a. Dropsondes

The global positioning system dropsonde (Hock and Franklin 1999) was designed to be dropped from high-altitude aircraft to measure pressure, temperature, winds and humidity while recording its position at up to 2 Hz. The typical fall time for the dropsonde (at a weight of 0.451 kg) from 200 hPa is roughly 13 min using a nylon parachute with an effective area of 0.0676 m². See Hock and Franklin (1999) and Wang (2005) for a detailed analysis of the dropsonde accuracy and error characteristics.

We subjectively examined all 435 BAMEX dropsonde profiles for basic quality control (soundings were complete, with absence of significant sensor wetting, deep superadiabatic layers, or erratic horizontal winds). The soundings were then classified based on structure of the low-level thermodynamics. This gave us the opportunity to evaluate the dataset before developing spatial criteria for classification.

b. Radar data

Radar reflectivity, from the Weather Surveillance Radar-1988 Doppler (WSR-88D; Crum and Alberty 1993), from the level III product suite was mosaicked using General Meteorological Package (GEMPAK; Koch et al. 1983). The algorithm for merging radar reflectivity used all base reflectivity data within the region of interest provided the analysis time and radar product time were within 5 min of each other. The radar reflectivity was interpolated to a 13-km grid by using the maximum value seen by any radar within the region of interest at the analysis time. The areas used for each intensive observing period (IOP) are illustrated in appendix B.

c. Sounding classification

Soundings were subjectively classified based on the sonde horizontal position at the 0°C level (GPS position when the temperature was 273.15 K; see appendixes A and B), then mapped onto a coarse-grid composite of level III radar reflectivity from the surrounding radars. We repeated the above step for the last GPS position or “surface” to make sure the sonde location had not appreciably changed because of sonde drift. GPS information was not present in 18 dropsondes, and an additional 19 dropsondes did not have GPS information within 2 km of the 0°C level. For dropsondes without GPS position data, the launch location was used and dropsonde drift could not be accounted for. For the other dropsondes the level that contained nonmissing latitude and longitude closest to the 0°C level was used. We also used the BAMEX radar imagery Web site (http://www.mmm.ucar.edu/imagearchive/bamex/) to verify all the dropsondes relative to radar reflectivity representation of MCSs.

The soundings were classified into 7 MCS subregions: environmental (G), leading line (F), transition region (E), stratiform rain region–center (D), stratiform rain region–gradient (C), stratiform region–rear (B), and wake (A) (shown schematically in Fig. 1; dataset characteristics of the identified MCS subregions are listed in Table B2).

The predominant MCS archetype in the dropsonde dataset was the leading-line trailing stratiform (LLTS) MCS. There were a few MCSs not resembling an LLTS MCS (Table 2). We were careful to take into account system temporal evolution of radar structures and departure from the trailing stratiform (TS) archetype. We attempted to maximize the population of dropsondes within each subregion to capture mean properties of MCS subregions. Although a few soundings per subregion may be classified differently or excluded depending on the analyst, the number of soundings in each category is large enough that switching a few soundings between categories is not expected to change the results appreciably.

The criteria for classifying the dropsondes into specific MCS subregions are as follows. We first define the leading convective line as a swath of high reflectivity greater than 40 dBZ with embedded values greater than 50 dBZ similar to the methods used to describe trailing stratiform MCSs by Parker and Johnson (2000) and references therein. We then define the stratiform rain region as a relative maximum in reflectivity (between 20 and 45 dBZ) located rearward of the convective line. The subregions are defined as follows:

• The wake region (A) included any dropsonde outside the 20-dBZ contour located in an area over which the MCS had previously passed but no more than 100 km behind the MCS.

• The stratiform rain back edge (B) included any dropsonde located to the rear of the stratiform rain reflectivity gradient region between the 20- and 30-dBZ contours.

• The stratiform rain reflectivity gradient (C) included any dropsonde located to the rear of major axis of the stratiform rain center within a region of strong reflectivity gradient.

• The stratiform rain center (D) included any dropsonde located within 10 dBZ of the major axis of the stratiform rain region and not located in a region of strong reflectivity gradient.

• The transition zone (E) included any dropsonde located in a relative minimum of radar reflectivity (lower than 35 dBZ) between the leading convective line and the stratiform rain center. Note this does not preclude the dropsonde from being located within a region of reflectivity gradient.

• The leading line (F) included any dropsonde that passed through the leading convective line aloft (revealed by saturation with respect to ice on the sounding), and any dropsonde located ahead of and within 50 km of the leading convective line. Note that no dropsonde penetrated deep convection below 6 km AGL.

• The prestorm region (G) included any dropsonde 50–250 km ahead of the MCS relative to its direction of motion and either within or not more than 50 km away from the radar swath of the MCS.

We have chosen to use seven categories based in part on the work of Smull and Augustine (1993), Stumpf et al. (1991), Schuur and Rutledge (2000), and Loehrer and Johnson (1995). The literature contains many soundings, mostly toward the rear of the MCS and a few within the transition or stratiform rain center. The environment close to the leading line of the MCS (F) is usually distinct from the prestorm environment (G) because of low-level moisture variations, fronts, or boundaries. The difference between regions B, C, and D may be less pronounced than between the other regions, but this has not been previously documented in great detail. Preferentially unique surface features such as mesoscale convective vortexes (MCVs), wake lows, and heat bursts form in these areas.

Additionally, the maximum in stratiform rain occurs in region D; it seems reasonable to investigate whether this is associated with distinct thermodynamic characteristics compared with regions B and C. The transition zone has been documented in electrification studies (Marshall and Rust 1993; Schuur and Rutledge 2000; Shepherd et al. 1996). The analysis presented here is a first step in understanding if there is any difference between the indicated MCS subregions. Statistical significance testing is complex when using correlated vertical profiles (profiles on the same day within the same MCS) of multiple variables (either derived or state). As many have found (Lucas and Zipser 2000; Jalickee and Ropelewski 1979), classification methods with soundings do not offer a sense of the statistical significance. The classifications require some validation, typically a subjective evaluation. The k-means cluster analysis approach was used early in this work but the classification is complicated by the choice of number of groups, which variables to use, how to standardize the variables, and if those variables can discriminate between the individual classes. In this context Lucas and Zipser (2000) found that this type of approach could distinguish between disturbed and undisturbed tropical soundings, though vertical wind profiles required a separate classification. Since there is not a well-established method for statistical significance testing in this context, we leave this aspect for future study.

d. Sounding composite construction

Sounding composites were constructed to preserve low-level thermodynamic features by using the closest observation to the ground and 0°C level as reference levels. The 0°C level was used since this was typically the level at which subsaturation occurred relative to the cloud layer aloft, which was saturated with respect to ice (Leary 1980). The vertical coordinate then ranged from zero (presumed ground; 0 km) to 1 (the 0°C level; 3.8 km) and to 3 (anvil layer; 11.4 km). State variables that were composited include potential temperature, water vapor mixing ratio, equivalent potential temperature, u, v, and calculated w wind components. Each sounding was linearly interpolated to conform to this vertical coordinate with a spacing of roughly 35 m. Once each sounding was interpolated to the new coordinate, averages for each MCS region were performed to arrive at composite vertical profiles. We then calculated standard deviations of the variables relative to the mean at each level.

e. Vertical velocity calculation

We obtained a fall speed profile by averaging vertical velocity over all profiles using an air-density vertical coordinate, motivated by the recognition that air density is directly related to aerodynamic drag. The sounding point values of fall speed were then binned and averaged to arrive at a fall speed–density relationship and a sixth-order polynomial was fit to the data (Fig. 2). Different bin widths and quality control thresholds (rejection of any point value more than 3σ above or below the mean) were applied and yielded very little change (less than 1% change at any level) to the estimate.

Next we examined individual profiles of vertical velocity. Thirty-eight soundings had large negative vertical velocity throughout the profile because the dropsonde parachute did not deploy correctly, was not functioning, or was no longer attached to the dropsonde (T. Hock 2005, personal communication). These profiles were removed from the composite vertical velocity profiles.

We repeated the same procedure for the Rain in Cumulus over the Ocean Experiment (RICO; Rauber et al. 2007) dropsonde data as a reference. The fall speeds of the dropsondes for the RICO and BAMEX datasets are similar (Fig. 2). The difference (Fig. 3) between the RICO and BAMEX calculation averages about 0.35 m s⁻¹ with the largest differences in the RICO cloud layer between 800 and 900 hPa (density of 0.9–1.05 kg m⁻³).

Confidence intervals (z; not shown) were computed at the 90% level:

View Expanded

where c is 1.645, σ is the standard deviation in each bin, and n is the number of samples per bin. Confidence intervals generally were below 0.12 m s⁻¹. The confidence intervals for the RICO dataset were generally lower than 0.06 m s⁻¹. We believe that the difference between the RICO and BAMEX datasets is the result of BAMEX dropsondes being flown in active MCS precipitation while RICO dropsondes were mostly flown outside shallow cumulus clouds with less variability of vertical velocity, weaker wind fields, and thus weaker turbulent motions. Documented vertical velocities (e.g., see Knupp 1985 Table 2) in strong precipitating downdraft cores can vary between 1 and 10 m s⁻¹, while updrafts in strong convection can vary between 2 and 60 m s⁻¹. In shallow convection (of the type observed in RICO) downdrafts vary between 0.5 and 6 m s⁻¹, while updrafts vary between 0.5 and 10 m s⁻¹. The difference in standard deviation between the two datasets is 0.5–1 m s⁻¹ (Fig. 4). The jump in standard deviation near 0.8–0.9 kg m⁻³ reflects the enhanced downdraft variability below the 0°C level.

We obtained further confidence that the signal extracted was indeed the air vertical velocity by examining the composited environmental soundings (i.e., outside of deep moist convection). The nonzero mean vertical velocity suggests an error of 0.1–0.2 m s⁻¹ (Fig. 5). The nonzero mean was partly due to a balance between (i) a few soundings being located in and around baroclinic zones with relatively deep ascent (0.10 to 0.25 m s⁻¹) and (ii) many soundings containing shallow low-level subsidence layers (−0.05 to −0.1 m s⁻¹). Our vertical velocity profiles in the vicinity of several BAMEX mesoscale convective vortices are qualitatively consistent with those obtained by Trier and Davis (2007) using different methods.

We then compared the relative frequency distribution of downdraft vertical velocity from the dropsondes (not shown) with that obtained from the National Oceanic and Atmospheric Administration (NOAA) P3 aircraft (Jorgensen 1984) during microphysical spirals (not shown). The data were not taken at the same spatiotemporal locations but are located in the stratiform rain region exclusively. Thus we cannot relate a specific dropsonde to a specific spiral. Rather we use these data to check the magnitude and vertical location of the downdrafts sampled. Both datasets show (i) an absolute maximum in downdraft frequency at a vertical velocity of ∼−0.5 m s⁻¹ extending from 1.25 to 6 km and (ii) a relative local maximum in frequency between −3 and −2 m s⁻¹ near 4 km. Thus it appears that the two datasets are very similar, which adds credibility to our technique.

f. State and derived variable statistics

The soundings were used to extract subregion-specific quantities such as maximum convective available potential energy and corresponding convective inhibition; 0–6-km bulk shear; lapse rates from 0–0.2, 0–0.3, and 0–0.5 km and the corresponding mean relative humidity for each layer; maximum dewpoint depression in the midlevel “mixed layer” along with lapse rate from the 0°C level to the inversion top or level of maximum temperature; mean relative humidity and virtual potential temperature from the 0°C level to the ground; and the lapse rate across an inversion layer.

a. Composite MCS

We constructed a composite cross section (Fig. 6) to elucidate the horizontal structure the soundings represent. The relative humidity field clearly shows the stratiform rain region and its rearward upward slope. Below the stratiform region the equivalent potential temperature is close to uniform in the vertical starting from the transition region (E) and extending rearward to the stratiform rain gradient region (C). Curiously there is a second region where the equivalent potential temperature field is almost uniform in the vertical (B), being deeper but having lower equivalent potential temperature than the structure seen below the stratiform rain region.

We computed the wind component relative to the average system motion by subtracting the average system motion of the leading line from the ground-relative wind components. This method should give us the general sense of the flow field but not the specific storm-relative wind field since we did not account for convective line geometry (thus we are unable to use terms such as rear inflow). Positive storm-relative u wind in the plane of the cross section extends along the gradient of the RH field (Fig. 6b; close to 625 hPa) and extends just to the rear of the stratiform rain region. The environment ahead of the MCS exhibits a region of high relative humidity between 900 and 800 hPa within the large vertical gradient of equivalent potential temperature that ascends into the composite MCS. This structure is consistent with the finding that MCSs tend to be uncoupled from the surface layer and feed on a slightly elevated region of instability (Trier et al. 2006).

b. Transition and stratiform regions

The composite soundings of the transition zone and stratiform rain region center are shown in Figs. 7 and 8, respectively. The soundings were converted from the height-based vertical coordinate to the skew T–log p diagram by averaging pressure, preserving the height-based vertical coordinate. The mean vertical velocity is downward from above the 0°C level to the surface in the transition zone and downward from the 0°C level to the surface in the stratiform rain region. The transition zone vertical motion differs from that of Biggerstaff and Houze (1993) probably because of the nature of the sampling volume of each respective measurement. Horizontal variability is negated within the radar sampling volume, though measures such as spectrum width can reveal just how turbulent the volume is, while the dropsonde falls through all or portions of up/downdrafts or both within an equal volume to that of the radar.

The lack of a composite isothermal melting layer can be attributed to the shallowness of an isothermal melting layer in individual profiles, the presence of dry adiabatic layers across the 0°C isotherm, and isothermal melting layers that do not occur at exactly 0°C. Storm and Parker (2007) identified a sounding during BAMEX in which sublimation accompanied by a downdraft may have caused the lack of isothermal layer near the 0°C level. Parker and Johnson (2004) noted a similar mechanism in numerical simulations of leading stratiform MCSs. Braun and Houze (1995) asked what impact melting layers have on both large- and small-scale features within a large number of MCSs. We note that these melting layers may be present to a lesser degree in the stratiform rain region, but McFarquhar et al. (2007) also observed a relative lack of isothermal melting layers, instead finding a lapse-rate transition near the 0°C level.

Biggerstaff and Houze (1993) found that the transition region contained deep descent with two relative minima in vertical velocity: one near the 0°C level and another aloft. It is well documented that soundings within the stratiform rain region are mostly dry below the 0°C level, since subsaturated downdrafts are present (Zipser 1977). We found that most soundings were subsaturated below the 0°C level, though some were close to saturated (defined as the mean relative humidity between the surface and 0°C level exceeding 87%).

We identified 6 saturated profiles out of 34 in the transition region (Table 3). The unsaturated profiles were similar to those documented previously in the transition zone (e.g., Schuur et al. 1991; Marshall and Rust 1993; Shepherd et al. 1996). It became apparent that proximity to the convective line, although important, was not the primary factor affecting the mean RH. As Braun and Houze (1994) point out, it is the combination of microphysics (aggregation, size sorting, collection and growth of precipitation particles) and relative flow that determines the difference in radar reflectivity between the transition and stratiform rain regions. The significance of the saturated profiles is unknown but may be related to the evolution of the transition region.

Thermodynamic variability within the transition region was greatest below the 0°C level for all soundings, with marginal variability in the anvil. Unexpected representations of the thermodynamic variability include multiple, nearly well-mixed layers in the lowest 150 hPa, near dry adiabatic surface layers topped by elevated stable layers, and surface-based stable layers of varying stability.

In general, the stratiform rain region and the transition region are similar. Both include surface-based stable layers and contain deep dry adiabatic lapse rates below the 0°C layer. However, a number of soundings contained multiple distinct layers below the 0°C level, often bounded by inversions (Figs. 9 and 10). These layers of differing wet-bulb potential temperature imply that air in this column has different origins. The lower inversion appears to be distinctly different thermodynamically compared to the surface and anvil (wet-bulb potential temperature difference of 1–2 K compared to the “onion-type” structure observed above). The thermodynamic characteristics of this low-level inversion correspond to the properties of the low-level inflow environment (either precipitation evaporation or forced descent in “up–down” downdrafts). W. C. Straka et al. (2007, personal communication) and Wakimoto et al. (2006) show that “up–down” downdrafts (Knupp 1987) were present in 3 case studies from BAMEX. The thermodynamic structures, observed in multiple soundings from these particular case study days during or just after the time of their analysis, were consistent with “up–down” downdrafts.

In these two soundings strong downdrafts were present in the deep dry adiabatic layer directly below the 0°C level. The moisture layers correspond to changes in the vertical velocity. The 10 June sounding (Fig. 9) also contains a strong downdraft at the top of the elevated inversion (850 hPa) with an updraft at the inversion base, potentially implying that the layer is mixing turbulently downward. The mixing ratio profile has a stair-step appearance across this interface, implying a transition between the reduced mixing ratio above and the enhanced mixing ratio below. Soundings taken later during this MCS show the two inversions eventually merge to produce a strong surface-based inversion (not shown). The 24 June sounding (Fig. 10) depicts a transition in vertical velocity near 800 hPa across the inversion and also across the 0°C level. This sounding also contains a melting layer near 0°C that is not isothermal. We make this assumption because below this layer the vertical velocity decreases to −2 m s⁻¹ throughout the layer below. Note that the lapse rate through a deep layer is 7.2 K km⁻¹ compared to 8 K km⁻¹ for Fig. 9. We will return to this point in section 4b.

a. Low-level thermodynamics

Wet-bulb potential temperature is approximately conserved and can be used as a thermodynamic tracer for air parcels that begin in the boundary layer and ascend dry and then moist adiabatically (Bosart and Nielsen 1993). The wet-bulb potential temperature difference between the 0°C level and surface averages about −2 to −4 K (Fig. 11). The convective-scale downdrafts modify the surface layer strongly when the gust front passes. Note the shift between the environment (G) and transition zone (E). Despite the rather large variability given by the outliers,¹ the 25–75th percentile ranges are small. The wide outliers may be due to the presence of up–down downdrafts, indicating that the origin of the surface layer air may be the result of mixing from multiple airstreams.

Given the linkage between the 0°C level and surface wet-bulb potential temperature, we explore the low-level lapse rates that characterize the surface layer. Cold pools usually are identified by negative buoyancy relative to the environment (Bryan et al. 2005). Wakimoto (1982) showed examples where the low-level sounding was cooled after thunderstorm passage in Florida and the soundings lacked a low-level isothermal or inversion layer. This prompted us to evaluate the low-level vertical temperature gradient. We examined the lowest 300 and 500 m vertical temperature gradient in each sounding within the MCS.

The vertical temperature gradient in the 300-m surface layer (Fig. 12a) indicates that the surface layer was statically stable especially rearward of the stratiform center. However, note the variability from statically stable to moist adiabatic with some outliers reaching superadiabatic. While this result seemingly defies explanation in the bulk sense, it suggests that the boundary layer is quite complex under typical MCS conditions, possibly affected by the vertical gradient of humidity, rainfall (thus the vertical profile of evaporation) and mixing via outflow circulations. The vertical temperature gradient in the lowest 500 m (Fig. 12b) is similar to that in the lowest 300 m except that the stability (positive vertical temperature gradient) has decreased. In other words, above the lowest 300 m there is a thermodynamic transition from stable to less stable conditions. This implies that the stable layers are typically confined to the lowest 300 m of the sounding when present. This is consistent with Houze (1977) and Johnson and Nicholls (1983) who found that stable layer tops occurred near 300 m within oceanic MCS wakes.

The rapid changes aloft in temperature and moisture caused by the mesoscale downdraft can act to increase the stability of the surface layer by warming aloft, but this impact is felt only toward the rear of the MCS. Within the transition and stratiform regions, the surface layer is controlled by mixing via internal circulations, convective-scale downdrafts, mesoscale descent and/or up–down downdrafts. Thus an area of future research is to examine what controls the properties of the lowest 0.5 km under MCS conditions.

b. Unsaturated downdraft characteristics

A consistent feature of MCSs analyzed here was the presence of a nearly dry adiabatic layer below the 0°C level. This layer was typically not well mixed as the mixing ratio was not constant. We calculated the vertical temperature gradient (Fig. 13) and mean RH (Fig. 14) from the 0°C level to the level of maximum saturated equivalent potential temperature. The median values of the vertical temperature gradient (indicated by the horizontal line within each bar of the box plot) are similar with the exception of the stratiform rain region center (D) and transition zone (E). Note that regions A and B are very similar, although the small sample size of B limits the robustness of this result. The stratiform rain center (D) has the lowest median lapse rates and is comparable to the transition zone. It appears the transition zone (E) has the most variability within the MCS. The stratiform rain region center (D) is slightly more variable in moisture but otherwise similar to the transition region. We infer that the transition and stratiform center regions are composed of mesoscale unsaturated downdrafts in the mean. In the sounding examples shown previously we noted layers of descent. The dropsondes have sampled a temporally varying structure and thus it is possible that the only regions that are not affected by the temporal variation of strong descent are the environment and wake regions. The variability of the vertical temperature gradient (given by the 25–75th percentiles) is lower in the wake region (A) than in the other MCS regions.

Typically the anvil exhibits saturation with respect to ice. Below the anvil, in the trailing stratiform rain region there is clear evidence of subsaturated mesoscale downdrafts. The bottom of the anvil can be marked by an isothermal 0°C layer containing melting small ice particles. McFarquhar et al. (2007) found that isothermal 0°C layers were not always present in observed MCSs but that there was at least a lapse-rate change across the 0°C level. The isothermal layer occurs when ice particles melt and has been shown to be present in the trailing stratiform regions of MCSs (Willis and Heymsfield 1989). The vertical location and depth of the 0°C layer found here are consistent with previous studies (e.g., Willis and Heymsfield 1989). The lack of “deep” isothermal layers between −4° and 3°C (Table 4) is not surprising in the environment or the wake of MCSs, but only 34% of the stratiform rain maximum region contained one such layer. This result changes very little by reducing the RH threshold. Possible explanations for this result include the following:

1. deep isothermal layers were not common or sampled;

2. ice was not melting (little evaporational cooling) and was being sublimated because of wide particle size distribution, or low relative humidity (McFarquhar et al. 2007);

3. there was some type of temperature sensor error (e.g., lack of response in near-saturated conditions or the dropsonde fell too quickly through the layer);

4. strong downdrafts transported ice rapidly through the 0°C layer so the isothermal layer could not form but ice could still melt; or

5. melting was occurring but no isothermal layer was present (McFarquhar et al. 2007).

The lack of a consistent isothermal melting layer could be confirmed if radar reflectivity data lack a horizontally homogeneous bright band in the MCSs assuming good coverage of each MCS examined here. The discontinuous structure (either in horizontal extent or depth) would be an indication of microphysical variability on the scale of the radar data. Such an extensive analysis of radar data is beyond the scope of this paper.

The lack of distinct isothermal 0°C layers in the stratiform region of the MCS has implications for the rear inflow jet. Weisman (1992) argued that horizontal buoyancy gradients determine whether rear inflow descends or ascends. He noted that inclusion of ice processes would increase the depth of the cold pool and modify the buoyancy field. It is not clear, however, how the buoyancy gradients would be modified. Yang and Houze (1995) have shown that ice microphysics has a qualitative and quantitative impact on MCS simulations. They found that nearly 25% of the intensity of the rear inflow jet was due to melting. Furthermore, the latent cooling due to melting did not initiate the mesoscale downdrafts but rather enhanced these downdrafts. Thus microphysical effects modulate the rear inflow jet.

c. Vertical distribution of wet-bulb potential temperature and RH

Contoured frequency by altitude diagrams (CFAD; Yuter and Houze 1995) depict the relative frequency distribution of a variable with height. Yuter and Houze (1995) used CFAD analysis to show the growth of stratiform precipitation in Florida cumulonimbus in radar reflectivity data. Here we construct CFADs using the dropsonde data to interpret the vertical structure variability within the MCS. In this section we combined dropsonde data in regions B–E.

The relative humidity profiles (Fig. 15) for both up- and downdrafts within the MCS have an approximate slope of 4% km⁻¹ from approximately 1 to 2.25 h, where h is the 0°C level. The model proposed by Leary (1980) yields a thermodynamic structure dependent on vertical velocity, initial RH, rain drop size distribution, and rainfall rates. However, their model does not reproduce the moisture structure of the observed soundings and the RH slope varies between 7% and 14% km⁻¹. The difference between our data and Leary (1980) may be the increased rainfall rates within MCSs during BAMEX. The environmental profile of RH has a slope of 3% km⁻¹, which represents the elevated mixed layer.

An interesting feature of the downdraft RH CFAD is the frequency extension located at and below the 0°C level (indicated by the horizontal line in Fig. 15a) between 85% and 95% RH. This feature likely represents evaporation into the unsaturated downdraft and was seen in a number of soundings as increasing mixing ratio in a nearly dry adiabatic temperature profile. As precipitation particles sublimate and/or evaporate the local mixing ratio increases and cooling leads to subsidence. Just below the 0°C level, the adiabatic warming does not balance the evaporative cooling rate, but as the air descends a balance is reached between adiabatic warming and cooling because of evaporation/sublimation. This imbalance should yield a nearly constant mixing ratio or decreasing RH. While this is the case on average, most dry adiabatic layers in downdrafts have a different mixing ratio profile, implying that microphysical variations (particle size distribution, number concentration; Srivastava 1987) likely play a major role in local vertical profiles.

The previous analysis of wet-bulb potential temperature showed similarities between the surface and 0°C level and anvil region. The CFAD for θ_w (Fig. 16) is relatively uniform, having a primary peak that decreases with increasing height. We can see that the surface and anvil (H greater than 2) θ_w are equal (i.e., the frequency peak is at a difference of 0°C) with differences occurring up to and above the 0°C level (i.e., two different thermodynamic regions). These apparently separate regions are consistent with Zipser (1977), who hypothesized that air from the environment was trapped and isolated within developing convection. It was later shown by Knupp (1987, 1988) that up–down drafts may be responsible for creating the unique thermodynamic properties of the air in the surface layer. In order for the low-level air to be comprised of nonunique θ_w air, the origination of parcels could be from ahead of the MCS (entrained into the gust front), from up–down downdrafts or from the stratiform region of the MCS. Zipser (1977) speculated that air from ahead of an MCS was found inside because of discrete propagation of convective elements. G. H. Bryan (2007, personal communication) has recently begun to explore the parcel trajectories of the air ahead of and within an idealized MCS in a fine-grid simulation.