Journal of Archaeological Science 35 (2008) 1389e1398 http://www.elsevier.com/locate/jas

Recent achievements in archaeomagnetic dating in the Iberian Peninsula: application to Roman and Mediaeval Spanish structures

M. Go´mez-Paccard a,*, E. Beamud b

a Research Group of Geodynamics and Basin Analysis, Department of Stratigraphy, and Marine Geosciences, Universitat de Barcelona, Campus de Pedralbes, E-08028 Barcelona, Spain b Research Group of Geodynamics and Basin Analysis, Paleomagnetic Laboratory (UB-CSIC), Institute of Earth Sciences ‘‘Jaume Almera’’, Sole´ i Sabarı´s, E-08028 Barcelona, Spain Received 18 May 2007; received in revised form 25 September 2007; accepted 8 October 2007

Abstract

Archaeomagnetic studies in Spain have undergone a significant progress during the last few years and a reference curve of the directional variation of the geomagnetic field over the past two millennia is now available for the Iberian Peninsula. These recent developments have made archaeomagnetism a straightforward dating tool for Spain and Portugal. The aim of this work is to illustrate how this secular variation curve can be used to date the last use of several burnt structures from Spain. Four combustion structures from three archaeological sites with ages ranging from Roman to Mediaeval have been studied and archaeomagnetically dated. The directions of the characteristic remanent magnetization of each structure have been obtained from classical thermal and alternating field (AF) demagnetization procedures, and a mean direction for each combustion structure has been obtained. These directional results have been compared with the new reference curve for Iberia, providing archae- omagnetic dates for the last use of the kilns. The inferred ages obtained from this study are consistent with the archaeological evidence available. Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: Dating; Archaeomagnetism; Iberian Peninsula; Secular variation; Earth’s magnetic field

1. Introduction the angle formed between the geographic north and the hori- zontal component of the geomagnetic field, ranging from Archaeomagnetism is mainly known by archaeologists as 0 to 360. I is the angle between the horizontal plane and a dating tool. This technique is based on two physical phenom- the geomagnetic field vector, it ranges from 90 to 90 and ena: first, the Earth’s magnetic field varies in space and ; is defined as positive downwards. The Earth’s magnetic field and second, certain archaeological structures heated to high intensity (F ) is usually measured in microTeslas (mT). temperatures can acquire a stable thermoremanent magnetisa- Measurements of TRM allow the determination of the di- tion (TRM) which is parallel and proportional to the ambient rection and intensity of the Earth’s magnetic field at the mo- magnetic field at the time of cooling. The mechanism of TRM ment of cooling of the burnt structure. Therefore, the past acquisition relies on the capacity of some naturally occurring secular variation (SV) of the Earth’s magnetic field can be de- minerals (principally iron oxides) to retain a permanent or termined for any given region by studying well dated and remanent magnetisation. heated archaeological structures of different ages. The archae- The direction of the surface geomagnetic field is usually de- ological structures suitable for this kind of studies include fined by the magnetic declination (D) and inclination (I ). D is kilns, furnaces, domestic ovens, thermal baths, burnt walls or soils, bricks, pottery, or any heated material that has ac- quired a TRM, recording the Earth’s magnetic field present * Corresponding author. Tel.: þ34 93 403 4888; fax: þ34 93 402 1340. at the time of cooling. It is worth pointing out that the direc- E-mail address: [email protected] (M. Go´mez-Paccard). tion of the ancient geomagnetic field can be obtained only

0305-4403/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jas.2007.10.005 1390 M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398 from structures found in situ since the last firing. The other 2006b). The Bayesian statistical approach allows the window finds, like ceramics or other displaced baked clays objects, width to be automatically adapted to the density of points can inform about the archaeointensity but not about the direc- along the time axis, making the points movable within their re- tion of the ancient geomagnetic field. Once determined, these spective dating error ranges. This method adds some prior SV reference curves can be used as a dating tool by comparing knowledge on the global nature of the curve to be estimated, the archaeomagnetic field information obtained from the ar- as it is assumed that the studied physical phenomenon varies chaeological material to be dated (direction and/or intensity in a smooth way. The Bayesian approach allows the fitting of the TRM) with the available SV curve of the Earth’s mag- of a spherical spline function based on roughness penalty to netic field for the region concerned. This analysis provides ar- the data in three dimensions (declination, inclination and chaeomagnetic dates for the last use of the archaeological time). Moreover, when available, this method can take into ac- structures, when the TRM was acquired. Obviously, this can count previous archaeologic/stratigraphic constraints during only be reliably applied in those regions for which reference the calculation process. The results are expressed as a mean SV curves are available. curve and an envelope (error) at a 95% confidence level. Several SV curves for the last millennia are available for The ‘‘real’’ curve will lie somewhere inside the error band. Europe, including France (Thellier, 1981; Bucur, 1994; Gallet The Bayesian SV curve obtained for Iberia (Go´mez-Paccard et al., 2002), Bulgaria (Kovacheva et al., 1998), Hungary et al., 2006b) spans approximately 2000 years, from 100 to (Ma´rton, 2003; Ma´rton and Ferencz, 2006), Germany 1959 AD. Several directions are available for each century (Schnepp and Lanos, 2005), Great Britain (Batt, 1997; Zana- throughout most of the record. However, there is still a need niri et al., 2007), Austria (Schnepp and Lanos, 2006) and Italy to increase the amount of data, especially between the 6th (Tema et al., 2006). These curves are currently used as a useful and 10th centuries AD and prior to 0 AD, where most data dating tool for archaeological studies in some of these coun- come from neighbouring countries. Despite these limitations, tries (see for example Kovacheva et al., 2004). these studies (Go´mez-Paccard et al., 2006a,b) finally made it Several archaeomagnetic studies carried out in Spain in possible to use archaeomagnetism as a dating tool for the Ibe- recent years allowed the establishment of the first SV curve rian Peninsula. for the Iberian Peninsula (Go´mez-Paccard et al., 2006b). The Obviously the reliability of the archaeomagnetic method aim of this work is to illustrate how this new curve can now used as a dating tool depends on the fidelity of the SV curve be used to date the last use of different burnt structures from used for comparison. The precision of archaeomagnetic dating Spain. To this purpose, four combustion structures from three depends on several factors and varies from case to case. It is archaeological sites with ages ranging from Roman to Medi- constrained by the rate of variation of the geomagnetic field aeval times have been studied and archaeomagnetically (declination and inclination), the envelope errors of the SV dated. curve, and the error of the archaeomagnetic direction of the structure to be dated. It is worth pointing out that the error 2. Archaeomagnetism in Spain of SV curve depends on the precision of the independent dat- ing of the archaeological structures on which the SV curve is Spain has a rich historical past that has been the subject of based, and the number of structures available for each period. intensive archaeological research, making this country a fa- This indicates not only the need to obtain more data but also vourable place to obtain detailed records of the variation of the need to study well dated materials in order to construct the Earth’s magnetic field during historical time. Despite this SV curves. Then, the success of the archaeomagnetic method remarkable potential, until recently, the number of archaeo- in Iberia requires a close collaboration between archaeologists magnetic studies for this region was limited. New studies and geophysicists from this region. This collaboration remains have increased the amount of data and it is now possible to re- fundamental to improve the method and the precision of trace the evolution of the geomagnetic field direction in this archaeomagnetic dating. region for at least the past two millennia. In order to document variation of the full geomagnetic vec- Go´mez-Paccard et al. (2006a) presented a compilation of tor, detailed and reliable measurements of the Earth’s mag- previously available archaeomagnetic directions for Spain netic field intensity are also needed. The establishment of an (Thellier, 1981; Pare´s et al., 1992; Oyamburu et al., 1996) intensity reference curve can also help to dating purposes, es- along with the results from 58 structures not studied previ- pecially in the case of displaced objects, where only the ar- ously, providing a catalogue of 63 archaeomagnetic directions chaeointensity, and occasionally the inclination can be spanning the last two millennia. In order to construct the SV obtained. Some archaeointensity studies have been performed curve for Iberia, Go´mez-Paccard et al. (2006b) compiled in Spain (Kovacheva et al., 1995; Nachasova et al., 2002a,b, a data set of archaeomagnetic directions from sites that fell 2007; Burakov et al., 2005, 2006; Burakov et al., in press; within a 900 km circle centred on Madrid. Madrid was used Go´mez-Paccard et al., 2006c). Despite these efforts, the exist- as the reference point because it is close to the geographic cen- ing archaeointensity data set for Iberia remains poor and only tre of the Iberian Peninsula (40.4 N, 3.7 W). The compiled eight archaeointensity values are available for the last two mil- data set, which includes 134 archaeomagnetic directions, lennia. More data are needed in order to reliably retrace the was treated by Bayesian modelling (Lanos, 2004) in order to evolution of the past geomagnetic field intensity for this re- obtain the first SV curve for Iberia (Go´mez-Paccard et al., gion, at least for the last thousand years. M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398 1391

3. Archaeomagnetic dating of four Spanish diameter), has a corridor as an entrance to the combustion archaeological structures chamber and lacks a central pillar or a grill on top. The kiln complex was interpreted as a ceramic production centre for Baked materials from four kilns sampled at three archaeo- building purposes. The ceramics recovered during the excava- logical sites in NE Spain (Can Xammar, Vila Caputxins, and tion were very scarce and did not allow a precise dating of the La Fornaca) have been studied and archaeomagnetically kiln complex. However, it was classified as a pottery produc- dated. In this section we present the results obtained and we tion centre of Late Roman age (3rd to 4th centuries AD). illustrate how the new SV curve for Iberia can be used as a dating tool. 3.2. The archaeomagnetic sampling

3.1. Archaeological background As stated above, archaeomagnetic dating consists in the comparison of an archaeomagnetic direction against the refer- The excavation works of the Can Xammar site (41.54 N, ence SV curve of the Earth’s magnetic field for the region. In 2.45 E) were promoted by the Culture Department of the order to obtain the archaeomagnetic direction, samples must Mataro´ Council and directed by E. Medina (ATICS). This ar- be oriented before they are taken from the archaeological chaeological site, which includes a kiln, is located in an area site. In the laboratory, TRMs are first measured in sample co- surrounded by the Middle Age fortress of the town of Mataro´. ordinates and then converted to geographic coordinates. For The Can Xammar kiln is a circular structure, about 2 m in di- this purpose, samples are oriented in situ using a magnetic ameter, built with a mixture of clay, sandstone and gravel and or solar compass. Caution must be taken to ensure that the covered by tiles in its upper part (see Fig. 1a). Not much inde- sampled objects have not experienced any post-baking move- pendent chronological information was available for this com- ment; otherwise the transformation from sample to geographic bustion structure. However, rebound ceramic pieces found coordinates would be completely meaningless. Several indi- within the combustion chamber of the kiln set the abandon- vidually oriented samples per structure must be taken in order ment of the structure around the 15th century. to minimize sampling ‘‘noise’’ and to average any systematic The Culture Department of the Mataro´ Council also pro- error introduced during sampling. moted the archaeological works in the roman villa of Vila Ca- In all the examples presented in this work, baked clay cy- putxins (41.55 N, 2.45 E). The excavation was directed by lindrical samples were collected using a portable electrical the archaeologist J. Herna´ndez (ACTIUM Patrimoni Cultural, drill with a water-cooled diamond bit, following the standard SL). This villa includes a square plan pottery kiln with a grill palaeomagnetic sampling method. For each kiln, about 10e supported by 3 arches, 2 of them reinforced by a central pillar, 12 independently oriented samples (10.8 cm3) were taken and a corridor as an entrance to the combustion chamber (see through the overall structure (walls, grill, and central pillars) Fig. 1b). Both the entrance corridor and the combustion cham- (Fig. 1). The in situ azimuth and dip of cores were measured ber were excavated into the geological terrain. They were with a compass coupled to a core orienting fixture. When pos- made up of a mixture of clay and sand, now partially rubefa- sible, samples were oriented in situ with a solar compass (Vila cient by the high temperatures during heating. The top grill Caputxins kiln). When either the structure features or the was built with planar cobbles and baked clays. The abandon- weather conditions hindered the use of the solar compass, ment of the kiln was tentatively established around the 1st to sample orientation was taken in situ with a magnetic compass 2nd centuries AD, based on the scarce pieces of ceramic ma- (Fornaca and Can Xammar kilns). terial found inside the combustion chamber. Further evidence for the age of the kiln arose from a clay deposit next to the kiln 3.3. Experimental methods and results which was used as a clay source area to produce the ceramics. By the end of the 2nd century AD the clay deposit was aban- Samples were demagnetized using the conventional doned and used as a dump as indicated by the ceramic pieces methods employed in archaeomagnetism: either heating the found within it (J. Herna´ndez, personal communication). specimen (thermal demagnetization) or exposing it to an alter- The archaeological works promoted by the Museum-Ar- nating magnetic field (AF demagnetization). Analyses con- chive of Vilassar de Dalt and conducted by the archaeologist sisted in stepwise demagnetization and measurement of the N. Rosello´ in the La Fornaca kiln complex (41.52 N, 2.37 remanent magnetization at each step. Remanent magnetization E) allowed the recovery of three kilns of roman age in an was measured on a superconducting rock (2G unusually excellent state of preservation. Kilns 1 and 2 are Enterprises) at the Paleomagnetic Laboratory of Barcelona circular structures, about 4 m diameter, with a central pillar (SCT UB-CSIC). supporting a grill. The kilns of the complex are made up of Some samples from La Fornaca and Vila Caputxins were a mixture of clay, sand and pebbles. The combustion chambers demagnetized both thermally (demagnetizer TSD-1, Scho¨n- show glass coat remnants indicative of high temperatures and stedt), and by alternating field (AF) (demagnetizer GSD-5, partial melt within the combustion chamber. The studied kilns Scho¨nstedt). In all cases, thermal demagnetization proved to are schematically shown in Fig. 1c. The third kiln, which be more effective than AF demagnetization. 10 to 16 demag- could not be sampled, differs in morphology from the other netization steps were applied to the samples up to 580 Cor structures of the complex, as it is smaller (about 2 m in 100 mT. Results of progressive demagnetization are 1392 M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398

Fig. 1. Sample location in the different kilns. (a) Can Xammar kiln. (a1) General view of the combustion structure. Location of images a2 and a3 is indicated. (a2) Kiln detail and location of samples 1 to 5 and 11. (a3) Kiln detail and location of samples 6 to 10. (b) Vila Caputxins kiln. (b1) Top view of the kiln and location of the archaeomagnetic samples. (b2) Front view of the combustion chamber entrance and location of the archaeomagnetic samples. (c) La Fornaca kiln complex. (c1) Ground plan of the kiln complex. (c2) Detail of kilns 1 and 2 and sample location. M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398 1393

(a) Can Xammar, XM-01 TH TH TH up/W 2A up/W 6A up/W 7A 400º 580º N 580º N 400º N 560º 560º 100º 300º 400º 300º 100º

300º

100º

NRM: 0.9A/m NRM: 1.9A/m NRM: 1.1A/m TH TH TH up/W 9A up/W 10A up/W 11A 560º N 300º N N 300º 100º 100º 560º 400º 100º 400º 400º300º 580º

NRM: 3.0A/m NRM: 2.8A/m NRM: 2.9A/m

(b) Vila Caputxins, VC-01 TH1B TH2A TH 3C up/W up/W up/W N N 560º N 540º 560º 570º 560º 540º

540º 520º 520º

520º 500º 500º 500º 450º 450º 450º 390º 390º

NRM: 13.6A/m NRM: 7.5A/m NRM: 12.8A/m TH 5B AF 7C TH 9B up/W up/W up/W N 540º N 570º N 520º 100 mT 500º 540º 80 mT 60 mT 480º 40 mT 520º 500º 390º 450º 390º

NRM: 3.6A/m NRM: 7.7A/m NRM: 8.2A/m

(c) La Fornaca, FR-01 and FR-02 AF TH TH up/W 01-1A up/W 01-8A up/W 01-9A N 100mT N N 80 540º 560º 500º 520º 60 500º 460º 50 40 480º 400º 30 460º 20 15 320º 400º 10 5 240º NRM: 2.1A/m NRM: 9.3A/m NRM: 0.7A/m TH TH TH up/W 02-3A up/ W 02-7A up/W 02-11B N N N 550º 560º 530º 530º 520º 500º 520º 500º 500º 460º 460º 460º

400º 400º 400º 240º

NRM:8.1A/m NRM: 7.7A/m NRM: 3.4A/m

Fig. 2. Orthogonal projection of the remanent magnetization vectors during progressive demagnetization for the different samples from (a) Can Xammar; (b) Vila Caputxins; (c) La Fornaca. Open (solid) symbols represent projections on vertical (horizontal) planes. TH indicates thermal demagnetization and AF alternating field demagnetization. 1394 M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398

0° 0°

10° 10°

XM-01 VC-01

N =7 N =9 Ds = 10.3° Ds = 357.4° Is = 58.1° Is = 55.3° α95 = 3.0° α95 = 3.1° 90° 90° 270° 90° 270° 90°

0 0°

10° 10° FR-01 FR-02

N = 12 N = 12 Ds = 359.2° Ds = 1.1° Is = 51.2° Is = 50.0° α95 = 3.3° α95= 4.9° 270° 90° 90° 270° 90° 90°

Fig. 3. Stereographic projection of the archaeomagnetic directions calculated for each sample, together with the mean direction and a95 error circles for each of the four structures studied. N indicates the number of independently oriented samples taken into account for the calculation of the mean; Ds and Is, mean declination and inclination in situ; a95 and k, 95% confidence cone of mean directions and precision parameter from Fisher statistics. represented in vector end point or Zijderveld diagrams (Zijder- More than 7 mean (independently oriented) sample directions veld, 1967), in which the direction and the intensity of the have been used in the calculation of each site’s mean direction magnetization is displayed on a single diagram by projecting (Fig. 3, Table 1). the vector onto 2 orthogonal planes (see Fig. 2). Natural Rem- The mean direction for each structure was calculated using anent Magnetization (NRM) of the studied samples ranged Fisher (1953) statistics and the concentration parameter k and from 0.7 to 36.4 A/m, which are typical values for thermorem- confidence factor a95 were obtained. The parameter k is a mea- anent magnetization acquired by baked clays. Most of the sure of the scatter of the individual sample directions around samples revealed a well defined single magnetic component the mean direction and a95 is the semi-angle of the cone pointing to the origin of coordinates. Only a few samples around the mean direction in which the true direction occurs showed also a viscous component which was usually removed with a probability of 95%,. Finally, each site direction has around 200 C(Fig. 2). The magnetization decay during heat- been recalculated to Madrid coordinates (40.4 N, 3.7 W) fol- ing occurred between 450 C and 560 C pointing to (titano)- lowing the Via Pole conversion method (Table 1). This method magnetite as the carrier of the magnetization. uses the declination and inclination of the studied site to cal- Characteristic remanent magnetization directions were cal- culate the latitude and longitude of its corresponding magnetic culated by principal component analysis (Kirschvink, 1980). pole (paleopole). The paleopole coordinates are then used to Mean directions for each site have in all cases been calculated calculate the new declination and inclination at the reference following a hierarchical structure (i.e. specimen / indepen- locality (Madrid). The new archaeomagnetic directions con- dently oriented sample / site). When more than one specimen verted to Madrid coordinates are then ready to be compared was available from an independently oriented sample, a mean to the SV reference curve for Iberia. sample direction was calculated using all the specimens or Results in situ and relocated to Madrid coordinates are the most reliable demagnetisation experiment was retained. shown in Table 1.

Table 1 Archaeomagnetic results

Name Lat. ( N) Long. ( E) n/NDs ( ) Is ( ) Dm ( ) Im ( ) k a95 ( ) Archaeomagnetic dating (years AD) Xamar, XM-01 41.54 2.45 11/7 10.3 58.1 9.7 56.5 407.0 3.0 [533, 754] and [1492, 1609] Caputxins, VC-01 41.55 2.45 12/9 357.4 55.3 356.7 54.6 270.6 3.1 [71, 437] Fornaca 1, FR-01 41.52 2.37 12/12 359.3 51.2 358.1 50.3 169.9 3.3 [117, 375] and [1439, 1447] Fornaca 2, FR-02 41.52 2.37 11/11 1.1 50.0 359.8 48.9 78.5 4.9 [132, 345] and [1252, 1497] Columns from left to right: Name, name of the structure; Lat. and Long., Latitude and Longitude of the site; n/N, number of specimens analysed (n)/independently oriented samples taken into account in the calculation of the mean site direction (N ); Ds and Is, declination and inclination in situ; Dm and Im, declination and inclination at Madrid (40.4 , 3.7 ); k and a95, precision parameter and 95% confidence limit of characteristic remanent magnetisation, from Fisher statistics; Archaeomagnetic dating, archaeomagnetic dates at 95% confidence level obtained by comparison of the combined probability density function with the secular variation reference curve for the Iberian Peninsula. M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398 1395

3.4. Comparison with the reference SV curve for Iberia: archaeological evidence, [117, 375] and [132, 345], indicate archaeomagnetic dating that both kilns were contemporaneous at 95% confidence level. This result is in agreement with the archaeological con- In archaeomagnetism the event to be dated is the last cool- straints that suggest that La Fornaca was a pottery production ing of the baked structures after heating at high temperatures. centre of Late Roman age (3rd to 4th centuries AD). The dating procedure consists of calculating the probability density functions of possible dates for both declination and 4. Conclusions and perspectives inclination. These probability density functions are then combined to obtain the most probable solution at a certain A first reference curve for the last two millennia is now confidence level. This can be done using the REN-DATE soft- available for the Iberian Peninsula. Therefore, the archaeo- ware, developed by P. Lanos, which is based on Bayesian sta- magnetic method has now reached a where it can be tistics (Lanos, 2004). In this study we calculated the possible ages at 95% confidence level. This procedure is illustrated in 25 archaeomagnetic error archaeomagnetic error Fig. 4 for the kiln sampled at Can Xammar. The probability 20 densities at the 95% level of possible dates on the basis of dec- 15 lination and inclination results for the Can Xammar kiln are 10 9.7 shown in Fig. 4. Several possible solutions are obtained for 5 each case. However, the combined probability density function shown in Fig. 5, indicates that there are only two possible so- 0 lutions ([533, 754] or [1492, 1609]). Of the two, the interval -5

[1492, 1609] is in agreement with archaeological evidence, Declination (°) -10 SV curve for Iberia as the ceramic rebounds suggested an abandonment of the -15 structure by the XV century. The interval [533, 754] is there- -20 fore not plausible for this structure and can be rejected. Thus, using both archaeomagnetism and scarce archaeological infor- -25 mation we can confirm and enhance archaeological informa- tion, dating the last use of the kiln between 1492 and 1609 AD, with a 95% confidence level. The same dating procedure was also carried out for the function other structures. The probability density functions and the 95% inferred dates are shown in Figs. 6e8. For the kiln sampled Probability density at Vila Caputxins archaeological site, only one solution was 0 200 400 600 800 1000 1200 1400 1600 1800 Age (years) obtained with a 95% confidence. Despite it being a wide range, the solution obtained [71, 437] is in agreement with 65 the archaeological evidence. The terminus ante quem (TAQ) of the structure is the 1st or 2nd centuries based on ceramics fragments found in the oven infill. Archaeological evidence 60 suggests that the use of the kiln was related to a clay deposit that was in use until the end of the 2nd century. The TAQ is 56.5 therefore set at 200 AD by archaeological evidence. This in- 55 formation, together with the archaeomagnetic dating allows

us to determine the last use of the kiln was within the range Inclination (°)

[71, 200]. SV curve for Iberia 50 In the La Fornaca kiln complex, two kilns were studied. The main purpose of this study was to establish whether or not the two kilns were contemporaneous and to determine 45 the TAQ for their use. From the results, presented in Table 1 and Fig. 3, we can first conclude that the two studied kilns in La Fornaca were contemporaneous, as the archaeomagnetic mean direction for kiln 1 and kiln 2 are statistically indistin- guishable with a 95% confidence level. function Results shown in Figs. 7 and 8 indicate two possible dates 95% Probability density for the first kiln: [117, 375] and [1439, 1447]; and, similarly, 0 200 400 600 800 1000 1200 1400 1600 1800 two dates for the second kiln: [132, 345] and [1252, 1497]. Age (years) The intervals [1439, 1447] and [1252, 1497] can be rejected Fig. 4. Probability density function (PDF) obtained by Bayesian statistics for on the basis of the archaeological evidence. The final dating the Xammar kiln for (a) declination and (b) inclination. All obtained by com- results, obtained from combining archaeomagnetism and parison with the SV curve for Iberia at 95% confidence level. 1396 M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398

Archaeomagnetic dating of Can Xammar kiln Archaeomagnetic dating of Fornaca kiln 1 PDF PDF Declination Declination PDF PDF Inclination Inclination

Possible dates: Possible dates: [533, 754] and [1492, 1609] [117, 375] and [1439, 1447] PDF PDF Combined Combined 95% 95% 0 200 400 600 800 1000 1200 1400 1600 1800 0 200 400 600 800 1000 1200 1400 1600 1800 Age (years) Age (years)

Fig. 5. Archaeomagnetic dating of the kiln of the Can Xammar archaeological Fig. 7. Archaeomagnetic dating of the first kiln of the La Fornaca archaeolog- site. The probability densities (PDF) obtained by Bayesian statistics for (a) ical site. The probability densities obtained by Bayesian statistics for (a) dec- declination; (b) inclination; and (c) combined. The final archaeomagnetic lination; (b) inclination; and (c) combined. The final archaeomagnetic dating dating at 95% confidence level obtained from the combined PDF is shown. at 95% confidence level obtained from the combined PDF is shown.

Archaeomagnetic dating of Caputxins kiln Archaeomagnetic dating of Fornaca kiln 2 PDF PDF Declination Declination PDF PDF Inclination Inclination

Possible dates: Possible dates: [71, 437] [132, 345] and [1252, 1497] PDF 95% PDF Combined Combined 95%

0 200 400 600 800 1000 1200 1400 1600 1800 0 200 400 600 800 1000 1200 1400 1600 1800 Age (years) Age (years)

Fig. 6. Archaeomagnetic dating of the kiln of the Can Caputxins archaeolog- Fig. 8. Archaeomagnetic dating of the second kiln of the La Fornaca archae- ical site. The probability densities obtained by Bayesian statistics for (a) dec- ological site. The probability densities obtained by Bayesian statistics for (a) lination; (b) inclination; and (c) combined. The final archaeomagnetic dating declination; (b) inclination; and (c) combined. The final archaeomagnetic dat- at 95% confidence level obtained from the combined PDF is shown. ing at 95% confidence level obtained from the combined PDF is shown. M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398 1397 used as a reliable dating technique for this region over most of Bucur, I., 1994. The direction of the terrestrial magnetic field in France during the last 20 centuries. Four kilns have been analysed and satis- the last 21 centuries. Phys. Earth Planet. Int. 87, 95e109. factorily dated using this new SV curve. The inferred ages Burakov, K., Nachasova, I., Najera, T., Molina, F., Camara, H., 2005. Geomag- netic Intensity in Spain in the Second Millennium BC. Phys. Solid. Earth, obtained from this study are consistent with the scarce Engl. Transl. 41, 622e633. archaeological evidence available. In addition, the archaeo- Burakov, K., Nachasova, I., Mata, C., 2006. Geomagnetic field intensity in the magnetic study has produced a more precise date range for first millennium BC from data on ceramics of the Los Villares archaeolog- the last use of the studied structures, demonstrating the poten- ical monument (Spain). Phys. Solid. Earth, Engl. Transl. 42, 942e950. tial of the archaeomagnetic method. The precision of the Burakov, K., Nachasova I., Lorrio A., Archaeomagnetic investigation of ce- ramic from settlement El Molon (Spain). Phys. Solid. Earth, Engl. Transl., archaeomagnetic dates obtained varies depending on the preci- in press. sion of the SV curve for the period in question and the error of Fisher, R.A., 1953. Dispersion on a sphere. Proc. Roy. Soc. London A 217, the mean archaeomagnetic direction of the structure to be 295. dated. Gallet, Y., Genevey, A., Le Goff, M., 2002. Three millennia of directional Archaeomagnetic research must continue in the Iberian variations of the Earth’s magnetic field in western Europe as revealed by archaeological artifacts. Phys. Earth Planet. Inter. 131, 81e89. Peninsula in order to extend the time range covered by the ref- Go´mez-Paccard, M., Catanzariti, G., Ruiz-Martı´nez, V.C., McIntosh, G., erence SV curve and improve its quality as a dating tool. It is Nu´nez,~ J.I., Osete, M.L., Chauvin, A., Lanos, P., Tarling, D.H., Bernal- most evident that the curve must be improved for the period Casasola, D., Thiriot, J. Archaeological Working Group, 2006a. A cata- previous to 200 BC and from AD 400 to 900. The latter cor- logue of Spanish archaeomagnetic data. Geophys. J. Int. 166, 1125e1143. ´ responds to the so-called Dark Ages, where few archaeomag- Gomez-Paccard, M., Chauvin, A., Lanos, P., McIntosh, G., Osete, M.L., Catanzariti, G., Ruiz-Martı´nez, V.C., Nu´nez,~ J.I., 2006b. The first archae- netic data are available in Europe. Another way to improve the omagnetic secular variation curve for the Iberian Peninsula. Comparison accuracy of this method is to date using not only directional with other data from Western Europe and with global geomagnetic field results, but also intensity data. The use of all the three mag- models. Geochem. Geophys. Geosyst. 7, Q12001, doi:10.1029/2006 netic parameters always leads to a better result, since more GC001476. ´ ´ constraints are involved. However, there is a lack of a well es- Gomez-Paccard, M., Chauvin, A., Lanos, P., Thiriot, J., Jimenez-Castillo, P., 2006c. Archaeomagnetic study of seven contemporaneous kilns from Mur- tablished reference curve for the variation of the intensity of cia (Spain). Phys. Earth Planet. Int. 157, 16e32. the Earth’s magnetic field over the last millennia in Western Kirschvink, J.L., 1980. The least-squares line and plane and the analysis of Europe. Future archaeomagnetic research must be devoted to paleomagnetic data. Geophys. J. Int. 153, 146e458. the determination of the full geomagnetic field vector. Kovacheva, M., Pare´s, J., Jordanova, N., Karloukovski, V., 1995. A new The establishment of reliable SV curves of the ancient geo- contribution to the archaeomagnetic study of a Roman pottery kiln from Calahorra (Spain). Geophys. J. Int. 123, 931e936. magnetic field for Iberia is not only of geophysical signifi- Kovacheva, M., Jordanova, N., Karloukovski, V., 1998. Geomagnetic field var- cance, but also a great help for . In this sense, it iations as determined from Bulgaria Archaeomagnetic data. Part II: The is important that archaeologists and archaeomagnetists collab- last 8000 years. Sur. Geophys. 19, 431e460. orate, not only to investigate new structures but also to revise Kovacheva, M., Hedley, I., Jordanova, N., Kostadinova, M., Gigov, V., 2004. and include any new archaeological information on the context Archaeomagnetic dating of archaeological sites from Switzerland and Bul- garia. Journal of Archaeological Science 31, 1463e1479. of the structures used as reference points for the SV curve Lanos, P., 2004. Bayesian inference of calibration curves: application to ar- established for Iberia. chaeomagnetism. In: Buck, C., Millard, A. (Eds.), Tools for Constructing : Crossing Disciplinary Boundaries, Vol. 177. Springer, Lon- Acknowledgements don, pp. 43e82. Ma´rton, P., 2003. Recent achievements in archaeomagnetism in Hungary. Geo- phys. J. Int. 153, 675e690. The authors are greatly indebted to the Museu Arxiu de Vi- Ma´rton, P., Ferencz, E., 2006. Hierarchical versus stratification statistical anal- lassar de Dalt, Patronat de Cultura de l’Ajuntament de Mataro´, ysis of archaeomagnetic directions: the secular variation curve for Hun- ACTIUM Patrimoni Cultural and ATICS who kindly allowed gary. Geophys. J. Int. 164, 484e489. us to publish the data. Thanks are also due to P. Lanos who Nachasova, I., Burakov, K., Bernabeu, J., 2002a. Geomagnetic field intensity variations in Spain. Phys. Solid. Earth, Engl. Transl. 38, 371e376. generously provided free access to REN-DATE software. Mi- Nachasova, I., Burakov, K., Bernabeu, J., 2002b. Archaeomagnetic studies guel Garce´s and Juan Cruz Larrasoana~ are gratefully acknowl- of the ceramic material from the ‘‘Cendres Cave’’ multilayer neolithic edged for their help during sampling and discussion of results. site. Geomagn. Aeron. 42, 808e813. The authors would like to thank two anonymous reviewers for Nachasova, I., Burakov, K., Molina, F., Camara, J., 2007. Archaeomagnetic their constructive comments. Thanks to O. Ferna´ndez-Bellon study of ceramic from the Neolithic Los Castillejos multilayer monument (Montefrio, Spain). Phys. Solid. Earth, Engl. Transl. 43 (2), 170e176. for revision of the English grammar. This work has been Oyamburu, I., Villalain, J.J., Osete, M.L., Zarzalejos, M., Blasco, C., 1996. Es- carried out with the aid of a Juan de la Cierva Fellowship tudio paleomagne´tico del yacimiento de Villa del Panuelo~ (Villamanta, (M.G.-P.) and was also funded by the CGL2005-00211 project Madrid). Geogaceta 20, 1044e1046. from the Spanish Ministry of Education and Science. The Pare´s, J.M., De Jonge, R., Pascual, J.O., Bermu´dez, A., Tovar, C.J., authors thank the Paleomagnetic Laboratory SCT UB-CSIC. Luezas, R.A., Maestro, N., 1992. Archaeomagnetic evidence for the age of a Roman pottery kiln from Calahorra (Spain). Geophys. J. Int. 112, 533e537. References Schnepp, E., Lanos, P., 2005. Archaeomagnetic secular variation in Germany during the past 2500 years. Geophys. J. Int. 163, 479e490. Batt, C.M., 1997. The British archaeomagnetic calibration curve: an objective Schnepp, E., Lanos, P., 2006. A preliminary secular variation reference curve treatment. Archaeometry 39, 153e168. for archaeomagnetic dating in Austria. Geophys. J. Int. 166, 91e96. 1398 M. Go´mez-Paccard, E. Beamud / Journal of Archaeological Science 35 (2008) 1389e1398

Tema, E., Hedley, I., Lanos, P., 2006. Archaeomagnetism in Italy: a compila- Zananiri, I., Batt, C.M., Lanos, P., Tarling, D.H., Linford, P., 2007. Archaeo- tion of data including new results and a preliminary Italian secular varia- magnetic secular variation in the UK during the past 4000 years and its ap- tion curve. Geophys. J. Int. 167, 1160e1171. plication to archaeomagnetic dating. Phys. Earth Planet. Int. 160, 97e107. Thellier, E., 1981. Sur la direction du champ magne´tique terrestre en Zijderveld, J.D.A., 1967. AC Demagnetization of rocks: analysis of results. In: France durant les deux derniers mille´naires. Phys. Earth Planet. Inter. Collinson, D.W., Creer, K.M., Runcorn, S.K. (Eds.), Methods in Paleomag- 24, 89e132. netism. Elsevier, New York, pp. 254e286.