Atmos. Meas. Tech., 12, 2595–2610, 2019
https://doi.org/10.5194/amt-12-2595-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
Planetary boundary layer height by means of lidar and numerical
simulations over New Delhi, India
Konstantina Nakoudi
1,2,3
, Elina Giannakaki
1,4
, Aggeliki Dandou
1
, Maria Tombrou
1
, and Mika Komppula
4
1
Department of Environmental Physics and Meteorology, Faculty of Physics, University of Athens, Athens, Greece
2
Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany
3
Institute of Physics and Astronomy, University of Potsdam, Potsdam, Germany
4
Finnish Meteorological Institute, Kuopio, Finland
Correspondence: Konstantina Nakoudi (knakoudi@phys.uoa.gr)
Received: 30 September 2018 – Discussion started: 20 November 2018
Revised: 28 March 2019 – Accepted: 30 March 2019 – Published: 6 May 2019
Abstract. In this work, the height of the planetary boundary
layer (PBLH) is investigated over Gwal Pahari (Gual Pahari),
New Delhi, for almost a year. To this end, ground-based mea-
surements from a multiwavelength Raman lidar were used.
The modified wavelet covariance transform (WCT) method
was utilized for PBLH retrievals. Results were compared
to data from Cloud-Aerosol Lidar and Infrared Pathfinder
Satellite Observation (CALIPSO) and the Weather Research
and Forecasting (WRF) model. In order to examine the dif-
ficulties of PBLH detection from lidar, we analyzed three
cases of PBLH diurnal evolution under different meteorolog-
ical and aerosol load conditions. In the presence of multiple
aerosol layers, the employed algorithm exhibited high effi-
ciency (r = 0.9) in the attribution of PBLH, whereas weak
aerosol gradients induced high variability in the PBLH. A
sensitivity analysis corroborated the stability of the utilized
methodology. The comparison with CALIPSO observations
yielded satisfying results (r = 0.8), with CALIPSO slightly
overestimating the PBLH. Due to the relatively warmer and
drier winter and, correspondingly, colder and rainier pre-
monsoon season, the seasonal PBLH cycle during the mea-
surement period was slightly weaker than the cycle expected
from long-term climate records.
1 Introduction
The planetary boundary layer (PBL) is the lowermost portion
of the troposphere, which experiences a diurnal cycle of tem-
perature, humidity, wind and pollution variations. PBL is a
key component of the atmosphere and of the climate system,
as it fundamentally affects cloud processes, as well as land
and ocean surface fluxes (Oke, 1988; Stull, 1988; Garratt,
1992). The PBL height (PBLH) is the most adequate param-
eter to represent the PBL. Therefore, it is usually required in
numerous applications, for instance in pollution-dispersion
modeling, where the upper boundary of the turbulent layer
acts as an impenetrable lid for the pollutants emitted at the
surface. The PBLH also appears as a mixing-scale height
in turbulence closure schemes within climate and weather
prediction models (Zilitinkevich and Baklanov, 2001). As
air pollution becomes more severe due to economic devel-
opment, particularly in developing countries (Wang et al.,
2009), high temporal and vertical resolution observations
of the PBLH are essential for weather and air-quality pre-
diction and research. Moreover, the PBLH is related to the
warming rate caused by enhanced greenhouse gas emissions
(Pielke et al., 2007). Several methods have been proposed to
estimate the PBLH, utilizing vertically resolved thermody-
namic variables, turbulence-related parameters and concen-
trations of tracers (Seibert et al., 2000; Emeis et al., 2004).
Different methods for the determination of the PBLH from
radiosondes have been compared and the associated uncer-
tainties have been estimated (Seidel et al., 2010; Wang and
Wang, 2014). Restrictions of radiosondes refer to the coarse
vertical resolution of standard meteorological data with re-
spect to boundary layer studies as well as the smoothing
due to the sensor lag constant bounded by the high ascent
rate of the radiosonde (Seibert et al., 2000). Remote-sensing
systems such as aerosol lidar (Boers and Eloranta, 1986;
Published by Copernicus Publications on behalf of the European Geosciences Union.
2596 K. Nakoudi et al.: Planetary boundary layer height over New Delhi, India
Davis et al., 2000; Lammert et al., 2006; Lange et al., 2014),
microwave radiometer (Cimini et al., 2013), wind-profiling
radar (Cohn and Angevine, 2000) and Doppler wind lidar
(de Arruda Moreira et al., 2018) are suitable for long-term
measurements of various atmospheric quantities with high
temporal resolution and can be used either independently
or synergistically to retrieve the PBLH. Space-borne lidar
systems provide the advantage of spatial coverage, although
for studies focusing on a particular area of interest, mea-
surements are constrained by the overpass frequency (Jor-
dan et al., 2010; McGrath-Sprangel and Denning, 2012; Lev-
entidou et al., 2013). Ceilometers are simple backscatter li-
dars which entail less operational cost. However, exploita-
tion of their full potential is on an early stage with limited
ceilometer-related studies (Münkel, 2007; Binietoglou et al.,
2011; Wiegner et al., 2014). Ceilometers have a high po-
tential to contribute to the PBLH climatology, within cer-
tain limits, but detailed investigation of open issues is still
needed, for example, into the treatment of incomplete over-
lap. Additionally, no adjustments can be typically made by
the user, contrary to the modified wavelet covariance trans-
form (WCT) algorithm. Hence, improvements on layer de-
tection algorithms are urgently needed to fully exploit the
potential of ceilometers. In elastic and Raman lidar systems,
the atmospheric aerosols are used as tracers and the PBLH
is indicated by a gradient in the range-corrected lidar signal
(Menut et al., 1999; Brooks, 2003; Amiridis et al., 2007; Mo-
rille et al., 2007; Baars et al., 2008; Engelmann et al., 2008;
Groß et al., 2011; Tsaknakis et al., 2011; Haeffelin et al.,
2012; Scarino et al., 2014; Summa et al., 2013; Korhonen
et al., 2014; Lange et al., 2014; Bravo-Aranda et al., 2016).
Weather and climate prediction models could alternatively be
used to determine the PBLH, especially for strong horizon-
tal inhomogeneity. However, inconsistencies in the definition
of the PBLH among the existing meteorological models also
result in significant differences in its calculation (Tombrou et
al., 2007).
New Delhi is one of the most densely populated cities
and the fifth most populous city in the world according to
United Nations population estimates and projections of ma-
jor urban agglomerations (https://esa.un.org/unpd/wup/, last
access: 11 April 2019). It is surrounded by the Thar Desert
to the west and the western Indo-Gangetic Plain to the north.
Particulate air pollution in this area is assumed to originate
from fossil fuel and biomass burning besides natural sources
such as desert dust (Hegde et al., 2007; Ramanathan et al.,
2007). The identification of the layer height within which
pollutants are trapped is particularly important in this pol-
luted area, since the largest and most persistent pollution
haze covers an area of about 10 million km
2
over southern
Asia (Nakajima et al., 2007; Ramanathan et al., 2007). Thus,
vertically resolved observations are indispensable to reveal
information regarding local air quality, climate change and
human-health-related issues.
Despite the importance of the area under investigation,
only a few ground-based measurements of aerosol vertical
profiles have been carried out, with most of the available data
accessed during short field campaigns (Lelieveld et al., 2001;
Nakajima et al., 2007; Ramanathan et al., 2007). In this study,
we investigate PBLH characteristics over New Delhi, In-
dia, based on 1-year-long ground-based lidar measurements.
The measurements were carried out from March 2008 to
March 2009 in the framework of the EUCAARI (European
Integrated project on Aerosol Cloud Climate and Air Qual-
ity Interactions) project (Kulmala et al., 2011). The aim of
this study is twofold: (1) to assess the efficiency and stability
of the modified WCT technique in retrieving the PBLH and
(2) to compare the PBLH derived from ground-based lidar to
independent data sources.
2 Measurement site
The lidar measurement site was located at Gwal Pahari
(28.43
◦
N, 77.15
◦
E, 243 m a.s.l.), which is situated in the
Gurugram (Gurgaon) district of Haryana state, about 20 km
south of New Delhi, India (Hyvärinen et al., 2010; Komp-
pula et al., 2012). The surroundings of the station represent a
semi-urban environment with agricultural test fields and light
vegetation. There were no major pollution sources, except
for the road between Gurugram and Faridabad about 0.5 km
to the southwest of the station, while only electric-powered
vehicles were allowed at the station area. Anthropogenic
sources in the greater region comprised traffic, city emissions
and power production (Reddy and Venkataraman, 2002a, b).
Meteorological parameters were measured at the meteoro-
logical station of Safdarjung Airport (28.58
◦
N, 77.21
◦
E,
211 m a.s.l.), New Delhi, which is located 18 km NE of Gwal
Pahari and was the closest climatological site to the lidar
measurement site.
During the measurement period, sunrise time varied be-
tween 05:45 and 07:15 LST, while sunset appeared between
18:15 and 19:15 LST. Solar noon appeared between 12:00
and 12:30 LST. Local time at New Delhi corresponds to
UTC+5.5 h. From now on, in this paper, UTC will be
adopted, to facilitate the comparison between lidar measure-
ments and numerical simulations.
Temperature and precipitation patterns can potentially re-
flect the state of sensible and latent heat fluxes within the
PBL as well as the exchange of moisture and momentum
with the Earth’s surface. Thus, climatologies of meteorologi-
cal parameters can be considered a valuable tool for assessing
the representativeness of the PBLH seasonal cycle with re-
spect to long-term measurements. Such a comparison is per-
formed in Sect. 4.3.2 based on the 30-year anomalies of max-
imum temperature and accumulated precipitation (Fig. 1).
Atmos. Meas. Tech., 12, 2595–2610, 2019 www.atmos-meas-tech.net/12/2595/2019/
K. Nakoudi et al.: Planetary boundary layer height over New Delhi, India 2597
3 Methodology and instrumentation
3.1 Ground-based lidar measurements
3.1.1 FMI–Polly
XT
lidar system
The measurements were conducted with a six-channel Ra-
man lidar called FMI–Polly
XT
(Finnish Meteorological In-
stitute – Portable Lidar sYstem eXTedend). The lidar system
was entirely remotely controlled via an internet connection,
with all the measurements, data transfer and built-in device
regulation being performed automatically. The instrument
was equipped with an uninterruptible power supply (UPS)
and an air conditioning system (A/C) to allow for safe and
smooth continuous measurements. A rain sensor was also
connected to the roof cover in order to assure a proper shut-
down of the instrument during rain.
FMI–Polly
XT
used a Continuum Inline III-type laser. The
pulse rate of the laser was 20 Hz and it delivered energies of
180, 110 and 60 mJ simultaneously (with external second and
third harmonic generators) at three different wavelengths,
i.e., 1064, 532 and 355 nm, respectively. A beam expander
was used so as to enlarge the beam from approximately 6 to
45 mm. The remaining beam divergence after expansion was
less than 0.2 mrad. The backscattered light was collected by
a Newtonian telescope, which had a main mirror with a di-
ameter of 30 cm and a field of view of 1 mrad. The output
of the instrument included vertical profiles of the particle
backscatter coefficient at three wavelengths, i.e., 355, 532
and 1064 nm (retrieved with the Klett method; Klett, 1981,
and Klett, 1985), extinction coefficient at 355 and 532 nm
(retrieved with the Raman method – Ansmann et al., 1990,
1992 – by using the Raman shifted lines of N
2
at 387 and
607 nm) and linear particle depolarization ratio at 355 nm.
The system vertical resolution was 30 m and the vertical
range covered the whole troposphere under cloudless con-
ditions. This is sufficient for PBL studies considering the
heights needed in this work. Engelmann et al. (2016) reports
a maximum vertical range of 40 km, which depends on the
capabilities (height bins) of the data acquisition. The FMI–
Polly
XT
lidar system is described in more detail in Althausen
et al. (2009) and Engelmann et al. (2016).
The incomplete overlap between the laser beam and the
receiver field of view, L–R (laser–receiver), restricted the ob-
servational detection range to heights above 200–300 m. This
was partly counterbalanced by the overlap correction func-
tion. In this study, overlap corrections were performed at
532 nm following the methodology proposed by Wandinger
and Ansmann et al. (2002). During the measurement cam-
paign, the L–R overlap was completed at 550–850 m, with
the estimation of the full overlap height performed five times,
since changes in the system could have affected the align-
ment between the laser beam and the receiving telescope op-
tical axes.
At nighttime, the configuration of FMI–Polly
XT
allowed
the determination of the residual layer height (RLH). The
study by Wang et al. (2016), which was performed at a station
of similar latitude, Wuhan, China, revealed that the RLH lies
mostly in the range 0.5–1.3 km, following a seasonal varia-
tion. Hence, for most of our nighttime cases, we considered
that the lidar system detected the RLH, which contained the
aerosol of the previously mixed layer. In particular, if a layer
top more than 500 m was detected between sunset and sun-
rise, it was associated with the RLH.
3.1.2 PBLH detection technique
The PBLH was derived from the 15 min averaged lidar
backscatter signals at 1064 nm using the WCT method
(Brooks, 2003) with modifications introduced by Baars et
al. (2008). The algorithm of the WCT method was applied to
6 h datasets. An overview of the lidar range-corrected signal
was made available by TROPOS (Leibniz Institute for Tro-
pospheric Research) and can be accessed at http://polly.rsd.
tropos.de/?p=lidarzeit&Ort=21?, last access: 11 April 2019.
The WCT method made use of the assumption that the PBL
contains much more aerosol load compared to the free tro-
posphere and, thus, a strong backscatter signal decrease can
be considered to be the PBLH. The covariance transform
W
f
(a, b) was based on the convolution of the range-corrected
lidar signal and the related Haar function (Baars et al., 2008).
This method was chosen because it allows larger adjusta-
bility than other techniques, as shown from previous stud-
ies (Baars et al., 2008; Korhonen et al., 2014). For instance,
the gradient technique involves an ambiguity in the choice of
the relevant minimum in the gradient that corresponds to the
PBLH (Lammert and Bösenberg, 2006). A first modification
by Baars et al. (2008) regarded the WCT threshold, which al-
lowed the identification of significant gradients and the corre-
sponding omission of weak gradients. The first height above
ground at which a local maximum of W
f
(a, b) occurred, ex-
ceeding the selected signal decrease threshold, was defined
as the PBLH. A second modification introduced by Baars et
al. (2008) was related to strong gradients in the lower parts
of the PBL (30–870 m) and the ability to exclude these parts
from the lidar data evaluation. In this work, the applicabil-
ity of the WCT technique under different meteorological and
aerosol load conditions is discussed (Sect. 4.1) in the con-
text of three case studies, and the stability of the WCT algo-
rithm is assessed as well (Appendix A). Additional cases,
in which the importance of a proper threshold and cutoff
zone are discussed, can be found in Nakoudi et al. (2018).
The WCT method also allows for the detection of clouds
by means of a negative threshold. Baars et al. (2008) found
that the cloud screening works well for a threshold of −0.1.
The cloud base is given 1 height bin below the altitude at
which W
f
(a, b) is lower than the chosen threshold value. The
WCT method has also been applied for the detection of cirrus
cloud base height over different geographical regions (Dion-
www.atmos-meas-tech.net/12/2595/2019/ Atmos. Meas. Tech., 12, 2595–2610, 2019
2598 K. Nakoudi et al.: Planetary boundary layer height over New Delhi, India
Figure 1. Maximum temperature and cumulative precipitation during the measurement campaign (black) and anomalies (blue) at New Delhi
on a monthly basis. Anomalies represent difference between the climatological values and the corresponding values during the measurement
campaign. Climatological values were obtained from World Meteorological Organization (http://worldweather.wmo.int/en/city.html?cityId=
224, last access: 11 April 2019) for the site of Safdarjung Airport.
isi et al., 2013; Voudouri et al., 2018). Uncertainties in the
retrieval of the PBLH mainly originated from the lidar signal
noise, which was lower at nighttime, the systematic error re-
lated to the estimation of the atmospheric molecular number
density from the pressure and temperature profiles as well as
the systematic error for overlap function. Furthermore, errors
were introduced by the operation procedure such as signal
smoothing and averaging by accumulating lidar returns. De-
tailed discussions on the overall relative errors of the Polly
and Polly
XT
lidar-derived aerosol properties can be found in
Baars et al. (2016) and Engelmann et al. (2016).
Daily mean and maximum PBLH corresponds to con-
vective hours (03:00–12:00 UTC). The hourly PBLH was
calculated from the 15 min lidar observations by averag-
ing of the three closest data points of the time considered
(e.g., 12:00 hourly height would be the average of the three
data points between 11:45 and 12:15 UTC). The seasonal cy-
cle study was based on the classification proposed by the
Indian Meteorological Department, i.e., winter (December–
March), pre-monsoon or summer (April–June), monsoon
(July–September) and post-monsoon (October–November)
(Perrino et al., 2011). However, the PBLH seasonal cycle was
examined during the winter, pre-monsoon and monsoon peri-
ods, as no sufficient data coverage was found during the post-
monsoon period (Sect. 3.1.3). The PBLH growth period was
determined following the guidelines of Baars et al. (2008).
More specifically, the PBLH growth period began when the
PBLH started to increase (typically 2–4 h after sunrise) and
was completed when 90 % of the daily maximum PBLH was
reached (typically between 08:00 and 10:30 UTC). Concern-
ing the daily evolution rate, this was determined through the
slope of a linear fit to the hourly PBLH (between the start and
the completion of the growth period). The evolution rate cal-
culation was restricted to cases in which at least four consec-
utive or three nonconsecutive hourly values were available.
Due to these restrictions, the evolution rate was determined
for 44 d.
Figure 2. Data coverage of lidar measurements calculated with re-
spect to total convective hours (from 4 h after sunrise to 1 h before
sunset) during the measurement days of the campaign.
3.1.3 Data coverage
During the 1-year-long measurement campaign, FMI–
Polly
XT
was measuring on 139 d. Due to technical problems
with the laser, the data coverage from September to January
was sparse. Furthermore, precipitation prohibited lidar mea-
surements, since the lidar system had to shut down. Hence,
sufficient data availability was achieved during 72 d. Multi-
ple aerosol layers appeared mainly between March and May,
whereas low clouds were present mostly in the monsoon pe-
riod, and both complicated PBLH detection. Additionally,
some technical issues arose due to photomultiplier supersat-
uration and signal problems. A lack of a significant decrease
in the backscatter profile was observed in only a few cases,
which was the first indication that the modified WCT method
can detect the PBLH efficiently, as long as the signal decrease
threshold was tuned properly. The data coverage is presented
on a monthly basis in Fig. 2. The highest PBLH detection fre-
quency was achieved in February, which can be attributed to
favorable meteorological conditions, with sparse low clouds
and hardly any rainfall events.
Atmos. Meas. Tech., 12, 2595–2610, 2019 www.atmos-meas-tech.net/12/2595/2019/
K. Nakoudi et al.: Planetary boundary layer height over New Delhi, India 2599
3.2 Space-borne lidar observations
Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Obser-
vation (CALIPSO) is an Earth science observation mission
that was launched on 28 April 2006. The vertical resolution
of the CALIOP (Cloud-Aerosol Lidar with Orthogonal Po-
larization) system is 30 m. The CALIPSO Level 2 aerosol
layer product provides a description of the aerosol layers, in-
cluding their top and bottom height, identified by automated
algorithms applied in the Level 1 data. A detailed description
of the aforementioned algorithms can be found in Vaughan
et al. (2004) and Winker (2006). In this study, a CALIOP
version V4-10 dataset was used. Currently, no operational
CALIOP PBL product is available.
More specifically we applied the CALIOP Level 2 Aerosol
Layer Product, which provides information on the base and
top heights of existing aerosol layers, reported at a uniform
5 km horizontal resolution. Leventidou et al. (2013) evalu-
ated the daytime PBLH derived by Level 2 Aerosol Layer
products over Thessaloniki, Greece, for a 5-year period, mak-
ing the assumption that the lowest aerosol layer top can be
considered to be the PBLH. The aforementioned method was
also applied over South Africa, revealing high agreement
with ground based observations (Korhonen et al., 2014). Dur-
ing the measurement campaign, the PBLH was accessed by
the space-borne lidar CALIOP, within overpass distances of
20 and 101 km from Gwal Pahari.
3.3 WRF atmospheric model
The Weather Research and Forecasting (WRF) model, ver-
sion 3.9.1 (Skamarock et al., 2005) was also applied in or-
der to determine the PBLH. The simulation domain was
centered at the lidar station in Gwal Pahari and three do-
mains with respective horizontal resolutions of 18, 6 and
2 km were used, where the two inner domains are two-way
nested to their parent domain. The third innermost domain
covers an area between 75.84–78.46
◦
E and 27.38–29.52
◦
N.
The output was provided every hour. On the vertical axis,
37 full sigma levels resolved the atmosphere up to 50 hPa
(≈ 20 km a.g.l.), with a finer grid spacing near the surface.
In this study, the Yonsei University scheme (YSU) (Hong et
al., 2006), in conjunction with the land surface model Noah
(Chen and Dundhia, 2001), was used for the estimation of
PBL height. In addition, the rapid radiative transfer model
(RRTM) scheme (Mlawer et al., 1997) for longwave radi-
ation and the scheme of Dundhia (1989) for shortwave ra-
diation were applied. A surface-layer scheme based on the
revised MM5 similarity theory (Jiménez et al., 2012) as well
as the Kain and Fritsch (1990, 1993) scheme for cumulus
parameterization were used. For microphysics, the scheme
proposed by Thompson et al. (2008) was considered. Re-
garding land use and soil types, the predefined datasets of
Moderate Resolution Imaging Spectroradiometer (MODIS)
with 21 land use classes were used. The initial and lateral
boundary conditions were derived from the National Cen-
ter for Environmental Prediction (NCEP) operational Global
Fine Analysis (GFS) with 1
◦
× 1
◦
spatial resolution and were
updated every 6 h. Sea surface temperature (SST) was ob-
tained from high resolution real-time global SST (RTG SST
HR), with a spatial resolution of 0.083
◦
× 0.083
◦
, which was
renewed every 24 h.
In the YSU scheme, the PBLH under unstable condi-
tions was determined as the first neutral level based on the
Bulk Richardson number (Ri) calculated between the lowest
model level and the levels above (Hong et al., 2006; Shin
and Hong, 2011). Under stable conditions, the Ri was set to
a constant value of 0.25 over land, enhancing mixing in the
stable boundary layer (Hong and Kim, 2008), whereas it was
a function of the surface Rossby number over the oceans, fol-
lowing the study of Vickers and Mahrt (2003). More specifi-
cally, the revised stable boundary layer (SBL) scheme (Hong,
2010) computed the exchange coefficients with a parabolic
function with height, as in the mixed layer, in which the top
of the SBL was determined by the Ri (Vickers and Mahrt,
2004). This led to a gradual and not abrupt collapse of the
mixed layer after the sunset due to the residual superadia-
batic layer near the surface even in the presence of negative
surface buoyancy flux. Within the frame of three case studies,
the default PBLH simulated from WRF was used to justify
the lidar PBLH.
4 Results and discussion
4.1 Applicability of the WCT method: case studies
It was found that in some cases the presence of multiple
aerosol layers and low clouds can pose difficulties in PBLH
detection (Sect. 3.1.3). However, these difficulties can be
dealt with the use of proper WCT threshold and cutoff val-
ues (Sect. 3.1.2). Three case studies of PBLH daily evolu-
tion were analyzed and evaluation with ancillary data sources
was performed so as to investigate capabilities and limita-
tions. First, the evolution of the PBLH under cloudless con-
ditions is discussed for 12 February 2009. Subsequently, a
2-day case with a multiple aerosol layer structure is pre-
sented for 1–2 March 2009. Finally, the diurnal development
of the PBLH is investigated in the presence of low clouds for
29 June 2008.
4.1.1 Cloud-free case: 12 February 2009
The PBLH during 12 February 2009 was characterized by
an almost constant daily growth rate (133 mh
−1
between
06:00 UTC and 10:00 UTC) with a maximum height of
950 m (Fig. 3). No aerosol layers were observed in the free
troposphere. Although gradients (yellow and red color) of
aerosol content appeared inside the PBL (06:00–12:00 UTC),
the default signal decrease threshold (0.05) was efficient.
However, later (12:00–18:00 UTC), in order to avoid strong
www.atmos-meas-tech.net/12/2595/2019/ Atmos. Meas. Tech., 12, 2595–2610, 2019
