A Building Information Modeling Approach to Integrate Geomatic Data for the Documentation and Preservation of Cultural Heritage

Non-destructive testing (NDT) techniques play an important role in the characterization and diagnosis of historic buildings, keeping in mind their conservation and possible rehabilitation. This paper presents a new approach that merges building information modeling (BIM) with environment geospatial data obtained by several non-destructive techniques, namely terrestrial laser scanning, ground-penetrating radar, infrared thermography, and the automatic classification of pathologies based on RGB (red, green, blue) imaging acquired with an unmanned aircraft system (UAS). This approach was applied to the inspection of the Monastery of Batalha in Leiria, Portugal, a UNESCO World Heritage Site. To assess the capabilities of each technique, different parts of the monastery were examined, namely (i) part of its west façade, including a few protruding buttresses, and (ii) the masonry vaults of the Church (nave, right-hand aisle, and transept) and the Founder’s Chapel. After describing the employed techniques, a discussion of the optimization, treatment and integration of the acquired data through the BIM approach is presented. This work intends to contribute to the application of BIM in the field of cultural heritage, aiming at its future use in different activities such as facility management, support in the restoration and rehabilitation process, and research.


Introduction
Built cultural heritage is delicate and, once lost, it is unrecoverable. It consists of architectural structures and other constructions that constitute the basic record of past human activities and, as such, should be protected [1]. The conservation and rehabilitation of historical buildings must be particularly cautious with the preservation of their singular characteristics. Hence, each such building must be fully examined in order to identify those unique characteristics: structure type, construction methods It should be emphasized that our proposal is to demonstrate that (i) datasets having diverse origins, and even nature, can be effectively included in a unique BIM model, and (ii) such a concentration of information can assist future users in obtaining a general 3D display and interpretation of the building and its condition at a given time. Note, however, that the paper does not cover the automatic extrapolation of additional information from the hosted information.
This paper is organized into six sections: (i) Section 2 provides a brief description of the Monastery of Batalha and the zone selected as the case study; (ii) Section 3 deals with the surveying techniques employed for characterization, diagnosis and generation of the 3D model, including the description of methodologies used for data acquisition and processing; (iii) Section 4 shows and discusses the results of these surveys; (iv) Section 5 presents the approach followed for data integration into a BIM model; (v) Section 6 summarizes the main conclusions.

The Monastery of Batalha
The Monastery of Batalha (MB), also called Monastery of Santa Maria da Vitória, is located in Batalha, Leiria, in the center of Portugal. It is inscribed on the UNESCO World Heritage List and its main styles are Gothic and Manueline, a Portuguese late gothic. The main part of the monastery, which is being covered by the present research, was built under Master mason Afonso Domingues between 1388 and 1402 and then under Master mason Huguet, who was responsible for the Founder's Chapel, until 1438 [56]. But new parts were added to the monastery (and others It should be emphasized that our proposal is to demonstrate that (i) datasets having diverse origins, and even nature, can be effectively included in a unique BIM model, and (ii) such a concentration of information can assist future users in obtaining a general 3D display and interpretation of the building and its condition at a given time. Note, however, that the paper does not cover the automatic extrapolation of additional information from the hosted information.
This paper is organized into six sections: (i) Section 2 provides a brief description of the Monastery of Batalha and the zone selected as the case study; (ii) Section 3 deals with the surveying techniques employed for characterization, diagnosis and generation of the 3D model, including the description of methodologies used for data acquisition and processing; (iii) Section 4 shows and discusses the results of these surveys; (iv) Section 5 presents the approach followed for data integration into a BIM model; (v) Section 6 summarizes the main conclusions.

The Monastery of Batalha
The Monastery of Batalha (MB), also called Monastery of Santa Maria da Vitória, is located in Batalha, Leiria, in the center of Portugal. It is inscribed on the UNESCO World Heritage List and its main styles are Gothic and Manueline, a Portuguese late gothic. The main part of the monastery, which is being covered by the present research, was built under Master mason Afonso Domingues between 1388 and 1402 and then under Master mason Huguet, who was responsible for the Founder's Remote Sens. 2020, 12, 4028 4 of 24 Chapel, until 1438 [56]. But new parts were added to the monastery (and others underwent important modifications) until 1530. From then until the beginning of the 19th century almost only conservation and restoration works were carried out. Between 1840 and 1900 the monastery suffered a significant intervention, which included the demolition of some parts, rebuilding some severely damaged elements, lowering by 65 cm the exterior pavement adjacent to the main (west) entrance, and also introducing some modifications to emphasize the gothic nature of the building, a regrettable practice of that period [57]. More recently, in the 1940s, another noteworthy intervention project took place.
The monastery was built with a compact sublithographic oolitic limestone (Bathonian stage) from the region [58]. Even though the intent in the 19th century rehabilitation intervention was to use the same material, different types of limestone, also from the region but weaker, were employed. As explained below in the results section, the GPR survey detected two different types of ashlar: one (probably) original and another one much more recent (probably applied in a restoration intervention). Unfortunately, there is no precise information about the type and date of application of the ashlars in the different parts of the monastery.
To assess the capabilities of each technique and to assure the representativeness of the gathered information, different parts and construction elements of the monastery were examined, namely (i) part of the monastery west façade, including some of its protruding buttresses, and (ii) some masonry vaults of the Church (nave, right-hand aisle and transept) and of the Founder's Chapel (see Figure 2). The west façade comprises three blocks of the monastery: the western aisle of the Royal Cloister, the Church (including its main entrance) and the Founder's Chapel.
Remote Sens. 2020, 12, x FOR PEER REVIEW 4 of 25 underwent important modifications) until 1530. From then until the beginning of the 19th century almost only conservation and restoration works were carried out. Between 1840 and 1900 the monastery suffered a significant intervention, which included the demolition of some parts, rebuilding some severely damaged elements, lowering by 65 cm the exterior pavement adjacent to the main (west) entrance, and also introducing some modifications to emphasize the gothic nature of the building, a regrettable practice of that period [57]. More recently, in the 1940s, another noteworthy intervention project took place. The monastery was built with a compact sublithographic oolitic limestone (Bathonian stage) from the region [58]. Even though the intent in the 19th century rehabilitation intervention was to use the same material, different types of limestone, also from the region but weaker, were employed. As explained below in the results section, the GPR survey detected two different types of ashlar: one (probably) original and another one much more recent (probably applied in a restoration intervention). Unfortunately, there is no precise information about the type and date of application of the ashlars in the different parts of the monastery.
To assess the capabilities of each technique and to assure the representativeness of the gathered information, different parts and construction elements of the monastery were examined, namely (i) part of the monastery west façade, including some of its protruding buttresses, and (ii) some masonry vaults of the Church (nave, right-hand aisle and transept) and of the Founder's Chapel (see Figure 2). The west façade comprises three blocks of the monastery: the western aisle of the Royal Cloister, the Church (including its main entrance) and the Founder's Chapel.  GPR and IRT surveys were performed on locations subject to past restoration interventions, in order to deepen our understanding of those interventions, which is required to improve future conservation and restoration plans. Moreover, the specific locations of GPR and IRT readings were chosen where better correlation with the characteristics of the façade could be obtained. In this sense, the GPR readings were made in order to obtain information about the internal structure of the different construction elements (such as walls, buttresses and roof). On the other hand, IRT was employed in the area where the classification of anomalies showed a larger variety ( Figure 2).
Three buttresses of the main façade have been investigated: one stiffening the south façade of the Founder's Chapel, and the other two corresponding to the Church. The one closer to the right of the main portal is aligned with the arcade that separates the nave from the side-aisle (that supports the clerestory, see Figure 2), while the one further right corresponds to the Church's south wall (one part of it forming the exterior façade and the other being connected to the Founder's Chapel).
The interior spaces of the Monastery of Batalha are all covered by ribbed vaults. Most of them are quadripartite vaults but there are also a few stellar vaults: the latter type was employed in the inner octagonal center of the Founder's Chapel (exterior view in Figure 2). All these vaults are covered by fill materials and limestone tiles (or plates).

Materials and Methods
This section includes the geomatic techniques employed in the surveys. The BIM-based approach developed is presented as a separate section after the results of these surveys and the generation of the 3D model of the monastery.

Unmanned Aircraft System (UAS) Survey: Methodology for Façade Orthophoto Generation
In this study, the UAS and their payload RGB cameras are synergistic used as a cost-effective and innovative inspection method that allows the assessment of the state of conservation of existing heritage structures and enhanced efficiency in damage classification. The orthophotos generated by processing the imagery obtained by the UAS and object-based image classification methods will support the automatic identification of the anomalies on the facades and rooftop of the heritage buildings, where access cannot be gained without the installation of a heavy, complex, expensive and time-consuming scaffolding.
The orthophoto of the main façade of MB was generated using an SfM (structure from motion) and multiview stereo (MVS)-based approach. This orthophoto was generated in Photoscan Pro v1.4 (Agisoft LCC, St. Petersburg, Russia) by using four steps of the standard workflow, which is represented in Figure 3. In step 1, a block of 324 images of the main façade of MB was acquired using a multirotor UAS (Phantom 4 Pro, DJI, Shenzhen, China) and a vertical flight path (the white circles in Figure 3b represent the positions of the image centers). In step 2, the BBA (bundle block adjustment) of the image block was performed in an arbitrary coordinate system. For scaling the model, 18 distances measured in the CAD (computer aided design) drawing of the façade were used (Figure 3a). This CAD drawing, supplied by MB's administration, was previously generated by vectorising an old 3D point cloud obtained by laser scanning, in which an RMSE (root-mean-square error) of about 1.5 cm was obtained. The final accuracy of the adjusted scale was 2.8 cm. This value is given by Agisoft Metashape and represents the RMSE of the 18 differences between the input distances (measured in the CAD drawing) and the estimated values (computed in the scaled model). In step 3, a dense 3D point cloud was generated (Figure 3c). In step 4, this point cloud was used to generate a textured model and subsequently the orthophoto (Figure 3d). A visual inspection was then carried out to identify and classify the anomalies on the main façade of the monastery. Taking into account the spectral resolution of the camera employed and the properties of the stone-reflected radiation, the pathology types described in Figure 4 were selected. Additionally, the automatic classification of the anomalies was performed using object-based image analysis (OBIA) and the k-nearest neighbor (KNN) classifier in the eCognition Developer v9.5 (Trimble GmbH, Munich, Germany) [59]. Image segmentation, the first step of OBIA classification, was undertaken using the multi-resolution segmentation algorithm with the following parameters: scale 65, shape 0.1, and compactness 0.4.

Ground-Penetrating Radar (GPR) Survey: Data Acquisition and Signal Processing
The GPR method was employed to assist in the characterization of the internal structure of the walls, piers, and roof, as well as to identify the different building materials, crucial information for investigating some of the detected pathologies.
As shown in Figure 2, eight zones (Z-1 to Z-8) of the main façade of MB were selected to be prospected with the GPR aiming to investigate the different building materials used for construction and reconstructions throughout history, namely to evaluate the use of recent stones, materials used in the buttress and the eventual connection with the main wall. Moreover, in order to evaluate different dome-filling materials, additional GPR measurements were carried out on the masonry roof of the Church-transept (1), nave (2) and, at a lower level, right-hand aisle (4)-and of the octagonal central and higher part of the Founder's Chapel (3), above its stellar ribbed dome. A visual inspection was then carried out to identify and classify the anomalies on the main façade of the monastery. Taking into account the spectral resolution of the camera employed and the properties of the stone-reflected radiation, the pathology types described in Figure 4 were selected. A visual inspection was then carried out to identify and classify the anomalies on the main façade of the monastery. Taking into account the spectral resolution of the camera employed and the properties of the stone-reflected radiation, the pathology types described in Figure 4 were selected. Additionally, the automatic classification of the anomalies was performed using object-based image analysis (OBIA) and the k-nearest neighbor (KNN) classifier in the eCognition Developer v9.5 (Trimble GmbH, Munich, Germany) [59]. Image segmentation, the first step of OBIA classification, was undertaken using the multi-resolution segmentation algorithm with the following parameters: scale 65, shape 0.1, and compactness 0.4.

Ground-Penetrating Radar (GPR) Survey: Data Acquisition and Signal Processing
The GPR method was employed to assist in the characterization of the internal structure of the walls, piers, and roof, as well as to identify the different building materials, crucial information for investigating some of the detected pathologies.
As shown in Figure 2, eight zones (Z-1 to Z-8) of the main façade of MB were selected to be prospected with the GPR aiming to investigate the different building materials used for construction and reconstructions throughout history, namely to evaluate the use of recent stones, materials used in the buttress and the eventual connection with the main wall. Moreover, in order to evaluate different dome-filling materials, additional GPR measurements were carried out on the masonry roof of the Church-transept (1), nave (2) and, at a lower level, right-hand aisle (4)-and of the octagonal central and higher part of the Founder's Chapel (3), above its stellar ribbed dome. Additionally, the automatic classification of the anomalies was performed using object-based image analysis (OBIA) and the k-nearest neighbor (KNN) classifier in the eCognition Developer v9.5 (Trimble GmbH, Munich, Germany) [59]. Image segmentation, the first step of OBIA classification, was undertaken using the multi-resolution segmentation algorithm with the following parameters: scale 65, shape 0.1, and compactness 0.4.

Ground-Penetrating Radar (GPR) Survey: Data Acquisition and Signal Processing
The GPR method was employed to assist in the characterization of the internal structure of the walls, piers, and roof, as well as to identify the different building materials, crucial information for investigating some of the detected pathologies.
As shown in Figure 2, eight zones (Z-1 to Z-8) of the main façade of MB were selected to be prospected with the GPR aiming to investigate the different building materials used for construction and reconstructions throughout history, namely to evaluate the use of recent stones, materials used in the buttress and the eventual connection with the main wall. Moreover, in order to evaluate different dome-filling materials, additional GPR measurements were carried out on the masonry roof of the Church-transept (1), nave (2) and, at a lower level, right-hand aisle (4)-and of the octagonal central and higher part of the Founder's Chapel (3), above its stellar ribbed dome.
A ground-coupled pulsed system manufactured by Malå Geoscience© (Malå, Sweden) was used, composed of a Proex control unit and three different antennas with central frequencies of 500, 800 and 2300 MHz ( Figure 5). The highest frequency provides greater resolution but lower signal penetration, while the lowest frequency provides poorer resolution but larger signal penetration. The setup used for data acquisition is summarized in Table 1. A survey wheel was attached to the antennas to control the trace-interval distance and measure the total length of the GPR profiles. Additionally, to guarantee positioning and accurate data integration, field marks were registered during data acquisition at the joints between stones.
Remote Sens. 2020, 12, x FOR PEER REVIEW 7 of 25 A ground-coupled pulsed system manufactured by Malå Geoscience© (Malå, Sweden) was used, composed of a Proex control unit and three different antennas with central frequencies of 500, 800 and 2300 MHz ( Figure 5). The highest frequency provides greater resolution but lower signal penetration, while the lowest frequency provides poorer resolution but larger signal penetration. The setup used for data acquisition is summarized in Table 1. A survey wheel was attached to the antennas to control the trace-interval distance and measure the total length of the GPR profiles. Additionally, to guarantee positioning and accurate data integration, field marks were registered during data acquisition at the joints between stones.  To enhance the extraction of information from the received GPR signals, they were first filtered and only afterwards the subsurface images with all interesting features were produced. The data was processed with the ReflexW software (Sandmeier geophysical research, Karlsruhe, Germany) [60] using the processing sequence described in Table 2. This filtering aimed to eliminate possible noise or interference with the signal, as well as to amplify the received signal (gain function) in order to mitigate possible losses or attenuations. To suppress the continuous component, a vertical or temporal filtering was applied (subtract-mean-dewow), determining and eliminating from each trace an average value based on the low energy of the last part of the wavelength. A horizontal or spatial filtering (background removal) was also applied to remove horizontal continuous low-frequency reflectors, which allowed estimating and removing an average value of all the traces in a time window. A band-pass filter (Butterworth) was then used to remove both low-and high-frequency noise in the vertical and horizontal directions. Finally, as presented in [61], migration processing (Kirchhoff) was used to suppress strong clutter aiming to improve the definition of the internal stone blocks.   To enhance the extraction of information from the received GPR signals, they were first filtered and only afterwards the subsurface images with all interesting features were produced. The data was processed with the ReflexW software (Sandmeier geophysical research, Karlsruhe, Germany) [60] using the processing sequence described in Table 2. This filtering aimed to eliminate possible noise or interference with the signal, as well as to amplify the received signal (gain function) in order to mitigate possible losses or attenuations. To suppress the continuous component, a vertical or temporal filtering was applied (subtract-mean-dewow), determining and eliminating from each trace an average value based on the low energy of the last part of the wavelength. A horizontal or spatial filtering (background removal) was also applied to remove horizontal continuous low-frequency reflectors, which allowed estimating and removing an average value of all the traces in a time window. A band-pass filter (Butterworth) was then used to remove both low-and high-frequency noise in the vertical and horizontal directions. Finally, as presented in [61], migration processing (Kirchhoff) was used to suppress strong clutter aiming to improve the definition of the internal stone blocks.
To transform the travel-time distance (ns) axis of the GPR image into a depth/distance (m) axis, the radar-wave velocity was previously calibrated for the two types of found ashlars (as explained before): the original one and the much more recent one. This calibration was performed, as in [62], by considering the travel-time difference measured in the GPR signal and the actual thickness of the stone. The radar-wave velocities obtained were 13.4 cm/ns for the original stones and 11.4 cm/ns for the more recent ones, which resulted in an average radar-wave velocity of 12.4 cm/ns. It is important to mention the technical difficulties experienced in data acquisition due to the excessive weighting of the antennas, especially with the 500 MHz antenna. As shown in Figure 5, a ladder was used to access the higher stonework of the façade and two operators were required in order to maintain both the GPR antenna and the survey wheel in permanent contact with the surface.

Infrared Thermographic (IRT) Survey: Thermal Image Acquisition and Processing
Infrared thermography (IRT) can be used for the investigation of historic structures, namely, to detect pathologies in the stone walls, during their heating or cooling phases, for example due to the presence of water that affects the surface temperature during the drying or wetting, or due to possible changes in the thermal diffusion of the wall [63].
The infrared images were captured with a thermographic Flir T335 camera (Flir Systems, Täby, Sweden). The inspection was carried out on the west façade of the Founder's Chapel (region highlighted in a blue rectangle in Figure 2). The capture of each single image was done as parallel as possible to the façade. Each single image was captured positioning the camera as perpendicularly as possible to the façade. The passive technique was used, profiting from the building façade heating or cooling according to the period of the day. Since the façade under study is facing west, the infrared images were taken in the morning to obtain acceptable weather conditions, when there was no direct solar radiation on the façade surface, avoiding heterogeneous heating of the stone.

Light Detection and Ranging (LiDAR) Survey: Data Acquisition and Processing
Light detection and ranging (LiDAR) scanning can be used to produce highly accurate 3D data and is becoming increasingly important for cultural heritage preservation, to capture building interiors [64] and provide digital documentation and deformation analyses of buildings [65][66][67]. Indeed, laser scanning captures the real condition of structures and its data can be used to create true as-built BIM models that can be relied on by engineers, architects, and contractors.
A tripod-mounted Faro Focus 3D X330 (Faro Technologies Inc., Stuttgart, Germany) [68], which is a phase-based TLS operating a laser of wavelength 1550 nm with an effective range of 330 m, was used for data acquisition (Figure 6a). A total of 58 scans were recorded in a complex geometry in the presence of occlusion. To capture the main façade of the monastery and the Founder's Chapel, 12 scans were performed outdoors and due to different object-to-scanner distances (ranging between 5 to 10 m), the computed surface density varies between 81, 428 to 20,285 points per square meter (pt/m 2 ). The remaining 46 scan positions were taken inside the Founder's Chapel, generating point clouds with a spatial resolution of 7 mm.
The point clouds from each scan were initially represented in the scanner's local coordinate frame. All the data were then aligned in a common, global coordinate system through a process known as registration (Figure 6b). For this purpose, plane targets fixed in common surfaces were used as control points in the Faro Scene software (Figure 6a). Besides, points on the floor were also coordinated in a local datum with a Global Positioning System (GPS). The Faro Focus 3D X330 was positioned on

Results and Discussion
This section includes the individual results obtained from the surveying techniques. The integration of the data into a BIM is presented and discussed in Section 5.

Results of the Anomalies Classification
The main aim of the process was to compare the results obtained from the classic visual inspection with an automatic procedure, and the level of degradation in some cases was defined based on the difference of the colors. The method allows us to obtain a quantification of the area affected by a certain pathology and, if repeated over time, the evolution in terms of area affected. Due to the low spectral resolution of the camera and the reflected radiation properties of the stone, only two of the six types of anomaly of Figure 4 could be automatically identified: patina and biogenic crusts. Actually, different severity levels of patina and biogenic crust stains can be distinguished, so that each of them is subdivided in two subclasses: patina 1 and 2 and biogenic crusts 1 and 2. For instance, patina 1 is the severest or more advanced (having a darker tone) while patina 2 is still in an initial state (lighter tone)-biogenic crusts were similarly classified. By visual inspection of the orthophoto, a stratified random sample of 358 points was manually identified and used for training the supervised KNN classifier (with 5 nearest neighbors). Figure 7a shows the classification results of façade pathologies obtained by the KNN classifier. These automatically identified pathologies are in agreement with the pathologies identified manually by the visual inspection of the façade orthophoto (Figure 7b), which took 104 h for an experienced operator.

Results and Discussion
This section includes the individual results obtained from the surveying techniques. The integration of the data into a BIM is presented and discussed in Section 5.

Results of the Anomalies Classification
The main aim of the process was to compare the results obtained from the classic visual inspection with an automatic procedure, and the level of degradation in some cases was defined based on the difference of the colors. The method allows us to obtain a quantification of the area affected by a certain pathology and, if repeated over time, the evolution in terms of area affected. Due to the low spectral resolution of the camera and the reflected radiation properties of the stone, only two of the six types of anomaly of Figure 4 could be automatically identified: patina and biogenic crusts. Actually, different severity levels of patina and biogenic crust stains can be distinguished, so that each of them is subdivided in two subclasses: patina 1 and 2 and biogenic crusts 1 and 2. For instance, patina 1 is the severest or more advanced (having a darker tone) while patina 2 is still in an initial state (lighter tone)-biogenic crusts were similarly classified. By visual inspection of the orthophoto, a stratified random sample of 358 points was manually identified and used for training the supervised KNN classifier (with 5 nearest neighbors). Figure 7a shows the classification results of façade pathologies obtained by the KNN classifier. These automatically identified pathologies are in agreement with the pathologies identified manually by the visual inspection of the façade orthophoto (Figure 7b), which took 104 h for an experienced operator.
For assessing the accuracy of the OBIA classification, a stratified random sampling was used ( Table 3). The overall accuracy of the automatic classification was high (76.5%) with a moderate agreement among the automatic classification and the reference data, given by a kappa coefficient k = 0.71. The producer's accuracy (pa) was lowest for class patina 2 (59%) and highest for patina 1 (88%) and the no pathology class (87%). The user's accuracy (ua) showed higher values for biogenic crusts 1 and 2 (respectively 92% and 84%) and much lower values for patina 1 (59%). For assessing the accuracy of the OBIA classification, a stratified random sampling was used ( Table 3). The overall accuracy of the automatic classification was high (76.5%) with a moderate agreement among the automatic classification and the reference data, given by a kappa coefficient k = 0.71. The producer's accuracy (pa) was lowest for class patina 2 (59%) and highest for patina 1 (88%) and the no pathology class (87%). The user's accuracy (ua) showed higher values for biogenic crusts 1 and 2 (respectively 92% and 84%) and much lower values for patina 1 (59%).   One of the advantages of using a super high-resolution orthophoto for mapping the façade pathologies is that it is possible to evaluate, in a simple and direct way, the area occupied by each (Table 4). The values presented in Table 4 allow is to conclude that about 51% of the study area presents biogenic crusts and the area values identified with both methods are very similar (automatic classification-660.51 m 2 , manual classification-658.66 m 2 ). Concerning the areas presenting Patina, a difference of about 37.6% was obtained between the two methods.  Figure 2). First, Figure 8a presents the data produced with the 500 MHz antenna in Zone 1 (façade buttress). Three different depths of the inner surface of the ashlars are identified: the first at~80-100 cm, the second at~150-170 cm and the third at~240-280 cm. Second, Figure 8b shows the 2300 MHz data obtained in Zone 2.1 (wall below lancet window of Founder's Chapel), while Figure 8c illustrates the 2300 MHz data produced in Zone 2.2 (buttress at the end of the arcade separating the nave from the right-hand aisle), showing the interpretation of the different stone's thicknesses. As seen in Figure 8b, a cavity (wall tomb) in the internal skin of the façade at Zone 2.1 causes (red ellipse) signal diffractions and complex pattern of reflections at the end of this profile (from 1.4 to 2.2 m), which were certainly produced by the cavity (gap) and the limestone arch above the internal wall tomb, respectively. Next, observing the 2300 MHz profile produced in Zone 3.1 (Figure 8d), it was realized that the bottom row of ashlars on the façade was recently restored using a stone~10 cm thick; whereas observing the profile acquired in Zone 3.2 (Figure 8e), it was inferred that the second and third rows (from the bottom of the wall) are composed of the original stones~50-60 cm thick. Figure 8f includes the radargram obtained with the 2300 MHz antenna in Zone 4, where both the original ashlars (thickness ranging from 40 to 70 cm) and those used in recent restoration interventions (thickness ranging from 2 to 12 cm) are detectable on the stonework of the façade. The thickness variation observed for a singular type of material is almost certainly caused by defects such as alveolization or by the unevenness of the inner face of the ashlars. Figure 8g presents the radargram obtained with the 2300 MHz antenna in Zone 5. As in Zone 4, it was possible to observe the presence of both original (~35 cm thick) and new (~8-10 cm thick) ashlars, the latter used for restoration. Regarding Zone 6, Figure 8h illustrates the radargram produced with the 500 MHz, showing three different layers of ashlar at~60, 130 and 200 cm in depth. The reflection observed at 200 cm depth at the beginning of the radargram is most certainly produced by the wall separating the Royal Cloister from the Church, which is perpendicular to the surveyed façade. Figure 8i presents the radargram obtained with the 800 MHz frequency antenna in Zone 7. Observing these radargrams, a single layer of ashlars~60 cm thick is identified. Finally, Figure 8j displays the radargram obtained with the 800 MHz frequency antenna in Zone 8, where both original (thickness ranging from 50 to 60 cm) and recent ashlars (thickness ranging from 5 to 10 cm) are distinguishable on the stonework of the façade. radargram obtained with the 800 MHz frequency antenna in Zone 8, where both original (thickness ranging from 50 to 60 cm) and recent ashlars (thickness ranging from 5 to 10 cm) are distinguishable on the stonework of the façade.  Figure 9 includes the GPR data acquired on the roof of both the Church and the Founder's Chapel (see Figure 2). First, Figure 9a presents the radargram obtained with the 2300 MHz frequency antenna through the roof of the Founder's Chapel (Figure 8b), showing the interpretation of the superficial limestone tiles and stronger reflections at the internal fill layers. Regarding this effect,  Figure 9 includes the GPR data acquired on the roof of both the Church and the Founder's Chapel (see Figure 2). First, Figure 9a presents the radargram obtained with the 2300 MHz frequency antenna through the roof of the Founder's Chapel (Figure 8b), showing the interpretation of the superficial limestone tiles and stronger reflections at the internal fill layers. Regarding this effect, previous maintenance works carried out on the roof of the Founder's Chapel, in which some cladding tiles were removed, revealed a fill layer made of lime mortar, stones and ceramic tiles. The scattering observed is certainly produced by the presence of such heterogeneous filling. Next, Figure 9b shows the radargram obtained with the 800 MHz frequency antenna on the roof above the right-hand aisle (second bay counting from the transept). It allowed for measuring the thickness of the cladding plates and of the internal fill layer (~20 cm and 80 cm thick, respectively), and to identify the top of the underneath ribbed vault (possibly its keystone). Finally, Figure 9c illustrates the radargrams obtained with the 800 MHz frequency antenna on the roof of the Church, above the nave and transept (see Figure 2), showing the cladding tiles and the internal fill layer.

Interpretation of the Produced GPR Images
Remote Sens. 2020, 12, x FOR PEER REVIEW 13 of 25 previous maintenance works carried out on the roof of the Founder's Chapel, in which some cladding tiles were removed, revealed a fill layer made of lime mortar, stones and ceramic tiles. The scattering observed is certainly produced by the presence of such heterogeneous filling. Next, Figure  9b shows the radargram obtained with the 800 MHz frequency antenna on the roof above the right-hand aisle (second bay counting from the transept). It allowed for measuring the thickness of the cladding plates and of the internal fill layer (~20 cm and 80 cm thick, respectively), and to identify the top of the underneath ribbed vault (possibly its keystone). Finally, Figure 9c illustrates the radargrams obtained with the 800 MHz frequency antenna on the roof of the Church, above the nave and transept (see Figure 2), showing the cladding tiles and the internal fill layer.  Figure 10 shows the images captured with the thermographic camera on the study area, and identifies the three wall spans t1, t2 and t3, bordered by the buttresses. Figure 10b shows the images captured on the wall span t3 (images P5 and P2) and bordering buttresses (images P1, P3, P4 and P6). In the areas closest to the ground (P1, P2 and P3) lower temperatures are observed (tones purple to blue in the thermogram), which are probably due to the presence of moisture, both absorbed from the ground (rising damp) and infiltrated in the façade, as confirmed by visual observation. The different colors are associated not only with different humidity levels but also with the material and pathology types-as they imply different absorption of solar radiation and, therefore, different energy emissions captured by the thermogram-and with the thickness of the limestone blocks. There seems to exist some match for areas with patina and also in the presence of stones resulting from recent restoration works in the upper middle part of P5 ( Figure 10). In addition, higher temperatures  Figure 10 shows the images captured with the thermographic camera on the study area, and identifies the three wall spans t1, t2 and t3, bordered by the buttresses. Figure 10b shows the images captured on the wall span t3 (images P5 and P2) and bordering buttresses (images P1, P3, P4 and P6). In the areas closest to the ground (P1, P2 and P3) lower temperatures are observed (tones purple to blue in the thermogram), which are probably due to the presence of moisture, both absorbed from the ground (rising damp) and infiltrated in the façade, as confirmed by visual observation. The different colors are associated not only with different humidity levels but also with the material and pathology types-as they imply different absorption of solar radiation and, therefore, different energy emissions captured by the thermogram-and with the thickness of the limestone blocks. There seems to exist some match for areas with patina and also in the presence of stones resulting from recent restoration works in the upper middle part of P5 ( Figure 10). In addition, higher temperatures (reddish color) are observed (1) near the pier buttresses, on the left-hand side of P5, which appear to be caused by the presence of the buttress, and (2) on the frieze at the top of P5, close to the window sill.

Three-Dimensional (3D) Model Generation and Structural Analysis
To create the 3D model the raw point clouds were first converted to a readable project file (RCP), using the Autodesk ReCap (San Rafael, CA, USA), which was then imported to a BIM environment using the Autodesk Revit software. The large amount of data contained in the indoor and outdoor point clouds allowed the actual dimensions of the surveyed walls and piers of the Founder´s Chapel to be determined, and subsequently to be introduced in a BIM Platform ( Figure  11). In addition, the point clouds were also used to evaluate the inclination of the inner columns of the Founder´s Chapel and therefore the deviations and alignments of the chapel.

Three-Dimensional (3D) Model Generation and Structural Analysis
To create the 3D model the raw point clouds were first converted to a readable project file (RCP), using the Autodesk ReCap (San Rafael, CA, USA), which was then imported to a BIM environment using the Autodesk Revit software. The large amount of data contained in the indoor and outdoor point clouds allowed the actual dimensions of the surveyed walls and piers of the Founder´s Chapel to be determined, and subsequently to be introduced in a BIM Platform ( Figure 11). In addition, the point clouds were also used to evaluate the inclination of the inner columns of the Founder´s Chapel and therefore the deviations and alignments of the chapel.

Three-Dimensional (3D) Model Generation and Structural Analysis
To create the 3D model the raw point clouds were first converted to a readable project file (RCP), using the Autodesk ReCap (San Rafael, CA, USA), which was then imported to a BIM environment using the Autodesk Revit software. The large amount of data contained in the indoor and outdoor point clouds allowed the actual dimensions of the surveyed walls and piers of the Founder´s Chapel to be determined, and subsequently to be introduced in a BIM Platform ( Figure  11). In addition, the point clouds were also used to evaluate the inclination of the inner columns of the Founder´s Chapel and therefore the deviations and alignments of the chapel.  The interior of the chapel was analyzed according to 8 vertical alignments corresponding to the 8 central columns of the Founder's Chapel, located between the windows of the octagon at the roof level. The coordinates of four points were taken in each vertical alignment at 3.7 m, 12.7 m, 13.6 m and 16.5 m from the floor (see Figure 12). In this figure, the lower level (3.7 m) dimensions are in natural scale and the others have a factor scale increase of 20 with respect to the coordinate difference to that lower level. The figure shows that the Founder's Chapel presents at its higher level a displacement to the south (maximum of 11.4 cm), east and west (maximum of 8.2 cm) sides, possibly due to misalignment during the construction stage or in consequence of seismic activity (the original roof collapsed during the Lisbon earthquake of 1755 [69]). The figure also shows that the part of the octagon on the side of the Church suffered smaller displacements possibly because of the structure's higher stiffness in that direction (it may even have been pushed away from the Church by its horizontal impulse). The results show an average column inclination of about 8 mm/m. These displacements are probably the main reason for the open joints above the windows. It is possible that either the roof collapse due to the earthquake of 1755 or the construction of the new roof may have worsened the octagon deformation. By comparing the results of future surveys with those now obtained, it will be possible to determine if this motion is still evolving or if it has already stabilized.
Remote Sens. 2020, 12, x FOR PEER REVIEW 15 of 25 The interior of the chapel was analyzed according to 8 vertical alignments corresponding to the 8 central columns of the Founder's Chapel, located between the windows of the octagon at the roof level. The coordinates of four points were taken in each vertical alignment at 3.7 m, 12.7 m, 13.6 m and 16.5 m from the floor (see Figure 12). In this figure, the lower level (3.7 m) dimensions are in natural scale and the others have a factor scale increase of 20 with respect to the coordinate difference to that lower level. The figure shows that the Founder's Chapel presents at its higher level a displacement to the south (maximum of 11.4 cm), east and west (maximum of 8.2 cm) sides, possibly due to misalignment during the construction stage or in consequence of seismic activity (the original roof collapsed during the Lisbon earthquake of 1755 [69]). The figure also shows that the part of the octagon on the side of the Church suffered smaller displacements possibly because of the structure's higher stiffness in that direction (it may even have been pushed away from the Church by its horizontal impulse). The results show an average column inclination of about 8 mm/m. These displacements are probably the main reason for the open joints above the windows. It is possible that either the roof collapse due to the earthquake of 1755 or the construction of the new roof may have worsened the octagon deformation. By comparing the results of future surveys with those now obtained, it will be possible to determine if this motion is still evolving or if it has already stabilized.

Data Integration into the Building Information Modeling (BIM) Model
This section describes the steps followed to merge into a BIM environment the geospatial data presented in the previous section, as well as the optimization and treatment of those data. The main objective is to make this model available for possible uses in the future, such as facility management, support to conservation activities, rehabilitation interventions, or further research.

Data Integration into the Building Information Modeling (BIM) Model
This section describes the steps followed to merge into a BIM environment the geospatial data presented in the previous section, as well as the optimization and treatment of those data. The main objective is to make this model available for possible uses in the future, such as facility management, support to conservation activities, rehabilitation interventions, or further research.

Methodological Approach for Data Integration
As mentioned in the previous section, as a first step of the development of the BIM model in Autodesk Revit, the LiDAR point clouds file (RCP) was imported into the BIM environment, and used to create the 3D walls of the façade, with the dimensions defined by the point clouds. In keeping with the object-oriented approach of BIM, this involved the creation of an initial wall object which was subsequently replicated by adjusting its properties. A similar approach was followed with the buttress piers (see Figure 13). In spite of the existence of several libraries of building elements in the Revit platform, specific objects had to be created for this project, such as the object walls, and the coating materials and their characteristics had to be introduced in the corresponding database.

Methodological Approach for Data Integration
As mentioned in the previous section, as a first step of the development of the BIM model in Autodesk Revit, the LiDAR point clouds file (RCP) was imported into the BIM environment, and used to create the 3D walls of the façade, with the dimensions defined by the point clouds. In keeping with the object-oriented approach of BIM, this involved the creation of an initial wall object which was subsequently replicated by adjusting its properties. A similar approach was followed with the buttress piers (see Figure 13). In spite of the existence of several libraries of building elements in the Revit platform, specific objects had to be created for this project, such as the object walls, and the coating materials and their characteristics had to be introduced in the corresponding database. To overcome the lack of interoperability between the software used to produce the classification map (eCognition Developer v9.5) and the Autodesk Revit (San Rafael, CA, USA), all the information was saved in JPEG format, which is supported by the software. Only afterward was the remaining information successively inserted into the model: the orthophoto images obtained with the UAS, the radargrams, the IRT images and the pathology classification maps (Figures 14 and 15).  To overcome the lack of interoperability between the software used to produce the classification map (eCognition Developer v9.5) and the Autodesk Revit (San Rafael, CA, USA), all the information was saved in JPEG format, which is supported by the software. Only afterward was the remaining information successively inserted into the model: the orthophoto images obtained with the UAS, the radargrams, the IRT images and the pathology classification maps (Figures 14 and 15).

Methodological Approach for Data Integration
As mentioned in the previous section, as a first step of the development of the BIM model in Autodesk Revit, the LiDAR point clouds file (RCP) was imported into the BIM environment, and used to create the 3D walls of the façade, with the dimensions defined by the point clouds. In keeping with the object-oriented approach of BIM, this involved the creation of an initial wall object which was subsequently replicated by adjusting its properties. A similar approach was followed with the buttress piers (see Figure 13). In spite of the existence of several libraries of building elements in the Revit platform, specific objects had to be created for this project, such as the object walls, and the coating materials and their characteristics had to be introduced in the corresponding database. To overcome the lack of interoperability between the software used to produce the classification map (eCognition Developer v9.5) and the Autodesk Revit (San Rafael, CA, USA), all the information was saved in JPEG format, which is supported by the software. Only afterward was the remaining information successively inserted into the model: the orthophoto images obtained with the UAS, the radargrams, the IRT images and the pathology classification maps (Figures 14 and 15).   Since this remaining information was georeferenced in an arbitrary coordinate system, all the information was co-registered inside the model using the dimensions defined by the point clouds or coordinates of common spatial elements (edges and corners), which are easily identifiable in the parametric model.
The strategy devised to integrate all the information into the model with their precise location and orientation was twofold: (i) the orthophoto images, the pathology classification maps and the IRT images, which are exterior information, were overlaid upon the model as wall and column claddings in a 3D plane, with the dimensions defined by the point clouds, while (ii) the GPR radargrams were positioned in the BIM 3D model using the field marks at the joints of the stones with their positions visible on the orthoimages.
The main difficulties found in these operations were caused by (i) the exterior wall and column surfaces pertaining to different plans and (ii) Revit software not allowing the simultaneous cladding of several coatings as different layers. To overcome these obstacles, a two-step procedure was conceived. First, the overlaid images and maps were cut out with the dimensions of the surfaces of those elements to which they were to be applied, to allow their insertion in the BIM model as coatings. Secondly, a virtual wall was generated for receiving each image or map as a coating, with all these virtual walls in the same position, and corresponding to the unique real wall. This solution allows the visualization of all images and classification maps in a unique BIM model, enabling their comparison and analysis ( Figure 15).
To include the GPR radargrams in the model with their actual direction, position and dimensions, several possible solutions were tested. The chosen one was to create for each radargram a fictitious wall-object belonging to the Revit wall family. Each of these Revit wall objects was then cladded with a radargram. This allows the visualization of the radargrams by simply hiding the façade wall ( Figure 16a). As an example, Figure 16b shows the 2300 MHz radargram corresponding to Zone 2.1 positioned in the BIM model.  Since this remaining information was georeferenced in an arbitrary coordinate system, all the information was co-registered inside the model using the dimensions defined by the point clouds or coordinates of common spatial elements (edges and corners), which are easily identifiable in the parametric model.
The strategy devised to integrate all the information into the model with their precise location and orientation was twofold: (i) the orthophoto images, the pathology classification maps and the IRT images, which are exterior information, were overlaid upon the model as wall and column claddings in a 3D plane, with the dimensions defined by the point clouds, while (ii) the GPR radargrams were positioned in the BIM 3D model using the field marks at the joints of the stones with their positions visible on the orthoimages.
The main difficulties found in these operations were caused by (i) the exterior wall and column surfaces pertaining to different plans and (ii) Revit software not allowing the simultaneous cladding of several coatings as different layers. To overcome these obstacles, a two-step procedure was conceived. First, the overlaid images and maps were cut out with the dimensions of the surfaces of those elements to which they were to be applied, to allow their insertion in the BIM model as coatings. Secondly, a virtual wall was generated for receiving each image or map as a coating, with all these virtual walls in the same position, and corresponding to the unique real wall. This solution allows the visualization of all images and classification maps in a unique BIM model, enabling their comparison and analysis ( Figure 15).
To include the GPR radargrams in the model with their actual direction, position and dimensions, several possible solutions were tested. The chosen one was to create for each radargram a fictitious wall-object belonging to the Revit wall family. Each of these Revit wall objects was then cladded with a radargram. This allows the visualization of the radargrams by simply hiding the façade wall ( Figure 16a). As an example, Figure 16b shows the 2300 MHz radargram corresponding to Zone 2.1 positioned in the BIM model. Since this remaining information was georeferenced in an arbitrary coordinate system, all the information was co-registered inside the model using the dimensions defined by the point clouds or coordinates of common spatial elements (edges and corners), which are easily identifiable in the parametric model.
The strategy devised to integrate all the information into the model with their precise location and orientation was twofold: (i) the orthophoto images, the pathology classification maps and the IRT images, which are exterior information, were overlaid upon the model as wall and column claddings in a 3D plane, with the dimensions defined by the point clouds, while (ii) the GPR radargrams were positioned in the BIM 3D model using the field marks at the joints of the stones with their positions visible on the orthoimages.
The main difficulties found in these operations were caused by (i) the exterior wall and column surfaces pertaining to different plans and (ii) Revit software not allowing the simultaneous cladding of several coatings as different layers. To overcome these obstacles, a two-step procedure was conceived. First, the overlaid images and maps were cut out with the dimensions of the surfaces of those elements to which they were to be applied, to allow their insertion in the BIM model as coatings. Secondly, a virtual wall was generated for receiving each image or map as a coating, with all these virtual walls in the same position, and corresponding to the unique real wall. This solution allows the visualization of all images and classification maps in a unique BIM model, enabling their comparison and analysis ( Figure 15).
To include the GPR radargrams in the model with their actual direction, position and dimensions, several possible solutions were tested. The chosen one was to create for each radargram a fictitious wall-object belonging to the Revit wall family. Each of these Revit wall objects was then cladded with a radargram. This allows the visualization of the radargrams by simply hiding the façade wall ( Figure 16a). As an example, Figure 16b shows the 2300 MHz radargram corresponding to Zone 2.1 positioned in the BIM model.  Another challenge was the inclusion of the metadata information associated with each radargram, orthophoto images, IRT images and the pathology classification maps. For example, for each radargram a fictitious wall-object belonging to the Autodesk Revit wall family was defined. But there is some more relevant information, the so-called metadata, such as how was that radargram was produced, what equipment was employed, what were the antennas' characteristics, etc. In fact, this metadata is required to better document each radargram and exploit its potential and integration in the main model, and also for correlation with future information. When a "BIM Family" object is added to a project 3D parametric model, the latter can be enriched with more information contained within the inserted object. This information can, therefore, be interrogated, edited, and updated later on. Thus, for each new BIM family created within the "Identity" parameter type, the corresponding information was associated, such as a brief description of the data, the survey date, the equipment used, the person responsible for producing the information, or a URL link with supplementary technical information. In addition, files in PDF format were also added with technical details and the attribute "Comment" filled with the main characteristics or analyses of each object to allow their access directly through the properties panel ( Figure 17). Another challenge was the inclusion of the metadata information associated with each radargram, orthophoto images, IRT images and the pathology classification maps. For example, for each radargram a fictitious wall-object belonging to the Autodesk Revit wall family was defined. But there is some more relevant information, the so-called metadata, such as how was that radargram was produced, what equipment was employed, what were the antennas' characteristics, etc. In fact, this metadata is required to better document each radargram and exploit its potential and integration in the main model, and also for correlation with future information. When a "BIM Family" object is added to a project 3D parametric model, the latter can be enriched with more information contained within the inserted object. This information can, therefore, be interrogated, edited, and updated later on. Thus, for each new BIM family created within the "Identity" parameter type, the corresponding information was associated, such as a brief description of the data, the survey date, the equipment used, the person responsible for producing the information, or a URL link with supplementary technical information. In addition, files in PDF format were also added with technical details and the attribute "Comment" filled with the main characteristics or analyses of each object to allow their access directly through the properties panel ( Figure 17). In brief, the Autodesk Revit platform used proved capable of easily receiving all the information, i.e., the images obtained from the surveys, the maps with the classification of anomalies and some corresponding metadata information, allowing its integration and visualization. Using a collaborative platform, like Autodesk360 (San Rafael, CA, USA), it is possible to share and visualize the model, as well as to access the associated information or link handover information from a web browser. In addition, the created model and objects can also be exported as IFC (Industry Foundation Classes) files. Nowadays, IFC is widely recognized as the leading interoperability standard and Autodesk supports buildingSMART, the organization that develops and maintains the IFC standard for both buildings and civil infrastructure allowing an opening collaboration between professionals.
With respect to the integration of the different types of captured data, the radargrams were revealed to be the most challenging but, as described above, the creation of fictitious walls perpendicular to the surveyed façade (and which may be imagined as drawers in this façade), In brief, the Autodesk Revit platform used proved capable of easily receiving all the information, i.e., the images obtained from the surveys, the maps with the classification of anomalies and some corresponding metadata information, allowing its integration and visualization. Using a collaborative platform, like Autodesk360 (San Rafael, CA, USA), it is possible to share and visualize the model, as well as to access the associated information or link handover information from a web browser. In addition, the created model and objects can also be exported as IFC (Industry Foundation Classes) files. Nowadays, IFC is widely recognized as the leading interoperability standard and Autodesk supports buildingSMART, the organization that develops and maintains the IFC standard for both buildings and civil infrastructure allowing an opening collaboration between professionals.
With respect to the integration of the different types of captured data, the radargrams were revealed to be the most challenging but, as described above, the creation of fictitious walls perpendicular to the surveyed façade (and which may be imagined as drawers in this façade), overcame all the integration difficulties and turned out to be the method that best suited these through-the-thickness records.
Conversely, the BIM approach has a major drawback since "historical" buildings are often composed of "non-standard" elements. Thus, despite the existence of several libraries of building elements in the Revit platform, historical buildings require the creation of new architectural elements, namely "new component families", which is an extremely time-consuming task. Figure 18, respecting the Founder's Chapel façade, illustrates a possible analysis of the combined representation of the OBIA classification, IRT and GPR imaging, showing the potential of the integration of data captured with different sensors (different ranges of penetration) into a BIM model (spatially overlapped). The areas highlighted with red circles can be interpreted as "anomalies" (different distribution of temperature, different time of reflection, etc., when compared with the surroundings). By analyzing the different information for the areas P5 and P6 of section t3 (see Figure 10), it may be concluded that zones with stones introduced during recent restoration works might show a lower temperature (IRT) and lower thickness of the blocks (GPR) and a "no pathology" status in the automatic classification. Also, regarding the central wall in section t1, the area having the lowest temperature (highlighted with a red circle) matches spatially the cavity detected with the GPR. Thus, the wall is less thick at this point, which corroborates that smaller thicknesses were associated with lower temperatures. overcame all the integration difficulties and turned out to be the method that best suited these through-the-thickness records. Conversely, the BIM approach has a major drawback since "historical" buildings are often composed of "non-standard" elements. Thus, despite the existence of several libraries of building elements in the Revit platform, historical buildings require the creation of new architectural elements, namely "new component families", which is an extremely time-consuming task. Figure 18, respecting the Founder's Chapel façade, illustrates a possible analysis of the combined representation of the OBIA classification, IRT and GPR imaging, showing the potential of the integration of data captured with different sensors (different ranges of penetration) into a BIM model (spatially overlapped). The areas highlighted with red circles can be interpreted as "anomalies" (different distribution of temperature, different time of reflection, etc., when compared with the surroundings). By analyzing the different information for the areas P5 and P6 of section t3 (see Figure 10), it may be concluded that zones with stones introduced during recent restoration works might show a lower temperature (IRT) and lower thickness of the blocks (GPR) and a "no pathology" status in the automatic classification. Also, regarding the central wall in section t1, the area having the lowest temperature (highlighted with a red circle) matches spatially the cavity detected with the GPR. Thus, the wall is less thick at this point, which corroborates that smaller thicknesses were associated with lower temperatures.  The proposed approach allows the integration of all the information obtained with several non-destructive techniques in a unique platform, proving the benefits in terms of interpretation and analysis and its usefulness in supporting and preserving built heritage data and documentation.

Conclusions
This work presents the results of a multidisciplinary survey for the inspection and characterization of some parts of the main wall façade of the Monastery da Batalha, a UNESCO World Heritage Site, using several non-destructive technologies: UAS, TLS, GPR and IRT. All the results produced were eventually integrated into a unique BIM model, providing a useful basis to permit their integrated analysis and interpretation. The most interesting findings from a diagnostic point of view were: • The super-high resolution orthophoto obtained by the cost-effective UAS technology allowed the automatic identification of several pathologies and the area that each pathology occupies on the façade.

•
The terrestrial laser scanning proved to be capable of generating a well-documented 3D model, as well as 2D plans, and the evaluation of the real condition of the structure, which plays an important role for heritage documentation purposes. It also permitted the tilt of the columns of the Founder's Chapel to be measured, a crucial information for assessing its stability.

•
The GPR allowed determining the depth of the most superficial layers (of blocks) of the structure, detection of possible cracks, voids, variations of the depth or type of material of the blocks, etc. Moreover, the use of different frequency antennas (combining different resolutions and different depths of penetration) in the same zone sometimes led to additional information.

•
The IRT information allowed detection of the presence of moisture, areas with patina and previous restoration works.
The proposed methodology intends to be a contribution to the application of the BIM approach to cultural heritage, aiming at the future use of the developed BIM model in different activities such as facility management, support in the restoration or rehabilitation processes and further research. Even though the development of a BIM model is a time-consuming task, it offers the ideal platform for the work and collaboration of multidisciplinary teams-however, in order to be truly useful, this model requires incessant updating and enrichment through additional information and data. The concentration of all the relevant information in a unique platform permitted by such a BIM model will strongly reduce the time required to access that information. Moreover, its visualization options, assisting in the interpretation of the requirements and results, the possibility of connectivity to other data systems and software, as well as the possibility of integration of simultaneous or non-simultaneous activities, also contribute to its added value.
In the study case presented, the integration in the BIM model of all the information obtained (the orthophoto images, radargrams, the maps with the classification of anomalies, thermographic images) was quite satisfactory. This integration and the visualization possibilities it offers will make this information easily available to support built heritage documentation and to assist in its preservation. The Revit platform proved to be capable of easily receiving all these additional non-standard types of information. The creation of the fictitious parallel walls perpendicular to the plane of the façade (the so-called drawers in the wall), turned out to be the method that best suits these through-the-thickness records and ultimately overcame all the integration difficulties. Furthermore, the set of collected images and the developed BIM model offer the possibility of creating a virtual reality (VR) environment which, thanks to hyper-realistic visual effects, can provide an accurate overall picture to every professional in the project team (architects, civil engineers, surveyors, mechanical engineers, contractors, conservators, etc.). Combining BIM and VR, the current state of the monument, as well as possible reconstruction or restoration alternatives, can be dynamically visualized, allowing end-users to interact with the virtual space and its components. In addition, a VR tour can be applied to a BIM model in order to check, from a facility management perspective, the maintenance schedule (7D/BIM model), or from a project team view, the reconstructions review (4D/BIM model) supporting responsible decision making.
Finally, it is worth emphasizing that many questions remain in some fields covered by the present work. These start with how the information determined by each individual technology can be integrated and compared. There is a need to investigate ways to integrate the results obtained with each individual technique based on the spatial co-registration of technologies without placing marks/targets on the constructions, particularly in monuments, due to the visual impact and possible degradation of the surface. The integration of all the information generated along the monument's life is another huge challenge, which will require the definition of clear rules for modeling and recording information, so that it can be looked up and used by other stakeholders later on.