Multi-Scale Transects of Three North American Drylines

: North American drylines are distinct air mass boundaries that have often been examined for their relation to the initiation of severe convective storms. Three cases of drylines occurring in synoptically quiescent environments are analyzed using data obtained from a single mobile platform in concert with data from operational synoptic and mesoscale observing systems. Very distinct moisture contrasts were noted in a nocturnal April case in mountainous terrain in the Trans-Pecos region of West Texas. The other two cases revealed multi-step moisture transitions within synoptically di ﬀ use moisture gradients. Their evolution over time suggests that such multi-step patterns may be associated with diurnal and geographic forcing transitions, as well as positioning of deep moist convection.


Introduction
Various investigations of the structure and evolution of drylines have been conducted since the initial description in 1958 [1]. As observed in North America, a typical dryline can be conceptually described as the surface boundary separating a moist subtropical air mass (largely originating over the Gulf of Mexico) from much drier vertically well-mixed air originating over high plateaus of the Southwestern United States and Northern Mexico [2]. Lack of a sustained baroclinic gradient or vertical continuity with mid-and upper-tropospheric waves helps to distinguish drylines from fronts. Fine-scale measurements of the dryline environment with aircraft, remote sensing (radar/lidar), and mobile observing systems [3] have focused in particular on factors related to the initiation of severe convective storms over Western Oklahoma and Northern Texas during spring and early summer, such as during the VORTEX-95 (Verification of the Origins of Rotation in Tornadoes Experiment) [4] and IHOP (International H 2 O Project) [5][6][7][8] field campaigns. Drylines and similar features have also been recognized to occur outside of this core domain [9,10] and often without initiating significant deep convection [11][12][13]. Relatively few studies have documented detailed characteristics of these types of drylines. Interactions of drylines with topography of the Rocky Mountains foothills in Southeastern Wyoming have been investigated [14]. However no intensive data have been published from research field campaigns across the southern extent of the dryline, which is commonly analyzed southward to the Mexican border.
It has been well established from peer-reviewed dryline case studies in the Southern Great Plains that a dryline commonly develops in the vicinity of a lee trough, moves eastward during day and regresses westward at night, and may have horizontal gradients of the mixing ratio that locally exceeds 1 g/kg per km in the immediate vicinity of the boundary [3]. Horizontal density contrast across the dryline has sometimes been determined to be consistent with density current theory at certain phases of the diurnal cycle [15], while the depth of the underlying moist air east of the dryline is recognized to be closely tied with a synoptic scale subsidence inversion and mid-tropospheric flow of dry air from the high plateaus of Northern Mexico [11,[16][17][18]. Accordingly, the air above such inversions is commonly described as an advected "elevated mixed layer". Much attention has been focused on conditions for convective initiation in association with drylines [19]. On the other hand, no systematic study has been made of drylines near the U.S.-Mexico border. In particular, seasonal variations of dryline occurrence in this region may be relevant to different mechanisms for cross-border severe thunderstorms.
During 2017, transects were made through three dryline environments with the Jackson State University (JSU) mobile meteorology unit (MMU). Through analysis of these (primarily) thermodynamic data in concert with operational datasets, the character and evolution of dryline features will be explored outside of classical dryline environments. All of the cases considered are "synoptically quiescent" according to the classification of [12], although one was transitioning to a more active environment at the end of the time considered. The synoptically quiescent character was evidenced by the absence of synoptic frontal features interacting with or in close proximity to the dryline. Key questions to be considered from these cases include: • Is a distinct dryline moisture gradient and structure maintained when the low-latitude nocturnal dryline retrogresses into the high terrain of the Trans-Pecos in far West Texas? • Could it sometimes be more appropriate to conceptualize the dryline as a "peripheral boundary" that separates different convective regimes rather than as a "forcing boundary" upon which air mass convection is preferentially focused and initiated? • Is there evidence to link the diurnal cycle of boundary layer processes to variations in broadness of the dew point gradient and evolution of multi-step dryline structures?

Description of Mobile Meteorology System
The JSU MMU has evolved incrementally from the earlier system that was described in [20,21]. The core measurements have remained temperature and humidity, using removable mounting systems on a standard passenger vehicle. For the cases reported here, temperature and humidity were measured with a Rotronics HygroClip HC2S3 (from Campbell Scientific, Logan, UT, USA) [22], with meteorological data and GPS position logged to a CR23X datalogger (from Campbell Scientific, Logan, UT, USA) at 2-s increments. Since February 2017, two HygroClips have been used concurrently, with one in a U-Tube shield (primarily for heavy rain) [23] and one in a Gill shield. Since there was very little rain encountered during the transects and experience has shown faster thermal responsiveness with the Gill shield, data from the Gill shield are preferred when available. Comparisons between sensors in the two shields for the current cases showed that mean temperature differed by less than 0.2 • C and relative humidity by less than 0.3%. Additional measurements include pressure (Vaisala PTB101B) for use in the calculation of potential temperature. soundings were also obtained with a mobile Windsond system [32]. This system uses a much smaller balloon with slower ascent rates compared to Vaisala and other operational radiosonde systems, and is primarily intended for lower tropospheric applications.

Operational HRRR Model Data
To complement observed datastreams (especially for a three-dimensional structure), archived data products from the operational 3-km HRRR (high-resolution rapid refresh) model [33][34][35] were used. Hourly data are available through the NOAA Air Resources Lab (ARL) READY (real-time environmental applications and display system) website [36,37]. Challenges in the depiction and identification of drylines in operational numerical model data have been considered [38]. Since the resolution and operational constraints limit the potential for the HRRR to describe details of dryline structure, these data will be utilized primarily for the larger mesoscale environment and approximate location of features consistent with drylines, such as concentrated moisture gradients in concert with lower tropospheric confluence patterns. Within these limitations, the operational model can provide spatial and temporal context in between observational datapoints as well as similarity to products that would be used in real-time forecasting and analysis.

Radar Data
Storm development and clear-air fine lines [39] are monitored using operational NOAA NEXRAD WSR-88D weather radar data when available. Primarily the base elevation reflectivity is used, but Doppler velocity and dual polarization parameters are also examined when useful. Although mosaic products were used for broader perspectives, only the Level II data from the NEXRAD radar sites (from the National Centers for Environmental Information) [40] were presented here.

Case 1: 8 April 2017 Alpine, Texas
During the period of 7 through to 10 April 2017, operational model guidance called for a dryline to develop in Western Texas ( Figure 1). The upper tropospheric pattern was characterized by a broad ridge across the central United States (not shown). A mobile transect with the JSU MMU was planned for the evening of 8 April across the southernmost section of the dryline. Including stops, the transect along U.S. highway 90 from San Antonio (96.8 W) to Alpine (103.6 W; Figure 2a) was conducted from 1603 UTC on 8 April until 0252 UTC on 9 April. The mobile data between Comstock (101.2 W) and Alpine will be emphasized since stops were minimal on this segment. The strong moisture contrast in the HRRR extended southward through the mountainous terrain of the Northern Coahuila state in Mexico. Maximum convective available potential energy (CAPE) of over 2100 J/kg was indicated by the HRRR around the Serrañías del Burro mountains, where severe convection is often preferentially initiated [41][42][43]. However skies remained generally clear across most of the region west of 100 W during the period. At 2049 UTC a Windsond lower tropospheric sonde was launched at Amistad Reservoir near Del Rio (29.  The sounding from Amistad Reservoir ascended up to 724 hPa before it was terminated, showing a very pronounced inversion between 856 to 841 hPa (1450-1630 m MSL; Figure 3). Data plots from the Windsond software (v. 2.90) show that the peak wind speed of 20 kt was just below the inversion base, while the wind shift from SSE to SW began near the speed minimum around 1750 m. It should be noted that the Windsond software indicates height MSL whereas SHARPpy uses height AGL. Comparison with the HRRR profile from 2100 UTC shows a similar inversion and wind shift. However the inability of the model to produce as abrupt of a transition results in the inversion base appearing about 200 m lower. The strengths and weaknesses of the HRRR above 700 hPa are similarly evident in comparison with the nearby (20 km southeast) KDRT radiosonde that was launched at 2307 UTC. Aside from the more diffuse inversion layer in the HRRR, the temperature and wind agreed fairly well. There are larger differences in dew point, particularly between 600 and 700 hPa. This appears to be related to the HRRR's overdevelopment of isolated deep convection in Northern Coahuila between 2200 and 2300 UTC whose mid-level moisture plume extended toward Del Rio. The thermodynamic contrast across the dryline is indicated observationally at 0000 UTC by comparison of the soundings from KDRT and KMAF (although nearly 300 km farther north) ( Figure 4). Radar coverage was inadequate to delineate the dryline south of about 31 N, but there was a fine line detected at 2245 UTC in the Permian Basin area and isolated precipitation echoes in the Serrañías del Burro. At 0000 UTC the HRRR indicated a narrow band of upward vertical motion at 0.85 sigma along the dryline even though it was inadequate to produce convective cloud [44], and a resulting relatively broad area of higher humidity at 0.90 sigma east of the dryline convergence ( Figure 5). The sounding from Amistad Reservoir ascended up to 724 hPa before it was terminated, showing a very pronounced inversion between 856 to 841 hPa (1450-1630 m MSL; Figure 3). Data plots from the Windsond software (v. 2.90) show that the peak wind speed of 20 kt was just below the inversion base, while the wind shift from SSE to SW began near the speed minimum around 1750 m. It should be noted that the Windsond software indicates height MSL whereas SHARPpy uses height AGL. Comparison with the HRRR profile from 2100 UTC shows a similar inversion and wind shift. However the inability of the model to produce as abrupt of a transition results in the inversion base appearing about 200 m lower. The strengths and weaknesses of the HRRR above 700 hPa are similarly evident in comparison with the nearby (20 km southeast) KDRT radiosonde that was launched at 2307 UTC. Aside from the more diffuse inversion layer in the HRRR, the temperature and wind agreed fairly well. There are larger differences in dew point, particularly between 600 and 700 hPa. This appears to be related to the HRRR's overdevelopment of isolated deep convection in Northern Coahuila between 2200 and 2300 UTC whose mid-level moisture plume extended toward Del Rio. The thermodynamic contrast across the dryline is indicated observationally at 0000 UTC by comparison of the soundings from KDRT and KMAF (although nearly 300 km farther north) ( Figure 4). Radar coverage was inadequate to delineate the dryline south of about 31 N, but there was a fine line detected at 2245 UTC in the Permian Basin area and isolated precipitation echoes in the Serrañías del Burro. At 0000 UTC the HRRR indicated a narrow band of upward vertical motion at 0.85 sigma along the dryline even though it was inadequate to produce convective cloud [44], and a resulting relatively broad area of higher humidity at 0.90 sigma east of the dryline convergence ( Figure 5).          The mobile transect from Comstock to Alpine traversed from a dissected plateau of around 400 m MSL up to mountain valleys at about 1400 m, with surrounding mountain peaks near Alpine of over 2000 m. Sunset was at 0113 UTC and the dryline retrogressed westward about 100 km by the time that it was intercepted below the slopes of the Del Norte Mountains about 10 km east of Alpine at 0244 UTC. Virtual potential temperatures (θ v ) from the mobile system along this segment have been adjusted to 0300 UTC by temporal linear interpolation of the changes (mainly from a diurnal cycle) observed at synoptic/mesonet stations along the route. This method to analyze long-distance mobile datasets was introduced in [20]. This had little impact near the dryline since the mobile observations were close to 0300 UTC. However, the time span of the full transect was long enough that significant nocturnal cooling (more than 4 • C) occurred at eastern locations. Adjusting to a standard time facilitates a "synoptic" spatial examination of the data, in order to depict variations within the moist air mass and to allow direct comparison with station observations. The use of stations from significantly different elevation than the mobile transect (in particular at 103.2 W) for temporal adjustment is acceptable because the longitudinal position relative to the dryline is expected to produce similar diurnal variations (note that much of the westward increase of θ v can be attributed to increasing elevation). No adjustment has been done with humidity measurements since dew points generally changed by less than 1 • C during the period except in association with dryline passage.
Mobile and station observations at 0300 UTC ( Figure 6) document the homogeneity of the air masses on each side of the dryline and surface flow patterns affecting the region. Between 101 and 103.5 W, the mixing ratio from the mobile system decreased from about 11 to 8 g/kg (potentially related to land use and soil moisture gradation; Figure 7a), while θ v increased by more than 7 K. At the dryline, a mixing ratio drop of about 6 g/kg was recorded within a distance of only a couple of km ( Figure 7b). Between the 3 and 4 g/kg isopleths, the gradient exceeded 5.5 g/kg per km. The virtual potential temperature in this area oscillated up and down for about 5 km, before dropping about 5 K over the remaining 5 km into Alpine. Returning eastward a half hour later, the moisture change had transformed into three steps with the easternmost boundary about 3 km farther west than before. The proximity of topographic features suggests that the cooling near Alpine may be more strongly forced by nocturnal drainage flow than by direct dryline influence. The rapidly changing fine-scale moisture pattern may also be impacted by interactions between the dryline and detailed (<1 km) topography of the surrounding Del Norte Mountains ( Figure 8). The HRRR did an adequate job of simulating mesoscale surface conditions, with good agreement at observing sites, but positioning the dryline about 15 km too far east and spreading the gradient out over several kilometers. The analyzed dryline from observations extended south to the Mexican border near Rio Grande Village and north across the Pecos River to Pyote.
A mobile sounding was launched at 0340 UTC within a couple km of the eastern edge of the dryline where the surface dew point was about 10 • C (9 g/kg mixing ratio). The wind rapidly veered from southeasterly to southwesterly through a surface-based inversion only about 350 m deep (Figure 9), as dew point dropped down to below 0 • C. The inversion and moist layer did not extend as high as the crests of the nearby Del Norte Mountains and other ranges. Above the inversion the temperature dropped at a steady lapse rate of around 10 • C/km up to a minor inversion at 587 hPa. A relatively moist layer was embedded within the dry air mass from 750 hPa up to the minor inversion. In spite of SSE surface winds and some qualitative agreement, the 0300 UTC HRRR profile at the Windsond site (but west of the HRRR dryline) only shows a radiation inversion and has significant differences in dew point. While the HRRR near-surface moisture differences are consistent with dryline displacement and a shallow radiation inversion, the discrepancies between 700 hPa and 500 hPa seem more likely due to very limited upper air data upwind in Northern Mexico rather than related to model depiction of the dryline. A mobile sounding was launched at 0340 UTC within a couple km of the eastern edge of the dryline where the surface dew point was about 10 °C (9 g/kg mixing ratio). The wind rapidly veered from southeasterly to southwesterly through a surface-based inversion only about 350 m deep (Figure 9), as dew point dropped down to below 0 °C. The inversion and moist layer did not extend as high as the crests of the nearby Del Norte Mountains and other ranges. Above the inversion the temperature dropped at a steady lapse rate of around 10 °C/km up to a minor inversion at 587 hPa. A relatively moist layer was embedded within the dry air mass from 750 hPa up to the minor inversion. In spite of SSE surface winds and some qualitative agreement, the 0300 UTC HRRR profile at the Windsond site (but west of the HRRR dryline) only shows a radiation inversion and has significant differences in dew point. While the HRRR near-surface moisture differences are consistent with dryline displacement and a shallow radiation inversion, the discrepancies between 700 hPa and 500 hPa seem more likely due to very limited upper air data upwind in Northern Mexico rather than related to model depiction of the dryline.
(a) A mobile sounding was launched at 0340 UTC within a couple km of the eastern edge of the dryline where the surface dew point was about 10 °C (9 g/kg mixing ratio). The wind rapidly veered from southeasterly to southwesterly through a surface-based inversion only about 350 m deep (Figure 9), as dew point dropped down to below 0 °C. The inversion and moist layer did not extend as high as the crests of the nearby Del Norte Mountains and other ranges. Above the inversion the temperature dropped at a steady lapse rate of around 10 °C/km up to a minor inversion at 587 hPa. A relatively moist layer was embedded within the dry air mass from 750 hPa up to the minor inversion. In spite of SSE surface winds and some qualitative agreement, the 0300 UTC HRRR profile at the Windsond site (but west of the HRRR dryline) only shows a radiation inversion and has significant differences in dew point. While the HRRR near-surface moisture differences are consistent with dryline displacement and a shallow radiation inversion, the discrepancies between 700 hPa and 500 hPa seem more likely due to very limited upper air data upwind in Northern Mexico rather than related to model depiction of the dryline. Using a combination of HRRR, Windsond, and surface observations, vertical cross-sections of at 2200 and 0400 UTC illustrate the relation of the moist air mass to regional terrain ( Figure 10). The cross-sections were analyzed using a subjective combination of mobile surface, fixed surface, radiosonde, and HRRR model data. The HRRR data from each model level were plotted at a 0.5 degree of longitude intervals and other strategic locations along the mobile track, and were adjusted Using a combination of HRRR, Windsond, and surface observations, vertical cross-sections of at 2200 and 0400 UTC illustrate the relation of the moist air mass to regional terrain ( Figure 10). The cross-sections were analyzed using a subjective combination of mobile surface, fixed surface, radiosonde, and HRRR model data. The HRRR data from each model level were plotted at a 0.5 degree of longitude intervals and other strategic locations along the mobile track, and were adjusted to reduce local biases noted by comparison with radiosonde data. It is assumed that observed soundings provide ideal vertical resolution where available, surface mobile data provide ideal horizontal resolution of near-surface features, and the HRRR model data are able to broadly depict upper-air patterns between radiosondes. The 0.5 degree spacing of profiles appeared adequate to capture the scales of relevant coherent structures noted in model level plots, except near the surface, dryline, and areas of organized convection (not crossing the transect). Profiles were chosen closer together in the proximity of the dryline for Figure 10b. Near the soundings and near the surface, much more weight was given to the observed versus model data so that the horizontal dryline gradient and vertical inversion gradient could be represented better than through model data alone. The 15 km spatial displacement of the dryline by the HRRR was accounted for. A similar procedure was utilized by [20]. Between 2200 UTC and 0400 UTC, the cross-sections show the westward upslope retrogression of the dryline. The lowest 500 m or so on the moist side generally have mixing ratios greater than 8 g/kg, while air to the west of the dryline is well under 4 g/kg. The moist layer is trapped below a subsidence inversion, although the HRRR model does indicate some upward transport within about 50-100 km to the east of the dryline boundary. Although there was only one mobile radiosonde system and one mobile observing system crossing the dryline (twice), the observations were able to conclusively show intense fine-scale dryline contrasts close to the Mexican border (far south of similar published data) in relatively high and complex terrain, as well as linkage of the dryline to the inversion and wind shear at heights comparable to the local terrain. It is believed that the 0340 UTC Windsond profile is both the closest that a sounding has been reported to a nocturnal dryline and the shallowest stable moist layer documented in association with a dryline. The most similar case study was that of [45], which considered nocturnal evolution of a dryline in the vicinity of Midland. Their case differed mainly in being over much smoother terrain and by the approach of a squall line and cold front.

Case 2: 9 April 2017 Snyder, Texas
The dryline was intercepted under very different circumstances on the afternoon of 9 April 2017 east of the Caprock Escarpment near Snyder, Texas. No radiosonde or clear-air radar data were collected in the area, and some surface data from the MMU were lost due to equipment malfunctions. However data from the HygroClip within the U-Tube shield were available along the most relevant segments. The first indication of a dryline was at 100.6 W about 120 km east of (below) the Caprock at 2135 UTC, along an eastward transect roughly 300 km farther north than the previous day's transect ( Figure 11). Evidence of a secondary weaker dryline was found farther east at about 2300 UTC, although the mobile track was less favorably oriented relative to it. Similar double dryline Although there was only one mobile radiosonde system and one mobile observing system crossing the dryline (twice), the observations were able to conclusively show intense fine-scale dryline contrasts close to the Mexican border (far south of similar published data) in relatively high and complex terrain, as well as linkage of the dryline to the inversion and wind shear at heights comparable to the local terrain. It is believed that the 0340 UTC Windsond profile is both the closest that a sounding has been reported to a nocturnal dryline and the shallowest stable moist layer documented in association with a dryline. The most similar case study was that of [45], which considered nocturnal evolution of a dryline in the vicinity of Midland. Their case differed mainly in being over much smoother terrain and by the approach of a squall line and cold front.

Case 2: 9 April 2017 Snyder, Texas
The dryline was intercepted under very different circumstances on the afternoon of 9 April 2017 east of the Caprock Escarpment near Snyder, Texas. No radiosonde or clear-air radar data were collected in the area, and some surface data from the MMU were lost due to equipment malfunctions. However data from the HygroClip within the U-Tube shield were available along the most relevant segments. The first indication of a dryline was at 100.6 W about 120 km east of (below) the Caprock at 2135 UTC, along an eastward transect roughly 300 km farther north than the previous day's transect ( Figure 11). Evidence of a secondary weaker dryline was found farther east at about 2300 UTC, although the mobile track was less favorably oriented relative to it. Similar double dryline structures have been previously reported from various studies [5]. The main synoptic difference from the previous day was the approach of an upper tropospheric trough into the Southwestern United States. Although some convective clouds and precipitation developed, they were not obviously connected to forcing attributable to the dryline itself or the approaching trough. As indicated in the moderate resolution imaging spectroradiometer (MODIS) imagery, the character and scale of convection was generally different west of and east of the western dryline boundary. Scattered rain showers were observed in the area, but none were intercepted by the MMU. Although the character of the convection was not indicative of direct forcing by the dryline boundaries, the moister air masses east of them were more favorable for moist convection. The observed cloud cover across North-Central Texas during the afternoon was probably an intersection of marginal forcing from an undetected weak short wave with diurnal heating of the moist air mass. It may be more appropriate to attribute the relative lack of productive convection to the west of the dryline position/passage than it is to attribute the precipitating convection to dryline forcing or initiation. In the absence of observed vertical moisture profiles or high-resolution numerical simulations, determination of convective instability mechanisms remains inconclusive. This sense of a dryline sometimes having an indirect role as a boundary between convective environments without contributing directly to trigger forced convection may be described as a "peripheral" boundary role, as opposed to a direct forcing role.
Atmosphere 2020, 11, x FOR PEER REVIEW 15 of 27 structures have been previously reported from various studies [5]. The main synoptic difference from the previous day was the approach of an upper tropospheric trough into the Southwestern United States. Although some convective clouds and precipitation developed, they were not obviously connected to forcing attributable to the dryline itself or the approaching trough. As indicated in the moderate resolution imaging spectroradiometer (MODIS) imagery, the character and scale of convection was generally different west of and east of the western dryline boundary. Scattered rain showers were observed in the area, but none were intercepted by the MMU. Although the character of the convection was not indicative of direct forcing by the dryline boundaries, the moister air masses east of them were more favorable for moist convection. The observed cloud cover across North-Central Texas during the afternoon was probably an intersection of marginal forcing from an undetected weak short wave with diurnal heating of the moist air mass. It may be more appropriate to attribute the relative lack of productive convection to the west of the dryline position/passage than it is to attribute the precipitating convection to dryline forcing or initiation. In the absence of observed vertical moisture profiles or high-resolution numerical simulations, determination of convective instability mechanisms remains inconclusive. This sense of a dryline sometimes having an indirect role as a boundary between convective environments without contributing directly to trigger forced convection may be described as a "peripheral" boundary role, as opposed to a direct forcing role. The MMU track extended northeastward from Odessa, Texas (102.4 W) at 1806 UTC on 9 April to Sulphur, Oklahoma (97.0 W) at 0323 UTC on 10 April. At the western dryline, the mixing ratio increased from roughly 4 to 8 g/kg (Figure 12), most of which occurred over about a 2 km distance. The MMU track extended northeastward from Odessa, Texas (102.4 W) at 1806 UTC on 9 April to Sulphur, Oklahoma (97.0 W) at 0323 UTC on 10 April. At the western dryline, the mixing ratio increased from roughly 4 to 8 g/kg (Figure 12), most of which occurred over about a 2 km distance. As might be expected during peak diurnal heating, there was no clear relationship between dryline position and θ v . The eastern dryline had a similar increment from 8 to 12 g/kg, but it appeared to be a broader gradient over about a 50 km distance. The HRRR indicated CAPE above 3000 J/kg at 2200 UTC, a broad moisture gradient in the vicinity, and generally confluent surface winds ( Figure 13). It failed to clearly distinguish the sudden transitions at the two dryline features, but it was consistent with the existence of two zones of enhanced moisture gradient. The combined observational evidence seems to point to a split from a single dryline into two drylines somewhere about 50 km south of the western dryline intercept. Other double dryline structures in published studies have not usually reported such similar sized mixing ratio changes across the two structures [5,7,47,48]. It is noted that published accounts of double dryline structures are all from mid to late afternoon, suggesting that they may be related to diurnal cycle transitions. This could be partly related to variable land surface forcing as the dryline tends to be quickly advancing eastward into different vegetation and soil moisture regimes at this time of day.
Atmosphere 2020, 11, x FOR PEER REVIEW 16 of 27 As might be expected during peak diurnal heating, there was no clear relationship between dryline position and . The eastern dryline had a similar increment from 8 to 12 g/kg, but it appeared to be a broader gradient over about a 50 km distance. The HRRR indicated CAPE above 3000 J/kg at 2200 UTC, a broad moisture gradient in the vicinity, and generally confluent surface winds ( Figure 13). It failed to clearly distinguish the sudden transitions at the two dryline features, but it was consistent with the existence of two zones of enhanced moisture gradient. The combined observational evidence seems to point to a split from a single dryline into two drylines somewhere about 50 km south of the western dryline intercept. Other double dryline structures in published studies have not usually reported such similar sized mixing ratio changes across the two structures [5,7,47,48]. It is noted that published accounts of double dryline structures are all from mid to late afternoon, suggesting that they may be related to diurnal cycle transitions. This could be partly related to variable land surface forcing as the dryline tends to be quickly advancing eastward into different vegetation and soil moisture regimes at this time of day.   The pattern of dew point between 31 and 35 N in the HRRR (not shown) evolved from a distinct boundary near 103 W in the early morning (1200 UTC) to a quite broad gradient at 1800 UTC, with a more distinct dryline to the north and south. Around 0000 UTC it appears to collapse back into a well-defined single dryline near 100 W, which retrogresses westward by 0300 UTC (within Texas). Sunset along the mobile track was at 0100 UTC. Observations are consistent with the general pattern evolution from the HRRR. Widespread middle and upper level clouds moved rapidly into the domain from the west around sunset, in association with a wave in the subtropical jet stream. At least for this particular transect, it seems that the double dryline pattern occurred as a transitional stage before the coalescence of the moisture gradient back into a single well-defined dryline before sunset.

Case 3: 12 June 2017 Midland, Texas
Except at higher latitudes (e.g., Wyoming), the frequency of drylines tends to dramatically decrease from the beginning to end of June. On the larger scale, this seems related to seasonal changes that are also associated with onset of the North American monsoon system around the beginning of July [8]. This regional moisture pattern change is anecdotally demonstrated by longitudinal variation of mixing ratio from multi-day mobile transects that were collected during mid-June and late June between central Texas and Southern California (Figure 14). While such multi-day continental scale mobile transects are quite non-standard in synoptic meteorology, they provide spatial structure to climatological patterns and have been utilized previously for air quality measurements [49,50]. The mobile transect of 12 June 2017 was conducted from Odessa, Texas (1840 UTC) to Carlsbad, New Mexico (2134 UTC), with the expectation of intercepting a weak dryline near 103 W at around 2000 UTC. The pattern of dew point between 31 and 35 N in the HRRR (not shown) evolved from a distinct boundary near 103 W in the early morning (1200 UTC) to a quite broad gradient at 1800 UTC, with a more distinct dryline to the north and south. Around 0000 UTC it appears to collapse back into a well-defined single dryline near 100 W, which retrogresses westward by 0300 UTC (within Texas). Sunset along the mobile track was at 0100 UTC. Observations are consistent with the general pattern evolution from the HRRR. Widespread middle and upper level clouds moved rapidly into the domain from the west around sunset, in association with a wave in the subtropical jet stream. At least for this particular transect, it seems that the double dryline pattern occurred as a transitional stage before the coalescence of the moisture gradient back into a single well-defined dryline before sunset.

Case 3: 12 June 2017 Midland, Texas
Except at higher latitudes (e.g., Wyoming), the frequency of drylines tends to dramatically decrease from the beginning to end of June. On the larger scale, this seems related to seasonal changes that are also associated with onset of the North American monsoon system around the beginning of July [8]. This regional moisture pattern change is anecdotally demonstrated by longitudinal variation of mixing ratio from multi-day mobile transects that were collected during mid-June and late June between central Texas and Southern California (Figure 14). While such multi-day continental scale mobile transects are quite non-standard in synoptic meteorology, they provide spatial structure to climatological patterns and have been utilized previously for air quality measurements [49,50]. The mobile transect of 12 June 2017 was conducted from Odessa, Texas (1840 UTC) to Carlsbad, New Mexico (2134 UTC), with the expectation of intercepting a weak dryline near 103 W at around 2000 UTC. At sunrise an intense dryline was located west of 104 W between the Pecos River and foothills of the Guadalupe Mountains. Cross-sections of potential temperature and mixing ratio were constructed from a combination of radiosondes (KMAF and KEPZ), HRRR, and surface observations at 1100 UTC along the vehicle track ( Figure 15). The top of the stable layer capping the moist air mass sloped downward from 1600 m MSL at Midland to intersect the topography at about 1100 m west of Carlsbad [51]. Mixing ratio increased westward from Midland, before suddenly dropping to less than 6 g/kg west of the dryline. Between 1500 and 3000 m, temperatures generally decreased to the west from a thermal maximum that was tilted upward to the east above the subsidence inversion. West of the dryline, very shallow surface inversions were consistent with nocturnal drainage driven by radiative cooling under light winds, while no well-defined surface inversion was present in the moist air at Midland. At sunrise an intense dryline was located west of 104 W between the Pecos River and foothills of the Guadalupe Mountains. Cross-sections of potential temperature and mixing ratio were constructed from a combination of radiosondes (KMAF and KEPZ), HRRR, and surface observations at 1100 UTC along the vehicle track ( Figure 15). The top of the stable layer capping the moist air mass sloped downward from 1600 m MSL at Midland to intersect the topography at about 1100 m west of Carlsbad [51]. Mixing ratio increased westward from Midland, before suddenly dropping to less than 6 g/kg west of the dryline. Between 1500 and 3000 m, temperatures generally decreased to the west from a thermal maximum that was tilted upward to the east above the subsidence inversion. West of the dryline, very shallow surface inversions were consistent with nocturnal drainage driven by radiative cooling under light winds, while no well-defined surface inversion was present in the moist air at Midland. After 1400 UTC, the dryline rapidly advanced eastward from Carlsbad. Subjective analyses of dryline position indicate that the HRRR dryline generally lagged about two hours behind (at 32 N) and up to around 100 km west compared to observations. Both the HRRR and observations showed a broad bulge of dry air between 1600 and 1800 UTC along the Pecos River south of Midland. The After 1400 UTC, the dryline rapidly advanced eastward from Carlsbad. Subjective analyses of dryline position indicate that the HRRR dryline generally lagged about two hours behind (at 32 N) and up to around 100 km west compared to observations. Both the HRRR and observations showed a broad bulge of dry air between 1600 and 1800 UTC along the Pecos River south of Midland. The progression of HRRR profiles at KMAF between 1700 and 1900 UTC (not shown) suggests that drying was mainly from diabatic heating of the boundary layer until the surface reached the convective temperature so that the capping inversion was removed. A similar progression is noted longitudinally in 1900 UTC HRRR profiles between Midland and Carlsbad. Mid-day convective clouds over high terrain northeast of Fort Davis developed in the vicinity of the dryline and developed a series of thunderstorms toward the northeast that reached the Fort Stockton area by 2100 UTC and KODO at 2303 UTC. The resulting outflow boundary advanced northward into the dry air west of the dryline and was well delineated in KMAF radar products ( Figure 16). A peak wind gust of 59 kt was observed at KODO at 2307 UTC, and 58 kt at KMAF at 2333 UTC. Public storm reports in the area indicated hail up to 5 cm and wind gusts up to 72 kt. Although the HRRR showed similar positions of the gust front, it seems to lag about one hour behind observations. The temporal evolution of temperature and dew point over 24 h at Andrews, Texas illustrates the passage of the dryline and outflow boundary ( Figure 17). A conceptual framework to consider the interaction of the thunderstorm outflow boundary with the dryline is to recognize the outflow boundary as primarily propagating by near-surface advection and the dryline(s) as primarily propagating by vertical mixing processes. Although dew point quickly rose following the outflow boundary, it remained fairly constant for a while before rising to values similar to those east of the dryline. A reasonable speculation could be that the shallow convective outflow cut underneath and mixed vertically with the deeper dry air behind the dryline. In this way an elevated residual dryline could persist south of the outflow boundary and then gradually mix into a more diffuse moisture gradient [52].  Figure 17). A conceptual framework to consider the interaction of the thunderstorm outflow boundary with the dryline is to recognize the outflow boundary as primarily propagating by near-surface advection and the dryline(s) as primarily propagating by vertical mixing processes. Although dew point quickly rose following the outflow boundary, it remained fairly constant for a while before rising to values similar to those east of the dryline. A reasonable speculation could be that the shallow convective outflow cut underneath and mixed vertically with the deeper dry air behind the dryline. In this way an elevated residual dryline could persist south of the outflow boundary and then gradually mix into a more diffuse moisture gradient [52].   The mobile transect sampled a broad but complex gradient of humidity between about 102.2 and 103.3 W, with temporally adjusted mixing ratio (to 1900 UTC) ranging from 15 down to 5 g/kg (dew point from 19 ℃ to 2 ℃). Spatial analysis determined that some of the complexity is probably due to track orientation with respect to the northern edge of the dry bulge ( Figure 18). In any case, a simplification of the transect data suggests that at least three major boundaries were embedded within the mixing ratio gradient (Figure 19), with some microscale oscillations as large as 5 g/kg. A similar three-step structure was noted temporally in the meteograms of Andrews and other stations. Although limited station distribution prevented spatial examination of the multiple steps, the similarity gives credence that they were real features that probably existed for at least a couple of hours. It seems that such broad gradients that are often described as "diffuse drylines" may be better understood to often contain multiple distinct moisture boundaries, with complex and turbulent mixing zones between them [53][54][55]. The variations noted in this case seem to bear characteristics of both the type 2 and type 3 dryline structures of [5]. It may also be noted however that no distinct wind shift was noted across the dryline at this time, again indicative that vertical mixing processes were more important to dryline maintenance and motion than horizontal advection and convergence [56,57]. The dew point gradient within the HRRR was similarly oriented, but much broader than the observations and with no indication of embedded boundaries. It was determined that inconsistencies between the HRRR and observations were too large in the afternoon and evening for a cross-section analysis to be useful. The mobile transect sampled a broad but complex gradient of humidity between about 102.2 and 103.3 W, with temporally adjusted mixing ratio (to 1900 UTC) ranging from 15 down to 5 g/kg (dew point from 19 • C to 2 • C). Spatial analysis determined that some of the complexity is probably due to track orientation with respect to the northern edge of the dry bulge ( Figure 18). In any case, a simplification of the transect data suggests that at least three major boundaries were embedded within the mixing ratio gradient (Figure 19), with some microscale oscillations as large as 5 g/kg. A similar three-step structure was noted temporally in the meteograms of Andrews and other stations. Although limited station distribution prevented spatial examination of the multiple steps, the similarity gives credence that they were real features that probably existed for at least a couple of hours. It seems that such broad gradients that are often described as "diffuse drylines" may be better understood to often contain multiple distinct moisture boundaries, with complex and turbulent mixing zones between them [53][54][55]. The variations noted in this case seem to bear characteristics of both the type 2 and type 3 dryline structures of [5]. It may also be noted however that no distinct wind shift was noted across the dryline at this time, again indicative that vertical mixing processes were more important to dryline maintenance and motion than horizontal advection and convergence [56,57]. The dew point gradient within the HRRR was similarly oriented, but much broader than the observations and with no indication of embedded boundaries. It was determined that inconsistencies between the HRRR and observations were too large in the afternoon and evening for a cross-section analysis to be useful.   Once well within the dry air, the MMU observed mixing ratio variations of about 1 g/kg on scales of 5-10 km. It is not known if these may be related to land use or soil moisture patterns. Spikes of mixing ratio to 10 g/kg (12 °C dew point) were noted on scales about 1 km or less in the immediate vicinity of a series of salt lakes southeast of Carlsbad.

Discussion and Conclusions
Each of the dryline cases examined includes some aspect that distinguishes from previously published studies. While largely consistent with previously documented cases, the data extend the breadth of features, forcing scenarios, and geographic context.

•
The maintenance of intense dryline contrasts in complex terrain near the U.S.-Mexico border was confirmed well after sunset, with a microscale mixing ratio gradient noted of around 5.5 g/kg per km. Not only is this in a previously unsampled region, but it is believed to be the first published account of "fine-scale" dryline details during nighttime.

•
Complex multi-step dryline patterns within what seem to be diffuse patterns on the synoptic scale have been documented in the afternoon, leading to further questions about the relationship of dryline evolution and movement to diurnal forcing and surface interactions. While drylines have sometimes been recognized as occurring in association with broader moisture gradients and as multiple steps or jumps [51,56,[58][59][60], little is known about relationship to diurnal and other influences.

•
In one case, afternoon convective development seemed only weakly influenced by the dryline. In another case severe convection developed, but the observational resources were too limited to adequately describe interactions between multi-step dryline patterns and convection. Consideration of the behavior of the dryline and of deep convective initiation in these cases contributed to the proposition that a conceptual distinction between peripheral and forcing roles of surface air mass boundaries such as drylines could be appropriate and useful. Once well within the dry air, the MMU observed mixing ratio variations of about 1 g/kg on scales of 5-10 km. It is not known if these may be related to land use or soil moisture patterns. Spikes of mixing ratio to 10 g/kg (12 • C dew point) were noted on scales about 1 km or less in the immediate vicinity of a series of salt lakes southeast of Carlsbad.

Discussion and Conclusions
Each of the dryline cases examined includes some aspect that distinguishes from previously published studies. While largely consistent with previously documented cases, the data extend the breadth of features, forcing scenarios, and geographic context.

•
The maintenance of intense dryline contrasts in complex terrain near the U.S.-Mexico border was confirmed well after sunset, with a microscale mixing ratio gradient noted of around 5.5 g/kg per km. Not only is this in a previously unsampled region, but it is believed to be the first published account of "fine-scale" dryline details during nighttime.

•
Complex multi-step dryline patterns within what seem to be diffuse patterns on the synoptic scale have been documented in the afternoon, leading to further questions about the relationship of dryline evolution and movement to diurnal forcing and surface interactions. While drylines have sometimes been recognized as occurring in association with broader moisture gradients and as multiple steps or jumps [51,56,[58][59][60], little is known about relationship to diurnal and other influences.

•
In one case, afternoon convective development seemed only weakly influenced by the dryline. In another case severe convection developed, but the observational resources were too limited to adequately describe interactions between multi-step dryline patterns and convection. Consideration of the behavior of the dryline and of deep convective initiation in these cases contributed to the proposition that a conceptual distinction between peripheral and forcing roles of surface air mass boundaries such as drylines could be appropriate and useful.
All non-operational data collected for the cases studied were obtained from a single mobile vehicle platform, with some logistical constraints. While this allowed some degree of flexibility, the spatial coverage of multiple platforms and complimentary data streams of research radar and lidar systems were not available. Future investigation of similar cases would benefit from deployment of at least two mobile mesonet systems, four strategically positioned radiosondes, remote sensing by a Doppler radar or lidar, and supplementary automated surface stations.
The mountainous terrain of southern New Mexico, Trans-Pecos Texas, and Coahuila represents a very different environment from where most dryline field campaigns have taken place. In spite of the logistical challenges of limited road networks, the high frequency of drylines in the region offers significant opportunity for new observational studies. With a greater density of initialization and verification data, modeling efforts at higher resolution than operational forecast models would be able to make greater progress at describing processes over the less accessible areas. Ideally, a serious study of the southern end of the dryline would focus observational and modeling efforts over a broad region from the Permian Basin into Northern Coahuila. Better understanding of the dryline in this region could have implications in the interplay of contributions to the seasonal transition (typically around the end of June) to Pacific monsoonal moisture in the Trans-Pecos. It is also unknown how extension of the dryline into Mexico may influence the development of supercell thunderstorms in the Serrañías del Burro that impact border communities with severe hail and flash flooding.
Over the moderate terrain east of Carlsbad, the direction of slope is opposite the generally monotonic slope (upward to the west) typical of the Western Great Plains that is often noted as contributing to dryline development [2]. It is unknown whether this reversed slope between the Pecos River and Llano Estacado may contribute to "jumps" in the dryline progression in this region.
Most observational and modeling studies of drylines have focused on synoptically active drylines during the late afternoon (2000 to 0000 UTC) when severe convection is most likely. Even within the most intensively studied domain, no comprehensive field campaign has observed the dryline environment over a full 24 h or more of diurnal forcing with mobile, radar/lidar, and supplemental radiosondes. Transitions between diffuse, multi-step, and well-defined dryline patterns relative to diurnal and synoptic forcing have yet to be thoroughly documented. As suggested by [7], advection and differential diabatic heating both contribute to the development and motion of drylines. However their relative roles at different stages of the diurnal cycle and between cases are not well documented or understood.