Atmospheric profiling based on aerosol and doppler lidar

Resumen

The observation and study of the Earth’s atmosphere have become of increasing concern because of the interest on weather prediction, air quality and climate change. Among all the atmospheric components and processes related to those topics, atmospheric aerosol particles (defined as solid and/or liquid particles suspended on the atmospheric air, excluding clouds) and the dynamic and turbulent properties of the atmospheric boundary layer (ABL, the lowermost part of the atmosphere, that is directly influenced by the Earth’s surface) represent two of the most active research fields due to the lack of knowledge about the uncertainties of their effects. Lidar (Light detection and ranging) is a key technique in atmospheric research because it provides atmospheric information with high spatial and temporal resolution. In this context, this thesis is the result of an effort of compiling, understanding and applying some of the most recent lidar techniques in the field of atmospheric profiling. The work done to this end includes instrumental set up, calibration and improvement, regular measurements and several field campaigns, algorithm development, knowledge of several specific and complex software and application to the analysis and interpretation of a variety of atmospheric situations. The experimental work is based on the use of two different lidar systems, namely a Raman lidar and a Doppler lidar. Some ancillary tools have also been used, as co-located ceilometer and photometers, simultaneous satellite-based lidar and HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) backward-trajectory model. The instrumentation is located at the UGR station in the city of Granada (southeastern Spain, 680 m above sea level, a.s.l.) and in Cerro Poyos station (1830 m a.s.l.) as ancillary mountain station at close distance for certain studies. They are part of the experimental observatory AGORA (Andalusian Global Observatory of the Atmosphere), included in ACTRIS (Aerosol, Clouds and Trace Gases Research Infrastructure) that is in the way to be constituted as a permanent European Research Infrastructure Consortium (ERIC). An important part of the measurements were taken in the framework of SLOPE I (Sierra Nevada Lidar Aerosol Profiling Experiment I) campaign. This thesis includes lidar measurements from two additional rural sites, an olive orchard in Úbeda (Spain) and a peatland in Rzecin (Poland), gathered respectively during AMAPOLA (Atmospheric Monitoring of Aerosol Particle Fluxes in Olive Orchard) and POLIMOS-2018 (Polish Radar and Lidar Mobile Observation System 2018) field campaigns. The multiwavelength Raman lidar system MULHACEN is used to retrieve vertical profiles of aerosol particle optical and microphysical properties. It emits pulsed laser radiation at 355, 532 and 1064 nm wavelengths and collects the elastic and inelastic backward-scattered radiation, also with depolarization capabilities. This spectral and polarization information is useful for retrieving size and shape properties of the aerosol particles and as an indicator of the aerosol type. The system is part of EARLINET (European Aerosol Research Lidar Network) in the frame of ACTRIS activities. The Raman radiation collected by MULHACEN corresponds to changes in the vibrational energy states of molecules, a widely used effect in aerosol Raman lidars although it has the disadvantage of a low Signal-to-Noise Ratio (SNR) that usually limits the retrievals to nocturnal measurements with 30-60 min time resolution. We have implemented a new setup in the UV of MULHACEN system in order to measure Raman lines corresponding to rotational energy states, enhancing the measured signal and diminishing the wavelength shift between elastic and Raman radiation (a key point for diminishing uncertainties due to spectral depencence). The rotational lines have some issues related to temperature dependence, but we demonstrate that, with an appropriate filter wavelength selection, this accounts for an additional uncertainty of less than 4 % on the retrieved aerosol optical properties. With this new setup, we have been able to retrieve aerosol extinction and backscatter coefficients profiles with 1-h time resolution during daytime and up to 1-min time resolution during nighttime. Nevertheless, this study has an exploratory nature within this thesis and the database used in the rest of the sections corresponds to the extended vibrational Raman measurements previoulsy obtained. The Doppler lidar system Stream Line emits pulsed infrared radiation at 1500 nm and measures the Doppler frequency shift in the backscattered radiation due to the aerosol particles movement with wind. This instrument is used for the first time at UGR station, and we have set it up to retrieve vertical profiles of wind field within the ABL and its turbulent properties. Thanks to this, the instrument has become part of Cloudnet in the frame of ACTRIS activities. The signal measured by the Doppler lidar is optimized for directly obtaining radial velocities, but the the signal intensity presents some artifacts that have to be corrected if this signal is to be used for further purposes. We have applied a correction algorithm developed at the Finnish Meteorological Institute (FMI), in the frame of ACTRIS, to correct for two kind of artifacts affecting background substraction, specifically certain range-independent step changes with time and time-independent residual structures at different altitudes. A third issue is the signal magnification around certain range and reduction at higher ranges due to the instrument optical system focal length, set at that range. We have proposed in collaboration with FMI a methodology to calculate experimental focal length and lens diameter as calibration parameters for our Doppler lidar. It consists of an iterative method based on comparing the Doppler lidar corrected signal with a co-located ceilometer (that needs no focus correction), and we have obtained the calibration parameters for our system with less than 20 % standard deviation. The corrected signal allows for estimating more advanced quantities as velocity errors and attenuated backscatter. The vertical profiles of aerosol optical and microphysical properties have been retrieved with several inversion algorithms from Raman lidar measurements. The starting point is the retrieval of particle bakcscatter and extinction coefficients at three wavelengths, and their derived intensive particle properties, by using Klett-Fernald method or Raman methods for elastic or inelastic signals, respectively, and depolarization method for polarization measurements. Our work is then focused on employing the set of optical properties to obtain more complex particle optical and microphysical properties (i.e., particle volume concentration, effective radius, complex refractive index and single scattering albedo) using a regularization algorithm. In particular, two different software tools developed at the University of Potsdam (Germany) have been used. The first software, called here UP, is built upon Mie model for spherical particles and can be used for forward or inverse calculations using simulated or measured inputs. The inversion algorithm is a hybrid regularization method based on explicitly solving the mathematical equations that relate the particle microphysical and optical properties according to the spherical model, obtaining the particle size distribution as output. The second software is called SphInX (Spheroidal Inversion Experiments) and is based on an extension of Mie model in two dimensions in order to account for non-spherical particles by assuming an ensemble of spheroids characterized by their volume-equivalent radius and aspect ratio. The regularization methods are similar to the one in UP, but SphInX uses a precalculated database for the calculation of the kernel equations for certain cases, due to the higher complexity of the model. An additional difference between the two tools is that depolarization measurements are only input for SphInX software, because UP assumes no depolarization by spherical particles. We have applied UP software for the characterization of long-range transported biomass burning particles from strong plumes that were detected during July 2013 at Granada and two more ACTRIS-EARLINET stations, namely Leipzig (Germany) and Warsaw (Poland). A deep analysis on the sources and transport paths of the particles has been performed using satellite observations and modeling tools, confirming the arrival of the smoke plumes to the studied stations several days after being emitted from forest fires in North America. The observed optical and microphysical properties correspond then to aged smoke particles and reveal their small and mostly spherical shape with weak absorption. SphInX software has also been used for the study of different types of transported aerosol particles. A first case with biomass burning particles has been selected in order to compare with UP software and assess for the impact of the additional depolarization measurement and the extended 2D model for particles. The second selected case corresponds to mineral dust particles from Sahara desert as well-known large and non-spherical aerosol type. The limited size range of the precalculated SphInX database due to lidar measured wavelengths has limited the application of this software to the submicrometric and micrometric part of the distribution, but the rest of the microphysical properties agree with the literature, with the added value of detailed information on particle shape according to spheroidal model. The minimum requirement to retrieve particle properties from multiwavelength Raman lidar systems with UP or SphInX software tools is to have 3 particle backscatter coefficient (at 355, 532 and 1064 nm) profiles and 2 particle extinction coefficient (at 355 and 532 nm) profiles (plus a particle depolarization profile in case of SphInX retrievals). Since this is not always the case, but global aerosol networks need as many and complete measurements as possible, we have proposed a methodology for systems with one missing channel. It is based on using spectral measurements from co-located star- or lunar-photometer to reproduce the missing profile, using Angström equation. This methodology can be applied using both UP and SphInX tools, although the results here refer only to UP. The proposed methodology has been tested with a sensitivity study in order to assess for the additional errors introduced in the microphysical retrievals. We have used UP software in forward mode to simulate different aerosol cases in terms of size distributions and complex refractive indices, and the suggested method has been applied for three different scenarios (according to the lidar channel that is missing). The same scenarios have been used for applying the methodology to real data measured with MULHACEN lidar at UGR station. All the results from simulated and real cases have been compared with the retrievals using the complete lidar setup without photometer information. Maximum deviations of 20 % have been found for simulated cases with an input error of 5 %, and deviations less than 40 % for real cases with input errors up to 40 %. The vertical profiles of wind field and turbulent properties within the ABL have been retrieved from vertical and scanning Doppler lidar measurements using several linked algorithms. We have used a software processing chain, the Halo lidar toolbox developed at the FMI, that includes the most recent methodologies to calculate vertical profiles of horizontal wind field and turbulent properties (namely vertical velocity statistical moments, wind shear and turbulent kinetic energy dissipation rate). The chain culminates with the combination of all the retrieved information to create a classification mask of the turbulence sources with temporal and vertical resolution within the ABL. We have used the horizontal wind product of the Halo lidar toolbox to carry out a statistical study over Granada using a two-year database of regular measurements. The data availability in terms of maximum analyzed altitudes for statistically significant results is limited to around 1000-1500 m above ground level (a.g.l.) due to the decreasing signal intensity with height that also depends on aerosol load. We have analyzed the differences and similarities in the diurnal evolution of the vertical wind profiles for different seasons, and diurnal and nocturnal wind roses have also been calculated for the whole dataset at three altitude intervals. Finally, we have studied the turbulent properties of the ABL with the corresponding Halo lidar toolbox products. We have evaluated the diurnal development of the ABL for two days from AMAPOLA campaign, one with clear-sky conditions and the other with clouds, observing their effect on the lower altitude reached by the ABL and on the top-down turbulent movements. For Granada and Rzecin sites, where the available databases were of two years and four months, respectively, we have been able to perform a statistical analysis of the main turbulent sources within those analyzed periods, with temporal (for each hour of the day) and height resolution. A seasonal distinction has been done only for Granada. Both sites show a clear convective activity during daytime at altitudes increasing with time, and a significant wind-shear driven turbulence during nighttime, in special in Rzecin. We have also found at both sites an important contribution of nocturnal turbulence with unknown sources (labelled as ‘intermittent’) although in Rzecin it is mostly concentrated around 600 m a.g.l., while in Granada it is present at all heights.