Aerosol and clouds are important atmospheric players influencing weather and climate. In contrast to long-lived gaseous components in the atmosphere, these components are short lived and feature a strong spatiotemporal variability. Aerosols act as cloud condensation nuclei and ice nucleating particles and are thus one major driver for cloud optical and microphysical properties and precipitation initiation. Because aerosol is emitted from various sources and is short-living, several aerosol types with different optical and microphysical properties exist in different heights of the atmosphere, influencing solar radiation and clouds in different ways. Therefore, the climate effects of aerosol directly and of aerosols on clouds (indirectly) are still very uncertain .

In order to better quantify the spatiotemporal distribution of aerosol and clouds as well as to improve the determination of their interaction, it is essential to observe aerosol and clouds, preferably in 4-D, but realistically round the clock and vertically resolved. Active satellite-based sensors such as CALIPSO Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation; , CloudSat , CATS Cloud-Aerosol Transport System; , and, planned for future space missions, Atmospheric Dynamics Mission Aeolus  and EarthCARE Earth Clouds, Aerosols and Radiation Explorer; cover the globe but with low temporal and spatial resolution. Thus, high-performance ground-based observations are also needed. Scientific networks such as Cloudnet  or the ARM (Atmospheric Radiation Measurement) Climate Research Facilities  use different ground-based instruments at the same location (supersite) to gather as much information as possible in high temporal and spatial (i.e. vertical) resolution but at specific locations only. Cloudnet, for example, uses as minimum instrumentation a cloud radar, a ceilometer (a simple backscatter lidar), and a microwave radiometer (MWR) to characterize the atmosphere above the supersite. Cloudnet delivers several products, ranging from calibrated measurements to microphysical cloud properties. Very well known and widely used is the Cloudnet target categorization , which classifies a series of different particle types in the observed atmospheric column (e.g. liquid droplets, ice crystals, aerosols). However, Cloudnet is not able to distinguish different aerosol types in its current state, which is a prerequisite to constrain aerosol–cloud-interaction studies and to improve the estimation of the radiative properties of aerosol.

Active remote sensing with lidar is a key technique for characterizing aerosol and allows capturing the atmospheric state on a vertically resolved basis usually covering the whole troposphere. For an intense characterization of aerosol type and properties, so-called multiwavelength lidars (MWLs) are applied using the synergistic information from different wavelengths, scattering mechanisms, and polarization states of the received light e.g. .

Optical aerosol properties have been widely investigated by using lidar profiles in low temporal resolution, applying the traditional Raman method , the Klett–Fernald method , and the depolarization method e.g. to determine the intensive properties (Ångström exponent, extinction-to-backscatter (lidar) ratio, particle depolarization ratio) of different aerosol types and their mixtures .

Based on such measurements, classification schemes for aerosol from high-resolution lidar measurements have been developed for space-borne lidars (CALIPSO, ; EarthCARE, ), airborne High Spectral Resolution Lidar (HSRL) measurements , some specific ground-based instruments (ARM, Darwin, Australia, ), and lidar networks focusing on the determination of mineral dust concentration in Asia (Asian Dust Network, AD-NET; ).

Due to recent advances in hardware, small sophisticated ground-based MWLs e.g. PollyXT lidar systems, , which can run unattended and autonomously 24 h a day, 7 days a week (24/7), have been developed and are deployed globally. Motivated by this technical progress, we aimed at developing a stand-alone tool for continuously running multiwavelength-polarization lidars for a basic categorization of the observed particles (targets) in analogy to the Cloudnet target categorization. With this tool, we want to obtain an estimate of the dominant type of backscatterer (molecules, aerosol types, clouds) which then can be used for further intensive studies and to complement the Cloudnet target categorization. For this approach we focus on the derivation of certain key parameters, which are not needed with high accuracy but are sufficient to perform a first estimate of certain particle types in the atmosphere. The basic lidar quantities used are the attenuated backscatter coefficients at 532 and 1064 nm (calibrated range-corrected lidar signal) and the calibrated volume linear depolarization ratio at 532 nm. These key parameters are highly useful as they are available for many continuously measuring lidar systems worldwide, e.g. the lidars within PollyNET , AD-NET, and the space-borne lidar CALIPSO. From these lidar parameters, further products have been developed to allow a first-guess particle typing.

To develop this tool and demonstrate the feasibility, potential, and limitations of this approach, we have used the unique data set obtained during the High-Definition Clouds and Precipitation for advancing Climate Prediction (HD(CP)2) Observational Prototype Experiment (HOPE; ) in western Germany. The MWL PollyXT  and the Cloudnet instruments (cloud radar, ceilometer, MWR) were deployed in the frame of the Leipzig Aerosol and Cloud Remote Observations System LACROS, next to each other at Krauthausen, Germany, continuously for 2 months in Spring 2013. PollyXT is a sophisticated, compact, scientific multiwavelength lidar to which the quality-assurance procedures proposed by EARLINET European Aerosol Research Lidar Network; are applied. Without such high-quality measurements, a proper aerosol characterization as described in the following is not possible. The collocation of the instruments makes the derived data set a perfect environment for developing an aerosol classification from MWL while the Cloudnet categorization can be performed in parallel.

For the HOPE data set, we perform a so-called absolute calibration on the lidar observations from automatically derived particle backscatter coefficient profiles  and derive atmospheric parameters in high temporal resolution, which allow us to estimate size and shape and finally type of the particles in the atmosphere. This basic typing can then be used for detecting different aerosol layers, for further research like on aerosol–cloud-interaction processes, as input for calibration procedures to automatically retrieve optical properties of the observed particles (e.g. ), and finally even for retrieving microphysical properties  which then may lead to an advanced particle categorization e.g. HETEAC, hybrid end-to-end aerosol classification model; .

The paper is structured as follows: first, the HOPE campaign, i.e. location and instrumentation, is briefly introduced in Sect. . The methodology to derive quantitative lidar parameters with temporal high resolution is explained in Sect. . Next, the new target categorization is introduced and intensively discussed by means of three case study days during HOPE in Sect. . The new methodology was applied on the complete HOPE data set and analysed in Sect. . Finally, conclusions are drawn in Sect. .

During the HD(CP)2 Observational Prototype Experiment (HOPE; ), the multiwavelength-Raman lidar  PollyIfTXT  was deployed at Krauthausen (50.879746∘ N, 6.414571∘ E, 110 ma.s.l.), near Jülich, western Germany, in April and May 2013 as part of the LACROS facility . A detailed description of the campaign together with the prevailing meteorological conditions can be found in .

PollyIfTXT system version labelled “IfT”, compare is an automatic, portable multiwavelength-polarization Raman lidar with automatic calibration procedures of the latest standards which was operated in 24/7 mode during HOPE. The lidar emits pulses of linearly polarized light at 355, 532, and 1064 nm at a repetition frequency of 20 Hz. The receiver has seven channels detecting the elastically backscattered light at the three aforementioned wavelengths, the cross-polarized light at 532 nm, and the vibrational Raman scattered light at 387, 407, and 607 nm. With PollyIfTXT, aerosol profiles can be obtained with 30 m vertical and 30 s temporal resolution. The full overlap between the laser beam and the receiver field of view is about 1500 m; thus, an overlap correction  is applied below this height. The lidar was operated in photon-counting mode. The system is pointed 5∘ off-zenith to prevent the detection of specular reflection by the planar planes of horizontally oriented ice crystals . A detailed description of the system including a quality assessment can be found in .

Furthermore, a cloud radar, a Doppler wind lidar, a ceilometer, and an AERONET (Aerosol Robotic Network) sun photometer were deployed next to the lidar as part of the LACROS facility. From these instruments, Cloudnet products  and AERONET products  are available. Because of radar scanning experiments during HOPE in Jülich, Cloudnet products which require vertically pointing measurements are sporadically unavailable for this campaign.

In modern multiwavelength lidars, a number of different receiving channels are installed to make use of as much information from the atmosphere as possible (elastic and Raman [inelastic] scattering, change in polarization state of emitted light due to scattering, etc.). In this way, high-quality aerosol products are obtained on a vertically resolved basis. However, because of the high background noise, Raman lidar observations during daytime are challenging. Therefore, for continuous (24/7) measurements, we concentrate on the use of channels for elastic backscattering, including depolarization. The key challenge to succeed with automated aerosol retrievals is the calibration of the lidar signals. There are two main tasks necessary before an automated aerosol target categorization can be performed: the calibration of the backscatter profiles and the calibration of the depolarization products.

For the typing of atmospheric features, i.e. the optical dominant scatterer type, three extensive (quasi-particle-backscatter coefficient at 532 and 1064 nm and volume depolarization ratio) and two intensive properties (quasi-particle Ångström exponent and quasi-particle depolarization ratio) are available to detect aerosol and cloud layers and to distinguish between those two and classify subtypes. The lidar-only attempt is made to categorize aerosols and clouds concerning different types in analogy to the Cloudnet classification. In the following, the methodology is described followed by an intensive discussion concerning the applicability by means of example cases of HOPE.

In this section, an overview of the aerosol conditions during entire HOPE is provided. The MWL PollyIfTXT was routinely operating at Krauthausen from 2 April 2013 to 31 May 2013. Thus, 2 full months of a spring season could be covered. An overview of the observations of the full campaign is given in the Appendix in Fig.  (April) and Fig.  (May) in terms of the quasi-particle-backscatter coefficients at 532 and 1064 nm (extensive properties), the quasi-particle Ångström exponent (532–1064 nm), and the quasi-particle depolarization ratio (intensive properties) as used for the categorization. As described in , the weather conditions during HOPE varied from periods with several warm and cold front passages interrupted by a few high-pressure systems with high-level cirrus clouds at the beginning of the campaign to more low-level convective cloud conditions later on.

Continuous MWL observations were available during the entire period with the exceptions of some short interruptions due to maintenance. During days of almost only precipitation (e.g. 16 May 2013), lidar observations are only sporadically available as the system stops measurements during precipitation events. Thus, calibrated lidar signals and the corresponding Ångström exponents and depolarization ratios are available for most of the time of favourable weather conditions and allow the typing of the particles according to the scheme described above.

The corresponding lidar target categorization for the entire HOPE campaign aiming on aerosol discrimination is shown in Fig.  together with the respective Cloudnet classification. The lidar target categorization reveals that aerosol was usually located from the ground up to 2 km height. Non-typed particles and low aerosol concentration were typically detected up to higher altitudes (4–5 km) showing that these regions are not appropriate for the Rayleigh fit procedure as already described above. Furthermore, it can be seen that the spring of 2013 at Krauthausen was dominated by low-level clouds and cirrus. Only on a few days clear sky conditions were observed. Comparing to the Cloudnet target categorization, it is confirmed that April and May was often dominated by deep clouds covering almost the whole troposphere. The lidar target categorization by definition only identifies the cloud bottoms in these cases, but this in good agreement with Cloudnet.

Interestingly, the intrusion of non-spherical particles was observed several times in the lowest 2 km until beginning of May (see lidar target categorization in Fig. ). We can only speculate that this might be local dust from open-pit mining, as intensively discussed for the 18 April case study, or pollen. After 10 May 2013, low-level clouds together with precipitation prevailed (see also Cloudnet target categorization), and thus it is reasonable that the local dust was too wet to be entrained into the air and/or the pollen season was over. These observations might be an interesting topic for future studies focusing on local aerosol emissions.

Furthermore, one sees that during HOPE the majority of the aerosol in the PBL was classified as small aerosol, as we would expect for an industrial and highly populated area. However, large aerosol was also observed occasionally, but mostly at the top of the PBL, indicating the importance of hygroscopic growth. Comparing again to Cloudnet, one sees that drizzle is often observed with radar while the lidar still detects aerosol. This interesting feature, discussed already for the presented case studies, was observed frequently and demonstrates the different sensitivity of the different instruments. Furthermore, it is found that Cloudnet does not detect as much aerosol with low concentrations due to the use of the ceilometer, which is not as powerful as the MWL.

To give an overview of the aerosol and also partly the cloud conditions during HOPE, a statistic of the classified scatterers for the entire troposphere for HOPE is shown in Fig. . Concerning typed aerosol (Fig. , top, left), the majority of the particles were classified as small aerosol (two-thirds). Large, spherical particles were observed 20 % of the time, while a mixture of non-spherical and spherical particles was observed in 9 % and large, non-spherical particles only in 3 % of the analysed pixels. As already discussed, these particles were mostly mixed from the ground into the atmosphere. Only on a few of the days, thin, lofted layers of Saharan dust were observed.

If one also takes into account the “non-typed particles/low concentration” class (Fig. , left, bottom) one sees that surprisingly 42 % of the particles are non-typed or of low concentration. But one has to take into account that this particle class can inherit every scatterer type (i.e., all types of clouds and aerosol) and that very low aerosol concentrations were almost always present above the PBL. Due to the conservative approach chosen, particles are only typed if enough information is available. Thus, often the 1064 nm channel, which is least sensitive for molecular scattering, detects particles while the other channels have a too-low SNR to be used for particle typing, which leads to a large number of pixels being classified as non-typed aerosol particles. However, the information given by this category is still very useful, as it makes clear that molecules are not the only ones that contribute to the light scattering, which is important when the target classification will be used for the determination of suitable calibration periods and regions with negligible aerosol scattering.

For the clouds identified during HOPE, a different picture was obtained (Fig. , top, right). Here, the “likely ice cloud” class is the dominant type, with 46 %. Due to the assumption made above aiming on aerosols (lidar ratio), the quasi-particle depolarization ratio is underestimated in cirrus and thus does not often exceed 35 %. Therefore, the clearly identified ice clouds make only a fraction of 6 %. However, we have to repeat that we do not aim at classifying clouds as we focus on aerosol, and the cloud information might be used only as a hint for the type of clouds for which further investigations are necessary. Water droplets are typed in 21 % of all cases and likely liquid clouds only three percent of the time. Non-typed clouds amount to 26 % of all cloud classes. We have to repeat that this cloud statistic is biased as the lidar can penetrate liquid clouds only by a few tens of metres. Above a detected liquid cloud, no typing is performed. In turn, the lidar can often penetrate cirrus clouds, and thus, in contrast to liquid clouds, ice crystals can also be detected well above the cloud base.

Altogether during the HOPE campaign, more than 1 million pixels in the troposphere of 30 m vertical and 5 min temporal resolution could be analysed. From these pixels, clean (i.e. molecular scattering dominating) atmosphere was observed in 29 %, clouds in only 7 %, aerosol in about 37 %, and “non-typed particles/low concentration” in 27 % of the analysed and feature-classified pixels (Fig. , bottom, right).

In this work, we have used absolutely calibrated lidar signals to categorize primary aerosol but also clouds in high temporal and spatial resolution. Two months of 24/7 observations from the multiwavelength-Raman-polarization lidar PollyIfTXT during the HOPE campaign have been analysed for that purpose. We have used the well-established Cloudnet framework to develop a lidar stand-alone classification. The Cloudnet equipment was operated continuously directly next to the lidar and has been used for comparison.

Automatically derived particle backscatter coefficient profiles  in low temporal resolution (30 min) have been used to calibrate the lidar signals. A daily mean lidar calibration parameter was derived with an accuracy better than 20 %. From these calibrated lidar signals, new atmospheric parameters in temporally high resolution (quasi-particle-backscatter coefficient) which require a priori information (assumptions) for attenuation correction have been developed. It was found that the newly developed procedure works well at 532 and 1064 nm, but deviations from the particle backscatter coefficients can be strong at 355 nm when the a priori information is not perfect. As a consequence for the particle typing, the quasi-particle coefficients at 532 and 1064 nm, its corresponding Ångström exponent, and the linear depolarization ratio at 532 nm are used for the classification.

By using thresholds obtained from multiyear, multisite EARLINET measurements, four aerosol classes (small; large, spherical; large, non-spherical; mixed, partly non-spherical) are defined. Thus, particles were classified by their physical features (shape and size) instead of by source as, for example, the well-known CALIPSO typing does. For source definition, additional information is needed, which has been out of the scope of this development, which has focused on a lidar stand-alone tool.

The bases of optical thick clouds (liquid droplets) can be identified using the Cloudnet approach applied to the MWL. Cirrus clouds/ice are identified by its highly depolarizing features. Furthermore, regions dominated by molecular scattering and regions of non-typed particles/low aerosol concentration are identified with the target categorization. The detection of molecular regions can be very useful for lidar calibration in the atmosphere.

By discussing three 24 h case studies, it was shown that the aerosol discrimination is very feasible and informative and gives a good complement to the Cloudnet target categorization. By analysing the entire HOPE campaign, almost 1 million pixel (5 min, 30 m) could be successfully classified with the newly developed tool from the 2-month data set. We found that the majority of the aerosol trapped in the PBL were small particles as expected for a heavily populated and industrialized area. Large, spherical aerosol was found mostly at the top of the PBL and close to cloud bases, indicating the importance of hygroscopic growth of the particles at high relative humidity. Interestingly, it was found that on several days non-spherical particles were mixed from the ground into the atmosphere. The origin of these particles remains unclear and needs further research. Lofted layers of Saharan dust as it is typical for spring in Germany were observed only sporadically and with low AOD during the investigated time frame of the HOPE campaign in spring 2013. Non-typed aerosol with low concentrations was found often above the PBL up to heights of about 4 km. Cloudnet was not able to identify these optically thin particle layers due to the lower sensitivity of the used ceilometer. The capability to detect cloud bases was compared to the Cloudnet feature mask, and the good agreement gives evidence that this feature could be used to apply robust cloud screening, which is often needed for lidar data retrievals, for example, for other automatic approaches such as the EARLINET Single Calculus Chain . Ice crystals were also often classified correctly but sometimes remained unclassified or were even falsely classified as aerosol as a consequence of multiple reasons (a priori information aiming at aerosol, low depolarizing characteristics in certain temperature ranges, etc.). This behaviour might be overcome when combining the lidar stand-alone target categorization with the Cloudnet target categorization as planned in ACTRIS-2

ACTRIS is the European Research Infrastructure for the observation of Aerosol, Clouds, and Trace gases: http://www.actris.eu/.

However, it is important to have a lidar stand-alone tool, as at the moment Cloudnet and automatic continuously running MWLs are operated only at three European stations, while stand-alone lidar systems are available at more than 25 EARLINET stations. We also consider the presented MWL approach for the classification of aerosol types as a prerequisite for the development of schemes for the identification of aerosol layers. Current retrievals, such as the STRAT algorithm , aim for providing aerosol layering information from lidar observations at one wavelength and can thus only identify a single layer even though it would actually consist of several layers of different types, such as smoke or dust. With this development, the integration of EARLINET and Cloudnet is ongoing and offers a high potential for future synergistic profiling of aerosols, clouds, and their interaction by combining modern state-of-the-art atmospheric instruments.