Browsing by Author "Roth, Matthias"

Now showing 1 - 2 of 2

Open Access
GOCE data analysis: optimized brute force solutions of large-scale linear equation systems on parallel computers
(2010) Roth, Matthias
The satellite mission GOCE (Gravity field and steady-state Ocean Circulation Explorer) was set up to determine the figure of the Earth with unprecedented accuracy. The sampling frequency is 1 Hz which results in a massive amount of data over the one year period the satellite is intended to be functional. From this data we can setup an overdeterminded linear system of equations to estimate the geopotential coefficients which are required for modelling the Earth's gravity field with spherical harmonics in the desired high resolution. The linear system of equations is solved "brute-force" which means that the normal equation matrix has to be kept in memory as a whole. The normal equations matrix has a memory demand of up to 65 GByte, hence we need a computer providing a sufficient amount of memory and fast multiple processors for the computations to get them done in a reasonable time. In this study, a program was written to compute the geopotential coefficients from simulated GOCE data, as GOCE real data were not available yet. As a first step, the program was optimized for the computations to become more efficient. As a second step, the program was parallelized to speed-up the computations by two different techniques: For a first version, the parallelization was done via OpenMP which can be used on shared memory systems which usally have a small number of processors. For the second version, MPI was used which is suited for a distributed memory architecture, hence can incorporate much more processors in the computations. In summation, we gained a huge boost in efficiency of the program due to the optimization. Furthermore, huge speed-up was gained due to the parallelization. The more processors are incorporated in the computation, the more the overall efficiency drops because of increasing communication between the processors. Here we could show that for huge problems the MPI version is running more efficient than the OpenMP version.
Open Access
Marine full tensor gravity gradiometry data analysis and Euler deconvolution
(2009) Roth, Matthias
The Euler Deconvolution is a standard tool of the geophysical prospection. It is a semi automated process and supports the interpretation of magnet field data, gravimetry data or gravity gradiometry data. It is used to estimate the location of the source of a magnet field anomaly or a gravity field anomaly. The obtained location is equivalent to a potential resource (e. g. petroleum, gas, ore, salt etc.) Although gravimetry data or gravity gradiometry data are used in geodesy frequently this method did not attract much interest in this branch of science. For this study thesis a real marine full tensor gravity gradiometry data set is given. The data set was recorded by Bell Geospace. The gravity gradiometer was mounted on a ship. The measurement area lies between the Shetland Islands and the Faroe Islands. This area was chosen for the measurements because of the basalt formations there. For geodesists the accuracy of data is of great interest. Therefore in a first step these data is run through an accuracy analysis. After the introduction of the theory of the Euler Deconvolution and its testing on synthetic data the Euler Deconvolution is carried out on the data set. The analysis of synthetic data shows that the Euler Deconvolution is suited to detect the location of a mass anomaly. If multiple mass anomalies are to be detected these mass anomalies must have a certain distance for this method to function reliably. If the distance between mass anomalies is too small the Euler Deconvolution tends to give false mass locations. In spite of the big noise on the gravity gradient data the Euler Deconvolution still shows to be useful on the real data set.