**6. Discussions and conclusions**

308 Earthquake Research and Analysis – Seismology, Seismotectonic and Earthquake Geology

anomalies in the days before the event of 7 May, while during all this month flow rates were comparable with the average values calculated for the period 1979-89, even if the rainfalls

Nevertheless, gas outpouring and water muddying have been noticed in the evening of 5 May, in the springs at Posta Fibreno, localized about 2 km before the hydrometric sections. Also in the hydrometric section of Sangro river at Ateleta, the hydrological anomaly was pointed out only in the value of the average streamflow level in May 1984. Consequently, for this earthquakes we have a total of 14 sites where hydrological effects have been noticed, taking into account that in 3 sites (the Rapido and Gari rivers at Cassino, the Gari river at S. Angelo in Theodice) the positive hydrological anomalies preceded the earthquake of 7 May. Figures 7a and 7b show the anomalous behaviour in two studied sites. In Fig. 7a we show an hydrological anomaly consisting of a strong increase in the streamflow rate of the Sangro River at Barrea, not imputable to the rainfalls. Figure 7b shows the daily streamflow level of Fucino channel, "a detail" of Fig. 7a. The two sites are very near each other. It shows how during April 1984 the sharp increase of 7 May cannot be imputed to the very small rainfalls

**Sangro river at Villetta Barrea 1984-05**

**1984-04**

**Monthly rainfall (mm) Runoff (mm)**

from January to May 1984 have been lightly higher than the average.

on 6 May.

**1981-01**

**0**

streamflow level of Fucino channel (b).

**6**

**12**

**18**

**Daily rainfall at Avezzano (mm)**

**24**

**30**

**36**

**1981-03**

**1981-05**

**1981-07**

**1981-09**

**1981-11**

**7**

**1982-01**

**1982-03**

**1982-05**

**11**

**1982-07**

**1982-09**

**1982-11**

**1983-01**

**1 4 7 10 13 16 19 22 25 28 31**

Fig. 7. Increase in the streamflow rate of the Sangro River at Villetta Barrea (a) and in daily

**1983-03**

**1983-05**

**1983-07**

May 7th and 11th, 1984 earthquakes

**1983-09**

**1983-11**

**1984-01**

**1984-03**

**1984-05**

**1984-07**

**1984-09**

**1984-11**

**0**

**Streamflow level of Fucino (cm)**

**20**

**40**

**60**

**80**

**100**

**120**

In the Southern Apennines we found that some earthquakes produced clear forerunner signals in various areas where geochemical and hydrological parameters were controlled.

These results seem to indicate that the hydrological phenomena are associated to the changes of the stress field, during and after an earthquake.

We analyzed the relations of primary tectonic effects with the local geomorphic and structural setting.

We have also applied for some category of secondary effects a statistical test to infer the presence of trends and a regression analysis based on least-squares method was performed. In particular, a simple bivariate scatter plot of two variables have been computed, specifically:

macroseismic intensities versus epicentral or fault distances of hydrological anomalies. Figure 4 shows for the 1930 earthquake the trend macroseismic intensity-distance from the epicentre. Very similar trends have been found for the 1980 and 1984 earthquakes. Hydrological phenomena occurred throughout the macroseismic field, and were the most numerous among the induced effects. They include flow increase both in spring and well, turbid water and drying up of springs, and even creation of new springs. Some variations in chemical parameters of the waters were observed at different locations, both inside and outside the epicentral areas. The data relative to the hydrological variations' distribution versus the distance from the epicentre show that the high concentration of phenomena lie between 25-80 km, about 20% are inside the epicentral area (0-25 km); few phenomena occurred at greater distances.

A dramatic increase in the springs' flow implies the deformation of major tectonic blocks, which influences deep aquifers. Examples are shown in figures 5d,e (Cassano Irpino springs) and figure 5f (Sanità spring near Caposele) during the 23 November 1980 earthquake.

A remarkable aspect of these anomalies is that they were observed at distances of more than 200 kilometers from the earthquake epicentral area. This suggest that the impact of the Irpinia earthquake on the hydrogeologic structure of the Southern Apennines was more important, in terms of both total number of recorded anomalies and for their epicentral and fault distance.

Today there are no valid earthquake precursors, but many effects are invoked as good forerunners: geophysical changes (vp/vs, telluric currents, electromagnetic effects), geochemical changes (chemical composition, pH, water temperature, gases like Rn, CO2) and hydrologic changes (piezometric levels, spring and stream flow).

Probably the simultaneous observation of all of these effects can constitute a sure forecasting. Many efforts and money are necessary for this purpose.

Taking into account that not always earthquakes are preceded by all of the above mentioned precursory phenomena, and that today's technology can provide probes for many geochemical and geophysical parameters at affordable cost, we maintain that a regional monitoring network can be installed in the Southern Apennines, in order to continuously control as many parameters as possible.

Hydrologic changes depend on both the structure of the aquifer and the strain that an earthquake induces on the area of the fault rupture.

Dobrovolsky et al. (1979) give a theoretical relation regarding earthquake magnitude, distance from the epicenter and volumetric strain. The "strain radius" Rs of a circle centered on the epicenter, in which precursor deformations and other physical phenomena occur, is given by:

$$\mathbf{R}\_{\ast} = 10 \,\, 0.43 \mathbf{M} \,, \quad \text{that is} \; \mathbf{R}\_{\ast} \approx \text{ e } \mathbf{M}$$

This exponential curve divides the areas where strain is lower than 10 -8 and is greater than 10-8. For strain = 10 -8 water level changes are only 1 cm. The data of some earthquakes in Irpinia (table 1 in Onorati and Tranfaglia, 1994; tables 2, 4 and 6 in Porfido et al., 2007) are plotted in figure 8 and the strain radius is indicated.

Fig. 8. (see text)

Montgomery and Manga (2003) suggest that the stream flow changes are attributable to liquefaction of valley bottom deposits. Papadopoulos and Lefkopoulos (1993) give an empirical maximum distance to the epicenter at which liquefaction can occur as a function of earthquake magnitude:

$$\mathbf{M} = -0.44 + 3 \, ^\ast \, 10 \, ^\ast \mathbf{D}\_\mathbf{e} + 0.98 \, \log \mathbf{D}\_\mathbf{e} \, \prime$$

where De is the distance to the epicenter in cm. Figure 9 shows distance to epicentre for hydrological changes in rivers versus magnitude of Irpinia earthquakes in 1930, 1980 and 1984 (Porfido et al., 2007; Onorati and Tranfaglia, 1994). The liquefaction curve determined by the above relation is reported in figure 9. Because many hydrological changes are at distances greater than the liquefaction curve determined for valley bottom deposits, they can be caused by preseismic fracturing of carbonate aquifers in the Apennine Chain.

To define the normal hydrodynamic behaviour of an aquifer it is necessary to develop some stochastic models of the input-output type. Even the simplest stochastic model provides a lot of information on the aquifer's structure and on the connections between hydrologic variables. For this purpose an analysis of the correlation between rainfalls and water levels and flow rates has been carried out.

Four examples are reported in figures 10a-d, which shows the cross-correlograms obtained calculating the coefficients of cross-correlation for various lags. For the Pomigliano well (figure 10a), seasonal variations are evident and the coefficients of cross-correlation between the precipitations and the water table are statistically meaningful for delays until 30 months. This is in agreement with the fact that the Pomigliano well is on alluvial water table with superficial feeding due only to precipitations. For the Bucciano well (figure 10b), the crosscorrelogram increases of significance until to the maximum value of 150 lags (that is about 250 days). The minor components can be due to surface feeding, but the lag=150 component is certainly due to deep feeding from the carbonate aquifer of Taburno mount. In figure 10c and figure 10d similar trends appear: seasonal correlations are very clear for Sanità Spring, fed only by carbonate aquifers of Picentini mounts; while for Tasso river the minor components indicate surface feeding, but the great lag=90 component is due to the deep feeding from Marsicano mount.

Fig. 9. (see text)

310 Earthquake Research and Analysis – Seismology, Seismotectonic and Earthquake Geology

This exponential curve divides the areas where strain is lower than 10 -8 and is greater than 10-8. For strain = 10 -8 water level changes are only 1 cm. The data of some earthquakes in Irpinia (table 1 in Onorati and Tranfaglia, 1994; tables 2, 4 and 6 in Porfido et al., 2007) are

Montgomery and Manga (2003) suggest that the stream flow changes are attributable to liquefaction of valley bottom deposits. Papadopoulos and Lefkopoulos (1993) give an empirical maximum distance to the epicenter at which liquefaction can occur as a function

M = - 0.44 + 3 \* 10-8 De + 0.98 log De , where De is the distance to the epicenter in cm. Figure 9 shows distance to epicentre for hydrological changes in rivers versus magnitude of Irpinia earthquakes in 1930, 1980 and 1984 (Porfido et al., 2007; Onorati and Tranfaglia, 1994). The liquefaction curve determined by the above relation is reported in figure 9. Because many hydrological changes are at distances greater than the liquefaction curve determined for valley bottom deposits, they

can be caused by preseismic fracturing of carbonate aquifers in the Apennine Chain.

To define the normal hydrodynamic behaviour of an aquifer it is necessary to develop some stochastic models of the input-output type. Even the simplest stochastic model provides a lot of information on the aquifer's structure and on the connections between hydrologic variables. For this purpose an analysis of the correlation between rainfalls and water levels

Four examples are reported in figures 10a-d, which shows the cross-correlograms obtained calculating the coefficients of cross-correlation for various lags. For the Pomigliano well (figure 10a), seasonal variations are evident and the coefficients of cross-correlation between the precipitations and the water table are statistically meaningful for delays until 30 months. This is in agreement with the fact that the Pomigliano well is on alluvial water table with

plotted in figure 8 and the strain radius is indicated.

Fig. 8. (see text)

of earthquake magnitude:

and flow rates has been carried out.

To conclude, the significant peaks can be attributed to the hydrodynamic behaviour of aquifers, and the delayed contribution from carbonate complex is evident. The delay of such peaks with respect to the start of the hydrological anomaly and/or its duration, can concur to define the space-time limits of the anomaly correlated with earthquake.

The study of the geochemical and hydrodynamic characteristics of aquifers is acknowledged to make a valid contribution to understanding the natural processes connected to earthquakes (King et al., 1981; King, 1985; Bredehoeft et al., 1987; Roeloffs et al., 1989; Kissin and Grinevsky, 1990; King et al., 1994; Quilty and Roeloffs, 1997; Ingebritsen and Sanford, 1999). Changes in the water-rock interaction are caused by the seismic stresses in the area where the tectonic deformation leads to the seismic event (Rojstaczer and Wolf, 1992, 1994; King and Muir-Wood, 1993; Quilty and Roeloffs, 1997; Roeloffs, 1998; Ingebritsen and Sanford, 1999; Manga, 2001).

Various mechanisms have been invoked to explain earthquake-related changes in water tables and in spring and stream discharges:

1. Large (as much as 20 m), near-field (probably <50 km from the epicenter) water level declines can sometimes be related to near-surface permeability enhancement due to ground motion (Rojstaczer and Wolf, 1992, 1994; Rojstaczer et al., 1995). These authors limit the validity of the relationship between seismic intensity and areas with water increases only to normal fault earthquakes.

Fig. 10. Cross-correlograms of (a) Pomigliano (lag 1 month), (b) Bucciano wells (lag 15 days), Sanità spring at Caposele (lag 1 month) and Tasso river at Scanno (lag 1 day).


The negative anomalies found in this work can be considered "rebound anomalies", which are the most common precursor reported by many authors and are related to increases in porosity and permeability caused by fracturing that precedes an earthquake (Roeloffs, 1988; Igarashi et al., 1992).

The total excess discharge (0.035 km3) caused by the Irpinia 1930 earthquake (Ms=6.7) of 11 streams (table 3) is comparable with the excess discharge of about 0.01 km3 for the Loma Prieta earthquake (Mw=6.9) (Rojstaczer and Wolf, 1992, 1994).
