Comprehensive geophysical prospection of the Roman and late antique city of Pollentia (Alcúdia, Mallorca, Spain)

An extensive magnetic survey has been carried out on a large part of the Roman and late antique city of Pollentia (Alcúdia, Mallorca, Spain), combined with ground penetrating radar (GPR) and electric resistivity imaging (ER Imaging) to obtain data for a better understanding of the ancient city.


| INTRODUCTION
The application of geophysical prospection to the study of the Roman past has already a long tradition in archaeological research (e.g., Gaffney & Gater, 2003;Witten, 2006).From initial localized studies, the improvement of the methods and equipment has prompted the application of geophysics to explore large areas of cities and other Roman sites using different techniques (Vermeulen, 2016).
The results have been increasing in the last few decades, and there are already many examples of extensive geophysical surveys.The city of Viroconium (Wroxeter) (Gaffney et al., 2000), using gradiometry, or Verulamium (Lockyear & Shlasko, 2017) are examples of this type of approach.The excellent results obtained at Falerii Novi provided an image of the entire Roman city using GPR (Verdonck et al., 2020).Also, in the Roman city of Interamna Lirenas in central Italy, and showed the potential of these surveys was evident (Verdonck et al., 2018).Geophysics has been also combined with other remote sensing methods in a more holistic approach (e.g., Hanson et al., 2019;Vermeulen, 2016).
The investigation of the topography and urbanism of the Roman city has a certain tradition, and several hypotheses for the layout and organization have been proposed, with special attention to the orientation of the structures (e.g., Doenges, 2005;Mar & Roca, 1998;Orfila, 2012;Orfila & Moranta, 2001).However, in many aspects, the extent of the archaeological remains uncovered is still limited.Geophysical prospection has been previously applied in the city.We are fortunate of counting on one of the first applications by Martin Aitken who in 1962 applied electrical resistivity to locate the continuation of the city wall of Sa Portella (Aitken, 1962).In the decade of 1990, Albert Casas and his team from the University of Barcelona helped to relocate some of the old excavations.In more recent years, several geophysical surveys have helped to obtain interesting data in different parts of the city (Ranieri et al., 2010(Ranieri et al., , 2016;;Trogu et al., 2011).However, these surveys were rather limited in extension and were planned to solve specific questions rather than to obtain an image of the city layout at large.In this sense, an extensive geophysical survey could help to provide data to fill the gaps of earlier partial geophysical investigations and to better understand the city, helping to locate areas of particular interest to plan future archaeological excavations.
The peculiarity of the city of Pollentia is its situation (Figure 1): on one hand, the ancient structures are not covered by a modern city, and on the other hand, it has densely populated and rapidly developing surroundings.Therefore, the conditions for magnetic measurements were partly rather disadvantageous due to the manifold sources of modern disturbances.Whereas most areas offered favorable surface conditions for investigation, in the North and the Northeast of the survey area some plots of land are still used as parking sites resulting in a significantly higher density of modern disturbances.Despite all modern interferences, the magnetic data give a surprisingly clear insight into the ancient urban structures.Following the magnetic survey, GPR measurements on selected areas were applied to broaden the understanding of the ancient urban structures.In addition, geoelectrical measurements were realized in the central part of the city.
At the Roman and late antique city of Pollentia (Alcúdia, Mallorca, Spain), more than 50 years after the first attempts of geophysical investigations by Martin Aitken a progressive geophysical prospection program was undertaken with the aim of developing an extensive geophysical survey with the intention to cover the entire city.The strategy combined full coverage with magnetometry, the use of ground penetrating radar (GPR) in selected areas, and electrical resistivity imaging (ER Imaging) in specific locations.The goal was to connect the results of excavations and previous geophysical measurements at isolated spots within a comprehensive data set to obtain the first full image of the structures of the entire city.

| THE ROMAN AND LATE ANTIQUE CITY OF POLLENTIA
The archaeological site of Pollentia is located in today's suburbs of Alcúdia, a municipality in the North of Mallorca.Situated on an elevation between the Bay of Pollença and the Bay of Alcúdia, the Roman and late antique city was of strategic importance in the navigation routes of the western Mediterranean and in the access to the island, also controlling the channel between Mallorca and Minorca.The terrain of the ancient city gently slopes to the South, towards the Bay of Alcúdia, where the main ancient port is suspected.A secondary port could have existed in the northern Bay of Pollença (Figure 1).
Pollentia was founded in 123 BC, but archaeological excavations proved that that location was already inhabited by the pre-Roman islanders.The central part of the Roman town, including the Forum was edified on a leveled area.The remaining terrain undulations were partly refilled to create a plain surface.The archaeological record also testifies to the existence of a calcareous layer of up to 0.30 m thickness underneath the Roman foundations.It can thus be assumed that the foundations of Pollentia follow targeted urban planning.
The most striking and still visible structures of the Roman and Late Antique city are the Forum where we find part of the forum square with the major Tuscan temple, two minor temples, an insula of tabernae (a block of shops and workshops) flanked by two porticoed streets west and east, and a porticoed street to the north with another insula of tabernae (Orfila et al., 1999).In this area, we also find the Early Byzantine fortification that protected the old forum area, and a large necropolis.To the north, the main visible remains are the residential area of Sa Portella with three houses (House of the two hoards, House of the bronze head and Northwest house), a porticoe street running east-west and a minor street north-south.Also in this area slightly further south, we find the remains of the House of Polymnia.To the south, the main standing and visible building is the Roman theatre in the southeast.The long-term excavations, initiated in 1923, also provided evidence for the existence of other residential and industrial areas including the existence of pottery kilns in the urban area.The results of the first geophysical investigations at Pollentia, carried out by Martin Aitken in 1962 (Aitken, 1962)  the area is free of modern buildings and agricultural areas although in general the entire area has been deeply changed by modern impacts since Alcúdia rapidly started to grow in the 80s of the 20th century.
Therefore, disturbed magnetic data caused by buried and open deposits of construction material and scrap metal or by fences and posts can especially be expected along the roads and pathways of the area.

| METHODOLOGY, EQUIPMENT AND INTERPRETATION RULES
The city of Pollentia survey project was performed to fill the gaps of previous isolated surveys and to extend the knowledge of the ancient city's structure from the better-known Forum to the partly known Roman settlement limits.We also wanted to explore these limits to bring new data over the full extension of Pollentia and its outskirts.
The medieval city of Alcúdia developed just a few meters to the north of the Roman city.The southern limit of the old medieval city wall that encloses the old part of the current village of Alcúdia is literally around 40 m away from the supposed northern limit of the ancient city.This has been the great fortune of Pollentia, left undisturbed in cultivation fields at the southern outskirts of the medieval, modern, and contemporary village.However, the characteristics of the actual landscape and the proximity to the modern city posed major problems to the geophysical exploration of the site.The contamination by recent metal and debris deposits, as well as the diverse vegetation cover and partly uneven surfaces affected both, the data collection, and the quality of the magnetic and GPR data sets.In addition, parts of the area of interest for prospection were not accessible for measurements.Certainly, some areas show the presence of a few modern constructions still dispersed on the site, such as the medieval chapel of Santa Anna or a series of constructions used by farmers like shelters, implement sheds, waterwheels, or simply masonry walls used to separate the different land properties.
The survey strategy was designed on the base of these conditioning factors, establishing a general coverage of the known archaeological area with magnetic gradiometer survey as a starting point.GPR and ER were then used to obtain additional information where the magnetic maps indicated areas of archaeological interest, or where contamination by recent structures or iron objects limited the information provided by magnetometry.
Despite these factors and the constraints given by the local geology, the successive survey campaigns have brought significant archaeological information, thanks to the combination of the information of complementary geophysical data sets and detailed interpretation work.
The magnetic survey covered a total area of 18 ha (Figure 2), providing the magnetic maps that guided the selection of areas where complementary data were needed.The magnetic data sets show significant variations in the contrast and intensity of anomalies caused by archaeological features along the different regions of the site, with a better definition of the building remains into the west and north of the forum.New insights into the extension of the Roman occupation were found towards the northwest of the known Roman city limits.
However, the influence of metal fences, water pipelines, and modern rubble blinds important part of the data, especially in the north of the site, today used as car parking.Other zones of the site, such as the area of Santa Anna and the south-eastern outskirts yielded more complex data with low magnetic contrasts, influenced by large-scale anomalies caused by geological entities.
The GPR survey covered 14 selected areas, with a total extension of 37,300 m 2 (Figure 2).As it will be shown in further sections, these surveys offered results of different quality.The main constraint for data processing and visualization came from the limited penetration related to dry and clayey soils.Moreover, poor ground contact of the antennae had to be taken on areas with high vegetation and stony or uneven surfaces.Another important issue in data treatment and interpretation came from the high contrast between the thin cultural layers and limestone bedrock outcrops in the southern area of the site (Ca 0 n Fanals necropolis).
Additionally, ER Imaging was applied in the center of the ancient city to investigate the anomaly of a possible street or ditch surrounding this area.
As observed on magnetic maps, the areas to the west and north of the forum show a better resolution of the building remains, pointing to a better conservation state of the archaeological layers.

| Magnetic prospection
A total surface of 18 ha was investigated (Figure 2) by magnetic prospection at the site of Pollentia for complete coverage of all accessible areas to identify large units of the Late Antique city with its fortifications as well as individual buildings and other archaeological features.
The survey included not only the supposed intramural part but also extramural areas, especially in the west of the forum.The magnetic survey was conducted by using the gradiometer array LEA MAX with 7 Förster fluxgate gradiometers of the CON650 type (Figure 3).The In a Geographical Information System (GIS), the magnetic data images were thoroughly examined for anomalies that might indicate archaeological features.To enable differentiation of the multitude of anomalies, the data were examined visually, using different data thresholds or grey scale dynamics.The general approach to classifying magnetic anomalies is to distinguish them respectively by means of their intensity, polarization, and geometric shape.As part of the first step, anomalies of unambiguously modern origin, indicating ferromagnetic objects, were separated and marked.These features, which usually form distinct dipole or multipole anomalies, can be recognized by high absolute values of the magnetic gradient above about 50 Nanotesla (nT), a second type of feature can be observed and recognized: irregularly shaped zones of diffusely increased or decreased magnetic gradient are attributed to shallow geological formations and outcrops of the bedrock.
The following step was to sort the remaining anomalies that were assumed to have an archaeological background.To structure these anomalies, several classes were introduced with corresponding causal physical structures.Linear features with negative values of the vertical gradient were associated with construction remains made of marès, the local biocalcarenite.The negative magnetic gradient values point to a predominant diamagnetism of the construction material.Positive anomalies indicate the presence of backfilled pits and ditches.Their fillings, a mixture of organic matter and settlement debris, are characterized by the presence of induced and remanent magnetization, resulting in increased values of the magnetic gradient.The induced magnetization is most likely due to biogenic magnetite.Stripes of slightly increased magnetic gradient can also be caused by compact sediments at the surface of ancient roads and streets, where material with an induced magnetization is accumulated.Therefore, it is possible to confuse them with backfilled ditches, so that interpretation also relies on additional data (resistivity, GPR) as well as attention to the archaeological context.Another feature typefireplaces, hearths, and kilnscan be recognized by magnetic anomalies of higher intensity 1962 for the evaluation of the measurements he made with a protonprecession magnetometer on the ancient city complex (Aitken, 1962).

| GPR prospection
The GPR surveys carried out on the Pollentia site have been based on area explorations to apply the time-slicing imaging technique (Conyers & Goodman, 1997).For all surveys, IDS GPR systems were used: in the early surveys, a two antennae Hi-Mod dual-frequency (200 MHz and 600 MHz) system was used at the Ca 0 n Fanals area.
Later works covering urban and suburban spaces of the site were carried out using a custom IDS system consisting of an array of five parallel monostatic 600-MHz antennae (Figure 4).All the GPR surveys used the same settings: a time window of 60 ns at an inline resolution of 35 scans per meter and a spacing of 0.2 m between profiles.All GPR measurements were carried out in summer under dry weather.Thus, only very slight changes in soil moisture are expected caused by differences in air humidity and not by precipitation.
The time-slice GPR data processing consists in the integration of single GPR profiles of known position, covering a given area, in a single 3D data matrix.This allows the production of images (time-slices) representing the reflection amplitudes in a selected depth range, typically in the form of consistent sequences from surface to increasing depths.These sequences could also be used to produce schematic maps by isolating only reflective structures above a reflection amplitude range (Goodman & Piro, 2013;Schmidt & Tsetskhladze, 2013).
These outputs (time-slice sequences and simplified vectorial maps) are the basis of the further interpretation process, which aims to produce comprehensible, classified interpretation maps that can be used to communicate and discuss the survey results in archaeological terms.
The datasets have been processed using the GPR-Slice software to produce the time-slice sequences.The data processing comprises two main phases: profile filtering and time-slice plot creation.As the main volume of data came from 600-MHz IDS antennae, a common processing string was applied, with some local variations due to the soil conditions at the moment of the surveys.Initially, the data was converted to the GPR-slice internal format (data conversion).After conversion, the individual raw profiles were filtered with background removal.Then, the resulting data were corrected using the function "bandpass and gaining" from GPR-Slice.This function consists in filtering the profiles using a bandpass filter and an AGC gain curve function.Eventually, the gain curve was set manually, since the aim of the AGC function in GPR-Slice is to propose a gain curve that compensates for the signal decay with depth.Using the AGC algorithm, several gain curves of individual profiles were calculated and evaluated to finally set up a general gain curve manually.The bandpass limits have been established according to the specific spectra of each dataset.In general terms, the low-cut values varied from 300 MHz to 340 MHz, while the high-cut values have been maintained at 900 MHz in all datasets.The creation of time-slice sequences consisted of the resampling of processed profiles in a series of 16 views from 0 to 32 ns, each one representing absolute amplitude maps of a time-lapse of 3.8 ns, or 32 samples.
The initial raw data processing aimed to clear raw data from frequency noises and to introduce a gaining curve that permits the creation of time-slice sequences that could reproduce the geometry of detectable archaeological remains.
The conversion from time to depth of data was made from examining hyperbola shapes on each dataset to obtain an approximation to propagation velocities.Due to the variability of soil moisture conditions during the survey campaigns, the measurements varied in a range of velocities between 0.08 m/ns to 0.1 m/ns.Since the latter value is predominant in the entire data set, all the depth estimations on the paper have taken a consistent velocity of 0.1 m/ns.The velocity analysis will be verified based on excavation results yet to be obtained.
The resulting time-slice sequences have been exported as raster maps and examined in the context of a GIS project to produce simplified vectorial interpretation maps.As some of the data showed important variations in contrast and short penetration ranges, alternative views have been produced to understand the disposition of detected features at increasing depths.In these cases, it has been particularly useful the overlay function of GPR-slice.It allows plotting accumulative maps from individual time-slices in a defined time-depth range.
The interpretation diagrams have been produced by analyzing the sequences and time-slices to identify the reflective, linear, or extensive features that could correspond to building remains or other archaeological elements.Low-reflection anomalies or heterogeneous response areas have been also considered important information in areas where modern fillings or debris deposits produced by agricultural works could cover or replace archaeological remains.In the northern and southern regions of the Roman city, the strong anomalies attributed to bedrock outcrops are also important, as excavations demonstrated the use of cuts into the geological base to create burials or wall foundations.

| Electrical resistivity imaging (ER imaging)
In addition to the magnetic and GPR prospection, ER Imaging was realized in the central part of Pollentia.These investigations aimed to support the interpretation of magnetic and GPR data at the assumed ditch or street structure, surrounding the forum.It was supposed that this enclosure had different functions during the existence of the late antique city.In the magnetic data, it manifests itself with increased intensity of the magnetic gradient, which could indicate a ditch filling and layers of compacted soil as found along roads and streets.The sequence was measured three times.The results were averaged when the standard variation was smaller than 0.5.Otherwise, the measurement was repeated.Before starting the actual measurements, a ground check was made to safeguard that the contact resistance did not exceed a value of 1.5 kΩ (kilo-Ohm).The measurements were carried out in the summer season during a hot and dry period.Thus, it was expected to register comparatively high resistivity values and low contrasts between the construction material and the surrounding soil.
All electrode positions were determined by means of an RTK-GNSS device, and are available in the coordinate system ETRS 89 UTM Zone 31 N (EPSG: 25831).Position and altitude values were used for the creation of the resistivity models.
The processing of the geoelectrical data was done using the inversion algorithm BERT (Boundless Electrical Resistivity Tomography).Since the data from the four profiles were unfitted for a 3D inversion, the 2D version of the program package was used.The numeric inversion of BERT is based on the Finite Element method (FE).Thereby, the application of non-structured elements like triangles (for 2D inversion) safeguards high flexibility to model arbitrary geometries as found in terrains of complex topography.Additionally, information from coring or other direct methods can be included in the calculation of the resistivity model (Günther et al., 2006).
The survey resulted in images of the subsurface resistivity distribution.Knowing the resistivity of different material types, it is possible to convert the image into a model of the underground consisting of different materials.However, since the resistivity of rocks depends on water saturation, chemical properties of the pore water, the structure of the pore volume, and temperature the range of values can be very broad (Hauck & Kneisel, 2008).Because of overlapping resistivity value ranges of different materials, the conversion into underground models might be ambiguous.In the case of the enclosing structure in the center of Pollentia, it is expected that a refilled ditch causes a decrease in electrical resistivity, while a layer of stones, bricks, and pottery fragments as found along a street, would cause increasing of the values.

| RESULTS: GENERAL CONSIDERATIONS
In the case of magnetic measurements in a suburban environment, as is the case with the investigation of Pollentia, the data are of course not only influenced by the ancient remains but by all possible factors resulting from the recent use of the area and its surroundings.The surveyed area in the north and the east are located very close to the buildings of modern Alcúdia, resulting in visually perceptible contamination with iron and other waste.Moreover, a large part of the open area is used as parking for visitors of the archaeological park and the town's center of Alcúdia with the consequence of severe contamination with iron and other sources of strong dipole and multipole magnetic anomalies (Figure 6).
The significance of the magnetic data in terms of the structures is further limited by the effects of several modern water pipelines, wire fences, and installations of the archaeological park such as footpaths and information boards for visitors.
The modern impact is less severe on the agricultural fields in the west of the ancient city, even if deposits of scrap metal, construction waste, and rubbish can be found throughout the entire investigation area.To make matters more complicated, the intensity of the magnetic field variations over the ancient structures are much weaker than those of the ubiquitous modern influences.The low intensity of the recorded magnetic anomalies over ancient remains is not only due to the magnetic material properties but can also be explained by the partly large thickness of the surface layers, which is particularly valid for the southern and western parts of the investigated area.
The datasets produced in the successive GPR surveys show differences in penetration and contrast attributed to changes in the environmental conditions (soil moisture, vegetation, temperature), but also to the different preservation levels or ground contact conditions imposed by the site.These differences are relevant in an archaeological context that lays at variable depths and conservation degrees depending on the zones, and especially in clayey soils, altered by centuries of cultivation.As shown in recent experiences, the sites are like the sites of Son Peret o (Mas Florit et al., 2021) or Firella in terms of geological and archaeological contexts, and produce in GPR a high contrast of reflection of the building remains or limestone bedrock rises, but at the same time tend to rapid attenuation of pulses, reducing the effective penetration ranges between 1.1 m and 1.4 m in dry conditions (Davis & Annan, 1989).In the case of Pollentia, the datasets show these characteristics, but also an added problem posed by the different ground surface conditions, from flat grass cover to rugged terrains with higher vegetation, compacted clay, or cultivation fields.
The GPR survey results obtained in Pollentia show a variable quality.In zones where the building remains are shallow and well preserved, as found to the west of the forum area the time-slice sequences allow to trace the outlines of buildings and streets.However, in areas with poorer preservation of buildings and rugged surfaces, the resulting time slices are noisy and tend to expose data processing artifacts and blurred images.
The electrical resistivity data reflect well the underground situation since no disturbing effects of near-surface conductive objects such as metal contamination were observed.Additionally, the quality and significance of the data are affected by the low lateral resolution conditioned by the electrode distance of 0.50 m and the vertical resolution decreasing with the depth.This means that statements can be made about the investigated archaeological structures down to depths of about 2.5 m.Although only data from 3 profiles are available, a correlation of low resistivity values with a limited penetration depth of the georadar waves can be observed.
The interpretation process of the obtained datasets has shown the importance of the complementary magnetic and GPR data.This is more evident in areas such as in the GPR zone 1 (Figure 7), where the rich evidence provided by both the magnetic data and the GPR timeslice sequences enable the production of detailed interpretation diagrams containing not only a geometric description of the remains of buildings, derived from GPR data but also significant qualitative information, such as the delimitation of areas affected by fire based on the analysis of the magnetic data.
In other areas, where the problems with data quality posed by the site conditions (contamination, poor penetration) affected more one survey method than others.This allowed us to at least obtain information on almost all explored areas from one of the applied methods.

| Urban area (GPR zone 1)
An area of nearly 1.2 ha in the west of the Forum, which was almost entirely covered by the magnetic survey, has a dense concentration of linear features where the structure of an urban mesh can be easily recognized (Figure 7a-c).Despite local variations in magnetic contrast, the structure of buildings disposed along three roads has been traced.
The magnetic data show abrupt increases in contrast interpreted as the product of areas affected by fire or concentrations of highly magnetic materials such as remains of hearths and ovens, ceramic pavements, or building material and debris layers.It should be noted that Aitken, in 1962, had already postulated the existence of ovens and accumulations of fired material based on his magnetic measurements in this area.
It is remarkable that some areas of zone 1 produced poorer definition images in the GPR time-slice sequences (Figure 7d).Examining data in more detail, these differences seem to come from a variety of causes.In region Z1A, in the north of the zone, the GPR data reveal spare linear features and an extensive reflective anomaly, where the F I G U R E 6 Overview of magnetic prospection results.
F I G U R E 7 Magnetic and GPR prospection in the central urban area (zone 1): data and interpretation.magnetic map shows a larger rectangular area with clearer limits and strong magnetic gradient variations.
The sum of both datasets induces us to interpret that this area could contain the remains of a building complex that was burnt down, but the poorer contrast of GPR data could imply that the remains could lay deeper than the surrounding features.
Another conclusion extracted from the comparison of the data sets is related to the character of the streets crossing the zone.Still in zone 1, a clearer segment is visible in the north of region Z1B, showing moderate positive magnetic gradient values (2.5 to 6 nT), and almost no reflective features in the GPR time-slices (Figure 7d,e).
These characteristics point to a non-paved street, probably made of compact soils, with no clear evidence of sewage canals.
Geoelectric measurements were planned and executed to add data and to improve the interpretation.The results of the ER profile, that crossed the assumed street structure show zones of increased resistivity between depths of 1 and 2 m, which can easily relate to the remains of walls at both sides of the street.However, the space between the buildings, i.e., the position of the assumed street is characterized by only weakly increased values of the resistivity observed within a thin layer of about 0.50 m.This is another hint to a nonpaved street only consisting of a compacted layer of soil, containing construction debris and organic remains, which are responsible for the increased intensity of the magnetic gradient (Figure 8).Similar results were found at two other geoelectric profiles in the east and Santa deposited to create a base for the assumed street.This feature reflects a late construction phase of the city.Therefore, it cannot be excluded that this street follows a back-filled ditch or a wall that has been removed and cleared out.
In contrast to these findings, the data corresponding to a segment of another road further to the south of zone 1, show a higher magnetic alteration along 30 m (region Z1C in Figure 7).The GPR data from this position reveal a highly reflective feature covering the whole width of the street and indicating the presence of a stone slab pavement.The strong magnetic gradient with values between 9 and 20 nT, could be due to underlying fillings used to level the pavement or ultimately by the shallow layers that filled the space of the road after the abandonment of the city.
5.2 | North of the Forum (GPR zones 8-9, 14) Another area of interest corresponds to an extension located north of the Forum (Figure 9a).Here, the surface still reflects the abandoned field divisions and some recent building features.The magnetic and GPR datasets put in evidence the preservation of the remains of buildings in the subsoil, but apparently in a less dense (or at least poorly conserved) urban structure (Figure 9b,d).
The main difficulty in interpreting the data of this area is the abundance of disturbances produced by metal objects and small buildings.In the north of this area, currently used as a car parking, the alteration caused by modern contamination and wire fences is even stronger.
Despite the strong dipoles produced by iron objects, the magnetic map allows us to recognize some positive, extensive anomalies interpreted as remains of the road mesh, clearly coherent with other segments detected in the west and south of the Roman city.Remarkably, negative and linearly arranged anomalies can be recognized even in strongly disturbed areas, where the GPR data confirmed the existence of walls (Figure 9c,e).
The GPR datasets collected in this region show a clear geometrical disposition of spare building remains, even in disturbed areas where the magnetic data are too much dominated by dipoles.
However, the GPR survey was far from providing an effective description of the archaeological remains alone.Firstly, the uneven surfaces, modern structures, pavements, or remaining field divisions, impeding regular contact of the antennae with the surface, were an obstacle to a more extensive coverage with GPR measurements.
Secondly, the GPR data processing and time-slice sequence generation also presented problems with contrast parameters.The clayey soils of the site tend to rapidly attenuate the GPR pulses with depth (Mas Florit et al., 2021;Verdonck, 2012), allowing penetrations that barely exceeded 1.2 m.At the same time, the local limestone (marès) produced strong reflections in the clayey media.In this context, one can expect a good definition of the building remains, mainly made of limestone, for a certain depth, but the results do not exactly match this assumption.
The datasets covering the discussed zone were taken with diverse conditions of vegetation cover, temperature, and humidity.
The time-slice sequences obtained from the surveys show different contrasts between soils and buildings in relatively shallow depths.In addition, some modern building features produce stronger reflections than the Roman building remains.As a result, the time-slices exhibit the reflective anomalies produced by walls with significant variations inside the same region.In that case, the interpretation diagrams were produced using harder contrast adjustments on the time-slice sequences to consider weak reflectors, which help to understand part of the structures of the detected building remains.

| South of the forum and Santa Anna
The region to the south of the forum and surrounding the medieval chapel of Santa Anna produced some of the poorest results for the exploration of the Roman city (Figure 10a).The magnetic map, covering most of the area, shows diffuse, positive anomalies which can be easily associated with the remains of the street mesh, but just some spare linear features of negative values related to the buildings' remains (Figure 10b,c).The GPR survey covered three limited areas inside this region where the time-slice sequences show noncontinuous groups of linear features and a scatter of isolated reflectors, probably produced by debris layers and cultivation activity (Figure 10d,e).
The coincidence of both methods in reproducing a lower contrast and weaker anomalies seems to point to a poorer conservation status of the building remains in this region, but as seen in the northern zones, the limited GPR penetration does not allow to exclude the preservation of archaeological structures below a depth of 1.2 m.A distinctive example of this is observed at a 140 m-long magnetic positive anomaly, that crosses the region from NE to SW.This structure has a similar response pattern and dimensions as the one detected surrounding the urban area, following the axes of the urban street mesh.The GPR data taken at the position of this magnetic anomaly show no clear features except some accumulations of dispersed reflectors and low reflection areas in the deeper time-slices.Thus, the magnetic anomaly seems to be produced by a highly magnetized soil layer.However, the lack of detection of significant anomalies in the GPR data suggests that the structure producing the magnetic anomaly could lay deeper than the depth range reached by the GPR 600-MHz system, or the materials causing the magnetic anomalies produce no significant electromagnetic variations.This assumption can be extended to other areas of the same region, where the magnetic anomalies, associated with street remains, found no correspondence in GPR data.
To evaluate the interpretation, based on magnetic and GPR data,  vertical dimensions or cut into the limestone.As the depth of the limestone bedrock is variable, the strong reflections from outcrops make it difficult to image subtler reflecting features that lay in a sedimentary context just some meters away (Figure 11e).
The magnetic dataset obtained in this area evidence similar interpretation problems, as the bedrock outcrops cause diffuse distributions of the magnetic gradient.The data are also affected by recent rubble deposits with metal objects (Figure 11c,d).
As shown in Figure 11e-g, the reflections produced by bedrock outcrops and structures immediately above them, are much stronger than other structures embedded in the clayey soil.As a result, building structures such as walls 1, 2 or 4 are poorly visible both, in the maps and in the vertical GPR profiles.Thus, it cannot be excluded that other structures are preserved in the area, since they may be masked by the strong reflections of the bedrock, which is comparatively at shallow depths.

| CONCLUSIONS
This paper summarises the results obtained in successive campaigns developed in the Roman and Late Antique city of Pollentia.The step from the geophysical investigation of isolated spots to an approach of complete coverage has resulted in significant progress in the archaeological investigation of the site of Pollentia.Even though the modern transformations of the terrain express manifold interferences in the magnetic data, the survey provided evidence of many archaeological structures, and large parts of the urban system of the ancient city can be reconstructed based on the magnetic data.The ancient street network and several insulae can largely be reconstructed based on the magnetic data.While a southern limit of the ancient city can be detected, the extent towards the north, the west and the east remains unclear.This large-scale magnetic approach combined with a progressive coverage of the most promising areas with GPR offers a good strategy to deal with the geophysics of entire cities.Dense georadar (GPR) measurements are suitable for a more detailed investigation of the structures that were first detected by means of magnetic prospection and this can be applied in the future to complement areas only explored by magnetometry.More detailed information on the buried structures can only be obtained by thorough comparisons between geophysical data and the results of archaeological excavations.
The data proved, as expected for a Roman city, the existence of a mostly regular street grid and dense building structures, especially north and west of the Forum.The continuation of the regularly developed ancient city towards the south could also be confirmed.Within 100 m south of the Forum, an important anomaly suggests the presence of a massive limiting structure was localized.However, this can only be confirmed and dated by means of direct archaeological interventions like test soundings or excavations.However, a distinct northern limit of the city area was not evident in the data.In the area northwest of the Forum and south of Sa Portella, a quarter of remarkably high magnetic anomalies caused by ovens, kilns or the remains of a destructive fire could be detected.It covers an area of almost 3,000 m 2 .The eastern and western limits of the ancient city also remain unclear.Concerning the area of the necropolis of can Fanals, despite the less favorable measuring conditions, the geophysical survey, both magnetic and GPR data reveal important features that help to outline the characteristics of this suburb in the southern limit of the city.
In short, the comprehensive geophysical of Pollentia has helped to gain insight into the urbanism of this important city of the Balearic Islands in the Roman and Late Antique periods.A further step of the investigation will include the comparison of all the geophysical data obtained with the current hypothesis concerning the urbanism of the city and its evolution.
also confirm this.However, many of these old excavations were covered again and are not visible nowadays.The excavated residential quarter at Sa Portella gives a good insight into the Roman constructions.The houses generally demonstrate the typical atrium and impluvium houses.The main street's width is 6.75 m.The ancient water supply network remains partly unclear, but the existence of wells across the city is well-known.The existence of an aqueduct from Ternelles to Pollentia has been considered.The necropoleis of Pollentia form another factful archaeological archive.One of the important burial sites is located south of the urban area at Can Fanals.By the end of the 3rd century AD, on a date that we can fix between 270 and 280, a massive fire destroyed different parts of the city.Burnt layers found in many parts of the urban area, notably in the Forum and the northern part of the residential quarters, indicate a major destructive fire at this time.Although this event was certainly dramatic for the inhabitants of Pollentia, the city persisted throughout Late Antiquity albeit its perimeter and population were considerably shrunk and experienced a deep transformation.By the middle of the 6th century AD, in the Byzantine period of the island, a fortification was built to transform the Forum into a fortified enclosure, configuring a kind of citadel.Remains of its walls can be observed at the northern side of the Forum, partially occupying the east-west road that closed the northern part of the forum area.A more detailed description of the urban structures at Pollentia and the archaeological conclusions drawn from excavations and other investigations is found in Cau & Chávez-Álvarez, 2003.Today, the urban area of the Roman and late antique Pollentia belongs, for the most part, to the different institutions that are part of the Consortium of the Roman city of Pollentia.Now the larger part of F I G U R E 1 Location of the city of the Roman and late antique city of Pollentia in Mallorca (Balearic Islands, Spain), and plan of the archaeological remains.
probes were mounted on a light and foldable cart at a lateral distance of 0.50 m and connected with a Real-Time Kinematic Global Navigation Satellite System (RTK-GNSS) consisting of two GNSS receivers allowing the investigation of irregular survey areas with trees and other obstacles.The sampling rate of the data logging was 40 Hz, resulting in an inline point distance between 0.03 to 0.05 m as the measuring speed was 1 to 2 m/s.The data were subjected to standard processing steps such as offset and drift correction using the script-based Eastern Atlas decoding and processing routines.The resulting data were merged into equidistant grid files, generated by means of the cubic spline interpolation, with a mesh size of 0.25 m and transformed into full-dynamic georeferenced Tiff images in the reference system ETRS 89 UTM zone 31 N (EPSG: 25831).

F
I G U R E 2 Roman and late antique city of Pollentia: geophysical survey areas (aerial image: PNOA 2019, www.ign.es).compared to the anomalies caused by pit and ditch filling.The heating of the material results in a strong thermoremanent magnetization, expressing dipole anomalies of high intensity.The here applied interpretation scheme corresponds well with the results of magnetic prospection at late antique rural sites of Mallorca, where similar features were found (Mas Florit et al., 2018, 2021).Of course, especially in the context of an ancient city complex, the recognition of individual structures is made difficult by manifold overlays of different settlement phases.It is noteworthy that this classification essentially corresponds to the interpretation scheme already established by Martin Aitken in

F
I G U R E 3 Magnetic measurements with the fluxgate gradiometer array LEA MAX.F I G U R E 4 GPR measurements with the IDS Hi Mod array of five 600-MHz antennae.
comparison with the GPR data shows stripes of low electromagnetic reflectivity along this feature, suggesting the existence of a nearly homogeneous filling, e.g. of a ditch.The geoelectrical measurements on three profiles in Pollentia were carried out by use of the multi-electrode device 4PLight with 60 electrodes (Figure 5).The electrode spacing of 0.50 m safeguarded a satisfying lateral resolution.The penetration depth of approximately 4 m resulted from a maximum spacing of the Wenner-alpha configuration of 6.5 m.The total length of the three profiles was 29.5 m.Measurements were carried out in a forward direction.Each electrode F I G U R E 5 Electrical resistivity measurements with 4PLight multi-electrode device.
Anna.These observations allow the assumption that the large structure surrounding the central part of the ancient city is related to a street made of compacted earth containing construction rubble and organic remains.However, the increased magnetic gradient values can alternatively be produced by a base layer of a geologic matter F I G U R E 8 Magnetic prospection and ER imaging in the central urban area: data and interpretation.
additional geoelectrical measurements were executed along one profile.The inversion results of the ER profile, crossing the discussed linear structure, show a slight resistivity decrease in shallow depth (0 to 1 m) indicating a backfill of a slight depression or ditch.The moderate F I G U R E 9 Magnetic and GPR prospection north of the Forum: data and interpretation.increase of resistivity, observed in depths between 1.25 and 2.5 m points to a more compacted layer, possibly to a pavement with a purposefully created substructure.Compared to the resistivity anomalies caused by building remains, the values observed at the combined ditch and street structure are lower, which is probably due to a simple pavement of compacted earth rather than to a stone plastering.

F
I G U R E 1 0 Magnetic and GPR prospection in the Santa Anna area: data and interpretation.