Stable isotopes show that earthquakes enhance permeability and release water from mountains

The study investigates how large earthquakes alter regional hydrogeological systems, focusing on why groundwater levels rose following the 2016 Mw 7.0 Kumamoto earthquake. Prior observations show widespread coseismic hydrologic responses, with proposed mechanisms including pore-pressure changes from static strain, fluid migration along ruptures, permeability increases due to cracking and seismic vibrations, and liquefaction or consolidation. Stable isotopes of water (δD and δ18O) can fingerprint water sources, but past isotopic studies lacked spatial and temporal coverage to resolve processes at watershed scale. The research question is whether the observed post-seismic groundwater level rise was driven primarily by increased contributions of mountain aquifer waters enabled by permeability enhancement, rather than by soil porewater infiltration, vertical aquifer mixing, river recharge, or deep fluid upwelling. The purpose is to use a comprehensive, regionally distributed isotopic dataset to identify the sources of waters responsible for post-seismic groundwater level changes and to propose a process-based model applicable to volcanic aquifer systems in active tectonic settings.

Coseismic water level changes have been attributed to four main mechanisms: (1) pore-pressure responses to static elastic strain; (2) fluid migration along seismic ruptures; (3) permeability changes from cracking and seismic shaking; and (4) pore-pressure changes due to liquefaction or consolidation. Previous isotopic studies have examined pre- and post-earthquake waters, but limited sampling density prevented regional-scale interpretation. Reports elsewhere have documented continuous water-level rise after earthquakes linked to new water contributions via enhanced permeability, with potential pathways including increased soil porewater infiltration from the vadose zone, inter-aquifer mixing through breached aquitards, and augmented inputs from adjacent mountain aquifers. Additional observations from the Kumamoto sequence include increased upstream river discharge near mountains, suggesting possible permeability increases in the headwaters. However, a comprehensive isotopic assessment, including evaluation of deep fluids and liquefaction influences at regional scale, had been lacking.

Study area and hydrologic context: The Kumamoto region hosts volcanic aquifers (pyroclastics, lavas, and Quaternary alluvium) overlying relatively impermeable basement. Two main aquifer systems exist: an unconfined aquifer (< ~90 m depth) and an underlying confined to semi-confined aquifer (~20–200 m thick), separated by an aquitard. Regional groundwater is recharged mainly in northern and eastern highlands (50–200 m elevation), flows south- and westward, and discharges largely at Lake Ezu within ~40 years. Mountain aquifers around the Aso caldera rim and Mt. Kinpo discharge as springs at the mountain foot (~200 m) and at higher elevations (~400–620 m). The 2016 earthquake sequence (Mw 6.2 foreshock and Mw 7.0 main shock) produced extensive ruptures (e.g., Suizenji faults) that crosscut groundwater systems. Observed hydrologic responses included rapid groundwater drawdown (max 4.74 m within 35 min) followed by recovery and prolonged rise in recharge areas, peaking 4–5 months and persisting at least 3 years.

Sampling design and datasets: A comprehensive isotopic dataset (total n = 1150) was compiled to characterize pre- and post-earthquake waters across the watershed.

• Pre-earthquake (n = 872): 135 monthly precipitation samples (2005–2016) to define local meteoric water lines (high-water April–September; low-water October–March); 500 soil porewater samples from five unsaturated-zone borehole cores in recharge areas (2012–2014; averaged top 10 m per core); 45 Shira River samples (Apr–Jul 2011); 45 mountain spring samples (Apr–Jul 2011); 43 unconfined groundwater and 134 confined groundwater samples from monitored wells (Nov 2009–Nov 2011). Groundwater isotopic stability was evaluated from monthly discharge-area sampling (n = 70) showing annual variability < ±0.12‰ (δ18O) and ±0.5‰ (δD).
• Post-earthquake (n = 201): Rivers (n = 11; Aug 2016–Apr 2017), springs (n = 30; Oct 2016 and Mar–May 2017), and groundwaters (n = 160) sampled in four campaigns (Jun–Aug 2016; Oct–Nov 2016; Mar–May 2017; Nov–Dec 2017) at the same sites used pre-earthquake. Additional deep-sourced samples included hot springs (n = 23; 180–1300 m depth; Jul–Aug 2018) to represent potential deep fluids, and one mountain aquifer sample from an active tunnel (182 m b.g.s., 582 m a.s.l.; Oct 2017).

Analytical methods: Water samples (except soil porewaters) were collected in 20-ml glass vials; soil porewaters were extracted by centrifugation (pF 4.2). δD and δ18O were measured with a continuous-flow gas-ratio mass spectrometer (Thermo Fisher Delta V Advantage) at Kumamoto University; analytical precision better than ±0.5‰ (δD) and ±0.05‰ (δ18O). Cross-check analyses were performed using cavity ring-down spectroscopy (Picarro L2120i/L2130i) on groundwater samples from Sep 2015, Aug 2016, and Mar 2017 with precision ±0.5‰ (δD) and ±0.1‰ (δ18O). Local meteoric water lines for high- and low-water seasons were constructed from precipitation data and used as references.

Data analysis and source attribution: Pre-earthquake isotopic fields were established for precipitation, soil porewaters, river waters, mountain springs (foot vs. high-elevation), and groundwaters. Post-earthquake groundwater compositions were compared against pre-earthquake fields to detect shifts toward specific sources. Potential deep fluid signatures were constrained using hot spring isotopic fields, including waters with δ18O-enriched, δD-invariant characteristics indicating high-temperature water-rock interaction, and seawater mixing near the coast. Recharge elevations of spring waters were estimated using the regional δD–elevation regression δD = −0.0164·h − 39.153 (h in m a.s.l.) to fingerprint likely mountain water contributions. Seasonal effects were controlled by comparing same-season datasets (e.g., Oct–Nov) and by cross-instrument checks; observed post-seismic shifts were robust to season, aquifer type, and analytical method.

• Groundwater isotopic compositions across recharge and lateral-flow/discharge areas shifted after the earthquake from a broad, mixed-source field to a narrow range closely matching pre-earthquake mountain foot spring water, independent of season or aquifer (confined/unconfined), except in stagnant areas where no significant change was observed.
• Post-seismic groundwater did not shift toward soil porewater, river water, or deep-fluid isotopic fields. Thus, vertical infiltration from the vadose zone, river recharge, inter-aquifer vertical mixing via ruptured aquitards, or deep fluid upwelling were not the dominant sources for the widespread water level rise.
• Mountain spring waters define two elevation-linked groups: mountain foot springs (δD ≈ −49.2 to −42.6‰; δ18O ≈ −7.79 to −6.79‰) with estimated recharge elevations ~210–613 m, and high-elevation mountain springs (δD ≈ −55.3 to −49.1‰; δ18O ≈ −8.76 to −7.79‰) with recharge elevations ~607–985 m. Post-seismic groundwater compositions predominantly approached the mountain foot spring field, indicating release from mid-elevation mountain aquifers on the western Aso caldera rim and Mt. Kinpo.
• Hydrologic responses exhibited two stages: (1) Immediate drawdown up to 4.74 m within 35 minutes after the main shock due to rapid drainage into newly formed Suizenji fault cracks; (2) Subsequent and prolonged groundwater level rise in recharge areas, peaking 4–5 months after the main shock, with anomalies up to ~11 m and persisting at least 3 years, driven by increased permeability and mountain water release.
• Mountain spring and river discharge increased within 1 day of the earthquake, consistent with rapid permeability enhancement and drainage from mountain aquifers. Isotopic shifts of mountain foot springs toward slightly more depleted values imply added contributions from higher-elevation sources post-seismically.
• Lateral transport from recharge “groundwater pools” down-gradient contributed to recovery of water levels near rupture zones over the annual hydrologic cycle, implying post-seismic preferential flow pathways and faster transit than suggested by pre-existing age tracers (few to tens of years).
• Deep fluid and river contributions were detectable chemically or microbiologically in localized zones, but isotopically negligible in volume at regional scale. Occasional isotopic signals near fault zones indicate some river water routing along preferential pathways.
• In stagnant plains and coastal areas affected by liquefaction, isotopic compositions remained unchanged while some unconfined wells showed immediate coseismic water level rise, consistent with pore-pressure mechanisms rather than source mixing.
• Estimated released mountain water volumes on the order of ~10^8 m3 contributed to regional water level recovery via enhanced permeability pathways.

The isotopic fingerprinting demonstrates that the dominant driver of post-seismic groundwater level rise was the coseismic enhancement of permeability in mountain aquifers, which released stored waters from mid-elevations to down-gradient aquifers. This addresses the central question by ruling out major volumetric roles for soil porewater infiltration, deep fluid upwelling, or widespread vertical mixing across aquitards. The two-stage regional response—rapid drawdown into new rupture-related voids followed by sustained recharge from mountains—provides a coherent hydrogeological framework consistent with observed water levels, river/spring discharge increases, and prior hydrologic modeling. Isotopic evidence localizes the primary flow pathways to fracture systems near the base of the western Aso caldera rim and Mt. Kinpo rather than high-elevation sources alone, while slight depletion trends in mountain foot springs indicate contributions from higher elevations. Lateral transmission through preferential pathways explains rapid recovery downstream and implies transient acceleration of basin-scale flow relative to pre-event residence times. The findings highlight the role of earthquakes in dynamically reconfiguring permeability and redistributing stored mountain water, with implications for water resources, geochemistry, microbiology, and temperature regimes. Given the prevalence of volcanic aquifers in tectonically active arcs, similar mechanisms and hydrogeologic responses are likely elsewhere.

Using a uniquely comprehensive, regionally distributed dataset of δD and δ18O (n = 1150), the study shows that the 2016 Mw 7.0 Kumamoto earthquake enhanced permeability in surrounding mountain aquifers, releasing large volumes of water that caused sustained groundwater level rises (~11 m) in down-gradient aquifers. Isotopic shifts of groundwater toward mountain foot spring compositions, together with the absence of shifts toward soil, river, or deep-fluid fields, identify mountain aquifers as the dominant source of additional water. A two-stage model emerges: immediate drawdown into rupture-created voids followed by months-to-years of recharge from mid-elevation mountain aquifers through newly created or reactivated preferential pathways, with lateral propagation across the basin. These insights provide a generalizable framework for anticipating and interpreting coseismic hydrologic changes in volcanic arc settings. Future work should quantify spatial-temporal evolution of permeability changes, integrate isotopic tracers with geochemical, microbiological, and temperature monitoring, refine source apportionment at fault-scale using multi-isotope/solute approaches, and couple observations with physically based models to predict post-seismic water redistribution and resource impacts.

• Source attribution relies on stable isotopes (δD, δ18O), which are powerful for volumetric fingerprinting but may not resolve small contributions detectable by chemical or microbiological tracers; localized increases in soil porewater, river water, and deep fluids were observed chemically/microbiologically but were volumetrically minor isotopically.
• Deep fluid influence was evaluated using hot spring waters as proxies collected in 2018; these may not perfectly represent all deep fluid inputs during 2016–2017.
• Isotopic data constrain pathways and sources indirectly; fracture connectivity and permeability changes are inferred from isotopic shifts and hydrologic responses rather than measured directly in situ.
• Stagnant areas affected by liquefaction showed pore-pressure-driven level changes without isotopic shifts; processes there are not captured by isotopic mixing analysis.
• While the anomaly persisted at least 3 years, longer-term evolution of permeability and recovery was not resolved within the study period.