Surveying the roofs of Rome

Journal of Cultural Heritage 13 (2012) 304–313

Available online at

www.sciencedirect.com

Original article

Surveying the roofs of Rome Lorenza Fiumi Institute for Atmospheric Pollution, IIA - CNR, c/o Consorzio per l’Università di Pomezia,Via Pontina Km. 31,400 - 00040 Pomezia, Rome, Italy

a r t i c l e

i n f o

Article history: Received 9 September 2011 Accepted 15 December 2011 Available online 11 January 2012 Keywords: Remote sensing Hyperspectral data Urban areas Roofs Materials and construction techniques

a b s t r a c t This study is aimed at investigating a portion of the city of Rome by means of remotely sensed Multispectral Infrared And Visible Imaging Spectrometer (MIVIS) data. A particular attention was devoted to building roofs described not only as the last defining touch given to a building, the aesthetic conclusion to a whole construction process, but also as expression and sign of a society’s level of civilisation, culture and technical skill. Through the classification of objects and materials, we propose to combine history and science as different ways of interpreting a city, in this case Rome, and to implement an image processing technique as an effective tool in urban planning. © 2012 Elsevier Masson SAS. All rights reserved.

1. Research aims The main research aims of the work are: • surveying the roofs of Rome through the processing of remotely sensed Multispectral Infrared And Visible Imaging Spectrometer (MIVIS) data; • combining history and science as different ways of interpreting a city, in this case Rome; • analysing the distribution of objects, materials and construction techniques in space and time, with special reference made to the continued use of a few materials in different historical eras; • implementing an image processing technique as an effective tool in urban planning actions for the promotion of environmental sustainability – to be used for the time being in view of the future development and introduction of new technologies and sensors. 2. Introduction The term “roof” properly indicates the covering on top of a building, the structure forming the upper waterproof finish, serving to protect against rain, snow, sunlight, wind, and extremes of temperature and which can be constructed in a wide variety of forms (flat, pitched, vaulted, domed, or in combinations, depending upon technical, economic, or aesthetic considerations) [1]. Since the most remote ancient times, this term was referred not only to the need for protecting people and their possessions against bad weather, but also to the concept of “shield”, “shelter”,

E-mail address: fi[email protected] 1296-2074/$ – see front matter © 2012 Elsevier Masson SAS. All rights reserved. doi:10.1016/j.culher.2011.12.003

“refuge”, designed to express something more than a technological unit. This is a metaphor of dwelling, which evokes and designates the home itself, becoming an integral part of the building envelope [1]. In fact, the Romans generally referred the term “tectum” to their homes roofed with curved clay tiles (imbrex and tegula). The Italian historian Giorgio Vasari defined it as an essential part of the overall image of a building, the harmonious and calibrated summation of a style, as intrinsic to it as the basements, the window and the fac¸ade [2]. Another key to understanding this feature is provided by Guido Nardi. He argues that roofing styles are one of the most universal signs adopted by human beings to define how they belong to a particular place [3]. In this connection, he cites a passage from an essay by Oliver Marc, [4], in particular if a house in its totality can be understood as an image of the Self, the roof represents the spirit in its aspiration to unity or its fidelity to origins. In the Mediterranean countries tiles and bricks represent one of the oldest roofing materials used for building roofs. Their origins are deeply rooted in the need to waterproof flat roofs, commonly used in the ancient Egypt and all over the Middle East. During the Renaissance, the great aristocratic families of Rome used to build opulent dwellings. According to Vasari, their characteristics with reference to roofs (materials and methods used) drew on the wealth and power of the Italian courts, the technicalscientific power of humankind over nature, roofs as the worthy conclusion to and correct proportioning of buildings [2]. They are emphasised by making them jut out beyond the supporting walls. From the early 16th century into the Baroque period, a hidden style of roofing became widely used, and balustrades were often mounted above cornices that crowned the entire building [5]. In Europe, the city’s ever-growing population and, as a consequence, the new dwelling requirements boosted architects to plan,

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313

starting from the mid-19th century, a particular kind of roof called mansard. This word is derived from a French architect, F. Mansart. It is a roof having two slopes on all sides with the lower slope steeper than the upper one [1]. At the beginning of the 20th century, with the Rationalism, the links to history gradually weakened, and the flat roof appeared alongside, and contrasted, with the traditional pitched roof. Le Corbusier is among the most important exponents of this new architecture in a rationalist and functionalist vein. Nowadays, we can choose from a wide variety of “roofing systems” according to the geometrical morphology of land surface, structure construction methods and materials used. Nevertheless, in addition to the new technologies that have led to the introduction on the market of new roof coverings made of light, opaque or transparent materials, offering sometimes impressive decorative solutions, such as corrugated iron roofing sheets, fibre-cement sheets, polycarbonate or textile fibres, the use of a few materials such as tiles and bricks has persisted. However, the increasing attention to environmental sustainability issues has led to use roof coverings made of natural or environmentally friendly materials. 3. Urban remote sensing The aforementioned introduction is a necessary prelude to a more detailed inspection of the roofs of a segment of Rome, expression and sign of a society’s level of civilisation, culture and technical skill. This is as true today as ever. Today, in Europe nearly 80% of the population lives in urban areas and, at world level, this percentage is equal to 50% [6–9]. In urban areas, remote sensing is still scarcely applied, however, prospectively, its potential of application is particularly important and potentially permits an analysis never carried out so far at such an operative level [10–12]. The images taken by MIVIS over the city of Rome allowed us to observe the city from above, to discover forgotten corners of the urban landscape, where historical squares, domes of churches, and in particular building roofs, can be easily identified on the basis of their geometry, colours and materials. A simple comparison between aerial photographs taken at a few decades’ distance suffices to demonstrate how the use of the soil has changed over the years, to assess the impact of such change on the territory and the surrounding environment and, therefore, to forecast the future growth dynamics, with the further opportunity to devise sustainable development models based on a fair trade-off between urban expansion and environment conservation [11,13,14]. Nevertheless, it is not always possible to obtain a detailed analysis to accurately identify objects and materials within urban areas because of spectral resolution limits [15–18]. Indeed, with the technological advancement of satellite systems, recent satellite sensors now produce exceptionally highresolution satellite images in a highly detailed scale (e.g., a pixel with a ground definition of a few centimetres). On the contrary, the spectral resolution, meaning the possibility to extend the survey simultaneously on different portions of the electromagnetic spectrum to improve the recognition of surfaces under study, still remains unresolved [19,20]. The spectral resolution commonly refers to the number of bands that comprises the range of spectral sensitivity of the system. This

305

expresses the ability of a system to distinguish two adjacent wavelengths to best separate spectral characteristics (recognition) of surfaces, overcoming those situations where their behaviour is very similar and ambiguous and, therefore, difficult to be detected. For this reason, hyperspectral remote sensing is a major expanding sector [16,18–21]. To this aim, the Daedalus AA5000 MIVIS instrument, acquired by the Italian National Research Council (CNR) within the framework of its LARA (Airborne Laboratory for Environmental Research) project, has been intensively operative [21]. A number of MIVIS deployments have been carried out in Italy and Europe in cooperation with national and international institutions on a variety of sites, including active volcanoes, coastlines, lagoons and oceans, vegetated and cultivated areas, oil polluted surfaces, waste discharges, and archaeological sites [9]. MIVIS is a modular scanning system constituted by 102 spectral channels that use independent optical sensors simultaneously sampled and recorded within the interval comprised between 0.433 and 12.70 ␮m (Table 1). This instrument, with four spectrometers designed to collect radiation from the Earth’s surface in the Visible (20 channels), Near-IR (eight channels), SWIR (64 channels), and Thermal-IR (10 channels), represents a second generation imaging spectrometer developed for its use in Environmental Remote Sensing studies across a broad spectrum of scientific disciplines. In addition to a very high spectral resolution, MIVIS also provides a high spatial resolution, with a pixel of 3 m × 3 m. This allows a detailed analysis when urban objects are to be identified and, in particular, when covering materials such as tiles and bricks, marble materials, asphalt, lead, copper, asbestos-cement, vegetated areas, bare soils are involved. The assessment of the potential of thermal channels (8.2–8.7 ␮m) allowed us to deduce considerations on thermal performances related to objects and materials. The study carried out by Fiumi and Rossi [22] over the city of Rome has given a strong emphasis on the potentiality of MIVIS data for analysing the degree of soil permeability in the urban setting. In this respect, the results obtained in the classification of hyperspectral data have shown the great potentialities of these tools applied to urban areas, complex situations with a high degree of fragmentation [9,11,12,15]. However, some key issues, such as the economic ones that could limit their use on a large scale, still remain unresolved. This has led the civil community to use less expensive technologies, which are very attractive also for systematic monitoring of the territory such as UAVs (Unmanned Aerial Vehicles). To cite only a few of the advantageous features of UAVs [23], these include:

• their flight performance: UAVs can operate in a wide range of operational altitudes (from 100 m to over 30,000 m) and have an elevated range of endurance (1–48 hours); • their adaptability to various typologies of missions (e.g., in remote areas); • their inexpensiveness.

However, several prerequisites must be satisfied to render UAV technology a viable, cost-effective and regulated alternative to existing resources. Major civil barriers include: the high costs of

Table 1 Multispectral Infrared And Visible Imaging Spectrometer (MIVIS) sensor technical details and spectrometer characteristics. Spectrometer

Spectral regions covered by the sensor

I II III IV

VIS (Visible) NIR (Near infrared) SWIR (Short-wave infrared) TIR (Thermal infrared)

Range (␮m) 0.43–0.83 1.15–1.55 2.0–2.5 8.2–12.7

Channels

Bandwidth (mm)

20 8 64 10

20 50 8 450

306

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313

the assurances for civil missions, lack of safe communication frequencies and regulatory issues [23–29]. The classification of remotely sensed MIVIS images allowed us to identify different spectral classes referred to materials used for building roofs and to deduce considerations on how the use of building techniques and materials has evolved and distributed in space and time, with a special attention to building roofs. Or the architectural composition of a building volume. It is necessary to define exactly this aspect because the present paper is mainly focused on this topic. 4. Study area The site taken into consideration for this study belongs to the territory of the city of Rome. The test area covers a north–south transect of 2.2 km by 26.3 km along the Tiber river. The airborne MIVIS imagery was acquired from an altitude of 1500 m. An overall view allows us to notice that there is a clear morphological distinction characterizing the areas built in different periods and the main roads (Fig. 1), even though it is represented in the range of small scales. In the middle of the scene, between the east and west loops of the Tiber river, it can be identified the historic centre of Rome: Trastevere and Vaticano. On the left bank of the Tiber river, behind Castel S. Angelo, are portions of Prati and Mazzini quarters. It follows Flaminio quarter, which is adjacent to Villa Glori Park and Tor di Quinto district. Finally, in the upper section of the MIVIS image, where the city becomes less densely built-up the further out from the historic centre, prevail the main roads connecting the city with the northern part of the capital. The city’s urban area is delimited by the ring-road called the Grande Raccordo Anulare, which circles the city centre at a distance of about 68.2 km and crosses the Tiber river and Via Flaminia, the historical consular road. Observing the study area, (starting) from the centre continuing southwards are two quarters: Monteverde and Garbatella. On the left bank of the Tiber river is Magliana district with its urban settlements; on the opposite bank of the river is the EUR quarter. The airborne MIVIS imagery used for the present study ended where the Roman campagna prevails over buildings. Towards the outskirts, the growing city gradually engulfs the natural countryside to which correspond light and bright colours correlated to natural environments, vegetated surfaces or to arable soils. On the contrary, colours darker as we move towards the downtown core where the city structure is more compact. 5. Materials and methods For this study, the MIVIS images were taken over the city of Rome on 19 June 2004. Herein are presented the aerial survey and the MIVIS images characteristics (Table 2). The MIVIS data were calibrated according to the procedures described by Bianchi et al. [30]. Not having reflectance measurements to the ground and information on the characterisation of the column of air between the sensor and the ground at our disposal, the calibration method known as IARR (international average relative reflectance), described in [30], was used. It consists in dividing the radiance spectrum of each pixel of the flight line by the average spectrum of the whole scene. This procedure is a variance of the so-called

Fig. 1. The study area taken with a synthesis in RGB by the Multispectral Infrared And Visible Imaging Spectrometer (MIVIS) sensor.

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313

307

Table 2 Aerial survey and Multispectral Infrared And Visible Imaging Spectrometer (MIVIS) images characteristics. Flight altitude (meters)

Resolution m/pxl

Starting local time

Resize n/pxl

Km2

1500

3×3m

11:35

755 × 8780 = 6.628.900

59.66

criterion known as “flat field calibration”, which approximately removes solar irradiance, atmospheric absorption, scattering effects and any other residual noise from the instrument [30,31]. The investigated area was assessed from a qualitative point of view by the visual analysis of each separate channel, according to specifications provided by Bianchi et al. The data, radiometrically corrected, were classified using the Spectral Angle Mapper (SAM) approach. It is an automated method for comparing image spectra to individual spectra or to a spectral library [32]. The algorithm determines the similarity between two spectra by calculating the “spectral angle” between them, treating them as vectors in n-D space, where n is the number of bands [15,33,34]. The advantage implied by the use of this algorithm is that the angle between the vectors does not change as the scene illumination changes. This characteristic is such that the spectral signatures belonging to the same material – but illuminated differently due to surface variability – are treated similarly by the SAM classification, as happens with the varying pitches of sloping roofs [32]. The SAM algorithm implemented in ENVI software [34] takes as input a number of training classes (training areas), or reference spectra from ASCII files, specific ROIs (Regions Of Interest), or spectral libraries. As far as this study is concerned, the input spectra were extracted from ROIs accurately identified in the MIVIS image. Inside each ROI, areas having different morphological characteristics were selected: flat surfaces, gently or steeply sloping areas, to best represent the heterogeneity of the area taken into consideration. In this phase of the method, 12 ROIs corresponding to other surfaces were identified. A brief description follows. 5.1. Tiles and bricks

capable of withstanding atmospheric corrosion. It widely reflects radiant energy in the whole spectral field, from ultra-violet to infrared [35]. 5.5. Pozzolan surfaces These are predominantly cinerary surfaces made of materials of volcanic origin, incoherent facies, of the Neapolitan Yellow Tuff. It is often used to make football and five-a-side football pitches. 5.6. Synthetic surfaces They are surfaces made of acrylic resins, polyester fibres and PVC especially used for sports ground flooring (tennis courts, soccer grounds, lawn bowling grounds, etc.) or for temporary structures particularly suitable for covering outdoor facilities during the winter months. 5.7. Roads They are surfaces covered by bituminous material consisting of a mixture of hydrocarbons of natural or pyrogenic origin, which has the linking function of joining inert elements, in order to give cohesion and stability to paved surfaces [35]. 5.8. Railway station areas Surfaces, which are covered with iron, rail tracks that guide a train. Tracks consist of two parallel steel rails, which are laid upon sleepers (or cross ties) that are embedded in ballast to form the railroad track.

They are made of a natural material, such as clay, with the addition of colouring additives. They represent one of the oldest roofing materials that are currently used mainly for the roofs of civil buildings [35].

5.9. Treed surfaces

5.2. Travertine and grits

5.10. Lawns

These are formations mainly of deciduous and evergreen species of trees (plane-trees, pines and ilexes).

Travertine is a sedimentary stone made essentially of calcite, deposited by calcareous waters. It has been used since the ancient times and its colour is whitish, slightly yellow or reddish. Its characteristics make it ideal for paving external surfaces and roofing buildings [35].

5.11. Bare soils

5.3. Bituminous surfaces

5.12. Water bodies

They are covering waterproofing membranes that are produced in sheets whose bitumen-polymer composition makes them waterproof and difficult to alter. Today it is one of the systems used for roofing industrial buildings given its cheapness, quickness and easiness in using it [35].

This means Tiber river, but it is also referred to small artificial lakes and surfaces dedicated to pools.

These are referred to surfaces with herbaceous vegetation.

Unvegetated areas, awaiting use (for sowing or urbanisation).

6. Data analysis and discussion 6.1. Confusion matrix

5.4. Metallic surfaces (sheets) Aluminium, and particularly alloys, differ from other metallic materials which have been used since the ancient times because they have only recently been introduced as building materials for roofing industrial buildings. Aluminium is a white silver material

Fig. 2 shows the classification results of the MIVIS image achieved by the SAM method (Table 3). The classification results were then evaluated by calculating the matrix of confusion [32,36–38], designed to verify or “to test” the performances of the classifier independently from the automated

308

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313 Table 3 Percent class distribution as shown in the classified Multispectral Infrared And Visible Imaging Spectrometer (MIVIS) image. Percent class distribution (Fig. 2) Classes

%

Tiles and bricks Travertine and grits Bituminous surfaces Metallic surfaces (sheets) Pozzolan surfaces Synthetic surfaces Roads Railway station areas Treed surfaces Lawns Bare soils Water bodies Unclassified

8.49 13.67 1.10 0.05 1.25 0.45 19.23 1.14 18.53 22.26 6.01 3.50 4.32

system. The values of the matrix, expressed in percentages, are the cases of agreement between classification and reference spectra. In the observed case, the input spectra were obtained from ROIs accurately identified in the scene, integrated with the direct field verification and the visual analysis of additive synthesis in RGB (Red, Green, Blue) and with Rome Capital coloured photomaps represented in the range of scales 1:500–1:2000. The confusion matrix, corresponding to the classification of Fig. 2, is shown in Table 4. This latter shows how the test pixel set was assigned to the different classes and it provides the classification accuracy for each class [36,38]. The values of the main diagonal of the confusion matrix represent the examples of agreement between the classification and input spectra, whereas the values found off the diagonal represent pixels classified incorrectly. Moreover, two kinds of errors can be calculated: errors of omission and errors of commission. The first ones are shown in the columns of the confusion matrix. They represent pixels that belong to the ground truth class but the classification technique has failed to classify them into the proper class. The second ones are shown in the rows of the confusion matrix. They represent pixels that belong to another class and that are labelled as belonging to the class of interest [38]. Results of the classification obtained were validated by means of new field campaigns. An analysis of the above mentioned table shows that: • the total classification accuracy obtained was equal to 84.68%; • most classes were extracted with an accuracy ranging from 62.10% to 100%, except for classes that cannot be easily discriminated due to the heterogeneity of their materials, such as “Railway station areas” (with an accuracy of 22.89%). The “confusion” between bituminous surfaces (65.46%) and roads (87.58%), is due to the fact that both the former (protective covering, sheaths, etc.) and the latter are mainly made of bituminous materials. In bituminous surfaces these materials are mixed with compatible synthetic products (atactic polypropylene), while in road paving bitumen is blended with inert materials as a binding compound [35]. As a consequence, when the road surface is in good condition, asphalt rises to the surface and thus the aforementioned classes tend to become spectrally confused. However, when inert materials rise to the surface, due to the wear state of the asphalt, the two classes tend to be separated. As shown by

Fig. 2. Materials classification using SAM (Spectral Angle Mapper) method applied to Multispectral Infrared And Visible Imaging Spectrometer (MIVIS) data.

1.96 8.16 2.05 3.04 1.41 3.68 0.45 12.15 0.59 11.,31 17.27 16.13 21.8 0.06 0 0 0 0 0 0 0.06 0 0.09 0.01 0 99.78 0.3 2.54 0.42 0 2.12 0 0 0 1.86 0.14 0.45 92.17 0 2.32 0 0 0.36 0 0 0 0 0.02 10.74 86.13 0.42 0 9.53 0 0 0.13 0.43 0 0 0.39 0 62.1 26.51 0.92 0 0 0 33.08 0.63 0 0 0 43.28 22.89 0 0 0.13 0 0.58 0 0.2 8.51 1.03 0 0.14 87.58 0 0.42 0.11 1.41 0.02 0 0 0 0 0 0 100 0 0 0 0 0 0 0 8.18 0 0 0 91.75 0 0 0 0 0 0.07 0 0 4 0 4.46 91.54 0 0 0 0 0 0 0 0 0 0 0 65.46 0 0 0 34.54 0 0 0 0 0 0.22 0.85 85.28 0.07 0.04 0 0.04 10.14 0 0.04 0.04 0.07 3.2 Unclassified 0.02 Tiles and bricks 77.47 Travertine and grits 0.55 Bituminous surfaces 0.24 Metallic surfaces (sheets) 0.66 19.35 Pozzolan surfaces Synthetic surfaces 0 Roads 0.25 Railway station areas 0.24 0.02 Treed surfaces Lawns 0.02 1.17 Bare soils Water bodies 0

Bituminous surfaces Travertine and grits Tiles and bricks Classes

Table 4 Confusion matrix calculated on the basis of trial areas properly identified.

Metallic surfaces (sheets)

Pozzolan surfaces

Synthetic surfaces

Roads

Railway station areas

Treed surfaces

Lawns

Bare soils

Water bodies

Total

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313

309

data in Table 4, water bodies (99.78%), metallic surfaces (sheets) (91.54%) and lawns (86.13%) are easy to identify in the study area. Tiles and bricks (77.47%) tend to be confused with bare soils (92.17%) mainly in the nearest neighbour areas characterized by the presence of large areas of clayey soils. Similarly significant is the spectral confusion between travertine and grits and metallic surfaces (sheets), due to the lead oxidation, which tends to become similar to travertines and the other light construction stones. This particular feature can be easily identified in some domes of the city [1]. 6.2. Distribution of the classified objects and materials In the MIVIS scene, the visual analysis of the classification (Fig. 2), allowed us both to identify, with a good level of detail, the distribution of the different spectral classes and to deduce considerations herewith reported. The first strong imprint of nature is that of the Tiber river, which occupies 3.50% of pixels classified as water bodies. With its meandering loops, the river has divided the scene into two sections. It is absolutely the most defined shape, made up of a single cluster of pixel, classified with an accuracy, which is equal to 99.78%. Immediately striking is the existence in the city centre, [39], adjacent to the great loop of the river, of a densely built-up area, showing up as spectrally compact and homogeneous, with a tangled skein of narrow streets. This authentic and distinctive heart of the city is characterised by the prevalence of pitched roofs mostly covered with tiles; these account for 8.49% of the classified surfaces. It corresponds to the heart of the Augustan city, followed by the city of the medieval, papal and pre-unification periods divided into rioni: Ponte, Parione, S. Eustacchio, Pigna, Regola, Campitelli, Borgo S. Angelo, Trastevere, Ripa, Trevi, Monti, Esquilino. The history of Rome spans more than twenty centuries and, for this reason, this part of the city reflects the stratification of several epochs. To the northeast side of the city, there is the Baroque style, which found its secular expression in the form of grand palaces built of stone and roofed with tiles. Baroque palaces were built around an entrance sequence of courts, anterooms, grand staircases, and reception rooms of sequentially increasing magnificence. An example of Baroque architecture and town planning can be found in Piazza del Popolo at the apex of a triangular area known as the Tridente which gets its name from three streets – Via di Ripetta, Via del Corso and Via del Babuino – that arrow out of Piazza del Popolo like a trident’s prong. A sequence of urban expansions dating back to pre- and postunification periods can be easily identified thanks to large buildings for government ministries: Viale Trastevere (Ministry of Education), Via Arenula (Ministry of Justice), Via Flaminia (Ministry of Marine). These imposing constructions are notable for their austere tone, regularity of shape and flat roofs covered with travertine and grits (13.67%). Large buildings for government ministries are visible alongside residential areas, with straight roads flanked by houses, in Prati district behind Castel S. Angelo; industrial buildings are introduced to the Testaccio district; the Janiculum Hill is the focus of new residences for the upper and middle classes. Other sub-urban areas are located in Aventino and Monteverde quarters. The city continued to expand southwards where highly urbanized areas are concentrated within the Garbatella quarter, along Via Ostiense and outside Porta San Paolo. These extensions to the city, often of mediocre architectural quality made by an army of speculating entrepreneurs referred to as “palazzinari” (posh block builders), are recognisable from the banal regularity of their chessboard streets, lack of green areas, use of flat roofs, and the boring repetitiveness of the houses. This is the culture of the 20th century, which sees the spread of white-painted plaster walls and with flat roofs that are a sort

310

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313

Fig. 3. Historic centre. Details of the classification.

of “fifth wall”. This was a building culture, at a time of feverish expansion, that was generally careless, largely ignored roofs as a significant feature, and laid waste to every tradition, including traditional skills in construction, thanks to the ease in handling the new materials (reinforced concrete, steel) [40]. As the eye moves towards the outskirts to the south and north of the MIVIS image, one can notice that roof coverings are characterized by a considerable heterogeneity of pixels. This spatial and spectral complexity is due to the variability of sizes, shapes and materials used. Bituminous surfaces (1.10%) and metallic surfaces sheets (0.05%) appeared on the scene. On the whole the image shows great variability from pixel to pixel to which corresponds a high degree of fragmentation of shapes and materials of surfaces. Treed surfaces (18.53%) are highly concentrated in the areas of Villa Borghese Park, Villa Glori Park, Janiculum Walk (Passeggiata del Gianicolo), defining the borders of the areas between the city dating back to the post-unification period and the suburbs. In the southern district of the city, named EUR, there is a wide green area surrounded by huge marble buildings built around an artificial lake. Finally, if we take a look to the north and south of the MIVIS classification, we can notice, as shown in Fig. 2, the limit to the city’s expansion and in particular the Grande Raccordo Anulare and the main roads. As a whole, roads account for 19.23% of the surfaces present in the study area. Moreover, the classification highlights a large number of lawns (22.26%) and bare soils (6.01%). These are grounds left free of constructions, the last remnants of the hybrid environment – simultaneously agricultural and urban, city married to campagna – that was Rome. More generally, the study area is intensely built-up achieving a percentage, which is equal to 44.13% that is the sum of all the following classes: • • • • •

tiles and bricks (8.49%); travertine and grits (13.67%); bituminous surfaces (1.10%); metallic surfaces (sheets) (0.05%); synthetic surfaces (0.45%);

• roads (19.23%); • railway station areas (1.14%). Ground free of buildings accounts for 48.05% that is the sum of all the following classes: • • • • •

bare soils (6.01%) lawns (22.26%); treed surfaces (18.53%); pozzolan surfaces (1.25%); the remaining surfaces belong to water bodies (3.50%) and unclassified classes (4.32%).

This last class has pixel values far exceeding certain thresholds assigned to the classes identified by the classification. 6.3. Sample study areas The peculiarity of the central area of the scene, characterized by a high concentration of tiles and bricks class, together with a high degree of fragmentation of another area located south of the flight strip in the Magliana district, led us to carry out a more thorough study (Figs. 3 and 4). 6.3.1. Historic centre Bordered to the south by Ponte Palatino and to the north by Ponte Umberto I, this study area has a percentage of tiles and bricks which is equal to 25.41% (more than one fourth of the surfaces), concentrated mainly on the right loop of the Tiber river, in the most ancient rioni of the city: Ponte, Regola, Parione and Sant’Eustachio. Pitched brick roofs, the standard design for centuries and which here show up in a repeated geometry of pixels, are also a feature of religious buildings. The MIVIS image shows a composition unit, “the heart of Rome”, with spatially contiguous and spectrally homogeneous pixels because of pitched brick roofs. Here MIVIS image provides an overview of Roman Architecture such as domes (cupole) and bell towers (campanili), squares (piazze) and fountains (fontane) with their rich symbolic significance, whose supremely sophisticated shapes can be easily identified.

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313

311

Fig. 4. The Magliana and EUR quarters. Details of the classification.

In the middle of the scene is the Pantheon, “temple of all the gods”, originally built by Marcus Agrippa and destroyed along with other buildings in the great fire of Rome in 80 A.D. It was subsequently rebuilt during the reign of the Emperor Hadrian. Remarkably well preserved, it is constructed mainly of a pozzolanbased concrete with a brick veneer. Its dome symbolizes the celestial sphere, the huge opening or ‘oculus’ not only serves as the temple’s light source, but is symbolic of the Sun. The opening in the centre of the dome is completely open to the elements providing filtered light into the building, which traverses the internal cornice and traces the celestial equator. The dome was first built out of wood and then in concrete mixed with inert materials and classified as travertine and grits. Its circular shape, the hole in the centre, the portico with a double-pitched brick roof, are described in detail in Fig. 3. Travertine and grits class, as shown in Fig. 3, accounts for 20.09% of the surfaces. But what immediately strikes the eye are travertinepaved surfaces of opulent dwellings, such as Palazzo di Giustizia (1910), Palazzo di Montecitorio, a sixteenth-century edifice which was restored several times, whose latest restoration dates back to 1918. The MIVIS classification shows that the slightly curved fac¸ade follows the morphology of the terrain. In the lower section of the image is the Tiber Island (Insula Tiberis) which lies in the middle of Tiber river: natural ford, it was essential to the erection of permanent installations on the surrounding heights of peoples (the Etruscans and the Sabines) who inhabited along the Tiber river in the 8th century B.C. [39]. It was entirely built in travertine and the Island profile suggested to arrange of the external perimeter in shape of war ship, with the embankments equipped for the moorings and with an obelisk as the main tree. The tiles and bricks in this historic cure place belong to roof coverings of the edifices still standing today. Among these, it is worth mentioning the Fatebenefratelli Hospital (brothers, engage in good works) whose origins date back to 1548; Saint-Giovanni Calibita church with its small baroque bell tower and the medieval tower; the church of St. Bartolomeo, which was erected in the 10th century. As a whole, the MIVIS image shows how roof coverings reflect a well-defined epoch and how the building style, consolidated over

the centuries, granted the privilege to use local materials until the 19th century. The age-old pitched roof covered with tiles is soon replaced by the flat roof paved in travertine tiles. This is used for almost all public buildings in the post-unification period, and spills over into housing located in Via Crescenzio, Via Alberico II, as shown in the upper left side of Fig. 3, or in Via delle Zoccolette, Via Arenula, in the lower section of the image. In fact, architectural rationalism advocated an emphasis on sharply defined linear forms, expression of a machinist civilization (civilisation machiniste). This was a period characterized by a strong belief in technological progress and a desire to break free of history and material resources [41]. From a first careful reading of the classification and the View Statistics Files, (Fig. 3), we can notice some important environmental aspects. If we add tiles and bricks (25.41%), travertine and grits (20.09%), roads (equal to 24.8%) classes to bituminous, (1.2%) metallic (0.18%), synthetic (2.10%) and railway station (0.85%) surfaces, we reach a total percentage of 74.63%. This refers to the impermeable surfaces, which are not able to soak up rainwater directly. This problem becomes particularly serious on the left loop of the Tiber river; however, this kind of situation should expect at least the correct functioning of the sewerage collection system, but, unfortunately, this occurs rarely in the city of Rome. As it is known, an indiscriminate use of asphalt and cement brings about the diffuse waterproofing of soils increasing the negative hydrologic effects both on the drainage of the rainwater, the microclimate due to the lack of vegetation surfaces, air oxygenation and underground water recharge. Finally, the impact of an excessive waterproofing of soils on local and regional climatic conditions is so high as to generate the phenomenon known as “urban heat island” [22]. In the study area there is a small number of permeable surfaces, which, as a whole, accounts for 15.7%: lawns (7.94%), treed surfaces (5.91%), bare soils (0.93%) and pozzolan surfaces (0.92%). This last class is almost non-existent, the remaining surfaces belong to water bodies class (6.03%) (Table 5). 6.3.2. The Magliana and EUR quarters The study area is divided into two sections by the course of the Tiber river: on the left, the Magliana district, on the right, a picture of the EUR district (Fig. 4).

312

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313

Table 5 Percent class distribution referred to the historic centre. Percent class distribution (Fig. 3)

Table 6 Percent class distribution referred to the Magliana and EUR quarters. Percent class distribution (Fig. 4)

Classes

%

Classes

%

Tiles and bricks Travertine and grits Bituminous surfaces Metallic surfaces (sheets) Pozzolan surfaces Synthetic surfaces Roads Railway station areas Treed surfaces Lawns Bare soils Water bodies Unclassified

25.41 20.09 1.20 0.18 0.92 2.10 24.8 0.85 5.91 7.94 0.93 6.03 3.64

Tiles and bricks Travertine and grits Bituminous surfaces Metallic surfaces (sheets) Pozzolan surfaces Synthetic surfaces Roads Railway station areas Treed surfaces Lawns Bare soils Water bodies Unclassified

1.45 12.48 6.3 8.7 0.26 0.24 26.39 1.13 22.45 13.08 2.18 3.94 1.4

The Magliana quarter developed spontaneously in the twenty years from 1965 to 1985, a period when unauthorised buildings and wildcat urbanisation were rampant. Today this area is characterized by a situation of messy urban growth. Owing to the high spatial and spectral morphological complexity, as shown in Fig. 4, it was not possible to easily detect surfaces. The spatial complexity is due to the variability of sizes and covering shapes, whereas the spectral complexity is due to a wide range of construction materials, which are very similar to each other. Asbestos-cement is an example. This material is composed of 90% cement and water and 10% of asbestos fibres. As it represents a serious danger for human health, Italy adopted Law No 257/92 laying down rules concerning the cessation of the use of asbestos. As shown by recent investigations carried out by the author of this paper [22], the Magliana area is particularly known for the substantial presence of asbestos-cement roofing located along Via Idrovore della Magliana (Fig. 4, at the left side of the image): over a surface of 519 hectares, 2,891 pixels were recognised as asbestos-cement, corresponding to 46,256 m2 , which make up 4.2% of the building roofs, mainly built for artisan use, depots and warehouses. The MIVIS classification shows that tiles and bricks class reduces drastically, representing only 1.45% of the surfaces except for roof coverings of luxury houses built in the sixties within the residential quarter of EUR, at the right side of the image. Although the background situation is decisively varied, significantly enough the use of tiles and bricks for quality buildings has persisted, in spite of a lower percentage incidence. This tradition is directly attributable to the use of locally available materials, which has down the centuries characterised construction methods and still constitutes a prerogative of today’s buildings. Travertine and grits surfaces show a decrease of 12.48%. On the contrary, roads surfaces increase dramatically, with percentages higher than 26.39%, and tend to be confused with bituminous surfaces (6.3%) belonging to industrial building roofs (Table 6). Besides the lack of homogeneity in the types and sizes of the pixels, there is a situation of messy urban growth with high percentages of waterproofed surfaces that are impossible or difficult to reverse. Among the negative effects of the cementification of soils, it should be pointed out that impermeable surfaces, particularly in the outskirts, mean that the land is unsuitable for other uses such as agriculture or forestry, and loses some of the functions specific to it; the natural corridors for communication and transport are, for example, interrupted, so that the original habitats and natural or semi-natural biotopes are compromised. The EUR quarter, in the extreme south of the city, is one of the most recent areas of Rome. Built for the Esposizione Universale

di Roma in 1942, this strange quarter of wide boulevards and huge linear buildings owes much of its look to the vision of the razionalisti (rationalists). Contrary to Magliana area, as already mentioned, the EUR quarter is characterized by a compact vegetation, treed surfaces (22.45%) and clusters of pixels classified as water bodies, made up of private swimming pools surrounded by greenery. This latter class is equal to 3.94% and it refers both to Tiber river and to the artificial lake of EUR. This was designed and constructed for the 1960 Rome Olympic games together with a great number of sports facilities [42], such as Sports Palace (Palazzetto dello Sport), designed by L. Nervi and M. Piacentini. New building materials appeared on the scene: synthetic surfaces (0.24%), especially used for sports ground flooring (tennis courts, soccer grounds, volleyball court flooring, roller-skating rinks and so on) or for temporary structures (pressostatic balls) for sports facilities [43]. As shown in Fig. 2, presently they cover the most recent sports facilities built along the Tiber river. As shown in Fig. 2, the surface classified as metallic surfaces (sheets) of the Auditorium Parco della Musica is worthy of note. It is a large public music complex on the north side of Rome, located between Villa Glori, Parioli hill and Villaggio Olimpico quarter, planned by Renzo Piano, Italy’s foremost architect, and inaugurated in 2002. This is an example of how “roofing systems” have undergone, over the last few years, strong transformations involving the exterior aspect. More recently, in line with the new environmental protection policies that make use of solar energy, roofs are becoming photovoltaic, made of panels with monocrystalline and polycrystalline silicon cells [44]. Scientists today reported the development of bio-based “intelligent roof coatings”. A team of recent Massachusetts Institute of Technology (MIT) graduates has developed colour-changing roof tiles that absorb heat in winter and reflect it in summer. Using phase-change polymer gel-filled tiles, the team is able to control the light energy transmission properties of the roofing material [44]. Concluding, the increasing attention to environmental sustainability issues is encouraging a widespread use of roofings made of natural materials or, more generally, eco-friendly materials. Proof of this are, for instance, green roofs [45]. In this perspective, surfaces showing up as travertine and grits (13.67%) add up to approx. 8 km2 . These might help upgrade the built-up environment if they were converted to grass and ponds. One can envisage more experimental measures aimed at safeguarding and enhancing the landscape complexity in order to compensate for and mitigate the visual impact of construction [45]. This could be an opportunity for the re-appropriation and rejuvenation of particular areas through environmental policies at the service of human beings.

L. Fiumi / Journal of Cultural Heritage 13 (2012) 304–313

7. Conclusions The MIVIS data acquired over the city of Rome allowed us to demonstrate, in an original and innovative manner, how it was possible to characterize and quantify the roofing surfaces and to deduce considerations on the aerial view of the city, and in particular, on how the use of building techniques and materials evolved in space and time. Contrasting with the old city, showing up as homogeneous and continuous in terms of roofing materials and styles, is a more random city. The pitch of brick-tile roofs, which has down the centuries characterised construction methods and is still used today, tends to merge into the landscape–which now resembles an obstacle course. Finally, one cannot help wondering why designers and builders devoted particular attention to roofs, to the skilful use of materials in their construction and to creating harmonious lines, where these were not visible to anyone. Nantas Salvataggio [46] suggests that we cannot exclude the idea that they did it to be noticed above the clouds and so gain some small fame in paradise. In conclusion it is worth observing that remote sensing activities are very rapidly evolving. Hence, what can now exclusively be detected with hyperspectral sensors – such as the airborne MIVIS sensor – will in a few years also be possible with more commonly available remote sensing equipment. Improvements in spectral resolution and the use of less costly shooting platforms such as UAVs [25–29] will no doubt open up new, significant prospects for systematic land control and monitoring with a consequent containment of costs. References [1] M.P. Belski, Particolari di progettazione, Centro Stampa Facoltà di Architettura, Politecnico di Milano, AA 1997-1998, no 181 Milano, 1987. [2] G. Vasari, Le vite de’ più eccellenti pittori, scultori et architettori italiani, Firenze, 1568. [3] G. Nardi, La copertura: archetipi costruttivi e cultura materiale, Costruire no 59 Settembre (1997), pp. 318–23. [4] O. Mark, Psychanalyse de la maison, Seuil, Paris, 1972. [5] G.P. Bellori, Le vite de’ Pittori, Scultori et Architetti moderni, Roma 1672, Edizione a cura di E. Borea, Torino, 1976. [6] EEA, European Environment Agency, technical report: core-set of indicators, Copenaghen. 2003. [7] EEA, European Environment Agency, L’ambiente in Europa: la terza valutazione, Copenaghen 2003. [8] EEA, European Environment Agency, Briefing 2006-04, La sovraccrescita urbana in Europa, 2003. ISSN:18302319. [9] R.L. Powell, D.A. Roberts, P.E. Dennison, L.L. Hess, Sub-pixel mapping of urban land cover using multiple endmember spectral mixture analysis: Manaus, Brazil, Remote Sensing Environ. 106 (2) (2007) 253–267. [10] A. Berk, L.S. Bernstein, G.P. Anderson, P.K. Acharya, D.C. Robertson, J.H. Chetwynd, Cloud and multiple scattering upgrades with application to AVIRIS, Remote Sensing Environ. 65 (1998) 367–375. [11] B.C. Forster, An examination of some problems and solutions in monitoring urban areas from satellite platforms, Int. J. Remote Sensing 6 (1985) 139–151. [12] C. Small, A global analysis of urban reflectance, Int. J. Remote Sensing 26 (2005) 661–681. [13] E. Zilioli, Appunti e spunti di telerilevamento, Arte stampa Daverio, 2001, pp. 171–4, ISBN: 8886596073. [14] M.K. Ridd, Exploring a V–I–S (vegetation–impervious surface–soil) model for urban ecosystem analysis through remote sensing: comparative anatomy for cities, Int. J. Remote Sensing 16 (1995) 2165–2185. [15] C. Small, High spatial resolution spectral mixture analysis of urban reflectance, Remote Sensing Environ. 88 (2003) 170–186. [16] G.J. Hay, K.O. Niemann, D.G. Goodnough, Spatial thresholds, image-objects, and upscaling: a multiscale evaluation, Remote Sensing Environ. 62 (1997) 1–19. [17] U. Heiden, K. Segl, S. Roessner, H. Kaufmann, Determination of robust spectral features for identification of urban surface materials in hyperspectral remote sensing data, Remote Sensing Environ. 111 (2007) 537–552. [18] M. Herold, M. Gardner, D. Roberts, Spectral resolution requirements for mapping urban areas, IEEE Trans. Geosci. Remote Sensing 41 (9) (2003) 1907–1919.

313

[19] G.F. Hepner, J. Chen, Investigation of imaging spectroscopy for discriminating urban land covers and surface materials. In Proceedings of AVIRIS Earth Science And Applications Workshop, Palo Alto, CA, 27 Feb–2 Mar 2001, pp. 306–20. [20] S. Roessner, K. Segl, U. Heiden, H. Kaufmann, Automated differentiation of urban surfaces based on airborne hyperspectral imagery, IEEE Trans. Geosci. Remote Sensing 39 (7) (2001) 1525–1532. [21] R. Bianchi, R.M. Cavalli, L. Fiumi, C.M. Marino, S. Pignatti, G. Pizzaferri, 1994/1995 CNR LARA Project airborne hyperspectral campaigns, Proceedings of Eleventh Thematic Conference And Workshops Applied Geologic Remote Sensing, Las Vegas, Nevada USA, 1996, pp. 301–10. [22] L. Fiumi S. Rossi, Dalla cartografia storica al telerilevamento: la città di Roma, ISBN9 788875 572136 Edizione CNR, Roma, 2007. [23] H. Bendea, P. Boccardo, S. Dequal, F. Giulio Tonolo, D. Marenchino, M. Piras, Low-cost UAV for post-disaster assessment, The International Archives of the Photogrammetry, Remote Sensing Spat. Inf. Sci. XXXVII (2008) 1373–1379, Part B8. Beijing Commission VIII, WG VIII/2. [24] K. Segl, U. Heiden, S. Roessner, H. Kaufmann, Fusion of spectral and shape features for identification of urban surface cover types using reflective and thermal hyperspectral data, ISPRS J. Photogrammetry Remote Sensing 58 (2003) 99–112. [25] U. Heiden, S. Roessner, K. Segl, H. Kaufmann, Analysis of spectral signatures of urban surfaces for their area-wide identification using hyperspectral HyMap data, Proceedings of IEEE-ISPRS Joint Workshop On Remote Sensing And Data Fusion Over Urban Areas; Rome, Italy, November 8–9, 2001, pp. 173–7. [26] L. Mejias, S. Saripalli, P. Campoy, G.S. Sukhatme, Visual servoing of an autonomous helicopter in urban areas using feature tracking, J. Field Robot. 23 (3–4) (2006) 185–199. [27] U. Coppa, A. Guarnieri, N. Milan, A. Vettore, Sviluppo di un sistema UAV a basso costo per il monitoraggio del territorio, Atti 14a Conferenza Nazionale ASITA Brescia 9–12 novembre 2010, pp. 653–8. [28] I. Colomina, M. Blázquez, P. Molina, M.E. Parés, M. Wis, Towards a new paradigm for high-resolution low-cost photogrammetry and remote sensing, IAPRS&SIS 37 (Part B1, Beijing) (2008) 1201–1206. [29] H. Eisenbeiss, The autonomous mini helicopter: a powerful platform for mobile mapping, IAPRS&SIS 37 (Part B1, Beijing) (2008) 977–983. [30] F.A. Kruse, G.L. Raines, K. Watson, Analytical techniques for extracting geologic information from multichannel airborne spectroradiometer and airborne imaging spectrometer data, Proceedings of International Symposium On Remote Sensing Of Environment, Fourth Thematic Conference On Remote Sensing For Exploration Geology, San Francisco USA, 1985, pp. 309–24. [31] E. Ben-Dor, F.A. Kruse, A.B. Lefkoff, A. Banin, Comparison of three calibration techniques for utilization of GER 63-channel aircraft scanner data of Makhtesh Ramon Negev, Israel, PE&RS 60 (11) (1994) 1339–1354. [32] R.H. Yuhas, A.F.H. Goetz, J.W. Boardman, Discrimination among semiarid landscape endmembers using the Spectral Angle Mapper (SAM) algorithm. Summaries of the Third Annual JPL Airborne Geoscience Workshop, 1 June, Pasadena, CA, Jet Propulsion Laboratory, 1992, pp. 147–9. [33] J.W. Boardman, F.A. Kruse, Automated spectral analysis: a geological example using AVIRIS data, North Grapevine Mountains, Nevada. Proceedings of Tenth Thematic Conference On Geologic Remote Sensing, San Antonio Texas USA, Vol. I. 1994, pp. 407–18. [34] ITT Visual Information Solutions, ENVI, Environment for Visualizing Images, version 4.4, 2008, Available online at: http://www.ittvis.com/envi/. (Accessed April 4, 2011). [35] A. Bruno, Materiali da Costruzione, Edizioni medicea, Firenze, 1981. [36] A.H. Strahler, C.E. Woodcock, J.A. Smith, On the nature of models in remote sensing, Remote Sensing Environ. (1986) 121–139. [37] M. Story, R.G. Congalton, Accuracy assessment: a user’s perspective, Photogrammetric Engineering and Remote Sensing 52 (3) (1986) 397–399. [38] G.M. Foody, Status of land cover classification accuracy assessment, Remote Sensing Environ., (2002) 185–201. [39] S. Rossi, Roma Struttura Dinamica Universale di Architettura, Collana diretta da Bruno Zevi, Officine Grafiche Artistiche, Torino, 1997, ISBN 88-86498-25-X. [40] G. Corzani, L’immagine panoramica della città, Costruire 59 (1997) 358–365. [41] A. Alauria, E. Nardi, Osservazioni sul rapporto tra precipitazioni atmosferiche e tetti, Available online at: http://www.copertureinlaterizio.it/COP/ UserFiles/File/BIBLIOTECA/Articoli/Tecnologia/77 46 55.pdf. (Accessed April 4, 2011). [42] B. Zevi, Un Pantheon schiacciato di cemento armato, Cronache di architettura, Bari, 1971, pp. 73–190. [43] E. Piccardo. Pneus, gonfiabili niente di nuovo. . .. però. Available online at: http://www.archphoto.it/images/. (Accessed April 4, 2011). [44] Tegole tecnologiche, ecologiche ed intelligenti per tetti. Available online at: http://www.3sulblog.com/tegole-tecnologiche-ecologiche-ed-intelligentiper-tetti-14043.html. (Accessed April 4, 2011). [45] M. Fiori, T. Poli, Coperture a verde della Maggioli Editore, Rimini, 2008. [46] N. Salvataggio, Tetti d’Italia Geografia e storia dei coppi in cotto attraverso le immagini di Mario De Biasi, Stampa grafiche Lema, Pordenone, 1990.