1
Scientific RepoRts | 6:32245 | DOI: 10.1038/srep32245
www.nature.com/scientificreports
Magma transfer at Campi Flegrei
caldera (Italy) before the 1538 AD
eruption
Mauro A. Di Vito
1
, Valerio Acocella
2
, Giuseppe Aiello
3
, Diana Barra
1,3
, Maurizio Battaglia
4,5
,
Antonio Carandente
1
, Carlo Del Gaudio
1
, Sandro de Vita
1
, Giovanni P. Ricciardi
1
, Ciro Ricco
1
,
Roberto Scandone
2
& Filippo Terrasi
6
Calderas are collapse structures related to the emptying of magmatic reservoirs, often associated with
large eruptions from long-lived magmatic systems. Understanding how magma is transferred from a
magma reservoir to the surface before eruptions is a major challenge. Here we exploit the historical,
archaeological and geological record of Campi Flegrei caldera to estimate the surface deformation
preceding the Monte Nuovo eruption and investigate the shallow magma transfer. Our data suggest a
progressive magma accumulation from ~1251 to 1536 in a 4.6 ± 0.9 km deep source below the caldera
centre, and its transfer, between 1536 and 1538, to a 3.8 ± 0.6 km deep magmatic source ~4 km NW of
the caldera centre, below Monte Nuovo; this peripheral source fed the eruption through a shallower
source, 0.4 ± 0.3 km deep. This is the rst reconstruction of pre-eruptive magma transfer at Campi
Flegrei and corroborates the existence of a stationary oblate source, below the caldera centre, that
has been feeding lateral eruptions for the last ~5 ka. Our results suggest: 1) repeated emplacement of
magma through intrusions below the caldera centre; 2) occasional lateral transfer of magma feeding
non-central eruptions within the caldera. Comparison with historical unrest at calderas worldwide
suggests that this behavior is common.
Dening and understanding the shallow transfer of magma at volcanoes is crucial to forecast eruptions, possibly
the ultimate goal of volcanology. is is particularly challenging at felsic calderas experiencing unrest, which
typically includes signicant changes in seismicity, deformation and degassing rates. In fact, caldera unrest is
particularly frequent, aects wide areas and its evidence is oen complicated by the presence of a hydrothermal
system: as a result, forecasting any eruption and vent-opening sites within an existing caldera is very dicult
1
.
In historical times only two felsic restless calderas have erupted: Campi Flegrei and Rabaul
2
. Campi Flegrei,
in the densely inhabited metropolitan area of Naples (Italy), is commonly considered one of the most dangerous
active volcanic systems. Campi Flegrei is a ~12 km wide depression hosting two nested calderas formed during
the eruptions of the Campanian Ignimbrite (~39 ka) and the Neapolitan Yellow Tu (~15 ka) (Fig.1; refs 3–6).
In the last ~5 ka, resurgence
7
[references therein], with upli > 60 m close to the central part of the caldera (the
Pozzuoli area), was accompanied by volcanism of the “III epoch” of activity (~4.8 to ~3.8 ka; ref. 7). Aer ~3 ka
of quiescence, several decades of increasing seismicity and upli preceded the last eruption at Monte Nuovo in
1538
7–9
. e most recent activity culminated in four unrest episodes between 1950–1952, 1969–1972, 1982–1984
and 2005-Present, with upli at Pozzuoli of ~0.7, ~1.7, ~1.8 and ~0.3 m, respectively
10,11
; the present unrest epi-
sode has been interpreted as being magma-driven
12,13
. ese unrest episodes are considered the most evident
expression of a longer-term (centuries or more) restless activity
4,10
. e post-1980 deformation largely results
from a magmatic oblate or sill-like source at ~4 km depth below Pozzuoli (e.g.
12
and references therein
13
); how-
ever, an important role for the hydrothermal system has been also proposed
14
[references therein].
Despite the restless activity of Campi Flegrei, the recent unrest episodes did not culminate in eruption, so
that any possibility to dene the pre-eruptive shallow transfer of magma (that is, from the magma reservoir to
1
Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Napoli Osservatorio Vesuviano, via Diocleziano 328,
80124 Napoli, Italy.
2
Dipartimento di Scienze Università Roma Tre, Italy.
3
Dipartimento di Scienze della Terra,
dell’Ambiente e delle Risorse, Università degli Studi di Napoli Federico II, Italy.
4
Dipartimento di Scienze della Terra,
Sapienza, Roma, Italy.
5
Volcano Science Center, US Geological Survey, Menlo Park, CA 94025, USA.
6
Dipartimento
di Matematica e Fisica, Seconda Università di Napoli, Italy. Correspondence and requests for materials should be
addressed to M.A.D.V. (email: mauro.divito@ingv.it)
Received: 17 February 2016
Accepted: 04 August 2016
Published: 25 August 2016
OPEN
www.nature.com/scientificreports/
2
Scientific RepoRts | 6:32245 | DOI: 10.1038/srep32245
the surface) at Campi Flegrei remains elusive. Indeed, this denition is a crucial step in order to identify and
understand pre-eruptive processes, and thus to make any forecast. To ll this gap, we focused on the last eruption
of 1538, reconstructing its pre-eruptive deformation pattern. For this, we exploited the unique historical, archae-
ological, geological and long-term geodetic record of the caldera to carefully determine the height variations (and
related errors) of 20 selected sites along its coastline (Fig.1 and Supplementary Table S1). e details of this com-
plex and multidisciplinary approach are provided in the Methods and in Supplementary Information sections.
In this paper a completely original data set is provided on the height variations within the Campi Flegrei
caldera in the last 2000 years, with the only exception for the previously known vertical displacements at site 30
(Fig.1). While our collected data span the last 2000 years of evolution of the caldera, in this study we focus only
on the deformation occurring ~300 years before the 1538 eruption, when the centuries-long subsidence of the
caldera reversed into upli.
Results
Elevation changes within the caldera. e integrated analysis of geomorphological, sedimentological,
paleontological, archaeological and historical data allowed a detailed and quantitative reconstruction of the evo-
lution of the ground displacements predating the Mt. Nuovo eruption along the coastline of the Pozzuoli Bay
(Fig.1). A representative example of the multidisciplinary procedure adopted for such a detailed description of
the historical elevation changes for the Capo Miseno area is included in the supplementary material.
Figure 1. Morphological and structural sketch map of the Campi Flegrei caldera. Circles: vents of the I
epoch (15-9.5 ka BP; shaded), II epoch (8.6-8.2 ka BP; densely shaded) and III epoch (4.8-3.8 ka BP; green)
of volcanic activity. Red: main faults and fractures; black line: Neapolitan Yellow Tu caldera; yellow ellipse
and cross: maximum uplied sector and caldera centre in the last 5 ka; green ellipse: projection of the quasi-
horizontal source (pressurized triaxial ellipsoid ~4000 m deep) of recent ground deformation
12
. Black squares:
selected benchmarks and relative number (names in Supplementary Table S1). ick blue line: cli of the La
Starza uplied marine terrace. inner blue line: 10 m bsl isobath. Inset: distance of the eruptive vents active in
the last 5 ka from the centre of the maximum uplied zone in the same period. Digital Terrain Model by INGV-
Osservatorio Vesuviano.
www.nature.com/scientificreports/
3
Scientific RepoRts | 6:32245 | DOI: 10.1038/srep32245
e general results of our analysis are summarized in Fig.2, which shows the historical elevation changes,
from 35 BC to Present, at the sites along the coastline of Pozzuoli Bay. ese data show that in 35 BC the coastline
extended outward into what is now the Pozzuoli Bay. However, since then all the area started to be aected by a
quick subsidence
4
, which resulted in progressive submersion of the coastline until 1251. e amount of subsid-
ence in the investigated area varies from place to place (Fig.2) and is well documented by the presence of geo-
morphological, sedimentological and paleontological indicators. A subsequent progressive emersion of the area
started during the 13
th
century, as suggested by historical and urban planning sources, archaeological evidence
and geological data (Supplementary Tables S1 and S2). e lower time limit for the caldera upli is given by his-
torical documents describing the Pozzuoli promontory as an island in 1251
15,16
, whereas at the end of the 13
th
and
beginning of the 14
th
century the previously submerged area around the promontory is reported as the location
of three new churches. Moreover, coeval urban studies testify to the expansion of Pozzuoli on new land formed
by the coastline regression, conrming the onset of a long-term upli
15,16
. Since sea-level variation in the last
2000 years has been on the order of 0.7 m on average
17
[reference therein], the much larger (see below for values)
Figure 2. Reconstruction of the elevation (m above the sea level in 1905 – rst leveling by Istituto Geograco
Militare) through time at 20 selected sites within the Campi Flegrei caldera, obtained integrating geological,
historical and archaeological data. e elevation is referred to the nearest benchmark or to archaeological
structures. e upli rate (cm/yr) calculated for the periods 1251–1400, 1400–1536 and 1536–1538 is reported
in green. e location of the sites is reported in Fig.1, from Miseno (benchmark 58) to Nisida (benchmark 194).
Error bars are in red. e diagrams highlight the dierence in elevation changes between the benchmarks close to
the caldera centre and those near Monte Nuovo. During 1251–1400 the caldera oor underwent a minor upli,
with maximum values in the Pozzuoli area. From 1400 to 1536 a general and sharp increase of the upli rate still
culminates in the Pozzuoli area. Immediately before the eruption (1536–1538) the upli reaches the highest rate
in the area of opening of the future vent (Mt. Nuovo). Data by the authors.
www.nature.com/scientificreports/
4
Scientific RepoRts | 6:32245 | DOI: 10.1038/srep32245
emersion of the area from the 13
th
to the 16
th
century was mainly due to the ground upli, with the maximum
values recorded in the Pozzuoli area.
e upli rate was quite low (0.3 to 1 cm/yr; Fig.2; Supplementary Table S1) from the middle of the 13
th
to
the end of the 14
th
century, and increased to 2.9 to 9.1 cm/yr from 1400 to 1536 (Fig.2; Supplementary Table S1).
During this latter time-span, all the coastal strip emerged in response to the generalized upli of the caldera oor,
whose maximum of 12.3 m has been recorded again in the Pozzuoli area (Fig.2; Supplementary Table S1). Since
the end of the 15
th
century this upli was accompanied by strong seismicity
9
. A new and stronger upli, with a
rate of 10 to 940 cm/yr (Fig.2; Supplementary Table S1), followed the previous one between 1536–1538, reaching
a maximum value of 18.8 m in the future vent-opening area (Mt. Nuovo; Fig.2; Supplementary Table S1). is
highest-rate upli was accompanied by very intense seismicity, which aected all the Pozzuoli area and was felt
also in the city of Naples
9
[references therein]. Indeed, all the historical sources coeval to the eruption report
an evident upli accompanied by seismicity and opening of fractures in the vent area during the two days that
preceded the eruption.
e error associated with the height estimate up to 1536 is relatively low (< 1 m): we therefore use this year to
distinguish the long-term caldera deformation from the short-term deformation preceding the eruption, from
1536 to 1538, reconstructed from several sources almost coeval with the eruption
18–22
. In the evaluation of the
short-term deformation we do not take into account for the deformation occurred approximately two days before
the onset of the eruptive activity, in order to exclude any contribution from the emplacement of the dike feeding
the eruption; however, due to the very rapid deformation, the evaluation of the total amount of deformation in
sites close to Monte Nuovo in the days before the eruption is aected by a larger error (up to 2.5 m).
Aer the eruption, all the recovered deformation data show a generalized renewal of the subsidence, with
maximum values in the Pozzuoli area.
e nature of the data (inferred from historical and archaeological records) makes it dicult to precisely infer
the amount and extent of horizontal deformation that accompanied the vertical deformation. However, records
of ground tilt oer an additional constraint on the deformation eld. To this aim, tilt changes between 1536 and
1538 have also been reconstructed (Supplementary Information and Supplementary Fig. S4).
Modelling. Our reconstruction shows that between 1251 and 1536 a general cumulative upli aected the
inner caldera, with a maximum value of 14 m in Pozzuoli (Fig.3a). e largest part of this deformation occurred
between 1400 and 1536, with a maximum of about 12 m in Pozzuoli (Fig.2 and supplementary Table S1).
erefore we used the data of this interval to model the source. e deformation from 1536 to 1538 (Fig.3b) is
centred on the area of the future eruption, with a maximum upli of ~19 m; in this period the upli along the
eastern Pozzuoli Bay shows a trend similar to that of 1251–1536 (Supplementary Table S1).
We model the caldera’s crust as a homogenous, isotropic, elastic half space. Once we include the data uncer-
tainty in our inversion model, our relatively simple approach allows us to get results that can be compared with
those of more complex, numerical models. erefore, while being aware of the limitations of our models (as
occurring in most, if not all, models of volcano deformation), we also emphasize that our models provide a
rst-order analysis with an estimated and acceptable error
23,24
.
We did not attempt to model any contribution to the deformation eld from the Campi Flegrei hydrothermal
system or from structural discontinuities associated with the caldera. e nature, extent, and permeability of the
pre-1538 hydrothermal system are highly uncertain, so any attempt to model the eect of magma accumulation
on it would have introduced a set of largely unconstrained variables. erefore, any modelling considering the
role of the hydrothermal system would have just introduced a higher set of non-constrained variables. Similar
considerations also hold in considering any pre-existing discontinuity in the modelling: our general knowledge
and data on the subsurface of Campi Flegrei are still too limited to include a reliable and univocal analysis taking
into account for pre-existing fractures.
Figure 3. Distribution of the surface upli preceding the Mt. Nuovo eruption. From 1251 to 1536 (a) the
upli aects the whole caldera, with a maximum in the Pozzuoli area. From 1536 to 1538 (b) the upli is centred
in the area of the future eruption (Monte Nuovo). Digital Terrain Model by INGV-Osservatorio Vesuviano.
www.nature.com/scientificreports/
5
Scientific RepoRts | 6:32245 | DOI: 10.1038/srep32245
Another limit of our analysis is that elastic deformation models have very similar near-eld vertical deforma-
tion for a range of source geometries
25
. Resolution of the geometry of a source would require the inversion of 3D
deformation data
24
that cannot be inferred from the existing historical and archaeological records.
e inversion of the deformation data suggests that the best-t source for the 1400–1536 upli is a radially
symmetrical intrusion (solutions from a sphere, spheroid or sill are very consistent regarding location, depth
and volume change – see Supplementary Table 2) in the caldera centre (Fig.4a), 1 to 2 km south of Pozzuoli at
a depth between 3.9 and 5.5 km and with a volume change of 0.93 to 0.95 km
3
(Fig.4b). e best-t solution for
the 1536–1538 upli and tilt change is a radially symmetrical intrusion lying 3.2 to 4.4 km beneath Monte Nuovo
(Fig.4a,b), with a volume change of 0.21 to 0.34 km
3
. A smaller, shallower intrusion beneath Monte Nuovo (depth
0.1 to 0.6 km) explains the large upli at four sites close to Monte Nuovo (Fig.4c). e volume change of this
second source (0.03 to 0.05 km
3
) is consistent with the volume of pyroclastic ejecta from the eruption
7
. Again,
Figure 4. (a) Source location (1400–1536 and 1536–1538) – the white line gives the position of the section
shown in (b); (b) Source depth (1400–1536 and 1536–1538); (c) deformation proles and best-t models.
Details of the best-t sources are available in Supplementary Table S3. A site by site comparison between the
data and the best t models is available in Supplementary Table 4. Error bars are 1 standard deviations. Digital
Terrain Model by INGV-Osservatorio Vesuviano.