Test measurements on our campus
On our campus, we estimated the initialization time of the GPS compass at the site with open sky. GPS signals from 7 to 11 satellites could be detected for about 5 min after the GPS compass was turned on (Fig. 2). However the latitude and longitude changed dramatically with the number of satellites, and the direction shifted gradually and was not stable. Therefore, we considered that it would take about 5 min to initialize the GPS compass and decided not to record data for the first 5 min.
We measured GPS and magnetic azimuths relative to the sun azimuth, as rotating the turntable of the orientation device by approximately 30°. After 5 min of initialization with the GPS compass, about 1800 directional data were acquired at 1 Hz for about 30 min for each heading, and the averages and rms were calculated for the 11 headings (Fig. 3, Additional file 1: Table S1). There seems to be no obvious dependence of the relative GPS and magnetic azimuths on the sun azimuth. Although the sun azimuth itself contains errors, the two differences are in the range of \(\pm \, 2.5^\circ\). The averaged rms of the GPS azimuth is \(0.39^\circ\), which is less than the value of \(0.75^\circ\) in the manufacturer’s specification. Averaged over the 11 headings, there is no significant discrepancy between the three pairs from the sun, GPS and magnetic azimuths (Additional file 1: Table S1). The averages and rms represent the accuracy and precision of the GPS compass in open sky, and those of the magnetic compass where strongly magnetized rock body is absent.
Temporal variations of the GPS azimuths relative to the sun compass and the numbers of satellites that were able to capture the GPS signal are shown in Fig. 2 for the open sky site and the site between buildings. In both cases, the elevation mask was set to 5°. Except for the first 5 min of initialization, the azimuths at the open sky site stayed within a few degrees from the sun azimuth, whereas the GPS azimuths at the site between buildings showed large deviations from the sun azimuth varying widely from 5° to 19°. The number of satellites was stable at 8 or 9 in the open sky site, whereas it kept fluctuating between 5 and 8 at the site between buildings.
Figure 4 shows the time-averaged GPS azimuths and satellite counts with their rms for the open sky site with an elevation mask from 5° to 45°, compared to the site between buildings with an elevation mask of 5°. The average deviations from the sun azimuth were less than 1° and the rms were small for the elevation masks from 5° to 35°, but at 45° the average deviation was increased to \(3.28^\circ\) and the rms was \(2.83^\circ\). The numbers of satellites were around 8 for the elevation angles from 5° to 25° and the variation was small. At 35° it was decreased to \(5\pm 1.5\), and it was only 4 and did not change with time at 45°. Despite the relatively large number of satellites \(6.2\pm 0.8\) at the site between buildings, the GPS azimuth deviation \(11.39^\circ \pm 2.97^\circ\) was extremely large. Although the site between buildings seems to be receiving signals from sufficient number of satellites, the directional deviation should be due to the multipath effect, in which false signals are reflected by the buildings rather than directly transmitted from the satellites to the GPS compass.
Orienting drill cores of historical lavas in Izu Oshima
At sampling sites in Izu Oshima, the GPS compass was turned on and stood by for approximately 5 min for initialization, then measurements were taken for about 5 min for each drill core to obtain an averaged azimuth. The elevation mask was set to 5°. Temporal variations of GPS azimuth and number of satellites at two sites situated on the coast and in forest are shown in Fig. 5. At the flat coastal site, the GPS azimuth varied less than \(1^\circ\) from the sun azimuth and the number of satellites remained constant at seven. At the forest outcrop, in contrast, the azimuth showed large variations of \(\pm\,10\)° and the number of satellites varied repeatedly between 4 and 6. The small number of satellites captured at the forest site is due to covering trees or the rock body that makes up the vertical outcrop. Multipath may induce the large deviations and variations of the GPS azimuth.
Time-averaged GPS azimuths and their rms were calculated for 35 drill cores at coastal and forest sites, and azimuths were not adopted if the rms were greater than 2° for the 5 min time series. The GPS azimuths were obtained for 22 of the 26 cores at the three coastal sites, but none of the 11 cores at the three forest sites yielded GPS azimuths (Additional file 1: Table S2). The difference between GPS and sun azimuths has the average of + 0.4° with a rms of 2.3°. Including magnetic compass and backsighting results at the other three sites where GPS compass measurements were not taken, the magnetic azimuth was \(-\,1.8^\circ \pm 5.2^\circ\) and the backsighting azimuth was \(-\,0.2^\circ \pm 1.5^\circ\) with respect to the sun azimuth (Fig. 6). The remaining three differences (GPS-backsighting, GPS-magnetic and magnetic–backlighting azimuths) were also determined. The three differences involving the magnetic azimuth showed larger deviations and larger rms than the other three differences. The GPS compass and backsighting have about the same accuracy and precision as the sun compass, but those of the magnetic compass are much poorer. The differences between the magnetic and sun azimuths are more than 10° in some cores, and even if taking averages for individual sites the site averages still exceed 2° at several sites. The historical lavas in Izu Oshima are strongly magnetized as shown by the natural remanent magnetization intensity of total average and rms \(23.0 \pm 11.8~\hbox {A/m}\) (Additional file 1: Table S2).