|From||Willem Jan van de Berg <firstname.lastname@example.org>|
|Date||Tue, 28 Mar 2017 13:42:03 +0200|
We seek to fill a 4-year PhD position on multiresolution flow-field processing for climate applications. This research project is a collaboration between the Department of Information and Computing Science (ICS) and the Institute for Marine and Atmospheric Research (IMAU), both at Utrecht University, The Netherlands.
The position is full time, 4 years and to start July to September 2017. All information, e.g. on how to apply, can be found here:
Project aims and setup:
Flow and vector-field design is a new and exciting topic that has gained considerable momentum in recent years in geometry processing. It has many applications: from climate analysis, through mass transportation, to computer graphics and animation. Many problems in measuring, reconstructing, and predicting flows are still unsolved. A major problem in both practice and theory is to work effectively with big data and multiple resolutions.
This project applies flow and vector-field design to climate research. More specifically, to upsampling of modelled precipitation fields over rough topographic terrains, such as ice sheet margins. Ice sheets and glaciers gain mass primarily by precipitation; hence, detailed spatial estimates are very useful. Conventionally, Regional Climate Models (RCMs) are used to estimate high-resolution precipitation fields, but computations at high resolution are very slow and storage intensive. Therefore, an alternative approach could be very valuable.
Precipitation over polar glaciated areas is predominantly topography, wind and temperature driven. Generally, climatological precipitation fields are largely determined by how much the topography interacts with the mean atmospheric flow. Hence, high-resolution precipitation fields should be deducible from low-resolution 3D wind and temperature fields, and high-resolution topography. As such, the vector-based upsampling technique must be geometry aware and adaptive. i.e. generate high-resolution fields only for regions with small spatial precipitation patterns.
The main goal of this project is to develop methodology to model multi-resolution scalar and vector fields to estimate high-resolution precipitation fields accurately and efficiently. We will extend these fields to adaptive resolution models. Consistency is a major consideration: the total flow volume must be preserved in each resolution level. In addition, upsampling and downsampling should preserve the topological features of the fields, e.g. sources, sinks, and vortices. Machine learning will be applied in order to learn the necessary adaptivity, and the correct way to upsample precipitation fields.
The successful candidate will study the state-of-the-art in multi-resolution and vector field modelling. Furthermore, the candidate will familiarize him/herself with climate simulations and the primary meteorological and cryospheric concepts. The research team will work together to develop approaches that are likely to give the desired results. The student will implement these multi-resolution approaches, and demonstrate their efficiency and accuracy on real-world data from climate models. Outcomes are analyzed by the team. The research encompasses the formulation of the underlying geometric methodology, targeting publications in relevant journals in computer science, and the demonstration of the use and effect of this methodology on climate modelling, targeting publications in relevant journals in climate research.
Willem Jan van de Berg
Institute for Marine and Atmospheric Research Utrecht
PO Box 80000, NL-3508 TA Utrecht, the Netherlands
Tel. +31 (0)30 2533273 Fax +31 (0)30 2543163
Go to: Periods · List Information · Index by: Date (or Reverse Date), Thread, Subject or Author.