Student projects

BSc topics

Impact of degrading data accuracy on airmass trajectories

Meteorological data output is generated with increasingly more complex numerical models and with increasingly larger spatial and temporal resolution. This results in strongly increased storage demands. To strike a balance between the computational expenses, storage demands and the model accuracy it is important to understand how these interact. For example, increasing the temporal or spatial resolution of a model together with a reduction of the precision of the output might result in essentially the same storage demands as before. In addition, it is not known how a reduced precision in the stored output data might affect further data analysis results and how this compares to other sources of uncertainty, e.g. the numerical schemes used or uncertainty due to inaccurate or sparse observational data.
In this project, we will explore this question for the computation of trajectories based on the recently released ERA5 data-set, which represents the "best guess" of the state of our atmosphere in the past.
Target group: BSc Computer Sciences/Atmospheric Physics

Contact: Dr. Annette Miltenberger

Impact of stochastic airmass trajectories on interpreting UTLS aircraft measurements

Interpreting aircraft measurements of trace gases and clouds often requires information on the air mass history that is typically derived by computation of backward trajectories. The computation of trajectories is, however, subject to several sources of uncertainty - not least in the wind field data. Recently, we have developed a stochastic trajectory calculation tool that allows to incorporate estimates of wind field uncertainty into the trajectory calculation.
In this project, you will use the new stochastic trajectory tool and aircraft data from some recent field campaigns in the North Atlantic region to quantify the uncertainty of airmass sources and history of observed air parcels. Thereby, the importance of stochastic trajectory data for the interpretation of aircraft measurements can be determined.
Target group: BSc Atmospheric Physics

Contact: Dr. Annette Miltenberger

Impact of secondary ice formation processes on hail forecasts

Hailstorms can cause substantial damage to infrastructure and agriculture. Forecasting these storms remains challenging not least due to large uncertainties in the involved cloud microphysical processes and their representation in numerical weather prediction model. Of particular importance for hail formation is the amount and size of ice particles in the clouds. One source of small ice crystals at relatively warm temperatures is secondary ice formation, i.e. the formation of ice crystals through processes other than freezing of liquid droplets/aerosols. Only recently model descriptions of some of these processes have become available and therefore insight in their impact on hail forecasts is still missing.
In this project, you will explore simulations of a supercell storm in the Munich area with and without representations of secondary ice formation. This will provide insight into the importance of the process for hail forecasting.
Target group: BSc Atmospheric Physics

Contact: Dr. Annette Miltenberger

Climatology of large-scale conditions during WCB occurrence in the North Atlantic

Extra-tropical cyclones and in particular the associated warm conveyor belt are important systems for cloudiness, precipitation, and moisture transport to the upper troposphere. The synoptic scale conditions during WCB occurrence varies and influences the efficiency of precipitation formation as well as moisture transport.
In this project, you will make use of an existing WCB climatology based on ERA5 and establish a climatology of large-scale conditions during these events. This climatology will later be used to investigate the relationship between large-scale conditions and the efficiency of moisture transport.
Target group: BSc Atmospheric Physics

Contact: Dr. Annette Miltenberger

 

Systematische Analyse der Unsicherheiten von Radar-/Satelliteretrievals mit AD

Die indirekte Beobachtung meteorologischer Grössen durch sog. remote sensing Methoden (z.B. Radar, auf dem Satellit) spielen eine zentrale Rolle in der Analyse von atmosphärenphysikalischen Prozessen und in der Verifikation von Wetter- und Klimamodellen.Die Ableitung der meteorologischen Grössen aus der Messung der Rückstreuung elektromagnetischer Wellen (d.h. dem „retrieval“) erfordert zahlreiche Annahmen, die einen z.T. nicht unerheblichen Einfluss auf die „Messwerte“ haben. Die assoziierten Unsicherheiten bzw. die Wichtigkeit der verschiedenen Annahmen / Parameter ist im Allgemeinen nicht bekannt. Die Methode der Algorithmic Differentation ermöglicht dies systematisch zu untersuchen und wäre daher ein wichtiger Baustein zur zukünftigen Verbesserung bzw. Unsicherheitsabschätzung der retrievals.

Contact: Dr. Annette Miltenberger

 

 

 

MSc topics

 

Analysis of uncertainties in radar-/satellite retrievals with algorithmic differentiation

The observation of meteorological quantities by remote sensing (e.g. radar or satellite) is an important source of information for example on clouds, cloud structures. The data are used in physical analysis, but also for the evaluation of weather and climate models. However, quantities of interest often have to be estimated from the observed backscattered radiation (“retrieval“). These retrievals often rely on many assumptions that can have a significant impact on the retrieved quantities, although these uncertainties are often not well characterised.
The method of algorithmic differentiation allows to directly assess the gradients of a model with respect to its parameters. In this project you will apply algorithmic differentiation to different retrieval algorithms for cloud related variables (satellite and radar products). The gradients of the algorithms are investigated for their spatiotemporal structure and relation to larger-scale meteorological conditions. Understanding and quantifying key sensitivities of retrieval algorithms is an important contribution for future improvements of retrievals and thereby quantitative information on cloud-related variables.
Target group: MSc Computer Sciences/Atmospheric Physics

Contact: Dr. Annette Miltenberger

 

Conservation of physical properties along airmass trajectories

Lagrangian analysis of meteorological data is a widely used technique in the scientific community, where one establishes a trajectory of an individual air mass through the atmosphere based on the wind fields. On top of the trajectory itself, one often considers the conservation or non-conservation of variables along the trajectories, e.g. potential vorticity, potential temperature, humidity or aerosol. The quality of trajectory data impacts whether the diagnosed (non-)conservation depicts physical processes as represented in the underlying data or is just a consequence of the accuracy of the trajectory scheme. For testing the conservation properties, one can use passive fields and deviations introduced by the trajectory scheme. Recently, test cases to analyse the impact of numerical diffusion on tracer transport have been proposed [doi:10.1016/j.jcp.2010.08.014].
In this project, we will implement and perform these test cases with the ICON model (the numerical weather forecast model run by the German Weather Service, Deutscher Wetterdienst) and use the output to perform trajectory simulations. Thereby we can analyse contributions from numerical diffusion and trajectory accuracy. We will further test the impact of data precision and spatio-temporal resolution of the wind field data on the trajectory accuracy.
Target group: MSc Computer Sciences/Atmospheric Physics

Contact: Dr. Annette Miltenberger

 

Impact of local flow features on the recurrence of organised convection in the Munich area

Hailstorms can cause substantial damage to infrastructure and agriculture. Organised convective storms occur in various places over Germany with local hotspots in the pre-Alpine region and the Swabian Jura. One area with high damage potential and frequent occurrence of organised convection is the area around Munich. Despite their high socio-economic impact, forecasting such events remains a major challenge.
In this project, you will analyse the conditions before, during and after the occurrence of organised convection in the Munich area. Thereby typical flow configurations can be identified and related to forecast performance of and error patterns in state-of-the-art modelling systems. The latter may provide insight into deficiencies in the modelling system and hint at necessary future developments for improving forecasts of high-impact weather events.
Target group: MSc Atmospheric Physics

Contact: Dr. Annette Miltenberger

 

Role of aerosol conditions for the predictability of heavy precipitation

Aerosol abundance and properties are intrinsically linked to cloud formation due to their role as cloud condensation nuclei and ice nucleation particles. The impact of aerosol variability on precipitation amounts and distribution remains, however, debated.
In this project you will explore ensemble forecasts of heavy precipitation events over Germany and investigate whether there are statistical links of forecast errors to aerosol conditions. On a weather timescales this would hint at significant aerosol-cloud interaction effects.
Target group: MSc Atmospheric Physics

Contact: Dr. Annette Miltenberger

 

Cloud biases in IFS / ICON-EU ensemble data and its relation to weather fronts

Cloud cover is a key quantity in weather forecasts with relevance for e.g. renewable energy production and radiative fluxes. Nevertheless, currently operational forecasting systems often struggle to predict cloud cover accurately. The physical processes are not always clear, which impedes developments of efficient remedies. In mid-latitudes synoptic-scale fronts are often related with extensive cloud systems. Biases in the representation of front-related clouds are therefore of particular interest
In this project you will combine satellite, reanalysis data, and IFS / ICON-EU ensemble forecast data to investigate cloud cover biases and error structures in regions adjacent to synoptic-scale weather fronts. Fronts will be identified with a recently developed machine learning based front detection algorithm.
Target group: MSc Atmospheric Physics

Contact: Dr. Annette Miltenberger

 

Sources and growth of moisture errors in the IFS ensemble

Clouds are still poorly forecast in many global and regional weather and climate model. In particular, it is not well understood how uncertainty in moisture fields either from an insufficient charactersation of the initial condition or due to mis-represented cloud physics propagates in ensemble forecasts and impacts cloud formation at later forecast times.
In our research group recently a tool for identifying and tracing moisture error features in IFS ensemble data has been developed. In this project you will use IFS ensemble data together with this tracking tool to establish a climatology of moisture error source and tracks. This climatology will then be used to investigate the consistency of errors across ensemble members as well as their connection to larger-scale meteorological conditions and initial condition uncertainty as estimated by data assimilation. This analysis will provide novel insight into the sources of moisture errors, error growth processes, and model deficiencies.
Target group: MSc Atmospheric Physics

Contact: Dr. Annette Miltenberger

 

Classification of cloud structures

Clouds consist of myriads of small particles. However, on larger scales pattern are formed via the interaction of different processes in an emergent way. It is not clear how this structure formation process really works, but obviously the environmental conditions (e.g. winds, temperature, humidity, aerosols) have a strong impact on the formation of different pattern as can be seen in observations. In this project, cloud structures as seen from satellite should be classified. Using machine learning algorithms, a large data set of satellite images (e.g. Meteosat) will be investigated in order to identify different classes of cloud pattern automatically. The results should be investigated statistically in order to obtain a first estimate about the frequency of occurrence of distinct cloud pattern. In a second step, the identified cloud structures should be combined with meteorological data sets (reanalysis data from ECMWF) in order to provide information about preferential conditions for different cloud pattern.

Contact: Prof. Peter Spichtinger

Target group: MSc Computer Sciences/Atmospheric Physics

 

Determination of radiative forcing for ice clouds using machine learning

Clouds crucially influence the energy budget of the Earth-Atmosphere system. They partly reflect or scatter incoming solar radiation back to space, thus less energy as compared to clear sky scenarios is transferred into the system (cooling, albedo effect). On the other hand, infrared radiation as emitted from Earth’s surface is partly absorbed by clouds and re-emitted, leading to warming of the system (greenhouse effect). For ice clouds, both effects are of comparable size. Thus, the resulting net effect depends on some parameters characterizing the environmental conditions. In this project, radiative transfer calculations using a very simple and idealized setup should be carried out for an ensemble of several parameter settings. From the results a new but simple model should be developed using machine learning techniques to describe the net radiative effect of ice clouds as a function of parameters. This new model then can be used for first estimations of the radiative effect of ice clouds for similar environmental settings.

Contact: Prof. Peter Spichtinger

Target group: MSc Computer Sciences/Atmospheric Physics

 

Impact of degrading data resolution on the quality of air mass trajectories

Meteorological data output is generated with continuously complexer numerical models with increasingly spatial and temporal resolution, resulting in strongly increased storage demands. To strike a balance between the computational expenses, storage demands and the model accuracy it is important to understand how these interact. For example, increasing the temporal or spatial resolution of a model together with a reduction of the precision of the output might result in essentially the same storage demands as before. In addition, it is not known how a reduced precision in the stored output data might affect further data analysis results and how this compares to other sources of uncertainty, e.g. the numerical schemes used or uncertainty due to inaccurate or sparse observational data.
In this project, we will explore this question for the computation of trajectories based on the recently released ERA5 data-set, which represents the "best guess" of the state of our atmosphere in the past.

Contact: Dr. Annette Miltenberger

Target group: MSc Computer Sciences/Atmospheric Physics

 

Conservation of physical properties along air mass trajectories

Lagrangian analysis of meteorological data is a widely used technique in the scientific community, where one establishes a trajectory of an individual air mass through the atmosphere based on the wind fields. On top of the trajectory itself, one often considers the conservation or non-conservation of variables along the trajectories, e.g. potential vorticity, potential temperature, humidity or aerosol. The quality of trajectory data impacts whether the diagnosed (non-)conservation depicts physical processes as represented in the underlying data or is just a consequence of the accuracy of the trajectory scheme. For testing the conservation properties, one can use passive fields and deviations introduced by the trajectory scheme. Recently, testcases to analyse the impact of numerical diffusion on tracer transport have been proposed [1].
In this project, we will perform these testcases with the ICON model (the numerical weather forecast model run by the German Weather Service, Deutscher Wetterdienst) and use the output to perform trajectory simulations. Thereby we can analyse contributions from numerical diffusion and trajectory accuracy. We will further test the impact of data precision and spatio-temporal resolution of the wind field data on the trajectory accuracy.

[1] doi:10.1016/j.jcp.2010.08.014

Contact: Dr. Annette Miltenberger

Target group: MSc Computer Sciences/Atmospheric Physics