Groundwater and Stream Composition

Abstract

Changes of groundwater chemistry have long been observed. We review some studies of the earthquake-induced changes of groundwater and streamflow composition. When data are relatively abundant and the hydrogeology is relatively simple, the observed changes may provide valuable insight into earthquake-induced changes of hydrogeological processes. Progress in this aspect, however, has been slow not only because systematic measurements are scare but also because of the distribution of chemical sources and sinks in the crust are often complex and unknown. Most changes are consistent with the model of earthquake-enhanced groundwater transport through basin-wide or local enhanced permeability caused by earthquake-induced breaching of hydrologic barriers such as aquitards, connecting otherwise isolated aquifers or other fluid sources, leading to fluid source switching and/or mixing. Because the interpretation of earthquake-induced groundwater and stream compositions is often under-constrained, multi-disciplinary approaches may be needed to provide a better constrained interpretation of the observed changes.

The water in wells is also more turbid than usual, and it emits a disagreeable odour. (Pliny the Elder, Natural History, Chapter 83).

9.1 Introduction

Changes in the turbidity, odor and taste of groundwater are probably among the earliest reported changes following earthquakes, as Pliny the Elder noted in his Natural History nearly two thousand years ago. Such changes and, more generally, the change of the chemical composition of groundwater, may be expected because earthquake-induced groundwater flow is effective in transporting solutes, and earthquake may open new passageways to connect fluids from different reservoirs to facilitate such exchange.

Progress in our understanding of these processes, however, has been slow partly because chemical measurements require discrete sampling of water and laboratory analyses are expensive and time-consuming. Hence quantitative data and systematic measurements for groundwater composition are even less abundant than that for groundwater temperature. In addition, the distribution of chemical sources and sinks in the crust is complex and largely hidden from observation, making the interpretation of the measured changes challenging.

On the other hand, in the cases where the chemical signatures of earthquake-induced changes are clear and well documented, they may provide evidence for the origin of the extra water released by the earthquakes, as will be shown later in this chapter. Significant progress has been made in the last decade to expand the observational record. Interesting are some apparent contradictions between the observations from nearby wells. Understanding these differences and contradictions will require greater effort in data collection and analysis.

This chapter reviews some significant observations on earthquake-induced changes of groundwater composition, particularly those made in the past decade, and their interpretation. Most observations were made in groundwater wells and natural springs because their relatively confined environment minimizes the influence of surface waters. The presence of flowing waters in streams prior to earthquakes makes similar studies in streams more difficult. But in regions such as California with extended dry seasons, earthquakes may occur when many streams have little flow or are dry. Under these circumstances, studies of stream water composition before and after the earthquake may provide valuable information about the earthquake’s effects on local hydrogeologic processes. For the convenience of discussion, we separate the discussion of the composition changes in stream water from those in groundwater wells and natural springs. When data are abundant and complex, we also categorize them into changes of the major elements, the trace elements, and the stable isotopes of oxygen and hydrogen. Many studies made measurements only after earthquakes; these are not reviewed here because the study of earthquake-induced changes requires measurements both before and after an earthquake.

Due to the scarcity of quantitative data and the complexity in the distribution of chemical reservoirs, interpretations of groundwater composition are often under-constrained. We end the chapter with a discussion on the need to integrate observational data for multiple types  of groundwater responses in constraining the interpretation of the earthquake-induced chemical compositions.

9.2 Groundwater Composition

9.2.1 Major Elements

More than 90% of the dissolved solids in groundwater can be attributed to eight ions, Na⁺, Ca²⁺, K⁺, Mg²⁺, ${\text{SO}}_{4}^{2 - }$, Cl^–, ${\text{HCO}}_{3}^{ - }$, and ${\text{CO}}_{3}^{2 - }$. Silica, SiO₂, is also often present as a nonionic species. Variations among the relative proportions of these ions in the groundwaters of different regions reflect their different sources.

Most earthquake-related studies of groundwater composition have focused on the search for precursory changes, and we discuss these in more detail in Chap. 13. In a few cases, however, the composition of groundwater was systematically monitored before and after earthquakes. Such changes can provide useful constraints on models of earthquake-induced groundwater flow. In this section we discuss the results of recent studies on earthquake-induced changes of the major element concentration in water in some wells and natural springs.

9.2.1.1 Iceland

The Tjӧrnes Fracture Zone (TFZ) in northern Iceland (Fig. 9.1a) is a transform fault zone that offsets segments of the Mid-Atlantic Ridge and is seismically active. Borehole HU-01 (Fig. 9.1b) is 1500 m deep and is near one of the fault zones. In the upper 1150 m, the basalt horizons are intercalated with sediments; below 1150 m, basalt dominates. Water inflow in this borehole occurs at depths of 500 m, 700 m, 730 m (sandstones), and 1220 m (fractured basalt) below the surface. Borehole HA-01 (Fig. 9.1b) is 101 m deep and water inflow occurs at 65 m, 81.5 m, and 95.7 m. Weekly samples were collected from the HU-01 well starting July 2002 (Claesson et al. 2004, 2007) and from the HA-01 well starting 2008 (Skelton et al. 2014; Andrén et al. 2016).

Fig. 9.1

(modified from Andrén et al. 2016)

a Map of Iceland showing the geographic locations of the epicenters of the studied earthquakes (red stars), the Mid-Atlantic Ridge (red dashed lines) and transform faults (black lines). Yellow areas mark the active volcanic systems. b Zoom in of the study area (Tjörnes peninsula) where the boreholes (red circles) are located

A M5.8 earthquake occurred on 16 September 2002 off the northern coast of Iceland, with epicenter ~90 km north of Husavik (Fig. 9.1a, b). Two consecutive earthquakes occurred on 21 October 2012 (M 5.5) and 2 April 2013 (M 5.3) (Fig. 9.1a); both are ~76 km from Husavik.

Figure 9.2 shows the extended time series of the major element compositions in the HU-01 and HA-01 wells from 2002 to 2018 (Skelton et al. 2019). Much discussion on the changes of water composition in these wells has been focused on the premonitory changes, which we review in Chap. 13. Here we only discuss the changes of water chemistry after the earthquakes. Claesson et al. (2004) reported that after the 2002 M5.8 earthquake, groundwater in the HU-01 well showed increased concentrations of Na⁺, Ca²⁺, K⁺, S, Si, Cl^–, and ${\text{SO}}_{4}^{2 - }$ by 12–19% in and decreased Na/Ca. Claesson et al. (2007) updated the earlier study and found that the chemical changes caused by the M5.8 earthquake recovered gradually over the subsequent two years before the trend was interrupted by a second rapid rise caused by other earthquakes near the end of 2004. The extended time series in Fig. 9.2 shows, however, that most of the changes in well HU-01 before 2002 and 2004 earthquakes were not repeated during the 2012 M5.5 or the 2013 M5.3 earthquakes, except those of Cl^– and ${\text{SO}}_{4}^{2 - }$. Also interesting are the differences between the chemical responses between the HU-01 well and the HA-01 well. The latter well shows no change in the concentration of K⁺ and only small changes in Ca²⁺ and Si, even though it shows a sharp increase in the concentration of Na, similar to the HU-01 well after the 2002 M5.8 earthquake. The greatest contrast between the two wells occurred in the concentrations of Cl^– and ${\text{SO}}_{4}^{2 - }$, which increased in the HU-01 well after the earthquakes but decreased in the HA-01 well. The reason for these differences is unknown, but is likely to be related to the different groundwater sources for the two wells, as revealed by the large differences between their background compositions (Fig. 9.2).

Fig. 9.2

(modified from Skelton et al. 2019)

Time series of weekly samples from 2002 to 2018, of a Ca²⁺, b K⁺, c Na⁺, d Si, e Cl⁻, f F⁻, and g ${\text{SO}}_{4}^{2 - }$ at boreholes HU01 (blue circles) and HA01 (green circles). The 2002, 2012, and 2013 earthquakes are marked by vertical dashed lines. Analytical errors are <2%

9.2.1.2 Japan

Groundwater chemistry has been continuously measured in central Kyushu long before the 2016 Mw7.0 Kumamoto earthquake, Japan. Before the earthquake, groundwater composition was relatively constant, but changed subtly afterwards (Fig. 9.3; e.g., Hosono et al. 2018; Hosono and Masaki 2020; Nakagawa et al. 2020). These authors found increased dissolved silica in many wells, similar to the findings in Iceland (Sect. 9.2.1.1). Contrary to the findings in Iceland, however, they found decreased concentrations of Cl^–, F^–, Na⁺, K⁺, Ca²⁺. They interpret these decreases to be caused by dilution from groundwater released from the surrounding mountains. Increased concentrations of nitrates, ${\text{SO}}_{4}^{2 - }$, and Mg²⁺ were interpreted to be due to leaching of contaminants and agricultural fertilizers from soils and sewage water pipe breaks. Finally, increases of Fe_total and Mn_total was interpreted to be due to leaching of marine clay by liquefaction in coastal areas.

Fig. 9.3

(modified from Nakagawa et al. 2020)

Map showing the locations of wells sampled by Nakagawa et al. (2020) in the Kumamoto area after the 2016 earthquake. Different ‘clusters’ of wells showed distinct chemical changes. Cluster 1 wells showed increased SiO₂ and ${\text{NO}}_{3}^{ - }$, but decreased F^–, Cl^–, Na⁺, K⁺, and Ca²⁺ concentrations. Cluster 2 wells showed increased SiO₂, ${\text{NO}}_{3}^{ - }$, Cl^–, ${\text{SO}}_{4}^{2 - }$, and Mg²⁺, but decreased F^– concentration. Cluster 3 wells displayed increasing Fe_total, Mn_total, and SiO₂, but sharp decreases of Cl^– and Na⁺ concentration. Cluster 3 wells showed increased Fe_totaland Mn_total concentrations. Cluster 4 wells showed decreased Fe_total and Mg concentrations. Cluster 5 wells showed increased SiO₂, ${\text{SO}}_{4}^{2 - }$, and Mg²⁺ concentrations but decreased Cl^– concentration. Cluster 6 wells showed increased ${\text{SO}}_{4}^{2 - }$, F^–, and Mg²⁺; other chemicals were basically unchanged

Hosono et al. (2018) also reported that new spring water inside the Aso caldera after the 2016 Kumamoto earthquake (Fig. 9.4) was characterized by high sulfate content and low lithium and boron stable isotope ratios (δ⁷Li = 2.6‰ and δ¹¹B = 1.4 to 2.6‰), consistent with meteorically-derived groundwater admixed with deeper fluids of hydrothermal origin. The occurrence of the new highly saline fluids in the northwestern plain where the earthquake caused major surface ruptures (Hosono et al. 2019) led the authors to suggest that these elevated hydrochemical fluxes migrated upward to the surface from greater depths along earthquake-generated fractures.

Fig. 9.4

a Maps showing the sampling sites in the Aso caldera watershed and dominant extensional fissures (black dots with arrows), and associated horizontal landsliding (area bounded by the black dashed curve) observed after the 2016 Kumamoto earthquake. b Map showing earthquake epicenters in the Aso caldera watershed before (1923–2016: black circles) and after the 2016 main shock (red circles). c Map showing spring and groundwater water chemistry. d Map showing river water chemistry based on data from 1968–1995. Locations of reported low resistivity zone for hypothesized melt finger in the deep crust and magma chambers beneath central volcanoes are shown with a yellow dotted curve and red dotted curves, respectively (from Hosono et al. 2018)

Koizumi et al. (2019) analyzed the composition of eleven springs in central Kyushu (see Fig. 9.5 for spring locations) after the 2016 Kumamoto earthquake and compared the results with the compositions before the earthquake. They found little change in the major element composition of the studied springs (Fig. 9.5); only the concentration of ${\text{NO}}_{3}^{ - }$ changed slightly just after the earthquake, which they attributed to leakage from surface sources.

Fig. 9.5

Chemical composition at eight springs (see Fig. 7.5 for spring locations) before and after the 2016 Kumamoto earthquake. Sampling date is shown above each hexa diagram. The precision of measurements is from a few percent to 10% (from Koizumi et al. 2019)

Ide et al. (2020) also compared the concentrations of major elements in many springs in central Kyushu after the 2016 Kumamoto earthquake with those measured in the year of 2009. They also found no remarkable difference between the water compositions before and after the earthquake (Fig. 9.6).

Fig. 9.6

(modified from Ide et al. 2020)

a Map showing the locations of studied natural springs in central Kyushu by Ide et al. (2020). Comparison of b major anion concentrations and c major cation concentrations in springs between the 2009 and 2017 sampling campaigns, showing no remarkable change of the major dissolved ions

The different conclusions by Hosono et al. (2018) and those by Koizumi et al. (2019) and Ide et al. (2020) may reflect the fact that the springs studied by Hosono et al. (2018) are inside the Aso Caldera, while most springs studied by Koizumi et al. (2019) and Ide et al. (2020) are outside the caldera and thus sample different groundwater sources.

9.2.1.3 Korea

Two consecutive earthquakes of M5.1 and M5.8, with strike-slip focal mechanisms and separated by less than an hour (Fig. 9.7), occurred on September 12, 2016, on the SE of the Korean peninsula. The second of these is the largest instrumentally recorded earthquake on the peninsula. Significant hydrological responses were reported (Kim et al. 2019; Kaown et al. 2020; Lee et al. 2020) and we discussed in Chap. 8 the response of the groundwater temperature to this earthquake. Here we discuss the response of the groundwater composition. Most measurements of groundwater composition took place after the earthquakes (Kim et al. 2019; Kaown et al. 2019). Fortunately, some wells (Fig. 9.7a) in the Korean national groundwater monitoring network measured the major element compositions both before and after the earthquake.

Fig. 9.7

(modified from Kaown et al. 2019)

a Locations of the studied area, the epicenter of the 2016 M5.8 earthquake, and the studied wells in SE Korea. b Concentrations of Na⁺, Ca⁺², ${\text{HCO}}_{3}^{ - }$ and ${\text{SO}}_{4}^{2 - }$ in the waters from the studied wells from 2014 to 2017. Measurements in the first three years were made before the earthquake; the one in 2017 was made after the earthquake

The hydrogeology of the studied area may be simply described as an alluvial sedimentary basin overlying a basement of Miocene to Cretaceous sedimentary formations and igneous rocks. At each of the groundwater stations (Fig. 9.7a) two wells are installed, one in the upper alluvial sediments, and the other in the deeper basement rocks. Figure 9.7b shows the composition of some major ions measured at different times. Measurements labeled 2014–2016 were made before the earthquake, while those labeled 2017 were measured after the earthquake. No clear changes of groundwater composition can be recognized in wells installed in the shallow sediments after the earthquake (left column of Fig. 9.7b). On the other hand, some clear changes of groundwater composition after the earthquakes were measured in wells installed in the basement rocks (right column of Fig. 9.7b). These include consistent post-seismic increase of the concentrations of Na⁺, Ca⁺², ${\text{HCO}}_{3}^{ -}$ and ${\text{SO}}_{4}^{2 - }$ (well KW5-2), consistent post-seismic decrease in the concentrations of the four same ions (well KW7-2), and clear post-seismic increases in Na⁺ and Ca⁺² (well KW6-2), but no change in ${\text{HCO}}_{3}^{ - 1}$ or ${\text{SO}}_{4}^{2 - }$. The inconsistent changes of groundwater composition after the earthquake among these wells suggest that the wells may be opened to different and isolated fractures in the basement rocks. They also illustrates the difficulty in understanding earthquake-induced composition changes.

9.2.2 Trace Elements

There is no universally agreed definition for ‘trace elements’ but the term is usually used for elements with concentrations below 0.1%. In this sense, whether an element may be a trace element or not depends on the lithology of the rock. For example, while Si is the most abundant element in beach sand, it may occur only in trace amounts in some carbonate aquifers.

9.2.2.1 Italy

Rosen et al. (2018) measured the chemical composition of four springs in the central Apennines of Italy before and after several earthquakes in 2016 to 2017. They found four springs at varying distances from the epicenters that all showed immediate post-mainshock increases in trace element concentrations but little change in major elements.

These springs are recharged by major aquifers hosted in Cenozoic to Mesozoic carbonate rocks that overly an evaporitic basal structure. The aquifers are characterized by two major flow paths: a shallower one with high flow rate and a deeper one with low to medium flow rates where seepage of deep mineralized fluid occurs along fractures. This dual-flow structure allows waters with different residence times to occur in the same aquifer.

The authors suggest that fluids enriched in trace elements may have been stored in fractures with slow flow and hence have long residence times. These fluids were expelled into the main flow paths after the earthquakes due to increased pore pressure and enhanced permeability. Rosen et al. (2018) also noticed that the response of the compositions of these trace elements to the later earthquakes is much weaker than that to the first (Fig. 9.8). They attributed the weaker response during the later earthquakes as the result of progressive depletion of the high solute fluids as the earlier shocks flushed such fluids stored in the fractures.

Fig. 9.8

(modified from Rosen et al. 2018)

a Location map of sampling sites of springs (blue circles) and epicenters of major earthquakes (stars). Colored contours show recharge areas for different springs. Red lines show surface traces of active faults. Time series of b trace metal concentrations and c major elements concentrations in four springs in central Apennines: Nerea spring (NER), Santa Susanna spring (SUS), Vicenna Riara spring (VIC), and Peschiera spring (PES). Vertical lines mark the times of the four major earthquakes; horizontal gray bands show pre-seismic values of Al, Cu, Pb, and Mn, where all springs had the same range. Dashed horizontal lines show the pre-seismic earthquake values of Sr and Rb at each individual spring

9.2.2.2 China

For completeness we also include a study of the change of concentrations of the rare earth elements. Shi et al. (2020) studied the concentrations of both the major elements and the rare earth elements in a groundwater well in SW China before and after a nearby M5.0 earthquake (Fig. 9.9a). The well is located in a tectonic graben bounded by active strike-slip faults and opens to an aquifer hosted in a sequence of Sinian sandstones and shales, which is confined above by Quaternary lacustrine clay, sands and gravels. Shi et al. (2020) found that the concentrations of the rare earth elements in the groundwater increased sharply after the earthquake (Fig. 9.9b). On the other hand, the concentrations of the major elements (Fig. 9.9c) and the δD and δ¹⁸O in this well showed no clear responses, similar to the findings of Rosen et al. (2018) in the central Apennines of Italy.

Fig. 9.9

a Location map of the Jiangchuan well and the epicenter of the 2018 M5.0 earthquake. Red lines show surface traces of active major fault zones. Time series of groundwater composition in the Jiangchuan well, SW China, of b the rare earth elements and c the major elements. The vertical arrows show the time of the 2018 M5.0 earthquake (from Shi et al. 2020)

Shi et al. (2020) attributed the different responses of the rare earth elements and the major elements in the SW China well to the small background noise in the measurements of the rare earth elements and the large background noise in the measurements of the major ions. Here we suggest the model proposed by Rosen et al. (2018) of a multi-flow system as an alternative interpretation. Because the Jiangchuan well is located in a geothermal area and its water temperature is ~34 °C (Shi et al. 2020), the well water is likely to be a mixture of shallow groundwater and deep geothermal waters that flow into the well by seepage from conductive fractures. The nearby 2018 M5.0 earthquake may have enhanced the fracture permeability, leading to an increased geothermal flow that in turn may have transported rare earth elements into the well but insufficient fluid to affect water isotopes and major elements.

9.2.2.3 Iceland

Claesson et al. (2004, 2007) reported the changes of some trace element concentrations (B, Li, Sr, Rb, Mo; Fig. 9.10a ) in the groundwater of borehole HU-01of Iceland (see Fig. 9.1b for well location) following a 2002 M5.8 earthquake (see Fig. 9.1a for epicentral location) and other smaller earthquakes in 2004–2005. The trace element concentrations show coseismic increases with the same pattern as those for the major elements.

Fig. 9.10

a Plot of water chemistry over time for groundwater samples from borehole HU-01, showing the percentage shifts after earthquakes in the concentrations of B, K, Li, Mo, Na, Rb, S, Si, Sr, Ca, Cl and SO₄. The timing of M > 5, M > 4 and M > 3 earthquakes are marked by red, green and blue lines, respectively. b δD versus δ¹⁸O for groundwater samples from borehole HU-01 showing data from 3 July 2002 to 18 September 2002 (grey circles), 25 September 2002 to 13 October 2004 (black circles) and 24 November 2004 to 6 January 2005 (white circles). The abrupt hydrogeochemical shift, which occurred within 2–9 days after the M 5.8 earthquake on 16 September 2002, and its recovery during the subsequent two years indicate that switching between or mixing of aquifers is the primary hydrogeochemical control. GMWL is the Global Meteoric Water Line (from Claesson et al. 2007)

The observation by Claesson et al (2004) is in contrast with the observations in the central Italian Apennines (Rosen et al. 2018) and in SW China (Shi et al. 2020) where only the trace element concentrations increased but those for the major elements remained unchanged. This contradiction is another example that shows that the interpretation of groundwater composition changes is challenging. Different types of data, in addition to the chemical data, may be required to better constrain the problem.

9.2.3 Stable Isotopes

Geologic processes often cause changes in the relative proportions of the stable isotopes of oxygen and hydrogen in groundwater. An important tracer to characterize the origin of the excess water is the isotopic composition of the post-seismic excess discharge. The differences in ¹⁸O and D relative to VSMOW emerge from the fractionation processes during the transport and precipitation of water vapor. The combination of decreasing temperature and increasing rainout with elevation results in water that is isotopically lighter, i.e., depleted in the heavier isotope, and the rate of this decrease with elevation is reported as the δ¹⁸O and δD lapse rates. Thus, water released from consolidation and from the pore waters in unsaturated soils would have a local isotopic signature, while water originating from high mountains would have a lighter isotopic signature, and water released from mid-crustal depths would have an isotopic signature with evidence of high temperature water-rock interaction (resulting in δ¹⁸O isotopic shifts towards enriched compositions leaving δD unchanged). Thus, the change in the stable isotope composition after an earthquake may be used to infer for the source of the new groundwater. In this section we review examples in Iceland, Taiwan and Japan, where interesting findings and significant conclusions have been obtained.

9.2.3.1 Iceland

The 2002 M5.8 earthquake in Iceland (Fig. 9.1a) caused changes in the oxygen and hydrogen isotope ratios of groundwater in well HU-01 (see Fig. 9.1b for well location). Claesson et al. (2004, 2007) reported these changes (Fig. 9.10b) and differentiated between two models for the changes of δ¹⁸O and δD in groundwater during and after earthquakes. In the first model, accelerated water-rock reactions are caused by an assumed increase in fresh mineral surfaces exposed to groundwater along newly formed cracks and fractures created by the earthquake, leading to rapid changes in groundwater composition. In the second model, rapid change in groundwater composition results from fluid-source switching or mixing of groundwater from a newly tapped aquifer containing chemically and isotopically distinct water, probably caused by unsealing of pre-existing faults and breaching of hydrologic barriers. The first model predicts that δD and δ¹⁸O of groundwater after the earthquake would move away from the Global Meteoric Water Line (GMWL), i.e., along the single-arrow light-gray curve at the bottom of Fig. 9.10b. The second model, on the other hand, predicts that δD and δ¹⁸O of the groundwater after the earthquake would change in a direction parallel to the Global Meteoric Water Line (GMWL), i.e., along the double-arrow path shown in Fig. 9.10b (Claesson et al. 2007).

The three sets of data for Iceland water samples, one collected shortly before and two after the M5.8 earthquake, show that the changes in δD and δ¹⁸O of the groundwater after the earthquake are nearly parallel to the GMWL (Fig. 9.10). The data are thus consistent with the model of source switching and/or mixing of groundwater with a newly tapped aquifer, but inconsistent with the model of accelerated water-rock reactions due to increased fresh mineral surfaces.

9.2.3.2 Taiwan

Groundwater samples were collected from a network of monitoring stations on the Choshui River fan for oxygen isotope analysis (small circles in Fig. 9.11) before and after the September 1999 Chi-Chi earthquake (Wang et al. 2005). At each station, two to five cluster wells were installed into different aquifers (Fig. 9.12) and the distribution of the isotopic composition of groundwater in each aquifer was determined independently. The upper five diagrams of this figure show the isotope compositions in the topmost aquifer (Aquifer I; see Fig. 9.12) before and after the Chi-Chi earthquake and their differences at several time intervals after the earthquake; the lower five diagrams of this figure show similar compositions in the third aquifer (Aquifer III; see Fig. 9.12). Before the Chi-Chi earthquake the distribution of δ¹⁸O in Aquifer I (Fig. 9.11a) increased from ~−5‰ near the coast to ~−8‰ near the upper rim of the alluvial fan. Shortly after the earthquake (October to December 1999), δ¹⁸O became more negative (Fig. 9.11b). The differences between Fig. 9.11a, b are given in Fig. 9.11c to show the earthquake-induced changes. Since the seasonal variation of δ¹⁸O is less than 0.4‰, all the changes more than 0.4‰ can be attributed to the earthquake; these areas are colored green to highlight the earthquake effect. The affected area covers a broad zone near the coast on both sides of the Chishui River, with a change up to 1‰. Changes persisted to the end of this analysis, nearly two years after the earthquake, even though the area and magnitude of the increased depletion slightly diminished over time (Fig. 9.11d, e). Wang et al. (2005) attributed this depletion in Aquifer I to an increased contribution from the Choshui River that discharges depleted δ¹⁸O water from the high mountains on the east.

Fig. 9.11

Oxygen isotope contours for Aquifer I in the Choshui alluvial fan from 1999 to 2001. The open circles are the sampled groundwater stations. At each station, several wells were installed in different aquifers. Red color denotes areas where δ¹⁸O values was below −10‰; green color denotes areas where δ¹⁸O was depleted by more than 0.4‰. a Absolute δ¹⁸O values in January–March, 1999, before the Chi-Chi earthquake. b Absolute δ¹⁸O values in October–December, 1999, shortly after the Chi-Chi earthquake. c Difference between (b) and (a). d Difference between measurements made in January-to-August, 2000, and (a). e Difference between measurements made in May–July 2001 and (a). The scale bars are in km. (a’) to (e’), similar to (a) to (e) but for Aquifer III (from Wang et al. 2005)

Fig. 9.12

(modified from Wang et al. 2005)

Hydrogeologic cross-section of the Choshui River alluvial fan. The thick vertical dashed lines represent boreholes. The numbers represent the measured δ¹⁸O in units of per mil for each sampling site after the Chi-Chi earthquake. The squares represent sites that had decreased oxygen isotope values after the earthquake, while the triangles represent sites that had increased oxygen isotope values. The double arrows indicate aquifers whose δ¹⁸O values converged to the same value after the earthquake

However, Aquifers II and III showed areas of depleted δ¹⁸O even before the earthquake; only the data from Aquifer III are shown here (Figs. 9.11a’–e’). Figure 9.11a ’ shows that before the earthquake, there was a large area of depleted δ¹⁸O on the northside of the Choshui River, probably due to recharge of groundwater from a higher elevation in the mountains to the east, transported to the aquifer through a subsurface abandoned river channel (Chang 1983). Shortly after the Chi-Chi earthquake (October–December 1999), the region of depleted δ¹⁸O expanded (Fig. 9.11b’), and the earthquake-induced change (the difference between Fig. 9.11a, b) is given in Fig. 9.11c’. Comparing Fig. 9.11c and c’ shows that the areas with a difference of δ¹⁸O more than 0.4‰ are nearly identical in these two aquifers. The area of depletion in Aquifer III also slowly diminished over time and persisted for >2 years to the end of the study (Fig. 9.11e’), similar to Aquifer I.

The spatial coincidence of locations with more than 0.4‰ depletion of δ¹⁸O in Aquifer I and Aquifer III led Wang et al. (2005) to suggest that there was vertical mixing of groundwater between different aquifers after the Chi-Chi earthquake. Figure 9.12 that shows the hydrogeological cross-section of the Choshui River fan, in which aquifers with a convergence of the δ¹⁸O values after the earthquake are connected with doubleheaded arrows. The suggestion of vertical mixing of groundwater between different aquifers is supported by subsequent studies that showed that the confinement between these aquifers was breached during the earthquake, which caused the initially different groundwater level in some aquifers to converge to the same level (Wang 2007) and their post-seismic tidal responses to become nearly identical after the Chi-Chi earthquake for an extended period (Wang et al. 2016).

The distribution of the change in δ¹⁸O composition shortly after the Chi-Chi earthquake (Fig. 9.11c, c’) shows entirely different patterns from that for the coseismic water-level change in the same aquifers (Fig. 6.4b, d). This difference suggests that that the change in isotope composition is unrelated to that that caused the coseismic water level change. This observation is consistent with the suggested mechanism of undrained consolidation for the groundwater level change (Wang et al. 2001) because undrained consolidation does not involve exchange of groundwater and thus is not expected to cause any change in groundwater composition. But it may be interesting to explain why an exchange of water source did not cause a change in water level. We consider the Peclet number for solute transport (Eq. 2.39), Pe = vL/D, where v is the linear velocity of the groundwater flow, L is the characteristic distance between different aquifers and D is the hydrodynamic dispersion coefficient. Appreciable change in solute concentration by advective transport would occur if Pe > 1, that is, if v ≥ D/L. A similar consideration shows that the effect of advective transport on the hydraulic head occurs if v ≥ (K/S_s)/L (Phillips 1991), where K is the hydraulic conductivity and S_s is the specific storage of the aquifer. Given the order of magnitude estimate of longitudinal dispersivity α = 1 to 10 m and a typical linear velocity of 10⁻⁸ m/s (Ingebritsen et al. 2006), D ~ αv is of the order of 10⁻⁸–10⁻⁷ m²/s, while K/S_s is of the order of 1 m²/s for the confined aquifers in the Choshui River fan (Tyan et al. 1996). Thus, the flow velocity required to significantly affect solute composition is 7–8 orders of magnitude smaller than that required to significantly affect the groundwater level. In other words, the amount of exchange of groundwater between Aquifers I and Aquifer III to cause the increased depletions of δ¹⁸O near the coast may be too small to cause an observable change in the groundwater level.

Finally, while the similarity between Aquifer I and Aquifer III in the locations of the areas with more than 0.4‰ depletion of δ¹⁸O suggests the occurrence of vertical mixing of groundwater after the Chi-Chi earthquake, it does not specify whether the groundwater flow was upward or downward, which may bear on the origin of the depleted water.  An objective criterion comes from the study of groundwater temperature change beneath the Choshui River fan after the Chi-Chi earthquake. In Sect. 8.3 we showed that an appreciable increase of groundwater temperature occurred along a broad area along the western coast after the earthquake. This increase of groundwater temperature implies an upward flow of groundwater in the coastal area after the earthquake. Given the Peclet number for advective heat transport (Chap. 2, Eq. 2.2.9) ${\text{Pe}} = qL/D_{h}$, where q is Darcy velocity, $D_{h} = K_{h} /\rho_{w} c_{w}$ is the thermal diffusivity, $K_{h}$ is the thermal conductivity, $\rho_{w} \;and\;c_{w}$, respectively, are the density and specific heat of water, advective heat transport becomes significant when $v\varphi = q \ge D_{h} /L$, where φ is porosity. Since φ is of the order of 10⁻¹ and $D_{h}$ is of the order of 10⁻⁶ m²/s, appreciable advective transport of heat occurs if $v\sim 10^{ - 5} L^{-1}$ m/s, which is ~3 orders of magnitude greater than that required to cause appreciable advective transport of solute. Thus, the hypothesis that upward flow caused the vertical mixing of water between Aquifers I and Aquifer III after the Chi-Chi earthquake is consistent with both the post-seismic change of groundwater temperature and that of groundwater chemistry. Quantitative modeling is clearly necessary to test this hypothesis and to estimate the amount of flow involved in the exchange process.

9.2.3.3 Japan

There are several studies of the changes in the isotopic composition of groundwater after the 2016 Kumamoto earthquake. Hosono et al. (2020a) analyzed a comprehensive dataset for isotopic compositions of groundwater, spring water and river water in the affected region before and after the earthquake. They found that all waters changed their δD and δ¹⁸O compositions towards more depleted values after the earthquake (Fig. 9.13). They also found that the composition of groundwater changed from resembling a mixture of multiple sources before the earthquake into a composition with a signature similar to the mountain foot spring waters after the earthquake (Fig. 9.13b–d), regardless of the sampling season, the aquifer types (confined or unconfined) and the areas of the aquifers, implying an increased post-seismic contribution of water from mountain aquifers.

Fig. 9.13

Changes in δD and δ¹⁸O compositions. a δD and δ¹⁸O compositions of groundwater, spring water and river water both before (April 2011–July 2011) and after (August 2016–May 2017) the main shock for river and spring waters for samples collected in different seasons. The blue, green and black contours show, respectively, the composition ranges of the mountain-foot springs, the high-elevation springs and groundwaters before earthquake. Compositions of hot spring waters and mountain aquifer water from ongoing tunnel construction for the samples collected after the main shock are also plotted. Springs (blue and green triangles) and river (yellow triangle) water samples obtained after the earthquake are shown in darker colors than samples from before the earthquake. Samples collected in all seasons for both aquifers (unconfined and confined aquifers) are plotted together. b Compositional changes of groundwater from confined aquifers collected in various seasons before (November 2009–November 2011) and after (June 2016–December 2017) the earthquake. Samples after the earthquake are shown in red, while those before the earthquake are shown in white. Error bars show the sizes of measurement errors. c and d, respectively, changes of groundwater compositions from the recharge area and from the discharge area (from Hosono et al. 2020a)

Ide et al. (2020) studied the change of water isotopes  in many springs in central Kyushu after the 2016 Kumamoto earthquake. They also found a regional decrease of δ¹⁸O of the spring waters (becoming more negative) after the earthquake (Fig. 9.14) by comparing their δ¹⁸O after the earthquake with those measured in the year of 2009. These authors attributed this change to the mixing of water with lighter isotopic composition released from the Aso Caldera into the regional groundwater that supplies most springs in the study area. The observed δ¹⁸O changes generally show greater absolute magnitudes closer to the Aso Caldera than those further away (Fig. 9.14), supporting their interpretations.

Fig. 9.14

Distribution of earthquake-induced changes in stable isotope ratio (δ¹⁸O) of spring water samples between 2009 and 2017 (Δδ¹⁸O = δ¹⁸O₂₀₁₇ − δ¹⁸O₂₀₀₉). Circles show spring location and the circle sizes show the relative change magnitude (from Ide et al. 2020)

9.3 Stream Water Composition

As noted at the beginning of this chapter, the detection of earthquake-induced changes of water chemistry in streams may be challenging because the signals in the new water may be diluted by the existing water in the streams. Thus earthquake-induced changes of stream water chemistry are rarely studied unless the amount of water in the streams before the earthquake is negligible. Such a situation, though unusual, occurred at least twice in central California in the past thirty years during the dry seasons when stream flow was either low or absent. Following the Loma Prieta earthquake, central California, on 17 October 1989, the discharge in some streams near the epicenter increased more than an order of magnitude over that before the earthquake (Fig. 9.15a). Rojstaczer and Wolf (1992) reported the stream discharge and the change of water chemistry at two gauging stations in the San Lorenzo drainage basin and compared these with earlier measurements (Fig. 9.15b, c) that had been documented on a biannual basis. The stream chemistry showed a marked increase in overall ionic strength after the earthquake, but the proportions of the major ions were nearly the same as those before the earthquake. The increased ion concentration decreased significantly over a period of several months after the earthquake (Fig. 9.15) together with the decrease in the excess stream discharge. By April 1990, the stream water chemistry had begun to approach the pre-earthquake conditions at both stations. The change of stream chemistry, together with a general cooling of the stream water by several degrees, led Rojstaczer and Wolf (1992) and Rojstaczer et al. (1995) to suggest that the additional stream discharge following the Loma Prieta earthquake was derived from groundwater from the drainage basin instead of from mid-crustal depths. This is the main argument used by Rojstaczer et al. (1995) to argue against the static strain model of Muir-Wood and King (1993).

Fig. 9.15

(modified from Rojstaczer and Wolf 1992)

a Discharge in the San Lorenzo River. Arrows show the occurrences of local precipitation. b Major ion chemistry in the stream water as a function of time at the San Lorenzo Park gauge, and c at the Big Trees gauge in the San Lorenzo drainage basin, central California

A second example is the South Napa earthquake in northern California (Fig. 7.3a), which occurred during a prolonged drought in California when many small creeks in the Coast Ranges were either dry or had little flow. These seasonal creeks (see Fig. 7.3a for locations) started to flow right after the earthquake (Figs. 7.3b, d and e). Since there was little or no water in the creeks before the earthquake, the new flows were not mixed with pre-existing waters and their composition represents the composition of the mobilized waters. For this reason, the composition of the new flows in the Napa and the Sonoma Valleys following the South Napa earthquake is particularly valuable.

Wang and Manga (2015) measured the δ¹⁸O and δD in the new discharges following the South Napa earthquake. Figure 9.16a shows that the stable isotopes of hydrogen and oxygen of the new waters define a linear relation on a δD versus $\delta^{18} {\text{O}}$ plot, parallel to, but slightly shifted to the left of, the global meteoric water line (GMWL). Wang and Manga (2015) interpret the  slight shift from GMWL to reflect differences in humidity and temperature that affect secondary evaporation as rain falls from clouds. The isotopic compositions of each flow, sampled at different times, cluster closely together, suggesting that each flow came from a distinct source of constant composition. Different flows, on the other hand, span a broad range of isotopic composition, suggesting that the different sources were recharged by meteoric water at different elevations. Also plotted are the isotopic compositions of the Napa River determined at various times of year from 1984 to 1987 (Coplen and Kendall 2000). From November to March, normally the rainy season, the isotopic composition of Napa River falls mostly close to the GMWL; during dry seasons, on the other hand, it becomes significantly heavier and falls to the right of the GMWL, reflecting the evaporation of river water and recharge from shallow groundwater or reservoirs in the valley during dry seasons.

Fig. 9.16

a Stable isotope data for the studied streams and spring in Wang and Manga (2015). Shown are measurements of δD versus $\updelta^{18} {\text{O}}$ for the new streams and Spencer Spring, the Napa River from 1984 to 1987, and three major perennial streams in foothills. Measurement errors are smaller than the size of symbols used. Solid line shows the GMWL. Data from this study define a local meteoric water line parallel to, but shifted slightly to the left of the GMWL. Notice that the isotopic compositions of each flow, sampled at different times, cluster together, while the isotopic compositions of different flows span a broad range along the local meteoric water line. During rainy seasons (normally November to March) the isotopic composition of the Napa River falls mostly close to the GMWL; during dry seasons, the Napa River composition becomes significantly heavier and falls to the right of the line due to evaporation and recharge by evaporated surface water. b Heiko Woith’s plot (personal communication) of stable isotope data from Wang and Manga (2015), Forrest et al. (2013) and Ingraham and Caldwell (1999). The label ‘GMWL hot’ on the dashed line suggests a mixing line between the meteoric water and geothermal water. GW: non-hydrothermal groundwater, MW: mixed hydrothermal/meteoric water, HC: hydrothermal groundwater from Calistoga; HSON: hydrothermal water from Sonoma; SW: saline water

Wang and Manga (2015) interpreted the isotopic signatures of the new flows (Fig. 9.16a) to suggest that the new flows originated from meteoric water that was stored as groundwater in the nearby mountains at different elevations, which was released by the South Napa earthquake through enhanced vertical permeability, like that suggested for the increased stream flows after the 1999 Chi-Chi earthquake. It is also significant that the isotopic composition of some perennial streams in foothills falls along the same local meteoric water line defined by the new flows (Fig. 9.16a). Since these streams are recharged by baseflow in the mountains during the drought, the similarity between their isotopic composition and that of the new flows supports the suggestion that the new flows originated from the groundwater in the nearby mountains. Wang and Manga (2015) also compared the composition of the new waters with the average composition of the perennial Napa River, which represents an averaged stream water composition in the valley. The spread of $\delta^{18} {\text{O}}$ in the new waters is from −6 to −7.5%₀ (Fig. 9.16a), while that of the Napa River water between May and October (before the rainy season) spreads from −5 to −6%₀. Thus, the earthquake may have caused an overall decrease of $\delta^{18} \text{O}$ by 1.5%₀ (more negative) from that on the valley floor. Assuming a global lapse rate of 2.1%₀ km⁻¹(Chamberlain and Poage 2000), Wang and Manga (2015) suggested that this difference in $\delta^{18} \text{O}$ corresponds to a difference in elevation of ~700 m, which may be compared with the difference in elevation between the valley floor (near sea level) and some mountains in the studied area, such as Cobb Mountain at 1440 m above sea level, and Mount Saint Helena and Hood Mountain over 762 m above sea level, consistent with the hypothesis that the new waters were released from the nearby mountains.

After the South Napa earthquake, many springs also began to flow in the nearby mountains. According to the owner of a local ranch (John Tuteur, 9/13/2014), “… the (largest) spring is a major enhancement of what used to be a seep. The opening is approximately 20–25 cm wide and 5–6 cm deep. When I last visited the spring the water was coming out of the opening in a shape that matched the opening. The water flows down a flat channel into the creek approximately 8–10 m below the spring outlet. The other springs which are pretty difficult to reach on foot are more like seeps that are flowing in thin sheets down the face of a cliff side approximately 2–4 m above the stream channel. There are four or five of those seeps in close proximity to each other.” Starting Sept. 14, 2014, Wang and Manga (2015) measured the discharge, temperature, isotopic and chemical composition of the largest hot spring (Spencer Spring, see Fig. 7.3a for location) at an elevation of ~200 m and away from any local water dams. Measurements continued until the end of July, 2020, with some interruptions in sampling caused by forest fires. Figure 9.17 shows that, while the discharge has decreased by more than an order of magnitude since the beginning of the measurements, temperature of the spring water has declined only slightly from 31 °C in the late 2014 to 30 °C in July, 2020, and the isotopic compositions has stayed nearly constant at −7.4‰ to −7.7‰ for $\delta^{18} {\text{O}}$, and at −46‰ to −48‰ for δD, suggesting that the spring water was supplied from a nearly constant source with little mixing of surface water, and that the decrease in discharge was probably due to a gradual clogging of the pathways connecting the source to the surface spring, which were opened by the earthquake. The nearly constant water temperature of the Spencer Spring is in contrast with the variable temperature of water in the streams, which ranges from 13 to 21 °C. The average surface temperature is ~15 °C and the regional average geothermal gradient is 46 °C/km; thus the new water may have come from depths greater than ~300 m beneath the surface assuming the average geothermal gradient at the spring site.

Fig. 9.17

Temperature, discharge, $\delta^{18} {\text{O}}$ and δD of water of the Spencer Spring of Napa Valley, California, following the 2014 South Napa earthquake to 2020 (diagram from the authors)

Curiously, the $\delta^{18} \text{O}$ and δD of the water in the Spencer Spring are the lightest among the new waters and plot on the extension of the dashed line in Fig. 9.15a. Forrest et al. (2013) also showed that fluids of inferred hydrothermal origin in the Napa and Sonoma Valleys are light in $\delta^{18} {\text{O}}$ and δD and remain close to the meteoric water line, suggesting that these hot springs are recharged by meteoric water. Heiko Woith (personal communication) plotted the data from Wang and Manga (2015), Forrest et al. (2013) and Ingrham and Caldwell (1999) together in Fig. 9.16b and labeled the dashed line that connects the new stream waters and the hydrothermal waters “GMWL hot”, hypothesizing that the isotopic compositions of the new waters were due to the mixing of groundwater released from nearby mountains with increased hydrothermal water released from depth by the earthquake. More work is clearly needed to better understand the origin of the new waters.

9.4 Need of Integrated Data to Interpret Composition Change

Discussions in the previous sections of this chapter have made clear that the interpretation of earthquake-induced changes in water composition is more challenging than that of other types of hydrological responses to earthquakes. Examples include the distinctly different responses of the concentrations of major elements in two geographically adjacent wells in Iceland (HA-01 and HU-01, see Fig. 9.1 for well locations) to the same earthquakes (Fig. 9.2; Skelton et al. 2019), the inconsistent responses in the groundwater compositions in the monitoring wells near the epicenter of the 2016 M5.8 earthquake in SE Korea, and the distinctly different responses between the concentrations of the major elements and trace elements in wells in the central Italian Apennines (Rosen et al. 2018) and in SW China (Shi et al. 2020), but, at the same time, the similar responses between the major and the trace element concentrations to earthquakes in the HU-01 well in Iceland (Claesson et al. 2004).

Difficulties in explaining these conflicting observations may be expected because the earthquake-induced changes of groundwater compositions are likely to result from the exchanges of groundwater with unknown and isolated groundwater sources that may be affected by different factors such as source depth, elevation, temperature and water-rock reactions. Ideally, the interpretation of the earthquake-induced changes of groundwater composition requires not only measurement of the water composition before and after the earthquake, but also simultaneous measurements of groundwater level, groundwater temperature, the hydraulic properties of aquifers and aquitards, and a detailed knowledge of the local hydrogeology. A major challenge is that most studies of earthquake-induced changes of groundwater composition are made without such complementary information. Another challenge is the difficulty in obtaining continuous measurements of groundwater composition in most wells. Often measurements of groundwater composition are made following an earthquake, but the necessary reference of composition before the earthquake is missing; or the post-seismic measurements are made once without continuous measurements, resulting in an absence of information for understanding the time-dependent processes.

In the unusual case where several types of data are available and densely distributed in the earthquake-affected area, an integrated dataset may be used to better constrain our interpretation of the earthquake-induced processes. This situation, unfortunately, is rare. Largely through serendipity, however, the interpretation of the changes of isotopic composition of groundwater in western Taiwan following the Chi-Chi earthquake (Sect. 9.4.2) may serve as an illustration of what is meant here by using integrated data for interpreting composition change. In a nutshell, analysis earthquake-induced changes in the δ¹⁸O composition from clustered wells on an alluvial fan near the earthquake epicenter (Figs. 9.11 and 9.12) revealed a striking similarity between the earthquake-induced changes in different aquifers separated by aquitards (Fig. 9.11c, c’), suggesting the occurrence of post-earthquake vertical mixing of initially isolated groundwaters among different aquifers (Wang et al. 2005). This same conclusion was drawnfrom the study of the post-seismic convergence of initially different groundwater level in different aquifers to the same level (Wang 2007) and the tidal analysis of groundwater level in the same area (Fig. 6.16; Wang et al. 2016), further demonstrating that the earthquake-breached confinement of the aquifers may have allowed vertical mixing of groundwater. Finally, the analysis of groundwater temperature in the same area demonstrated an enhanced basin-wide groundwater flow after the earthquake, with increased downward flow in the foothills and increased upward flow from depth near the coast (Fig. 7.3; Wang et al. 2013), which is also consistent with the suggestion of the post-earthquake basin-scaled mixing of groundwater. Hence, we may conclude that the earthquake-induced change of the δ¹⁸O composition in western Taiwan was due to an increased basin-wide groundwater transport from the Taiwan western foothills to the coast following the Chi-Chi earthquake and a vertical mixing of groundwater between different depths. Obviously, this hypothesis requires further field tests and coupled numerical simulations, but the fact that it is consistent with a variety of observations makes the hypothesis better constrained.

Following the 2016 Kumamoto earthquake, there has been a tremendous increase of high-quality data for water composition, groundwater level and groundwater temperature (e.g., Hosono et al. 2020b). It may be timely to integrate these data in a multidisciplinary interpretation to advance our understanding on earthquake-induced changes of groundwater temperature and composition.

9.5 Concluding Remarks

Groundwater chemistry has long been used as an important tracer for understanding hydrogeological processes in general, and their changes after earthquakes in particular. In this chapter we discussed some existing studies of the earthquake-induced changes of the groundwater composition and streamflow composition. Even though the data for these changes are much less abundant than those for the changes in groundwater level and stream discharge, they have provided valuable information to constrain models of earthquake-induced hydrogeological processes. Most changes are consistent with the model of earthquake-enhanced groundwater transport through basin-wide or local enhanced permeability. The enhanced permeability may breach hydrologic barriers such as aquitards, connecting otherwise isolated aquifers or other fluid sources, causing fluid source switching and/or mixing. Studies of these processes may be important not only for better understanding natural transport processes but also for better understanding earthquake-induced contamination of groundwater by surface water, as reported in central Kyushu, Japan, following the 2016 Kumamoto earthquake.