Observations of the Evolution of the Nocturnal and Convective Boundary Layers and the Structure of Open-Celled Convection on 14 June 2002

1. Introduction

The International H₂O Project (IHOP_2002) took place in the southern Great Plains in the summer of 2002 (Weckwerth et al. 2004). The boundary layer evolution (BLE) component aimed to better understand the distribution of water vapor in the quiescent boundary layer and its evolution from the early morning through the development of the daytime convective boundary layer (CBL). These missions are the first time the boundary layer has been studied with such a dense dataset.

The structure of the boundary layer over land during synoptic-scale high pressure conditions has a well-defined diurnal evolution and has been described in terms of the vertical profile of virtual potential temperature θ_υ (Stull 1988) (Fig. 1). During the early morning, the stable nocturnal boundary layer (NBL), characterized by an increase of θ_υ with height, occupies the lowest few hundred meters and is overlaid by a neutrally stratified residual layer (Fig. 1a). Shortly after sunrise thermals begin to form, which transport warm air from the surface upward and cause entrainment of warm air from above the nocturnal inversion into the NBL. This convective overturning leads to the gradual erosion of the stable layer (Garratt 1994; Angevine 2008). Mechanical turbulence generated by large wind shear in the NBL can also lead to mixing. Recent work has suggested that shear-driven turbulence plays a larger role than convective turbulence in the warming and destruction of the NBL (e.g., Angevine et al. 2001; Lapworth 2006). As the stable layer is mixed out, a shallow CBL (or mixed layer) develops at the surface, where θ_υ is approximately constant with height (Fig. 1b). The new CBL grows through the residual layer as a result of entrainment and mixing until it reaches an overlying stable layer or capping inversion, which inhibits further vertical development (Figs. 1c,d).

There have been many observations of horizontal variability in humidity within the CBL due to thermals and horizontal convective rolls (e.g., Warner and Telford 1963, 1967; Crum et al. 1987; LeMone and Pennell 1976; Weckwerth et al. 1996). However, there have been few observations of the variability in humidity during the early morning prior to the formation of thermals and of the vertical and horizontal structure of the water vapor field during the transition from the NBL to the CBL. A common observation is the decrease of humidity with height in the convective boundary layer (e.g., Mahrt 1976; Crook 1996; Linne et al. 2007). It has been suggested that the vertical gradient is a result of entrainment of dry air at the top of the boundary layer outweighing the moistening by surface fluxes and the vertical transport of this moist air (Mahrt 1976). Mahrt (1991) suggested that the diurnal evolution of the boundary layer could be characterized by two regimes. The moistening regime occurs during the early morning and is associated with large surface evaporation. The drying regime later in the morning is associated with entrainment at the top of the boundary layer after the NBL has been eliminated. For example, Grossman and Gamage (1995) observed a period of drying during the rapid growth of the CBL as it encountered an elevated mixed layer (EML). Penetrative convection caused the downward mixing of drier air from the EML. Horizontal advection can also play a large role in the humidity structure of the boundary layer (Mahrt 1976; LeMone et al. 2002).

The organization of boundary layer convection often occurs in the form of two-dimensional horizontal convective rolls (cloud streets) or three-dimensional cells and is frequently observed in satellite images, particularly over the oceans (Atkinson and Zhang 1996). Atkinson and Zhang (1996) presented a comprehensive review of mesoscale shallow convection and summarized the characteristics of rolls and cells. They concluded that in general rolls tend to occur in environments of moderate wind shear (1–10 m s⁻¹ km⁻¹) whereas cellular convection usually occurs in a weakly sheared environment (less than 2 m s⁻¹ km⁻¹). However, more recently, Weckwerth et al. (1999) investigated the evolution of rolls over land and showed that they formed in even minimal wind speed and shear conditions. They found that roll formation was controlled solely by the magnitude of the buoyancy flux. Roll dissipation was controlled by the convective instability of the boundary layer and its evolution into either open cells or unorganized random convection was controlled by the magnitude of the wind speed.

Over the oceans, open-celled convection is recognized by the honeycomb or hexagonal organization of the clouds, associated with ascending motion in a ring around the outside and descending motion in the center (Agee et al. 1973; Agee and Dowell 1974). The convection is driven by heating from below as cold air moves over relatively warm water. The average diameter and convective depth for open cells is 30 and 2 km, respectively. There have been few observations of open-celled convection over land similar to that seen over the oceans. However, several early radar studies investigating the development of the clear-air convective field over land reported similar hexagonal structures to those over the oceans, but on a much smaller scale (e.g., Hardy and Ottersten 1969; Konrad 1970). The cells observed by Hardy and Ottersten (1969) had diameters between 5 and 10 km with depths of 1–2 km and were detectable because of the scattering from insects within the updrafts of the thermals. The cells were composed of updrafts around the edges and downward motion in the center. The cells observed by Weckwerth et al. (1999) had diameters of ∼2.4 km and those reported herein were 3 km on average.

This paper describes the development of the boundary layer over an essentially flat land surface from before sunrise to the time when it became a mature convective boundary layer, using a synthesis of datasets gathered with multiple instruments on 14 June 2002. The early morning boundary layer is characterized by a deep moist layer and oscillations associated with gravity waves. The growth phase of the convective boundary layer is characterized by the evolution from horizontal convective rolls to open cells. Simultaneous measurements from radars, lidars, and aircraft provide insight into the structure of the cells and in particular, the horizontal and vertical distribution of humidity. This is one of the few studies showing horizontal maps of the moisture in the CBL from an airborne water vapor lidar.

The structure of the paper is as follows. The layout and description of instrumentation are described in section 2 and the general meteorological situation in section 3. Observations of the evolution of the early morning boundary layer are presented in section 4, the development of the convective boundary layer in section 5, and the characteristics of the open cells in section 6. A summary of the observations are given in section 7. All times are in Central Standard Time (CST, UTC − 6 h), local solar noon was 1242 CST, and all heights are given as above ground level (AGL) unless otherwise stated.

2. Details of instrumentation

A map of the location of ground-based instrumentation and aircraft flight tracks on 14 June 2002 is shown in Fig. 2. The Homestead Profiling Site (HPS) accommodated the following instruments within 50–100 m of each other: the National Aeronautics and Space Administration Goddard Space Flight Center (NASA GSFC) Scanning Raman lidar (SRL) and Holographic Airborne Rotating Lidar Instrument Experiment (HARLIE), the University of Massachusetts Frequency Modulated-Continuous Wave (FMCW) radar, and the National Center for Atmospheric Research (NCAR) Integrated Sounding System (ISS).

3. General meteorological situation

On 14 June 2002, a ridge of high pressure was oriented north-northwest to south-southeast across the IHOP_2002 domain and there were no active convective systems or large-scale boundaries present. A band of stratocumulus clouds covered the southwest region of the domain during the early hours of the mission (Fig. 3a) but advected southward to leave a mostly cloud-free domain throughout the day (Fig. 3b). Pyranometer data from the MIPS surface station shows that solar radiation was intermittently reduced because of cloud cover during three periods: 0815–0855, 1110–1135, and 1300–1320 CST (Fig. 4a). The stratocumulus band and subsequent filaments to the north of the band were responsible for the reduction in radiation during the morning. A few very small cumulus clouds were visible in satellite images during the early afternoon.

The evolution of the wind profile is shown in Fig. 4b. Relatively strong (6–7 m s⁻¹) low-level northeasterly winds developed at 0630 CST, persisting until 0840 CST and then decreasing to ∼2 m s⁻¹ by 1000 CST. As the wind speed decreased the wind direction veered to the east, then to the southeast, and to the south by 1200 CST. The structure of the developing convective boundary layer (CBL) was influenced by the evolution of the winds. The winds aloft (1–1.5 km) were generally from the north or northeast, with a similar decrease from 7 to 2 m s⁻¹ between 0400 and 1000 CST.

In situ measurements from the King Air and P3 illustrated the mesoscale variations in temperature and moisture across the domain during the morning between approximately 0545 and 1140 CST (not shown). The northern leg, AB (see Fig. 2), was cooler than the central and southern legs, CD and EF, respectively. Surface analyses showed that these variations were caused by cooler air advecting into the northeast of the domain during the morning between 0600 and 0900 CST (Couvreux et al. 2009).

There was also a large east–west gradient in humidity, which intensified with time. The humidity was greatest in the west and decreased gradually toward the east, in contrast to the climatological gradient that characterizes the Great Plains during the spring and summer months (Dodd 1965; Hoch and Markowski 2005). Couvreux et al. (2009) investigated the nature of this mesoscale gradient in humidity and found that it was a result of the distribution of precipitation from previous convective events, soil moisture content, and advection.

4. Evolution of the early morning boundary layer

The evolution of the early morning boundary layer on 14 June 2002 in the area around the HPS was observed in detail by a combination of sensors (Fig. 5). The height of the NBL is often difficult to define but is usually taken to be the top of the nocturnal inversion or the height of a low-level wind speed maximum (Mahrt 1979; Stull 1988). The 0415 CST sounding shows that the wind speed increased sharply in a shallow layer from the surface to approximately z = 50 m where it reached a maximum of 4.8 m s⁻¹. It then decreased with height up to 200 m. The maximum was similar at 0530 CST, but located at z = 110 m. The increase of θ_υ with height in this layer indicates the stability associated with the NBL. The figure shows that the water vapor mixing ratio q decreased with height within and above the NBL to minimum values at approximately z = 120 m at 0415 CST and 175 m at 0530 CST, then increased to maximum values at 230 and 330 m, respectively. This is in contrast to the profiles observed by Mahrt (1979) where q increased with height in the region of the low-level jet and then decreased rapidly above.

A deep moist layer (q > 8 g kg⁻¹) extended from the lowest range gate (315 m) of the SRL to an altitude of 1.1 km from the start of observations at 0415 CST until just after sunrise at 0530 CST. This was probably a residual layer, connected with the previous day’s mixed layer.

The magnitude and structure of the profile of q is consistent between the SRL and the radiosondes. Both datasets show that there was a gradual decrease in q with height through the residual layer from ∼10 g kg⁻¹ at z = 300 m to ∼8 g kg⁻¹ at 1 km and that it was embedded with several layers of relatively moister air. At 0415 CST, there were maxima in the radiosonde profile at z = 230, 500, and 710 m, but there was only one maximum in the SRL profile at 610 m. At 0530 CST, there were maxima in the radiosonde profile at z = 330 and 600 m and in the SRL profile at 390 and 670 m. These differences are expected because the radiosondes were not launched directly above the SRL and they drifted horizontally with the wind by typically 600 m during the first 800 m of ascent.

There was significant drying in the upper half of the residual layer from 0525 CST, coinciding with sunrise and a change in the wind direction from the north-northeast to the northeast (Fig. 4b), advecting drier air into the domain. The depth of the residual layer, determined from the sharp gradient in q from the SRL, decreased sharply from z = 1.1 km to approximately 0.7 km by 0530 CST. The mixing ratio decreased rapidly with height above 0.7 km before increasing again within a shallow elevated moist layer between 1.1 and 1.4 km. The elevated layer coincided with a region of higher wind speed, backing in the wind direction, and a stable layer.

The sensible heat flux became positive at 0615 CST, 50 min after sunrise (Fig. 5). Angevine et al. (2001) defined this time as the crossover and found it to occur on average 1.3 h after sunrise. The latent heat flux also began to increase rapidly at this time. The development of a CBL began soon after crossover, but its depth, z_i, is difficult to define because of the data gap below the minimum range of the SRL. However, there is an indication of the growth of the CBL from three point measurements: z_i = 60 m at 0702 CST measured by the WCR, 100 m at 0730 CST measured by the ISS sounding, and 394 m at 0815 CST observed as a sharp transition in the entire day’s time series of vertical velocity from the P3. Angevine et al. (2001) defined the CBL onset to be the time when the nocturnal inversion has been eroded and found that the median onset time occurred 2.7 h after sunrise. In this case, the decrease of θ_υ with height (Fig. 5) shows that the nocturnal inversion had been fully eroded by 0730 CST, 2 h after sunrise. It is possible that the strong surface winds added some mechanical mixing and enabled the faster erosion of the NBL.

During the early growth of the CBL from 0630 to 0830 CST the latent heat flux increased from 30 to 160 W m⁻². However, there was a decrease in q with time at all levels within the lowest 300 m. The ISS surface station measured a decrease in q from 11.2 to 10.5 g kg⁻¹ and TAOS measured a decrease at z = 100 m from 10.6 to 9.3 g kg⁻¹. The drying is likely to be due to a combination of vertical transport of moisture by thermals, entrainment of drier air from above, and advection of drier air by the northeasterly flow. The remaining residual layer above the shallow CBL also experienced drying due to advection. The SRL measured a decrease in q from 9.8 to 9 g kg⁻¹ at z = 460 m. The 0730 CST sounding shows that q decreased with height from the surface to a minimum at approximately z = 230 m and then increased to a peak at the top of the residual layer at 600 m. This is in contrast to the mature CBL where the humidity usually decreases with height from the surface to the top of the CBL (Mahrt 1976), as will be shown for this case in section 5. The peak in q also coincided with a layer of enhanced wind speed.

Throughout the whole time period shown in Fig. 5 the vertical profile of humidity within the residual layer was also being perturbed by oscillations, which suggest the presence of gravity waves. The MIPS wind profiler observed similar structures during the same time period (Fig. 6). Several elevated layers of the increased signal-to-noise ratio (SNR) can be seen between 0545 and 0800 CST, the clearest ones being centered at altitudes of 0.5 and 0.8 km between 0545 and 0630 CST (Fig. 6a). Although it is difficult to distinguish individual peaks and troughs in the SNR field because of the lower resolution of the profiler than the SRL, the corresponding vertical velocity field reveals the oscillations very clearly (Fig. 6b). Alternating periods of upward and downward motion (maximum of +0.5 m s⁻¹ and minimum of −0.4 m s⁻¹) can be seen from the first gate at 143 m to the maximum extent of the data at 1.5 km. Convective updrafts and downdrafts start to dominate the signal below 500 m from 0645 CST onward, but the gravity wave oscillations are still present at higher altitudes. The average period of oscillation between 0545 to 0800 CST estimated from the vertical velocity data is approximately 19 min, which equates to an average frequency of 8.6 × 10⁻⁴ Hz. The Scorer parameter (Lin 2007) was calculated using the frequency information and sounding data and the analysis showed that the atmosphere was conducive to gravity wave propagation.

5. Evolution of the convective boundary layer

The continued development of the CBL after it had eroded the NBL was observed simultaneously by multiple instruments at the HPS. Figure 7 shows the mixing ratio field between 0600 and 1300 CST from the SRL, TAOS, and ISS radiosondes. The additional lines and symbols represent the depth of the CBL derived from the MAPR, FMCW, HARLIE, WCR, ISS, and the SRL itself. The top of the CBL only becomes easily discernible once it erodes through the residual layer at 0930 CST. It increases in altitude between 0930 and 1300 CST from approximately z = 750–1600 m, equating to a growth rate of 6.7 cm s⁻¹. This is comparable with other remote sensing observations of the growth of the CBL (e.g., Boers and Eloranta 1986; Angevine et al. 1994; Cohn and Angevine 2000). The growth is a result of entrainment of air above the CBL by overshooting thermals. Throughout this time period the SRL mixing ratio at the top of the CBL is characterized by alternating regions of moist domes and dry intrusions, which will be discussed further in section 6.

During the growth of the CBL between 0830 and 1300 CST, the humidity continued to decrease within the lowest 300 m AGL. The latent heat flux increased steadily from 160 W m⁻² at 0830 CST to a peak of 250 W m⁻² at 1145 CST, then leveled off at this value until it started decreasing an hour later. Advection of drier air was probably dominant until the wind veered to the southeast at 1100 CST. However, strong convective mixing and the vertical transport of moisture was probably the main mechanism for the drying at low levels between 1100 and 1300 CST. The SRL measured a decrease of mixing ratio with height throughout the CBL during this period as a result of entrainment of dry air at the top of the CBL, consistent with previous studies (e.g., Mahrt 1976; Crook 1996).

The magnitude of mixing ratio between the majority of overlapping datasets agrees to within 1 g kg⁻¹, which is within the range of accuracy of the instruments. It is unreasonable to expect any better agreement, particularly between the ground-based and airborne instruments, as they are not sampling the same volume of air because of the differences in time and space of the measurements.

Quantitatively defining the top of the CBL has been the subject of much research since it is an important parameter in applications such as pollution dispersion modeling. Reviews of remote sensing methods to derive the CBL depth (or mixed layer height) can be found in Seibert et al. (2000) and Emeis et al. (2008), for example. The altitude of z_i, and the top and bottom of the entrainment zone, were estimated from profiles of θ_υ from the ISS soundings, based on the definition given in Wyngaard and LeMone (1980) and Stull (1988). One-minute averages of z_i were determined from HARLIE aerosol backscatter profiles using an automated algorithm that employs the Haar wavelet technique documented by Davis et al. (2000). The algorithm finds the center of the aerosol gradient at the top of the CBL. The layer identified by the algorithm between altitudes of 0.6 and 0.8 km from 0600 to 0930 CST is not the CBL top, but corresponds with the top of the residual layer identified by the SRL. The maximum gradient in mixing ratio at the top of the CBL for each vertical profile of the SRL was used to construct its time series of z_i. The reflectivity data from MAPR and FMCW was much noisier; therefore, z_i was subjectively determined by estimating the peak in reflectivity every 5 min, based on the technique documented by Angevine et al. (1994). They suggested that the peak in SNR occurs just above the inversion base where the humidity gradients are strongest. The WCR value of z_i was derived from the reflectivity field using the method described in Geerts and Miao (2005) and Miao et al. (2006).

The average rate of growth of the CBL was comparable for all the instruments (Fig. 7). The time series of z_i from the SRL matched well with the instantaneous values determined from the radiosondes. The FMCW and MAPR mirrored the undulations in the SRL and although the absolute values of z_i were slightly lower, they were within the range of the entrainment zone, giving confidence in any one of the measurements. The HARLIE estimates of z_i were consistently higher than the SRL and the peaks and troughs did not correlate well with the other instruments. The main reason for the discrepancy is likely to be because the lidar scans at 45° elevation and the data are averaged over 360°. The WCR values of z_i are the lowest of all the instruments. The differences in absolute CBL height are mostly due to the sampling characteristics of the remote sensors, as well as the temporal resolution and the details of the algorithms used to define the CBL.

Continuous 1° elevation plan position indicator (PPI) scans from the DOW radar revealed a striking evolution of convective features throughout the morning and early afternoon (Fig. 8). The first clear-air signals were observed from 0700 CST, but they had no particular organization (Figs. 8a,b). The radar echoes developed into horizontal convective rolls 40 min later at 0740 CST (Figs. 8c,d). The rolls were most distinguishable at 0830 CST after increasing in spacing and size as the CBL depth increased (Figs. 8e,f). The rolls began to evolve into open-celled convection from 1000 CST, soon after the winds in the lowest 500 m had decreased to 2 m s⁻¹ and veered to the east (Fig. 4b, section 3). The cells were particularly well defined by 1200 CST (Figs. 8g,h). This is consistent with previous observations of the development of the CBL, for example, those over Florida presented by Weckwerth et al. (1999). The undulations observed by the SRL at the top of the CBL and mirrored in the MAPR and FMCW time series of z_i between 1100 and 1300 CST (Fig. 7) are, for this case, a result of the open-celled organization of the CBL. The following section presents a unique insight into these cellular structures by comparing collocated measurements from the DOW radar, MIPS wind profiler, SRL, and aircraft.

6. Collocated measurements of open-celled convection

Examples of the open-celled convection observed by the DOW radar are shown in Fig. 9. The cells are hexagonal-shaped regions of high reflectivity surrounding a low reflectivity core. The high reflectivity is caused by large backscatter from insects that have been forced together by convergence and swept into the updraft branches of the cells (Wilson et al. 1994; Achtemeier 1991). Similarly, the low-reflectivity regions are associated with a broader region of downward motion causing the dispersion of insects. Cells existed with a variety of diameters, ranging from 2 to 4 km, comparable to other observations such as those of Konrad (1970) and Weckwerth et al. (1999). The examples in Fig. 9 illustrate a few well-defined cells; however, there was considerable variability in size and shape.

The MIPS wind profiler was located approximately 10 km east/16 km south of the DOW and 5 km west/10 km south of the ISS (Fig. 2). The wind profiler also observed structures associated with the open-celled convection (Fig. 10) as the cells advected with the main flow over the profiler. The top of the CBL was delineated by a layer of high SNR that slowly increased in height from an altitude of 0.7 km at 1000 CST to 0.9 km by 1100 CST and 1.1 km by 1215 CST. The rate of growth of the CBL was comparable to the other datasets shown in Fig. 7. Similarly, the increase was not linear, there were large variations associated with thermals within the cellular structures. The large backscatter was a result of the sharp gradient in humidity at the top of the CBL and the mixing between thermals and the environment (Angevine et al. 1994; White et al. 1991).

The inverted U-shaped echoes at 1115, 1127, 1139, and 1215 CST stand out very clearly and represent regions where this mixing is taking place. The minima in reflectivity below the echoes are due to the homogeneity in humidity within the cores of the thermals. Using the observed average wind speed at the altitude of the echoes (approximately 1–2 m s⁻¹ between z = 0.6 and 1.1 km, Fig. 4b) and the time taken for the features to pass over the instrument (∼13 min, assuming a steady state), their widths were between 800 and 1600 m. Such echoes, of comparable scales, were first observed by backscatter radars over 40 years ago in the CBL (Hardy and Ottersten 1969; Konrad 1970), and within cumulus clouds (Harper et al. 1957) where they were termed “mantle echoes.”

Collocated data from the wind profiler and the DOW radar are presented in Fig. 11 to show that the mantle echoes were associated with the updraft branches of the open-celled convection. The mantle echoes at 1115, 1127, and 1139 CST were all associated with updrafts (black contours) that started at the lowest level of the wind profiler data (∼200 m) and extended into the low SNR region beneath the top part of the echoes (Fig. 11f). Most of the updrafts extended upward all the way to the underside of the echoes. The echo at 1127 CST consisted of two updraft cores. The corresponding DOW scans show that the wind profiler was located in a region of high reflectivity corresponding to the edge of one of the open cells (Figs. 11b–d). It remained within the updraft branch of the cell as the cell advected toward the northeast.

Columns of high SNR extend from approximately 200 to 500 m within the updrafts. It is likely that these signals were dominated by Rayleigh scattering associated with plumes of insects being transported vertically by the updrafts, similar to the measurements of the WCR. Bragg scattering is unlikely because these are solid signals without minima at their centers like the mantle echoes. The humidity of the thermals was probably still quite homogeneous at these lower altitudes.

The updrafts were estimated to be between 540 and 900 m in width, using a wind speed range of 1.5–2.5 m s⁻¹ (Fig. 4b) and a time to pass over the wind profiler of approximately 6 min. The mean width of the updraft plumes, calculated from the WCR using the method described in Miao et al. (2006), was found to be 0.34z_i and largely independent of height. Using the CBL depth of z_i = 1 km at this time gives smaller updraft widths of 340 m. The wind profiler widths are likely to be slightly overestimated because of the assumption made in the calculations and the lower resolution of the instrument compared with the WCR.

The wind profiler observed a broad region of downdraft between 1139 and 1200 CST, which was associated with high SNR throughout the whole depth of the CBL (Fig. 11f). The high SNR is most likely caused by Bragg scatter resulting from inhomogeneities in the refractive index of the descending air. The wind profiler was located under a low DOW reflectivity region within an open cell at this time (Fig. 11e). As the cell advected toward the northeast, the wind profiler remained under the low-reflectivity region. Conversely, the downdraft observed at 1100 CST was only a very narrow feature below z = 500 m and was surrounded by updrafts on either side (Fig. 11f). The wind profiler was located beneath the boundary between the high and low DOW reflectivity regions (Fig. 11a). A series of DOW images helped to clarify that as the cell moved to the northeast the wind profiler was only briefly within the downdraft region. The downdrafts were stronger toward lower altitudes and were strongest below 400 m.

Large variations in the humidity structure of the CBL were measured by the SRL during the time of open-celled convection. Collocated data from the SRL and the DOW radar are shown in Fig. 12. The top of the CBL was characterized by domelike structures of high humidity separated by narrower regions of drier air. The domes are similar in shape to the mantle echoes seen by the wind profiler (Fig. 11); however, they are much broader. Their widths are estimated to be between approximately 3 and 4.5 km, using a wind speed range of 2–3 m s⁻¹ at the top of the CBL (measured by ISS radiosondes) and a time period of 25 min. It is also possible to distinguish a finer-scale structure to the domes themselves, where there are alternating thin columns of higher humidity and drier air. The moist columns are likely to be associated with individual thermals.

The SRL observed a narrow intrusion of dry air at 1039 CST when it was located under a boundary between high and low DOW reflectivity (Figs. 12a,g). Similarly a decrease in the height of the CBL and a zone of drier air was observed at 1236 CST, while the SRL was located under a low reflectivity region (Figs. 12e,g). At 1109 and 1251 CST, the SRL observed narrow moist columns of air that extended from 300 m to almost the top of the CBL. The SRL was located under regions of high reflectivity at the time (Figs. 12b,f). Similarly, the domes of humidity at 1139 and 1209 CST also corresponded to relatively higher zones of reflectivity (Figs. 12c,d,g). It is interesting that the open-cell organization observed by the radar in the lowest few hundred meters of the CBL corresponds so well with the features observed by the SRL at the top of the CBL. The humidity domes correspond to the updraft branches of the cells, where thermals transport moist air upward and the regions between the domes correspond to downdraft zones within the cells where the convective motions entrain drier air down into the CBL.

In situ measurements of q′, θ′, and w from the King Air and reflectivity from the DOW radar are shown in Figs. 13a–d. Seven updrafts, labeled I to VII, were observed during this transect. Most of them were associated with positive fluctuations in θ confirming that these features were thermals, transporting warm air upward. Most of the updrafts correspond to a region of enhanced reflectivity associated with the edges of the open-celled convection (Fig. 13e). The sequence of updrafts and downdrafts between 19 and 25 km east of the DOW is particularly distinct. The updrafts correspond to the edges of cells and the downdrafts correspond to the low reflectivity center of the cells. The updrafts are not strongly correlated with q′, although there is an indication of relative maxima at some of these times. The variations in q′ are generally broader than those in θ′ and w, which may explain the weaker correlation. There is a stronger correlation between q′ and w in the downdraft zones, representing the entrainment of dry air into the CBL.

Mixing ratio from LEANDRE II is shown in Fig. 13f. Despite the restricted range and reduced resolution of LEANDRE II compared to the DOW radar and King Air in situ measurements, the horizontal moisture fields show variations that may be associated with the open cells. The regions of high and low humidity are in approximately the same location as the regions of high and low reflectivity, respectively, but are sometimes broader. For example, the minimum in q located between x = 22 to 24 km and y = −5 to −4 km corresponds to an area of low reflectivity. Similarly, the maximum in q close to updraft V and the minima in q on either side correspond to maxima and minima in reflectivity.

Transects of reflectivity and vertical velocity from the WCR for the same time period are shown in Fig. 14. The WCR echoes, like those from the DOW radar, are due to scattering from insects. Geerts and Miao (2005) and Miao et al. (2006) showed that the plumes of high reflectivity correspond to the vertical displacement of the insects by thermal updrafts. The velocity field reveals the vertical motions associated with updrafts and downdrafts. The WCR reflectivity plumes correspond to the regions of enhanced DOW reflectivity. Both the in situ measurements and the WCR show that updrafts I to V were narrow whereas updraft VI was much broader. The WCR shows this broad updraft existed over the whole depth of the cell (Fig. 14). The center of the cells correspond to areas of depressed CBL height and very few insects, indicative of subsidence.

A second set of collocated King Air and DOW radar observations are shown in Fig. 15. Updrafts I, II, and III correspond to positive fluctuations in θ and q and enhanced values of DOW reflectivity. The θ fluctuations were similar widths to the updrafts (∼500 m); however, the q fluctuations were generally broader than the updrafts (∼1 km) as in Fig. 13. These observations correspond well with the LES simulations performed by Couvreux et al. (2005) for this case. They found that the CBL was characterized by larger scales for water vapor than for potential temperature and vertical velocity. One possible mechanism is the detrainment of moisture as thermals ascend in the CBL, similar to the process around cumulus clouds described by Perry and Hobbs (1996).

The reflectivity structures observed by the DOW radar were more unorganized at this time. Similarly, the corresponding reflectivity and velocity plumes observed by the WCR were inhomogeneous in both the horizontal and vertical (Fig. 16). More distinct cellular structures were observed along the second half of the transect. Updrafts IV and VI were clearly associated with the north–south-oriented edges of a hexagonal-like cell, large θ′, and relative maxima in q′. This larger cell consisted of smaller structures and updraft V corresponds to the edge of one of these features. Updraft VII is associated with a separate area of high reflectivity to the east of the large cell.

A narrow region of dry air was observed at 19.7 km east of the DOW (Fig. 15a), adjacent to updraft V. It corresponded to a region of low reflectivity in the DOW data, a downdraft of −1.4 m s⁻¹, and a θ′ value of +0.2 K. The WCR observed a narrow reflectivity and strong velocity echo associated with updraft V, followed by a region with only a few weak echoes and downward motion (Fig. 16). Couvreux et al. (2005, 2007) showed that in the middle of the boundary layer vertical velocity was positively skewed, representing a few strong and narrow thermals. However, the distribution of humidity was negatively skewed, reflecting the presence of narrow intrusions of dry air, which they termed dry tongues. The dry tongues are caused by moist thermals penetrating into the overlying inversion (entrainment zone) and then sinking back down, pulling dry air down with them into the boundary layer. The dry intrusion observed by the King Air is an example of one of these dry tongues.

The horizontal moisture field from LEANDRE II is shown in Fig. 15f. There is a weaker correlation between the moisture and reflectivity fields for this section; however, there is still an indication that the lidar is able to resolve water vapor variability on the scale of the open cells.

The results in this section are summarized in the schematic diagram shown in Fig. 17. The open cells are characterized by high DOW reflectivity branches around their edges and low DOW reflectivity centers (feature 1). The long axis of the cell edges consist of multiple thermals carrying warm, moist air upward. However, it is not clear how many thermals span the short axis (i.e., the width) of the cell edges. The hexagonal shape of the cells is idealized and there were no instances where the instruments observed a cell edge directly along its short axis. The updrafts measured by the King Air and the MIPS wind profiler were 500–900 m wide (feature 2, solid lines), but the humidity field measured by the King Air was broader, on the order of 1 km (feature 2, dashed lines). Along the open cell edges, at the top of the CBL the thermals overshoot the capping inversion and entrain air down into the CBL. The entrainment and mixing of environmental air with the air inside the thermals results in the mantle echoes observed by the MIPS wind profiler (feature 3). The mantle echoes were larger than the thermals, having widths of 800–1600 m. The edges of the open cells were also collocated with dome-shaped structures of higher humidity at the top of the CBL, with scales of 3–5 km (feature 4). It appears that the domes are accumulations of several columns of higher humidity from individual thermals.

7. Summary

There have been few documented observations of the variability in humidity during the early morning and of the vertical and horizontal structure during the transition from the NBL to the CBL. This case study on 14 June 2002 utilized simultaneous measurements from a large number of ground-based and airborne instruments to build a comprehensive picture of the evolution of the boundary layer, from shortly before sunrise through to the early afternoon. In particular the observations describe the vertical and horizontal distribution of humidity within the NBL and residual layer, and during the growth of the CBL.

The boundary layer observed on this day largely followed the classic diurnal evolution described by Stull (1988) and Garratt (1994). During the early morning there was a shallow NBL, and a deep moist residual layer that persisted for several hours before it became engulfed by the deepening CBL (Fig. 5). Advection of drier air into the region resulted in a minimum in the humidity profile in the middle of the residual layer and a peak at the top of the layer. Throughout the morning, the NBL and residual layer were perturbed by gravity waves, observed most clearly by the SRL, but they did not appear to have any affect on the development of the CBL.

A CBL started to develop about an hour after sunrise, but its growth was not well observed until it had mixed out the residual layer. After this time, the top of the CBL was clearly distinguishable because of the sharp gradient in temperature, humidity, and aerosol concentration separating the CBL from the free atmosphere (Fig. 7). Measurements of the altitude of the growing CBL by several different instruments were compared. There was a good correspondence between the average rates of growth compared with previous reported observations. Differences in the absolute altitudes of the CBL could mostly be explained by the different sampling characteristics of the instruments.

During the late morning and early afternoon, humidity in the CBL gradually decreased because of a combination of advection and mixing. There was a decrease in humidity with height, consistent with previous findings, but in contrast to the profile observed in the residual layer earlier in the morning. There were also significant horizontal variations in humidity. The SRL observed columns of higher humidity extending over a large fraction of the CBL associated with thermals. These were interspersed with intrusions of drier air. The King Air measurements showed that the humidity structures associated with thermals were broader than the widths of the updrafts, consistent with the modeling studies of Couvreux et al. (2005).

The organization of the CBL evolved from horizontal convective rolls into open-celled convection. Collocated measurements from the DOW radar, MIPS wind profiler, and SRL enabled a detailed analysis of the structure of the open cells. This is the first time the relationship between thermals, mantle echoes, CBL top, and open cells has been investigated. The results, summarized in Fig. 17, showed that the high reflectivity branches were associated with thermals transporting warm, moist air upward and the low-reflectivity centers were associated with drier, descending air. The reflectivity structures observed by the DOW radar in the lowest few hundred meters appear to be connected to the thermal processes and the humidity structures throughout the depth of the CBL.

The evolution of the thermal structures as they ascend in the CBL is not investigated in this paper. It is not clear why the humidity field of thermals is broader than the vertical velocity and temperature fields. A detrainment mechanism similar to that observed in cumulus clouds (Perry and Hobbs 1996) may be responsible. This is one of the most comprehensive datasets of open-celled convection over land. It would be interesting to further investigate whether such clear-air convective patterns may lead to deep moist convection, as sometimes observed for horizontal convective rolls (e.g., Weckwerth et al. 1996; Weckwerth 2000). Numerical simulations would aid in addressing these outstanding questions, as well as detailed analyses of two other BLE missions on 21 and 25 June during IHOP_2002.