The knowledge of the raindrop size distribution is key for characterizing precipitation. It is however still a challenge to retrieve it with radars. Several polarimetric and spectral techniques are proposed for cm-wavelength radars (weather radars). What about the mm-wavelength radars (cloud radars), which have a better spatial and time resolution and can still measure light and moderate rain? Knowing that 90% of the rain volume in Europe comes from rainfall rates between 0.1 mm/h and 10 mm/h, this is worthwhile to investigate. The goal of this thesis is to retrieve 1 of the 3 parameters of the modelled gamma raindrop size distribution, the median volume diameter (D₀), during stratiform rainfall events using a slantwise profiling dual-frequency polarimetric cloud radar. Focus is given to phase measurements, which are not affected by attenuation. Simulations show that the differential backscatter phase (δ_co) strongly depends on D₀. At mm-wavelength, backscattering and propagation effects need to be disentangled first. To achieve this, an algorithm to detect and characterize Rayleigh plateaus is proposed and implemented. After the application of this algorithm, a methodology to estimate the differential backscatter phase and its error is given. The 95% confidence interval of δ_co is estimated with the re-sampling method bootstrapping. 
Using simulation results, an attempt is made to find combinations of D₀ and the raindrop size distribution shape parameter μ that match with the confidence interval of δ_co. The confidence interval of δ_co restricts D₀, but not μ in most cases. This proposed technique is applied for both the 35 and 94 GHz frequency band of the new cloud radar at Cabauw (Ruisdael Observatory site near Utrecht). The resulting 95% confidence intervals of D₀ with 35 and 94 GHz and their overlap are compared with in-situ disdrometer measurements of the mass-weighted mean diameter (D_m) which is closely related to D₀. The median volume diameter retrieved with the 35 and 94 GHz frequency bands both shows a normalized cross correlation coefficient of 0.845 with the measured D_m of the disdrometer. Therefore, the cloud radar seems to have the capability to provide the detailed variations of the raindrops mean/median diameter like a local disdrometer, but at different heights. Nonetheless, the values differ. The disdrometer provides higher values than the cloud radar. One possible explanation is the inability of the disdrometer to measure raindrops smaller than 0.25 mm and the expected underestimation of the number of raindrops with sizes between 0.25 and 0.375 mm. However, because D₀ values retrieved from 35 GHz data are also higher than the ones at 94 GHz, further research, which can use all the methodologies proposed in this master thesis work, is needed to examine the quantitative values of the median volume diameter retrieval. 
These techniques can be implemented for all the single-frequency cloud radars (94 GHz) of the national Ruisdael Observatory (cloud and precipitation profiling mobile station and Lutjewad site above Groningen).

Airport operations are highly dependent on safe weather conditions for their daily operations. Especially the visibility conditions are crucial. Low Visibility Procedures (LVP phase M, A, B and C) are defined at Schiphol Airport, resulting in capacity and runway use restrictions. These Low Visibility Pro­cedures are based on a post­processing model of a Numerical Weather Prediction (NWP): the Terminal Aerodrome Forecast Guidance (TAFG). This model provides a probability for the occurrence of a given LVP phase, based on thresholds of the Runway Visual Range (visibility including back luminescence at airports) and the ceiling (cloud base height). Currently, the TAFG is operational for three NWP models (HARMONIE, HIRLAM and ECMWF). This research found that the HIRLAM TAFG has a better performance than the HARMONIE TAFG. This result is caused by the larger training data­set that is used for the HIRLAM TAFG (20 years) compared the training data­set of for the HARMONIE TAFG (3 years). The ECMWF TAFG has a better performance for the severe LVP phases B and C. This is remarkable, since ECMWF is not often used for airport operations. However, the standard error of the performance indicator of the ECMWF TAFG is higher compared to the other models and, therefore, a larger data­set should be studied to confirm these findings.

Radar rainfall nowcasting stands for the prediction of rainfall amounts and intensities over the next 6 hours by means of statistical extrapolation of radar measurements. It is the principal ingredient for modern flood forecasting and early warning systems. Radar forecasts are generated by identifying and tracking rainfall cells in radar images and extrapolating them along the main direction of motion based on the assumption of Lagrangian persistence. The latter means that the rainfall cells do not physically evolve over time. Therefore the key to improving predictions is to anticipate the growth and decay of rainfall cells during the next few hours. Studies have shown that estimating growth and decay based on past radar images is nearly impossible. Therefore, in this thesis, a new approach for predicting growth and decay based on physical guidance from a numerical weather prediction model is investigated. The idea is that numerical weather prediction models are better at anticipating changes in the atmosphere than radar and thus, should also contain information about the growth and decay of rain cells. To test this hypothesis, two simple machine learning models have been trained to learn the relationship between the observed growth and decay in radar images and the output parameters from the Dutch mesoscale numerical weather prediction model HARMONIE. The models were trained on data in the summer of 2019 around the city of Rotterdam for two area sizes, $20\times 20$ km$^2$ and $60\times 60$ km$^2$. The first model is static and assumes that all predictions from HARMONIE are correct. Unfortunately, this turns out to be too optimistic as HARMONIE often places the rain cells at the wrong locations with the wrong intensities and trends. As a result, no meaningful relationship between HARMONIE outputs and growth/decay in radar could be learned. The dynamic model overcomes this issue with the help of an additional classifier. The classifier predicts whether the information from HARMONIE can be trusted or not. Then, a regression model predicts the magnitude of growth and decay (only for the trusted cases). Out of 308 analysed cases, the classifier labelled 237 HARMONIE predictions as being untrustworthy, thereby removing a lot of bad cases. However, 98 of these were false negatives, meaning that they could have been used to predict growth and decay. On the other hand, only 71 cases were labelled as containing useful information, from which 27 cases were false positives. Despite these errors, the dynamic model was able to improve the root mean square error on the predicted growth/decay by $27$\% for these 71 cases. Also, the correlation between the predicted and actual growth/decay increased from near zero to $0.488$. This is encouraging and clearly highlights the potential of this approach. Still, some important challenges remain. In particular, results show that the dynamic model often underestimates the magnitude of growth and decay. Also, when the entire validation data set is considered, including all unusable cases, the improvement with respect to the static model is only $6$\%. This is mainly caused by the poor performance of HARMONIE and the large number of unusable cases. However, performance could be improved further by building a better classifier capable of identifying all good cases with a low number of false negatives.

Stratiform precipitation is one of the most important precipitation systems in the mid-latitudes. To gain understanding of the melting layer, which is part of a stratiform precipitating system, an algorithm which is able to select melting layer data from large datasets is required. The goal of this thesis work is to create and deliver a program which is able to performthis work. For development the available data was separated into three groups: no melting layer, only melting layer and the rest. Based on the second group, typical melting layer signatures are analysed. In particular the different signatures (reflectivity, polarimetric and Doppler) occur at different heights. Based on the knowledge gained from the analysis and focussing on the reflectivity and polarimetric signatures, three different approaches were taken to detect and characterise the melting layer fromthe data. The first approach is based on an existing method in literature, this approach acts as the reference method. The second approach is based on image processing techniques, while the third approach is based on machine learning techniques. The second approach is later abandoned because of limitations of the techniques investigated. The third approach, machine learning, is the main contribution of this thesis work. For analysis of the performance of the different algorithms an annotated test dataset is created which represents the entire dataset. The performance is determined by the melting layer detection probability using a confusion matrix and by the determination of errors of the melting layer boundaries. The reference method proved to show an extremely low false positive rate (0%) on the test dataset. This means that if the method detected a melting layer, it is almost certain that there is one. The overall detection probability was 80%. The method fails in detecting the melting layer when the peak reflectivity is below the threshold (30dBZ) used in the method. The detected upper melting layer boundary of the reference method is on average 365 m lower compared to the ground truth. The lower boundary is on average 149 m higher than the ground truth. This means that the reference method only selects a part of the entire thickness of the ML. The correlation between the lower boundary and the ground truth is higher than the upper boundary (0.987 vs. 0.965). The proposed machine learning method has a detection performance of almost 94%, which is higher than the reference method, but some false positives occur (3%). The upper boundary is on average 20 m above the ground truth, while the lower boundary is 66 m lower than the ground truth. The machine learned method thus overestimates the thickness of the ML. The correlation between the lower boundary and the true boundary is 0.976 and the correlation between the upper boundary and the true boundary is 0.966. The upper boundaries of both methods have a very similar correlation, while the lower boundary of the machine learning method has a lower correlation. However, despite the lower correlation of the lower boundary and the introduction of false positives, the machine learning method is an improvement over the reference method since it has a significantly higher detection probability and it captures the entire thickness of theML much better. It is therefore suited forML analysis.

In the context of climate modelling, convective clouds in tropical regions play a major role in determining the climate sensitivity. The vertical transport of mass and energy associated with this type of clouds is often represented with so called mass-flux parameterization schemes. In this work the aim is to evaluate the relationship between mass-flux and large scale environmental conditions using observations in a tropical region, over a period of 13 wet seasons. A uniquely comprehensive data set from the C-band polarimetric radar (CPOL) in Darwin, Australia, is used to estimate vertical velocity inside precipitating convective clouds. Ultimately, mass-flux is derived over a domain size similar to that of a general circulation model (GCM) grid box. Five parameters (RH_500, CAPE, CIN, ­Omega_500 and Chi_crit) are selected to describe environmental conditions and with these, the magnitude and shape of mass-flux is analysed. Chi_crit appears to be the most valid parameter to represent both the shape and magnitude of mass-flux. All other selected parameters strongly influence only one of the two aspects of the profile. Additionally, fractional entrainment is retrieved from mass-flux profiles and partitioned into two terms: one dependent on area fraction and the other on vertical velocity. Under all environmental conditions, the layer between 4.5 and 7 km experiences detrainment. It can be inferred that a stable layer, known as the freezing level, is present at 4.5 km. Below the freezing level, the vertical velocity and the mass-flux shape are most relevant to determine entrainment rates leading to the conclusion that the vertical velocity should not be disregarded when parameterizing convection.

Accurate predictions of the extinction and scattering properties of the atmosphere are important for climate research and interpreting satellite data. This study introduces a model (called the H-model) that calculates the scattering coefficients and scattering enhancement factors based on in situ measurements of the dried ambient aerosol. A disadvantage of using dried aerosol measurements is that they do not correspond with the ambient conditions, as they are measured at a relative humidity below 40% and thus the particles are assumed to contain no water. Measurements of aerosol chemical composition do not contain water mass concentrations and measurements of the particle size distribution do not include water. To solve this problem, the H-model uses ISORROPIA, a thermodynamic equilibrium model, to estimate the expected amount of aerosol water content and growth factor g(RH) of aerosol particles for any given temperature and relative humidity (RH). With this information, the conversion between dry and enhanced relative humidity can be made. The chemical composition measurements can be complemented with the estimated aerosol water concentrations and the particle size distribution can be recalculated based on the growth factor for any given RH. In addition, the growth factor is also calculated by using k-Köhler theory and compared to the results of ISORROPIA. The findings of this sub-study show that the growth factors calculated by both approaches (ISORROPIA and k-Köhler theory) are similar as they significantly correlate. ISORROPIA, however, is more sensitive to small chemical changes which makes it more appropriate for the H-model. The calculated growth factors are used in the H-model to estimate changes in the chemical composition and particle size distribution of the aerosol particles at enhanced relative humidity. Subsequently, the H-model uses MIE theory to estimate the scattering properties of the particles at a specific relative humidity. By doing so, the scattering properties can be calculated at dry and enhanced RH, making it possible to calculate scattering enhancement factors. Finally, the H-model is validated by comparing the calculated scattering properties to measured scattering properties of a (humidified) nephelometer. To do so, in situ measurements from the CINDI campaign in 2009 and the TROLIX campaign in 2019 at Cabauw are used. The findings of this validation show that the results from the H-model do not yet accurately match the measurements. That being said, a strong correlation is observed between the calculated and the measured scattering properties. This shows that the H-model is able to capture changes in the particle size distribution and chemical composition while calculating the enhancement factors. It can be concluded that the results from the H-model are promising but need further work to close the gap between the calculations and measurements. The H-model makes multiple simplifications and assumptions which could be improved upon, thereby increasing the precision of the results as well. Furthermore, to fully conclude the findings of this study, the measurements of the SMPS and the nephelometers should be calibrated. A better statement can then be made about the accuracy of the comparison between the scattering properties calculated by the H-model and measured by the nephelometers.

Lidar systems can provide vertically resolved measurements of the physical and optical properties of the atmospheric particles with high spatiotemporal resolution. In this study, four case studies were analyzed and studied in order to identify the dominant aerosol type over the city of Nicosia in Cyprus: 05 April and 07 April 2015 and 17 April and 21 April 2016. More specifically, vertical profiles of the extinction and backscatter coefficient, Ångström exponent, particle linear depolarization ratio and lidar ratio were manually retrieved from the lidar signal for each case study. For the first case, the main aerosol load was observed in the very lower atmosphere, between 0.25 and 1.25 km. The observed aerosol optical properties indicated the strong presence of maritime aerosols. In the second case study, the aerosol layer was thicker and it was observed between 1.6 and 7 km. Analysis of the optical properties showed that the predominant aerosol types within the PBL were mainly maritime aerosols and aerosols from local sources while higher in the atmosphere the aerosol load consisted of dust mixtures. The third case study, on 17th April 2016, was a relatively clean day and a thin aerosol layer was observed between 0.25 and 2.0 km consisting mainly of maritime and urban aerosols. 21st of April 2016, which was the fourth case study, was a dust event case study. The aerosol load, accumulated between 2.0 and 5.0 km, was characterized as a purely dust layer. The results obtained with the manual retrieval method, were compared with automatically retrieved profiles (provided by the Leibniz Institute for Tropospheric Research) and are in good agreement. More specifically, for the vertically resolved extinction profiles at 532 nm the correlation coefficient values ranged between 0.9784 and 1. The correlation coefficients for the backscatter profiles at 532 nm and 1064 nm ranged from 0.9975 to 0.9986 and from 0.9937 to 0.9996 respectively. The linear particle depolarization ratio profiles also correlated less well with the automatically retrieved one with R ranging from 0.8781 to 0.9889 while similarly, for the volume depolarization ratio profiles the correlation coefficient ranged between 0.7054 to 0.9603.

Wind is an important indicator of circulation related processes in the atmosphere. Accurate wind in- formation is used as an input for weather and circulation models. Wind data with a high temporal and spatial resolution are useful for research on the microphysics of the atmosphere, which is the area of application for the three-beam Transportable Atmospheric Radar (TARA) located at Cabauw Ex- perimental Site for Atmospheric Research (CESAR) in the Netherlands. Although some comparisons were performed using radiosondes, no thorough quality assessment of the wind estimation results has been done so far. During this research, the quality assessment is performed for the estimation of the wind speed and direction during the Analysis of the Composition of Clouds with Extended Polarization Techniques (ACCEPT) campaign, which took place in October and November 2014. This research shows that the wind estimation is working well during precipitation and a TARA-based criteria for the data selection to guarantee the quality of the wind retrievals is identified in the form of the coefficient of variation for the mean Doppler velocities.
The quality assessment of the results of the TARA wind estimation is done by comparing the wind retrievals to the measurements of the meteorological tower at a height of 200 m above the surface at minute resolution. For the assessment to be performed, an effective way of removing clear air measurements is needed, which is done for the whole ACCEPT-campaign using a rough selection of time steps including rain. This approach showed that the wind estimation is working well during precipitation and shows that an improved data selection is needed.
Several approaches towards improving the data selection for the wind estimation of TARA are per- formed. The most successful is the use of the coefficient of variation, which is defined as the standard deviation divided by the mean. This coefficient is calculated for the mean Doppler velocities, which are the input for the wind estimation algorithm. Comparing the results of the coefficient of variation method to the one based on rain selection shows that both return good results. This leads to the conclusion that the coefficient of variation is useful to improve the data selection.