Thirty hydrological responses induced by the M8.0 Tokachi-oki earthquake in 2003 were observed at hot spring wells and an undersea coal mine in Hokkaido, Japan. Most of the decreases and increases in groundwater levels or discharge rates can be explained as a poroelastic response to earthquake-induced volumetric strain inferred from a fault model determined by dense static GPS observation. In five wells, observed groundwater-level changes and inferred volumetric strain steps induced by the Tokachi-oki earthquake are consistent with groundwater-level changes that are proportional to the inferred volumetric strain steps induced by four large earthquakes in 1993–1994.
 In this paper, we report hydrological responses in 29 hot spring wells, one observation well and a coal mine in Hokkaido associated with the M8.0 Tokachi-oki earthquake in 2003. Around this area, many hydrological anomalies were observed in response to four large earthquakes in 1993 and 1994 [Akita and Matsumoto, 2001]. Furthermore, a dense static GPS observation network recorded the detailed crustal deformation in Hokkaido area associated with the Tokachi-oki earthquake (Geographical Survey Institute, http://www.gsi.go.jp/WNEW/PRESS-RELEASE/2003/0926-2a.gif, 2003, hereinafter referred to as GSI, 2003). We show how hydrological anomalies respond to coseismic strain step inferred from the fault model determined by static GPS observation. We also compare the hydrological anomalies in response to the Tokachi-oki earthquake with those in response to the four large earthquakes in 1993 and 1994 to show the relationship between the hydrological anomalies and earthquake-induced volumetric strain steps.
2. Hydrological Data
 We collected hydrological responses from 30 wells and discharge rate from one undersea coal mine. The purpose of measuring the groundwater level or discharge rate of the wells is mainly to monitor hot spring resources. Figure 1 shows the location of the wells and coal mine. Table 1 gives a detailed description of the wells. Nineteen wells are observation wells to monitor the static groundwater level. Groundwater of two wells flows out without pumping, and that of nine wells is pumped out and used (Table 1).
Table 1. Description of Wells, Earthquake-Induced Volumetric Strain in Each Well at the Depth of 500 m, and Hydrological Response to the Tokachi-oki Earthquake in 2003a.
Depth of Well (m)
Screened Depth (m)
Graphs of coseismic hydrological changes are shown in the online auxiliary material. Qs: Quaternary sedimentary rocks, Qv: Quaternary volcanic rocks, Ps: Pliocene sedimentary rocks, Pv: Pliocene volcanic rocks, Ms: Miocene sedimentary rocks, Mv: Miocene volcanic rocks, PTs: Pre-Tertiary sedimentary rocks.
Observation Wells (Static Groundwater Level: cm)
Ps, Ms, PTs
Ps, Ms, PTs
Pumping Wells (Pumping Groundwater Level: cm)
Ms, Mv, PTs
Ps, Ms, PTs
Ps, Ms, PTs
Ps, Ms, PTs
Flowing Wells (Discharge Rates: l/min)
Undersea Coal Mine (Discharge Rates: l/min)
 Depths of the wells range from 50 m to 1500 m, and aquifers are confined. Groundwater levels and discharge rates are observed with accuracies of 5–10 mm and 0.1–0.5 l/min, respectively. Groundwater level or discharge rate at each well is recorded every 5 or 10 minutes except in KC and KS3 (once a day) and CR (once every 1–4 weeks). All data are stored at each observation site and collected every one month to one year except for KC and CR.
3. The Tokachi-oki Earthquake in 2003 and Hydrological Responses
 The Tokachi-oki earthquake occurred on 26 September 2003 at 4:50 JST. The epicenter, depth and magnitude of the earthquake were 41.7617°N and 144.0783°E, 42 km and M 8.0, respectively [Japan Meteorological Agency, 2003]. Geographical Survey Institute (GSI) inferred a fault model of the earthquake (Figure 1 and TAB-A1 of the online auxiliary material) by observing crustal deformation through its static GPS observation network (GEONET). GSI inferred that the moment magnitude of the earthquake was Mw 8.0 (GSI, 2003). The inferred fault model and distribution of aftershocks [Japan Meteorological Agency, 2003] show that the earthquake is an interplate thrust earthquake.
 Coseismic changes in groundwater levels or discharge rates associated with the Tokachi-oki earthquake were observed in 29 of the 30 wells and one coal mine (Table 1 and FIG-A1, FIG-A2, FIG-A3 and FIG-A4 of the online auxiliary material). Coseismic increases in groundwater levels and discharge rates were observed at wells in the southeast part of Hokkaido. The maximum increase in groundwater level was 430 cm in OB4 (Table 1). In KC, the total groundwater flow rate from the undersea coal mine changed from 2,400 l/min to 3,900 l/min. A 400 cm of coseismic increase in groundwater level was observed in the pumping well KY (Table 1 and FIG-A3 of the online auxiliary material).
 In the other area of Hokkaido, coseismic decreases in groundwater levels and discharge rates were observed except SK and YC in Hakodate city. The maximum decrease in groundwater level was −59 cm in AB. In HR, only an oscillatory change was observed, and there was no persistent change in the groundwater level.
4. Hydrological Responses and Coseismic Strain
Figure 1 shows the distribution of static volumetric strain induced by the Tokachi-oki earthquake using the fault model shown in TAB-A1 of the online auxiliary material. The strain distribution and strain steps at the wells (Table 1) are estimated by the computer program, MICAP-G [Naito and Yoshikawa, 1999] whose core routine was coded by Okada . We also show the observed coseismic increases and decreases of groundwater levels or discharge rates in Table 1. A contraction area is inferred in the southeast area of Hokkaido and poroelastically expected to cause a coseismic increase of groundwater level or discharge rate. Observed coseismic increases of groundwater levels and discharge rates at eight wells and a coal mine matched the calculated coseismic contraction in the southeast area of Hokkaido. In particular, the inferred coseismic contraction is 3.02 micro volumetric strain in CR, whereas the discharge rate became more than twice (225 l/min) its original (80–100 l/min) before the earthquake.
 In the other area of Hokkaido, coseismic dilatation is estimated by the fault model, and coseismic decreases in the hydrological responses are expected. There are 22 wells in this area, and observed groundwater levels and discharge rates in the wells are coseismic decreases except at HR, OB4, SK and YC. In OB1, the inferred coseismic dilatation is 1.90 micro volumetric strain, whereas the groundwater level decreased 1.3 m just after the earthquake.
 In conclusion, 12 coseismic increases and 18 decreases in groundwater levels or discharge rates were observed in 29 wells and one coal mine in response to the Tokachi-oki earthquake. Twenty-seven of the 30 hydrological anomalies can be explained as a poroelastic responses to the earthquake-induced contractional or dilatational strain.
Akita and Matsumoto  reported coseismic hydrological responses induced by the Kushiro-oki (Mw 7.7), Hokkaido-nansei-oki (Mw 7.7), Hokkaido-toho-oki (Mw 8.3) and Sanriku-haruka-oki (Mw 7.6) earthquakes in 1993–1994 (Figure 2 and TAB-A2 of the online auxiliary material). They also illustrated the relationship between the groundwater-level steps and coseismic volumetric strain steps in eight wells and showed that the coseismic steps in the groundwater level were proportional to the coseismic strain steps.
 We added the observed changes in groundwater levels and the inferred volumetric strain steps in response to the Tokachi-oki earthquake at OB1, SP1, SP2, SR1 and YN for figures made by Akita and Matsumoto . We newly collected changes in groundwater level in response to EQ1 – EQ4 at OB4. Figure 3 shows the relationship between the groundwater-level steps and inferred strain steps at OB1, OB4, YN, SR1, SP1 and SP2 wells. The observed coseismic changes in groundwater level at OB4 have completely different sign from inferred ones from poroelastic theory. The possible reasons may be flow from shallow aquifer to screened aquifer or leakage from shallow aquifer to the well, because water temperature decreased after the Tokachi-oki earthquake at OB4. On the contrary, relationship in OB1, YN, SR1, SP1 and SP2 wells between the observed groundwater-level changes and inferred strain steps in response to the Tokachi-oki earthquake support that coseismic groundwater-level changes are proportional to the coseismic strain steps in the five wells.
 We inferred strain sensitivities at OB1, OB4, YN, SR1, SP2 and SP2 wells using M2 tidal constituent (period: 12.4206 hours) in order to compare strain sensitivity determined by coseismic water-level changes (Figure 3). We used the computer program GOTIC2 [Matsumoto et al., 2001] to calculate theoretical M2 amplitude and phase shift of volumetric strain produced by earth tide and oceanic tidal loading. Responses of groundwater level to M2 constituent at the six wells are determined by the computer program BAYTAP-G [Tamura et al., 1991].
Table 2 shows strain sensitivities estimated by M2 tidal constituent and by coseismic response. The strain sensitivities estimated by M2 at OB1, SP1 and SP2 wells are plausible, because extracted phase shifts in groundwater level tw are almost the same as theoretical phase shifts in volumetric strain tt. Furthermore, strain sensitivities determined by coseismic response are within 2.6 times as large as that by using M2 at OB1, SP1 and SP2 wells. These results support that coseismic changes in groundwater level at OB1, SP1 and SP2 wells are proportional to coseismic strain steps.
Table 2. Calculated Amplitudes and Phase Shifts of M2 Volumetric Tidal Strain, Responses of Water Level to M2 Tidal Constituent, Wells' Strain Sensitivities Using M2 Tidal Constituent and Inferred Wells' Strain Sensitivities Using Coseismic Strain Steps (Figure 3)a
Amplitude (10−8) [Phase Shift (deg.)]
Negative phase shifts denote phase lags. “–” denotes that we could not infer well's strain sensitivity using coseismic strain step.
Vol. strain by M2 earth tide, te
Vol. strain by M2 oceanic tidal loading, to
Vol. strain by earth + ocean tide, tt = te + to
M2 amplitude of water level, tw
8.25 ± 0.13 [−6.4 ± 0.9]
3.42 ± 0.10 [−6.7 ± 1.7]
2.64 ± 0.43 [−5.2 ± 9.2]
18.70 ± 1.41 [72.7 ± 5.6]
47.07 ± 0.22 [48.1 ± 0.3]
7.66 ± 0.97 [−8.4 ± 7.2]
Strain sens. by M2 tide, ws = tw/tt (mm/10−8)
Strain sens. by coseismic strain (mm/10−8)
 On the other hand, extracted phase shifts in tw at SR1 and YN wells differ from those in tt (Table 2). The reason of discrepancy in phase shift is that distances from SR1 and YN wells to the coast are only 900 m and 50 m, respectively. The strain sensitivity by M2 might be overestimated because groundwater level at SR1 and YN wells might be affected by fluid pressure of the ocean and flow from the ocean that GOTIC2 cannot estimate. Nevertheless, strain sensitivities determined by using M2 are only 2.5–4 times larger than those by coseismic response at SR1 and YN wells. The relationship may roughly support that coseismic changes in groundwater level are proportional to coseismic strain steps because we recognize the overestimation of strain sensitivity using M2 in SR1 and YN.
Sato et al.  described groundwater-level changes associated with the Tokachi-oki earthquake in 42 wells. Forty of the 42 wells are located more than 1000 km away from the earthquake, and other two wells are located near AB and NW. They observed coseismic decreases in the groundwater level in these two wells, that were consistent with our observation.
 Hydrological anomalies at 29 wells and one coal mine in Hokkaido Prefecture were observed after the M8.0 Tokachi-oki earthquake in 2003. Twenty-seven of the 30 changes in groundwater levels or discharge rates can be explained as a poroelastic responses to inferred volumetric strain after the Tokachi-oki earthquake.
 Observed groundwater level steps associated with the Tokachi-oki earthquake and four large earthquakes in 1993–1994 around the Hokkaido area are proportional to the inferred earthquake-induced coseismic strain steps in OB1, YN, SR1, SP1 and SP2 wells. Strain sensitivities determined by coseismic responses in groundwater level are consistent with those estimated by M2 tidal strain in OB1, SP1 and SP2 wells.
 Tomo Shibata and Tetsuya Takahashi collected hydrological data in several wells. The Department of Health and Welfare of Hokkaido Government, Japan Nuclear Cycle Development Institute and many municipalities in Hokkaido provided the data. We used the software package GMT [Wessel and Smith, 1998] for some of the figure. We thank two anonymous referees for their valuable comments.