PBL Height Retrievals during ASKOS Campaign †

: This study analyzes the structure of the Planetary Boundary Layer (PBL) at Mindelo, Cabo Verde, where the ASKOS Campaign took place from 2021 to 2022. Datasets from ground-based remote sensing instruments and radiosondes are used to derive the PBL height, by applying the Wavelet Covariance Transform (WCT), Threshold (TM), and Gradient Method (GM). Two case studies are described in detail, one with a signiﬁcant dust load (23 September 2022) and one with relatively less dust load (12 September 2022). In the ﬁrst case, the PBL top is found lower, and the methods used for the retrievals are characterized by larger uncertainties. In the second case, a higher and more convective PBL is observed. Additionally, results are compared with ECMWF outputs, establishing good agreement.


Introduction
The ASKOS experiment [1] is the ground-based component of the Joint Aeolus Tropical Atlantic Campaign (JATAC) and was carried out at Cabo Verde during the summers and autumns of 2021 and 2022.The primary aim of ASKOS was the collection of a novel dataset of synergistic measurements in the region, to address a wide range of scientific objectives, including the processes affecting dessert dust transport and the effect of dust on boundary layer dynamics.Within ASKOS, a full ACTRIS (Aerosol Cloud and Trace Gasses Research Infrastructure) aerosol and cloud remote sensing station was established at Cabo Verde, Mindelo city (16.87 • N, 24.99 • W) on the island of São Vicente, where intense ground-based remote-sensing and UAV-based airborne in situ measurements were conducted.
The purpose of this study is to map the Planetary Boundary Layer (PBL) [2] at the São Vicente island of Cabo Verde and investigate the capabilities of the models to accurately retrieve the PBL height of this coastal site.Additionally, we will investigate PBL height characteristics in the ASKOS dataset under different dust load conditions.The experimental site is a tropical area far away from the western coast of the African continent and is affected by Saharan dust and marine aerosols.These conditions shape an advantageous field for investigating the effect of aerosol presence on PBL evolution, with high-quality atmospheric observations.Furthermore, very shallow PBL heights can occur at coastal locations [3].The continuous monitoring of the boundary-layer top with active remote sensing has been verified in past studies using several methods [4][5][6][7][8].Studying PBL using lidars can suffer from many restrictions related to weather conditions, range, and accuracy, but it can also provide high temporal and spatial resolution.
The manuscript is structured in three sections.Section 2 describes the data and the methodology in detail.Section 3 presents two case studies analyzed in detail, one with a thick dust layer present on the lower troposphere and one with relatively less dust load.A statistical analysis for a short time period is also performed, in order to depict the difference between the two cases.In Section 4, the main conclusions are summarized.

Data and Methodology
In this study, we used measurements from the September 2022 ASKOS operations.More specifically, collocated PollyXT Raman Lidar [9] and Halo Wind Doppler Lidar [10] profiles were used to derive the diurnal PBL height.Radiosonde profiles were also analyzed to provide the dynamic structure of the lower troposphere and, furthermore, evaluate the remote sensing measurements.Additionally, the observed results were then compared to the ERA5 Re-Analysis dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF) [11], at a 0.25 • × 0.25 • resolution with 137 levels [12].
The methodologies used to identify the PBL top are depicted in Figure 1: Wavelet Covariance Transform (WCT) [13]; Gradient Method (GM) [14]; and Threshold Method (TM) [3].height characteristics in the ASKOS dataset under different dust load conditions.The experimental site is a tropical area far away from the western coast of the African continent and is affected by Saharan dust and marine aerosols.These conditions shape an advantageous field for investigating the effect of aerosol presence on PBL evolution, with highquality atmospheric observations.Furthermore, very shallow PBL heights can occur at coastal locations [3].The continuous monitoring of the boundary-layer top with active remote sensing has been verified in past studies using several methods [4][5][6][7][8].Studying PBL using lidars can suffer from many restrictions related to weather conditions, range, and accuracy, but it can also provide high temporal and spatial resolution.
The manuscript is structured in three sections.Section 2 describes the data and the methodology in detail.Section 3 presents two case studies analyzed in detail, one with a thick dust layer present on the lower troposphere and one with relatively less dust load.A statistical analysis for a short time period is also performed, in order to depict the difference between the two cases.In Section 4, the main conclusions are summarized.

Data and Methodology
In this study, we used measurements from the September 2022 ASKOS operations.More specifically, collocated PollyXT Raman Lidar [9] and Halo Wind Doppler Lidar [10] profiles were used to derive the diurnal PBL height.Radiosonde profiles were also analyzed to provide the dynamic structure of the lower troposphere and, furthermore, evaluate the remote sensing measurements.Additionally, the observed results were then compared to the ERA5 Re-Analysis dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF) [11], at a 0.25° × 0.25° resolution with 137 levels [12].
The methodologies used to identify the PBL top are depicted in Figure 1: Wavelet Covariance Transform (WCT) [13]; Gradient Method (GM) [14]; and Threshold Method (TM) [3].The WCT method is applied on the 30 min averaged Water Vapor Mixing Ratio (WVMR) product of the PollyXT Lidar and on the attenuated backscatter coefficient (BSC) of both the PollyXT and Halo Wind Lidar [4].The TM is applied on the 1 h averaged Turbulent Kinetic Energy dissipation rate (TKE) product of the Halo Lidar with the threshold value of 0.00015 m 2 s −3 [3], and the GM is applied to potential temperature and relative humidity radiosonde profiles in order to acquire the PBL top [14].The WCT method is applied on the 30 min averaged Water Vapor Mixing Ratio (WVMR) product of the PollyXT Lidar and on the attenuated backscatter coefficient (BSC) of both the PollyXT and Halo Wind Lidar [4].The TM is applied on the 1 h averaged Turbulent Kinetic Energy dissipation rate (TKE) product of the Halo Lidar with the threshold value of 0.00015 m 2 s −3 [3], and the GM is applied to potential temperature and relative humidity radiosonde profiles in order to acquire the PBL top [14].

Results
Two cases with different dynamic characteristics and aerosol loads are presented in detail: 12 and 23 September.

Case Study: 12 September 2022-Light Dust Load
On 12 September 2022, an elevated dust layer was present above approximately 1.5 km, with an Aerosol Optical Depth (AOD) of 0.45 at 500 nm (Appendix A).This aerosol condition represents a common day in Mindelo.
In Figure 2, all the retrievals of that case are cumulatively depicted.Orange shading represents the dust layer, and blue shading represents the turbulence.Overall, the results for PBL height show good agreement during the entire day (after 17:00 UTC Halo measurements were not available).PBL WVMR_PollyXT is only available during 00:00-06:00 and 20:00-24:00 UTC because of the nighttime operation of the PollyXT water vapor channel.Northwestern winds dominate below 1 km (not shown), except for 2:00-4:00, when they become north winds.At this time space, TKE is dampened, and the TM method does not suggest the existence of a mixing layer, while PBL WVMR_PollyXT seems to be overestimated compared to the other observations at the beginning of the day.On the other hand, during the daytime, a shallow PBL is observed for all methods.That limited daytime evolution is an indicative feature of coastal areas.The radiosonde launched at 10:00 captures the highest PBL top at 1.1 km.PBL ECMWF is in good agreement with the rest of the retrievals.

Results
Two cases with different dynamic characteristics and aerosol loads are presented in detail: 12 and 23 September.

Case Study: 12 September 2022-Light Dust Load
On 12 September 2022, an elevated dust layer was present above approximately 1.5 km, with an Aerosol Optical Depth (AOD) of 0.45 at 500 nm (Appendix A).This aerosol condition represents a common day in Mindelo.
In Figure 2, all the retrievals of that case are cumulatively depicted.Orange shading represents the dust layer, and blue shading represents the turbulence.Overall, the results for PBL height show good agreement during the entire day (after 17:00 UTC Halo measurements were not available).PBLWVMR_PollyXT is only available during 00:00-06:00 and 20:00-24:00 UTC because of the nighttime operation of the PollyXT water vapor channel.Northwestern winds dominate below 1 km (not shown), except for 2:00-4:00, when they become north winds.At this time space, TKE is dampened, and the TM method does not suggest the existence of a mixing layer, while PBLWVMR_PollyXT seems to be overestimated compared to the other observations at the beginning of the day.On the other hand, during the daytime, a shallow PBL is observed for all methods.That limited daytime evolution is an indicative feature of coastal areas.The radiosonde launched at 10:00 captures the highest PBL top at 1.1 km.PBLECMWF is in good agreement with the rest of the retrievals.

Case Study: 23 September 2022-Heavy Dust Load
On 23 September 2022, a few low-level cumulus clouds and a significant aerosol load with AOD that exceeded 0.7 at 500 nm (Appendix A)were present.A thick dust layer penetrates at lower altitudes below 1 km, affecting the evolution of PBL.In Figure 3, there are some alienated points above 1 km before 6:00 and after 15:00 that do not correspond to PBL values.These 'outliers' of PBLBSC_PollyXT, PBLBSC_Halo, and PBLWVMR_PollyXT are the results of WCT, detecting elevated aerosol layers instead of the PBL top.However, the TM at TKE dissipation rate and ECMWF result in a more effective PBL representation than WCT.

Case Study: 23 September 2022-Heavy Dust Load
On 23 September 2022, a few low-level cumulus clouds and a significant aerosol load with AOD that exceeded 0.7 at 500 nm (Appendix A) were present.A thick dust layer penetrates at lower altitudes below 1 km, affecting the evolution of PBL.In Figure 3, there are some alienated points above 1 km before 6:00 and after 15:00 that do not correspond to PBL values.These 'outliers' of PBL BSC_PollyXT , PBL BSC_Halo, and PBL WVMR_PollyXT are the results of WCT, detecting elevated aerosol layers instead of the PBL top.However, the TM at TKE dissipation rate and ECMWF result in a more effective PBL representation than WCT.
As mentioned also in the previous case, no sharp daytime evolution is observed.On the contrary, lower PBL values are recorded during the convective hours, relative to 12 September.PBL ECMWF is in good agreement with PBL TKE_Halo during all day and with all the results during 10:00-18:00.The radiosondes of 05:22 and 19:38 UTC capture As mentioned also in the previous case, no sharp daytime evolution is observed.On the contrary, lower PBL values are recorded during the convective hours, relative to 12 September.PBLECMWF is in good agreement with PBLTKE_Halo during all day and with all the results during 10:00-18:00.The radiosondes of 05:22 and 19:38 UTC capture PBL heights that are in good accordance with remote sensing (PBLBSC_PollyXT, PBLBSC_Halo, PBLWVMR_Pol-lyXT, and PBLTKE_Halo) and model (PBLECMWF) results.

Comparison of the Two Cases
In both cases, the measured PBL presents several characteristics that resemble a Marine Atmospheric Boundary Layer (MABL): it is a relatively shallow layer with no significant variabilities on the top.Table 1 shows the average and standard deviation of PBLBSC_PollyXT, PBLBSC_Halo, PBLTKE_Halo, and PBLECMWF.

Time Space (UTC)
PBLBSC_PollyXT PBLBSC_Halo PBLTKE_Halo PBLECMWF 12 September 10:00-14:00 1  During the period 10:00-14:00 UTC (that corresponds to 9:00-13:00 local time), the standard deviation for both cases varies between 21.5 and 86.4 m, indicating low variability that is connected with the limited daytime evolution.The study on 23 September presented lower PBL retrievals and also smaller uncertainties, compared to 12 September during that time period.This result is a combination of the dust amount and the dust layer level.On 23 September, the layer was thick and infiltrated in lower levels (approx. 1 km), reducing the amount of solar radiation [15] that reaches the surface and also capping the top of the PBL (Appendix A).Moreover, the lower troposphere is less turbulent than on 12 September and is, therefore, more stable.

Conclusions
Our PBL results using WCT and TM are in good agreement with the PBL derived from radiosonde profiles, indicating trustworthy references for the detection of layering.BSC and WVMR are proportional to the aerosol concentration and constitute a possible indicator for PBL detection.However, when an elevated aerosol layer approaches the surface, it is likely that WCT results in big variabilities for the detection of layering, since this elevated layer is captured instead of PBL (23 September 02:00-04:00 and 20:30-22:00).

Comparison of the Two Cases
In both cases, the measured PBL presents several characteristics that resemble a Marine Atmospheric Boundary Layer (MABL): it is a relatively shallow layer with no significant variabilities on the top.Table 1 shows the average and standard deviation of PBL BSC_PollyXT , PBL BSC_Halo , PBL TKE_Halo , and PBL ECMWF .During the period 10:00-14:00 UTC (that corresponds to 9:00-13:00 local time), the standard deviation for both cases varies between 21.5 and 86.4 m, indicating low variability that is connected with the limited daytime evolution.The study on 23 September presented lower PBL retrievals and also smaller uncertainties, compared to 12 September during that time period.This result is a combination of the dust amount and the dust layer level.On 23 September, the layer was thick and infiltrated in lower levels (approx. 1 km), reducing the amount of solar radiation [15] that reaches the surface and also capping the top of the PBL (Appendix A).Moreover, the lower troposphere is less turbulent than on 12 September and is, therefore, more stable.

Conclusions
Our PBL results using WCT and TM are in good agreement with the PBL derived from radiosonde profiles, indicating trustworthy references for the detection of layering.BSC and WVMR are proportional to the aerosol concentration and constitute a possible indicator for PBL detection.However, when an elevated aerosol layer approaches the surface, it is likely that WCT results in big variabilities for the detection of layering, since this elevated layer is captured instead of PBL (23 September 02:00-04:00 and 20:30-22:00).
Mindelo is an area directly influenced by the continuous ocean-atmosphere exchange of heat, moisture, and momentum.Thus, the measured PBL presents MABL characteristics.No sharp daytime evolution or big vertical variabilities are observed on our relatively shallow PBL retrievals.
The presence of dust layers directly affects the formation of PBL.When a thick dust layer hovers over Mindelo (23 September), less short-wave solar radiation reaches the surface.As a result, a shallower PBL is formed, compared to a day with a lighter aerosol load (12 September).
A possible future study is to perform statistical analysis for the retrieval of PBL heights on the entire dataset of the ASKOS campaign for the years 2021 and 2022.Classification of the results according to the meteorological conditions and aerosol layers' presence is also envisaged.
Mindelo is an area directly influenced by the continuous ocean-atmosphere exchange of heat, moisture, and momentum.Thus, the measured PBL presents MABL characteristics.No sharp daytime evolution or big vertical variabilities are observed on our relatively shallow PBL retrievals.
The presence of dust layers directly affects the formation of PBL.When a thick dust layer hovers over Mindelo (23 September), less short-wave solar radiation reaches the surface.As a result, a shallower PBL is formed, compared to a day with a lighter aerosol load (12 September).
A possible future study is to perform statistical analysis for the retrieval of PBL heights on the entire dataset of the ASKOS campaign for the years 2021 and 2022.Classification of the results according to the meteorological conditions and aerosol layers' presence is also envisaged.
AOD measurements for 23 September 08:00-16:00 UTC are not available, but taking into consideration the VLDR from PollyXT, one could say that it was a day with a heavy aerosol load.
In Figure A2, the global horizontal irradiance of shortwave radiation is presented.The overall solar radiation that reached the surface on 12 September was more relative to 23 September.AOD measurements for 23 September 08:00-16:00 UTC are not available, but taking into consideration the VLDR from PollyXT, one could say that it was a day with a heavy aerosol load.
In Figure A2, the global horizontal irradiance of shortwave radiation is presented.The overall solar radiation that reached the surface on 12 September was more relative to 23 September.

Figure 1 .
Figure 1.Conceptual diagram of the application of the (a) Wavelet Covariance Transform method (WCT): wavelet function (red line) is derived using a signal (black line), and a local maximum is identified (purple line); (b) Gradient Method (GM): the gradient (red line) of a vertical profile (black line) is calculated, and a sharp change is located (purple line); (c) Threshold Method (TM): while the signal is lower than a threshold value, it is considered to be within PBL.

Figure 1 .
Figure 1.Conceptual diagram of the application of the (a) Wavelet Covariance Transform method (WCT): wavelet function (red line) is derived using a signal (black line), and a local maximum is identified (purple line); (b) Gradient Method (GM): the gradient (red line) of a vertical profile (black line) is calculated, and a sharp change is located (purple line); (c) Threshold Method (TM): while the signal is lower than a threshold value, it is considered to be within PBL.

Figure 2 .
Figure 2. Boundary layer height observed on 12 September 2022: PBLWVMR_PollyXT (blue triangles) is retrieved with WCT on WVMR product of PollyXT Lidar; PBLBSC_PollyXT (purple circles) is retrieved with WCT on 1064 nm BSC product of PollyXT Lidar; PBLBSC_Halo (black circles) is retrieved with WCT on BSC product of Halo Lidar; PBLTKE_Halo (maroon diamond) is retrieved with TM on TKE product of Halo Lidar; and PBLECMWF (green square) is the output of ECMWF model.Orange shading corresponds to Volume Depolarization Ratio (VLDR) at 532 nm channel of PollyXT Lidar, and blue shading corresponds to TKE dissipation rate measured with Halo Lidar.

Figure 2 .
Figure 2. Boundary layer height observed on 12 September 2022: PBL WVMR_PollyXT (blue triangles) is retrieved with WCT on WVMR product of PollyXT Lidar; PBL BSC_PollyXT (purple circles) is retrieved with WCT on 1064 nm BSC product of PollyXT Lidar; PBL BSC_Halo (black circles) is retrieved with WCT on BSC product of Halo Lidar; PBL TKE_Halo (maroon diamond) is retrieved with TM on TKE product of Halo Lidar; and PBL ECMWF (green square) is the output of ECMWF model.Orange shading corresponds to Volume Depolarization Ratio (VLDR) at 532 nm channel of PollyXT Lidar, and blue shading corresponds to TKE dissipation rate measured with Halo Lidar.

Figure A2 .
Figure A2.Global horizontal irradiance of shortwave radiation measured with a pyranometer for 12 September 2022 (blue line) and 23 September 2022 (orange line).

Table 1 .
Average and standard deviation of PBL height retrievals with each product and method.

Table 1 .
Average and standard deviation of PBL height retrievals with each product and method.