It is also interesting to compare the substorm signature in the polar magnetosphere with that in the magnetotail. The GEOTAIL spacecraft was located around (-6.4,-8.8,-2.4) () in GSM coordinates at 0500 UT. Figure 2 shows the data of POLAR and GEOTAIL in the same time frame. The bottom panel shows data from the low-energy plasma detector (LEP) on board GEOTAIL [ Mukai et al., 1994]. The GEOTAIL spacecraft was located within the plasmasheet throughout the interval of the figure: Before 0500 UT, the ion density was rather high and the ions were not moving much (as shown in the ion bulk velocity data), which are features of the plasmasheet ions. The gradual temperature decrease from 0300 UT to 0500 UT can be explained in terms of the plasmasheet thinning during the loading phase of the substorm: Because of the thinning, the relative distance of the spacecraft from the center of the plasmasheet increased, which caused the decrease in temperature. The ion density gradually increased from 0300 UT to 0430 UT and then decreased until 0500 UT. The decrease is consistent with the above-explained increase in the spacecraft distance from the center of the plasmasheet during the loading phase. The increase until 0430 UT may have been caused by the compression of the plasma sheet during the loading phase, which overcame the effect of the relative motion of the spacecraft away from the plasmasheet center. Another possibility is the dawnward motion of the spacecraft: As a spatial structure of the plasmasheet, its density increases with deceasing distance from the flank magnetopause [e.g., Lennartsson and Shelley, 1986]. After 0500 UT, the ion temperature jumped up, and there was a burst of earthward and dawnward ion flow with the duration of 3 min. Thus, at first sight, GEOTAIL data appear to suggest that the substorm onset was 0500 UT. There is a 52 min lag from 0408 UT, the initial onset time on the ground (line B). A possible way to explain this difference is the Y position of GEOTAIL. That is, because GEOTAIL was located at , or at 3.6 hour MLT, dawnward propagation of substorm signal, from the onset region, might have taken several tens of minutes to reach the GEOTAIL position. Nagai [1982, Figure 12] reports that the east-west propagation speed of the substorm onset region, or so called the current wedge, is [min/MLT hour]. We note in Figure 2 that the multiple onsets are more apparent at the 327 geomagnetic longitude chain than in any other longitudes, thus the onsets are likely to have happened in the premidnight sector, around 22 hour MLT. Then, the propagation time from 22 hour to 3.6 hour MLT is estimated to have been min. The observed time lag (52 min) is larger than 40 min, but at least some of the lag may be explained in terms of the east-west propagation of the substorm heating of the plasmasheet. If the event B did not reach the position of GEOTAIL but the event C (0435 UT) did, then the lag is 25 min, which is well within the range of 17-40 min. In relation to this, we note that POLAR was similarly distant from the expansion onset region (presumably in the near-Earth tail) to that of GEOTAIL. Thus the onset signal would have spent similar time to propagate to POLAR, but the time delay from the initial ground onset to the maximum at POLAR (from 0408 to 0430 UT) was smaller than that at GEOTAIL (from 0408 to 0500 UT). Thus the delay for POLAR may not be explained in terms of the propagation of the onset signal. Still another possibility is that neither event B nor C reached the position of GEOTAIL, but the event D did. However we do not prefer this possibility because event B is expected to have been large, judging from its observability at many local times (top panel of Figure 2) and from the fact that the at POLAR decreased after the event B (middle panel).