The algorithm presented in this paper was developed to retrieve ice water content (IWC) profiles in cirrus clouds. It is based on optimal estimation theory and combines ground-based visible lidar and thermal infrared (TIR) radiometer measurements in a common retrieval framework in order to retrieve profiles of IWC together with a correction factor for the backscatter intensity of cirrus cloud particles. As a first step, we introduce a method to retrieve extinction and IWC profiles in cirrus clouds from the lidar measurements alone and demonstrate the shortcomings of this approach due to the backscatter-to-extinction ambiguity. As a second step, we show that TIR radiances constrain the backscattering of the ice crystals at the visible lidar wavelength by constraining the ice water path (IWP) and hence the IWC, which is linked to the optical properties of the ice crystals via a realistic bulk ice microphysical model. The scattering phase function obtained from the microphysical model is flat around the backscatter direction (i.e., there is no backscatter peak). We show that using this flat backscattering phase function to define the backscatter-to-extinction ratio of the ice crystals in the retrievals with the lidar-only algorithm results in an overestimation of the IWC, which is inconsistent with the TIR radiometer measurements. Hence, a synergy algorithm was developed that combines the attenuated backscatter profiles measured by the lidar and the measurements of TIR radiances in a common optimal estimation framework to retrieve the IWC profile together with a correction factor for the phase function of the bulk ice crystals in the backscattering direction. We show that this approach yields consistent lidar and TIR results. The resulting lidar ratios for cirrus clouds are found to be consistent with previous independent studies.
The importance of clouds for the climate system has been
extensively discussed during the last decades
Recent advances in satellite observational systems, particularly the Cloud
Profiling Radar (CPR;
Lidar systems have proven to be powerful tools to study even the most
tenuous cloud layers
In spite of their undenied importance, lidar systems are not the only tools
to study cirrus clouds. Another important source of information are
measurements from passive TIR radiometers, which are also performed from the
ground or from space, e.g., the Imaging Infrared Radiometer (IIR) aboard
CALIPSO
However, in recent years synergistic approaches using independent sets of
measurements in a common retrieval framework have become more and more
popular. Examples that could be cited here are the raDAR/liDAR (DARDAR)
algorithm to retrieve ice cloud properties from the synergy of the CPR and
CALIOP measurements
The algorithm proposed in this paper also establishes a synergy between lidar
and TIR radiometer measurements, although our lidar is a simple micro-pulse
lidar and does not possess depolarization channels. In contrast to
The paper is organized as follows: Sect.
The data used in this study originate from the measurement platform of the Laboratoire d'Optique Atmosphérique (LOA) situated on the campus of the University of Lille in northern France. This platform is equipped with, amongst other instruments, an elastic-backscatter micro-pulse lidar and a TIR radiometer.
The lidar is a Cloud and Aerosol Microlidar (CAML) CE370
The radiometer is called Conveyable Low-Noise Infrared Radiometer for
Measurements of Atmosphere and Ground Surface Targets (CLIMAT)
It should be noted that due to the larger FOV of the TIR radiometer compared to the FOV of the lidar, the two instruments do not see exactly the same cloud area. This difference also depends on the altitude of the cloud. As in almost all remote-sensing algorithms, we assumed a homogeneous cloud in the instrument FOV and did not take into account any uncertainty due to sub-pixel heterogeneity.
The algorithm presented here was developed to retrieve profiles of particle
extinction and IWC from measured lidar backscattering profiles. We propose a
method to simultaneously retrieve a profile of aerosol extinction in the
layers close to the ground and a profile of IWC inside cirrus layers.
However, our main focus is the characterization of the microphysical
properties of cirrus clouds. There are many techniques to invert the lidar
equation including the classical Klett–Fernald method
The relationship between the range-resolved backscattered power,
The retrieval of particle optical properties from elastic lidar measurements
alone is challenging since there is an intrinsic ambiguity between the
effects of backscattering and extinction arising from the combination of
scattering and absorption processes in the atmosphere. In Eq. (
In this study we are focusing on the retrieval of cirrus cloud properties.
Thus, the backscatter-to-extinction coefficient for aerosols is assumed to be
constant and is fixed to 64 sr. This value originates from the Optical
Properties of Aerosols and Clouds (OPAC) database
The backscatter-to-extinction ratio for cirrus clouds is calculated using the
definition of
The ensemble model of
We obtain the single-scattering properties (scattering coefficient,
absorption coefficient and asymmetry parameter) for each cloud layer from the
parametrization of
Since our algorithm seeks to retrieve the IWC for cirrus cloud layers, the
extinction required in the lidar equation (Eq.
As mentioned above, we apply an optimal estimation method to invert the lidar
equation following
Following
To find the best estimate of the state vector
The application of this theoretical framework to the lidar retrieval problem
requires the definition of all necessary elements described above. The state
vector
As mentioned in Sect.
Following
The Jacobian
It should be noted that in the case of opaque cirrus clouds that completely
attenuate the lidar signal, the size of the measurement vector and
consequently the size of the state vector are reduced. In this case, only the
altitudes until full attenuation of the lidar signal are considered. Thus,
the size of the Jacobian is reduced as well and it contains only
Following
As mentioned above, vector
It should be noted that the characterization of the errors related to the
BV2015 parametrization is very challenging. Thus, no error for the
microphysical model is currently taken into account. However, this issue
needs to be addressed in future studies, and an evaluation of the uncertainty
arising from the
To start the iteration, a first guess is required and in this study we chose
to use the a priori state vector as a first guess. In order to reach faster convergence of
the algorithm, the elements of the a priori state vector for the layers close to the ground where aerosols
are present are calculated from a one-step solution of the lidar equation
following the approach of
Forward model of the
Figure
Figure
The retrieval results for the whole afternoon of 30 November 2016 are shown
in Fig.
Figure
Retrieval results for 30 November 2016, 15:00 to 20:00 UTC.
Our retrievals strongly depend on the phase function in the backscattering
direction, which defines the backscatter-to-extinction ratio. As explained in
Sect.
Examples of phase functions for different degrees of particle
heterogeneity. Black line: phase function for a bulk ice crystal with a
smooth surface; blue line: introduction of some heterogeneity; green line:
maximum degree of heterogeneity (particle roughness, air bubbles). The red
line represents the phase function obtained from the parametrization of
Figure
Having that in mind, we tested the influence of the backscatter-to-extinction
ratio in our algorithm. Figure
As discussed in Sect.
Dependence of the retrieved IWC on the backscatter-to-extinction
ratio for
Since TIR radiances are sensitive to the integrated properties of the cloud,
in particular the IWP, we can use them to constrain the amount of ice in the
cloud and hence the backscatter-to-extinction ratio. CLIMAT radiometer
measurements (see Sect.
The normalized radiance (emitted by a source of brightness temperature
Figure
The aim of our method is to use these TIR radiances to constrain the
backscatter-to-extinction ratio correction factor (
Same as Fig.
Figure
However, the curves shown in Figs.
The results presented in this section show that the ensemble of measurements
should be used to find a retrieval that corresponds best to all available
information. As mentioned above, the optimal estimation method is a
well-adapted tool to use different kinds of measurements in a common
retrieval framework. The results shown here confirm that the TIR radiances
provide an additional constraint for the amount of ice inside the cloud, which
strongly depends on the backscatter-to-extinction ratio. Therefore, the phase
function in the backscattering direction can be constrained by the TIR
radiometer measurements under the assumption that the single-scattering
albedo is accurately known (see Eq.
The synergy algorithm is an expansion of the lidar-only algorithm, which
integrates the TIR radiometer measurements in the optimal estimation method.
The new state vector contains, in addition to the elements of the previous
state vector given by Eq. (
The measurement vector, initially containing the logarithm of the calibrated
range-corrected lidar signal, is expanded by the measured radiances from the
two channels of the CLIMAT instrument discussed above:
The forward model for the lidar is the same as in the case of the lidar-only
algorithm, given by the lidar equation in the form of Eq. (
The forward model for the TIR radiances is the abovementioned radiative
transfer model LIDORT
The Jacobian of the synergistic algorithm contains, in addition to the
Jacobian of the lidar-only algorithm, two new rows for the sensitivity of the
TIR forward model to each state vector parameter and one new column for the
sensitivity of the forward model to the new state vector element
The sensitivities of the TIR radiances to the extinction profile outside the
cloud are set to zero since they are assumed to be small. The correction
factor
As in the lidar-only algorithm, the variance–covariance matrices in the
synergy algorithm are also considered to be diagonal. Concerning the lidar,
they are defined in the same way as in the lidar-only algorithm (see Sect.
This section presents some preliminary results of our new algorithm. Figure
TIR forward model and measured normalized radiances
(
The retrieved value for the correction factor
Lidar forward model of the
The application of the synergy algorithm to the profile measured at 16:20 UTC
on 30 November 2016 results in a factor
TIR forward model and measured normalized radiances
(
Retrieval results of the synergy algorithm for 30 November 2016,
15:00
to 20:00 UTC.
Finally, Fig.
However, between 16:00 and 18:18 UTC the COT obtained from the synergy algorithm
underestimates the COT derived from the transmission method for most of the
retrievals (except around 18:12 UTC). It should be noted that the COT obtained
from the transmission method is an effective COT
However, the multiple-scattering factor alone cannot explain the
inconsistency between the COT retrieved with the synergy algorithm and the
COT derived from the transmission method. Another possible reason for this
discrepancy may arise from the uncertainty in the transmission method itself
because it depends on a good characterization of the molecular signal above
the cloud and a good estimation of the cloud-top altitude. These parameters
are related to rather large uncertainties due to the quite noisy
micro-pulse lidar signal in the high altitudes of cirrus clouds. Furthermore,
the discrepancy between the two COTs could also originate from a potential
bias in the TIR radiometer measurements due to an inaccurate temperature
correction as mentioned in Sect.
Despite these limitations, the retrievals of the lidar ratio shown in Fig.
Nevertheless, this new synergistic algorithm suggests that using information from both, active in the visible part of the electromagnetic spectrum and passive in the TIR part, allows us to obtain new information on bulk ice optical properties, especially on the amount of ice and its capability to backscatter the visible light. Moreover, it allows us to test existing microphysical models, particularly the BV2015 model and its original representation of bulk optical properties as a function of the in-cloud temperature and IWC. The results of this study point out the overall good coherence of the BV2015 model but also its limitations in representing all the different measured profiles, especially due to the poor representation of the exact backscattering characteristics of the bulk ice.
In this paper a method to retrieve IWC profiles of cirrus clouds from the synergy of ground-based lidar and TIR radiometer measurements has been presented. The algorithm is based on optimal estimation theory and combines the visible lidar and TIR radiometer measurements in a common retrieval framework to retrieve profiles of IWC together with a correction factor for the backscatter intensity of bulk ice cloud particles.
As an initial step, an algorithm to retrieve IWC and extinction profiles (outside the cloud) from the lidar measurements alone was developed. Due to the backscatter-to-extinction ambiguity arising from the combination of scattering and absorption processes in the atmosphere, assumptions are required for the backscatter-to-extinction ratio, and the retrieval results strongly depend on these assumptions. As a consequence, the challenge is to find ways to reduce the uncertainties in the retrieval arising from insufficient knowledge of the backscatter-to-extinction ratio.
To overcome the backscatter-to-extinction ambiguity, we showed in a second step that it is possible to use TIR radiances to constrain the backscatter-to-extinction ratio defined as the product of the single-scattering albedo and the phase function in the backscattering direction. The latter has not yet been fully characterized and is associated with large uncertainties. Moreover, it strongly depends on the characteristics of the particles composing the cloud. However, the BV2015 microphysical model links the optical properties of cirrus clouds directly to the IWC without the need for assumptions about the particle shape and PSD. This model allows us to obtain the single-scattering albedo and the asymmetry parameter (from which the phase function is parametrized) as a function of IWC and in-cloud temperature alone. Our algorithm benefits from the fact that TIR radiances are sensitive to the integrated IWC over the whole cloud (IWP) and that the IWC of each layer governs the optical properties via the microphysical model. That means the backscatter intensity of the ice crystals is constrained by the TIR radiances under the assumption that the single-scattering albedo is represented sufficiently accurately in the microphysical model. Consequently, our synergy algorithm retrieves a profile of IWC together with a correction factor for the phase function of the ice crystals in the exact backscattering direction, which is assumed to be constant over the entire cloud profile. Hence, the integration of the TIR radiances into the optimal estimation framework allows us to retrieve the lidar ratio although we use backscattering profiles from a simple micro-pulse lidar.
It is important to note that the same microphysical model has been used to compute the bulk ice optical properties (i.e., the scattering and absorption coefficients as well as the asymmetry parameter and the phase function) for all wavelengths considered in this study. The consistency of this microphysical model over a large portion of the electromagnetic spectrum ranging from the visible to the infrared ranges has been tested in numerous studies. Nevertheless, the parametrization of these optical properties as a function of IWC and temperature may introduce some uncertainty. However, a personal communication from Anthony J. Baran (2018) suggests that the error introduced by such a parametrization is rather small (smaller than 5 %). Thus, we believe that the results presented in this paper are robust and mainly point out the misrepresentation of the phase function in the exact backscattering direction, which is a key result of this study.
Another achievement of our algorithm is the integration of information from
the whole atmospheric profile, accessible thanks to the active lidar
measurements, in the forward modeling of the TIR radiances. Most common
retrieval algorithms for passive sensors assume a homogeneous cloud and
include only information about the cloud altitude from active measurements in
the radiative transfer calculations. The synergy between the lidar and the
TIR radiometer measurements established in this paper allows us to account for
the profile of IWC in the radiative transfer model. Furthermore, the
extinction of aerosols that may be present in the atmosphere is included in
the TIR forward model although further information on the aerosol type is
required. In this study, the aerosol optical properties were fixed to a
predefined aerosol model and an improvement of our method would be to better
characterize the properties of the aerosols that are actually present during
the measurement. It is worth noting that the high vertical resolution of the
radiative transfer calculations in the TIR is possible thanks to the
numerical efficiency of the radiative transfer model LIDORT discussed in
Sect.
The results for the case study discussed in Sect.
Furthermore, when regarding ground-based TIR radiometer measurements, a good
characterization of the surrounding atmosphere, especially the water vapor
profile, is crucial since the TIR radiances are very sensitive to water
vapor,
which is spatially and temporally highly variable. Hence, the ECMWF
reanalysis profiles used in this study, which are available for four time
steps at 00:00, 06:00, 12:00 and 18:00 UTC and have a spatial resolution of 1
Finally, the quality of the measured TIR radiances plays an important role.
For the case study presented here, the temperature correction of the
sensitivity of the instrument results in a quite large uncertainty because of
a large temperature difference between the temperatures during the
measurement and the calibration. It was shown in Sect.
Nevertheless, the first results obtained from this algorithm are promising and we showed that our method allows us to converge at the same time towards the measurements of two very different instruments. However, these results have to be confirmed in future studies for other measurement periods and measurement sites.
Data used in this paper are available upon request to the corresponding author.
FH, LCL, FP, and GB conceived the method, developed the retrieval algorithm and discussed the results. FH and LCL analyzed the data and prepared the figures. FH wrote the paper. BD conducted the calibration of the TIR radiometer. TP processed the lidar data. All co-authors reviewed the paper.
The authors declare that they have no conflict of interest.
The authors thank the Région Hauts-de-France, the Ministère de l'Enseignement Supérieur et de la Recherche (CPER Climibio) and the European Fund for Regional Economic Development for their financial support. The authors thank the CaPPA project (Chemical and Physical Properties of the Atmosphere) funded by the French National Research Agency (ANR) through the PIA (Programme d'Investissement d'Avenir) under contract ANR-11-LABX-0005-01 and by the Regional Council “Hauts-de-France” and the European Funds for Regional Economic Development (FEDER). The ACTRIS-FR research infrastructure is acknowledged for financial support. Edited by: Alexander Kokhanovsky Reviewed by: two anonymous referees