a. Dryline definition for this study

It is useful to define what is meant by “dryline” within the context of this study. It will become obvious in the material that follows that the dryline is not in every case simply a single narrow band of large moisture gradient. There exists in some cases a dryline zone that may contain more than one band of large moisture gradient, and the magnitudes of moisture gradients and characteristics of other variable fields (e.g., wind) may vary from one band to another. The term dryline will be used to denote what is perceived to be the band of moisture gradient possessing the largest gradient contained within the zone. This band will usually (but not always) be the easternmost one within the zone. The dynamical justification invoked here for this easternmost location is that the motion of the dryline is largely controlled by vertical mixing processes, and that no other process is evident for establishing a strong gradient at the easternmost location. Other gradients farther to the west may represent either a former location of the dryline where discontinuous displacement has taken place (as detailed in Part I), or the result of intermediate vertical circulations that tend to concentrate gradients. From the analysis found in Part I it is clear that the vertical mixing process is quite complex and not well understood, but appears to occur over preferred horizontal scales within the zone rather than continuously in time and space.

Since the dryline by any definition is identified by gradients in moisture, it may or may not be characterized by convergence in the wind field. For identification of the dryline in this study surface analysis of dewpoint temperature is employed. In regions where surface stations are widely spaced, aircraft, radar, or visible satellite data may be used to locate more precisely the dryline within a region where it is known to exist based on surface moisture measurements. Since radar and visible satellite systems do not sense moisture variables, their use for identification generally assumes a coincidence between water vapor gradients and convergence in the wind field.

b. Observing systems employed

A key observing system in this analysis was the National Weather Service (NWS) Weather Surveillance Radar 1988-Doppler (WSR-88D) located at Twin Lakes (KTLX), about 20 km east-southeast of central Oklahoma City. This radar was one of the few WSR-88D network radars operating in the United States at this time. It was very fortunate, therefore, that the dryline was located within range of this radar, permitting observation of boundary layer clear air return associated with the dryline and other features. Clear air return was present within approximately 100 km of the radar on this day. In these very early months of radar operation, archival procedures were not routine. For this study the importance of the archival issue is that no data are available in the very early afternoon (the first image is available at 1935 UTC [1335 central standard time (CST)] and archived data begin at 2027 UTC). The radar data were also used to determine the location and early evolution of convective precipitation.

Another observing system vital for this analysis included the National Oceanic and Atmospheric Administration (NOAA) P-3 research aircraft, which flew through mesoscale features collecting in situ observations. The combination of ground-based KTLX Doppler observations and airborne in situ data collection focusing on the same mesoscale features provided unique mesoscale insights that could not have been obtained by either system alone. Satellite observations also contributed to the analysis in several ways. As cumulus clouds developed along much of the dryline, location and motion of the dryline could be determined from visible images where coverage from other systems was absent. The spatial relation of clear air return to visible clouds could be determined from the combination of radar and satellite observations. Satellite infrared measurements allowed estimates of surface temperature, which influences stability and therefore both storm initiation and dryline motion (through vertical mixing). Other important sensors included a pair of the National Severe Storm Laboratory's mobile laboratories equipped with Cross-Chain Loran Atmospheric Sounding System (CLASS) ballooning capability (Rust et al. 1990), and a network of National Center for Atmospheric Research (NCAR) Portable Automated Mesonetwork (PAM) stations (Brock et al. 1986).

c. Previous research on convective initiation at boundaries

Discussions of previous work relating to dryline motion and to dryline redevelopment were included in Part I and will not be repeated here. Many studies have included treatments of convective initiation along boundaries in general. Wilson and Shreiber (1986) documented in a Colorado climatology that 79% of thunderstorms initiated within their study area occurred near convergence boundaries and that boundary intersections were preferred locations for convective initiation. Kessinger and Mueller (1991) showed preferential convective development occurred in a Florida study at the intersection of an outflow boundary and horizontal convective rolls. Crook et al. (1991) carried out numerical experiments that showed vorticity centers and accompanying updrafts developed owing to the intersection of boundary layer thermal instabilities in convergence zones in the lee of topographic obstacles. Wilson et al. (1992) used radar to document enhanced updrafts leading to convective cloud formation at the intersection of convergence lines and horizontal convective rolls in Colorado. In another Florida study Wakimoto and Atkins (1994) observed the initiation of cumulus congestus at the intersection of the sea-breeze front with horizontal convective rolls. Atkins et al. (1995) documented for Florida cases the development of enhanced updrafts and convective clouds at the intersection of sea-breeze fronts and horizontal convective rolls for both onshore and offshore flow regimes.

A number of studies have specifically focused on initiation of convective clouds or thunderstorms along drylines. McCarthy and Koch (1982) attributed enhanced moisture convergence at particular locations along the dryline to perturbations associated with gravity waves whose source was in the dry air. Bluestein et al. (1988) concluded that localized differences in low-level diabatic heating might have produced pressure falls in the dry air that influenced storm development for a dryline storm near Canadian, Texas. In another case (Bluestein et al. 1990) cumulus congestus and a small cumulonimbus formed near the intersection of an outflow boundary and the dryline owing to locally enhanced convergence and local deepening of the moist layer. Sanders and Blanchard (1993) found that localized lifting of a capping inversion as a result of shearing instability-induced waves allowed for storm formation along a small sector of a western Kansas dryline. Differences in landscape and land use were implicated by Hane et al. (1997) in the formation of a cloud line in the dry air whose intersection with the dryline produced enhanced convergence leading to development of tornadic storms in northwest Oklahoma. Atkins et al. (1998) observed reflectivity maxima and (in some cases) cloud formation at the intersections of horizontal convective rolls with the dryline in a Texas panhandle case. In an investigation of thunderstorm initiation Weiss (2000) found that storms formed in a region of enhanced vertical motion near the intersection of the dryline with an outflow boundary.

d. Clear air return

Clear air return from weather radars has been attributed to scattering from index of refraction gradients and from insects or birds. Early studies resulted in a division of opinion between the index of refraction source (e.g., Atlas 1960) and the particulate source (e.g., Crawford 1949). More recent studies seem to point toward insects as the cause (Hardy and Katz 1969; Wilson et al. 1994). In a study utilizing dually polarized Doppler radar (Achtemeier 1991), clear air return was observed in association with a gust front. Since the return did not extend above a certain level in upward moving air, it was hypothesized that insects were falling or flying downward to avoid cooler temperatures at higher altitudes. The dual-polarization measurements obtained in this case were consistent with this hypothesis, which would also explain the persistence of lines of clear air return. Recent studies of horizontal convective rolls by radar (Christian and Wakimoto 1989) and of clear air return in a variety of environments (Wilson et al. 1994) have concluded that clear air echoes in the mixed boundary layer result from scattering from insects, while return just above the mixed layer is due to refractive index gradients. In this study, which focuses on boundary layer processes, the explanation involving clear air return owing to insects is assumed. Quasi-linear features known as “thinlines” appear in the clear air radar return in a number of the figures that follow; moreover, thinlines are routinely used both in this work and elsewhere (by forecasters and researchers) to identify mesoscale boundaries.

e. Observation of drylines by radar

A number of investigations have included radar observations that associate thinlines with the dryline. Installation of the WSR-88D network, an increased density of radars in dryline-prone areas, and increased sensitivity of those radars, have resulted in more frequent observation of thinlines identified with the dryline. Fujita (1970), in an investigation of a tornadic storm near Lubbock, Texas, identified a dryline in the Texas panhandle using thinline observations from the Amarillo WSR-57 radar. At that time drylines were referred to as “dry fronts,” and since this one was observed to move northwestward (at about 9 m s⁻¹) toward dry air, Fujita referred to it as a “moist front.” In other investigations of severe storms in the Texas panhandle, Bluestein et al. (1988, 1989) identified westward moving drylines from thinlines observed by the Weather Surveillance Radar-1957 (WSR-57) at Amarillo. More recently, Wakimoto et al. (1996) utilized observations from the NCAR Electra Doppler Radar (ELDORA) to investigate the motion field in the dryline environment; identification of the dryline in that case also was based upon thinline observations. Atkins et al. (1998) in a study of finescale dryline structure utilized both airborne and ground-based radar to identify thinlines associated with the dryline. In that case, in addition to the prominent thinline associated with the dryline, other thinlines intersecting the dryline at varying angles were found to be associated with horizontal convective rolls; moreover, clouds were found in some cases to develop in enhanced reflectivity areas at the dryline–roll intersections.

f. Issues to be examined

In this paper the following subjects will be addressed: (a) the mesoscale dryline environment will be examined in detail within the area where eastward dryline redevelopment took place, (b) the nature of thinlines at the dryline and west of the dryline observed by the KTLX radar will be examined through analysis of in situ aircraft data, (c) the relation of clear air return observed by radar to convective clouds observed by satellite will be investigated to better understand initiation of clouds at the dryline, and (d) the question of why severe convective storms developed along the dryline in two particular areas, but not in other areas, will be addressed. These issues are examined in order to provide a better understanding of dryline convective initiation processes so that disruptive weather associated with dryline storms can be better forecast over short time periods.

a. Early afternoon dryline environment

During the day on 16 May 1991 a dryline moved generally eastward across portions of Kansas, Oklahoma, and Texas, under the influence of a translating synoptic disturbance. In Part I the detailed motion of the dryline throughout the day was described. By early afternoon the dryline had advanced to central Oklahoma where it exhibited a synoptic-scale eastward bulge (Tegtmeier 1974). As shown in Fig. 1a (1900 UTC), the dewpoint change in the immediate vicinity of the dryline was about 10°C, and total change across the region was about 20°C. The dryline was located within a diffuse trough of low pressure, with winds generally from the southwest to the west of it. To the east of the dryline winds were from the south-southeast south of the bulge and more backed north of it. By 1 h later (Fig. 1b) the dryline in the northern portion of the domain had advanced steadily northeastward, as evidenced by the changes at Dodge City (near the dryline–frontal intersection). In the southern part of the region (northern Texas) the dryline had progressed slowly eastward. The apparent very rapid eastward motion in central Oklahoma was shown in Part I to be at least partially a result of a decrease in dryline characteristics in one location and development in another location farther east.

The P-3 research aircraft flew a series of long legs at approximately 930 hPa during the 1920–2040 UTC period over a portion of the area in which the eastward redevelopment took place. In Fig. 2, an analysis of the dewpoint field is shown based on the aircraft data. Also shown are flight-level winds along the track during this period. Since the total period associated with this track segment is 80 min, there is some distortion of the orientation of areas of large moisture gradient. For example, areas to the north would be farther west relative to other features if observations were simultaneous. Additionally, there is uncertainty in the analysis because there are large areas without data, where small-scale variability could have been present. The analysis was carried out with a focus on smoothness in such areas.

It is clear that there are multiple regions of enhanced moisture gradient in the east–west direction, and that there is variability in the number of regions, strength of gradient, and distance between regions in the north–south direction. Complex structure on small scales, similar to that shown in Fig. 2, was documented in another dryline case by Hane et al. (1997). Crawford and Bluestein (1997) have also inferred multiple moisture gradients in the dryline environment on a number of days on the basis of time series from mesonetwork stations. The dryline itself (as defined above) is collocated with the easternmost band of large dewpoint gradient and the wind shift. In the area north of Enid (END) where the aircraft crossed the dryline the wind shifted strongly, from south-southwesterly in the dry air to southeasterly in moist air. In the crossing area east-northeast of Oklahoma City (OKC) the shift was more subtle, from southerly in the dry air to south-southeasterly in the moist air. In the area where the eastward redevelopment took place the earlier location of the dryline can be identified with the area of large dewpoint gradient near OKC; a wind shift across the sharp moisture gradient is subtler in this area.

b. Aircraft observations within thinlines

An issue important to nowcasters is how well the dryline can be identified by thinlines in clear air radar return, especially in the case where multiple thinlines are present. In the 1930–2200 UTC period on this day the P-3 aircraft penetrated thinlines near the 930-hPa level (250–400 m AGL) on numerous occasions. The first of these occurred at about 1933 UTC and is illustrated in Fig. 3a. The clear air reflectivity (at 1935 UTC, and whose areal extent is approximated in Fig. 2) in this figure contains numerous short line segments in addition to the prominent arc along the eastern edge of the reflectivity field. Two longer and more persistent lines are marked by A and B in Fig. 3a. Short line segments oriented roughly northeast–southwest to the west of thinline A are likely produced by horizontal convective rolls (HCRs), owing to their spacing and orientation along the mean boundary layer wind (Weckwerth et al. 1997). The mean boundary layer wind calculated from a sounding taken approximately 60 km to the northwest at 2103 UTC was from 222° at 13 m s⁻¹. The inset at the lower left shows the aircraft observations of water vapor mixing ratio and east–west wind component (u) measured in crossing line A. Just after 1933 UTC there is a sharp increase in vapor mixing ratio from about 8 to about 10 g kg⁻¹, but the accompanying decrease in u-component wind is quite gradual (weak convergence). Changes in both mixing ratio and the u component are quite gradual in crossing line B (lower-right inset, Fig. 3a) in an area with an apparent weakness in the line of clear air return at the crossing point.

The third inset in Fig. 3a (upper right) shows the data traces as the aircraft crossed the dryline. There is an abrupt increase in mixing ratio from about 15 to 17 g;thkg⁻¹ beginning at about 1948 UTC, accompanied by a decrease in the u (east–west) component from about +1 to −3 m s⁻¹. The actual changes normal to the dryline are somewhat more pronounced than those depicted, since there is a large along-dryline component to the aircraft track at this time. Taking into account the ground speed of the aircraft and the height above ground (about 250 m), and assuming a 30° angle between aircraft track and dryline orientation, negligible changes along the line, and constant u-component divergence from the ground to flight level, this translates into an average vertical velocity of about 0.7 m s⁻¹ over a 2.1-km distance normal to the dryline.

Data from three more aircraft track segments are shown in Fig. 3b along with the radar reflectivity field at 2027 UTC. There are no radar images available between 1935 and 2027 UTC; therefore, this image followed the displayed aircraft data by 15–33 min. The aircraft crossed the dryline toward the west in the 1956–1957 UTC period, as indicated by the lower inset in Fig. 3b. Mixing ratio decreased abruptly from about 17 to 13 g kg⁻¹ and the u component increased sharply from about −4 to 0 m s⁻¹, indicating convergence. About 2 minutes later in the same inset there is a second sharp decrease in mixing ratio from 13 to 10 g kg⁻¹, followed by a more gradual increase. This is accompanied by very little change in the u component, and appears to be associated with a weaker thinline just west of the dryline. The second flight segment in Fig. 3b includes the period from 2002 to 2006 UTC (upper right). It includes a sharp drop in mixing ratio from about 11 to 8 g kg⁻¹ just before 2004 UTC, accompanied by a small increase in the u component. Though there is little clear air return evident at this location, it appears likely that these gradients are associated with line A (Fig. 3a), which has become less distinct during the 1935–2027 UTC period. Line B, indicated in Fig. 3a, is apparently a transient feature, as it is not identifiable at the later time. The third inset in Fig. 3b includes the period from 2008 to 2012 UTC (upper left) and indicates a rather sharp decrease in mixing ratio from about 9 to 6 g kg⁻¹ beginning at 2009 UTC. This is superimposed on a gradual increase in the u component and an absence of clear air return. Some azimuths from the radar may be more likely to include clear air return because of a preferred orientation of targets relative to wind direction.

Although there is no clear air return (owing to the great range) in the area where the aircraft crossed the dryline to the northwest of END (see Fig. 2), it is of interest for comparison to examine gradients in this area also. Following an aircraft data systems failure during a period of about 10 min, measurement resumed at about 2032 UTC, just prior to a dryline crossing from dry to moist side. Figure 4 shows the time series (top and bottom) of moisture and wind observations along two track segments whose locations are indicated in the middle panel, which is a subarea of Fig. 2. Dryline crossings centered at 2033:10 and 2038:30 UTC are characterized by abrupt and relatively large changes in mixing ratio and u component. Changes at the two crossings are quite similar in magnitude and distance over which they take place; mixing ratio changes are in the 6–7 g kg⁻¹ range and the changes for the u component are 8–10 m s⁻¹. Taking into account changes in both u and υ components of the wind, and if it is again assumed that there are no changes along the line, that divergence is constant between the surface and flight level, and that the dryline is oriented northwest–southeast, the mean vertical motion is estimated to be just over 1 m s⁻¹ over a distance of 2.9 km normal to the dryline. Thus, stronger vertical motion over a broader line-normal region is present in this location compared to that estimated in the region east-northeast of the radar. The relative maximum in mixing ratio centered at 2041 UTC and the secondary convergent signature that occurs just before 2040 UTC will be addressed in more detail in a later section. Gradients in both moisture and wind appear to be both larger and more concentrated in the northern region where no eastward redevelopment occurred.

An additional flight track (returning to the area of clear air return in central Oklahoma) is shown in Fig. 5. This track begins west of the original dryline location and extends eastward through the location of the redeveloped dryline. A distinct thinline, clearly discernible about 30 km west of the radar, marks the location of the dryline prior to eastward redevelopment. This thinline became more sharply defined with time in the late afternoon. Reflectivity associated with it does not extend far to the north, but the increase in moisture in the 2151–2155 UTC inset in Fig. 5 is likely an indication of a northward extension of the water vapor gradient along it. This increase in mixing ratio from about 7.5 to 10 g kg⁻¹ takes place in two abrupt steps separated by about 5 km. The accompanying decrease in the u component takes place over a relatively large distance, indicative of relatively weak convergence. Unfortunately there were no aircraft measurements across this thinline southwest of the radar, where it is very sharply defined in the reflectivity field at this time.

There is once again a well-defined thinline along the eastern edge of the reflectivity field where the dryline is located. Data from the aircraft crossing of this thinline–dryline in the 2200–2201 UTC period is shown in the 2158–2202 UTC inset in Fig. 5. An abrupt 4 g kg⁻¹ increase in mixing ratio and a 4 m s⁻¹ decrease in u component occurred during this crossing. Another diffuse thinline (northeast of the radar) intersects the 2158–2202 UTC flight segment just south of where the thinline merges with the dryline. This crossing is characterized by a gradual change in both mixing ratio (increase) and u component (decrease). Yet another more diffuse thinline, noted in Fig. 5 east-southeast of the radar, is 20–25 km west of the dryline and roughly parallel to it.

c. Thinline characteristics

The thinline associated with the dryline itself is characterized by some of the higher reflectivity values in the clear air field. The exception to this is the thinline southwest of the radar that is gradually redeveloping dryline characteristics during the latter part of the afternoon (see Part I for details). Aircraft penetrations of the dryline–thinline reveal large changes in both moisture (4 g kg⁻¹ or more) and across-line winds (5 m s⁻¹ or more) over a few kilometers. Other thinlines to the west in drier air are characterized by varying gradient magnitudes. All thinlines are associated with gradients in moisture and wind, and the density of clear air return within the lines appears closely correlated with the degree of convergence present. For example, thinlines that appear diffuse are generally associated with lesser implied convergence than a sharply defined line (keeping in mind that sharpness is radar range dependent), and may be associated with a wide range of moisture gradient magnitudes. The sharply defined thinline southwest of the radar at 2200 UTC was not sampled by the aircraft at this time and, upon preliminary examination, appears to be an example of a sharply defined thinline west of the dryline. However, as noted in Part I (e.g., Fig. 13), this feature was strengthening in the latter part of the afternoon, and was analyzed to be the relocated dryline by 0000 UTC. Surface observations at the Amber PAM site indicated that the thinline passed at about 2330 UTC and included a mixing ratio gradient comparable to that measured during the dryline crossing by the research aircraft at 1947 UTC (Fig. 3a).

The aircraft observations also included some areas of moisture gradient where no thinlines were present; these areas also lacked a convergent signature. The only exceptions to these generalizations occurred to the northwest of the radar, where convergent signatures on several occasions were not associated with any apparent thinline. The absence of well-defined thinlines in this area may have resulted from a lack of scatterers or less than optimal orientation of scatterers owing to the wind direction. It appears that direct observation of the scattering source in the dryline environment is needed to resolve the questions concerning the relation among thinline sharpness, convergence, and moisture gradients.

In the analyses described here and in those reported in Part I, there has been an implicit assumption that the thinlines observed in the zone between the two prominent thinlines (dryline and “earlier dryline”) developed as a part of the discreet propagation process. It is also conceivable, however, that these thinlines would have been present in the absence of the dryline zone. Heterogeneities in land use, etc., could produce such convergence lines, or they could simply exist owing to horizontal convective roll circulations. The availability of radar observations earlier in the afternoon in this case would have aided in addressing this point. Examination of (for example) vegetative greenness images (sensed by satellite) did not reveal any obvious relation between these lines and land use characteristics. It does seem likely that if these intermediate thinlines were preexistent features, the circulations associated with the more prominent thinlines would have at least affected their spacing, intensity, vertical extent, etc. This overall question is a prime candidate for scrutiny in future field programs.

A summary of characteristics found from the crossing of thinlines is included in Table 1. All events but those numbered 9 and 10 are depicted by time series in Fig. 3, 4, and 5. The time of each event was determined from estimating when the most rapid moisture change was occurring. The 1-s aircraft data were first smoothed to eliminate small-scale features by averaging each value over 30 s. Gradients in mixing ratio and east–west wind component were calculated over a 30-s interval (about 3.6 km). This interval was chosen because it was approximately the largest during which individual changes were of the same sign throughout the period in all cases. Gradients calculated without first smoothing the data (not shown) were also calculated. Gradients in the unsmoothed case were larger, but the smoothed values were deemed more representative since subjectivity was necessarily employed in defining the beginning and ending times of individual changes in the unsmoothed case. The next to last column in Table 1 contains estimates of the percentage of local area around the crossing point covered by clear air return. No estimates are given for events 1 and 2, which were embedded in ground clutter, nor for events 7 and 8, which were at long range from the radar. The percentage of area covered by clear air return was estimated by inspection of an elongated area encompassing about 150 km² centered on the aircraft track's intersection with the thinline. The degree of sharpness of the thinline as a whole (if present) is also indicated in the last column of the table.

Scatterplots in Fig. 6 illustrate the relationships of measurements shown in Table 1. For all events in the table the relation of convergence magnitude to the absolute value of mixing ratio gradient is shown in Fig. 6a. There appears to be a positive relation between magnitudes of moisture gradient and convergence for this relatively small sample. There is a suggestion that the moisture gradient need be above a threshold to be associated with convergence. Five of the six most strongly convergent events derive from dryline crossings. In Fig. 6b a plot of convergence versus reflectivity coverage is shown. The sample size is very small, but there is some indication of a positive relation between convergence magnitude and clear air reflectivity coverage. The outlier in this case is event 4 (small convergence, large coverage). Examination of the unsmoothed data in this case reveals that a stronger convergence signature was greatly reduced by the averaging process. A robust relation between these two quantities should not be expected, since clear air return depends not only on convergence in the wind field, but also the availability and orientation of targets. A scatterplot of moisture gradients versus reflectivity coverage (not shown) revealed a weaker relationship than exhibited in Fig. 6b.

d. A broad view of moisture and wind gradients across the dryline region

In the region near the radar where eastward redevelopment took place, the transition from moist to very dry air occurred in a series of steps (also described in Part I) over a distance of approximately 120 km normal to the dryline. The largest step in moisture was at the dryline, while other steps in the drier air to the west were generally smaller and only in some cases coincided with a radar-observed thinline. Aircraft-measured convergence was also strongest along the dryline and generally exhibited a positive relation with the strength of clear air return both at the dryline and in thinlines west of the dryline. In contrast, in the area to the north where no eastward redevelopment took place, the moisture transition occurred in two steps spanning about 50 km. The strength of gradients in moisture and across-line wind, and the inferred vertical motion at the dryline appear to have been significantly stronger than that farther south. Convective clouds, and eventually severe storms, developed along local sections of the dryline within both areas in spite of these differences. In the next section analyses of available data are examined in an attempt to explain why local areas along the dryline were more favorable environments for the development of significant deep convection.

a. Broad and small-scale views of dryline convective cloud development

Visible satellite images from Geostationary Operational Environmental Satellite-7 (GOES-7) were available at approximately 30-min intervals on this day. An image at 1931 UTC is shown in Fig. 7. Most of the clouds in the image are relatively low-based cumuli in the moist air east and north of the dryline organized in streets (e.g., in northeast Oklahoma, where not obscured by a midlevel band near the eastern border) and in cellular patterns (e.g., in Kansas at top of panel left of center). West of the dryline only cirrus are present. In roughly the northern half of the panel the dryline location can be identified by the clear–cloudy boundary. Convective clouds forming along the dryline in some locations appear to be separated from fields of clouds in the moist air, but in most locations the dryline convection appears to be at the edge of the cloud field.

A zoomed visible satellite image at 2001 UTC is shown in Fig. 8. This image was chosen because it was nearest in time to two research aircraft crossings of the dryline (at 1948 and 1956 UTC). Past investigators (e.g., NSSP Staff Members 1963; Ziegler and Hane 1993) of convective initiation along the dryline have proposed that in some cases the first convective clouds may develop within moist billows or water vapor turrets that originate within an elevated layer of moisture ahead of the dryline (see Ziegler and Hane 1993, Fig. 14) rather than at the dryline itself. Based upon the locations of clouds in relation to crossing points, it appears that there were convective clouds forming at or within a few kilometers of the dryline in this case. No measurements are available at this time to determine the presence of billows, but it is evident from both Figs. 7 and 8 that cloud streets associated with convective boundary layer rolls were present in the moist air. Whether the billows observed in the past represent intersections of horizontal convective roll circulations (sometimes producing cloud streets) with the aircraft-derived vertical cross section is a question worthy of future investigation.

In order to compare the locations of clear air reflectivity features (which appear to be associated with boundary layer convergence and upward motion) and convective clouds from visible satellite images, the two fields are combined in Fig. 9. Satellite images were available at 30-min intervals; therefore, these two times (2131 and 2201 UTC) were the two that just preceded the time when the first precipitation echo appeared (2206 UTC). It would be interesting also to compare these fields at other locations along the dryline, but the KTLX radar was the only WSR-88D in operation in the area. In order to eliminate as much error as possible in placing the visible clouds on the radar image, the satellite image grid was checked and adjusted if needed based upon identification of geographical features (reservoirs, large lakes, etc.) in each image. The horizontal displacement error in locating the cloud features following adjustment is estimated to be 1–2 km. Parallax was neglected owing to the low altitude of these clouds. There were no visible clouds (within the domain shown) west of the dryline, which is identified by the arc-shaped line of clear air return that is farthest east from the radar. In general, clouds were not included in the figure that were more than 10 km east of the dryline. At both times most of the visible clouds were coincident with elements within the line of clear air return. It is highly unlikely that the clouds themselves caused the clear air return, since there was clear air return in many locations where no visible clouds were present. The fact that the clear air return from the dryline and visible clouds were coincident is further evidence that visible clouds were forming at the dryline rather than ahead of the dryline on this day. Visible clouds in the figure that are immediately east of the dryline are likely cumulus congestus that developed at the dryline earlier and whose upper portions were advected toward the northeast. Unfortunately, owing to the coarse time resolution in the satellite images, it is not possible to track individual clouds.

b. Thunderstorm initiation in east-central Oklahoma (sector S1)

c. Lack of convective development along an extended section (S2) of the dryline

Farther north along the dryline (sector denoted S2 in Fig. 1a) there was no thunderstorm development in an along-line distance of more than 200 km. Moreover, along a portion of this segment there was very little convective cloud development during most of the afternoon. An example of a relatively cloud-free segment of the dryline that is about 60 km long can be seen in Fig. 7 in north-central Oklahoma. There is also an area that extends about 20 km into the moist air northeast of this dryline segment that is free of cloud streets associated with horizontal convective rolls. This pattern at and ahead of the dryline persisted through much of the afternoon as the dryline moved northeastward. In Fig. 16 isochrones depicting dryline motion are plotted from 1800 to 2100 UTC. The white portion of each isochrone represents locations where significant convective cloud development was occurring along the dryline, while the black portion shows where there was little or no development occurring. The background field in the figure is again the surface temperature sensed by satellite at 2201 UTC. It can be seen that during this part of the afternoon there was a close relationship between cooler surface temperatures and a lack of cloud development along a segment of the dryline, presumably owing to lessened sensible heating and destabilization compared to other nearby dryline locations.

This analysis is extended a short distance eastward in Fig. 17, where the same type of isochrone analysis from 1800 to 2200 UTC is superimposed upon a KTLX reflectivity image from 0533 UTC (the night before). The storm complex in northwest Oklahoma that contributed to the thermal pattern of Figs. 15 and 16 is shown along with the estimated envelope of moderate to heavy rainfall (heavy dashed lines) that occurred between 0230 and 1130 UTC, based upon reflectivity images from KTLX and a WSR-57 radar in Wichita, Kansas. Observed rainfall amounts from cooperative sites are also included. The cloud-free area along the dryline continued into the late afternoon, although by 2100 UTC convective cloud development began to occur within the northern half of the previous night's precipitation envelope.

d. Initiation of the thunderstorms near the Kansas–Oklahoma border (sector S3)

Thunderstorms also developed along the dryline in extreme northern Oklahoma (within sector S3 in Fig. 1b). Intensifying storms moved into southern Kansas and one complex produced tornadoes generally along a line from 40 km southwest to 30 km east of Wichita. This area is not within the region where mesoscale observing systems were concentrated, but satellite, aircraft, and long-range radar data provide the reason for storm development at this along-line location. Storm initiation occurred shortly after 2000 UTC at the intersection of the dryline and a preexisting cloud line that extended into the dry air and appeared to have derived its energy from the boundary layer. In Fig. 18 a visible satellite image at 1901 UTC shows the cloud line extending about 35 km south-southwestward from its intersection with the dryline. The superimposed surface temperature analysis shows that the cloud line at this time existed above the relatively cool envelope of air resulting from effects of the antecedent precipitation.

Isochrones indicating the motion of this cloud line are shown in Fig. 19 along with the location of the cloud line–dryline intersection point (circles) at each time. The location of the dryline at 2026 UTC and the reflectivity from the thunderstorm complex at 2027 UTC (in an early stage) are also indicated. The cloud line was first evident in the 1831 UTC visible GOES-7 image (not shown) as an area of scattered cumuli over the cool envelope (also indicated in the figure) near the location where heaviest rainfall had occurred on the previous night. By 1901 UTC the area had moved eastward and taken on the linear appearance shown in Fig. 18. Between 1930 and 2000 UTC the cloud line–dryline intersection moved over a relatively warm surface (also indicated in Fig. 19), and it was shortly after this that the first echo occurred. It is therefore possible that thermal instability associated with this relative maximum in surface temperature aided in thunderstorm initiation, adding to the enhanced low-level convergence inferred at the intersection of the cloud line and dryline. Once the convection grew to greater heights it moved faster than the dryline along a track toward the north-northeast.

The intersection of boundaries as a preferred location for convective initiation is well documented in other cases (e.g., Kessinger and Mueller 1991; Crook et al. 1991; Wilson et al. 1992; Wakimoto and Atkins 1994; Atkins et al. 1995; Hane et al. 1997; Weiss 2000). The spatial arrangement of dryline and cloud line in this case bears superficial resemblance to that found by Hane et al. (1997). This case differs from that one and others in that 1) the cloud line in this case appears to have been initiated through a strong influence of excessive evaporation from the surface, and 2) both the dryline and cloud line were moving at an appreciable rate.

Further evidence of enhanced convergence inferred at the cloud line–dryline intersection is presented in Fig. 20. The research aircraft flew one leg that crossed both the dryline and the cloud line within 15 km of their intersection point. All features in the figure are referenced to the time that the aircraft crossed the cloud line (2040 UTC); that is, the locations of the cloud line and dryline are interpolated to 2040 UTC, and the KTLX reflectivity is at 2039 UTC. Wind vectors and dewpoint temperatures are plotted along the aircraft track at 20-s intervals between 2036 and 2046 UTC and were smoothed over 20 s. Aircraft altitude varied from approximately 240 to 360 m above the terrain in flying at a nearly constant 929 hPa. The mean ground speed of the aircraft was approximately 115 m s⁻¹; therefore, the spatial separation of plotted data is about 2.3 km. As the aircraft flew toward the southwest across the dryline the wind shifted from southeasterly to southerly over a distance of less than 5 km. Crossing of the cloud line resulted in a shift from southerly to southwesterly, again over less than 5 km. These two convergent signatures (assuming negligible along-line gradients) were accompanied by distinct moisture gradients. The overall decrease in dewpoint in crossing the dryline was about 8°C, followed by an increase of about 3°C just east of the cloud line. It therefore appears that the cloud line existed in a region with convergence and enhanced moisture in the boundary layer below it. Based on the aircraft observations near the cloud line and assuming well-mixed conditions in the subcloud layer, cloud base is estimated at 1.75 km AGL. For reference, the same estimate 3–5 km east of the dryline along the aircraft track yields cloud bases at 1.05 km.

The remaining questions dealing with initiation in this area are how convective clouds developed in the dry air in the first place and by what mechanism they evolved into a line. Though speculation could be offered concerning the role of evaporation versus sensible heating in cloud formation (Segal et al. 1995), the effect of landscape variability on development of convective clouds (Rabin et al. 1990), and the existence of a preexisting mesoscale boundary, the existing observations are too limited to support firm explanations. It does seem quite likely, however, that high evaporation rates over a very moist surface were influential in the formation of the initial cloud field. Future field experiments that investigate similarly structured environments for storm initiation will hopefully obtain measurements necessary to address this issue.

a. Conclusions

A low-level aircraft survey pattern over a 150-km-scale domain revealed that the dryline environment moisture gradients in the east–west direction occurred over about 60 km in the north, where the dryline was progressing steadily northeastward, and over about 120 km elsewhere. The number of separate regions of moisture gradients also varied in the along-line direction.

Data from aircraft penetrations of thinlines near and coincident with the dryline were examined with an emphasis on line-normal moisture gradients and convergent signatures in the wind field. In general, the thinlines that were most well defined (e.g., the one identified with the dryline) in the reflectivity field were found to contain both sharp moisture gradients and distinct convergent signatures. Less distinct thinlines were associated with convergent signatures and moisture gradients of smaller magnitude. It was also observed that significant moisture gradients were in some instances not accompanied by convergent signatures and thinlines. Therefore in this case the presence of a moisture gradient was not a sufficient condition for the existence of a thinline. Both moisture gradients and convergence may be necessary for thinline existence; moreover, in this case there is a positive relation between convergence strength and thinline prominence.

Based upon comparison of satellite and radar data in east-central Oklahoma, it appears that the development of convective clouds nearest the dryline occurred in this case at or within a few kilometers of the dryline (thinline), rather than say 5–10 km ahead of the dryline. In east-central Oklahoma three thunderstorm cells developed along a 20-km section of the dryline during a 20-min period. Each successive cell developed further southward and reached its precipitating stage closer to the dryline with time.

Thunderstorm development was just to the south of a synoptic-scale dryline bulge (within sector S1); furthermore, dryline motion was toward the northeast to the north of the initiation area and toward the northwest to the south, resulting in little dryline motion along the section of the dryline where storms developed. Initiation to the south of the bulge may have benefited from the fact that convective elements tended to move parallel to the dryline, and thereby resided in moist and upward-moving air for a relatively long period.

A potential reason for the favored location of thunderstorm initiation along this section of the dryline is enhanced boundary layer convergence. Higher clear air reflectivity levels along this section suggest strong convergence. Further, there is some evidence of backed winds east of the dryline where storms formed, owing to locally lowered pressure in the dry air to the west. Aircraft and satellite observations confirm that there was a locally warm region in the boundary layer just to the northwest of the area where thunderstorm initiation occurred. This warming through the boundary layer would lead to locally lowered pressure.

Farther to the north along the dryline (sector S2) there was an area where cumulus convection was absent or diminished. Satellite surface temperature observations revealed an elongated region of relatively cool surface temperatures from west-central through northern Oklahoma. This cool region resulted from significant thunderstorm rainfall during the previous night, and exhibited near coincidence with the sector along the dryline where convection was inhibited.

In a second area where a cluster of severe thunderstorms formed southwest of Wichita along the Kansas–Oklahoma border (sector S3), storm initiation occurred at the intersection of the dryline with a cloud line that extended into the dry air. The cloud line was evident over a period of 3 h from visible satellite images, and originated over the location of heaviest rainfall on the previous night. Aircraft observations show convergence and elevated moisture values within the line. The exact mechanisms responsible for development of clouds in the dry air and for line formation are unknown.

b. Discussion

With reference to forecasting thunderstorm initiation, this study has highlighted several issues. For one, the dryline may redevelop during its forward progression (or retreat; see Part I), rather than move continuously. This obviously affects the location where storms might develop. The along-line location where storms develop is highly dependent upon along-line varying environmental conditions near the dryline, and in the dry air in particular. Along-line differences in heating rates in the dry air can affect the distribution of the boundary layer pressure field, which in turn influences variations in the wind field in both moist and dry air and strength of convergence along the dryline. Recent precipitation appears to reduce boundary layer heating rates sufficiently to retard convective cloud development along the dryline. Lines of surface-based clouds may develop in the dry air, and their intersection points with the dryline can provide preferred locations for storm development.

Though this analysis has illustrated a number of structures and causal mechanisms that have not been described previously, it has also served to emphasize a list of unresolved questions. Some of these are as follows: (a) What are the detailed dynamic and thermodynamic processes by which the dryline redevelops as observed here? (b) What are the detailed relationships between clear air return along the dryline and convective cloud development? (c) What is the mesoscale structure of the wind and pressure fields in areas along the dryline where there is localized heating in the dry air? (d) What are the mechanisms acting to produce cloud line development on the dry side of the dryline in some cases? (e) Do secondary thinlines exist independent of the discreet propagation process, and what mechanisms are operative in the division of moisture gradients and convergence between the dryline itself and these secondary lines in the dry air? This is only a sampling of questions that could be posed based on this analysis.

To resolve these issues and others, comprehensive observations and carefully designed numerical simulations are needed. The value of in situ measurements by research aircraft in the dryline environment is clear from all the structures that have been elucidated in this analysis and others. Coordinated flight patterns by two aircraft would help provide information on several scales. One aircraft might, for example, focus on the larger scale and provide soundings in key areas, while the other might concentrate on small-scale features. The ability of airborne Doppler radar to measure the motion field in clear air (Wakimoto et al. 1996) provides an opportunity to observe finescales over an area with decreased concern about evolution of structures. Other instrumentation now available that might be applied to probe structures in the dryline environment include differential absorption lidar (DIAL) to measure temperature and pressure remotely from aircraft, an interferometer for measuring vertical cross sections of temperature and humidity below an aircraft (Smith et al. 1990), and lidar based on Raman scattering (Melfi 1972; Strauch et al. 1972) to measure atmospheric moisture. Additionally, the existence of high quality surface data from the Oklahoma Mesonetwork or similar networks to be established would add key information on small temporal and spatial scales. Nonhydrostatic numerical simulations of this case or of idealized cases would contribute insight toward resolution of many of the issues mentioned above.