Simon Flower

A technique for estimating the absolute vector geomagnetic field from a marine vessel

We have developed a technique to estimate the absolute strength and direction of the geomagnetic field from a marine vessel. This technique will be of value in the study of marine magnetic anomalies, directional drilling or geomagnetic field modelling. One of the main difficulties in this operation is to correct the data for the magnetic field generated by the vessel itself. Assuming the vessel susceptibility is isotropic, we show that by turning the ship through 360° at a place where the strength of the field is known we can estimate the local direction of the magnetic field. Once this is known, the vessels field at any attitude can be robustly estimated and the measurements of the full magnetic field vector made in a normal surveying mode can be corrected. The ambient magnetic field estimates at the turn locations have proved to be very accurate. In normal surveying mode, these estimates are not as good since they are directly dependent on the accuracy of the vessel attitude measurements. However, when the technique is applied on real data, the total intensity field estimates have a very low level of noise showing that the vessel signal has been well accounted for.

INDIGO: Better Geomagnetic Observatories Where We Need Them

Good magnetic observatories are needed more than ever for global modeling and navigation. Magnetic satellite missions, once said to be the death of ground based observations, are now demanding quality data from fixed observations points on the Earth.

Making fortran legacy code more functional: using the BGS* geomagnetic field modelling system as an example

This paper presents an application case study of the British Geological Surveys (BGS) Geomagnetic Field Modelling System code. The program consists of roughly 20 000 lines of highly-tuned F<scp>ortran</scp> MPI code that has a runtime of about 12 hours for a signal execution cycle on a cluster utilising approximately 100 CPU cores. The program contains a sequential bottleneck that executes on a single node of the cluster and takes up to 50% of the overall runtime. We describe an experiment in which we rewrote the bottleneck F<scp>ortran</scp> code in S<scp>a</scp>C, to make use of auto-parallelisation provided by the S<scp>a</scp>C compiler. The paper also presents an implementation of a foreign-function interface, to link the S<scp>a</scp>C kernel with the F<scp>ortran</scp> application. Our initial performance measurements compare the S<scp>a</scp>C kernel performance with the F<scp>ortran</scp> bottleneck code; we also present results using an O<scp>pen</scp>MP Fortran implementation. Our figures show that the S<scp>a</scp>C-based implementation achieves roughly a 12.5% runtime improvement, and outperforms the O<scp>pen</scp>MP implementation.