The results of subsurface gas monitoring by application of gas chromatography (GC) to the gas composition of bubbles associated with groundwater for seismogeochemical studies are reported. An automated gas monitoring system was used to determine gas compositions in a 500-m borehole at the Haruno Crustal Movement Observation Site (HOS), central Japan during period 1, from December 1999 to December 2000. The average ± two standard deviation (2SD) compositions of gases in this period were He = 82 ± 29 ppmV, H 2 = 170 ± 62 ppmV, Ar = 0.05 ± 0.07%, N 2 = 50 ± 8%, and CH 4 = 45 ± 6%. A new automated gas monitoring system equipped with a micro-GC was installed in the borehole at the HOS, and gas bubbles from the borehole were monitored during period 2, from December 2006 to March 2007. The average ± two standard deviation (2SD) compositions of gases in this period were He = 8 ± 7 ppmV, H 2 = 13 ± 15 ppmV, Ar = 0.6 ± 0.3%, N 2 = 66 ± 7%, and CH 4 = 14 ± 14%. The gas concentration ratios (He/Ar, H 2 /Ar, N 2 / Ar, and CH 4 /Ar) fluctuated significantly over time and repeatedly showed abrupt spike-like increases during period 2. The gas compositions obtained in period 1 and 2 were markedly different. Over the period from 2006 to 2007, the gas bubbles were depleted in He, H 2 , and CH 4 of deep origin, but enriched in Ar and N 2 of atmospheric origin. This difference can be interpreted as being due to an irreversible change of the aquifer/gas system. The present deep component in the HOS gas is estimated to have composition He = 63 ppmV, H 2 = 37 ppmV, Ar = 0.17%, N 2 = 63%, and CH 4 = 37%. The new monitoring system is able to analyze the gas composition using a smaller volume of sample gas and with greater precision than the previous system. During the 3-month monitoring period 2, the separation capacity of the capillary column of the micro-GC was sufficiently maintained to determine gas-chromatographic peak areas for the five gaseous species examined. This study confirms that the new monitoring system with micro-GC is promising for continuous subsurface gas monitoring for earthquake prediction studies.