Water drainage from Kushiro Coal Mine decreased on the day of all M ≥ 7.5 earthquakes and increased thereafter

The amount of water drainage from Kushiro Coal Mine in Hokkaido, Japan decreased on the day of all M ≥ 7.5 earthquakes with epicenters within 300 km of the mine during the monitoring period and increased after these earthquakes. This is a valuable finding which would give us a clue to understand pre- and post-seismic rockmass behaviors and contribute for progress in earthquake prediction in future.


Results
Seven M ≥ 7 earthquakes, including two giant (M ≥ 8) earthquakes, occurred during the monitoring period within 300 km from the coal mine and were catalogued in a database by the Japan Meteorological Agency ( Fig. 1 and Table 1). The amount of water drainage during the entire monitoring period were shown in Fig. 2. It gradually decreased with time and, after the sudden decrease due to the abandonment of the drainage from Cretaceous, became almost constant. The amount of water drainage in the period of 15 days before and after all M ≥ 7 earthquakes were shown in Fig. 3. The regular variations in a 1-week period typically seen in (c), (e) and (g) occurred because some pumps in the mined out areas were shut down on Sunday for safety reasons and did not exhibit any variation in water inflow. The amount of water drainage is larger for Taiheiyo Coal Mine (Fig. 3a-c) than that for Kushiro Coal Mine ( Fig. 3d-g). This is because Kushiro Coal Mine quit the drilling to Cretaceous thereby abandoned the deeper mining areas as stated above. It can be seen that the amount of the water drainage decreased on the day of earthquake and increased afterward for earthquakes #1 (Fig. 3a), #2 (Fig. 3b) and #5 (Fig. 3d) although the decreased and increased pattern was   not obvious for #5. The pattern was checked (see Methods) for the entire monitoring period and only those for #1 and #2 were detected.

Discussion
The Geospatial Information Authority of Japan provides daily coordinates of GPS targets in Japan from 1996. Therefore, it is not possible to investigate the GPS results for the earthquakes #1-#3 (1993)(1994). However, the change in daily coordinates for #4-#7 were shown in Fig. 4 and displacements due to the earthquakes were found for #5 (Fig. 4b), #6 (Fig. 4c) and Tohoku 2011 (Fig. 4e).
The displacements for #5 earthquake were clear and that in longitude shows an opposite motion, which might be related to the pre-seismic decrease in drainage, just before #5 (Fig. 5a).
For earthquake #7, even the largest permanent displacement in longitude was unclear (Fig. 5b). Displacements for Tohoku 2011 earthquake were mainly transient and far-field ones in which permanent displacements were very small. The above characteristics are in harmony with the drainage observation in which the decreased and increased pattern was not obvious but observed for #5 and that was not observed for #7 or Tohoku 2011 earthquakes. The plots of epicentral or hypocentral distance versus magnitude (Fig. 6) imply that large earthquakes tend to have greater effects, although the Tohoku 2011 earthquake neither affected the amount of water drainage nor caused the permanent surface displacements because it was far from the mine. From the plot, more distinct decrease and increase pattern was expected for #5. The unclear pattern for #5 would be due to the decrease of the sensitivity of the drainage amount to the crustal deformation as a consequence of the abandonment of drainage from Cretaceous formation. It is estimated that the change in the drainage amount would have been mainly caused rather by the drainage of the fossil water from Cretaceous than the groundwater from the surface.
Orihara et al. 2 postulated that the mechanism of the change in the water level and temperature at Goyo Hot Spa due to the Tohoku 2011 was a decrease in pore pressure due to dilatancy and the formation of new water paths    as a result of slow slippage before the main shock. The cause of the change in the drainage amount from Kushiro Coal Mine cannot be clarified at this stage because the 1-day time resolution of the drainage data as well as the GPS solutions are not small enough and also more data including geological, geo-hydrological, geochemical etc. are needed. However, it is estimated that a decrease in groundwater pressure due to dilatancy before the earthquakes and an opening of discontinuities due to stress relief after the earthquakes would have played a significant role as a part of the mechanisms for the changes in the drainage. The decrease in groundwater levels were observed a month and a half before Izu-Oshima-Kinkai earthquake 1 and three months before Tohoku 2011earthquake 2 . It would be very difficult to predict when earthquakes will occur from those observations. On the other hand, the decreases in drainage at Kushiro Coal Mine began on the day of the each earthquake and it would rather be easier for immediate predictions of large earthquake occurrences.
There are many deep mines in the world and it is expected that they can work as much larger sensors for pre-seismic changes of rockmass conditions by recording the amount of drainage etc. without spending huge costs than newly drilling small diameter drill holes from the ground. The authors fully recognize that the pre-seismic decrease itself cannot be used to predict earthquake at this stage and hope that data accumulation from observations at a shorter sampling time at deeper mines will contribute to clarify the drainage pattern at large earthquakes and to develop earthquake prediction methods in future.

Methods
Detection of the decrease and increase pattern. Firstly, the average values Q ave1 and Q ave2 for the period 14 of days prior and succeeding a day are defined (Fig. 7). Then the pre-seismic drainage decrease ΔQ 1 is defined as where Q p is the drainage amount on the day. The post-seismic drainage increase ΔQ 2 is defined as ΔQ 1 and ΔQ 2 were calculated for the each and every day in the monitoring period except for the first and last 14 days (Fig. 8), and the average values and standard deviations for ΔQ 1 and ΔQ 2 were calculated. The days which satisfied that they exceeded certain threshold values were selected as the days which showed the decrease and increase pattern. Setting the threshold values as the average value minus the standard deviation and the average value plus three times the standard deviation for ΔQ 1 and ΔQ 2 , respectively, it was found that only the days for #1 (Fig. 8b) and #2 (Fig. 8c) satisfied the threshold values for the entire monitoring period. The pre-seismic decrease is obvious for #1 but not for #2. This would be mainly due to the too slow time resolution. Faster sampling, for example, at each hour of drainage amount is desired for better results.

Data
Earthquake database was from the website of the Japan Meteorological Agency (http://www.data.jma.go.jp/svd/ eqdb/data/shindo/index.php). GPS solutions were downloaded from the Geospatial Information Authority of Japan (https://terras.gsi.go.jp/sso_login.php). The solutions were calculated using Bernese GNSS Software (http:// www.bernese.unibe.ch) from the whole raw GPS data for each UTC day and given as daily values. They can be considered as average daily coordinates of the site.