cyclone center were obtained at approximately 0600 and 1800 UTC 22 November and 0600 UTC 24 November. Both observations on 22 November correspond to times during which the precipitation and cloud liquid water distributions at the intersection between warm and cold fronts were relatively uniform in extent and modest in magnitude, while the cloud content and precipitation were more spatially nonuniform on 24 November, as was previously shown in the results from AMSR-E (Fig. 4).
Examination of the CloudSat profiles of reflectivity, combined with equivalent potential temperature θe computed from ECMWF temperature and water vapor interpolated to the CloudSat track, are depicted in Fig. 5. At 0600 UTC 22 November, CloudSat observed cold-frontal convective clouds south of 35°N, while to the north it intersected the warm frontal clouds located within the northern end of the WCB airstream (Fig. 5a). The θe profiles within this cross section depict the effect of evaporating precipitation at low levels: θe is nearly constant from the surface to approximately 2 km in the region just north of the surface warm front. The frontal region itself can be seen in the moist isentropes slope from near the surface at 36°N to a height of nearly 10 km near 43°N. The largely stratiform nature of the cloud in the warm frontal region is reflected in the high degree of spatial continuity in the radar reflectivity (and by extension, the cloud content) north of 36°N. Retrieved precipitation rates at this time were relatively low (less than 10 mm h−1), with a single region near 37°N where the rain rate was large enough to cause full attenuation of the CloudSat signal.
At 1800 UTC, CloudSat again intersected the northern portion of the WCB, but, in contrast to the 0600 UTC overpass, did not observe the cold-frontal convection. Stratiform cloud and precipitation in advance of the surface warm front (between 36° and 39°N) was far more extensive at this time than at 0600 UTC, and the cloud features along and above the frontal zone itself were still quite uniform (Fig. 5b). The vertical gradient of θe in the frontal zone had increased in magnitude, and the region of constant θe associated with evaporating hydrometeors was confined to the lowest 1–2 km. An increase in cloud content within the frontal zone relative to 0600 UTC can be inferred from the increase in reflectivity values, which were greater than 15 dBZe through a large portion of the frontal zone, with some values below 5000 m likely underestimated due to attenuation. Precipitation rates at this time were nearly uniformly larger than at 0600 UTC with a large portion of the frontal region producing precipitation rates large enough that the CloudSat reflectivity signal was fully attenuated at the surface.
In the 36 h between 1800 UTC 22 November and 0600 UTC 24 November, a significant change occurred in the thermodynamic and cloud structures within the warm frontal zone. We see the result of this transition in the CloudSat observations at 0600 UTC 24 November, as it again passed near the cyclone center in a trajectory very similar to the overpass at 0600 UTC 22 November. CloudSat again intersected convection along the cold front (south of 41°N), though the convection appears to be more intense than at earlier times, with larger reflectivity values and indication of full radar attenuation below 2 km above the surface (Fig. 5c). The surface warm front is still clearly evident at and north of 41°N, and the vertical gradient of θe in the lowest levels of the troposphere remains relatively large; however, the cloud fields and thermodynamic structure above the low-level frontal zone are significantly different. The reflectivity structure exhibits a high degree of variability in both the vertical and horizontal at this time, as the cloud structures have become more convective and isolated. The presence of convection is borne out in the vertical structure of the θe field, the vertical gradients of which have become much weaker above the frontal zone. In some regions, θe is nearly constant with height, indicating the presence of saturated vertical mixing. This can be seen in the shallow precipitating region between approximately 40° and 41°N, and is especially evident between 41.5° and 43.5°N at levels between 2 and 4 km. This layer is known to be characterized by strong latent heating, primarily due to the conversion between vapor and liquid (Joos and Wernli 2012). Also of note are the shallow-topped elevated-base convective features between 44° and 46°N. These extend no higher than 7 km above the surface, are upright, have bases above 2 km, and are not producing precipitation at the surface. The origin of these features is not clear, and their properties beg further study with a numerical process model. It is notable that the significant differences in cloud vertical structure seen in the CloudSat radar reflectivity are not reflected in the MODIS cloud-top pressure field (Fig. 3f).
Overall, in contrast to the reflectivity structures observed in the warm frontal region on 22 November, more individual and narrow cloud structures were observed on 24 November than deep and wide cloud structures. Precipitation retrievals indicate the signal was fully attenuated over a smaller range of latitudes, with regions containing large precipitation rates corresponding generally to regions characterized with weak vertical gradients of θe. Comparison between Figs. 5b and 5c indicates that the heavy precipitation at 1800 UTC 22 November was stratiform in nature, while precipitation at 0600 UTC 24 November was convective.
c. AMSR-E 89-GHz scattering index
Horizontal context for CloudSat’s nadir-pointing curtain observations may be obtained via the examination of the ice scattering signatures in higher-frequency AMSR-E microwave channels. Relative to clear air and cloudy regions containing primarily liquid cloud particles, ice scattering typically appears as a vertically polarized brightness temperature (TB89V) reduction due to volumetric scattering of microwave radiation by precipitation-sized frozen hydrometeors (e.g., Spencer et al. 1989; Petty 1994; Weng and Grody 2000; Bennartz and Petty 2001; Kulie et al. 2010). In addition, ice scattering introduces differences in vertically–horizontally polarized brightness temperatures (TB89V-H), as well as the 89-GHz scattering index (S89; Petty 1994). The S89 linearly combines vertical and horizontal polarization brightness temperature observations with similar values (observed or modeled) in cloud-free regions of the nearby environment. It does so by calculating the brightness temperature reduction due to ice scattering as compared to a nonscattering atmosphere that returns the same polarization difference. In theory, S89 values in nonprecipitating clouds do not typically exceed about 10 K, with significant scattering associated with S89 values larger than ~10 K (Petty 1994; Bennartz and Petty 2001). As such, elevated S89 values serve as valuable proxies for the column-integrated ice content. Elevated S89 values are also usually associated with higher surface precipitation rates; however, this relationship is complicated by variability in ice particle shape, size, and distribution that varies with precipitation type (Bennartz and Petty 2001; Bennartz and Bauer 2003).
Compared with the surrounding environment, clouds in the WCB exhibit TB89V reductions (Figs. 6a,d,g) on the order of ~30–60 K, and are closely associated with regions of surface precipitation shown earlier in Figs. 4c,4f, and 4i. Decreased TB89V values to the north of the WCB on 22 November (Figs. 6a,d), and to the north/northwest of the WCB on 24 November (Fig. 6g), are due to lower water vapor amounts and relatively clear skies, under which conditions ocean surface signatures may propagate through to the top of the atmosphere. The ocean surface signature can be clearly seen in the TB89V-H (Figs. 6b,e,h), where polarization differences larger than 50 K are indicative of the highly polarized ocean surface signal. In contrast, reduced TB89V-H values in cloudy regions correspond to significant ice cloud content and perhaps surface precipitation in the WCB and attendant fronts.
Despite the complicated relationship with surface precipitation rate, the S89 figures for the 22 and 24 November WCB event effectively illustrate the extent and evolution of the frozen cloud shield associated with the WCB, and grossly display a similar trend as is observed in the CloudSat reflectivity profiles. Maximum S89 values routinely exceed 30 K in the warm frontal region for all three overpasses, indicating significant scattering throughout the evolution of the WCB. The region over which the S89 field exceeds 10 K is rather expansive and slightly more homogeneous in the 22 November overpasses (Figs. 6c,f) compared to the narrower and more convective scene observed on the 24 November overpass (Fig. 6i). Note also the enhanced S89 values on 24 November between 40° and 45°N near the eastern U.S. coast in the region of the occlusion.
Figure 7 illustrates S89 relative frequency of occurrence (histogram) statistics for the three AMSR-E WCB overpasses. In generating this plot, we restricted our analysis to pixels for which S89 values exceeded 10 K, and with AMSR-E surface precipitation rate retrievals greater than 2 mm h−1. As previously noted, significant S89 values associated with the WCB are observed on all three AMSR-E overpasses; however, changes in S89 histograms indicate subtle ice microphysical differences (e.g., ice particle size, shape, and/or distribution) between the various WCB phases. For instance, the S89 histogram peak shifts from ~20 to ~30 K between the first (solid) and second (dashed) 22 November AMSR-E overpasses. While the peak S89 value shifts significantly, note that the first overpass contains a larger percentage of S89 occurrences exceeding ~45 K compared to the second overpass. The maximum S89 value for both of the 22 November overpasses is ~75 K. The 24 November (dash–dot) AMSR-E overpass S89 histogram peak (~20 K) is similar to the first 22 November overpass. Note, however, the markedly larger fraction of extremely high S89 values (>45 K) compared to the earlier overpass times. This tangible increase in the fraction of intensive scattering events is indicative of systematically larger ice particles associated with the more convectively active 24 November WCB phase. The systematic scattering changes depicted in Fig. 7 reflect the evolving ice microphysical composition of the cloud structures associated with the WCB.
5. Implications for satellite storm sampling frequency
Our analysis of this case study highlights several known benefits of repeated temporal sampling with multiple diverse remote sensing instruments. Specifically, observations of the warm frontal portion of the storm in approximately the same storm-relative location at multiple times revealed a distinct transition from stratiform to convective cloud structures. Collocated analysis of the thermodynamic environment provides the context for the observed cloud features, and confirms the presence of reduced static stability in regions with cloud features that appear convective. The fact that significant changes in the vertical cloud structure were observed in the CloudSat radar reflectivity and were not detectable in the passive infrared observations from MODIS, is indicative of the need for both active and passive observations of clouds in ETCs. However, repeated observations of the same storm-relative frontal position with active remote sensing instruments are rare. When CloudSat ground tracks were compared with cyclone locations in a database containing all extratropical cyclones that have occurred since the 2007 calendar year, it was found that 50% of the time CloudSat passed within 200 km of the cyclone center only once during the evolution of the storm. In 25% of the cyclone cases, there were no CloudSat overpasses within 200 km of the storm center at all. The implication is that it is difficult to use CloudSat to examine, in a systematic fashion, the time evolution of frontal clouds in an extratropical system. Such an analysis is more straightforward over land in the continental United States and Europe, where there is ample coverage from ground-based radar. However, operational radar systems are optimized for rain, and may not detect the subtle transitions in cloud structure that higher-frequency radars, like CloudSat, can observe.
The current A-Train constellation and CloudSat frequency of frontal intersect has enabled a robust series of composite analysis (Naud et al. 2010, 2012, 2014, 2015); however, composites tend to smear out some of the details observed in individual case studies like ours, and an exploration of the cloud-scale processes that occur in and around midlatitude cyclones will require measurements with a much higher temporal frequency. While instruments on board current geostationary satellites can provide frequent observations, they are limited to passive visible and infrared channels, which can only observe cloud-top properties similar to what we observed with MODIS in this case study. As noted above, visible and infrared observations are not capable of detecting changes in internal cloud structure. Sequential active microwave observations are not feasible using large platforms in polar orbit, and it is possible that a new paradigm (e.g., constellations of small satellites flying in formation) may be required. Such observing systems are currently in active development, and are being pioneered in the Cyclone Global Navigation Satellite System (CYGNSS) mission (Ruf et al. 2016). We also wish to point out that any conclusions we may draw about physical processes, even with repeated observations of the same storm-relative location, are circumstantial without quantitative estimates of latent heating rates and vertical motion. As such, it is our assertion that a critical need from future observing systems will be either direct estimates of vertical motion (i.e., from measurements of the Doppler spectrum) or indirect estimates made using high temporal frequency observations of the cloud fields themselves (e.g., rapid temporal samples of the cloud radar reflectivity).
6. Summary and conclusions
Cloud processes in extratropical cyclones are known to exert an influence on storm dynamics and frontal structure. While the overall synoptic dynamics of marine extratropical cyclones have been well understood for decades (e.g., Shapiro and Keyser 1990), recent studies have begun to explore cloud processes and mesoscale frontal evolution occurring in ETCs and their warm conveyor belts. In particular, Flaounas et al. (2016) observed two ETCs in the Mediterranean region during the HyMeX field campaign, one of which produced embedded convection that developed within the WCB. Unlike our case, the storm analyzed by Flaounas et al. (2016) exhibited convection that occurred south and west of the cyclone center. Other recent case studies from various field campaigns have examined the microphysical processes within ETCs and their impact on frontal evolution (Crosier et al. 2014; Dearden et al. 2014, 2016; Lloyd et al. 2014; Vaughan et al. 2015). Unlike our case, most of these studies have focused on cold frontal evolution, which exhibits different dynamics than a marine warm front. Using a late November 2006 case that was observed multiple times by various instruments aboard the polar-orbiting A-Train satellite constellation, we were able to detect a significant transition in the mesoscale characteristics and distributions of clouds and precipitation within the warm frontal region at the poleward end of the warm conveyor belt. This transition was clearly visible in subsequent CloudSat overpasses of the warm front, but was not visible in passive infrared imagery.
While cloud-top observations showed that the cloud shield produced by the anticyclonically curving portion of the WCB remained quite uniform throughout the WCB’s evolution, observations from AMSR-E (Figs. 4, 6, and 7) and CloudSat (Fig. 5) indicate a transition was occurring below the cloud top. The distribution of water vapor and cloud liquid water in the warm frontal zone were more spatially uniform on 22 November. Between 22 and 24 November, the cloud content became increasingly concentrated in small regions with greater intensity as the cyclone evolved. Cloud water and rain rates were no longer strongest at the warm front, but rather spread out throughout the WCB in intermittent pockets of convection. Details of the vertical structure in these convective regions were captured in the CloudSat radar reflectivity profiles and associated ECMWF data, which showed that the thermodynamic structure within the WCB had become more convective and unstable (Fig. 5c) compared to the thermal stratification observed along the leading edge of the conveyor belt on the 22 November (Figs. 5a,b). Examination of the high-frequency scattering signature from AMSR-E (Figs. 6 and 7) confirmed an increase in the highest ice scattering values over this time period.
Taken together, the results lead us to conclude that a mesoscale stratiform to convective transition occurred within this extratropical cyclone. Summarizing what we observe of the stratiform–convective transition in this case, we note the following. The transition happens later in the cyclone life cycle, and is characterized by shallow convection rooted in the warm frontal zone, with destabilization evident in the lower troposphere The 89-GHz scattering index indicates an increase in the amount of ice hydrometeors aloft in the region of the warm front and in the occluded sector, reflecting the transport of water from the lower to the upper troposphere. Taken together, the results indicate a possible increase in the lower-tropospheric latent heating maximum in the warm frontal region. As was shown in Posselt and Martin (2004), positive diabatic heating in the midtroposphere is associated with PV redistribution; creating a reduced upper-tropospheric PV anomaly and a positive anomaly in the lower troposphere (Methven 2015). The warm frontal region in which we observe the transition from stratiform to convection is displaced to the east of the cyclone center, and it is unlikely that the latent heating–induced positive PV anomaly is contributing directly to the strength of the cyclone vorticity maximum. However, we do note that the surface pressure minimum expands to the east with time while the sea level pressure minimum remains nearly stationary (Figs. 1 and 2). In this case, the influence of the stratiform–convective transition appears to be twofold; extending the low-level trough toward the north and east, and perhaps also restricting the eastward movement of the upper-tropospheric wave via creation of a reduced upper-tropospheric PV anomaly. It is also possible that the lower-tropospheric positive PV created via latent heat release may be advected into the region of the cyclone center, enhancing the cyclone circulation (e.g., Schemm and Wernli 2014). Further exploration of the specific mechanisms will require numerical simulation (e.g., as in Posselt and Martin 2004; Schemm et al. 2013; Schemm and Wernli 2014).
It is difficult to determine how or why this transition occurred without use of numerical simulation. However, we may speculate from the fact that the weak θe gradient was located just above the warm frontal surface, that destabilization was likely caused by lower-tropospheric processes rather than cold air advection at upper levels. The source regions for the water vapor transported poleward in the WCB remained consistent throughout its evolution, while relatively cold and dry continental air was ingested into the storm at low levels from the north and west. Sustained latent heating within the warm frontal region or the surface sensible and latent heat fluxes north and west of the cyclone center could have played a role in the convective transition. Previous research has shown that extreme marine cyclogenesis can occur in the winter seasons east of continental landmasses with large sea surface temperature gradients in the immediate vicinity (Sanders and Gyakum 1980). Extreme thermodynamic air–sea differences can increase the baroclinicity and instability within the boundary layer (Cione et al. 1993; Vukovich et al. 1991), leading to large surface fluxes of latent and sensible heat release (Neiman and Shapiro 1993). Given that latent and sensible surface fluxes are primarily driven by large air–sea thermodynamic differences along with strong surface winds, both of which are featured in ETCs like the one presented in this case study, we hypothesize that these surface fluxes contributed to the convective evolution of the warm front and will be the main focus in future work. It is not clear how often such a transition happens in the warm frontal region of oceanic (or continental) cyclones. We are currently conducting composite studies of extratropical frontal structure (e.g., Naud et al. 2012, 2015), but as mentioned above, composites have a tendency to obscure the details of frontal cloud horizontal and vertical structure. It is possible that recent and near-future field campaigns [e.g., the Hydrological Cycle in the Mediterranean Experiment (HyMeX; www.hymex.org) or the North Atlantic Waveguide and Downstream Impact Experiment (NAWDEX; www.nawdex.org)] may help to shed light on the prevalence of convection in warm frontal regions. Numerical modeling studies have indicated large values of latent heating in the WCB in mature oceanic cyclones (e.g., Posselt and Martin 2004; Joos and Wernli, 2012). However, given the lack of in situ or remote surface observations within this cyclone, it is not possible to conclude at this time how either surface fluxes or elevated latent heating could have affected this cyclone. Numerical modeling studies are perhaps best suited for a process-based study, and we leave such an analysis for future work.
The authors thank Dr. Catherine Naud for her assistance with the analysis of this storm and for providing feedback on and assistance with the A-Train data. Dr. Mark Kulie and Samantha Tushaus provided the 89-GHz scattering plots, assisted with interpretation of the results, and provided valuable feedback on the manuscript as a whole. The comments given by Dr. Heini Wernli and two anonymous reviews are appreciated. The research described in this manuscript was supported by NASA CloudSat/CALIPSO Science Team Grants NX10AM20G and NNX13AQ33G.