a. Radar editing

Two Doppler on Wheels (DOW) radars (Wurman et al. 1997) and the mobile Doppler X-band polarimetric (XPOL) radar (Wurman 2001) (3-cm wavelength, 0.95° beamwidth), and the Shared Mobile Atmospheric Research and Teaching Radar (SMART-Radar) (5-cm wavelength, 1.5° beamwidth) (Biggerstaff and Guynes 2000) collected data between 1928 and 2118 UTC. Each radar scanned a ∼180° sector toward the center of the IOR. The DOW and XPOL radars scanned elevations from 0.5° to 14.4° above the horizontal with 0.5° increments below 3.5° and approximately 1.0° increments above 3.5°. They collected a volume every 90 s. The SMART-Radar collected a volume every 180 s at elevation angles of 0.5°, 0.8°, 1.7°, 2.9°, 4.4°, 6.0°, 7.7°, 9.5°, 11.5°, 13.7°, and 21.7°. To minimize second-trip echoes, DOW2 and DOW3 frequently changed the pulse repetition frequency (PRF), leading to varying degrees of oversampling in the azimuthal direction. Typical beamwidth at the center of the IOR is 200 m, but, when accounting for oversampling, the data spacing is reduced to about 130 m. Typical data spacing in the radial direction is 75 m.

The radars typically were synchronized so that each volume was observed by the radars over the same period. Because CI studies require observations before precipitation forms, clear-air returns are needed. All four radars can detect clear-air returns given a sufficient insect population, and, on this particular day, sufficient returns were present up to at least 1.5 km above ground level (AGL) and a radar range of 20 km.

A semiautomated approach using NCAR's SOLO software program (Oye et al. 1995) was undertaken to remove ground clutter echoes from the data based on concurrent high reflectivity and low velocity. The Bargen–Brown (1980) scheme was used for most of the velocity dealiasing. Manual dealiasing, however, was required where the automated scheme failed, such as near large velocity gradients in the radial direction.

Before the radar data can be synthesized on a common earth-relative grid, several corrections must be applied. First, because mobile radar data are collected in a truck-relative coordinate system, the data azimuths must be made earth-relative by adding an offset based on the truck orientation. This offset is found using a newly implemented solar alignment technique customized for mobile radar platforms (Arnott et al. 2003). Because the XPOL and the SMART-Radar did not perform an alignment scan on 10 June 2002, ground clutter echoes were matched with known tower locations and power lines to determine the radar orientation.

Next, a correction was applied to account for movement of features during the time required to collect the volume such that the resulting volume represents data valid at a central time. Without this correction, features will obtain an artificial tilt in the direction of motion. For a reference frame speed of greater than 2 m s⁻¹, this error will exceed the grid spacing of this study; thus, we correct for this distortion by moving the observations to their location at the center time of the volume using the velocity of the most steady reference frame (U₀, V₀) as defined by Matejka (2002).

Six (U₀, V₀) pairs of approximately (3, 5) m s⁻¹ were determined and applied for approximately 20-min intervals. We estimate the error in (U₀, V₀) to be less than 0.5 m s⁻¹, which represents a maximum displacement error of 22.5 m, much less than the grid spacing of 100 m. We account for vertical changes in the reference frame velocity by determining (U₀, V₀) pairs at three vertical levels and linearly interpolating between those levels.

b. Objective analysis and multi-Doppler synthesis

Once the data were transformed to a common reference time, REORDER software (Oye and Case 1995) was used to create Barnes objective analyses (Barnes 1964) for all four radars. Koch et al. (1983) propose the following criteria for analysis grid spacing Δg:

View Expanded

where Δd is the data spacing. The upper limit is imposed to resolve the observed wavelengths sufficiently on the grid, and the lower limit is imposed to minimize noise in derivative fields. Because azimuthal and vertical data spacing depend on distance from the radar and vary with each radar, no single grid spacing will satisfy (1) at all ranges and times. Thus, one must optimize the grid spacing such that features in the primary interest region are well resolved while not compromising the resolution too severely closer to the radar or creating very noisy derivative fields farther from the radar. We chose Δd based on the 1° separation between elevation scans and on the beamwidth at the average distance between a radar and the center of the primary interest region, leading to 100-m grid spacing in x, y, and z.

The Barnes smoothing parameter, κ, was based on the Pauley and Wu (1990) recommendation that κ = (1.33Δd)² = 0.264 km². If κ is too large, important features will be smoothed out, whereas if it is too small, wavelengths less than twice the data spacing will not be damped sufficiently.

Finally, NCAR's Custom Editing and Display of Reduced Information in Cartesian Space (CEDRIC) software was used to produce multi-Doppler syntheses. Kessinger et al.'s. (1987) overdetermined dual-Doppler approach was applied to estimate the horizontal wind components using four radars, and an interative technique was subsequently used to determine the vertical velocity from the mass continuity equation. In particular, beginning at the surface with vertical velocity, w = 0 m s⁻¹, w at the next level (level 2) is estimated from integration of the layer-average divergence. This estimate for w is used to adjust the estimated horizontal divergence at level 2, and a new estimate for w at level 2 is obtained. This iteration continues until a domainwide average of the change in the density-weighted vertical velocity between iterations reaches a specified cutoff (0.01 kg m⁻² s⁻¹). Because data are sparse in the southern region of the IOR, only regions to which three or four radars contribute to the syntheses are plotted in forthcoming figures. The regions with three or four contributing radars vary slightly during the deployment, but the approximate areas are outlined in Fig. 1.

Time series of vertical velocity observed by the King Air are compared to time series of vertical velocity calculated in the multi-Doppler syntheses (Fig. 2). Comparing in situ observations with radar-derived quantities is not straightforward because radar quantities are representative over the radar volume time and over a grid cell. The ground-relative speed of the King Air was approximately 100 m s⁻¹, such that the King Air moved roughly one grid length (100 m) every second. Hence, the 1-Hz data recorded by the King Air has similar spatial resolution as the synthesized radar data. The radar data, however, represent a 90-s period. Thus, some time averaging of the King Air 1-Hz data is warranted and we applied a 10-s (1 km) running mean. Further difficulties in the comparison arise from inherent damping of small wavelength amplitudes when objectively analyzing radar data to aircraft point measurements. Nevertheless, observed and radar-derived vertical velocities compare well and are highly correlated (see Fig. 2). Thus, we are confident that the vertical velocities in forthcoming analyses are reasonable.

a. Synoptic conditions

A slowly sagging cold front was draped across the Great Plains into southern Colorado on the afternoon of 10 June 2002 (Fig. 3). Although the temperature and moisture gradients across the front are small, the wind shift from southerly to north-northeasterly clearly depicts its location. At 850 mb, a trough west of Kansas produced strong south-southwest flow over the analysis region and the moisture axis extended across central Kansas (Fig. 4, valid at 1200 UTC). By 0000 UTC on 11 June 2002, the moisture axis pushed into eastern Kansas (not shown).

An IOR was deployed surrounding the cold front north of Dodge City near Ness City, Kansas (box in Fig. 3). Although only cumuli developed along the cold front in the IOR during the deployment, deep convection was initiated along the cold front just west of the IOR. An M-GLASS sounding taken within 5 km from the initiation location about 90 min prior to initiation (Fig. 6) shows little difference from the sounding taken within the grid domain an hour later (Fig. 5). Because the superadiabatic layer is so shallow, and because the soundings have a history of inaccurate surface temperature readings, a parcel path based on the surface temperature may underestimate the amount of CIN present. Therefore, for the 2044 UTC sounding, a parcel with the mean CBL temperature at 1200 m AGL plus one standard deviation (to capture the effects of thermals) recorded by the King Air ahead of the cold front was used for the CIN, LCL, and LFC calculations in Fig. 5. Aircraft data were not available near the 1935 UTC sounding, so a parcel with a CBL temperature recorded by the sounding at approximately 1200 m plus the standard deviation recorded by the King Air was used for the calculations in Fig. 6. Both soundings indicate small amounts of CIN, and neither is obviously more favorable than the other for initiating convection.

b. Kinematic structure of the front

Scientists with DOW2 and DOW3 noted widely scattered cumulus humilis during the first 30 min of the deployment on 10 June 2002. These are also evident in the 1-km visible satellite imagery (Fig. 7a). At this time, DOW2 detected a reflectivity thin line indicative of the cold front and a secondary, weaker line to its east. Smaller, weaker reflectivity maxima are present west of the front. By 2000 UTC, cumulus towers were developing in the IOR while the two thin lines approached each other (Fig. 7b). The clouds grew in size and number along the front as indicated by the clouds filling in along the cold front in the satellite imagery (Fig. 7c) until 2030 UTC when they became widely scattered (Fig. 7d). Simultaneously, the cold front thin line became more diffuse and the secondary thin line moved north out of the IOR. By 2103 UTC, however, towering cumulus clouds redeveloped in the southwest corner of the IOR and deep convection was initiated just outside of the IOR. Even though the base reflectivity from DOW2 does not show any sign of initiation at this time, upper-level reflectivity maxima indicative of rain formation in the growing cumulus clouds at 5.7 km AGL were first observed at 2055 UTC on the edge of the DOW2 9.4° scan (Fig. 8). Indeed, by 2121 UTC, the base radar reflectivity confirms that convection had been initiated and the fully developed storm with reflectivities nearing 50 dBZ entered the IOR (Fig. 7f). Similarly, the small cell just north of the main storm initiated on the edge of the DOW2 scans.

Multi-Doppler syntheses at 0.1 and 1.3 km AGL illustrate the finescale structure of the convergence and horizontal wind fields associated with the cold front, linear reflectivity maxima west of the front, and secondary thin line (Fig. 9). The domain shown in Fig. 9 and in subsequent radar synthesis figures is the entire IOR unless otherwise noted. As expected, convergence maxima typically coincide with reflectivity maxima. The reflectivity thin line, enhanced convergence, and a wind shift mark the cold front position. In addition, four linear reflectivity and convergence maxima behind and roughly perpendicular to the front (Fig. 9a) resemble HCRs in that convergence weakens with height (Fig. 9b) up to the top of our domain, suggesting their circulation may encompass the depth of the boundary layer. Further, these features align with the boundary layer vertical wind shear (shown later), as Ferrare et al. (1991) note. Unlike some previous studies on HCRs, however, the observed features do not align with the mean boundary layer wind; indeed the mean boundary layer wind is nearly perpendicular to these features (not shown). Perhaps some discrepancy exists because the vertical shear is predominantly speed shear in the studies where HCRs align with the mean wind, while on 10 June 2002, and in the case of Ferrare et al. (1991), the vertical shear has a directional component that may lead to a closer alignment between HCRs and the shear vector.

These reflectivity maxima may alternately be explained as reflections in the boundary layer of gravity waves in the weak capping inversion aloft. The estimated Brunt–Väisälä frequency (N) in this inversion layer is 0.02 s⁻¹. Using

View Expanded

where λ is the horizontal spacing of the features (3 km) gives a phase speed (c) of 9.6 m s⁻¹. These features appear to move little, and the wind speed and direction in the stable layer is 10 m s⁻¹ at 213°. Hence, if these features are propagating with a phase speed of 9.6 m s⁻¹ from the northeast, little net movement would result. HCRs in the boundary layer could be a source of the gravity waves, possibly making these features a reflection of gravity waves and HCRs. Unfortunately, no data are available through these features above the boundary layer to confirm the presence of gravity waves in the capping inversion. Thus, because we cannot diagnose with certainty the origins of these linear reflectivity and convergence maxima, we will refer to them from hereon as linear reflectivity maxima (LRMs).

Reflectivity and convergence rings ahead of the cold front indicative of the observed open-cell convection also are evident (Fig. 9a). The secondary thin line observed in the raw reflectivity fields is present in the multi-Doppler syntheses as a convergence line with a south to southwest wind shift across it.

Near the upper boundary of the multi-Doppler domain (Fig. 9b), the convergence and reflectivity signatures are weaker and the wind direction varies less across the location of the surface cold front and secondary convergence line. Further, there is less evidence of the LRMs in the reflectivity field as the low-level convergence has been replaced by divergence at this level. Although divergence indicates some weakening of the vertical velocity with height, positive vertical velocity extends to the top of the domain over the LRMs, the cold front, the secondary convergence line, and the boundary layer convective cells.

c. Evolution of convergence along the front

The along-line variability of the depth and strength of convergence along the cold front at this time are extraordinary. To complicate matters, the cold front and boundary layer structure also varied significantly in time. Normally, such transient and small-scale features might not be trusted, but the fine spatial and temporal resolution, volume-to-volume temporal continuity, and favorable comparisons with aircraft data validate their existence. Convergence at 100 m AGL for four different times is shown in Fig. 10. Figure 10a is similar to Fig. 9a except without DOW2 reflectivity. At this time, convergence is evident along the entire length of the cold front. Note, however, that convergence along the cold front and secondary boundary is spatially variable with local maxima embedded in the background field. Between the two boundaries is a region of divergence. As the two convergence zones approach each other (Fig. 10b), the LRMs remain well defined, but the open-cell convection becomes less distinct and the divergent region between the two zones becomes smaller.

By 2037 UTC (Fig. 10c), however, convergence along the cold front becomes markedly fractured. High temporal resolution reflectivity imagery (not shown) indicate that most of the secondary convergence line was advected by the southerly mean flow out of the IOR. The LRMs and open-cell convection also are less distinct. Finally, when deep convection is initiated just west of the IOR (Fig. 10d), the convergence field is more continuous than at 2037 UTC, but less so than at 2010 UTC. Reasons for the observed changes in convergence are discussed in the next section.

d. Evolution of the frontal circulation

We now consider why the upward motion pattern was more fractured after 2020 UTC than before. Vertical wind shear in the IOR indicates a horizontal vorticity imbalance across the front (Fig. 11). Here vertical wind shear is defined as the difference between the wind vector at 1.4 km AGL and the wind vector at 100 m AGL. The orientation of the cold front is approximated by the thin oblique lines in order to calculate a front-normal shear vector. To attain the optimal vorticity balance of RKW theory (Rotunno et al. 1988), the front-normal shear vectors ahead of the front should have the same magnitude, but opposite direction, as those behind the front. This balance clearly is not present at most locations along the front during the deployment (Fig. 11), indicating that this front may not provide optimal lifting and that updrafts will tilt back over the front.

To assess the evolution of the frontal circulation as a whole, a coordinate transformation is performed such that the new x axis (x′) is roughly parallel to the cold front and the new y axis (y′) is perpendicular to the cold front. Horizontal winds are also transformed such that u is along x′ and υ is along y′. Using a piecewise function to approximate the front, u, υ, and w were averaged along the front. The averaging reduces the data to only the y′ and z dimensions, yielding an average vertical cross section across the front (Fig. 12). It is clear in these cross sections that the vertical shear between 1.4 and 0.1 km AGL ahead and behind the front are unbalanced. Average upward motion associated with the cold front is much stronger at 1946 UTC (Fig. 12a) than that at 2037 UTC (Fig. 12b). Reasons for this weakening in time are addressed in the next section.

e. Thermodynamic structure and evolution of the front

The University of Wyoming King Air flew across the cold front at several elevations between 2010 and 2103 UTC. Data at an average of 170 m AGL (905 mb) and 1230 m AGL (803 mb) [standard deviations of height AGL (pressure) are 14 m (3 mb) and 33 m (1 mb), respectively] are shown in Figs. 13a and 13b, respectively. The north–south component of the wind (V) indicates clearly the location of the cold front at 170 m AGL (point B), which agrees with the multi-Doppler synthesis frontal location (Fig. 14). The significant minimum in mixing ratio (point C), is likely caused by downward vertical motion advecting dry air from higher in the boundary layer (Figs. 13a and 14), and is likely transient in time because it is not evident at upper levels on a later pass (Fig. 13b). Although potential temperature is about 1°C less behind the front than ahead of it, mixing ratio is slightly higher behind the front. Virtual potential temperature perturbations are negative west of the cold front and are positive east of the cold front (Fig. 13c). Furthermore, the virtual potential temperature perturbations decrease and become negative east of the secondary convergence line (between points D and E) indicating that this convergence line is associated with density gradient.

To explore reasons why the cold front weakened, we turn to the frontogenesis equation, which is given by

View Expanded

(Bluestein 1993), where θ is potential temperature. For an overall estimate of frontogenesis on 10 June 2002, we assume no variations in the x direction (perpendicular to the path of the King Air). Furthermore, (∂θ/∂z) ≈ 0 in the nearly dry adiabatic boundary layer. Equation (3) then simplifies to

View Expanded

where υ is the wind component along y (in this case, along the King Air path). Using the 170 m AGL data from the King Air valid between 2011 and 2016 UTC, the first term is 2.3°C h⁻¹ km⁻¹. This term is positive, as expected due to convergence along the front. The average vertical cross sections across the front, however, indicate a weakening of the frontal circulation with time; therefore, the diabatic terms must be negative.

To verify the sign of the diabatic terms, data from the P-3 are examined. The P-3 flew parallel to the front, twice behind it and twice ahead of it (Fig. 1). Data were collected at 550 m AGL (on average) and are shown in Fig. 15. During its flight behind the front between 1938 and 1940 UTC, the average potential temperature was 313.85 K (Fig. 15, left panel, thin line), while ahead of the front between 1954 and 1959 UTC the average potential temperature was 315.48 K (Fig. 15, right panel, thin line), indicating a 1.63-K drop in potential temperature across the boundary, similar to data collected by the King Air.

The next flight behind the boundary (2018–2020 UTC) indicates an average 1.51-K increase in potential temperature (Fig. 15, left panel, thick line) compared to the flight behind the boundary 40 min earlier. Ahead of the front, the average potential temperature increased only 0.40 K (Fig. 15, right panel, thick line).

This warming behind the front also was observed at the surface by a mobile mesonet probe (Fig. 16). Data for three passes along the same longitude are shown. For each pass, the cold front was just south of 38.555°N. Although there are few data points ahead of the front, warming behind the front in time is clear, with an average 1.1°C increase, similar to that observed by the P-3. This warming behind the cold front is represented in the diabatic heating term of F, and is frontolytic. Hence, the weakening frontal circulation we observe in Fig. 12 is likely caused by the enhanced warming behind the cold front.

f. Structure and evolution of vertical velocity and vertical vorticity

In addition to temporal and spatial variability in convergence, the cold front also displayed significant variations in vertical velocity and in vertical vorticity (Fig. 17) over this time period. As expected, convergence along the cold front (Fig. 10a) indicates rising motion aloft (Fig. 17a). Vertical velocity along the secondary convergence line is generally weaker than that along the cold front. Rising motion associated with the LRMs and open-cell convection is also evident.

The horizontal wind shear across the cold front (see the wind vectors in Fig. 10) is associated with a zone of positive ζ along the cold front (Fig. 17). In contrast, very little vertical vorticity is present along the secondary convergence line. Four striking ζ maxima (misocyclones) are present along the front at this time. All four misocyclones are near an intersection of an LRM with the cold front, similar to Wilson et al. (1992) and Atkins et al. (1995). Furthermore, as also noted by previous researchers, the w maxima along the cold front typically occur in association with, but offset from, a misocyclone (e.g., Kingsmill 1995; Lee and Wilhelmson 1997). For instance, in the case of misocyclones 1 and 3, the w maximum is northwest of the misocyclone, whereas for misocyclones 2 and 4, the w maximum is due north of the misocyclone.

Vertical cross sections of misocyclones 1, 2, and 3 reveal that the misocyclones are deep features, extending at least to the top of the domain (Fig. 18). For misocyclones 1 and 3, ζ is maximized above the lowest radar elevation indicating that the vorticity may be produced through tilting. Near the surface, ζ in the misocyclones is stronger than the ambient ζ along the cold front suggesting the importance of stretching in this region. Computing the vorticity tendency terms along trajectories through a misocyclone may help clarify the origins of the misocyclones and will be addressed in a future paper.

The evolution of w and ζ is similar to that of convergence. At 1946 UTC (Fig. 17a), four misocyclones are present along the cold front, but +w is fractured only slightly (between misocyclones 1 and 2 and near misocyclone 4). At 2010 UTC, the convergence lines approached each other and the number of misocyclones along the front decreased. Misocyclone 2 has intensified, but it is aligned along the front, and does not break the upward motion pattern significantly. On the other hand, misocyclone 4 has created a kink in the boundary. The only break in upward motion along the front at this time is over this kink. Misocyclone 3 dissipated and misocyclone 1 propagated out of the IOR. Notice that upward motion is more continuous along the front at this time than at other times.

Starting at 2020 UTC, several misocyclones begin to merge, and the axes of the mergers tend to be perpendicular to the front. Much like a vortex sheet rolling up, several significant kinks in reflectivity (not shown) develop as a consequence of these misocyclones (Marquis et al. 2004), coincident with large fractures of upward motion along the cold front (Fig. 17c). This continues until the end of the deployment when a very large merger of misocyclone 6 with ζ associated with LRMS behind the cold front creates an occluded appearance in the reflectivity field (not shown) and a large kink and break in upward motion along the cold front (Fig. 17d).

g. Parcel trajectories and evolution of cumulus clouds

Parcel trajectories are computed to estimate how the evolution of convergence and upward motion affect the cumulus field. At two different times, parcels at 100 m AGL are chosen every 1 km in a region straddling the cold front (Figs. 19a and 19b). The parcels are traced forward and backward in time over the entire period they are in the IOR. Parcels entering the frontal circulation have strong along-frontal components to their motion. During the first hour of the deployment, parcels entering the convergence zone ascend along a mostly vertical path (Fig. 20a), but after 2020 UTC, many fewer parcels leave the top of the domain (Fig. 20b).

The decrease in the number of parcels able to exit the top of the domain is directly related to the fractured pattern of upward motion along the cold front that results from the superposition of the weakening front and embedded misocyclones. For instance, when upward motion along the cold front is nearly continuous, parcels entering the frontal zone retain upward motion as they move horizontally along the front, allowing the parcels to rise for a long period of time. When the upward motion becomes discontinuous, however, parcels moving along the cold front enter regions of varying vertical forcing, causing them to rise initially, but then to decelerate, and possibly even descend, as they encounter regions of downward forcing. If, however, parcels move with the same speed as the w maximum, they may also reach the top of the domain. Vertical velocity maxima associated with misocyclones tend to have this trait even if w is not continuous along the boundary and may, therefore, be preferred locations for initiation when upward motion is discontinuous. The w maximum associated with misocyclone 6 is an example of such a location. Parcels chosen at 100 m AGL in enhanced convergence associated with misocyclone 6 retain upward forcing as the misocyclone moves along the front (Figs. 19c and 19d). Therefore, for maximum parcel displacement, parcels must remain in regions of upward forcing for a significant time by either moving through a region with continuous upward forcing or by moving at the same velocity as a maximum in upward forcing.

The discontinuous upward motion along the weakened cold front causes the cumulus cloud field to become scattered and more shallow as well. Cumulus congestus clouds deepen along the cold front between 2000 and 2015 UTC (Fig. 21a). At 2020 UTC (Fig. 21b), the viewing angle of the northernmost camera changes, focusing on the northernmost portion of the cold front where clouds tended to be more shallow. Hence, it is unclear if the clouds become more shallow between 2015 and 2020 UTC, or if the apparent decrease in cloud depth is due to the change in the viewing angle. Nevertheless, between 2020 and 2037 UTC, the cumulus clouds clearly become more shallow (Fig. 21c).

Photogrammetic analyses (Rasmussen et al. 2003) of the cloud fields indicate a decrease in areal coverage of the cumulus clouds (Fig. 22). Cloud bases are contoured assuming an LCL of 3 km. Note that w and ζ are shown 10 min prior to the time of the cloud analysis. This is to account for the fact that parcels at the level shown in a 2–3 m s⁻¹ updraft at the top of the domain require approximately 10 min to travel the additional 2.2 km to cloud base. Thus, we expect the cloud evolution to lag the w evolution by about 10 min. Grid cells in which 50% or more pixels contain cloud base are contoured. It must be noted that although these analyses indicate only where there are definitely clouds, regions outside the 50% contour do not necessarily indicate a lack of clouds. For instance, clouds may block each other from the camera view, cloud base may be indistinct, or the clouds may be too distant to determine cloud base. Despite this caveat, the photogrammetric analyses are supported by 1-km visible satellite imagery showing similar dissipation in cloud coverage, and scientists with DOW2 and DOW3 noted this dissipation in their visual observations during the deployment.

Thus, the weakened front and resulting discontinuous upward motion after 2020 UTC diminish the areal coverage and depth of cumulus clouds along the cold front because parcels traveling along the cold front experience less dwell time in regions of upward forcing. Nevertheless, scattered cumulus clouds remain. These clouds likely are associated with the w maxima that advect with the flow (e.g., maxima associated with the misocyclones) such that parcels retain upward motion. For instance, misocyclone 6 is associated with strong vertical motion, and trajectories in this region show that parcels in this upward motion reach the top of the domain (Fig. 20b). Unfortunately, misocyclone 6 was outside the cloud photogrammetry domain, but visible satellite imagery does indicate a cloud near this area.