We present the first regional map of CO2 Earth degassing from a large area (most of central and south Italy) derived from the carbon of deep provenance dissolved in the main springs of the region. The investigation shows that a globally significant amount of deeply derived CO2 (10% of the estimated global CO2 emitted from subaerial volcanoes) is released by two large areas located in western Italy. The anomalous flux of CO2 suddenly disappears in the Apennine in correspondence to a narrow band where most of seismicity concentrates. Here, at depth, the gas accumulates in crustal traps generating CO2 overpressurized reservoirs which induce seismicity.
 High fluid pressure at depth can play a major role in triggering earthquakes by reducing fault strength and potentially controlling the nucleation, arrest, and recurrence of earthquake rupture [Cox, 1995; Rice, 1992; Sibson, 1992, 2000]. On a global scale it is known that CO2 discharges are associated to seismically active zones [Barnes et al., 1978]. However, this finding remains at a general level because the unavailability of regional maps of earth degassing.
 A previous study [Chiodini et al., 2000] showed that a large portion of the inorganic carbon dissolved in the groundwaters circulating in the central Apennines is derived from a deep, mantle-related source. The published chemical and isotopic compositions of 62 main springs are here integrated with new data collected during 2001–2002 on 77 springs of the southern Apennine (Figure 1). The springs were selected based on their high flow rates in order to have a data set that is highly representative of the groundwaters of the region. A total flow rate of 221000 l s−1 (58% of the total water discharge of the Apennine [Cardellini., 2003, and references therein]) has been sampled and analyzed for major chemical components and for the isotopic composition (δ13CDIC) of dissolved inorganic carbon (DIC).
 The geochemical data of the springs are here used to elaborate a regional map of CO2 Earth degassing. Our first aim is to investigate the relationships between Earth degassing and the geological features of Central Italy, i.e., quaternary volcanism, extensional tectonics and intense seismic activity, and our second is to estimate the total deeply derived CO2 involved in the degassing process in this non-volcanic area.
2. Origin of the Carbon Dissolved in Groundwaters
 A fraction of DIC derives directly from the dissolution of the carbonate rocks, mainly limestone and subordinately dolostone, which host the Apennine aquifers [Chiodini et al., 2000]. Considering that (i) calcite (or dolomite) dissolution leads to an equi-molar increment of DIC and Ca (or Ca + Mg) in the solution, and (ii) minor amounts of Ca derive from the dissolution of gypsum locally present in the aquifers hosting rocks, the amount of DIC from carbonates dissolution (Ccarb) is computed for each sample by Ccarb = Ca + Mg − SO4. The carbon derived from other sources (Cext) is given by DIC - Ccarb. In order to investigate the origin of Cext, its isotopic composition (δ13Cext) is computed by the equationδ13CDIC × DIC = δ13Cext × Cext + δ13Ccarb × Ccarb setting the isotopic composition of the carbon from carbonate mineral dissolution (δ13Ccarb) at 2.21‰ ± 0.66‰ (mean and standard deviation of 567 samples of Apennine carbonate rocks, [Cardellini., 2003, and references therein]). If the water is affected by CO2 degassing and/or carbonate minerals precipitation prior to sampling then the used equations underestimate Cext and overestimate δ13Cext because of isotopic fractionation during carbon sinks from the solution. However, these uncertainties are considered negligible over the range of DIC and PCO2 values for virtually all the samples [Chiodini et al., 2000]. In Figure 2 the variables Cext and δ13Cext indicates the presence of two carbon sources. The samples with relatively low carbon contents have an isotopic composition (δ13Cext = −21.4 ± 2.9‰) which is in the range of soil CO2 from vegetated areas, i.e., from −15‰ to −30‰ [Deines et al., 1974; Robinson and Scrimgeour, 1995]. The first source is therefore the soil CO2 of organic origin which is dissolved during the infiltration of the meteoric waters in the recharge areas. The samples of relatively high Cext fall in the hyperbolic domain (grey area in Figure 2) that represents the compositions of infiltrating waters which have had variable amounts of CO2 with δ13C from −5‰ to +1‰ added to them. This δ13C range coincides with the isotopic signature of the CO2 of deep provenance emitted in Italy from active volcanoes, geothermal fields, and the many cold gas emissions located in the western sector of the region (Figure 1). Figure 2 shows that most of the groundwaters circulating in the Apennines are affected by variable CO2 inputs from the deep source. The total amount of carbon transported by the Apennine groundwaters is 1.2 × 1011 mol/y. This value is the sum of the carbon transported by each spring (computed by multiplying DIC by the flow rate) divided by a factor 0.58, i.e., the sampled fraction of Apennine groundwater total flow rate. Minor amounts of this carbon came from the dissolution of carbonates (33.2%) and from the biogenic CO2 of the soils (23.5%), whereas the greatest part derives from the deep source (43.3%) pointing out the relevance of non-volcanic Earth degassing processes in Apennines.
 Further evidence for the double origin of Cext in Apennine groundwaters is given by the probability plots of δ13Cext (Figure 3a) and log Cext (Figure 3b). Both the diagrams show the existence of two populations: the first, population A, is represented by the samples where the ‘external’ carbon derives only from biogenic CO2 (background), while the second, population B, is represented by the samples in which variable amounts of deeply derived CO2 are added during groundwater circulation. Figure 3b shows that a Cext of 0.004 mol/kg is about the maximum value expected for the dissolved carbon if only a biological carbon source is present. Since this value well separates the two populations, it has been selected as the cutoff value for the elaboration of a map of the probability that a deep source of CO2 exists in the studied area.
3. Mapping and Quantification of CO2 Earth Degassing
Figure 4a shows the probability that at any location Cext is higher than the cutoff value. The map has been drawn using an algorithm of sequential Gaussian simulation [Deutsch and Journel, 1998], recently applied to investigate soil CO2 diffuse degassing in volcanic areas [Cardellini et al., 2003]. The data set on which the geostatistical approach is based has been integrated with data from literature relative to other Apennine springs, springs from smaller carbonate aquifers of the western part of Italy, and the average values of the volcanic aquifer of Mt. Albani. In each case the selection was based on the high flow rates of the springs (flow rates >10 l/s).
 Because the cutoff is the maximum value expected for the dissolved carbon if only a biological carbon source is present, the map in Figure 4a represents the probability that a deep source of CO2 exists at any location. The highest probabilities (>0.5) define two large anomalies located on the Tyrrhenian side of the Italian peninsula. The northern structure partially overlaps the Tuscany, Roman magmatic province (TRDS, Tuscan Roman degassing structure) while the southern structure relates to the Campanian volcanism (CDS, Campanian degassing structure). The two volcanic provinces are characterized by quaternary potassic and ultrapotassic magmas rich in fluids with high CO2/H2O ratios [Foley, 1992]. Geochemistry of the magmas is consistent with the melting of a mantle source metasomatized by the addition of subducted crustal material [Peccerillo, 1999]. We suggest that the TRDS and CDS reflect the degassing process of this metasomatized uprising mantle. In the western sectors of TRDS and CDS, which are characterized by outcrops of impermeable formations, most of the gas is likely released by the numerous, widespread gas emissions [Rogie et al., 2000, Figure 1], while in the eastern sectors the gas is dissolved by groundwaters circulating in the Apennine aquifers and it is subsequently released to the atmosphere, concurrently with the deposition of travertines (Figure 1), during the sub-aerial circulation of the waters.
 The total amounts of deeply derived carbon released from TRDS and CDS have been estimated to be 1.4 × 1011 mol/y and 0.7 × 1011 mol/y respectively. The computation has been done by (i) calculating the specific flux (i.e., mol y−1 km−2) of Cext affecting the hydrogeological basin of each spring, (ii) using these data to simulate the Cext flux over the entire area, and (iii) subtracting the contribution from the organic fraction of Cext. The total CO2 released by TRSD and CDS (2.1 × 1011 mol/y) is globally significant, being ∼10% of the estimated present-day total CO2 discharge from subaerial volcanoes of the Earth [Kerrick, 2001]. This result suggests an underestimation on CO2 globally released by the Earth, because unquantified processes of CO2 Earth degassing from non-volcanic environment affect almost all tectonically active areas of the world.
4. Carbon Dioxide Earth Degassing and Seismicity
 Earth degassing and seismicity in central and south Italy [Gruppo di Lavoro CPTI, 1999; Istituto Nazionale di Geofisica e Vulcanologia (INGV), Catalogo della sismicità strumentale, ftp://ftp.ingv.it/bollet/, 2003] is compared in Figure 4a. The map highlights an intriguing coincidence among seismicity and the eastern boundary of the CO2 degassing areas. Earthquake epicenters concentrate in a belt that separates two distinct crustal domains: the Tyrrhenian hinterland, where both the TRDS and the CDS are located, and the Adriatic foreland where deeply derived CO2 is virtually absent. A conceptual model to explain this relation between Earth degassing process and the seismic zone of the Apennines is proposed based on the 40 km deep CROP03 (deep seismic reflection profile) cross section of central Italy [Collettini, 2002]. A marked upwelling of the mantle exists in the Tyrrhenian hinterland (Figure 4b). In this zone, mantle fluids may enter the ductile lower crust at near lithostatic pressure [Kennedy et al., 1997], infiltrate upwards through the interconnected network of extensional fractures and normal faults and generate the high CO2 domains observed at the surface. The degassing process is evident in the TRDS, since here the extension started several million years ago, allowing the development of a mature set of faults and fractures. Focal mechanisms of earthquakes suggest that also the Apennine belt is undergoing extension but only since very recent times [Frepoli and Amato, 1997; Mariucci et al., 1999]. This explains the lack of a well-developed system of interconnected fractures, and of clear degassing at the surface. This source of CO2 does not exist in the Adriatic foreland, which is characterized by a thicker crust. Following the information from the CROP03 profile, we hypothesize that CO2 from the mantle intrudes the deeper, now inactive, thrusts affecting the Moho and the upper crustal levels at the boundary between the Tyrrhenian hinterland and the Adriatic foreland. Here, at depths ranging from few kilometers to about 15 km b.s.l., the thrusts are dislocated by the active, low-angle normal faults, where the seismicity of Apennines is concentrated. In this zone, the arrangements of the deeper thrusts and low-angle normal faults describe outward verging structures which could act as traps (structural seal) in which CO2 may accumulate and generate overpressurized reservoirs. Independent data suggest the presence of overpressurized CO2 reservoirs at the boundary between the TRDS and the non-degassing zone: the wells of S. Donato and Pieve S. Stefano (Figure 4a) encountered a CO2 pressure of about 98 MPa and 67 MPa at depths of 4750 and 3700 m b.s.l. respectively (i.e., ∼0.8 of the lithostatic weight); the seismogenetic low angle faults, that detach in this area at depths from 5 to 15 km, can move at depth only in response to high fluid pressure [Kennedy et al., 1997; Collettini, 2002; Ghisetti and Vezzani, 2002]; the recent seismic crises that occurred at Colfiorito in 1997 have been recently interpreted as caused by pressure diffusion from deep overpressurized CO2 zones to shallow reservoirs at hydrostatic fluid pressure [Miller et al., 2004].
 Overpressurization in these structures, fed by mantle derived CO2, is in our opinion, the primary trigger of Apennine earthquakes. We suggest that mantle degassing has an active primary role in the geodynamics of the region. Understanding the mechanism of mantle degassing, its flux through the crust, and its contribution to crustal deformation and earthquake generation is a major goal of modern Earth science in our region, and deserves attention in many other tectonically active regions of the world.
 This research was supported by GNV-INGV and MURST-PRIN “GEOCO2”.