A dropsonde dataset is analyzed to quantify the necessary thermodynamic conditions for tropical cyclogenesis by evaluating the properties that distinguish developing tropical disturbances from nondeveloping disturbances, and by describing the temporal evolution of the developing inner core. The dataset consists of 2204 dropsonde observations from 12 developing disturbances and 245 from four nondeveloping disturbances. These disturbances are the cases with the best pregenesis sampling from field programs between 2005 and 2010, and include those investigated by three coincident field programs during 2010: the NASA Genesis and Rapid Intensification Processes (GRIP) and NCAR/NSF Pre-Depression Investigation of Cloud Systems in the Tropics (PREDICT) experiments, as well as NOAA’s Intensity Forecast Experiment (IFEX). Composite analyses indicate clear differences between developing and nondeveloping disturbances: developing disturbances exhibit greater moisture and a higher humidity at midlevels (above 800 hPa) than nondeveloping, and while the developing inner core experiences some midlevel moistening and stabilization as genesis nears, nondeveloping disturbances become progressively drier and more convectively unstable. Developing disturbances also exhibit some important characteristics in their inner core within 2 days of genesis: the low to midtroposphere (below 500 hPa) approaches near-saturation, a mid- to upper-level warm temperature anomaly develops and progressively deepens toward the low levels, and a low-level (below 900 hPa) cool, dry anomaly develops and is removed by the day of genesis. Overall the results support one proposed pathway to tropical cyclone formation in which an initially stronger midlevel vortex, in a moist, humid environment, precedes primarily low-level intensification within a day of genesis.
An understanding of the physical processes related to tropical cyclone formation, as well as the detection of a genesis event, has long suffered from the lack of routine observations over the ocean. In the absence of field campaigns, which provide critical in situ data, the temporal and spatial coverage of conventional observations (such as rawindsondes and IR and passive microwave satellite sensors) are insufficient to properly assess the processes occurring during tropical cyclogenesis on a wide range of scales. As a consequence, few details have been presented on the temporal evolution and vertical structure of developing tropical disturbances. Although numerical models are convenient for process-based studies of tropical cyclogenesis, because of the lack of observations in the pregenesis stage, results are often not subject to rigorous verification.
Fundamentally, tropical cyclogenesis is described as a two-stage process: 1) “large-scale formation,” in which the environment becomes primed for genesis; and 2) “inner-core formation,” in which predominantly mesoscale processes contribute to circulation intensification (McBride 1995; Karyampudi and Pierce 2002). Climatological studies (Gray 1968, 1975; McBride and Zehr 1981) have identified the necessary large-scale properties for tropical cyclogenesis to proceed: an environment containing concentrated areas of lower-tropospheric cyclonic relative vorticity and convergence, relatively high moisture content, high sea surface temperatures (>26.5°C) and relatively deep oceanic mixed layer, and low vertical wind shear (<10 m s−1 over a deep layer, 850–200 hPa). However, given the high frequency that such environmental conditions exist in the tropics, tropical cyclogenesis is not guaranteed; in fact, is a relatively rare event.
Whether genesis occurs is most often linked to how convective systems evolve within a disturbance (stage 2). One pathway to tropical cyclone formation requires increased stabilization in the low troposphere (Bister and Emanuel 1997; Raymond and Sessions 2007; Davis and Ahijevych 2012), whereby circulation is initially maximized in the middle troposphere. Consistent with thermal wind balance, the low to middle troposphere in the developing inner core is characterized by a cool temperature anomaly (cf. the environment) and the upper troposphere is characterized by a warm temperature anomaly. Hypotheses differ, however, on how an initially cold core circulation maximized at midlevels transitions to a warm core circulation maximized at low levels. Some studies suggest that on the mesoscale, low-level spinup occurs through the vertical penetration of the midlevel circulation due to merging mesoscale convective vortices (Ritchie and Holland 1997; Simpson et al. 1997). Others suggest that the low-level circulation intensifies by the downward advection of the midlevel circulation once the low levels become cooled and saturated within the sustained stratiform rain region of a mesoscale convective system (Bister and Emanuel 1997). Raymond and López Carrillo (2011) and Raymond et al. (2011) hypothesize that an initially stronger midlevel circulation in a moist, near-saturated troposphere supports a bottom-heavy mass flux profile, which subsequently favors preferentially low-level mass flux and convergence (spinup). Regardless of the hypothesis, in this pathway, observational (Raymond et al. 1998) and modeling (Bister and Emanuel 1997; Nolan 2007) evidence suggests that genesis should not be considered an instantaneous event; rather, the formation process requires “priming” or “incubation” in which humidification of the middle troposphere to “near saturation” by convective detrainment is necessary for genesis to occur. In a contrasting hypothesis, modeling (Hendricks et al. 2004; Montgomery et al. 2006) and observational studies (Reasor et al. 2005; Sippel et al. 2006) have identified the aggregation, merger, or axisymmetrization of deep, intense cyclonically rotating convective cores [i.e., vortical hot towers (VHTs)] as a pathway to amplify preexisting background vorticity and spinup a low-level circulation.
Genesis cases have received disproportionately less attention during field programs than mature tropical cyclones; however, over the past couple of decades a number of field programs have not only emphasized investigating the mature tropical cyclone structure, but also the formation of a tropical cyclone. Field programs that had some genesis-related objectives include the Tropical Experiment in Mexico in the east Pacific in 1991 (TEXMEX; Bister and Emanuel 1997), Tropical Cyclone Motion (TCM) experiments in 1992 and 1993 in the west Pacific (Harr et al. 1996; Ritchie and Holland 1997), the National Aeronautics and Space Administration (NASA) Tropical Cloud Systems and Processes (TCSP; Halverson et al. 2007) experiment in the east Pacific and Caribbean in 2005, the NASA African Monsoon Multidisciplinary Analyses experiment (NAMMA; Zipser et al. 2009) in the east Atlantic in 2006, and Tropical Cyclone Structure 2008 [(TCS-08); part of The Observing System Research and Predictability Experiment (THORPEX) Pacific Asian Regional Campaign (T-PARC)] experiment in the west Pacific (Elsberry and Harr 2008). In addition, since 2005, the National Oceanic and Atmospheric Administration (NOAA) has also utilized their two WP-3Ds and their G-IV for genesis-related research missions for the Intensity Forecast Experiment (IFEX; Rogers et al. 2006, 2013). Flights sampling the pregenesis stage in these programs, however, were often too infrequent to adequately describe the temporal evolution of the developing inner core.
Although the aforementioned field campaigns achieved varying success in sampling tropical cyclogenesis, the triagency effort during the summer of 2010 served as a culmination of all previous attempts at sampling genesis processes. NASA’s Genesis and Rapid Intensification Processes field campaign (GRIP; Braun et al. 2013), the National Center for Atmospheric Research/National Science Foundation (NCAR/NSF) Pre-Depression Investigation of Cloud Systems in the Tropics (PREDICT; Montgomery et al. 2012), and NOAA’s IFEX sought to obtain high temporal and spatial resolution observations in nondeveloping, developing, and rapidly intensifying disturbances (GRIP and IFEX only) located in the Atlantic basin and Gulf of Mexico. The unique observing strategies of PREDICT–GRIP–IFEX (PGI; triagency) involved consecutive and coordinated aircraft missions. Participating dropsonde-equipped aircraft include the NASA DC-8 (GRIP), NCAR/NSF G-V (PREDICT), NOAA P-3s and G-IV, and U.S. Air Force (USAF) C-130s. Collaborative investigations during PGI include the rapid intensification and mature stages of Hurricane Earl, the nonredevelopment of Tropical Storm Gaston, the genesis of Tropical Storm Matthew, and perhaps most impressive, the entire life cycle of Hurricane Karl starting 4 days before genesis, to rapid intensification in the Bay of Campeche and landfall near Veracruz, Mexico.
A few recent studies have presented results from PREDICT and GRIP investigations of developing and nondeveloping disturbances (Davis and Ahijevych 2012; Smith and Montgomery 2012; Wang 2012; Komaromi 2013). Smith and Montgomery (2012) and Wang (2012) analyze Karl, Matthew, Gaston for the day-to-day thermodynamic characteristics within the inner-core region, while Komaromi (2013) similarly compares thermodynamic characteristics of developing and nondeveloping disturbances by compositing G-V dropsonde data from all PREDICT flights. Davis and Ahijevych (2012) focus not only on the thermodynamic aspects of Karl, Matthew, and Gaston, but also the evolution and alignment of the low- and midlevel circulations, as well as their relationship to the vertical shear and location of convective systems. Likewise, Komaromi (2013) also compares the evolution of the composite vertical profile of radial and tangential wind for developing and nondeveloping PREDICT disturbances.
This study differs from all those just cited by using not only all available dropsonde data from the PGI cases, but also from genesis cases from several other field campaigns since 2005. For example, neither Smith and Montgomery (2012) or Komaromi (2013) use more than 269 dropsondes between Karl and Matthew, while this study uses 461. As a step toward advancing our understanding of the necessary and sufficient conditions for tropical cyclogenesis, this study analyzes an unprecedented collection of 2204 dropsonde observations from 12 developing cases and 245 from 4 nondeveloping cases. Using composite vertical profiles, the dataset is analyzed to quantify thermodynamic properties that distinguish developing from nondeveloping disturbances, and to characterize the temporal evolution of the inner core of those disturbances at all levels of the troposphere. The results will attempt to corroborate or disprove a few recent hypotheses: that the inner core of a developing disturbance progressively moistens, that the troposphere must achieve near saturation for genesis to proceed, and that stabilization (cool temperature anomaly at low levels and warm anomaly at upper levels) and an initially stronger midlevel circulation precedes low-level spinup (genesis).
Although the same PGI cases as examined by Smith and Montgomery (2012), Davis and Ahijevych (2012), Wang (2012), and Komaromi (2013) are included, this study offers an analysis based on the complete dropsonde datasets from those cases, as well as observational support from seven additional well-sampled developing disturbances. The goal of this study, therefore, is to complement their efforts by offering similar analyses based on a more robust dataset, particularly in the low to middle troposphere, by describing the variability of observed thermodynamic conditions in developing disturbances during the pregenesis stage using additional cases, and by serving as a review of the predominant hypotheses regarding tropical cyclogenesis pathways, garnered by other investigations of recently observed genesis cases.
2. Data and methodology
a. Case selection
The dropsonde dataset consists of dropsondes from developing disturbances from 2005 to 2010, and some nondeveloping disturbances from 2010 (see Table 1); all dropsondes from PGI and USAF aircraft for cases during August–September 2010 are included in the dataset (five developing and four nondeveloping). The best-sampled genesis cases from PGI are Karl and Matthew. In Karl, 29 flights were conducted among PGI and USAF aircraft; 13 (of which 12 are dropsonde equipped) prior to genesis and 16 (12) after genesis. Likewise, five flights of dropsonde-equipped aircraft investigated pre-Matthew, while two were flown after genesis. Five additional cases [Tropical Depression (TD)-2 and Bonnie (2010); Danny (2009), Kyle and Fay (2008)], in which NOAA and USAF investigated the pregenesis stage, are included. TD-2, which originated near the Yucatan Peninsula and tracked into the Gulf of Mexico, is one of the best non-PGI genesis cases included as NOAA accomplished five consecutive G-IV and P-3 flights into the pregenesis environment. Likewise, Fay (2008) is also well sampled as three NOAA P-3 flights investigated the disturbance in the day prior to, and on the day of, formation. In addition, Typhoon Nuri, sampled during TCS-08 in the west Pacific by the Naval Research Laboratory (NRL) and USAF aircraft (three pregenesis flights), and Tropical Storm Gert (two pregenesis flights by the NOAA P-3), sampled during TCSP and IFEX by NOAA and USAF in the Caribbean, are also included.
Although the majority of dropsondes in the dataset are associated with developing disturbances (total drops are 667 before development and 1537 after), some well-sampled nondeveloping disturbances (sample size of nondeveloping profiles is 245) investigated by the 2010 field campaigns are included in the dataset. PREDICT (five flights) and GRIP (two flights) successfully sampled the nondeveloping remnants of Gaston (hereafter, ex-Gaston) over the course of 3 days, providing unprecedented sampling of a null genesis case. Three other disturbances, PGI-27, -30, and -48,1 were also investigated by the G-V only and are included in the nondeveloping composites. One of the nondeveloping cases sampled by NOAA (11 August) and GRIP (17 August) in the northern Gulf of Mexico, ex-TD5, was excluded from this analysis given its unique location compared to the other nondeveloping disturbances investigated in this study, that it was located in a highly sheared region associated with a midlatitude frontal zone, and that 6 days passed between aircraft sampling.
Prior to 2010, nearly every mission by USAF aircraft was considered for inclusion in the dataset; however, whether the disturbance consisted of a single reconnaissance flight during the pregenesis stage (those flights were often the reason for upgrading the investigated disturbance to a tropical depression or tropical storm) or consisted of multiple C-130 reconnaissance flights, many flights are excluded because of insufficient spatial coverage (less than three dropsondes) and a lack of temporal continuity. On the other hand, cases in which at least one P-3 or G-IV flight investigated the disturbance prior to TD formation are included, considering that those aircraft fly at a higher altitude and provide a greater sample of dropsondes. Not all disturbances from recent field campaigns are included. For instance, another case from TCSP, Eugene, is excluded as it has since been documented that NASA ER-2 and NOAA P-3 flights did not investigate Eugene’s incipient vorticity center (Kieu and Zhang 2009). Likewise, NAMMA cases are also excluded as, in addition to the lack of temporal continuity, a clear link between the waves investigated by the DC-8 and the formation of tropical cyclones downstream has not yet been established (Zawislak and Zipser 2010). Only one NAMMA flight investigated a wave during genesis (Helene), and dropsonde data from that flight are limited. So while the current dropsonde dataset does not consist of every dropsonde that has sampled the pregenesis environment, many of the historically best-sampled genesis cases are already included.
Although genesis is often loosely defined (Montgomery and Smith 2012), in this study, genesis is defined by the TD classification by the National Hurricane Center (NHC). For each case in the dataset, dropsondes from flights after genesis are also included. Although only a few disturbances, specifically those investigated by the PGI campaigns, offer extensive sampling more than two days before genesis, in most cases critical in situ information from aircraft missions flown within a day or two of genesis were available to the NHC to help identify that formation was nearing, or had already occurred. Therefore, for the majority of cases (Earl and Gaston are exceptions), one can have more confidence in the declared genesis time, given that decisions were made from less subjective aircraft in situ data.
b. Vorticity maxima tracking
For each disturbance, the “center” is defined by the vorticity maximum (VM) manually tracked in the National Centers for Environmental Prediction (NCEP) Final (FNL) operational global model analysis. A product of the Global Forecast System (GFS), the NCEP FNL is prepared 1 h after GFS initialization, every 6 h on a 1° × 1° grid, and contains 26 mandatory pressure levels from 1000 to 10 hPa. Relative vorticity fields are spatially smoothed for tracking, using a Gaussian smoother, so as to remove spurious grid scale vorticity maxima, or “bull’s-eye” features, that are most likely not real. Likewise, unlike previous attempts at relative vorticity tracking (such as Kerns et al. 2008), which filter vorticity for a time-scale representative of an easterly wave (2–7 days), no time filtering is applied for this study. Since some disturbances investigated in this study do not appear to directly originate from an easterly wave, and since tropical cyclogenesis is often linked to mesoscale circulations produced by organized convective systems that vary at a time scale less than that of the easterly wave, time filtering has not been applied.
Using the analyses, relative vorticity maxima exceeding a threshold of 2 × 10−5 s−1 [consistent with Zawislak and Zipser (2010)] are tracked at four levels: 925, 850, 700, and 600 hPa (results will only be shown for the 850-hPa tracks). VM tracking begins when the VM exceeds the designated threshold in each 6-hourly analysis during a 24-h period, with the initial time of the tracking being the beginning of that period. Tracking ends when the VM drops below the threshold, can no longer be consistently tracked, makes landfall (anywhere but the Caribbean islands), or becomes extratropical. Consistent with the marsupial paradigm (Dunkerton et al. 2009), the pouch center is also tracked at the same pressure levels by evaluating the total wind field in the comoving (Lagrangian) framework; the comoving field is computed by subtracting the zonal phase speed of the wave from the zonal u wind field. Whereas the early literature presenting the marsupial paradigm (Dunkerton et al. 2009; Montgomery et al. 2010; Wang et al. 2010) computes wave phase speed representative of the entire lifetime of the wave, this study updates the phase speed daily. Using a Hovmöller diagram of the meridional υ wind, the phase speed is identified as the zonal progression of the wave trough axis (transition between the southerly winds on the east side and the northerly winds on the west side of the trough; “υ = 0” meridional wind line) from 0000 to 0000 UTC.
Tracking the υ = 0 line is often unreliable for waves where there is an apparent horizontal tilt to the wave trough (in other words, the trough axis is not classically oriented north–south). In these scenarios, a wave axis is difficult to deduce because winds on the western side of the trough may not turn northerly. At times when the wave axis is not well defined, the phase speed is instead calculated by using locations of the VM at each level. Waves such as those just described, or any waves that have characteristically high phase speeds (~10 m s−1), tend to not exhibit a pouch (Wang et al. 2012). For these reasons, the pouch center was not located during some periods of tracking. Given that the results for the VM and pouch (when present) do not obviously differ, only the analyses for the VM center will be shown.
c. Dataset description
Each dropsonde is interpolated onto 19 vertical pressure levels: every 25 hPa between 1000 and 900 hPa, then every 50 hPa between 900 and 200 hPa. Variables given directly from the sensors on the dropsonde include time, latitude, longitude, temperature, pressure, height, and relative humidity. Derived variables include the u- and υ-wind components (and thus, wind speed and wind direction). Other variables are computed and include the following: potential temperature θ, equivalent potential temperature θe [computed using Bolton’s (1980) formula], saturated equivalent potential temperature θes, virtual potential temperature θυ, water vapor mixing ratio w, saturated water vapor mixing ratio ws, specific humidity q, dewpoint temperature Td, and virtual temperature Tυ. All dropsonde data have been quality controlled using the Atmospheric Sounding Environment (ASPEN) software (online at http://www.eol.ucar.edu/data/software/aspen/aspen). Additional quality control has been manually applied to GRIP and PREDICT dropsondes by staff at the NCAR Earth Observational Laboratory (EOL). Manual quality control includes visually examining all profiles of temperature, relative humidity, wind speed, and vertical velocity for outliers (including those outliers as a result of interference from the transmitter antenna and erroneous aircraft initialization), adjusting data for GPS signal loss or pressure, temperature, humidity (PTU) oscillations, and identifying dropsonde profiles that may be classified as “fast falls” where the parachute fails to deploy (for fast falls, wind speed, and wind direction are set to missing).
Each dropsonde’s radial distance from the VM/pouch in the nearest analysis time at 925, 850, 700, and 600 hPa are also included. This allows the user to readily differentiate dropsondes at various radial distances from the VM/pouch center. The distribution of samples versus the radial distance from the 850-hPa VM for all dropsondes in developing and nondeveloping disturbances is illustrated in Fig. 1. In both nondeveloping and developing cases, the samples are distributed equally across all quadrants within 3°, and in developing cases, somewhat biased to the north beyond 3°. Figure 1 also shows the sample distribution for those dropsondes before and after formation and indicates that samples are well distributed around the inner core in both stages.
Only the DC-8 and G-V (93 drops, or 14% of the 667 pregenesis total) contribute to the pregenesis data sample from 2–4 days prior to genesis, with the inner core (within 3° of the VM) contributing to nearly half (46%) of the sampling. Since these aircraft fly at a higher altitude (above 10 km, or approximately 300 hPa), the upper levels (e.g., 300 hPa) and low levels (e.g., 850 hPa) are sampled equally. Within 2 days of genesis, there is an increase in the sample size (574 drops); besides the DC-8 (15% of total within 2 days of genesis) and G-V (19%), there are additional contributions from the NOAA G-IV (29%), NOAA and NRL P-3s (33%), and USAF C-130s (4%). Considering that the P-3s and C-130s fly at a lower altitude (typically below 4 km, or approximately 600 hPa), the increase in sample size is somewhat biased toward the mid- and low levels. As a consequence, in contrast to 2–4 days prior, within 2 days of genesis there are slightly more (30%, or 190 drops) samples at 850 than at 300 hPa. However, despite the fact that the mid- and low levels are somewhat preferentially sampled, the majority of drops (359; 63% of the total within 2 days of genesis) are from higher-altitude aircraft (i.e., the DC-8, G-V, and G-IV), thus upper-level sampling is also robust. Likewise, despite the addition of dropsondes from the P-3s and C-130s, where patterns tend to be focused closer to the center, within 2 days of genesis the inner core (328 drops; 57% of the total) is only somewhat preferentially sampled compared to the environment.
a. Developing and nondeveloping composite profiles
Figure 2 shows the composite θe, water vapor mixing ratio (WVMR), relative humidity (RH), and θ for developing and nondeveloping disturbances in the dataset, and also distinguishes the composite profiles for the inner core (within 3° of the VM center) and environment (3°–7°). Developing disturbances are further separated by pregenesis and postgenesis stages. Figure 2 clearly illustrates some important differences in the thermodynamic environment of developing and nondeveloping disturbances. For θ, only a small difference is observed below 600 hPa between nondeveloping disturbances and the pregenesis stage of developing disturbances; however, as one should expect, the mean θ in the postgenesis inner core is warmer at upper levels (above 600 hPa) than prior to formation. The most distinguishing differences, however, appear in the moisture variables. Although developing and nondeveloping disturbances exhibit similar moisture (and thus θe) profiles at low levels, in the inner-core composite, the pre- and postgenesis stages of developing disturbances exhibit a greater midlevel (above 700 hPa) WVMR compared to nondeveloping disturbances (by approximately 1–2 g kg−1). The composite RH also illustrates an important difference between the inner core of developing and nondeveloping disturbances; nondeveloping disturbances have a lower RH in the mid- to upper troposphere (above 700 hPa) than developing disturbances (by approximately 10%–20%). When compared to the pregenesis stage, the postgenesis WVMR within the inner core increases by up to 1 g kg−1 from the pregenesis composite, while the entire θe profile has increased approximately 5 K. In addition, the θe lapse rate is reduced after genesis, a sign of stabilization at low levels. The difference between the surface θe and midlevel θe minimum during the pregenesis stage (~15 K) and in nondeveloping cases (~20 K) corroborates the differences presented in Smith and Montgomery (2012) and Wang (2012) for developing Karl and Matthew (~15 K) and nonredeveloping Gaston (~25 K).
Although the midlevel RH in the inner core during the postgenesis stage is slightly greater (by a few percent) than during the pregenesis stage, the postgenesis stage RH in the environment is slightly less; a possible consequence of subsidence drying outside the inner core. Overall, the environment is characteristically less humid and cooler than the inner core for developing cases; in fact, the environmental properties in developing cases are fairly similar to those observed in nondeveloping cases. Not unexpectedly, nondeveloping cases exhibit the least difference between inner core and environmental properties.
Results from this composite study, as well as the one presented by Komaromi (2013), indicate clear differences between the thermodynamic conditions in developing and nondeveloping disturbances. Working under the assumption that the variables are normally distributed and that dropsondes from each flight can be treated as independent (both likely valid), a Student’s t test was performed on each variable in Fig. 2 at each level to determine if the means of the developing and nondeveloping samples are significantly different from each other. For the moisture variables (WVMR, RH, and θe), at all levels, the developing and nondeveloping samples are significantly different at the 95% level. Not surprisingly, given the small differences in the potential temperature below 600 hPa (Fig. 2), the means of the developing and nondeveloping samples are only significant at the 95% level above 600 hPa. The results from the significance testing, however, must not be considered conclusive; the sample sizes of developing and nondeveloping disturbances are too small to make any definitive conclusions. So while the nondeveloping composite WVMR and RH profiles are clearly different from the developing composite profiles at midlevels, given that only four disturbances contribute to the nondeveloping composite (which is much less than the 12 disturbances contributing to the developing composite), one must not generalize this result for all nondeveloping cases.
b. Thermodynamic evolution of developing disturbances
Figure 3 shows the time evolution of the vertical profile of θe anomaly for all genesis cases included in the dropsonde dataset. Data are composited in 1-day periods from 4 days before genesis to 4 days after. The anomaly is defined by the difference between the mean profile of the inner core (0°–3°) dropsondes for each 1-day period, and the mean profile of environmental (between 3° and 7° of the VM center) soundings from all flights. Figure 3 illustrates an important characteristic of developing tropical cyclones; within the inner core, their mean low- to midlevel (above 900 hPa) θe is higher than that of the surrounding environment and increases slightly each day, while the low levels (below 900 hPa) exhibit little increase until after formation.
Figure 4 similarly illustrates the time evolution of θ and indicates that the warm core is developing at mid- to upper levels (above 500 hPa) as many as 3 days before genesis, and extends downward with time. Given that Karl is the only disturbance sampled more than 3 days prior to genesis, the warm θ anomaly observed at low to midlevels (900–500 hPa) 3–4 days before genesis in the composite reflects that pre-Karl is initially warm core; however, within 3 days of genesis pre-Karl transitions to the cold core through the midlevels (verified independently of the composite). Within 3 days of genesis, the warming, which is statistically significant at the 90% level, is maximized between 400 and 200 hPa in the composite. This location is similar to the one identified in analyses of Karl and Matthew in Davis and Ahijevych (2012), and in the composite analysis of Komaromi (2013). The magnitude of the warm anomaly (~1.0°–1.5°C within 2 days of genesis) only differs slightly from the magnitude identified in Davis and Ahijevych (2012) (~0.5°–1°C) and that identified in Komaromi (2013) (~2°C); however, this may simply be a reflection of differing methodologies in computing the mean reference profile. For this dataset, there is little difference in the magnitude and location of the warm anomaly even when the analysis in Fig. 4 is redefined for a 1° inner core (not shown).
Given the relatively small magnitude of the θ anomaly compared to θe, the positive, increasing (albeit slowly) θe anomaly at midlevels is mostly attributed to an increase in moisture content. This conclusion is confirmed when looking at the inner-core WVMR anomaly (Fig. 5); the positive low- to midlevel (above 950 hPa) WVMR increases slightly each day (less than 0.5 g kg−1 day−1). This midlevel moistening trend, however, is only statistically significant (at the 90% level) between 850 and 700 hPa, where the greatest WVMR increase occurs within 3 days of genesis. Otherwise, above 700 hPa, the moistening trend appears to be somewhat less and is not statistically significant. When the inner core is redefined as 1° from the center (rather than 3°), there is a more noticeable increase in the inner-core θe and θ anomaly in the 2–4 days before genesis (although data are primarily for Karl). Not unlike the composite from Komaromi (2013), and consistent with an individual case study of Karl (Wang 2012), the WVMR anomaly within 1° shows a more noticeable daily increase at midlevels, closer to 1 g kg−1 day−1, prior to genesis than what is observed in the 3° analysis (0.5 g kg−1). However, is the anomaly’s daily increase a result of an increase in the inner-core moisture, or a decrease in that of the environment? A time series of the dropsonde mean midlevel WVMR (700–400 hPa; Fig. 6) indicates that the increase in the anomaly can be attributed to both a slight increase in mean midlevel WVMR within the inner core (1°) and a slight decrease at larger radii (3°–5°).
Figure 7 shows the time series of the 1° inner-core RH. Humidification appears to occur at all levels, most notably at midlevels (750–500 hPa), between 2–4 days prior to genesis; however, this result likely only applies to pre-Karl since dropsonde data during this period are primarily from that disturbance. The composite analysis for within 2 days of genesis indicates that humidification slows as genesis nears, and thus does not support a “progressive” humidification during formation; disturbances tend to already be sufficiently humid 2 days prior to genesis. Given that more disturbances are included in the composite within 2 days of formation, the result during this period is more robust; however, a significance test indicates that, from 3 days prior to genesis to 1 day prior, the only statistically significant (at the 90% level) humidification occurs between 925 and 700 hPa. The 1° composite time series (Fig. 7) also indicates that within 2 days of genesis, although the midlevels (750–500 hPa) only approaches near saturation (RH is approximately 85%–88%), the RH does achieve near saturation at low levels (below 750 hPa). These results agree in principle with Nolan (2007), who hypothesizes that the troposphere must achieve near saturation for genesis to proceed, and apparently agrees with the hypothesis of Bister and Emanuel (1997), which requires humidification of the low levels.
Figure 8 shows the composite time series of relative vorticity from the NCEP FNL within 3° of the center. To obtain the coincident NCEP FNL profile, each dropsonde is assigned the nearest-neighbor NCEP FNL analysis grid point after the dropsonde location is moved the appropriate distance for the analysis time, based on zonal and meridional phase speed of the parent wave. Consistent with the observed warm anomaly aloft (above 700 hPa) and cool anomaly at low levels (below 800 hPa), the relative vorticity is initially stronger at midlevels (750 hPa) 3 days prior to formation. Likewise, given the fact that pre-Karl (the only disturbance with samples during this period in the composite) is initially warm core 3–4 days prior to formation (θ anomaly; Fig. 4), the relative vorticity is initially maximized at low levels (below 850 hPa). Within 2 days of formation, while the relative vorticity steadily increases at all levels, the low levels (below 850 hPa) increase more rapidly until the relative vorticity maximum is located near the surface a day after genesis. The results of the NCEP FNL model analysis corroborate observational evidence from Komaromi (2013), who similarly show the time evolution of relative vorticity, except computed from PREDICT dropsondes only. In their results relative vorticity is also initially broadly maximized between 900 and 600 hPa, and within 24 h of genesis, is predominantly maximized at 800 hPa. In another study of three PGI cases (Karl, Matthew, and ex-Gaston), Davis and Ahijevych (2012) support the conclusion that developing disturbances exhibit initially faster growth in the midlevel circulation, prior to low-level circulation spinup (which predominantly occurs within a day of genesis). In their study, however, they approximate the circulation tendency with the time evolution of the observed vertical profile of tangential wind speed, computed from the dropsondes. Although depending on the perspective one may interpret this result as supporting the bottom-up or top-down pathway, this result must be interpreted more simply; the midlevels exhibit the stronger vortex prior to formation and low-level spinup is required for genesis.
Another important observation from Fig. 4 is the persistent cool anomaly (~0.5°–1.0°C) at low levels. The anomaly, which extends from the surface to 900 hPa, is a common feature for the VM and pouch, and both inner-core definitions (1° and 3°). A negative (dry) anomaly is even apparent in θe (Fig. 3; approximately 2.0–2.5 K) and WVMR (Fig. 5; approximately 0.5 g kg−1). The anomaly reaches its greatest vertical depth a day before genesis, and then gradually shallows until it is confined only to the near surface after genesis. Subsequently, one should expect increased static stability at low levels; the mean vertical profile of temperature lapse rate in the 4 days prior to genesis for the 3° inner core (not shown) confirms a slight decrease in the lapse rate in the 1000–950-hPa layer as genesis nears. A significance test of the mean θ, θe, and WVMR profiles indicates that the observed low-level cooling and drying from 2 days prior to genesis to 1 day prior is statistically significant (at the 90% level). The observed anomaly appears to support observational results from TEXMEX (Bister and Emanuel 1997), as well as numerical cloud model results of Raymond and Sessions (2007), who both identify cooling in the low troposphere as an important step in cyclogenesis. By analyzing a few of the same developing systems as this study (Karl, Matthew, ex-Gaston, and Nuri), Montgomery and Smith (2012) and Smith and Montgomery (2012) dispute the anomaly in their analyses, while Davis and Ahijevych (2012) and Komaromi (2013) confirm it. The magnitude of the cool, dry anomaly (approximately 0.5°–1.0°C, 0.5 g kg−1) corroborates the composite of Komaromi (2013); however, the vertical depth varies among the studies, which may simply be a consequence of the mean profile used. As Davis and Ahijevych (2012) note, Montgomery and Smith (2012) most likely do not detect the cool anomaly since they do not confine their analysis to the inner core (either 1° or 3°) and use a longer time-averaged reference profile.
c. Thermodynamic evolution of nondeveloping disturbances
Four nondeveloping cases are included in the dropsonde dataset: ex-Gaston, PGI-27, PGI-30, and PGI-48. Although the nondeveloping sample size is significantly less than the developing, some conclusions can still be made regarding the essential differences between the time evolutions of the inner core of developing disturbances compared to nondeveloping disturbances. Figure 6 shows the scatter of midlevel (700–400-hPa-layer average) WVMR values of each developing and nondeveloping dropsonde. Note that in Fig. 6, with the exception of ex-Gaston, nondeveloping samples are plotted as “before genesis.” Although plotted on the same abscissa as developing samples, the abscissa for PGI-27, -30, and -48 refers to the number of days prior to the conclusion of flights into each disturbance; therefore, “zero” on the abscissa is defined as the next closest-in-time FNL analysis hour after the last flight ends. Ex-Gaston, although considered nondeveloping, is plotted “after genesis” since all flights occurred after Gaston had weakened. Figure 6 indicates that the greatest difference between the midlevel WVMR in developing and nondeveloping samples is within the 1° inner core within 2 days of formation. In fact, a Student’s t test reveals that the midlevel WVMR is significantly different (at the 95% level) from nondeveloping disturbances as many as 4 days prior to formation. Even when the analysis is confined to the environment (3°–5° from the center) the midlevel WVMR (as well as 600-hPa RH; Fig. 9) in developing disturbances is still significantly different (at the 90% level) from nondeveloping samples as many as 4 days prior to genesis.
In contrast to PGI-48 (one flight), PGI dedicated at least two flights in each of ex-Gaston (eight flights), PGI-27, and PGI-30. The time series from these cases, particularly the best-sampled case (ex-Gaston), indicates that at all radii (up to 5°) the midlevel WVMR (Fig. 6) and RH (600 hPa; Fig. 9) generally decrease each day in nondeveloping disturbances. Likewise, the difference between the surface θe and the midlevel θe minimum (Fig. 10) generally shows an increase over time, an indication that the low troposphere may be becoming more convectively unstable in nondeveloping disturbances (this result may not only be interpreted as a decrease in the midlevel θe, but also as a result of an increase in the surface θe). Komaromi (2013) supports this conclusion as he shows that the composite CAPE in nondeveloping disturbances (+336 J kg−1 relative to the PREDICT mean CAPE) exceeds that in both the genesis (−171 J kg−1) and TC stages (−42 J kg−1), but cautions that additional instability does not appear to make the environment more favorable for genesis. Smith and Montgomery (2012) also found the highest CAPE in ex-Gaston (as compared to developing Karl, Matthew, and Nicole); however, they conclude that instability does not alone distinguish developing and nondeveloping disturbances.
Not surprisingly, given that similar cases are examined in this study, the composite results for the nondeveloping disturbances confirms the results presented in Komaromi (2013), and corroborates the reasons for Gaston’s nonredevelopment concluded by Davis and Ahijevych (2012), Smith and Montgomery (2012), and Wang (2012). Despite a progressive increase in the convective instability (Smith and Montgomery 2012), Gaston is apparently unable to redevelop because of consequences of progressive drying above 700 hPa and the misalignment of the low- and midtropospheric vortex (Davis and Ahijevych 2012). As stated before, although the differences between developing and nondeveloping disturbances within the inner core (0°–3°) appear to be significant for the cases examined, one should heed caution in generalizing this result since the nondeveloping cases analyzed in this study may not be representative of the variations observed across all null cases.
A compilation of results from Raymond et al. (1998), Nolan (2007), Raymond and Sessions (2007), Raymond and López Carrillo (2011), and Raymond et al. (2011) indicate how, thermodynamically, a midlevel vortex is important for tropical cyclogenesis. Initially, a stronger midlevel circulation is observed; this translates to an upper-level warm temperature anomaly over a low-level cold anomaly (Fig. 4), and thus increased stabilization in the low troposphere. Raymond et al. (2011) suggest that if the troposphere is moist and the saturation fraction (or column relative humidity—the fraction of precipitable water to saturation precipitable water) is sufficiently high (>0.80), a consequence of this thermodynamic profile is increased rainfall rates. In turn, that leads to a preferentially bottom-heavy mass flux profile and a tendency for vorticity convergence (which must overcome frictional spindown) to spin up the low-level (0–1 km) circulation. Although the authors do not explicitly relate the mass flux profile to precipitation characteristics, they suggest that the bottom-heavy mass flux profile is reminiscent of the profile expected in VHTs (Hendricks et al. 2004; Montgomery et al. 2006). Nonetheless, questions remain as to the consequence of low-tropospheric stabilization on precipitation characteristics (such as rainfall area, convective fraction, and intensity), as well as the reason for the bottom-heavy mass flux profile.
Results from the composited dropsonde dataset, however, support this pathway to genesis; an initially stronger midlevel (around 750 hPa) vortex (Fig. 7), in an environment characterized by a cold anomaly below 800 hPa and warm anomaly above 700 hPa (Fig. 4), increased static stability, and high RH (Fig. 6), precedes the low-level intensification (Fig. 7). One could further speculate that the relatively cool, dry (2.0–2.5 K according to θe anomaly, Fig. 3, and 0.5 g kg−1 WVMR anomaly, Fig. 5) air observed within the inner core at low levels 1–2 days prior to genesis suggests that latent heat fluxes from the ocean are no longer able to maintain the boundary layer against the cumulative effect of convective downdrafts. However, given the positive feedback between surface wind speed and latent heat flux from the ocean, when the low-level wind speed (and relative vorticity) intensifies the day before formation, not coincidentally, the cool, dry anomaly at low levels begins to warm and moisten, and is completely removed by the day of formation.
The reason for the low-level intensification in the 1–2 days prior to genesis has not yet been elucidated. Presumably during this period, low-level convergence replaces or overcomes competing low-level divergence expected from cold pools formed from convective downdrafts. Komaromi (2013) provides evidence for low-level divergence occurring during his 48–72-h pregenesis composite, as the radial wind is +0.5 to +1.0 m s−1 below 850 hPa. Although Raymond et al. (2011) suggest that the initiation of low-level vorticity convergence and the spinup that follows is related to convective-scale processes (VHTs), another possible explanation is that on the mesoscale, warming aloft (such as that observed in the composite θ anomaly; Fig. 4) can produce, through hydrostatic adjustment, surface pressure falls that enhance low-level convergence (Zhang and Zhu 2012). The development of the inner-core upper-level warm anomaly prior to genesis may be related to latent heating aloft (the composite θ anomaly profile is characteristically stratiform), or as Zhang and Zhu (2012) conclude, be a consequence of compensating subsidence in the convective region (Fritsch and Chappell 1980).
It is important to note that the composite profiles 1–4 days before genesis are predominantly influenced by data from the five best-sampled pregenesis cases included in the dataset: Karl, Matthew, Nuri, Fay, and TD2 (all had three or more flights prior to genesis). Although the magnitude and depth differs in each case, the θe, θ, and WVMR anomalies described in the composite time series are common among all the best-sampled cases in the dataset. The other, less-sampled cases, such as Danny, Bonnie, Gert, and Kyle, do exhibit some of these features; however, given the limited spatial and temporal sampling (in all cases, only 1–2 flights were dedicated to the pregenesis stage; all within a day of formation), no conclusion is made as to whether they support or differ from the genesis pathway exhibited by the composite. With the exception of Fay, a cool, dry anomaly is observed at low levels (below 850–900 hPa) in each of the best-sampled cases individually, while all of the cases exhibit a positive θe and WVMR (moisture) anomaly at the mid- to upper levels (above 850 hPa). Although a pregenesis warm anomaly at upper levels (above 500 hPa) is consistently observed in all cases, below 500 hPa, the θ anomaly is somewhat more variable. Some disturbances, such as TD2 and Karl, exhibit a warm core that extends to the midlevels (850–500 hPa) prior to genesis, while others, such as Matthew and Nuri, are strictly cold core at midlevels until genesis. Regardless, not one of the best-sampled disturbances examined in this study exhibits a maximum in relative vorticity below 850 hPa within 3 days of genesis.
The objective of this study was to quantify some of the necessary thermodynamic conditions for tropical cyclogenesis by evaluating the properties that distinguish developing from nondeveloping disturbances and the temporal evolution of inner-core thermodynamic characteristics. We use a comprehensive dropsonde dataset that not only includes cases investigated during the PREDICT–GRIP–IFEX (2010) field programs, but also many of the best-sampled genesis cases since 2005. The dataset comprises dropsondes from 12 developing disturbances (667 pregenesis and 1537 postgenesis samples) and 4 nondeveloping disturbances (245 samples).
Results indicate that the pregenesis stage of developing disturbances exhibit higher WVMR and RH at midlevels (above 800 hPa) than nondeveloping disturbances, particularly within 2 days of formation; the differences are less pronounced at low levels (below 800 hPa). Likewise, in contrast to the developing inner core, nondeveloping disturbances become progressively drier at midlevels, while the difference between the surface θe and midlevel θe minimum increases, an indication that the low troposphere may be becoming more convectively unstable. Despite these clear differences, however, one should caution generalizing the conclusions from the developing and nondeveloping comparison since the observed properties of the nondeveloping cases included may not be representative of the variability possible across all null cases.
Consistent with some previous observational and numerical studies of tropical cyclogenesis (Wang et al. 2008; Montgomery et al. 2010; Montgomery and Smith 2012; Smith and Montgomery 2012; Wang 2012), the developing inner core is characterized by a mean midlevel (700–400 hPa) θe (WVMR) that is higher than the surrounding environment, and that anomaly increases slightly each day before formation; within 2 days of genesis, “progressive moistening” quantitatively translates to an increase in WVMR at midlevels of less than 0.5 g kg−1 day−1. Likewise, within 2 days of genesis RH has reached near saturation (85%–90%) in the low troposphere (below 800 hPa), and approaches near saturation in the midtroposphere (800–500 hPa); further humidification is not observed. These results suggest that, at least for the best-sampled developing cases included in the dataset (Karl, Matthew, Nuri, TD2, and Fay), the inner core appears primed (high moisture content and RH) for formation as many as 2 days prior to genesis.
Finally, the composite results—and for each of the best-sampled cases individually—confirm a pathway to genesis in which, in an environment characterized by cooling at low levels (below 800 hPa) and warming at mid- to upper levels (above 700 hPa), the intensification of the low-level vortex is preceded by an initially stronger midlevel vortex (750 hPa). However, even though the results generally support the hypothesis that a midlevel vortex in an already moist, humid environment may be necessary for formation (Raymond and López Carrillo 2011; Raymond et al. 2011), we concede that other pathways to genesis likely exist and should be examined in future observational and modeling studies.
This research is funded by NASA Grant NNX09AC44G through the NASA Hurricane Science Research Program (HSRP), which is under the leadership and support of Dr. Ramesh Kakar, NASA headquarters. The dropsonde dataset would not be possible without the efforts of the GRIP, PREDICT, and IFEX science teams, as well as the U.S. Air Force Hurricane Hunters. This research is part of the lead author’s Ph.D. dissertation and has benefited from the comments from his committee: Drs. Steven Krueger, Zhaoxia Pu, Jim Steenburgh, and Jeffrey Halverson. The authors also appreciate helpful comments from two anonymous reviewers, as well as suggestions provided by Dr. Chuntao Liu and Dr. Adam Varble for improving the methodology and figures in the manuscript.
The PGI number represents the pouch number designated during the real-time tracking of pouches during the PREDICT field program (Montgomery et al. 2012).