IHOP_2002 boundary layer flights

Some 34 h of combined radar and in situ data were collected in the mature CBL in Kansas and Oklahoma in late-spring 2002 aboard the University of Wyoming King Air (WKA) aircraft. In all cases, the synoptic conditions were rather quiescent, and the 700–500-hPa surfaces were essentially barotropic, the CBL well capped, and the skies mostly cloud free.

The WKA carried a gust probe and measured temperature and humidity at a high frequency (25 Hz). High-rate humidity was measured both with a Lyman-α hygrometer and a Li-Cor, Inc., 6262 infrared absorption probe. Aircraft soundings from the ground to ∼500 m above the mixed layer were conducted regularly. The mature CBL was examined along three fixed tracks, each about 60 km long, over relatively flat terrain. During a single flight, a track was flown 12–18 times in either direction, between approximately 1300 and 1600 local solar time (LeMone et al. 2005, manuscript submitted to J. Hydrometeor., hereinafter LEM05). One of these tracks was oriented north–south in the Oklahoma Panhandle (the “western track”), over grassland, winter wheat, and sagebrush. The “central track” was east–west oriented in south-central Kansas, over mostly grassland and winter wheat. And, the “eastern track,” located 100 km east of the central track, contained slightly undulating land of grassland and crops. This track is the same as that used in Cooperative Atmosphere–Surface Exchange Study (CASES)-97 (LeMone et al. 2000). The WKA flew straight legs over the three tracks at several constant heights, ranging from ∼70 m AGL to just above the CBL. Above about 400 m AGL the flight legs followed a constant pressure level, which departed a little from a constant height AGL.

This study uses 10 flight legs along the western track on 29 May, 7 legs along the central track on 6 June, and 6 legs along the eastern track on 17 June 2002. On each of these days the wind was southwesterly, and the mean wind speed at the lowest flight legs was 4.9, 5.3, and 7.6 m s⁻¹ respectively. The mesoscale 900-mb temperature gradient over the flight tracks pointed to the west at about 9, 3, and 6 × 10⁻⁶ K m⁻¹, respectively, producing a northerly thermal wind of 2.7, 0.9, and 1.8 m s⁻¹ km⁻¹, respectively. The observed wind shear within the CBL or across the CBL top was relatively weak (<5 m s⁻¹ km⁻¹), and scanning ground-based radar images did not reveal any evidence for horizontal convective rolls, except on the windiest day (17 June). On that day, some linear organization is apparent (LEM05), as revealed by radar and visible satellite imagery (Fig. 2 in GM05).

The vertical air motion inferred from the WKA gust probe data is referred to as w. The aircraft motion is removed from the gust probe raw winds by means of aircraft speed and attitude parameters obtained from the inertial navigation system (Lenschow 1972). Further corrections are made based on multiple global positioning system receivers on the WKA. Potential temperature (θ) was derived from the reverse-flow thermometer and static pressure probe, and the water vapor mixing ratio (r) was from the Lyman-α hygrometer, after it was calibrated by a chilled mirror dewpoint sensor, which is more accurate but has a slow response time.

Wyoming Cloud Radar data

The radar used here, the Wyoming Cloud Radar (WCR), is a 3-mm (95 GHz) multiantenna Doppler radar (Pazmany et al. 1994; for details about the WCR see information online at http://atmos.uwyo.edu/wcr/). The key radar configuration used for this study is the profiling mode, with fixed antennas looking up and down from the aircraft. The pulse width is 250 ns for the flight legs studied here, thus the gate spacing is 37.5 m. Data are sampled at 15-m range intervals. The first reliable radar gates are centered at 120 (nadir) and 105 (zenith) m from the aircraft, resulting in an ∼225-m blind zone, approximately centered at flight level. For lowest flight legs (at about 70 m AGL), only the zenith antenna can be used, while for most flight legs in the upper CBL or above the CBL, only the nadir antenna is useful.

The along-track sampling rate corresponds with Δx ≅ 4 m depending on the speed of the aircraft and the radar processor load (which affects the sampling frequency, ranging between 25 and 35 Hz). All in situ variables are synchronized to the WCR reflectivity and vertical velocity data at this frequency. The beamwidth (half-power width) is 9–14 m at a range of 1 km for the nadir and zenith antennas, respectively; thus the along-track dimension is oversampled at most ranges of interest.

The WCR antennas have a sensitivity threshold of about −27 dBZ at a range of 1 km. Between echo plumes (Fig. 1), the density of scatterers is often rather low, and thus the returned power is close to the minimum detectable signal; for one flight over the western track, on 29 May 2002, the probability density function of reflectivity peaks at about −17 dBZ in the lower CBL, decreasing to −27 dBZ in the upper CBL (GM05). WCR data analyzed here are subjected to a simple echo detection filter: the power yielding the minimum detectable reflectivity is set to be the mean system noise plus two standard deviations. All data points with received power below this threshold are eliminated. A small but significant fraction of the CBL between plumes is below this threshold, thus the Doppler velocity spectrum is not available for the entire CBL.

The WCR Doppler velocities are corrected for aircraft motion (Leon et al. 2006). For the flights examined in this paper, Doppler velocities of the earth's surface indicate that the mean antenna-pointing angles are within 1° of the vertical axis in both along- and across-track directions, such that during level flight, the radial velocities are very close to true vertical motion. The instantaneous antenna-pointing angle is usually slightly off vertical, resulting from aircraft attitude changes in turbulent air. However, it is important that the average offset from vertical be close to zero, because while contamination of the (known) horizontal aircraft motion into an off-vertical radar beam can be extracted, the contamination by the horizontal wind cannot, at least not exactly, because the horizontal wind can only be estimated, for example, from a nearby sounding. The accuracy of the WCR-inferred vertical velocity is discussed more in GM05.

The WCR vertical velocities were further adjusted for the vertical motion of scatterers relative to the air, using Eq. (2) in GM05. GM05 find that insects tend to subside (as measured by the WCR) in the presence of updrafts (as measured by the WKA gust probe). They show that this downward bias is a strong function of the vertical air motion. The adjusted WCR vertical velocities have a leg-mean uncertainty of 0.2 m s⁻¹ (GM05), although instantaneous errors can be much larger.

Radar-inferred CBL depth

Echo plumes clearly mark the vertical extent of the CBL (Fig. 1). This enables an objective determination of the CBL depth z_{i_WCR} from the WCR reflectivity profile. The scatterer density decreases rapidly across the CBL top. Also, the radar noise, expressed in reflectivity units, increases with the square of the radar range. For these reasons we found z_{i_WCR} to be best defined as the level at which the zenith-beam WCR reflectivity reaches a minimum.

The high-resolution reflectivity profiles do contain local minima, because the insect density and size distribution can vary considerably among adjacent resolution volumes (section 2b). Therefore, to determine z_{i_WCR}, reflectivity data are first averaged in the along-track direction for 10 s (∼830 m). Further smoothening occurs in the radial direction from 15 to 45 m, which is close to the physical gate spacing (37.5 m). All reflectivity averaging occurs in native units (mm⁶ m⁻³, not dBZ).

In regions with very weak echoes between plumes, the resulting z_{i_WCR} values can be unrealistically small, for example, less than 50% of the typical z_{i_WCR} over plumes. Therefore, a case-specific threshold value for the profile-mean reflectivity within the CBL is assumed (e.g., −25 dBZ) for a z_{i_WCR} estimate to be valid. To improve a few z_{i_WCR} estimates, two more conditions are imposed: the signal-to-noise ratio must exceed 6 dB at z_{i_WCR}, and the percentage of good points in the averaging domain (250–350 points in the along-track direction and 3 in the radial direction) must exceed 30% at the level just below z_{i_WCR}. Values of z_{i_WCR} are shown as white plus signs in Fig. 1; the CBL depth and its local variations appear to be captured well.

Echo plumes and CBL depth

An example of the variation of z_{i_WCR} is shown in Fig. 2, for a sequence of seven flight legs along the western track on 29 May 2002. This example was chosen because of the remarkable north–south gradient in CBL depth. Clearly, z_{i_WCR} is less at the southern end of the track, where the soils were locally saturated and large puddles were present because of recent rains (LeMone et al. 2005; GM05). The CBL depth also increases slightly with time as the day progresses into the early afternoon.

Aside from this regional trend, the scatter of z_{i_WCR} estimates around the mean at a given latitude in Fig. 2 is at least in part because of convective activity in the CBL—plumes of rising motion tend to concentrate insect scatterers and yield a deeper CBL, and compensating subsidence and/or downward entrainment tends to suppress the CBL top. The correlation coefficient between detrended z_{i_WCR} and WCR reflectivity (averaged over corresponding 10-s intervals) is +0.46, +0.23, and +0.16 on average for all flight legs used in this study for 29 May, 6 June, and 17 June, respectively. Cross sections such as that in Fig. 1 confirm this trend. The correlation between echo-top height and WCR vertical velocity is generally positive as well, but weaker, possibly because of a time lag between updrafts and CBL top up thrusts.

To ascertain that the echo plume top z_{i_WCR} corresponds to the CBL top, we compare it with the thermodynamically defined CBL depth (z_{i_TH}) from a sounding released at “Homestead” in the middle of the western track on 29 May. The depth z_{i_TH} is the top of the mixed layer, that is, the base of the layer marked by a rapid increase in potential temperature θ. This level also marks the base of a rapid decrease in mixing ratio q, in this case. The z_{i_TH} value corresponds well with z_{i_WCR} values at this location and time of day (Fig. 2). Backscatter lidar data have long been used to determine z_i, because the CBL top is marked by a rapid decrease of backscatter signal with height (Ferrare et al. 1991; Marsik et al. 1995; Cooper et al. 1996; White et al. 1999; Davis et al. 2000, Cohn and Angevine 2000). The z_{i_WCR} values also compare well with backscatter lidar z_i estimates (z_{i_lidar}). These are derived from the high-resolution Doppler lidar (HRDL) on board an aircraft flying on the same track, well above the CBL (Fig. 2). The lidar reveals the same north–south gradient in CBL depth and the gradual deepening during the course of the day, although early on it seems to underestimate the CBL depth over the wet southern part of the track. Both the mean and the standard deviation of z_{i_lidar} values closely match those of the z_{i_WCR} values. One distinction is that the local CBL deepening trend during the duration of this intercomparison is larger in the lidar data than in the radar data.

In short, on fair-weather days with a well-defined CBL and sufficient scatterers, the WCR accurately measures the CBL depth and its small-scale variability. The average z_{i_WCR} values, based on all 23 flight legs, are 1031, 1049, and 1083 m on 29 May, 6 June, and 17 June, respectively. Only for the last three legs on 17 June does this height differ significantly from z_{i_TH}, which is estimated at 1240 m for that period (LEM05). More scatterers were present on 29 May (Figs. 1 and 2) than the two other days; on 29 May the mean reflectivity between 400 and 500 m AGL was 2.9 dB higher than on 6 June and 4.2 dB higher than on 17 June for all flight legs used in this study. On the last three legs of the 17 June flight, the scatterer density was insufficient to define z_{i_WCR} in the interplume regions, but was still sufficient to define echo and updraft plumes and to examine their properties.

Echo plume definition

Conditional sampling techniques have been used to define and characterize CBL thermals (e.g., Lenschow and Stephens 1980; Young 1988). We use a similar technique to distinguish echo plumes from their environment based on radar reflectivity. An echo plume is defined as a region of vertically averaged radar reflectivity (Z) that is some threshold value above the flight-leg-mean radar reflectivity (Z_m). The vertical average comprises 14 gates of radar reflectivity for each profile (i.e., a depth of 210 m). Normally the first seven nadir beam gates and the first seven zenith beam gates are used. In this case the WCR data are centered at the level of in situ data, and the averaging depth (including the 225-m-deep blind zone) really represents 435 m. For very low (high) flight levels in the CBL, all 14 gates had to be drawn from the zenith (nadir) beam. (Very low is defined as less than 0.3z_{i_WCR}, and very high is more than 0.7z_{i_WCR}.)

This averaging depth mainly represents a compromise between capturing a sufficient depth and avoiding drawbacks that come with covering the entire CBL from the ground to z_{i_WCR}. These drawbacks arise when echo plumes occupy only part of the CBL or are tilted, in which case a large averaging depth exaggerates the plume width. Consider the case of a shallow, incipient thermal. The vertically averaged reflectivity (mm⁶ m⁻³) is strongly dominated by the highest values in the profile, thus it is likely that if the averaging depth were to cover the entire CBL, this incipient thermal would be labeled as an echo plume, even if the flight level is well above the actual plume and in situ measurements do not characterize this thermal.

A low-pass filter is applied to Z, with a cutoff wavelength of 120 m, in order to remove the contribution of the turbulence within the inertial subrange, which is not driven by buoyancy. Young (1988) chooses 0.1z_i as the low-pass filter cutoff wavelength for this purpose. In our case, 0.1z_i is close to 120 m. This horizontal distance also matches the vertical distance between the in situ data and the first radar data. We use a finite impulse response filter based on a Kaiser's window with a rather gradual cutoff in the frequency domain.

The filtered reflectivity time series for each flight leg further is detrended. We define echo plumes based on the criterion that the vertically averaged along-track-filtered and detrended reflectivity (Z_df) be equal to or greater than a threshold value, which is a certain value ΔZ (ranging from 0 to 5 dB) greater than Z_m, over a distance of at least 120 m. The value ΔZ is referred to as the (threshold) plume strength. Background or “interplume” regions are defined as the segments where Z_df is less than the specified threshold. Plumes less than 120 m wide are not counted as echo plumes, thus they are included in the background.

An example of the filtering and thresholding procedure is shown in Fig. 3 for a flight leg flown at ∼800 m above the surface over the western track. The procedure nicely captures the echo plumes apparent in Fig. 3b, especially those that are continuous across the blind zone, that is, the objectively defined echo plumes are very close to those resulting from visual inspection of the cross section. This method is applied to all 23 flight legs used in this study.

Echo plume size characteristics

Echo plumes have a range of widths and spacings. We normalize these horizontal dimensions by the CBL depth, estimated as the leg-mean value of z_{i_WCR}, except for the three last legs on 17 June, where we used z_{i_TH} from WKA soundings (1240 m). The likelihood of finding a plume, or an interplume region, of a given size roughly decays exponentially with size (Fig. 4). The most common size is the smallest one resolved. The widest plumes are about 2.5 km, which is a small fraction of the total length of the flight leg. A useful parameter for the observed distributions is the “half-width,” which is the distance at which the best-fit exponential curve attains 0.5 times the value of the smallest resolvable scale (comparable to the “half life” of unstable isotopes). Assuming a plume strength ΔZ = 1 dB, the half-width of echo plume distribution is about one-third of that of interplume regions, that is, the space between the end of one plume and the beginning of the adjacent plume. The true plume spacing, measured from/to the center of adjacent plumes, is the sum of the interplume spacing and two halves of plumes, and its reciprocal is the number of plumes per unit distance. Using the half-width numbers given in Fig. 4a as representative values of plume and interplume widths, the true plume spacing (0.70 + 2 × 0.21/2 = 0.91) is close to the CBL depth itself; in other words there is about one plume per kilometer.

Echo plumes tend to be more narrow than the CBL depth, thus their aspect ratio (height: width) tends to be larger than 1.0. Clearly, plumes become narrower and the interplume regions wider when a more restrictive plume definition is used. At a plume strength ΔZ = 1 dB, 21% of the CBL is in plume and 79% is interplume. This is an average for data from 23 flight legs examined here, but the numbers are very consistent among these 3 days. At ΔZ = 3 dB, the in-plume percentage is 12%; at ΔZ = 5 dB it is 6%. The echo plume size distribution contines to resemble an exponential curve, and the plume half-width for the best-fit exponential curve decreases more or less linearly over this range of plume strengths (Fig. 4b).

The 2D structure of plumes can be described as well, using basically the same method used to define plumes based on Z_m (section 3a), but employing reflectivity data at all heights. First, the reflectivity is averaged every three profiles and every three radar gates, yielding a grid with a resolution of about 12 × 45 m². Then the reflectivity is low-pass filtered in the along-track direction, with a cutoff wavelength of 120 m, at each height separately; thus the smallest resolved 2D plume is 120 × 45 m². Plumes are defined as segments with some threshold reflectivity above the flight-leg-mean reflectivity at each height. This is illustrated in Fig. 5 for a threshold of +1 dB. Echo plumes evident in Fig. 5a are captured well in the binary analysis (Fig. 5b). They are spatially rather continuous entities, even across the blind zone, although some plumes do not occupy the full depth of the CBL.

We can use this 2D plume definition to estimate the variation with height of the typical plume width and interplume spacing; changes in the width and especially the number of plumes with height can affect flight-leg-mean plume thermodynamic anomalies. These anomalies, discussed below, will be plotted as a function of flight level, and the interpretation assumes that the population of plumes is roughly the same at each level. Our analysis over 3 days shows that the normalized plume width tends to increase slightly with height within the CBL (Fig. 6), as has been observed elsewhere for thermals (Kaimal et al. 1976; Lenschow and Stephens 1980). The spacing between plumes also slightly increases with height, at least in the lower half of the CBL (Fig. 6). Thus, the number of plumes must decrease with height, specifically from about 5 to 4 (10 km)⁻¹ between the lower and the upper CBL (not shown). This can be due either to anvil merger or to shallow plumes not (yet) reaching the CBL top. Examples of both can be found in the cross sections.

Thermodynamic and kinematic properties of echo plumes

We now evaluate in-plume anomalies of several in situ variables, namely, mixing ratio (r′), potential temperature (θ′), virtual potential temperature (θ′_υ), and vertical air velocity (w′). All of these variables are low-pass filtered (using the same minimum wavelength of 120 m) and detrended before computing the in-plume anomalies, denoted with a prime (′). These anomalies are shown in Fig. 7 in terms of plume strength ΔZ.

In the absence of liquid or solid water, and ignoring pressure perturbations, the buoyancy B (m s⁻²) can be partitioned in a temperature term and a water vapor term (e.g., Doswell and Markowski 2004),

View Expanded

where the overbars refer to detrended leg-mean values. Thus, the buoyancy force is linearly proportional to θ′_υ. For 10 flight legs on 29 May, echo plumes are generally buoyant (θ′_υ ≅ +0.12 K) (Fig. 7a). The leg-to-leg variability is high however, which is mainly attributable to flight level within the CBL, as discussed below. The buoyancy is largely because of excess water vapor in plumes, that is, the second term in Eq. (1) (r′ ≅ +0.6 g kg⁻¹). The plumes are not anomalously warm, in fact at low plume strength, plumes tend to be slightly cooler than the background (θ′ ≅ −0.02 K). All anomalies increase as the plume strength ΔZ increases from 0 to 5 dB; that is, more-defined echo plumes are more buoyant.

The plume buoyancy is remarkably similar for seven flight legs on 6 June (Fig. 7b), and for six flight legs on 17 June (Fig. 7c); θ′_υ is positive and gains about +0.5 K as ΔZ increases in the range of 0–5 dB. However, on these days the buoyancy is only partly because of excess water vapor in the plumes (0.2–0.4 g kg⁻¹); the water vapor term explains ∼55% of the buoyancy on 6 June and ∼40% on 17 June. The remainder of the buoyancy is because of excess warmth in the plumes (θ′ < 0.1 K). The mean mixing ratio was highest on 17 June, and only on this day were a few very small cumuli present in the upper CBL. One unexpected observation is that the water vapor excess in the plumes slightly decreases with increasing plume strength on this day; it is not clear why this is, but the 17 June sample was the smallest in size (fewest flight legs and shortest leg length). Especially at high plume strength, few plumes were sampled.

The air tends to rise in plumes (Fig. 7), consistent with the positive buoyancy there, on all 3 days. Updrafts tend to be stronger on days with higher plume buoyancy, although the differences in mean plume buoyancy and updraft strength between the 3 days are small. The increase of updraft strength with plume strength is consistent with the corresponding buoyancy change. Is the observed buoyancy, ∼0.12 K at 1-dB plume strength, sufficient to explain the observed updraft speed? The mean value of w′ at the 1-dB plume threshold ranges from 1.0 m s⁻¹ on 29 May to 1.5 m s⁻¹ on 17 June. The height over which an undiluted parcel with the observed buoyancy (unopposed by pressure perturbation gradient forces) gains an updraft of 1.0–1.5 m s⁻¹ is only 132–296 m, that is, far less than the CBL depth. In other words, the magnitude of the plume buoyancy appears sufficient to produce the observed updrafts. The magnitudes of θ′_υ and w′ in Fig. 7 are similar to those for thermals as documented by Lenschow and Stephens (1980). Their profiles (shown in their Fig. 4) are normalized by the surface fluxes, but upon scaling the variables, they find values of r′of ∼0.3 g kg⁻¹, θ′ of ∼0.15 K, and w′of ∼1.0 m s⁻¹ in thermals.

Variation of plume characteristics with height in the CBL

In section 3c, average properties of echo plumes were examined. We now describe the variation of plume properties inferred from individual flight legs as a function of height in the CBL. Water vapor excess in plumes tends to be height independent (Fig. 8a), except near the CBL top, where some large water vapor anomalies occur, presumably because of the entrainment of dry free-tropospheric air into the CBL between plumes. This is more pronounced on 29 May than on the other 2 days. That can be explained by the large drying across the CBL top on 29 May; sounding data indicate a vertical mixing ratio difference of 4–5 g kg⁻¹ across an ∼200-mdeep stable layer above the CBL (Fig. 3 in GM05), which is about 2 times as large as on 6 and 17 June.

Similarly, θ′ tends to be positive at low levels and become negative in the upper CBL (Fig. 8b). Unfortunately, none of the flight levels fall between 0.8z_i and 1.0z_i, but the rapid decrease of in-plume θ′ in the highest flight levels indicates that high-θ air from above the CBL entrains rather far down between plumes, at least to 0.7z_i.

In a well-mixed BL, capped by a drier free atmosphere, the in-plume θ′ tends to decrease with height because d/dz > 0, and vice versa for r′, because dr/dz < 0 across the CBL top (the overbar refers to a basic state). Edwards (2004) confirms these vertical trends for 29 May by plotting θ and r from surface stations along the flight track and from all WKA flight legs, as a function of height AGL, and averaged for 1-h time blocks. These opposite vertical trends imply that θ′_υ is less height dependent [see Eq. (2)]. The buoyancy is expected to be due more to temperature at low levels and to water vapor higher in the CBL (Nicholls and LeMone 1980). Indeed, comparing Figs. 8a, 8b and 8c, it is clear that at lower levels plume buoyancy (θ′_υ ≅ +0.15 K) is largely because of excess warmth (θ′ ≅ +0.10 K), while at higher levels (above 0.6z_i) the plumes lose their warmth (θ′ < 0) yet remain positively buoyant for some time because of their excess humidity (r′ > 0). We find that below 0.6z_i, while θ′ does decrease slightly with height, r′ remains rather constant, and as a result buoyancy decreases slightly, consistent with the conceptual image of an entraining thermal. Extrapolated values of θ′_υ become negative close to 0.8z_i (Fig. 8c).

The plume vertical velocity profile (Fig. 8d) is broadly consistent with the plume buoyancy: average plume updrafts strengthen with height from an extrapolated 0 m s⁻¹ at the ground to a maximum near 0.5z_i, where the plume buoyancy starts to rapidly decrease. The vertical profiles of r′, θ′, and w′ correspond well to those in Lenschow and Stephens (1980).

Because echo plumes are both anomalously moist and warm (at least in the lower CBL) and tend to be associated with updrafts, they must contribute to the upward flux of both heat and moisture, simultaneously, even under the observed heterogeneous land surface conditions (e.g., LeMone et al. 2003). To assess the vertical profile of heat fluxes by the coherent eddies that produce echo plumes, we take the products θ′w′ (a surrogate for sensible heat flux by plumes) and r′w′ (latent heat flux). The former slightly increases up to 0.4z_i and decreases with height above this level, becoming negative above 0.6z_i (not shown). The product r′w′, on the other hand, rises at all levels with flight data by a factor of 4 between 0.1z_i and 0.7z_i. This may imply that the latent heat flux by the plume-generating eddies rapidly increases with height, whereas the sensible heat flux decreases with height above 0.4z_i. This is in agreement with flight-level turbulent flux measurements (LEM05). Also, the ratio of latent heat to sensible heat LE/H ∼ L r′w′/(C_pθ′w′) is larger than 1 (Fig. 9a) and increases with height, indicating that the latent heat flux by coherent eddies increasingly exceeds the sensible heat flux. This too is consistent with airborne eddy flux measurements. Here L is the latent heat of condensation and C_p is the specific heat of air under constant pressure.

In summary, this basic characterization suggests that echo plumes correspond with the conceptual image of CBL thermals. Clear-air thermals lose their buoyancy gradually at first, and rapidly in the upper part of the CBL (Lenschow and Stephens 1980; Young 1988). The negative buoyancy of plumes near z_i reflects both the overshooting/detrainment of plumes, and subsidence/entrainment of free-tropospheric air between the plumes. Significant variance exists in the properties of individual echo plumes, which may be due in part to a time lag between the dynamics and the radar reflectivity signal (which results from insect convergence). Such a time lag becomes important if it is significant relative to the typical lifespan of a thermal.

Updraft plume definition

We now define updraft plumes and compare them with reflectivity-defined plumes. Observational studies (e.g., Young 1988; Khalsa and Greenhut 1985) and numerical simulations (e.g., Berg 2002) have used vertical air velocity as the indicator variable in the definition of CBL thermals. This study has the advantage that the vertical air motion can be estimated not just at flight level but also in the vertical dimension by means of WCR Doppler data above and below the aircraft. Thus we use the method discussed in section 3a using radar vertical velocity instead of reflectivity. WCR vertical velocities were averaged over the 14 closest range gates—7 above and 7 below the aircraft, low-pass filtered (using the same threshold wavelength of 120 m), and detrended (w_df). An “updraft plume” is defined as a continuous section of flight leg at least 120 m wide with w_df equal to or greater than a threshold value, which is chosen to be the flight-leg mean (w_m) plus a certain value ΔV. The parameter ΔV is referred to as the updraft plume strength, and it ranges between 0.0 and 1.0 m s⁻¹.

Updraft plume size characteristics

The definition of updraft plumes is illustrated in Fig. 10. Reflectivity-defined plumes for the same cross section (Fig. 3) do mostly coincide with updraft plumes, both in width and location. The size distribution of updraft plumes decays exponentially, as for echo plumes, but updraft plumes tend to be narrower than echo plumes (Fig. 11). Clearly plume width strongly depends on plume definition (i.e., the assumed plume strength), but even under the weakest definition of updraft plumes (ΔV = 0), the half-width of these plumes (0.19z_i) is less than the half-width of echo plumes (0.21z_i at ΔZ = 1 dB). The half-width of updraft plumes decreases roughly linearly with increasing plume strength ΔV (Fig. 11b), as for echo plumes, but the decay is more gradual than for echo plumes (Fig. 4b). For very well-defined plumes, say those that occupy only 6% of the CBL (i.e., for ΔZ = 5 dB and ΔV = 1 m s⁻¹, approximately), both updraft plumes and echo plumes are about equally narrow, with a best-fit plume half-width of about 0.14z_i.

The spacing between echo plumes is significantly larger than the updraft plume spacing, that is, updrafts are more numerous. Clearly, as the definition of an updraft plume becomes more restrictive, the spacing between updrafts increases. The difference between typical echo and updraft plume spacing applies for all echo and updraft plume strengths, although the ratio of echo: updraft interplume half-width decreases from 2.5 for the weakest plumes (ΔZ = 0 dB and ΔV = 0 m s⁻¹) to 1.5 for the strongest plumes (ΔZ = 5 dB and ΔV = 1 m s⁻¹). The difference in spacing is evident also in comparing Fig. 3 with Fig. 10. Clearly large sections exist without clear echo plumes, but with some updraft plumes.

In short, updrafts tend to be both narrower, and more numerous than echo plumes. One explanation is that reflectivity is the result of a history of updrafts in the CBL, that is, that the insect response time is relatively long in comparison with the lifespan of a thermal. Possibly the updrafts without clear echo plumes are very young ones. It is conceivable that insect concentrations only rise gradually in a developing updraft, because it is unlikely that insects move horizontally toward this updraft. Evidence exists that concentrations increase because insects, carried by air mass convergence, find themselves caught in an updraft and then oppose this rising air motion. Assuming a steady-state 2D circulation of representative strength, Geerts and Miao (2005b) find that insect concentrations settle in updrafts after 30–60 min. It would take about the same time for insects to disperse again after that circulation has been turned off.

The dimensional properties of updraft plumes can be compared indirectly with results from other studies. Melling and List (1980) and Young (1988) also found updraft widths in the CBL to be lognormally distributed. However, Young mentioned that the lognormal distribution might be caused by removing small drafts by the smoothing process. In our case, the plume width distribution peaks in the first bin (0.1–0.2z_i) for both echo plumes (Fig. 4a) and updraft plumes (Fig. 11a). The average number of updraft plumes per unit distance and the plume width compare well with estimates by Lenschow and Stephens (1980) for corresponding updraft sizes, if their normalized data are scaled by observed surface fluxes. They use flight-level data, in particular, vertical velocity and thermodynamic data, to characterize a rather shallow CBL.

Thermodynamic properties of updraft plumes

Thermodynamically updraft plumes are broadly similar to echo plumes (Fig. 12). Updraft plumes contain excess water vapor (although some 0.1–0.2 g kg⁻¹ less than echo plumes on the three respective days), and about 0.04–0.11 K in excess warmth, and these anomalies generally increase as the updraft plume becomes more defined. As a result, updraft plumes are positively buoyant on all 3 days, with an average magnitude of 0.09 K at ΔV = 0 m s⁻¹ increasing to 0.12 K at ΔV = 0.5 m s⁻¹ and 0.15 K at ΔV = 1.0 m s⁻¹. The contribution of θ′ to buoyancy [first term in Eq. (1)] is generally somewhat larger for updraft plumes than for echo plumes (Fig. 7). It is smallest on 29 May (∼40%) and largest on 17 June (ranging from 68% to 80%, depending on updraft definition), matching what we found for echo plumes. The consistency of thermodynamic properties of updraft plumes with those of echo plumes is confirmed by the positive reflectivity anomalies found coincident with updraft plumes, and the tendency of those anomalies to strengthen as ΔV increases.

The vertical variation of thermodynamic and kinematic properties of updraft plumes is also similar to that for echo plumes. The profiles shown in Fig. 13 display the properties for any updraft (ΔV = 0). Clearly the anomalies are stronger and more consistent for more restrictively defined updraft plumes. In the lower and middle CBL (between 0.1z_i and 0.7z_i), r′ is scattered, revealing no trend, while θ′ shows a negative trend, consistent with Nicholls and LeMone (1980). Above 0.7z_i, the anomalies reveal the influence of the drier, potentially warmer free troposphere. At very low levels, updraft plumes tend to have more excess warmth (θ′ > 0.2 K), and thus θ′_υ is highest near the ground. The decrease of positive buoyancy with height in the CBL suggests that these updrafts are entraining thermals.

Interestingly, the upwelling IR temperature anomaly within updraft (or echo) plumes is also highest for the lowest flight levels, about +0.7 K for echo plumes (Fig. 9b) and +0.4 K for updraft plumes. This temperature is measured by a nadir-pointing Heiman radiometer and is a measure of the underlying land or “skin” surface temperature. (For these low-level flight legs, the contribution of the interstitial air to the upwelling IR is insignificant.) This suggests that the spatial organization of plumes may well be affected by the heterogeneity of the land surface in this largely agricultural region, at least at low levels (LEM05).