Master theses topics

External Projects


Topics for master thesis projects in the experimental group on radiation and remote sensing

Ground supersites and weather models: Cloud climatology in model and reality for three locations in Germany

Cloudnet is a network of ground stations for the observation of clouds set up across Europe. The standard Cloudnet suite of instruments comprises a cloud radar, a ceilometer, and a microwave radiometer to provide a cloud observation supersite. Using these instruments macro- and microphysical cloud parameters like cloud vertical distribution, ice and liquid water and rain differentiation, or liquid and ice water path/distribution are derived continuously.

Details of topography are a challenge for each weather model. The weather model version closest to real measurements should be a model reanalysis, tying model physics to a variety of observations. For this project the systematic climatological comparison of the Cloudnet products to the optimum model cloud representation in weather model reanalysis is targeted. It will be analyses with respect to the three different Cloudnet sites' data sets with increasing topographic contrasts: 1) Jüllich, a low level site at the south-eastern edge of the Beneleux/Northern German lowlands - 2) Munich at 500 m height in the pre-alpine foothills - and 3) Mount Zugspitze/ UFS Schneefernerhaus at about 2600 m height. Reanalysis data on different spatial resolution will be evaluated: ECMWF ERA interim on 80 km resolution obviously will not be able to represent cloudiness in mountainous terrain. In comparison COSMO-REA6 on 6 km and COSMO-REA2 on 2 km resolution is much more likely to reproduce large parts of the real measured cloud climatology.

Data and Knowhow: ECMWF ERA interim publicly available, COSMO-REA6 and 2 available at Uni Köln. Cloudnet data publicly available.


The thesis is jointly offered by Universität Köln and LMU. Responsible Tobias Zinner

Ground supersites and weather models: High resolution modeling in mountainous terrain

Details of topography are a challenge for each weather model. On standard weather model resolution of several kilometers, topographic features like the Zugspitze range and the nearby Inn valley are smoothed out. Atmospheric flow, and related cloud development, will not have to follow the strong vertical gradients of reality. This will lead to deviations of forecasted flow and weather phenomena. On high spatial resolution mountain ranges are represented much more realistically. In this project the ability of two model setups to represent weather, cloudiness and flow in mountainous terrain will be evaluated. Setup 1 is a standard numerical weather model setup at xxx km spatial resolution, setup is a specialized high resolution LES setup run nested in a weather model.

Cloudnet is a network of ground stations for the observation of clouds set up across Europe. The standard Cloudnet suite of instruments comprises a cloud radar, a ceilometer, and a microwave radiometer. Using these instruments macro- and microphysical cloud parameters like cloud vertical distribution, ice and liquid water and rain differentiation, or liquid and ice water path/distribution are derived continuously. For this study a comparison of the Cloudnet products to specific weather simulations at different spatial resolution is targeted. Two different Cloudnet sites' present different challenges with respect to topography. 1) Munich at 500 m height in the pre-alpine foothills - and 2) Mount Zugspitze/ UFS Schneefernerhaus at about 2600 m height.

Data and Knowhow: WRF and ICON-LEM modeling experience available at Uni Köln (ICON-LEM) and LMU (WRF). Cloudnet data publicly available.


The thesis is jointly offered by Universität Köln and LMU. Responsible Tobias Zinner

Air chemistry and aerosol modelling: Estimating and monitoring emissions of air pollutants using Lagragian modeling and inverse methods

Man-made air pollution is driven by emissions of trace gases and particles from traffic, industry, agriculture and households. But who is responsible for which fraction of the observed pollution? How accurate are current emission estimates for certain polluter categories? Can we monitor their emissions in quasi-realtime using observations and modelling? These are the science questions you would answer in your thesis.

Linking sources to observed air pollutant concentrations is a challenging task, and various methods exist to determine these contributions. One way would be to look at the amount of fuel sold, combine it with traffic counts and estimates of the emissions per kilometer driven to arrive at a map of traffic emissions, for example. This emission inventory is „bottom-up“, created through the combination of all known sources. However, there are a number of shortcomings with this approach - unknown sources, uncertain activity counts, incorrect emission estimates.

In this thesis you would pursue a different road and create a „top-down“ emission inventory. Instead of summing contributions from individual sources, you would use observations of pollution levels made by research aircraft during recent field campaigns and combine these with backtrajectory calculations to connect observations with source regions. Bayesian inversion, a statistical optimization procedure would then lead to an optimized emission inventory based on observations. Such an inventory is highly valuable for emissions monitoring and air pollution control, aids in estimating the quality of existing bottom-up inventories and will help in identifying the most active contributors to bad air quality.

As you can see from the topic description a fair amount of modelling / computer work is required, so I would expect you to have used a Linux computer before and are somewhat familiar with Python or a similar programming language.

Responsible Christoph Knote

Air chemistry and aerosol modelling: Development of a dust resuspension module in a chemistry-transport-model

Fine particulate matter („Feinstaub“) is a major air quality hazard, especially in urban settings like Munich. Numerical models have been developed to understand the sources and transformation of particles in the atmosphere, and to develop efficient mitigation strategies. These regional or global scale models are based on numerical weather prediction models (like WRF or COSMO) and were extended by a description of particulate matter and trace gases. Particulate matter, its size distribution and chemical composition, is a prognostic quantity in such models they hence describe its complete lifecycle from emission (think Diesel exhaust) to deposition on surfaces (the dirt on your window sill). Currently those models do not consider that the dust deposited might be resuspended (e.g. through traffic). In this work you would develop a module to consider the resuspension of dust depending on factors like wind speed, relative humidity, precipitation and location which could then be implemented in chemistry-transport-models like WRF-Chem. Depending on your interests you could evaluate your developments on the regional to continental scalem but also on the local scale to estimate the effect of this paramterization on curbside measurements of particulate matter concentrations.

Responsible: Christoph Knote

Air chemistry and aerosol modelling: Air quality impacts of wild fires

Wildfires are often responsible for episodes of severe air pollution and elevated levels of particulate matter ('Feinstaub') in diverse areas of the world - Southern Europe, North America, or Russia, to just name a few. What makes the burning of biomass burning so extraordinarily bad for air quality? Is the chemistry in biomass burning plumes responsible for high ozone episodes downwind? In the context of an upcoming field campaign you would use an existing Lagrangian trajectory modeling system to predict the dispersion of wild fire plumes in North America, and develop a method to estimate the emission source strength of these wild fires based on observations taken from research aircraft, ground observations and mobile sensors. Your work would require familiarity (or lack of fear to work) with Linux computers and reasonable programming skills, and interest in environmental modelling aspects.

Responsible: Christoph Knote

Radiation and weather modelling: Synthetic satellite images for improved aerosol forecasts

Some of today’s most important environmental concerns are related to the presence of aerosols in the atmosphere. Dust, sand, smoke and volcanic aerosols affect the safe operation of transport systems and the availability of power from solar generation and influence the formation of clouds and rainfall. Moreover, not only the increasing concentration of greenhouse gases but also changes in the aerosols affect climate change and the extent of the aerosol impact is still uncertain. Numerical weather prediction models have made significant progress in accurately predicting the transport of aerosols, changes in their composition, and their effects on the dynamics of the atmosphere. However, large uncertainties exist in the initial and boundary conditions. Often neither the initial aerosol distribution nor the strength of aerosol sources is well known. These problems can be addressed by assimilating more and better aerosol observations. Aerosol information with high spatial and temporal resolution and excellent coverage is contained in solar channel geostationary satellite images.

Only recently radiative transfer (RT) methods for the visible spectrum have been developed by our group that are sufficiently fast to use solar channels in operational data assimilation systems. So far, information about water and ice clouds has been be assimilated. An important challenge in extending these methods to include aerosols is that considerably more parameters are required to describe them, compared to clouds. To make a look-up table based RT method feasible for aerosols, the number of parameters has to be reduced. The aim of this master thesis is therefore to compute synthetic satellite images from model data including aerosols and to investigate the following questions:

These investigations will be an important step towards improved aerosol forecasts and a better understanding of the influence of aerosols on weather and climate.

Responsible Leonhard Scheck, Bernhard Mayer

Radiation and weather modelling: Analysis of a Temporal / Spectral Integration Scheme for Radiative Transfer Calculations in Numerical Weather Prediction Models

Due to the computational complexity of radiative transfer computations, atmospheric models usually try to minimize the computational effort for this important boundary condition, the only energy source in the physical system. They sub-step the radiation module, i.e. call the radiation routines only every n-th timestep. A too coarse temporal sampling strategy, may however have implications for the evolution of clouds and the model dynamics in general. Preliminary results for a parameterization of the temporal sampling have shown promising results. The parameterization uses a simple regression model, fitting the error growth in heating rates and radiative fluxes to estimate the next point in time where radiative transfer has to be computed.

The thesis involves the testing and further development of this adaptive integration scheme and the assessment of its applicability and limitations for various model types, ranging from high-resolution LES to Global Circulation Models.

Responsible Fabian Jakub, Bernhard Mayer

Radiation and weather modelling: Hat Strahlung einen Einfluss auf konvektive Bewoelkung?

Die Antwort darauf ist eine klares „ja“: Konvention wird durch Absorption solarer Strahlung am Erdboden ausgeloest. Die Frage ist, welche Relevanz die einzelnen Prozesse (Absorption und Emission solarer und thermischer Strahlung am Boden und in der Atmosphaere) haben. Und diese Frage versuchen wir, in verschiedenen Projekten im Verbund mit externen Partnern zu beantworten, um mittelfristig die Vorhersage von Wetter und Klima zu verbessern. In dieser Arbeit sollen dazu Simulationen konvektiver Bewoelkung mit dem an der Uni installierten ICON-LES durchgefuehrt werden. ICON wird in Zukunft zur Basis der DWD Wettermodellierung. Dazu muessen die im Rahmen von zwei Doktorarbeiten (Klinger, Jakub) neu konzipierten Strahlungsroutinen auf das Dreiecksgitter des ICON-LES angepasst und getestet werden. Dies ist eine anspruchsvolle und sehr interessante Aufgabe, die unter anderem die Entwicklung eines statistischen Monte-Carlo-Modells auf einem Dreiecksgitter sowie einer einfachen deterministischen Loesung des thermischen Strahlungstransports umfasst. Die zugrunde liegenden Methoden sind bei uns etabliert, aber eben „nur“ auf einem Rechtecksgitter. Die Arbeit ist zentrales Thema des Lehrstuhls fuer Experimentelle Meteorologie und wuerde die Arbeitsgruppe massgeblich voranbringen.

Responsible Bernhard Mayer, Carolin Klinger, Fabian Jakub

Radiation and weather modelling: Radiative transfer in next-generation weather models

The next-generation German weather model ICON is planned to be run on 100m-resolution. On this scale 3D-radiative transfer influences cloud formation. Current weather models (e.g., COSMO) do not account for these effects as they simulate radiation only in very simplified 1D way. In the framework of HDCP2, the High Definition Clouds and Precipitation for Climate Prediction Project, we are currently developing a 3D Radiative Transfer code that is fast enough to drive dynamical processes. The basic concept is to use a combination of linear algebra with pre-calculated transfer coefficients.

At the moment. these transfer coefficients are calculated with a simple Monte Carlo ray-tracing model and stored in a look-up table. Due to the complexity of radiative transfer and the many influencing variables this table is big and its use is computationally expensive. Earlier approaches using neural-networks and adaptive high order polynomial regression have shown promising results but need further thought. Your part is to develop a fast yet flexible parameterization for the generation of the transfer coefficients replacing the mentioned approach.

This work is a crucial step to enable high resolution model runs with realistic radiative forcing. The goal is to further our understanding of cloud-radiation interactions which will advance short term weather prediction as well as reduce uncertainties in climate projections. We hope for a candidate with appreciation for radiative transfer, applied mathematics and programming.

Responsible Fabian Jakub, Bernhard Mayer

Radiation and climate modelling: Parameterization of sub-grid cloud-radiative effects in climate models

Clouds remain the largest source of uncertainty in climate models. This is mainly due to the fact that current climate model resolution of several tens of kilometers is too coarse to resolve clouds properly. Clouds are usually treated in terms of „cloud fraction“ only in climate models. A detailed, subgrid scale, description is not available. Radiation is usually calculated in simple 1D approximation. It was shown by our group that cloud shape has a distinct effect on solar and thermal radiation. E.g. absorption and emission of radiation at cloud edges can cause significant heating and cooling locally and thus changes in the development of shallow cloud fields (Klinger et al., 2017, Jakub and Mayer 2017). However, as climate models can not predict the size and distribution of clouds, these radiative effects cannot be calculated so far.

Shallow cloud field properties, such as the distribution of liquid water, cloud size, height and the distance between clouds have been found to show characteristic statistical behavior. Cumulus cloud size follows, for example, a power law. Based on these findings, Feingold et al., 2017 developed a heuristic model for shallow clouds (cumulus and stratocumulus). This fast, statistical model allows to simulated a variety of cloud fields.

Based on these cloud fields, we aim to find a statistical relationship of the heating and cooling rates of clouds which can then be included, along with the heuristic model of Feingold et al., 2017 in a climate model to solve the above explained sub-grid cloud disability of climate models.

The task for a master thesis would be: First, extend the heuristic model of Feingold et al. (2017) by another dimension (cloud distance). Second, perform a set of 1D and 3D radiative transfer calculations for the cloud fields of the heuristic model. Third, find a statistical relationship between cloud field properties (size, height, distance, liquid water content) and the heating/cooling rates as well as upwelling and downwelling fluxes.

Responsible Carolin Klinger, Bernhard Mayer

Cloud remote sensing: Determine the Influence of Radiatively Induced Surface Inhomogeneities using High-Resolution Satellite Imagery

Current research efforts (Jakub and Mayer 2017) suggest that 3D effects of radiative heating may lead to convective organization resulting in the formation of cloud streets. The above mentioned study examines the cloud-radiative-surface feedback in an idealized enviroment without complex wind fields and surface heterogeneities.

A subsequent step is to investigate if this kind of organization can be found in the complex environment of earths' atmosphere. To this end, this master thesis aimes at a determination of a possible observable correlation between radiative heating and the organization of shallow and deep convection in the real world.

The thesis likely encompasses the tasks of

Responsible Fabian Jakub, Bernhard Mayer

Sunphotometers: Extracting Atmospheric Parameters from Spectrometer Measurements and Comparison with Established Methods

A major challenge for remote sensing of the Earth surface is the influence of the atmosphere on the airborne and satellite data. For dark surfaces like open water, more than 90 % of the signal can be caused by the atmosphere. To correct the contribution of the atmosphere, its variable components (aerosol concentration, aerosol type, water vapour, ozone) must be measured. The determination of these parameters often requires in-situ measurements, because of their high spatial variability. These measurements are usually performed with dedicated sunphotometers, such as Microtops or CIMEL. The main goal of this thesis is to develop a measurement concept and evaluation method for extracting atmosphere components from measurements with any spectrometer which has a suitable spectral range and resolution. Furthermore this method should be validated in respect to established measurement protocols and hardware (in particular AERONET/CIMEL).

This thesis will be conducted in cooperation with the German Aerospace Center (DLR). Place of work and supervision will primarily be at the DLR site at Oberpfaffenhofen near Munich.

A background in meteorology, physics, or a related subject is required. Programming skills in an object oriented programming language, basic understanding of the operation principles of a spectrometer and willingness to take part in measurement campaigns at different sites in Germany are desired.

Responsible Tobias Kölling, Sebastian Riedel (DLR), Peter Gege (DLR)

Cloud remote sensing using polarized multi-angle observations

A new mission to measure cloud and aerosol properties is currently prepared by NASA-JPL. The satellite instrument will measure polarized radiances from a large number of viewing directions at several wavelengths from the UV to the NIR spectral region. Data from the airborne demonstrator of this instrument (AirMSPI, is already available. The aim of this thesis is develop an algorithm to retrieve parameters of the cloud droplet size distribution from these measurements. The algorithm shall be validated by simulating the observations using the 3D radiative transfer model MYSTIC. The retrieval algorithm may then be applied to the synthetic data, and the retrieved results may be compared to the real input. In particular the impact of 3D effects shall be investigated using this method. Part of the work can be performed at NASA-JPL in Pasadena, USA, if the student is interested to go abroad.

Responsible: — Claudia Emde

Ground-based remote sensing: Cloud and aerosol properties from simple polarized observations

Aim of the thesis is to use an existing off-the-shelf camera system to determine the degree of polarization of sky and clouds and to study how much information about cloud cover, optical thickness, cloud phase, droplet size, width of the size distribution, etc. can be obtained from such observations. At MIM we have a digital camera equipped with a polarizer. This work is meant as a pre-study for the cloud spectrometer which could be equipped with polarization capabilities in the future. Raw data of the camera system has to be used to determine the degree of polarization. Advantage of this quantity - which is a ratio of several images - is that there is no need for absolute calibration of the camera pictures. Questions to be answered: How accurate is the derived degree of polarization? Can we detect the thermodynamic phase of clouds from polarization images of their flanks or bottoms? Can we derive aerosol optical thickness or aerosol type from the clear sky?

Claudia Emde, Tobias Zinner

Ceilometer: Spatial and temporal variability of aerosol and cloud parameters

Backscatter signals of ceilometers of the „Ceilonet“ of the German Weather Service can be provided for several stations and years. They shall be used to determine mixing layer heights (with COBOLT developed in the framework of a phd-thesis at MIM), cloud base heights and sort of a cloud cover (in contrast to the conventional cloud cover that considers the whole hemisphere every few hours, a ceilometer measures continuously in zenith direction). Determine and compare local climatologies of different cloud- and aerosol parameters at different sites, and (optionally) with independent data sets, e.g. satellite based climatologies (literature) or numerical models.

Responsible: Matthias Wiegner

Trace gases: Imaging DOAS

This project is about the interpretation of differential optical absorption spectroscopy (DOAS) measurements in an imaging setup in order to determine the horizontal and vertical distribution of NO2 concentrations in cities. DOAS uses differential absorption structures of different trace gases like ozone or NO2 to derive their concentrations from spectral measurements. The DOAS method can be applied in different setups, both in an active and a passive mode, depending on whether an artificial light source or the sun light which had been scattered and partly absorbed in the earth’s atmosphere is used. In this setup the telescope of the instrument can be pointed in different directions and can generate an image of so called slant column densities when scanning in two dimensions (called imaging DOAS). Slant column densities represent concentrations integrated along all possible light paths. Those light paths have to be simulated using a radiative transfer model in order to interpret the slant column densities derived for a given scene correctly. Since the measurement setup aims at measuring NO2 concentrations in cities, the 3D radiative transfer model has to be able to include the geometry of buildings and other objects within the field of view, so the model to be used, MYSTIC, has to be extended accordingly.

The project includes the following steps:

  1. Familiarization with the 3D radiative transfer model MYSTIC
  2. Extending the model to include vertical surfaces with a defined albedo
  3. Modeling the building forming the observed scene (e.g. the Hong Kong skyline) to include in the MYSTIC simulation
  4. Learn the basics of DOAS
  5. Retrieval of slant column densities of already existing DOAS measurements using standard DOAS software
  6. Combining the DOAS results with the model output to form a consistent image of concentrations.

Responsible: Mark Wenig and Claudia Emde

Estimation of visibility from image sequences

In this project image sequences have to be analyzed in order to derive visibility range information. Different landmarks at different distances from the camera, e.g. high-rise building or church spires, can be used to derive optical parameters that depend on the visibility. Those parameters could be the local variability or the difference of the intensity of the object compared to the view in the sky close to the object, and can be compared to aerosol optical depths that have been measured at the same time.

The project includes the following steps:

  1. Automatic adjustment of the images from the sequence in case the camera shifted
  2. Chose useful objects in the field of view
  3. Determine intensity differences between objects and the sky
  4. Derive local structure parameters, e.g. local variability
  5. Explore further optical parameters using digital image processing operators
  6. Analysis of the dependency of the derived parameters on aerosol optical depths measured independently
  7. Simulation of the optical parameters using a radiative transfer model and comparison with parameters derived from the image sequence
  8. Installation of more cameras and implementation of a real-time evaluation

Responsible: Mark Wenig and Matthias Wiegner

Cloud remote sensing: Simulation of cloudbow and glory for MSG/SEVIRI

Particular optical phenomena like the cloudbow and the glory can be exploited for the investigation of cloud properties that are usually not accessible using passive optical measurements. However, they probably require constant cloud optical properties over large areas or high spatial resolution measurements. The SEVIRI radiometer aboard the geostationary Meteosat Second Generation (MSG) satellite has a sampling distance of 3 km at the subsatellite point and has therefore a rather coarse spatial resolution.

Aim of this thesis is the realistic simulation of SEVIRI radiances for these special sun-satellite geometries in order to check whether cloud properties derived from SEVIRI are sensitive to these phenomena.

Responsible: Luca Bugliaro

Alte Themen, usw.

Januar 2018