1. Introduction
The planetary boundary layer (PBL) is the lowest portion of the atmosphere that is directly influenced by Earth’s surface. The PBL follows a diurnal cycle, growing upward during the daytime due to turbulent fluxes from the surface and reaching a peak height in late afternoon before mixing from the surface ceases and a shallow nocturnal stable boundary layer develops after sunset (Stull 1988). However, complex structures and evolutions that deviate from this conceptual model are frequently observed (e.g., Clarke 1969; Agee and Hart 1990; Cooper and Eichinger 1994; Kristovich et al. 2003; Pullen et al. 2008; De Tomasi et al. 2011; Takle et al. 2019). Because the top of the canonical PBL represents the depth through which convective mixing freely occurs, its height and characteristics can have a significant impact on many socioeconomically high-impact phenomena, including convection initiation (e.g., Crook 1996; Browning 2007), smoke and pollutant transport (e.g., Dabberdt et al. 2004; Clements et al. 2007; Miao and Liu 2019; Yuval et al. 2020), and wind energy production (e.g., Draxl et al. 2014; Carvalho et al. 2014; Surussavadee 2017).
Despite its importance, real-time measurements of the PBL, including the PBL top, remain scarce in the United States, as standard radiosondes are typically launched only twice daily at distantly spaced sites and more specialized measurements, such as 915-MHz profilers (e.g., White 1993; Angevine et al. 1994; Bianco and Wilczak 2008) and mobile platforms (e.g., Wagner et al. 2019), are expensive and not widely available. One exception to this is weather radars, which can exploit Bragg scattering to infer the top of the PBL. Bragg scattering occurs due to fluctuations in the atmosphere’s refractive index at scales of one-half the radar wavelength (Knight and Miller 1993), which frequently occur at the top of the PBL as turbulent entrainment mixes air in the PBL with that of the free atmosphere above it. Rabin and Doviak (1989), and later Heinselman et al. (2009) and Elmore et al. (2012), observed rings of enhanced radar reflectivity (Z) on plan position indicator (PPI) scans that expanded outward during the day and were hypothesized to be evidence of Bragg scattering associated with the PBL top. However, this signature was not always observed, and could be easily obfuscated by scattering associated with biota.
Polarimetric weather radars allow for clearer differentiation between backscatter due to Bragg scattering versus biota. Owing to their highly oblong shapes, biota have very large intrinsic values of differential reflectivity, ZDR, while Bragg scatter, which is assumed to be isotropic, has an intrinsic ZDR of 0 dB (Melnikov et al. 2011, 2013; Richardson et al. 2017). Thus, in a background field of biota, layers with additional scattering contributions from Bragg scatter manifest as local minima in the ZDR field. Originally explored through specialized scanning strategies (Melnikov et al. 2011, 2013), Banghoff et al. (2018; hereafter B18) exploited this concept and explored daytime PBL evolution via quasi-vertical profiles (QVPs; Kumjian et al. 2013; Ryzhkov et al. 2016) for use with operational scanning strategies. QVPs are formed by azimuthally averaging high-elevation radar scans, which permits the detection of small gradations in the data by significantly reducing noise and the construction of time–height cross sections for tracking the evolution of features of interest.
The proof-of-concept findings in B18 were encouraging, with favorable comparisons to PBL heights estimated from radiosondes. Should this approach prove robust and reliable for a wide variety of seasons and geographical locations as suggested, it offers the promise of real-time PBL height estimates as well as the ability to construct climatologies (Stensrud et al. 2024) and improve PBL parameterization schemes through both validation and assimilation (e.g., Tangborn et al. 2021; Mykolajtchuk 2021; Eure et al. 2023) from an existing operational network. A potential limitation of this approach is that the scope of B18’s initial quantitative validation was limited to central Oklahoma using late-afternoon/early-evening radiosondes whose launch time (roughly 2300 UTC) corresponds to a time with an already mature daytime PBL, and specifically focused on days without precipitation in which unambiguous PBL heights were able to be gleaned from both QVPs and radiosondes. Heinselman et al. (2009) reported biases in the Z-based PBL height that varied with the diurnal cycle and were improved when incorporating an additional insolation term (Elmore et al. 2012), suggesting the relation between the observed Bragg scatter signature and the actual PBL height may be variable and depend on factors that affect surface fluxes, such as local land surface and vegetation type. In addition, B18 and Melnikov et al. (2013) observed many noncanonical Bragg scatter signatures that deserve further exploration and contextualization. Finally, while the azimuthal averaging in QVPs aids in the detection of minute changes in ZDR potentially associated with Bragg scatter, spatial heterogeneities in the height and magnitude of the Bragg scatter signature that are frequently observed (Melnikov et al. 2011) will be obscured.
The goals of this study are to explore these outstanding issues using specialized boundary layer profiling measurements from the Collaborative Lower Atmospheric Mobile Profiling System (CLAMPS; Wagner et al. 2019). CLAMPS provides high vertical- and temporal-resolution data of low-level atmospheric thermodynamics and kinematics that can be collocated with WSR-88D sites to provide a direct comparison with PBL height estimates using polarimetric QVPs (Banghoff et al. 2018). More specifically, we seek to use CLAMPS data to validate the aforementioned methodology using independent point measurements and expand upon the work of B18 by contextualizing the full diurnal PBL evolution rather than only examining synoptic radiosonde sampling times. We also seek to study the impact of distance from the radar and different geographic and climatic regions on QVP-inferred PBL heights and identify any consistent failure modes/regimes.
2. Methodology
a. IOP overview
Data collection was performed from 24 August 2020 through 24 September 2020. During the first part of this period (24 August 2020–3 September 2020), both CLAMPS platforms were deployed in central Oklahoma: CLAMPS-2 was stationed at the University of Oklahoma’s Max Westheimer Airport (35° 14′13.2″N, 97° 27′46.8″W) collocated with NOAA’s Norman, Oklahoma, research S-band WSR-88D radar (KOUN), while CLAMPS-1 was located at the Kessler Atmospheric and Ecological Field Station (KAEFS; 34° 59′2.4″N, 97° 30′57.6″W), 29.6 km south-southwest of KOUN. This arrangement was designed to study the impact of distance from the radar and the impact of azimuthal averaging on estimating potentially heterogeneous PBL heights. For the second half of the IOP (4–24 September 2020), CLAMPS-2 remained collocated with KOUN while CLAMPS-1 was moved to the NWS Forecast Office in Shreveport, Louisiana,1 collocated with NOAA’s S-band KSHV WSR-88D radar (32° 27′7.2″N, 93° 50′31.2″W), with the deployment ending one day earlier on 23 September 2020. This arrangement was designed to study the impact of different climatic regions and is depicted in Fig. 1.
MODIS satellite data from 6 Sep 2020 showing a regional view of land typology variability across the IOP domain. Insets show closer views of the KOUN and KSHV subdomains and the positions of each CLAMPS relative to the NEXRAD site. The dotted and solid rings at ranges of 13 and 25 km indicate the approximate averaging areas at 1.0 and 2.0 km AGL from the 4.5° QVP, respectively.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
The early part of the IOP was characterized by 500-hPa ridging building in from the southwest, with weak flow aloft and clear skies over Oklahoma. Hurricane Laura (2020) passed to the southeast of the initial IOP data collection region on 26–27 August 2020. By 29 August 2020, a positively tilted trough formed over the Intermountain West and gradually moved eastward, with a stationary front draped across the IOP data collection region and widespread precipitation from 31 August 2020 through 1 September 2020. During the second half of the IOP, another 500-hPa ridge built in from the southwest with clear skies and high pressure across both IOP regions. On 9 September 2020, a 500-hPa cutoff low centered over the Four Corners region resulted in widespread precipitation before moving into the northern Great Plains, with otherwise quiescent weather through much of the remaining IOP. Beginning on 20 September 2020, Tropical Storm Beta (2020) brought additional precipitation to Louisiana that lasted for the remainder of the IOP, with low clouds and fog at the end of the IOP in Oklahoma.
b. CLAMPS data and quality control
CLAMPS is a mobile boundary layer research instrumentation platform and a joint effort between the University of Oklahoma and the NOAA/OAR National Severe Storms Laboratory. Each CLAMPS platform features a Doppler lidar (DL), an atmospheric emitted radiance interferometer (AERI), and a microwave radiometer (MWR). The DL conducted 70° PPI scans every 10 min for VAD calculations and was pointed at zenith otherwise. Radiative measurements from the MWR and AERI (i.e., brightness temperatures and radiances, respectively) were also collected and used as input to a temperature and moisture profile retrieval algorithm (Turner and Blumberg 2019) with data available every 10 min. From the CLAMPS, vertical profiles of vertical velocity w, vertical velocity variance
PBL heights are determined from the CLAMPS data using the fuzzy-logic algorithm described in Smith and Carlin (2023). Briefly, this approach expands upon the DL-based fuzzy-logic algorithm of Bonin et al. (2018) by additionally incorporating thermodynamic retrievals from the AERI and MWR. The algorithm works in two steps. In the first step, direct measurements of ongoing mixing (e.g., vertical velocity variance) are used to determine a first-guess PBL height. In an effort to aid in detecting nocturnal stable boundary layers, temperature inversion information retrieved from the AERI and MWR is incorporated using sunset- and sunrise-time-dependent weights. In the second step, additional indicators that mixing has occurred (e.g., isentropic potential temperature profiles) are used to refine the first-guess height. The PBL height estimates are then smoothed using an hourwide triangular-weighting sliding window and interpolated to 10-min intervals.
Although we are comparing the radar-derived PBL heights to those derived from CLAMPS, the CLAMPS estimates are also prone to error and not intended to be considered the definitive truth. The instruments onboard CLAMPS make it an optimal platform for observing the evolution of the PBL, but some assumptions, such as thresholds and (time dependent) weights for the membership functions in the fuzzy-logic algorithm (Smith and Carlin 2023), are still required for determining a PBL height and will not necessarily be optimal for all locations and times. Further, the CLAMPS’ ability to detect the PBL top can be limited if there are insufficient aerosol scatterers through the full depth of the PBL (Smith and Carlin 2023). In this IOP, sufficient signal-to-noise ratio (SNR) often extended only up to approximately 1.5 km AGL. An evaluation of our data suggests this was not prohibitive in the cases examined but cannot be ruled out more broadly. Finally, there is no universal definition for PBL top in the meteorological or adjacent communities; this lack of universal definition leads to even greater ambiguity in noncanonical PBL scenarios. These issues with characterizing the boundary layer are instrument agnostic (Seidel et al. 2010; Smith and Carlin 2023). Despite these considerations, the data collected by the CLAMPS represent a robust source for contextualizing the QVP signatures.
CLAMPS data are sorted into four categories for each 10-min period: missing, suspicious/erroneous, data that corresponds to complex and/or atypical PBL evolution, and data that corresponds to clear-cut canonical PBL evolution (e.g., Stull 1988). Periods corresponding to the first two categories are excluded, with data from the latter two included unless otherwise stated. This filtering is done separately for the kinematic (i.e., DL) and thermodynamic data (i.e., MWR and AERI), with the attendant PBL heights removed if either of the data streams are missing/erroneous. After this initial filtering, 68.0% of CLAMPS-1 data and 54.5% of CLAMPS-2 data are available for comparison, respectively (Table 1).
Availability of usable data from each platform.
c. WSR-88D data and quality control
Radar data were collected and analyzed from the KSHV and KOUN WSR-88D radars. We collected data at KOUN in clear-air mode [volume coverage pattern (VCP) 32; Crum et al. 1993], with volume update times of approximately 10 min and a maximum elevation angle of 4.5°, for the duration of the IOP to maximize sensitivity and emulate operational scanning strategies in clear air. In contrast, KSHV was in operational usage during the IOP, with variable VCPs and volume update times throughout the IOP depending on the contemporaneous needs of forecasters.
We replicate the QVP filtering and generation approach of B18. We use the elevation angle closest to 4.5° for consistency regardless of VCP, although a higher elevation angle would result in a smaller averaging area likely to be more representative of local conditions. This elevation angle combined with typical radome heights of 5–30 m AGL (NOAA Radar Operations Center 2022) results in the lowest available data being at ≈175–200 m AGL. The approach proposed in B18 determines PBL height by taking the height of the minimum ZDR at each scan time prior to smoothing using a running average. However, because we are looking at all times (i.e., not only during periods when the signature was unambiguous), this approach can often result in PBL heights that are noisy and clearly erroneous. In an attempt to address this, we modify the approach for determining the PBL height by adding a continuity component and a confidence flag. The height of the ZDR minimum (up to some user-defined height threshold where data become noisy) is taken as an initial guess as in B18. If the minimum ZDR is located at the lowest QVP level (e.g., potentially due to ground clutter, or a PBL height below the lowest observable data level), the PBL height is deemed unable to be unambiguously determined and the data are assigned a negative confidence flag; 81.0% of KOUN data and 76.0% of KSHV are retained (Table 1). Next, all “troughs” (defined as ZDR minima with a prominence greater than −0.05 dB that are at least 3 data points wide and 4 data points away from the next trough) are found using the SciPy (Virtanen et al. 2020) “find_peaks” function. If any of the identified troughs are within ±300 m of the PBL height estimate from the preceding volume scan, that point (or, in the case of multiple troughs meeting this criterion, the most prominent trough) is taken to be the PBL height and assigned a positive confidence flag. This is done to reduce noise and aid the algorithm in following continuous Bragg scatter layers even if a larger absolute minimum exists elsewhere in the column. Finally, the data are subjectively analyzed and assigned a third confidence flag if the PBL height estimate is deemed unlikely to represent the actual surface-based PBL height. The use of these flags allows for filtering the algorithm’s output to various confidence levels. Here, “group A” data denote all of the radar-based PBL heights that are able to be unambiguously determined (i.e., those assigned a positive confidence flag), while the more-selective “group B” denotes only radar-based PBL heights that are believed to correspond to a surface-based PBL and pass the subjective analysis test. The final estimated PBL height is interpolated to a 10-min grid to match that of the interpolated CLAMPS data.
d. Combining CLAMPS and WSR-88D data
After performing the aforementioned quality control procedures separately for the radar and CLAMPS data, the interpolated 10-min data are time-matched and compared. Comparisons are only retained if there are valid data from both the respective radar and CLAMPS platforms. This process results in valid comparisons for 40%–50% of the IOP data in the default group A dataset, while only approximately one-quarter of the data in the more-selective group B dataset meet these criteria. Despite over half of the data being unusable either due to ambiguity in the low-level radar data or other missing/erroneous data, there are still approximately 1800, 700, and 1500 matched data points available for the KOUN/CLAMPS-2, KOUN/CLAMPS-1, and KSHV/CLAMPS-1 comparisons, respectively (Table 1).
A representative example comparison between the QVP-derived PBL heights and those observed from the CLAMPS platforms is shown in Fig. 2. At the beginning of the period, the QVP-derived PBL heights are still elevated near 1100 m due to residual turbulent mixing from the previous day, while the CLAMPS-derived PBL heights are near the ground below the radar’s lowest observable height. After approximately 0130 UTC (i.e., immediately after sunset), the QVP-derived PBL heights follow a layer of localized ZDR minima that gradually descends, believed to be indicative of mixing within the residual layer as hypothesized in B18 and discussed further in section 4a. Because it is deemed to not be representative of the (likely) surface-based PBL, the overnight hours are flagged and shaded gray. Meanwhile, the CLAMPS-derived PBL heights experience some fluctuations overnight, with a sudden burst of mechanical turbulence between 0600 and 0800 UTC associated with an outflow boundary from nearby convection (not shown) that temporarily drives up the observed boundary layer; otherwise, the surface-based PBL remains shallow and deepens very gradually through the overnight period. Around 1200 UTC (i.e., sunrise), a local minimum in ZDR appears at the lowest-available QVP data level, and thus was flagged in light gray due to the ambiguous nature of whether this signal represents ground clutter, surface-based PBL growth, or both. From 1500 UTC onward, surface-based PBL growth appears much more clearly in the radar data (in agreement with Heinselman et al. 2009), with a clear layer of locally reduced ZDR ascending and reaching a peak height of roughly 1200 m by 2100 UTC before leveling off. During this period, data from both CLAMPS are in good overall agreement with the QVP-derived heights but feature more variability and less-smooth growth. This is to be expected, as the CLAMPS platforms are capable of observing finer-scale undulations and localized bursts of growth in the PBL that are smeared out from the azimuthal averaging of QVPs. Indeed, there is some suggestion that periods with larger variability between the CLAMPS platforms are reflected as deeper layers of Bragg scatter in the QVP (as hypothesized in B18) than during periods of closer agreement and more homogeneity (e.g., 1800 and 2100 UTC vs 1900 UTC).
ZDR QVP from 30 Aug 2020 at KOUN with PBL heights estimated from KOUN (solid), CLAMPS-1 (dotted), and CLAMPS-2 (dashed) overlaid. Accompanying data quality flags are included, where black, gray, and light gray bars signify suspected reliable estimates, estimates believed to be due to something other than the PBL, and ambiguous out-of-range estimates, respectively, as described in the text. On the abscissa, sunset and sunrise are denoted with circles with down and up arrows, respectively, along with solar midnight and noon.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
3. Results
a. Bulk statistics
Figure 3 shows the overall evolution of the PBL heights determined from each of the platforms and includes all available data (i.e., even the ambiguous PBL height estimates at the lowest radar data levels). The bulk evolution observed from both of the CLAMPS platforms follows the canonical PBL evolution described in Stull (1988) (Figs. 3c,d). After sunset, a shallow nocturnal stable PBL forms and slowly deepens, reaching median depths of approximately 100 and 200 m by sunrise (1200 UTC) for CLAMPS-2 at OUN and CLAMPS-1 at SHV, respectively, although appreciable variability exists. While the nocturnal PBL depth is usually ≤400 m at OUN, an examination of individual cases shows instances of sudden spikes of estimated PBL depth due to mechanical mixing in the lowest 1 km, likely associated with remnant convection or the overnight passage of bores (e.g., 30 August 2020 shown in Fig. 2; 22 September 2020). In contrast, CLAMPS-1 at SHV features more frequent variability through the entire overnight period, with PBL depths that frequently reach 500 m. Thermodynamic retrievals from the AERI indicate that, on many nights, the boundary layer remains very moist, little-to-no low-level temperature inversion forms, and weak turbulent mixing persists, reflecting a more muted diurnal PBL cycle. After sunrise, both sites feature variable but steady growth of the PBL. The median PBL depth from CLAMPS-2 at OUN peaks slightly higher (≈1.2 vs 1.1 km) and later in the day (2200 UTC vs 1900 UTC) than CLAMPS-1 at SHV. After 2300 UTC, in the hour preceding sunset, the PBL height rapidly descends to the surface as convectively generated turbulence decays (e.g., Bianco and Wilczak 2008) and the nocturnal stable boundary layer begins to form due to radiative cooling at the surface.
24-h time–height normalized histograms (shading) and hourly medians (black lines) of PBL heights observed at (a) KOUN, (b) KSHV, (c) CLAMPS-2 stationed at OUN, and (d) CLAMPS-1 stationed at SHV with all available data included. Gray shading indicates overnight data where the majority of the detected PBL estimates were flagged as unlikely to actually correspond to the PBL. Data are normalized by the maximum bin count within each hourly bin. On the abscissa, sunset and sunrise are denoted with circles with down and up arrows, respectively, along with solar midnight and noon.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
In contrast, the QVP-derived PBL heights differ appreciably from the CLAMPS-derived PBL heights in the overnight period. Both KOUN and KSHV feature irregular histograms between 0000 and 1200 UTC, with three primary clusters (Figs. 3a,b). The first (region “1” in Figs. 3a,b) is at the lowest available data levels, particularly for KOUN. The reduction in ZDR associated with these radar-based PBL height estimates is likely primarily due to the influence of ground clutter, but the CLAMPS-1 at SHV samples many nontrivial surface-based boundary layers that last throughout the night (Fig. 3d) indicating that some of the radar-based estimates in this cluster could be legitimate. Other clusters appear higher aloft at both sites, with a clustering of points above 1.5 km (region “3”) and another that begins around 1.0 km and gradually descends (region “2”). This descent occurs more clearly at KSHV and appears to indicate mixing at the top of the residual layer from the night before that persists throughout the overnight. The irregularity of the histogram above 1.25 km at KOUN (Fig. 3a; region “3”) appears to be due to both legitimate elevated residual layers and contamination by clouds and precipitation (discussed further in section 4b), which were more readily detectable due to the sensitivity of the clear-air mode VCP used for KOUN. The inability for the radar to observe the lowest ≈175–200 m made it much more likely that these other features with ZDR minima not associated with the nocturnal shallow PBL would be selected by the algorithm. After sunrise, the PBL heights at KOUN and KSHV grow steadily from 1400 UTC onward, reaching peak median values of just over 1.0 km at both sites around 2100 UTC. PBL heights remain elevated through the rest of the period as mixing at the top of the PBL results in continued Bragg scatter (Bianco and Wilczak 2008) and the lack of radar data below ≈200 m prevents the sensing of any surface-based nocturnal boundary layer generation.
The root-mean-square-deviation (RMSD) and mean bias for each of the platform combinations are provided in Table 2. When including all of the unambiguous radar-derived PBL heights (i.e., group A), RMSD values are quite large at all three sites, ranging from 758 m at KSHV/CLAMPS-1 to 989 m at KOUN/CLAMPS-1. The radar-derived PBL heights are also overestimated when compared to the CLAMPS-derived heights, with biases reaching 656 m at KOUN/CLAMPS-1. These large errors and biases are due in large part to discrepancies between the estimates during the overnight, as the large majority of CLAMPS-derived overnight PBL heights are below 500 m but the radar-derived heights frequently extend up to or beyond 2.0 km. These time-dependent discrepancies are apparent in Fig. 4, which shows the CLAMPS- versus radar-derived PBL heights colored according to time. The radar-derived PBL heights for the 1200–0000 UTC period (i.e., daytime) have appreciable variance but are generally well-correlated with the CLAMPS-derived heights. After 0000 UTC, a dramatic shift occurs at all three platforms as the points collapse leftward. When including only the points that are believed to correspond to surface-based PBL (i.e., group B), the RMSD is appreciably reduced, decreasing by 45%, 38%, and 35% at KOUN/CLAMPS-2, KOUN/CLAMPS-1, and KSHV/CLAMPS-1, respectively (Table 2). While these RMSD values are higher than those of other studies that verified against radiosonde observations (Elmore et al. 2012; B18) or human judgements (Bianco and Wilczak 2008) of PBL height, this is somewhat expected given the localized and highly resolved nature of PBL growth in the CLAMPS data over the full diurnal cycle compared to QVPs. More notably, the biases are much lower in group B compared to group A, ranging from just +68 m at KSHV to +220 m at KOUN/CLAMPS-1. These represent decreases from the group A biases of 71%, 66%, and 86% at KOUN/CLAMPS-2, KOUN/CLAMPS-1, and KSHV/CLAMPS-1, respectively. While these metrics represent noteworthy improvements over the group A numbers, the group B statistics still include a number of points of very large disagreement between the radar-derived and CLAMPS-derived PBL heights primarily occurring during the early evening transition (Fig. 5), as the CLAMPS-derived heights become indicative of nocturnal stable PBLs while the radar-based heights are still sensing residual Bragg scatter at the top of the previous day’s PBL that takes a few hours to decay (Bianco and Wilczak 2008).
RMSD (m) and mean bias (m) of the radar-derived PBL heights compared to the CLAMPS-derived PBL heights.
Scatterplots comparing PBL heights estimated from CLAMPS and the corresponding collocated WSR-88D radar for all unambiguous classifications (i.e., group A), shaded according to time of day in UTC.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
As in Fig. 4 but for the group B dataset.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
While the overall biases are quite low in group B, it is instructive to examine the diurnal evolution of the biases following Heinselman et al. (2009). In their study, azimuthally averaged Z was used to infer radar-based PBL heights and compared to PBL heights derived from radar wind profilers. Figure 6 shows the hourly mean difference between the radar-derived PBL heights and the CLAMPS-derived ones. Here, we limit our period of focus to 1300–2300 UTC (i.e., from approximately one hour after sunrise to one hour before sunset) and use only data deemed to be from the clear-cut group B data for a more direct comparison with the procedures used in Heinselman et al. (2009). The findings are in good qualitative agreement with those observed in Heinselman et al. (2009): the radar-based PBL heights are overestimated through midmorning and are subsequently underestimated through the afternoon compared to those from CLAMPS, although the afternoon biases are larger in our dataset than in Heinselman et al. (2009). Differences between the studies emerge after 2100 UTC as large positive biases appear in our radar-based estimates as the CLAMPS-derived PBL heights begin to descend. The overall PBL evolution observed in Heinselman et al. (2009) in central Oklahoma appears similar to that observed at KOUN here. The average maximum PBL height occurred at around 2200 UTC, as in this study, although they observed an average maximum PBL height of ≈1.6 km compared to our 1.2 km; this difference is likely due to their study period occurring during the climatologically warmer period of July and early August compared to our IOP during late August and September. The RMSD and bias for this dataset are 342 and −58 m compared to 150 and 27 m reported in Heinselman et al. (2009), respectively. As previously discussed, this larger RMSD is likely in part a consequence of comparing point measurements of fluctuations in local PBL evolution to spatially averaged ones.
Radar-derived hourly median PBL heights minus CLAMPS-derived hourly median PBL heights, with the findings of Heinselman et al. (2009) shown for comparison.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
b. Impact of distance from radar
One of the goals of this project is to evaluate the impact distance from the radar has on the representativeness of the radar-based PBL heights. For a 4.5° elevation angle and assuming a standard atmosphere, the averaging radius at 1.0 km AGL is ≈13.0 km and at 2.0 km AGL is ≈25.0 km (Fig. 1). Therefore, the implicit assumption made in the B18 methodology is that the field obtained by averaging out the heterogeneities during the QVP processing will be representative of the PBL height at the center point. In contrast, for the typical range of PBL heights during this period, CLAMPS-1 at KAEFS was barely within or just outside of the 2-km AGL averaging area for QVPs from KOUN.
To evaluate whether PBL heights derived from KOUN are better correlated with those derived from the collocated CLAMPS-2 or the CLAMPS-1 located nearly 30-km away at KAEFS, data points from the group B dataset with valid data from all three platforms are identified and matched. Unfortunately, due to sporadic data outages at either of the CLAMPS platforms, only 130 data points from 6 different days meet this criteria. RMSDs computed with respect to KOUN are similar (468 and 488 m for CLAMPS-1 and CLAMPS-2, respectively), with corresponding biases of −34 and +121 m. An examination of the PBL heights on these days, however, reveals many instances in which the radar-derived heights are poorly correlated with PBL heights derived from both of the CLAMPS, with the CLAMPS platforms agreeing better with each other than KOUN. Thus, the similarity of the RMSD of each CLAMPS platforms is not necessarily reflective of the representativeness of the radar-derived PBL heights but rather the poor agreement of the radar-derived PBL heights with both CLAMPS-derived estimates for the days examined. For this reason and the limited sample size, it is difficult to draw definitive conclusions about the representativeness of the radar-derived PBL heights at various distances from the radar.
c. Impact of region
The overall performance of the radar-derived estimates is similar at both radar sites (Table 2), but noteworthy differences in the QVP characteristics from each radar exist (as found in B18). The most notable discrepancy in this study is the degree to which clear ZDR minima associated with the PBL top were apparent. Out of 27 usable days at KOUN (i.e., those without precipitation with identifiable PBL growth), 19 feature clear ZDR minima throughout the diurnal cycle and another 4 feature clear ZDR minima for at least a portion of the daytime hours. In contrast, out of 16 usable days at KSHV, only 8 (50%) feature clear minima associated with the top of the PBL. The other 8 feature no identifiable ZDR minima associated with the top of the PBL. Instead, the PBL features elevated (e.g., 3–4 dB) but near-constant values of ZDR up to the top of the apparent PBL, above which ZDR rapidly increases. In these instances, the observed PBL heights from the collocated CLAMPS-1 track alongside this abrupt change in ZDR, providing external validation that this transition is indeed associated with the top of the PBL. An example of this is shown in Fig. 7. Starting approximately an hour after sunrise (1300 UTC), the PBL grows steadily upward, with the ZDR rapidly increasing from 3 to 4 dB in the boundary layer to ≥6 dB immediately above the PBL top. The lack of clear minima negatively impacts the ability for the radar to estimate the PBL height using the B18 methodology; it is not until mid–late afternoon (2100–2200 UTC) in Fig. 7 that the radar-based estimate rises to where CLAMPS has determined the PBL height to be. A somewhat analogous signature with homogeneous ZDR values within the PBL was reported in B18 at KTLX, but those instances featured very low values of ZDR within the PBL attributed to Bragg scatter in the absence of biota. The elevated values of ZDR within the PBL observed here preclude the absence of biota as an explanation.
As in Fig. 2, but for KSHV on 14 Sep 2020.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
Characterizing the magnitude of the ZDR trough due to Bragg scatter is challenging due to uncertainty in how to define the “background” values of ZDR at the edges of the Bragg scatter layer. The bottom of the layer is generally well-characterized by a pronounced shift to decreasing ZDR values with height, while the top of the Bragg scatter layer is often ill-defined, often transitioning gradually into very high ZDR values associated with lofted biota. Herein, we define the magnitude of the ZDR depression (ΔZDR) with respect to the value at the bottom of the layer, although results are qualitatively similar when using an average ZDR value from the bottom and subjectively determined top of the layer. For the aforementioned cases in which no ZDR trough was apparent, a ΔZDR value of 0 dB is assigned.
Figure 8 shows a comparison of the “bottom-relative” ZDR trough magnitude compared to the observed change in RH over 500 m at the top of the PBL. Overall, there is a clear increase in the magnitude of the ZDR decrease for larger RH gradients, supporting the hypothesis that the prominence of the QVP Bragg scatter signature depends on the environment. The differences between KOUN and KSHV are clear for the IOP dates: KOUN has a median ΔZDR magnitude of −1.32 dB and 500-m RH decrease of −36.27%, while KSHV has a median ΔZDR magnitude of only −0.22 dB and 500-m RH decrease of −10.75%. Crucially, the cases at KSHV that exhibited no ZDR minimum and a stepwise change in ZDR at the PBL top all had 500-m RH changes smaller than −13%, with a mean value of just −2.7%. Qualitatively similar results are found when looking at the average ΔZDR within the layer, reflecting the notion that more heterogeneous Bragg scatter-layer heights within the QVP domain will present as deeper, but less prominent, ZDR depressions due to the azimuthal averaging. Caveats with this analysis include the fact that 1) it was sometimes difficult to define a reference ZDR value to compare against; 2) the radiosonde represents a quasi-vertical profile at a single location while the radar QVP is a spatial average; and 3) the clarity of the Bragg scatter signature in ZDR depends on the relative scattering contributions of biota (see Fig. 4 of B18). Nevertheless, the above analysis suggests that changes in Bragg scatter due to environmental variability (e.g., continental vs tropical airmasses) need to be considered when using the QVP-based approach to, for example, build PBL climatologies; should only days with unambiguous ZDR minima be included rather than the step changes we observe, the resultant climatologies may be biased toward days with larger RH gradients between the PBL and free atmosphere, with differential impacts between regions.
Scatterplot comparing the magnitude of the ZDR depression associated with the top of the PBL to the change in RH at the top of the PBL from the coincident soundings at KOUN (blue) and KSHV (red).
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
4. Considerations for QVP-based PBL-height detection
a. Detection of elevated residual layers
As mentioned previously, many of the most common and prominent discrepancies between the radar- and CLAMPS-derived PBL heights occur overnight and have been hypothesized to be due to elevated residual layers (B18). In the canonical PBL diurnal cycle (Stull 1988), the surface-based stable boundary layer begins to form after sunset, but residual turbulence may slowly descend and take hours to decay at what was the top of the convective boundary layer from the previous day. Layers of reduced ZDR, while much less prominent than during the daytime, are often observed emanating from the top of the previous day’s PBL and last through much of the overnight. A subjective examination of each site’s QVPs reveals suspected overnight residual layer signatures on 58% and 53% of nights at KOUN and KSHV, respectively. From a practical standpoint, these scenarios present difficulties for an algorithm based on ZDR minima in the column to detect the surface-based PBL height, particularly when these layers are shallow and present below the lowest available radar data and feature weak/nonexistent Bragg scatter. Despite this, detection and observation of the evolution of the residual layer is important in and of itself for understanding the turbulent decay within this layer as well as the evolution of the next day’s boundary layer (e.g., Blay-Carreras et al. 2014).
An example of probable residual layer detection is shown in Fig. 9 for 18 September 2020 at KOUN. On this day, smoke from wildfires in the western United States advected into the region overnight, with periodic reductions in visibility down to approximately 11 km reported at the Norman, Oklahoma, Automated Surface Observing System (ASOS) site (OUN) and periodic reports of smoke at the manually observed Oklahoma City, Oklahoma, ASOS site (OKC). Smoke aerosols are optimal for providing robust DL backscatter. Coincident with sunset (0031 UTC), the overall ZDR field decreases in magnitude due to different overnight biota characteristics, but a layer of subtly lower ZDR persists, beginning at roughly 1300 m and gradually descending to approximately 500 m by 0900 UTC before increasing to around 750 m between 1000 and 1200 UTC (Fig. 9a). The DL backscatter signal clearly shows the presence of two distinct layers, with one layer of aerosol trapped in the residual layer between the surface and 1500 m, another layer of smoke initially near and above 2000 m, and apparent aerosol scavenging in between. The radar-detected PBL height follows the descent of this lower layer closely, bolstering the hypothesis that this overnight reduction in ZDR is indeed due to residual mixing. By midmorning, the ascent of the daytime convective PBL tracks closely with the growth of the PBL seen in the smoke aerosols, as the scavenging layer is eroded and the two distinct layers begin to merge. A similar dual-layer structure was observed on 16 September 2020 at KOUN during another period of smoke (not shown).
Time–height series of (a) ZDR from KOUN and (b) backscatter intensity [defined as signal-to-noise ratio (SNR) + 1] from CLAMPS-2 on 18 Sep 2020. PBL heights derived from KOUN and CLAMPS-2 are overlaid in both panels. On the abscissa, sunset and sunrise are denoted with circles with down and up arrows, respectively, along with solar midnight and noon.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
b. Detection of clouds
On a number of days, sudden and discontinuous decreases in ZDR are observed that appear to correspond to the presence of clouds. One example on 20 September 2020 at KSHV is shown in Fig. 10. From 0000 to 0900 UTC there is some evidence that the radar is subtly picking up on the presence of a slowly ascending residual layer between roughly 1000 and 1200 m AGL as described in section 4a amid widespread biota with ZDR values exceeding 3 dB. Beginning around 0900 UTC, two layers of high backscatter intensity and subsequent extinction of the signal are observed by the DL (Fig. 10b), in strong agreement with cloud heights derived from ASOS ceilometer measurements (Fig. 10a). Each of these cloud base layers are collocated with a reduction in ZDR above them, with the higher layer exhibiting a stronger reduction amid more reported cloud coverage. Widespread cloudiness ceases again after 1200 UTC, after which a more standard reduction in ZDR associated with the growth of the PBL is observed. At 2300 UTC, more clouds are reported at and above 2000 m, with another attendant drop in ZDR from that point upward. In this case, the continuity requirement imposed on the radar-derived PBL heights prevented the algorithm from erroneously jumping upward in height.
As in Fig. 9, but for 20 Sep 2020 at KSHV and CLAMPS-1. White circles in (a) correspond to cloud base heights reported at the SHV ASOS station, with sizes corresponding to the reported degree of cloudiness (e.g., few, scattered, overcast).
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
While the apparent good agreement between the appearance and magnitude of these ZDR drops and the observed lidar backscatter intensity seems to support the notion that these are indeed evidence of the radar detecting clouds, it is difficult to know with certainty. Cloud droplets are spherical with an intrinsic ZDR of 0 dB, high ρhv, and can have a similar magnitude of Z as Bragg scatter, making disentangling the contributions from each difficult without additional measurements (Knight and Miller 1998; Miller et al. 1998). Knight and Miller (1998) reported on Bragg scatter and subsequent “mantle” echoes due to entrainment within/above burgeoning clouds themselves. The 1200 UTC 20 September 2020 sounding from KSHV observed an extreme moisture gradient immediately above this shallow cloud layer, with RH decreasing to 22% at 2100 m AGL; this supports the hypothesis that Bragg scatter due to refractive index fluctuations within/above the cloud may be contributing to all or part of the observed signature. While there are other days that the DL backscatter intensity field indicates clouds that have no discernible presence in the ZDR QVP, clouds are also frequently observed to be collocated with the top of the PBL through its diurnal cycle, further complicating their disentanglement. More investigation is needed regarding the role that detection of nonprecipitating clouds may be playing in the ZDR signatures, but suffice it to say that such signatures have the potential to degrade the skill of any PBL detection algorithm based on local minima of ZDR if caution is not exercised, as seen between 0600 and 1300 UTC in Fig. 10a.
c. Detection of ground clutter
Ground clutter occurs when the atmosphere’s refractivity profile causes part—or, in rare instances, all—of the radar beam to scatter off of ground-based targets and has historically been considered undesirable contamination of weather radar data. The desire to minimize its impact in part motivated the decision to use the highest-available (clear-air mode) elevation angle when generating the QVPs in this and other work. Despite this, ground clutter is apparently present during many of the cases at both sites, particularly at night; it is subsequently difficult at times to find the top of the PBL by simply finding the height of the minimum ZDR in the QVP column and motivated us to modify the B18 approach by adding a continuity constraint to the search algorithm, which helped address, but did not completely alleviate, this issue.
While we are often unable to scan low enough to observe the nocturnal stable boundary layer as discussed in section 2d, we hypothesize that ground clutter itself may actually have utility for retrieving information relevant to the PBL. For example, methods have been developed for estimating the vertical refractivity profile from ground clutter (Fabry et al. 1997; Besson and Parent du Châtelet 2013; Feng et al. 2016). Similarly, Sakar et al. (1992) and Bibraj et al. (2020) showed that the vertical refractivity gradient caused by nocturnal temperature inversions is related to the degree of ground clutter on radar. On multiple occasions, we observe overnight ground clutter signatures that grew overnight (either in depth, magnitude of the ZDR reduction, or both). Two examples of this are shown in Fig. 11 at KOUN. On 7 September 2020, a reduction in ZDR believed to be associated with ground clutter appears around 0300 UTC (i.e., approximately two hours after sunset) and gradually intensifies throughout the overnight hours (Fig. 11a). The attendant nocturnal temperature inversion retrieved from the AERI (Fig. 11c) similarly forms and intensifies throughout the overnight hours, reaching a peak magnitude between 0900 and 1200 UTC, in agreement with the ZDR field, before more typical mixing-based Bragg scatter signatures begin to dominate during the daytime. Another example is shown in Figs. 11b and 11d for 16 September 2020. In this case, no ground clutter signature is apparent until 1000 UTC, when a sudden but minor reduction of ZDR occurs at the lowest available data levels (Fig. 11b). Similarly, the AERI thermodynamic retrievals (Fig. 11d) show no low-level temperature inversion forming until 1000 UTC and lasting until 1300 UTC, in agreement with the apparent ground clutter. Beyond the unavoidable lack of low-level radar data due to beam heights increasing with distance from the radar, these examples suggest that the ground clutter signature in ZDR QVPs may have some utility for at least estimating the magnitude of the surface-based temperature inversion associated with the nocturnal stable boundary layer. That said, the potential for reliably implementing such a retrieval and translating it to nocturnal PBL height remains purely speculative.
Time–height series of (a),(c) ZDR from KOUN and (b),(d) vertical temperature gradient retrieved from the AERI from CLAMPS-2 for (left) 7 Sep 2020 and (right) 16 Sep 2020. The KOUN-derived and CLAMPS-derived PBL heights are overlaid in both panels for each case. On the abscissa, sunset and sunrise are denoted with circles with down and up arrows, respectively, along with solar midnight and noon.
Citation: Journal of Applied Meteorology and Climatology 63, 7; 10.1175/JAMC-D-23-0231.1
d. No detection of Bragg scatter
In some instances, Bragg scatter is simply not apparent in the QVPs despite clear-cut canonical PBL growth indicated in the CLAMPS data. One such case is shown in Fig. 11b, where a reduction in ZDR due to Bragg scatter only becomes apparent in the afternoon (i.e., after 1800 UTC) despite PBL growth driven by convective mixing (not shown). An examination of 1800 UTC radiosonde data from the nearby Southern Great Plains Atmospheric Radiation Measurement (ARM; Stokes and Schwartz 1994) site in Lamont, Oklahoma, for both this and another case absent of Bragg scatter (29 August 2020 at KOUN) reveals pronounced temperature and moisture inversions at the top of the PBL. As such, it is not immediately clear why no Bragg scatter signature is detected given a strong vertical refractivity gradient and turbulent mixing occurring, although the deleterious effects on automated QVP-based PBL detections are obvious. Other studies (B18; Comer et al. 2023) similarly report an unexplained absence of observable Bragg scatter; in B18 Bragg scatter was absent on 1 out of every 7 days examined, making this a not-uncommon situation.
5. Summary
In this study, a proposed methodology for estimating PBL height from ZDR QVPs described in B18 is evaluated using collocated specialized boundary layer profiling measurements aboard two mobile CLAMPS platforms. CLAMPS platforms were deployed in central Oklahoma near and collocated with the KOUN radar and in Shreveport, Louisiana collocated with the KSHV radar in August–September of 2020. The goals of the study were to
-
further explore and contextualize some of the signatures observed in B18,
-
evaluate the reliability of the method through the full diurnal cycle rather than at standard radiosonde launch times,
-
investigate the impact of distance from the radar on the representativeness of the radar-based PBL heights, and
-
determine the impact of geographical/climatological region on the characteristics of observed Bragg scatter signatures.
Overall, our results support the findings of B18 that the method is reliable for capturing the general diurnal growth of the daytime convective PBL. However, local fluctuations in PBL growth observed by the CLAMPS are not reflected in the QVP, resulting in instantaneous RMSDs of nearly 500 m at each radar site compared to the collocated CLAMPS platforms. Insufficient data was available to quantify the impact of distance from radar compared to the azimuthally averaged QVPs. Radar-based PBL estimates are lower than those from CLAMPS by, on average, roughly 200 m during the daytime, in agreement with past studies. In the hour preceding sunset (i.e., after 2300 UTC), the radar- and CLAMPS-derived PBL estimates diverge appreciably. Ongoing Bragg scatter at the top of the daytime boundary layer/forming residual layer results in a continued ZDR reduction and a persistent PBL height estimate. CLAMPS data, however, indicate the cessation of mixing and formation of a low-level nocturnal stable PBL resulting in a rapid decrease in PBL height. This hints at the challenge of confidently defining a PBL during transition regimes (i.e., morning and early evening).
Detection of the boundary layer overnight was more challenging, with the radar-based PBL often identifying Bragg scatter within the elevated residual layer. Low-level nocturnal stable PBLs are often obscured by the radar’s cone-of-silence and ground clutter, although instances are identified in which the presence of ground clutter itself appears to be a reliable indicator of temperature inversion intensity and, subsequently, the nocturnal stable PBL. The detection of residual layers is itself worthwhile but must also be taken into account when constructing climatologies. Apparent clear-air-mode detection of clouds can also create difficulties for a PBL detection algorithm based solely on the minimum ZDR in the profile, although there is outstanding uncertainty regarding whether such signatures are due to cloud hydrometeors or Bragg scatter associated with clouds themselves. It is also demonstrated that the environment is correlated with the intensity of the Bragg scatter signature, with stronger vertical moisture gradients resulting in, on average, deeper ZDR depressions and the absence of radar-indicated Bragg scatter in more moist environments lacking strong refractivity gradients, which should be taken into account when selecting cases to include when building climatologies. Still, other instances are observed of an apparent absence of Bragg scatter despite refractivity gradients and strong turbulent mixing and canonical PBL growth observed by CLAMPS.
CLAMPS and QVPs both provide estimates of PBL height but do so with drastically different observation and estimation principles; there is no reason to expect the QVP-based PBL estimates to mirror those of CLAMPS through the full diurnal cycle. However, the use of specialized measurements for observing PBL evolution like the ones CLAMPS provides allows for a more thorough look at the potential strengths and limitations of the PBL information provided by QVPs. A key takeaway of this study is the challenge of automating such a detection algorithm when looking at noncanonical real-world data. Even with smoothing and the continuity component described in section 2c, manual analysis was often required to work through noncanonical cases and ZDR minima due to phenomena other than PBL-driven Bragg scatter, with some cases remaining indeterminate. The development of new methods for reliably and automatically identifying layers of Bragg scatter from QVPs (e.g., Comer et al. 2023; Stouffer et al. 2023) is essential and should be further explored. In addition to the need for more data, particularly in other regions and seasons, further efforts should be made to improve filtering and the robustness of the detection algorithm in order to be able to scale up the amount of radar data processed and generate representative, comprehensive PBL climatologies. Future work exploring machine learning approaches that incorporate holistic datasets of polarimetric QVPs (potentially including other variables such as spectrum width, which may hold additional information about turbulence at the top of the PBL), regularly scheduled soundings, and specialized data sources such as CLAMPS may prove fruitful in combating many of these outstanding challenges of automation.
Acknowledgments.
Funding was provided by NOAA/Office of Oceanic and Atmospheric Research under NOAA–University of Oklahoma Cooperative Agreements NA16OAR4320115 and NA21OAR4320204, U.S. Department of Commerce. The authors wish to thank the National Weather Office in Shreveport, Louisiana, for graciously hosting the CLAMPS platform on-site, the CLAMPS team for aiding with deployments, Charles Kuster and the Radar Operations Center for help operating KOUN, the CIWRO administration for helping with deployment logistics amid COVID-19 restrictions, the three anonymous reviewers whose feedback improved this manuscript, and the CIWRO Director’s Discretionary Research Fund for selecting and funding this project.
Data availability statement.
The CLAMPS observations associated with this campaign are available on a THREDDS server at https://data.nssl.noaa.gov/thredds/catalog/FRDD/CLAMPS/campaigns/PBLtops/catalog.html, and the corresponding PBL depth detection algorithm is available at a Github repository at https://github.com/OAR-atmospheric-observations/bliss-fl. Archived KSHV and KOUN WSR-88D data are available from the NSSL THREDDS server at https://data.nssl.noaa.gov/thredds/catalog/PARR/2020/Jacob-Carlin/pbl-tops.html and archived radiosonde data are available from the University of Wyoming at https://weather.uwyo.edu/upperair/sounding.html.
While not discussed in this study, CLAMPS data were made available in real time to forecasters at both NWS offices to capitalize on this research-to-operations exchange.
REFERENCES
Agee, E. M., and M. L. Hart, 1990: Boundary layer and mesoscale structure over Lake Michigan during a wintertime cold air outbreak. J. Atmos. Sci., 47, 2293–2316, https://doi.org/10.1175/1520-0469(1990)047<2293:BLAMSO>2.0.CO;2.
Angevine, W. M., A. B. White, and S. K. Avery, 1994: Boundary-layer depth and entrainment zone characterization with a boundary-layer profiler. Bound.-Layer Meteor., 68, 375–385, https://doi.org/10.1007/BF00706797.
Banghoff, J. R., D. J. Stensrud, and M. R. Kumjian, 2018: Convective boundary layer depth estimation from S-band dual-polarization radar. J. Atmos. Oceanic Technol., 35, 1723–1733, https://doi.org/10.1175/JTECH-D-17-0210.1.
Besson, L., and J. Parent du Châtelet, 2013: Solutions for improving the radar refractivity measurement by taking operational constraints into account. J. Atmos. Oceanic Technol., 30, 1730–1742, https://doi.org/10.1175/JTECH-D-12-00167.1.
Bianco, S., J. M. Wilczak, and A. B. White, 2008: Convective boundary layer depth estimation from wind profilers: Statistical comparison between an automated algorithm and expert estimations. J. Atmos. Oceanic Technol., 25, 1397–1413, https://doi.org/10.1175/2008JTECHA981.1.
Bibraj, R., B. A. M. Kannan, K. R. Rao, and K. C. Saikrishnan, 2020: An analysis of anomalous propagation parameters and its effect on the intensity of clutter in weather radars. Mausam, 71, 11–20, https://doi.org/10.54302/mausam.v71i1.2.
Blay-Carreras, E., and Coauthors, 2014: Role of the residual layer and large-scale subsidence on the development and evolution of the convective boundary layer. Atmos. Chem. Phys., 14, 4515–4530, https://doi.org/10.5194/acp-14-4515-2014.
Bonin, T. A., B. J. Caroll, R. M. Hardesty, W. A. Brewer, K. Hajny, O. E. Salmon, and P. B. Shepson, 2018: Doppler lidar observations of the mixing height in Indianapolis using an automated composite fuzzy logic approach. J. Atmos. Oceanic Technol., 35, 473–490, https://doi.org/10.1175/JTECH-D-17-0159.1.
Browning, K. A., and Coauthors, 2007: The Convective Storm Initiation Project. Bull. Amer. Meteor. Soc., 88, 1939–1955, https://doi.org/10.1175/BAMS-88-12-1939.
Carvalho, D., A. Rocha, M. Gómez-Gesteira, and C. Silva Santos, 2014: Sensitivity of the WRF model wind simulation and wind energy production estimates to planetary boundary layer parameterizations for onshore and offshore areas in the Iberian Peninsula. Appl. Energy, 135, 234–246, https://doi.org/10.1016/j.apenergy.2014.08.082.
Clarke, J. F., 1969: Nocturnal urban boundary layer over Cincinnati, Ohio. Mon. Wea. Rev., 97, 582–589, https://doi.org/10.1175/1520-0493(1969)097<0582:NUBLOC>2.3.CO;2.
Clements, C. B., and Coauthors, 2007: Observing the dynamics of wildland grass fires: FireFlux—A field validation experiment. Bull. Amer. Meteor. Soc., 88, 1369–1382, https://doi.org/10.1175/BAMS-88-9-1369.
Comer, C. L., B. Stouffer, D. J. Stensrud, M. Kumjian, and Y. Zhang, 2023: Automated detection of boundary layer depth using dual-polarization radar observations. 40th Conf. on Radar Meteorology, Minneapolis, MN, Amer. Meteor. Soc., 13A.6, https://ams.confex.com/ams/40RADAR/meetingapp.cgi/Paper/426182.
Cooper, D. I., and W. E. Eichinger, 1994: Structure of the atmosphere in an urban planetary boundary layer from lidar and radiosonde observations. J. Geophys. Res., 99, 22 937–22 948, https://doi.org/10.1029/94JD01944.
Crook, N. A., 1996: Sensitivity of moist convection forced by boundary layer processes to low-level thermodynamic fields. Mon. Wea. Rev., 124, 1767–1785, https://doi.org/10.1175/1520-0493(1996)124<1767:SOMCFB>2.0.CO;2.
Crum, T. D., R. L. Alberty, and D. W. Burgess, 1993: Recording, archiving, and using WSR-88D data. Bull. Amer. Meteor. Soc., 74, 645–654, https://doi.org/10.1175/1520-0477(1993)074<0645:RAAUWD>2.0.CO;2.
Dabberdt, W. F., and Coauthors, 2004: Meteorological research needs for improved air quality forecasting: Report of the 11th prospectus development team of the U.S. Weather Research Program. Bull. Amer. Meteor. Soc., 85, 563–586, https://doi.org/10.1175/BAMS-85-4-563.
Davison, J. L., R. M. Rauber, and L. D. Girolamo, 2013: A revised conceptual model of the tropical marine boundary layer. Part II: Detecting relative humidity layers using Bragg scattering from S-band radar. J. Atmos. Sci., 70, 3025–3046, https://doi.org/10.1175/JAS-D-12-0322.1.
De Tomasi, F., M. M. Miglietta, and M. R. Perrone, 2011: The growth of the planetary boundary layer at a coastal site: A case study. Bound.-Layer Meteor., 139, 521–541, https://doi.org/10.1007/s10546-011-9592-6.
Doviak, R. J., and D. S. Zrnić, 2006: Doppler Radar and Weather Observations. 2nd ed. Dover Publications, 562 pp.
Draxl, C., A. N. Hahmann, A. Peña, and G. Giebel, 2014: Evaluating winds and vertical wind shear from Weather Research and Forecasting model forecasts using seven planetary boundary layer schemes. Wind Energy, 17, 39–55, https://doi.org/10.1002/we.1555.
Elmore, K. L., P. L. Heinselman, and D. J. Stensrud, 2012: Using WSR-88D data and insolation estimates to determine convective boundary layer depth. J. Atmos. Oceanic Technol., 29, 581–588, https://doi.org/10.1175/JTECH-D-11-00043.1.
Eure, K. C., P. D. Mykolajtchuk, Y. Zhang, D. J. Stensrud, F. Zhang, S. J. Greybush, and M. R. Kumjian, 2023: Simultaneous assimilation of planetary boundary layer observations from radar and all-sky satellite observations to improve forecasts of convection initiation. Mon. Wea. Rev., 151, 795–813, https://doi.org/10.1175/MWR-D-22-0188.1.
Fabry, F., C. Frush, I. Zawadzki, and A. Kilambi, 1997: On the extraction of near-surface index of refraction using radar phase measurements from ground targets. J. Atmos. Oceanic Technol., 14, 978–987, https://doi.org/10.1175/1520-0426(1997)014<0978:OTEONS>2.0.CO;2.
Feng, Y.-C., F. Fabry, and T. M. Weckwerth, 2016: Improving radar refractivity retrieval by considering the change in the refractivity profile and the varying altitudes of ground targets. J. Atmos. Oceanic Technol., 33, 989–1004, https://doi.org/10.1175/JTECH-D-15-0224.1.
Heinselman, P. L., P. L. Spencer, K. L. Elmore, D. J. Stensrud, R. M. Hluchan, and P. C. Burke, 2009: Radar reflectivity-based estimates of mixed layer depth. J. Atmos. Oceanic Technol., 26, 229–239, https://doi.org/10.1175/2008JTECHA1091.1.
Knight, C. A., and L. J. Miller, 1993: First radar echoes from cumulus clouds. Bull. Amer. Meteor. Soc., 74, 179–188, https://doi.org/10.1175/1520-0477(1993)074<0179:FREFCC>2.0.CO;2.
Knight, C. A., and L. J. Miller, 1998: Early radar echoes from small, warm cumulus: Bragg and hydrometeor scattering. J. Atmos. Sci., 55, 2974–2992, https://doi.org/10.1175/1520-0469(1998)055<2974:EREFSW>2.0.CO;2.
Kristovich, D. A. R., N. F. Laird, and M. R. Hjemfelt, 2003: Convective evolution across Lake Michigan during a widespread lake-effect snow event. Mon. Wea. Rev., 131, 643–655, https://doi.org/10.1175/1520-0493(2003)131<0643:CEALMD>2.0.CO;2.
Kumjian, M. R., A. V. Ryzhkov, H. D. Reeves, and T. J. Schuur, 2013: A dual-polarization radar signature of hydrometeor refreezing in winter storms. J. Appl. Meteor. Climatol., 52, 2549–2566, https://doi.org/10.1175/JAMC-D-12-0311.1.
Melnikov, V. M., R. J. Doviak, D. S. Zrnić, and D. J. Stensrud, 2011: Mapping Bragg scatter with a polarimetric WSR-88D. J. Atmos. Oceanic Technol., 28, 1273–1285, https://doi.org/10.1175/JTECH-D-10-05048.1.
Melnikov, V. M., R. J. Doviak, D. S. Zrnić, and D. J. Stensrud, 2013: Structures of Bragg scatter observed with the polarimetric WSR-88D. J. Atmos. Oceanic Technol., 30, 1253–1258, https://doi.org/10.1175/JTECH-D-12-00210.1.
Miao, Y., and S. Liu, 2019: Linkages between aerosol pollution and planetary boundary layer structure in China. Sci. Total Environ., 650, 288–296, https://doi.org/10.1016/j.scitotenv.2018.09.032.
Miller, M. A., J. Verlinde, C. V. Gilbert, G. J. Lehenbauer, J. S. Tongue, and E. E. Clothiaux, 1998: Detection of nonprecipitating clouds with the WSR-88D: A theoretical and experimental survey of capabilities and limitations. Wea. Forecasting, 13, 1046–1062, https://doi.org/10.1175/1520-0434(1998)013<1046:DONCWT>2.0.CO;2.
Mykolajtchuk, P., 2021: Diagnosing a missed convective initiation forecast by assimilating GOES-16 satellite radiances and WSR-88D radar observations. M.S. thesis, Dept. of Meteorology and Atmospheric Science, The Pennsylvania State University, 64 pp.
NOAA Radar Operations Center, 2022: NEXRAD technical information. Accessed 4 February 2024, https://www.roc.noaa.gov/WSR88D/Engineering/NEXRADTechInfo.aspx.
Pullen, J., J. Ching, D. Sailor, W. Thompson, B. Bornstein, and D. Koracin, 2008: Progress toward meeting the challenges of our coastal urban future. Bull. Amer. Meteor. Soc., 89, 1727–1731, https://doi.org/10.1175/2008BAMS2560.1.
Rabin, R. M., and R. J. Doviak, 1989: Meteorological and astronomical influences on radar reflectivity in the convective boundary layer. J. Appl. Meteor., 28, 1226–1235, https://doi.org/10.1175/1520-0450(1989)028<1226:MAAIOR>2.0.CO;2.
Richardson, L. M., J. G. Cunningham, W. D. Zittel, R. R. Lee, R. L. Ice, V. M. Melnikov, N. P. Hoban, and J. G. Gebauer, 2017: Bragg scatter detection by the WSR-88D. Part I: Algorithm development. J. Atmos. Oceanic Technol., 34, 465–478, https://doi.org/10.1175/JTECH-D-16-0030.1.
Ryzhkov, A. V., P. Zhang, H. Reeves, M. R. Kumjian, T. Tschallener, S. Trömel, and C. Simmer, 2016: Quasi-vertical profiles—A new way to look at polarimetric radar data. J. Atmos. Oceanic Technol., 33, 551–562, https://doi.org/10.1175/JTECH-D-15-0020.1.
Sakar, S. K., P. K. Pasricha, A. B. Ghosh, H. N. Dutta, M. V. S. N. Prasad, and B. M. Reddy, 1992: High resolution refractivity measurement and its influence on radar clutter extent. Indian J. Radio Space Phys., 21, 5–9.
Seidel, D. J., C. O. Ao, and K. Li, 2010: Estimating climatological planetary boundary layer heights from radiosonde observations: Comparison of methods and uncertainty analysis. J. Geophys. Res., 115, D16113, https://doi.org/10.1029/2009JD013680.
Smith, E. N., and J. T. Carlin, 2023: A multi-instrument fuzzy logic boundary-layer top detection algorithm. EGUsphere, 2023-2050, https://doi.org/10.5194/egusphere-2023-2050.
Stensrud, D. J., C. L. Comer, B. STouffer, Y. Zhang, and M. Kumjian, 2024: Evolution of monthly mean convective boundary layer depths detected from WSR-88D observations. 24th Symp. on Meteorological Observation and Instrumentation, Baltimore, MD, Amer. Meteor. Soc., 12.3, https://ams.confex.com/ams/104ANNUAL/meetingapp.cgi/Paper/437557.
Stokes, G. M., and S. E. Schwartz, 1994: The Atmospheric Radiation Measurement (ARM) program: Programmatic background and design of the cloud and radiation test bed. Bull. Amer. Meteor. Soc., 75, 1201–1222, https://doi.org/10.1175/1520-0477(1994)075<1201:TARMPP>2.0.CO;2.
Stouffer, B., C. L. Comer, D. J. Stensrud, M. Kumjian, and Y. Zhang, 2023: Exploring convective boundary layer depth and entrainment zone properties with dual-polarization radar observations. 40th Conf. on Radar Meteorology, Minneapolis, MN, Amer. Meteor. Soc., 6B.1, https://ams.confex.com/ams/40RADAR/meetingapp.cgi/Paper/426181.
Stull, R. B., 1988: An Introduction to Boundary Layer Meteorology. Kluwer Academic, 666 pp.
Surussavadee, C., 2017: Evaluation of WRF near-surface wind simulations in tropics employing different planetary boundary layer schemes. 2017 Eighth Int. Renewable Energy Congress (IREC), Amman, Jordan, Institute of Electrical and Electronics Engineers, 1–4, https://doi.org/10.1109/IREC.2017.7926005.
Takle, E. S., D. A. Rajewski, and S. L. Purdy, 2019: The Iowa Atmospheric Observatory: Revealing the unique boundary layer characteristics of a wind farm. Earth Interact., 23, https://doi.org/10.1175/EI-D-17-0024.1.
Tangborn, A., B. Demoz, B. J. Carroll, J. Santanello, and J. L. Anderson, 2021: Assimilation of lidar planetary boundary layer height observations. Atmos. Meas. Tech., 14, 1099–1110, https://doi.org/10.5194/amt-14-1099-2021.
Turner, D. D., and W. G. Blumberg, 2019: Improvements to the AERIoe thermodynamic profile retrieval algorithm. IEEE J. Sel. Topics Appl. Earth Obs. Remote Sens., 12, 1339–1354, https://doi.org/10.1109/JSTARS.2018.2874968.
Virtanen, P., and Coauthors, 2020: SciPy 1.0: Fundamental algorithms for scientific computing in Python. Nat. Methods, 17, 261–272, https://doi.org/10.1038/s41592-019-0686-2.
Wagner, T. J., P. M. Klein, and D. D. Turner, 2019: A new generation of ground-based mobile platforms for active and passive profiling of the boundary layer. Bull. Amer. Meteor. Soc., 100, 137–153, https://doi.org/10.1175/BAMS-D-17-0165.1.
White, A. B., 1993: Mixing depth detection using 915-MHz radar reflectivity data. Eighth Symp. on Observations and Instrumentation, Anaheim, CA, Amer. Meteor. Soc., 248–250.
Xu, S., C. Su, X. Wu, T. Luo, and X. Li, 2020: A simple method to estimate the refractive index structure parameter (C n 2) in the atmosphere. Publ. Astron. Soc. Pac., 132, 084501, https://doi.org/10.1088/1538-3873/ab9451.
Yuval, Y., Y. Levi, U. Dayan, I. Levy, and D. M. Broday, 2020: On the association between characteristics of the atmospheric boundary layer and air pollution concentrations. Atmos. Res., 231, 104675, https://doi.org/10.1016/j.atmosres.2019.104675.
