Introduction
Relative humidity (RH) is a crucial parameter for atmospheric research, as
it represents the current state of water vapor and ambient air related to
saturation. Changes in RH may influence atmospheric optical properties such
as visibility, which is often reduced due to RH variations in the atmosphere
(Tang et al., 1981). Moreover, increased RH in the atmosphere may influence
the physical properties of aerosols, causing condensation onto their
surface, which subsequently triggers their hygroscopic growth. Not only does
this growth affect the direct scattering of radiation (Hanel and Zankl, 1979;
Hegg et al., 1996; Zieger et al., 2013), but also the process of cloud
condensation nuclei (CCN) formation (Charlson et al., 1992; Petters and
Kreidenweis, 2007; Wex et al., 2008; Mochida, 2014). Previous studies have
found that even minor changes in RH may affect processes such as cloud
formation or precipitation (Kulmala et al., 1993; Tomkins, 2003;
Sherwood et al., 2010; Altaratz et al., 2013). The interest in the role of
RH in the modification of aerosol, precipitation and cloud microphysics,
including CCN formation, has recently increased mainly due to the crucial
role of aerosol–cloud interactions in climate change (Fan et al., 2007;
Veselovskii et al., 2009; Zieger et al., 2013; Granadoz-Munoz et al., 2015;
López and Ávila, 2016).
In addition, RH measurements are frequently used for evaluation studies
aiming to predict the formation of clouds (Heerwaarden and Arellano, 2008)
and aircraft contrails (Radel and Shine, 2007). No less important are the
significant uncertainties in the estimation of global climate change
parameters using climate modeling (Schneider et al., 2010). Usually these
uncertainties are associated with RH variations, since water vapor acts as a
global constraint in the climate system (Sherwood et al., 2010). Despite
advancements in satellite remote sensing of water vapor, continuous
quantification of RH in the low troposphere remains challenging. Vertical
resolution of spaceborne measurements of water vapor and temperature
constrains the accuracy of RH retrieval close to the ground (with spatial
resolution of ∼1–2 km) (Wulfmeyer et al., 2015). RH
observations are based on water vapor and temperature measurements which are
together frequently referred to as thermodynamic atmospheric profiling. A
comprehensive description of the modern techniques for thermodynamic
profiling by different instruments is given by Wulfmeyer et al. (2015),
who have outlined the advantages, disadvantages and
uncertainties of each instrument. Here, we present a brief description of
different ground-based techniques for RH vertical profiling. One of the most
frequently used instruments for RH vertical profiling are radiosondes.
Radiosoundings provide vertical profiles of RH with spatial resolution of a
few meters, and relatively high accuracy (±4–5 %, depending on the
time of the day) (Miloshevich et al., 2009). More automated RH vertical
profiling is usually performed using passive and active remote sensing
sensors which are able to measure both water vapor and temperature vertical
distribution. The Atmospheric Emitted Radiance Interferometer (AERI) is an
example of a passive sensor that can be used for water vapor and temperature
quantification, based on atmospheric radiance measurements at the 15 µm
CO2 band. Profiles of RH are provided using a combination of AERI water
vapor and temperature datasets from the ground up to 3000 m, with a temporal
resolution of 10 min (Feltz et al., 1998; Knuteson et al., 2004). The AERI
system is, however, limited by coarse spatial resolution and often AERI cannot
detect sharp and strong inversion layers (Mattis et al., 2002). Microwave
radiometric measurements of temperature and humidity can provide water vapor
and temperature vertical profiles as well. A microwave radiometer performs
multifrequency measurements of brightness temperatures at high temporal
resolution (∼1 s) and high accuracy from the surface (0.6 K)
up to the middle troposphere (1.5 to 2 K) (Hogg et al., 1983; Ware et al.,
2003). Nevertheless, radiometer measurements suffer from coarse spatial
resolution and substantial uncertainties in the retrieval of humidity and
temperature at heights greater than 4000 m, where only 5 % of the
independent information originates from radiometer measurements themselves (Rose
et al., 2005). Active remote sensing instruments such as atmospheric lidar
instruments have the ability to obtain high-resolution measurements of RH. To this end,
two major techniques can be employed: the differential absorption lidar
(DIAL) and the Raman lidar techniques. Each technique has its unique
advantages and limitations. The DIAL water vapor profiling is based on the
ratio of two elastic backscatter signals at two adjacent wavelengths and is
affected by the temperature dependence of the water vapor molecular
absorption, which is greater than 1–2 % (Wulfmeyer et al., 2015).
Additionally, the presence of strong aerosol gradients may result in high
systematic uncertainties that exceed the requirements of most desired
applications (Theopold and Bosenberg, 1993). The Raman lidar technique for
RH vertical profiling is based on the vibrational Raman scattering, which
can be combined with the rotational Raman scattering to provide also the temperature
vertical profiles (Arshinov et al., 1983). When a laser beam is emitted to
the atmosphere at 355 nm, the use of a Raman lidar at 387 nm (Raman shifting
by atmospheric N2) and at 407 nm (Raman shifting by water vapor)
enables the humidity-dependent parameter to be derived, which is subsequently
normalized to the mixing ratio of water vapor. The most important
constraints for current lidar instruments make most of them not applicable for water
vapor measurements during daytime (due to high atmospheric background
levels) and in the lowest several hundred meters of the troposphere (due to
geometrical optics limitations).
Despite the relatively high performance of remote sensing instruments with
regard to RH vertical profiling, we still suffer from a lack of a consistent
robust method for continuous RH vertical profiling. One step forward in RH
vertical profiling without the technological improvement of sensors is to
use synergistic approaches, as proposed by Turner et al. (2000). This study
presented the synergistic retrieval of RH based on a Raman lidar-retrieved
water vapor mixing ratio and temperature profiles from the AERI instrument.
Such synergy allowed the profiling of RH with high temporal
resolution to be performed. Another example of a synergetic approach towards RH vertical
profiling has been presented by Nagel et al. (2001), where the authors
approached RH profiling by combining lidar-derived humidity and temperature
measurements from radiosoundings (launched every 10 min). Moreover, Wang
et al. (2011) demonstrated that density and the water vapor mixing ratio can be
combined with temperature observations from a collocated rotational Raman
lidar to provide RH vertical profiles. They showed relatively good agreement
between lidar-retrieved and radiosonde observations, with a bias of up to
10 % in the lowest 2000 m of the troposphere. A step forward in
synergistic approaches towards accurate RH vertical profiling has been
performed by Navas-Guzman et al. (2014). They demonstrated the method for RH
vertical profiling based on the combination of Raman lidar humidity and
temperature measurements from a collocated microwave radiometer, together
with air density profiles taken from a standard atmospheric model scaled to
near-ground density measurements (COESA, 1976). This combined retrieval
method resulted in an increased accuracy for continuous lidar-derived RH
measurements in comparison with other remote sensing techniques. The
resultant mean absolute deviation in RH compared to radiosonde data based on
lidar–radiometer retrievals varied from 6 to 7 % from 1000 to 5000 m,
respectively. Recently, Barrera-Verdejo et al. (2016) once again tested
lidar–radiometer combination perspectives for water vapor studies. They
developed a new approach for lidar–radiometer synergy for absolute humidity
(AH) vertical profiling using an optimal estimation method. They combined
multifrequency brightness temperature observations from a microwave
radiometer and mixing ratio observations from a Raman lidar to retrieve
high-resolution profiles of AH. Their results proved that the combination of
lidar and radiometer data can reduce the theoretical error by a factor of 2
in the lower troposphere when water vapor information is retrieved. All of
these aforementioned methods and approaches have shown the perspectives of
synergistic approaches for RH vertical profiling using collocated remote
sensing instruments. More recently, Schutgens et al. (2017) presented
promising results from the combination of spatially collocated observations
and model simulations, pointing out that high-resolution model simulations
can serve as a robust data source.
Based on the recommendations of Wulfmeyer et al. (2015) and Schutgens et al. (2017),
in our study we used two synergistic approaches for RH vertical
profiling. The first approach is based on the synergy of lidar and
radiometer instruments, while the second method is based on lidar and
numerical simulations from the Weather Research and Forecasting (WRF) model
output. Our approaches use a combination of datasets, including water vapor
mixing ratio from a Raman lidar, temperature profiles from radiometer,
high-resolution simulations from the WRF model and air density profiles from
the US Standard Atmosphere (1976) (COESA, 1976). Datasets were acquired
during the HygrA-CD (Hygroscopic Aerosols to Cloud Droplets) campaign
conducted in Athens, Greece, from May to June 2014 (Papayannis et al., 2017).
The main scope of this paper is to show the effectiveness of the two
synergistic approaches in comparison with single-instrument observations of
RH from microwave radiometer and RH single simulations from WRF. As a second
objective, we determine the effectiveness of these approaches according to
crucial requirements applied for thermodynamic profiling techniques
formulated and generalized by Wulfmeyer et al. (2015). These requirements
can be applied for RH vertical profiling and include several points, among
which the accuracy, spatial resolution and the minimum–maximum range of
measurements.
Instruments and models
Experimental site
The datasets used in this work were collected during the HygrA-CD
experimental campaign organized in the greater Athens area in the period 15 May–22 June 2014.
The aim of the campaign was to bring together various
instruments for atmospheric measurements in order to improve our current
understanding of the impact of aerosols on clouds near the top of the
planetary boundary layer (PBL). During the campaign period a variety of
remote sensing and in situ instruments provided an important record of data
on aerosols, clouds and local meteorology conditions. Among the five
measurement sites involved in the HygrA-CD campaign, most of the data were
obtained at the National Technical University of Athens (NTUA)
(37.97∘ N, 23.79∘ E, 212 m a.s.l.) and the National
Center of Scientific Research Demokritos (DEM) (37.99∘ N,
23.82∘ E, 275 m a.s.l.) (Papayannis et al., 2017).
Multiwavelength lidar
The aerosol and ozone lidar system (EOLE) multiwavelength Raman lidar system
located at the campus of NTUA emitted pulses at three wavelengths: 355, 532
and 1064 nm, with energies per pulse of 240, 260 and 300 mJ, respectively,
with a 10 Hz repetition rate. A receiving Cassegrainian telescope (primary
mirror of 300 mm diameter and 600 mm focal length) was used to
simultaneously receive the elastic backscattered lidar signals and the Raman
ones (387, 407, 607 nm). The full overlap of the system is achieved at
∼300 m from the lidar system (Kokkalis et al., 2012). Since
the Raman signals are relatively weak, the Raman lidar measurements were
performed only at nighttime under clear-sky conditions. The Raman-derived
vertical profiles of the water vapor mixing ratio were calculated for 26 days
of the campaign with different temporal scales (2 min, 1 h and 2 h),
which were selected depending on the various instruments' intercomparison:
2 min averaged data were used for comparisons with integrated radiometric
values, and 2 h averaged data for the intercomparison with the radiosonde
data, depending on the signal-to-noise ratio (SNR) of the lidar signals, as
outlined as the minimum required resolution for effective thermodynamic
profiling in the review from Wulfmeyer et al. (2015). The signal detection
at the vibration Raman channels of 387 and 407 nm gives the possibility
to retrieve the water vapor mixing ratio profiles defined as the ratio μ of the mass of water vapor to the mass of dry air (gkg-1)
(Goldsmith et al., 1998) as extensively used in the last 2 decades for both
daytime and nighttime measurements (Whiteman et al., 2006, 2010; Adam et al.,
2007; Leblanc et al., 2011).
μ=CPWV(R)PN2(R)exp-∫0RαN2(dr)drexp-∫0RαWV(dr)dr,
where PWV is the detected lidar signal at the water vapor channel,
PN2 is the detected Raman signal at the nitrogen channel and C is the
calibration constant (see Sect. 3.1). The exponential part of Eq. (1) takes
into account the ratio of the atmospheric transmission at 387 (αN2) and 407 nm (αWV) (Weitkamp, 2005). This difference in
transmission is mainly contributed by Rayleigh scattering and can be
calculated using temperature and pressure profiles taken from the US
Standard Atmosphere 1976 (COESA, 1976) and range-independent Rayleigh
scattering cross sections at appropriate wavelengths (Bucholtz, 1995). The
standard profiles of pressure are assumed to be accurate for our purposes
since the uncertainties introduced by their use are lower than 5 %. The
difference between profiles of pressure obtained from radiosoundings and
standard pressure profiles did not exceed 0.055 kgm-3 (4.5 %). A
more detailed description of the calibration procedure and the analysis of
water vapor observations for the EOLE system can be found in Landulfo et al. (2009) and Mamouri et al. (2008), respectively, as well in Sect. 3.1 below.
Microwave radiometer
The HATPRO-G2 microwave radiometer consists of several components: two
receiver units (22.24–31.4 and 51.3–59 GHz) with the relevant receiving
optics, the ambient load, the internal scanning mechanism, the electronics
and the data acquisition system (Rose et al., 2005). The microwave
radiometer used in this study is manufactured by Radiometer Physics GmbH
and belongs to the National Institute of R&D in Optoelectronics
(Bucharest, Romania). The HATPRO-G2, installed at NTUA, was calibrated
before the observation campaign according to the procedure of radiometer
absolute calibration using liquid nitrogen (Liljegren, 2002). The
atmospheric radiation is measured at seven channels located in the K band,
along the wing of the water vapor absorption line (22.35 GHz), and seven
channels located in the V band (Westwater, 1965), along the oxygen
absorption complex (center is around 60 GHz) (Westwater et al., 2005). The
vertical profiles of water vapor and temperature are inverted from observed
brightness temperatures by using statistical regression algorithms, based on
long-term datasets of collocated radiosoundings. The radiometer measurements
used in this work have a temporal resolution of 15 s and a height-dependent
vertical resolution: 200 m from 0 to 2000 m, 400 m from 2000 to 5000 m and
>500 m for heights above 5000 m (Löhnert et al., 2004; Rose et
al., 2005; Mashwitz et al., 2013). The radiometric measurements (the data
provided are integrated water vapor, IWV and vertical profiles of AH, RH
and temperature) were continuously performed from 15 May to 20 June 2014,
except on 12 June where they were only available from 00:00 to 05:55 UTC and
from 20:38 to 23:59 UTC, due to a technical shutdown. Radiometer-related
random errors are analyzed based on previous studies which had determined
that the error related to the systematic bias for AH retrievals using the
regression method below 4000 m equals 0.8 gm-3. The random error of
0.5 K is taken into account in the PBL (up to 1000 m) and
1.7 K between 1000 and 6000 m (Güldner, 2013; Crewell et al., 2001;
Liljegren et al., 2005; Löhnert and Maier, 2012).
Sun photometer
The sun photometer is a passive remote sensing instrument that retrieves
columnar atmospheric aerosol properties during daytime while pointing at the
sun. For this work, we used aerosol columnar optical properties retrieved
from a CIMEL CE-318-NEDPS9 sun photometer (Holben et al., 1998), which is a
member site of the NASA AERONET (Aerosol Robotic Network) and was located on
a nearby site around 5 km from NTUA (Papayannis et al., 2017). The sun
photometer performs direct sun and sky measurements of solar radiances at
eight wavelengths (340, 380, 440, 500, 675, 870, 1020 and 1640 nm). Aerosol
optical depth is retrieved from direct sun measurements, while diffuse sun
measurements are used by the inversion algorithms for columnar microphysical
properties (Dubovik and King, 2000; Dubovik et al., 2006). Finally,
measurements at 940 nm are used to estimate the integrated amount of
precipitable water which is used in our work for the purposes of microwave
radiometer validation and calibration.
Radiosoundings
During the campaign 17 high-resolution radiosondes (Vaisala RS92-SGP) were
launched by the Hellenic National Meteorological Service (HNMS) at the
Hellinikon airport (37.88∘ N, 23.73∘ E).
Vertical profiles of temperature were measured with an uncertainty of
0.3–0.4 ∘C, relative humidity with an uncertainty around 4 % and
height uncertainty around 20 m (Nash et al., 2011). Here we use only the
radiosoundings launched at 00:00 UTC, since only nighttime lidar
measurements were considered for water vapor profiling. When high vertical
resolution radiosonde data were not available, we used the radiosounding
data (pressure, temperature, humidity) with sparser vertical resolution of
about 50–700 m (low-resolution radiosoundings) obtained from the
University of Wyoming website
(http://weather.uwyo.edu/upperair/uamap.shtml).
WRF model configuration
The WRF model (Skamarock et al., 2005) is a numerical weather prediction
system designed for both atmospheric research and operational forecasting
needs. Here, we use WRF model version 3.4.1 with the Advanced Research WRF
(ARW) dynamical core. Three one-way nested domains are configured, with the
finest domain on a 1 km × 1 km grid over the greater Athens area. This
high resolution is deemed sufficient to make simulations comparable with
observational measurements in the complex area of Athens. Moisture
parameters were simulated with 39 vertical levels at 50–100 m vertical
spacing and with a 1 h writing period of the results. Daily simulations
were run with a 36 h forecast cycle, including an allowance of 12 h for
model spin-up. More details about the configuration of the WRF model during
the HygrA-CD campaign are described in Banks et al. (2016).
Evaluation of individual retrievals
RH calculation requires profiling of water vapor, temperature and pressure.
In this paper we used Raman lidar, a microwave radiometer and the WRF model to
calculate the RH profiles. The calibration and evaluation of water vapor and
temperature measurements are further presented.
EOLE calibration for water vapor measurements – validation of lidar
calibration results
The lidar calibration is required since the measured quantity cannot be
directly referred to as the water vapor mixing ratio. There are several methods
to calibrate a Raman lidar system for water vapor profiling: by comparing
with radiosonde observations, with collocated instrument observations
(Whiteman et al., 1992; Foth et al., 2015) or by using hybrid normalization
by calibrating lamps (Leblanc and McDermid, 2008). Here we apply the
most commonly used method of the calibration using collocated radiosondes
data within specific altitude ranges depending on the lidar signal-to-noise
ratio for two main reasons. Firstly, calibration using passive remote
sensors may result in incomplete sampling of the water vapor column observed
by lidar. Secondly, the use of calibrated lamps can be challenging and may
lead to the increased amount of unexpected errors that are difficult to
observe and compensate for (Whiteman et al., 2011). To derive the water vapor
mixing ratio (μ), a calibration constant (C) has to be estimated (see
Eq. 2) taking, as well, the ratio of the vibrational Raman lidar signals at
407 (PWV) and 387 nm (PN2):
μ=CPWV(R)PN2(R).
The mean value of C is calculated from the bottom to the top layers of
six
high-resolution radiosoundings, taking into account the altitudes where its
standard deviation is less than 10 % (Table 1). In our case C was
estimated to be equal to 23.65, with a weighted standard deviation of 9.5 % (see
Fig. 1, left panel). The values of the calibration constant with appropriate
deviations for certain altitudes where averaging was performed (calculated
from a series of radiosonde data), are presented in Table 1.
Calibration constant mean values for each case of
high-resolution radiosounding.
Date
Lidar time
Radiosonde
Bottom
Top
Constant mean
σ
(UTC)
launcha
(km)
(km)
(±σ)
(%)
15 May 2014
21:00–22:00
00:00
1.00
2.04
22.52 ± 2.23
9.94
17 May 2014
19:00–20:00
00:00
1.00
4.00
24.51 ± 2.47
10.00
18 May 2014
22:00–23:00
00:00
1.00
2.57
23.41 ± 2.27
9.78
20 May 2014
22:00–23:00
00:00
2.40
4.24
24.00 ± 2.29
9.70
21 May 2014
21:00–22:00
00:00
2.26
3.34
21.58 ± 2.03
9.50
1 Jun 2014
23:00–00:00
00:00
1.00
4.04
25.89 ± 2.27
8.84
a All radiosondes are launched on the following day after the Raman lidar
measurements.
Furthermore, we compared the Raman lidar data with collocated low-resolution
radiosoundings to check the sanity of the estimated C value (the one shown
in Table 1). The low-resolution calibration constant is 24.36 with a weighted
standard deviation of 8.8 % (Fig. 1., right panel) and the difference
between the mean calibration constants is negligible. Relatively low
differences with referenced radiosounding directly approves the validity of
our calibration to meet the requirements (uncertainty should not exceed
10 %) from Leblanc and McDermid (2008). Therefore in this study we used
calibration constant from high resolution radiosounding (C=23.65±2.28).
Variations of the water vapor calibration constant
calculated based on low-resolution (LR) and high-resolution
(HR) radiosoundings. Green bars: calibration constant variations, red solid line: mean
calibration constant, dashed lines: maximum and minimum calibration
constants.
As outlined by Leblanc et al. (2011), the calibration stability of a water
vapor Raman lidar has to be insured by collocated water-vapor measurements.
Therefore, an intercomparison of the IWV between the Raman lidar, sun
photometer and microwave radiometer was performed. First, we compared the
IWV from collocated radiometer and sun photometer measurements. Secondly,
the lidar-derived IWV values (2 min resolution) were compared with the IWV
values retrieved from the radiometer data at the same time with a temporal
difference of less than 30 s. The intercomparison was done in two steps because
the Raman lidar provided the water vapor mixing ratio only during nighttime,
while the sun photometer measured only during daytime. In the first step,
the sun photometer and radiometer datasets were intercompared, while in the
second step the radiometer and lidar datasets were intercompared. The sun photometer
uncertainty in IWV calculation was considered to be 1.2 kgm-2 (Wang,
2008), while that of the microwave radiometer to be 0.8 kgm-2 (Rose et
al., 2005). In our analysis we used in total 36 measurement pairs. The
determination coefficient (R2) was estimated to be equal to 0.89 and only in
one case was noted as inconsistent (Fig. 2, left panel). The average bias
(b) of the radiometer data regarding the sun photometer was estimated to be equal
to 1.02. The intercomparison between lidar and the microwave radiometer was
performed with some additional assumptions as in our study EOLE was not able
to provide reliable data below 750 m, due to geometric optical limitations
of the system (see Sect. 3.2). An upper limit of the integration range of
9000 m has been chosen, such as to have Raman lidar signals with
SNR > 3 in order to derive accurate
integrated water vapor values. The agreement between lidar-derived IWV
radiometer-derived IWV was very high (R2=0.98) as shown in Fig. 2
(right panel). Also the Raman lidar retrievals were not significantly biased
from the microwave radiometer results. In most of the cases the absolute
difference between the IWV values derived from these two instruments was not
higher than 1. We therefore inferred that the Raman lidar calibration can be
considered reliable and the system is suitable to perform accurate water
vapor mixing ratio measurements.
IWV intercomparison between different instruments:
radiometer–sun photometer (a) and radiometer–lidar (b) with the R2 coefficients provided.
We performed another intercomparison to validate both WRF simulations and
instrumental measurements of water vapor. We analyze the absolute difference
(%) between high-resolution radiosonde observations and each measurement
and simulation technique (lidar, microwave radiometer, WRF simulations)
separately. From Fig. 3 we can see that in the lowest tropospheric layer,
lidar demonstrates the poorest agreement with radiosounding measurements due to
incomplete geometrical overlap of the signal. In the lowest 500 m, the
median difference between radiosounding and lidar water vapor measurements
is 151.6 %. Previous studies have shown that the overlap region of the
EOLE system is around 300 m (Kokkalis et al., 2012). However, when water
vapor parameters are retrieved, the overlap issue may affect lidar
measurements between 500 and 1000 m due to different overlap characteristics
of the Raman channels at 387 and 407 nm. So, we investigated the bottom
layer for lidar water vapor measurements for our study by averaging the
radiosounding mixing ratios from 7 days and calculated the standard
deviation from the mean water vapor mixing ratio value. The resultant
standard deviation (0.58) is used to apply the data quality test. We set
the altitude threshold based on 3 standard deviation so that heights where the
mean difference between lidar and radiosounding data exceeded 1.74 are not
used. Based on this analysis we determined the bottom height of 750 m, above
which the lidar measurements of the mixing ratio can be accurate enough to be
used for RH retrievals. Then, when we analyze the agreement with
radiosounding measurements between 750 and 6000 m, lidar shows the best
results. Lidar–radiosounding median difference is 11.8 %, while
WRF–radiosounding median difference equals 28.6 % and
radiometer–radiosounding median difference reaches 86.6 % (Fig. 3). As
expected, the radiometer demonstrates reasonable agreement with reference
radiosoundings only in the lowest 4000 m since above this layer, only 5 % of
retrieved observations originate from the radiometer itself.
Absolute difference (%) between the mixing ratio of water
vapor from radiosoundings and lidar (green), microwave radiometer (red) and
WRF simulations (blue). All six high-resolution radiosounding cases from the
HygrA-CD campaign are used (15, 17, 18, 20, 21 May and 1 June).
Evaluation of temperature profiles
We analyzed the agreement and systematic bias of the available atmospheric
temperature profiles compared to the 6 high-resolution radiosondes data,
mentioned in the previous section. Only MWR (microwave radiometer) and WRF model simulations were
applied, since the EOLE lidar system is not capable of temperature
measurements. Just like humidity, physical temperature in the atmosphere is
related to the brightness temperature of the object. Therefore, the basic
principle of temperature retrieval by the radiometer is close to an analogical
principle as applied to humidity profiling. Temperature varies vertically at
the same rate for both radiometer measurements and model simulations for the
considered dates, and there is no strong inversion by any instrument or
simulation. Both the WRF model simulations and MWR measurements show high
agreement with radiosoundings (R2=0.99 and 0.98, respectively).
However, some minor temperature inversions measured by the radiosoundings
are not retrieved by the WRF model or the radiometer measurements (Fig. 4).
For most of the cases, the agreement between the WRF model simulations and
radiosoundings is better than the agreement between microwave radiometer and
radiosoundings. In particular, the difference of temperatures between WRF
model simulations and radiosoundings does not exceed 2 ∘C in the
lowest region of the troposphere. WRF–radiosounding mean absolute bias
(0.45 ∘C) is lower than radiometer–radiosounding mean absolute
bias (1.84 ∘C). Radiometer results meet our expectations,
according to Sect. 2.3, as the random error of temperature equals
1.7 ∘C. Considering the very high statistical agreement between
radiometer and WRF simulations with radiosoundings, the low systematic bias
and the fair agreement with literature values for the radiometer, we further used
these temperature datasets as input observations for RH calculation. The
role of the temperature uncertainty effects in resulting RH vertical
profiling techniques are considered in the Sect. 4.2.3.
Temperature profiles intercomparison for all heights. All
six high-resolution radiosounding cases from the HygrA-CD campaign are used
(15, 17, 18, 20, 21 May and 1 June): radiosoundings versus radiometer (a,
red) and radiosoundings versus WRF model simulations (b, blue).
Combined algorithms of relative humidity vertical profiling:
lidar–radiometer and lidar–WRF methods
We used two combined algorithms of RH vertical profiling: lidar–radiometer
and lidar–WRF methods. The basics of RH calculations are explained below and
presented along with description of both methods.
Lidar–radiometer combination for relative humidity vertical
profiling
The synergistic lidar–radiometer method (LD-MWR hereafter) is based on
two input datasets. The first input dataset is based on lidar measurements of
the water vapor mixing ratio. The second input dataset is taken from water vapor and
temperature measurements from the radiometer. AH profiles for lidar are
calculated from a simple conversion formula in which mixing ratio profiles are
multiplied with air density profiles (taken from the US Standard
Atmosphere). We separated the atmospheric column into three different
regions. The range of heights in lowest region (0–750 m) is dictated by
lidar overlap limitations and only radiometer measurements are applied below
750 m. In the middle region (750–2000 m) we combined mixing ratio
measurements from the lidar with temperature and humidity measurements from
the microwave radiometer in order to retrieve combined measurements of RH.
The radiometer-retrieved RH profiles are interpolated to lidar measurements
and then averaged with appropriate lidar measurements, which resulted in
combined RH profiles. The combination acts to smooth the transition from the
region where lidar is “blind” to the region where only lidar is used for
water vapor measurements (>2000 m). In order to retrieve the RH
profile, we calculate the saturation density of water vapor (es) (Eq. 3a)
which can be defined as the ratio between the molecular mass of water (M=18 g) and gaseous constant of water vapor (R=0.0623) multiplied by
the atmospheric temperature (T) taken from radiometer observations at the
appropriate heights. Then, the ratio between (M) and (RT) is multiplied by
the empirical value of the water vapor density (ps), the calculation of
which implies the use of temperature (T) as well (Eq. 3b).
es=MRT⋅ps(T)ps=0.61078⋅7.501e17.2694⋅T238.3+T
RH calculation includes the implementation of a simple conversion formula as
the ratio between AH from lidar and saturation vapor density calculated from
Eq. (3a).
Lidar–WRF combination for relative humidity vertical profiling
Using the lidar–WRF (LD-WRF) combined method for RH profiling we apply the
same principles and formulas of the RH calculation as for the LD-MWR method.
One difference is that instead of microwave radiometer measurements, we use
WRF simulations in the same three layers of the troposphere chosen for analysis.
Both synergetic methods are used simultaneously and compared with collocated
radiosoundings.
Evaluation of the relative humidity profiles
Microwave radiometer and WRF simulations versus high-resolution radiosounding
We determine the effectiveness of each considered RH vertical profiling
method and examine radiometer measurements (MWR), WRF model simulations,
LD-MWR and LD-WRF methods. At first, only MWR and WRF model simulations are
analyzed versus RH vertical profiles calculated from radiosoundings (Fig. 5).
In some cases, MWR measurements do not depict some finer-scale humidity
features in the middle troposphere (18 and 21 May, 1 June). The results are
somewhat ambiguous in the first 1000 m where the agreement between
radiosoundings with MWR or WRF simulations depend on every case. No clear
pattern is identified to see whether MWR or WRF simulations agree with the
radiosoundings or not. For instance, the agreement with radiosoundings is
very similar for MWR and the WRF model in the case of 15 and 20 May, where
the difference for both techniques does not exceed 10 % between 100 and
1000 m. Moreover, some higher humidity layers detected by radiosondes are
not seen by both MWR measurements and WRF simulations (20 May, at 4500 m, 21 May at 5000 m,
1 June between 4000 and 5000 m). These deficiencies that are
indirectly seen from the bias calculation can be alleviated by
addressing the combined algorithms.
Vertical profiles of relative humidity from WRF model
simulations (blue), microwave radiometer (red) and radiosoundings (grey).
Determination coefficients (R2) between various measurements
techniques or simulations of relative humidity profiling and high-resolution
radiosounding for different height regions. LD: lidar, MWR: microwave
radiometer, WRF: WRF model.
R2
MWR
WRF
LD-MWR
LD-WRF
Height regions
Low
Mid.
Full
Low
Mid.
Full
Low
Mid.
Full
Low
Mid.
Full
15 May
0.79
0.79
0.82
0.55
0.82
0.86
0.79
0.91
0.91
0.55
0.92
0.82
17 May
0.88
0.78
0.78
0.92
0.81
0.77
0.88
0.95
0.94
0.92
0.96
0.91
18 May
0.86
0.20
0.21
0.94
0.82
0.77
0.86
0.92
0.91
0.94
0.92
0.85
20 May
0.52
0.58
0.32
0.40
0.21
0.21
0.52
0.91
0.88
0.40
0.90
0.88
21 May
0.96
0.91
0.88
0.89
0.84
0.87
0.96
0.91
0.87
0.89
0.91
0.93
1 June
0.27
0.94
0.87
0.05
0.90
0.90
0.27
0.96
0.96
0.05
0.90
0.90
Mean
0.71
0.70
0.65
0.63
0.73
0.73
0.71
0.93
0.92
0.63
0.92
0.88
Synergetic methods versus high-resolution radiosounding
The LD-MWR and LD-WRF methods are intercompared to the RH vertical profiling
provided by high resolution radiosoundings and the results are shown for 6 days (15 May to 1 June). In general, both methods agree quite well with the
radiosonde data from 1000 to 6000 m height; however, significant differences
are evidenced. For example on 15 and 17 May the differences of both methods
to the radiosonde data remain quite large between 500 and 3000 m. On 20 and
21 May the agreement of both methods with the RH radiosonde data is very good
in the height region from 1000 to 6000 m. Finally, on 1 June both methods
show the best agreement with the radiosonde data. In Tables 2 and 3 the
efficiency of these approaches is further analyzed by addressing a
statistical analysis of the datasets obtained.
Vertical profiles of RH from combined methods for LD-WRF
(blue) and LD-MWR (red) against high-resolution radiosoundings (grey).
We calculated the R2 and the resulting mean bias of RH vertical
profiling obtained by each technique and the radiosonde data (Tables 2 and 3).
The R2 values for the combined methods are obviously similar to the
associated single methods (e.g., LD-MWR with radiometer and LD-WRF with WRF
model simulations) in the lowest region. Interestingly, analysis of the
middle region shows that the use of lidar-derived humidity data in the
combined methods drastically increases the agreement with radiosoundings in
comparison to the MWR and WRF model simulations alone. For instance, the
mean R2 value between MWR and radiosoundings in this region is 0.70,
while the replacement of low-resolution humidity data from the MWR to lidar
humidity data improves R2 (R2=0.93). However, the
WRF model-simulated RH in the same region shows reasonable agreement with
radiosounding data (R2=0.73). The agreement increases when we
combine WRF model-simulated temperature and lidar mixing ratio in this
region (R2=0.92). Both examples demonstrate that the role of the
lidar data becomes more effective in heights above 1000 m in combination
with measurement data (modeling data) obtained (simulated) within the PBL. The
highest R2 values (R2=0.92) in the full considered
tropospheric column are observed when the LD-MWR method is used.
The radiometer shows
the lowest mean bias (7.09) in the low region, while the WRF model shows a
poorer agreement with radiosoundings (12.32). We observe a significant
improvement in the middle and full region when we address the combined
algorithms. For example, the microwave radiometer when used alone shows a
mean bias of 13.63 and 11.61 in middle and full region, respectively, while
the LD-MWR combined method decreases the mean bias to 4.08 and 3.37,
respectively. A similar improvement is noted with the WRF model simulations
as well. Alone, the WRF model-simulated RH shows mean biases of 7.50 in the
middle region and 7.44 in the full region. The synergistic use of lidar
water vapor data in the full region decreases the mean biases down by nearly
75 % to the value of 1.29 in the middle region and 1.73 in the full
region.
Mean bias between various measurements
techniques or simulations of relative humidity vertical profiling and
high-resolution radiosounding measurements for different height regions. LD: lidar, MWR: microwave radiometer, WRF: WRF model.
Mean bias
MWR
WRF
LD + MWR
LD + WRF
Height regions
Low
Mid.
Full
Low
Mid.
Full
Low
Mid.
Full
Low
Mid.
Full
15 May
9.20
15.30
11.95
7.60
2.29
0.93
9.20
3.44
1.71
7.60
2.20
0.85
17 May
1.29
0.59
0.33
17.11
1.31
1.21
1.29
6.42
5.35
17.11
0.09
2.42
18 May
2.49
12.88
11.45
27.98
8.18
10.90
2.49
6.88
6.27
27.98
0.98
2.99
20 May
15.6
24.22
23.00
8.85
24.31
22.19
15.60
2.97
4.70
8.85
0.50
0.78
21 May
13.6
13.77
10.00
6.06
1.68
2.28
13.69
2.94
0.66
6.06
0.65
1.39
1 June
0.30
15.06
12.96
6.33
7.27
7.14
0.30
1.88
1.58
6.33
3.32
2.00
Mean
7.09
13.63
11.61
12.32
7.50
7.44
7.09
4.08
3.37
12.32
1.29
1.73
Synergetic methods versus
low-resolution radiosounding
At last we apply both synergetic methods in the comparison against the
low-resolution radiosounding data, when high-resolution radiosoundings were
not available. The LD-MWR and the LD-WRF method show a R2 value of 0.66
and 0.65, respectively, for heights between 100 and 6000 m. The agreement
between the combined methods of RH profiling and radiosoundings is again
quite satisfactory for the June period, as presented in Fig. 7. However, some
disagreement is found between 2500 and 4000 m on 2 June, and between 100
and 1500 m on 12 June, probably due to missing data from the radiosonde.
Vertical profiles of RH from combined methods: LD-WRF
(blue) and LD-MWR (red) versus radiosounding (grey) for the dates when only
low-resolution radiosoundings are available during June 2014.
We calculated again the R2 values between each method of RH profiling
(MWR, WRF model simulations, LD-MWR, LD-WRF) against low-resolution
radiosoundings (Table 4). In the lowest region the WRF-based methods show
the highest agreement with radiosoundings. The R2 for both separate WRF
simulations and LD-WRF method equals 0.70, while R2 for the MWR-based
measurements equals 0.59 for both cases (MWR and LD-MWR). In the middle
region, the use of lidar data contributes to an improvement in the agreement
with radiosounding for both MWR and WRF methods. In this case the
MWR-related R2 is improved from 0.54 (MWR) to 0.68 (LD-MWR) and the
WRF-related R2 is increased from 0.62 (only WRF) to 0.71 (LD-WRF). In
addition, if we ignore two cases where the lidar signal was noisy, the
LD-MWR agreement remarkably improves (R2=0.76), while the radiometer
alone remains nearly constant (R2=0.55). Finally, if we consider all
the heights from 100 to 6000 m one can summarize that both combined methods
(LD-MWR and LD-WRF) show the highest statistical agreement in comparison
with single methods. The R2 value of LD-MWR equals 0.66 in this case
(and is up to 0.73 when the noisy lidar data are not taken into account), while the use of radiometer data alone results in a mean R2 of only
0.49 for these heights. The LD-WRF method has a R2 value of 0.66 at
these heights (0.73 if lidar noisy data are excluded), while the WRF model
simulations alone result in similar R2 values (0.65). The uncertainties
of RH profiles associated with temperature variability are discussed in
detail for both methods in Sect. 4.4.
Determination coefficients between various measurements
techniques or simulations of relative humidity profiling and radiosounding
for different height regions (low-resolution radiosounding cases). LD: lidar, MWR: microwave radiometer, WRF: WRF model.
Det. coef.
MWR
WRF
LD + MWR
LD + WRF
Height regions
Low
Mid.
Full
Low
Mid.
Full
Low
Mid.
Full
Low
Mid.
Full
22 May
0.34
0.63
0.62
0.70
0.76
0.75
0.34
0.85
0.82
0.70
0.85
0.87
23 May
0.95
0.88
0.77
0.95
0.92
0.93
0.95
0.97
0.97
0.95
0.97
0.94
24 May
0.77
0.44
0.53
0.83
0.18
0.76
0.77
0.34
0.45
0.83
0.35
0.11
25 Maya
–
0.19
0.28
–
0.76
0.75
–
0.34
0.33
–
0.36
0.35
26 May
0.96
0.89
0.70
0.87
0.90
0.89
0.96
0.96
0.96
0.87
0.97
0.96
28 Maya
0.20
0.78
0.67
0.61
0.37
0.40
0.20
0.22
0.30
0.61
0.29
0.34
2 June
0.68
0.46
0.44
0.74
0.56
0.54
0.68
0.62
0.61
0.74
0.61
0.58
5 June
0.64
0.56
0.54
0.62
0.68
0.67
0.64
0.89
0.64
0.62
0.89
0.87
12 June
0.22
0.10
0.14
0.44
0.81
0.77
0.22
0.93
0.92
0.44
0.92
0.89
14 June
0.59
0.88
0.89
0.68
0.84
0.78
0.59
0.96
0.96
0.68
0.97
0.92
15 June
0.91
0.21
0.04
0.66
0.13
0.05
0.91
0.55
0.40
0.66
0.56
0.31
18 June
0.24
0.43
0.29
0.55
0.51
0.50
0.24
0.56
0.55
0.55
0.72
0.72
Mean
0.59
0.54
0.49
0.70
0.62
0.65
0.59
0.68
0.66
0.70
0.71
0.66
a Days when the lidar signal-to-noise ratio is not high enough in the
middle troposphere.
Mean and maximum absolute uncertainty of RH synergetic
retrieval for two combined methods (lidar–radiometer and lidar–WRF) for
different height regions. Data from 25 May are excluded from this analysis due to the signal-to-noise
ratio. The lidar system is registered to adequately assess the uncertainties
resulting from final RH profiles for this day.
Uncertainties
LD + MWR (mean)
LD + WRF (mean)
LD + MWR (max.)
LD + WRF (max.)
Height regions
Low
Mid.
Full
Low
Mid.
Full
Low
Mid.
Full
Low
Mid.
Full
15 May
2.63
3.41
2.59
1.62
0.79
0.90
2.67
9.51
9.51
1.99
2.15
2.15
17 May
2.94
8.05
6.56
2.00
1.73
1.76
2.79
10.86
10.86
2.16
2.36
2.36
18 May
2.43
5.42
4.36
1.79
1.03
1.13
2.27
13.34
13.34
2.17
2.32
2.32
20 May
3.00
5.36
4.22
1.22
1.19
1.19
1.62
9.49
9.49
1.37
1.99
1.99
21 May
1.58
2.58
2.02
1.55
0.55
0.69
2.22
8.16
8.16
1.84
1.88
1.88
22 May
2.25
5.12
4.12
1.67
1.19
1.26
2.40
8.60
8.60
2.03
2.10
2.10
23 May
2.44
5.70
4.60
1.39
1.24
1.26
1.75
9.62
9.62
1.63
2.11
2.11
24 May
2.50
9.33
7.73
1.78
2.00
1.97
1.91
14.08
14.08
2.12
3.11
3.11
26 May
2.26
4.62
3.68
1.46
1.06
1.11
1.81
8.98
8.98
1.56
1.97
1.97
28 May
1.41
6.33
5.29
0.95
1.57
1.49
1.35
13.81
13.81
1.09
4.25
4.25
1 June
2.77
5.70
4.56
1.71
1.21
1.28
2.48
10.06
10.06
1.89
2.23
2.23
2 June
2.33
7.46
6.13
1.79
1.43
1.47
2.82
14.33
14.33
2.08
2.54
2.54
5 June
2.94
5.24
4.13
1.40
1.12
1.16
1.97
10.85
10.85
1.55
2.38
2.38
10 June
3.51
5.51
4.28
1.58
1.17
1.22
2.43
8.92
8.92
1.85
1.92
1.92
12 June
2.48
3.73
2.88
1.33
0.77
0.85
1.92
9.01
9.01
1.53
1.84
1.84
14 June
2.31
4.82
3.85
1.54
1.10
1.16
1.40
8.97
8.97
1.84
1.98
1.98
15 June
2.53
7.47
6.12
1.37
1.84
1.78
2.08
9.42
9.42
1.70
2.25
2.25
18 June
1.86
5.28
4.31
1.22
1.21
1.21
1.63
8.72
8.72
1.35
2.29
2.29
Abs. mean
2.45
5.46
4.34
1.55
1.19
1.22
Max
2.82
14.33
14.33
2.17
4.25
4.25
Role of temperature random error in relative humidity vertical
profiling based on lidar–radiometer and lidar–WRF methods
Points of measurements were not depicted in Figs. 5, 6 and 7
since all measurements and simulations are interpolated to high-resolution
measurements of water vapor mixing ratio. Error bars were also not shown
since it is challenging to compare instrumental-related errors from
radiometer brightness temperature noise and WRF-model systematic errors. To
alleviate such gaps and to conclude our analysis we examine the uncertainties
of each synergistic RH vertical profiling method. The largest uncertainties
in synergetic methods can be most likely caused by the errors in temperature
input. For the MWR we take temperature systematic bias values from
the literature as described in Sect. 2.3. Due to the difference in the performance of
the radiometer at different altitudes, as shown before, we use a random error of
0.5 ∘C in the altitudes below 1000 m and 1.7 ∘C in the
layers above 1000 m. Since WRF model uncertainties cannot be addressed in
the same manner as the radiometer random error, we use for the
WRF-simulation radiosounding a mean absolute bias of 0.45 ∘C
from our results (shown in Sect. 3.3) to use as temperature random error in
our simulations. Therefore, we calculate the RH profiles based on both
combined methods with different temperature inputs (minimum, mean and
maximum according to abovementioned temperature random errors) where minimum
and maximum temperatures are used to calculate the minimum and maximum RH
values for each method. In such a manner, difference between the resulting
maximum and minimum RH represents the final uncertainty of our RH profiling
due to temperature random errors. When low tropospheric layers are
considered, as we expected, the synergetic methods show improved performance
as the temperature variations are not high for both MWR and WRF simulations.
As a result, the LD-MWR method results in a mean uncertainty of 2.45 % of
RH due to temperature variations, where the RH uncertainty does not exceed
2.82 % in the lowest layer for LD-WRF, respectively. The LD-WRF method in
the lowest layer results in a mean uncertainty of 1.55 % for the RH
vertical profiling where the maximum possible uncertainty equals
2.17 %. The RH results show higher discrepancies in the layer between 1000
and 6000 m for the LD-MWR method due to an increased temperature-related error
propagation. The mean LD-MWR uncertainty of RH profiling in this layer is
5.46 %. Meanwhile, the LD-WRF method shows high performance in the middle
layer based on a calculated mean uncertainty of RH equals 1.19 %. The
results of both methods seem reasonable based on the mean values of RH
uncertainties. However, by analyzing the maximum uncertainty of RH we
can see that temperature deviations in input data may affect our RH vertical
profiles very significantly. For instance, when radiometer temperatures
input varies by 1.7 ∘C, we face larger error propagations (of the
order of 10 %) in the middle troposphere observations (see Table 5).
Despite this, average uncertainties for both LD-MWR and LD-WRF are reasonable,
and some atmospheric layers can still be reconstructed with rather high
uncertainties in RH (maximum can reach 14.33 % for LD-MWR). This
phenomenon is evidently not seen in the LD-WRF method, as we assumed a
temperature simulation uncertainty to be uniform and rather low along the
atmospheric column. Considering this assumption, the resulting uncertainties
from RH profiling using the LD-WRF method in the middle troposphere does not
exceed 4.25 %. We also analyze full considered atmospheric height from 100
to 6000 m. Here LD-WRF shows quite accurate RH profiling performance;
average uncertainty of RH resulting from temperature error propagation is
1.22 % and maximum is 4.25 %. The LD-MWR method shows reasonable results
according to the mean uncertainties (4.34 %) for full considered
heights; however, for some atmospheric layers the maximum uncertainty from this method
was also high (of the order of 13 %).
Conclusions
In this study, we addressed two synergistic methods for RH vertical
profiling based on combined lidar–radiometer and lidar–WRF datasets. The
first method exploited water vapor mixing ratio profiles derived from lidar
measurements in combination with temperature measurements from a collocated
radiometer as input datasets for RH calculation. The second method combined
water vapor mixing ratio profiles derived from Raman lidar data with
high-resolution simulations from the WRF model. We showed the advantages and
disadvantages of these methods for RH vertical profiling for case studies
during the HygrA-CD campaign. We evaluated both methods according to current
thermodynamic profiling requirements regarding accuracy, spatial resolution
and measurement range analysis of observations. Prior to the main analysis,
the EOLE Raman lidar system had been successfully calibrated for water vapor
measurements by using collocated high-resolution radiosonde data and
yielding an optimal calibration constant (C=23.65±2.28). The sanity
of the high-resolution calibration constant was checked by the comparison
with a second calibration constant calculated from low-resolution radiosonde
data (C=24.36±2.14). The ability of each instrument to retrieve
water vapor parameters has been checked via IWV intercomparison. IWV
intercomparison showed very high agreement between all comparable
instruments (R2=99 for lidar–radiometer and R2=0.89 for sun
photometer-radiometer).
We determined the lowest and the upper threshold heights for the water vapor
vertical profiling using Raman lidar. Based on the absolute difference
between the mean calibration constant C and 3 standard deviation we
concluded that below 750 m the combined methods should rely solely on
radiometer observations and WRF simulations. The upper threshold layer for
water vapor measurements has been determined accordingly at 6000 m. In the
atmospheric layer constrained by these boundaries (750–6000 m), the lidar
water vapor retrievals are the most accurate and reliable. The use of
high-resolution (7.5 m) water vapor measurements gave advantages in terms of
resolution and accuracy for both synergetic methods between 1000 and 6000 m.
We exploited high-resolution collocated radiosoundings as a reference for
our RH observations. In terms of accuracy, we demonstrated the significant
improvement from single (radiometer or WRF simulations) to synergetic
methods of RH profiling. The R2 value between the WRF model alone and
high-resolution radiosoundings was found to be equal to 0.73, while the use of the
LD-WRF method improved this agreement slightly (R2=0.88). Similarly, the
use of the radiometer alone resulted in a mean R2 of 0.65 against
high-resolution radiosoundings, while the LD-MWR method showed remarkable
improvement (R2=0.92). We observed a similar improvement by using the
LD-MWR method when low-resolution radiosoundings were employed (launched for
12 days during the campaign). The mean R2 value between radiometer RH
profiles and low-resolution radiosounding RH profiles improved from 0.49 to
0.66. When low-resolution radiosondes were used, the LD-WRF method did not show
any improvement in comparison with RH stand-alone WRF model simulations. No
improvement in this case stems rather from the lack of high-resolution
reference measurements than from the issues associated with the combined method.
In order to understand the role of temperature variations in resulting RH
profiles, we examined the temperature-related errors for both LD-MWR and
LD-WRF methods. Temperature random errors for the radiometer were taken from the
literature (±0.5 ∘C for 0–1000 m and ±1.7 ∘C
for 1000–6000 m) and for WRF simulations from a systematic
bias of ±0.45 ∘C calculated in our study. According to the
mean RH uncertainties (between 100 and 6000 m), both methods showed a good
performance since the LD-WRF and LD-MWR mean temperature-related
uncertainties equal to 1.22 and to 4.34 %, respectively. However, in
future studies, the mean uncertainties of the LD-MWR method should be
carefully applied as the sources of systematic errors since LD-MWR
temperature-associated uncertainties in some cases may result in
significant error propagation (up to 14.33 %).
During the HygrA-CD campaign, both combined methods showed an improved
performance in terms of spatial resolution and accuracy in comparison to
single observations and high-resolution WRF simulations. This improvement is
remarkable in the height range between 1000 and 6000 m. When RH observations
are required, both combined methods can be valuable in terms of accuracy and
spatial resolution of the measurements with reasonable limitations. In
future studies, LD-MWR and LD-WRF methods can be beneficial for the steadily
increasing number of atmospheric stations possessing both microwave
radiometer and Raman lidar systems.