kommt noch

Achtung: Die Navigationslinks auf dieser Seite funktionieren nicht, weil das nicht der Rahmen ist, der zu diesem Skript gehört. Wenn das Skript ordnungsgemäß dort installiert ist, wo es später laufen soll, funktionieren auch die Links in der dortigen Umgebung.

LINUX Cluster Project

Geographie und geographische Fernerkundung


  • Name: Lehrstuhl für Geographie und Geographische Fernerkundung
  • Address: Luisenstraße 37, 80333 München
  • Project Proposal Date: 2017-02-27 17:13:10


Project A: Das Projekt beinhaltet die Weiterentwicklung eines Klassifikationsverfahrens (Modified Nearest Neighbor) zur überwachten Klassifikation vor allem im Bereich der Satellitenfernerkundung. Kern des Verfahrens ist ein globales Optimierungsverfahren - Simulated Annealung - das z.B. über MPI parallelisiert werden soll. Desweiteren sind sogenannte Cross-Validierungs Verfahren nötig, um die Güte des Verfahrens abzuschätzen. Dieses basiert auf Monte-Carlo Verfahren, die ebenfalss parallesiert werden sollen. Projekt B: We apply a gap-filling scheme that attempts to address the problem of blending current observations and supplementary information when predicting the deterministic component of missing data in net CO2 flux time series. The starting point for this approach is to assume light, temperature and time as the dominant driving forces controlling the net canopy CO2 flux. The challenge then is to extract the relevant linear or nonlinear relationships between net CO2 flux and each of these drivers, simultaneously, and without having to assume their parametric form a priori. We achieve this by optimizing a three-dimensional hypersurface for net CO2 flux within the CO2 flux-light-temperature-time space. This hypersurface is constructed using piece-wise polynomials (cubic splines), hence allowing a certain degree of control over the rates of change within the hypersurface while retaining the objectivity of the semi-parametric approach. A stochastic component is added to the modelled signal component to fill the measurement gaps. The net CO2 flux is finally disaggregated into CO2 uptake and respiration by inter- and extrapolating the light independent night fluxes in the flux-time-temperature-space. 3.By providing and preparing the climate station and ERA-interim data, the statistical distributions of climatic factors on the selected locations are calculated. In further analysis, first an emphasis on the temperature as an important factor and it is designed to answer the following questions: 1. Does ERA-interim data have the true or approximate distributions with the climate station data? 2. How big are the differences or bias between ERA-interim and climate station data? 3. Which data is restricted to certain regions, and is there some bias correction procedure? 4. How long is the time series? How to choose a representative or stationary distribution? The shorter the required period, the lower the costs for the commercial use of data! 5. The ERA-Interim data are available in a maximum spatial resolution of 1 °. Since the 2.5 ° resolution is clearly favorable, should also be reviewed their usefulness! Project C: Extreme flood events and extensive dry periods in the recent past have shifted the perception of the scientific community and public, leading to an intensified focus on investigating hydrological extreme events and how they might be affected by climate change. Therefore, new and improved prediction tools as well as management strategies have to be developed. The ClimEx project (Climate change and hydrological extreme events – risks and perspectives for the Bavarian water management) under the leadership of Prof. Dr. Ralf Ludwig (Ludwig-Maximilians-University Munich) in collaboration with the Leibnitz Rechenzentrum (LRZ) Garching, the Bavarian state office for the environment (LfU) and several Canadian partners intend to assess the impact of climate change on hydrological extreme events through a highly sophisticated hydro-climatic model chain established for Bavaria. A large ensemble of high resolution regional climate model (RCM) data for central Europe will be produced, with the assistance of the LRZ, at the supercomputing facility SuperMUC. This ensemble data, as well as further available RCM data (Euro-CORDEX), form the basis for the high resolution hydrological modeling of all Bavarian catchments (ca. 102.000 km²). With the large amount of available climate model data a detailed statistical analysis of rare hydrological extreme events is possible. Furthermore, a new method for ‘virtual perfect prediction’ will be developed for climate change impact assessment. According to the large amount of available data this method shall identify patterns within the data which induce hydrological extreme events using different lead times. This offers the opportunity to simultaneously analyze different scenarios of control measures for damage reduction. The hydrological model setup will require substantial computational resources due to the high spatial and temporal resolution and the applied physically based algorithms, which are needed for the reproduction of peak and low flows. Furthermore, the parallelized investigation on management scenarios contributes to the high computational demand. Hence, ClimEx also intends to employ the computing capacity of the LRZ’s HPC SuperMUC for hydrological modeling for the first time in Bavaria. Projekt C (Josef Schmid / Ralf Ludwig): Am Department für Geographie der Ludwig-Maximilians-Universität startete am 1. Juni 2015 unter der Leitung von Prof. Dr. Ralf Ludwig und in enger Zusammenarbeit mit dem LRZ (Prof. Dr. Dieter Kranzlmüller, LMU) das Projekt „Klimawandel und hydrologische Extremereignisse – Risiken und Perspektiven für die bayerische Wasserwirtschaft (KlimEx)“. Ein Schwerpunkt besteht in der Anwendung eines großen Ensembles an regionalen Klimamodellen für die hydrologische Modellierung aller relevanten bayerischen Flusseinzugsgebiete (ca. 102.000km²). Dieses Ensemble besteht aus 50 transienten Modelläufen des Regionalen Klimamodels CRCM5, angetriebne von 50 Membern des Globalen Klimamodel CanESM2. Die erwartete Datenmenge beträgt ca. 500TB.