ANNALS OF GEOPHYSICS, VOL. 49, N. 1, February 2006

The natural areas of Province detected by airborne remotely sensed data

Rosa Maria Cavalli (1), Lorenzo Fusilli (1), Anna Guidi (2), Simone Pascucci (1), Stefano Pignatti (1)(3), Ludovico Vannicelli Casoni (4) and Maria Vinci (2) (1) Laboratorio Aereo Ricerche Ambientali (LARA), IIA-CNR, Tor Vergata (RM), (2) Provincia di Roma - Dipartimento II, Roma, Italy (3) Istituto Metodologie Analisi Ambientale (IMAA), CNR, Tito Scalo (PZ), Italy (4) Provincia di Roma - Dipartimento I, Roma, Italy

Abstract Rome Province with its 4 million inhabitants is one of the Italians areas with the largest urban expansion, main- ly concentrated around the capital city. The uncontrolled urbanization of the past has heavily marked the land- scape, especially Rome countryside and coastline. However many zones have exceeded the anthropic pressure- without serious consequence since the sensitivity towards environmental protection has grown in recent years. Rome Province Administration has devoted special attention to the improvement and protection of its naturalistic heritage by means of a series of administrative actions, cultural initiatives and projects for environmental educa- tion. In this perspective a three-year agreement was concluded with CNR LARA focused on the study of natural vegetation by means of MIVIS (Multispectral Infrared Visible Imaging Spectrometer) remotely sensed data. This study distinguished and mapped the most important natural forests, shrub and herbaceous formations, assessed the health conditions of the arboreal vegetation, identified the areas with little water supply, and measured some environmental parameters, like temperature and surface humidity. The results achieved highlight the large botan- ical and naturalistic assortment and the complexity of the study-area.

Key words airborne hyperspectral remote sensing surface elements, is a piece of environmental in- Ð environmental monitoring Ð maximum likelihood formation and represents an effective resource vegetation mapping for management and policy purposes for a wide range of human activities. The location and rates of forest structural change and the degree to 1. Introduction which landscapes respond to human distur- bances require extensive investigations (Lam- The importance of mapping, quantifying and bin, 1998; Borak et al., 2000). Among the large monitoring changes in the physical characteris- variety of forested ecosystems the Mediter- tics of forest cover has been widely recognized ranean comprises a vegetation form and distri- as a key element in the study of global change bution with worldwide importance being of spe- (Nemani and Running, 1996). Land cover, i.e. cial interest to global change research in terms the composition and the characteristics of land of biodiversity, desertification, water resources and urbanization (Hill et al., 1995; Moreno and Oechel, 1995). The new generation high resolution optical Mailing address: Dr. Lorenzo Fusilli, Laboratorio Ae- sensors allow new applications in Earth obser- reo Ricerche Ambientali (LARA), IIA-CNR, Via del Fosso del Cavaliere 100, 00133 Tor Vergata (RM), Italy; e-mail: vation for land cover mapping by delivering [email protected] new methodologies for generating land cover

187 Rosa Maria Cavalli et al. information products with high accuracy. For 2. Study area the remote sensing community vegetation map- ping is a considerable objective for the study The area investigated is of particular interest and monitoring of ecosystems. The rapid and for both its botanic-naturalistic variety still in- cost-effective application of remote sensing to tact and its historical-archaeological heritage. map vegetation is one of the important motiva- Local authorities, in particular Rome Province tions for its utilization in land use planning to Administration, are properly improving and replace more time consuming and expensive safeguarding this area by means of administra- field surveys. In this context, an increasing tive provisions, cultural initiatives and projects number of agencies as well as private compa- for environmental education. nies with large land holdings currently use veg- The area (fig. 1), located in the E-NE por- etation maps derived from satellite and airborne tion of Rome Province (), extends from data (Congalton et al., 1993). the Valley and the lightly rolling undula- Rome Province Administration has long de- tions of the Roman countryside, towards the voted its efforts to enhancing the value and pro- sharp carbonate massif near Tivoli, and east to tecting its historical naturalistic heritage and the mountainous ridge of the Latium Apen- hence it has implemented some specific Land nines. This territory has already been safeguard- Management Plans for its Natural Reserves ed to a great extent by establishing several pro- (Nature 2000 Project of the European Commu- tected areas, some important like the Natural nity). The purpose of these Management Plans Reserves of Nomentum, Macchia di Gattaceca, is to represent a reference framework from Mt. Catillo, Mt. Soratte, and the Regional Parks which to undertake land management policy ac- of Mts. Lucretili and Mts. Simbruini. The high tions and stimulate bio-compatible activities in environmental value of this area is also proved the respect and the safeguard of the environ- by the presence of numerous Sites of European ment. They also aim to organize all geographic, Community Interest (SIC) and Special Protect- urban and environmental data in a homoge- ed Areas (ZPS) instituted according to the neous context, to carry out a careful analysis 92/43/CEE ÇHabitatÈ and 79/409/CEE ÇBirdsÈ and intervention planning. The necessity has provisions. thus arisen to make a detailed ÇMap of the From the point of view of flora, the area Main Vegetation ClassesÈ (scale 1:10000) of comprises several wood formations which are the Natural Reserves recently established, and one of the prevailing land covers. Besides the of the north-eastern area of Rome Province ter- arboreal component, the herbaceous and shrub- ritory (scale 1:25000) where a detailed homo- by formations, presenting a wide ecological and geneous cartography of vegetation has never floristic differentiation, are also well represent- been compiled. ed. According to the Latium Phytoclimate Map The considerable extension of the area to in- (Blasi, 1994), the above plant formations fall vestigate, the difficulty in approaching inacces- into the following phytoclimatic units: sible places, together with the biodiversity of Ð Temperate bioclimatic region: unit 2, vegetation, led to the use of airborne remote summits of lower carbonate massif with lower sensing as a means of investigation. The cartog- montane thermotype and upper humid om- raphy deriving from the Corine Land Cover at brotype; units 3 and 4, Apennines intramontane 1:100000 scale (1994) obtained from Landsat valleys with submontane thermotype and from satellite images (ground pixel resolution of 30 upper humid to lower hyperhumid ombrotype; m) was not able to meet the required carto- unit 6, Sabina piedmont with hilly thermotype graphic and thematic details. Rome Province and from subhumid to lower humid ombrotype. Administration therefore started a triennial Ð Temperate bioclimatic transition region: agreement with CNR-IIA/LARA to study the unit 7, Middle Tiber valley between Orte and main vegetation covers by means of MIVIS Monterotondo with thermotype from hilly to (Multispectral Infrared Visible Imaging Spec- Mesomediterranean and inferior humid om- trometer) airborne sensor. brotype.

188 The natural areas of Rome Province detected by airborne remotely sensed data

Fig. 1. Location of the study area on Rome Province with MIVIS swaths overlaid.

The whole area shows formations belonging Arid prairies are diffuse as well, being the con- to the Mediterranean environment, like ever- sequence of the prolonged exploitation of green forests with a prevailing presence of mountain pastures; they represent the most Quercus ilex, and the typical Mediterranean favourable habitat for the growth of those vege- maquis. This Mediterranean maquis which in tation species typical to arid environments on some spots (mainly in the area of Mt. Catillo) is carbonatic lithologies. Rocky environments are characterized by the peculiar chorologic eastern of particular interest since they also host numer- component with an abundant presence of Styrax ous endemisms and botanic rarities. officinalis (protected species by the Regional Law 61/74), Cercis siliquastrum, Paliurus spina-christi, Carpinus orientalis (Montelucci, 3. Data and processing methods 1972, 1984). In the submontane climatic type (between MIVIS data campaign, conducted in June 200 and 600 m a.s.l.) deciduous species mixed 1998, covered an area of about 168225 ha, cor- with Quercus pubescens, Quercus cerris and responding to about 31% of the Province’s Ostrya carpinifolia are present, at mountain lev- whole surface. It included 28 runs recorded with el (between 600 and 1600 m a.s.l.) Fagus sylvat- relative altitude of about 1700 m, NNW/SSE- ica is present with a fragmentary distribution oriented flight lines, parallel to one another, with while in mountainous areas it is thickly spread approximately 25-30% overlapped laterally (fig. (Montelucci, 1978). The articulated morphology 2). The pixel ground resolution, due to the rela- and diffuse surface karstification bring about the tive flight altitude, is about 3.4 m. presence of humid depressions covered by mes- The MIVIS is composed of 4 spectrometers ophyll prairies with different floristic types. which measure the electromagnetic Earth radia-

189 Rosa Maria Cavalli et al.

Fig. 2. Mosaic of MIVIS flight lines.

Table I. MIVIS spectral characteristics.

Spectrometer Spectral region Bands number Spectral range (nm) Band width (nm) I Visible 20 0.43-0.83 0.02 II Near infrared 8 1.15-1.55 0.05 III Medium infrared 64 2.0-2.5 0.009 IV Thermal infrared 10 8.2-12.7 0.34-0.54

tion with 102 spectral bands in the visible, near not produce any meaningful improvements in infrared, middle infrared and thermal infrared re- the classification process and they increased gions, with 2 mrad of IFoV (see table I). processing time. The dataset used was the first 28 bands (I MIVIS data have been radiometrically cali- and II spectrometer) covering the spectral range brated using references inside the sensor, and 0.43-1.55 nm because this spectral region cor- the Radiance Factor was measured on optical responds to the maximum reflectance of vege- bench; this operation converted the sensor’s tation and encompasses the major spectral fea- electromagnetic radiation, expressed in Digital tures related to the biophysical characteristics Numbers (DN), into radiance [nW/(cm2 nm sr)]. of the plants. The bands of the III and IV spec- MIVIS images were then corrected for the path trometers were not employed because they did radiance atmospheric component by means of

190 The natural areas of Rome Province detected by airborne remotely sensed data simulations performed with the MODTRAN the system acquisition geometry and the air- code (Berk et al., 1989), since such effect on a borne navigation data. The topographic varia- scene produces classification errors. tions of the countryside territory were instead The geometric correction was carried out corrected by means of the Digital Elevation through a software developed in IDL language Model (DEM) derived from satellite interfer- by LARA (Avanzi et al., 1997, 2006) which en- ometry techniques with a resolution of 30 m abled us to correct the airborne attitude plat- (courtesy of ESA-ESRIN). The residual errors form during flight for panoramic distortions, due to the not perfect reconstruction of the air- (due to yaw, pitch, roll, drift, altitude and speed plane’s trajectory and ground elevation were variations) and the morphological variations of minimized using the Ground Control Points the Earth surface. These corrections (fig. 3) (GCP) extracted from the Regional Technical were performed taking into consideration both Map (scale 1:10000). The image warping was

Fig. 3. Geometric correction: (left) uncorrected image; (right) geometrically corrected image.

191 Rosa Maria Cavalli et al. accomplished in one single phase by means of and minimize the drawbacks occurring during a ÇNearest NeighbourÈ re-sampling. image mosaiking, the area was divided into 3 This pre-processing chain methodology was sectors, each one with its own morphologic and developed and then tested on the four Natural environmental characteristics. The first sector is Reserves, as their vegetation complexity consti- constituted by the mosaic of runs 1 to 7, charac- tutes a suitable test bench to experiment and terized by mainly plain and hilly areas. The sec- improve the different techniques adopted. ond one is formed by the mosaic of the runs 8 to In view of the botanic and naturalistic im- 18, mostly characterized by sub-mountain and portance of these Natural Reserves, the Map of mountain ranges. The third one comprises the Prevailing Vegetation Classes was produced at mosaic of runs 19 to 28, especially characterized 1:10000 scale (Cavalli et al., 2001), while the by mountain and high relief surfaces. These are- corresponding map referring to the entire air- al divisions also minimized the problem of the borne remotely sensed area is at 1:25000 scale. non uniform distribution of ground truths in all In addition to the Map of the Main Vegetation the scenes for each spectral class. Classes, other specific thematic maps have been To facilitate the image supervised classifica- produced concerning some environmental and tion, a first discrimination of the main homoge- phytosanitary indicators like soil humidity, dry neous land cover classes was defined by photo- vegetation, NDVI (Normalized Difference Veg- interpreting MIVIS images (see table II) to cre- etation Index) and NDWI (Normalized Differ- ate uniform spectral classes. ence Water Index; Gao, 1996). Informative layers were then produced by digitizing on screen their limits (so obtaining vector files), covering the following classes: 4. Implementation of the Map of the Main arable lands, permanent crops, urban areas, Vegetation Classes quarries, waters, natural environments. Vector layers so obtained were transformed into masks To manage the great quantity of available da- and used to create thematic subsets over which ta as best as possible, a relatively simple method- the usual techniques of image processing can be ology has been defined, which could be repro- applied for the supervised extraction of spectral ducible and applicable to all remotely sensed classes. In particular to perform the ÇMap of runs. In particular to reduce data processing time Prevailing Vegetation ClassesÈ, a mask of natu-

Table II. False Color Composites (FCC) and Black and White (BW) images used for photo-interpretation. A-com- binations of bands corresponding respectively, for vegetation, to the two bands with higher spectral absorption (red and blue colours) and the one with major reflectivity (green colour): it better approaches the natural colour repre- sentation; B-combinations of channels corresponding respectively to the maximum reflectance plateau in the Near- Infrared, the water absorption peak and chlorophyll peak; hence, this FCC highlights vegetation covers, discrimi- nating between agricultural lands and spontaneous vegetation; C- and D-BW images which allow to emphasize less easily detectable covers such as bare soils, vegetation sparse or in relative dryness conditions.

FCC Red Green Blue BW images

A Ch. 13 Ch. 7 Ch. 1 (0.6 nm

192 The natural areas of Rome Province detected by airborne remotely sensed data

Table III. Definitions of prevailing vegetation classes.

Forest formations Shrubs formations Herbaceous formations Anthropics areas

Evergreen woods with prevalence Prevailing deciduous High mountain pastures Arable lands of Quercus ilex; Quercus suber and evergreen sclerophyll with Sesleria tenuifolia. woods with Styrax officinalis shrubs. and Castaneasativa (only for Catillo Mt. Natural Reserve). Mixed woods with dominant Shrubs with Crataegus sp. Mountain and Permanent crops presence of Quercus cerris; Prunus sp. Spartium sub-mountain pastures. (only for Natural mixed woods with local junceum, Cytisus sp.; Reserves) prevalence of Quercus robur shrubs with Erica arborea, (only for Nomentum and Cystus salvifolius Macchia di Gattaceca and Spartium junceum Natural Reserves). (only for Catillo Mt. Natural Reserve). Mixed woods with dominant Mountain and high Mesophylous Urban areas presence of Quercus pubescens. and high mountain shrubs pastures. with prevailing presence Juniperus sp. Mixed woods with dominant Thermophylous Ornamental presence of Ostrya carpinifolia. pastures with green Ampelodesmos mauritanicus; garigues. Castanea sativa woods. Bare surfaces; Quarries rock outcrops. Fagus sylvatica woods. Riparian Vegetation with dominant Waters presence of Populus alba and Salix alba. Conifers woods with dominant Unclassified presence Pinus nigra. Thickets in evolution; thickets with Cercis siliquastrum, Styrax officinalis, Pistacia terebinthus and Acer monspesulanum (only for Catillo Mt. Natural Reserve).

ral vegetation was created for each sector using In order to define correct vegetation classes the Çnatural environmentsÈ informative layer; the naturalists of Rome Province carried out this mask enabled us to improve the classifica- several phyto-sociological surveys in situ (in to- tion of vegetation classes as much as possible tal 160 field surveys performed on the Region- and to remarkably shorten the processing time. al Technical Map at scale 1:10000) according Therefore, this selective approach does classify to Braun-Blanquet’s method that was further in- areas of vegetation included in the natural envi- tegrated with previous works accomplished in ronments, so removing urban areas, arable the surveyed area (Brandizzi et al., 1974; Ab- lands, etc. bate et al., 1981; Avena et al., 1990; Centro Re-

193 Rosa Maria Cavalli et al. gionale per la Documentazione dei Beni Cultur- The classification accuracy of the thematic ali e Ambientali, 1993). The main physiognom- map was estimated by confusion matrix and K ic classes defined by the naturalist are reported coefficient (see table IV). Overall accuracy is in table III. The in situ observations enabled us computed by dividing the total correct pixels to identify spectrally homogeneous classes on (sum of the major diagonal) by the total number the airborne remotely sensed images, useful for of pixels in the confusion matrix. Kappa analy- defining 133 Regions Of Interest (ROI, as im- sis is a discrete multivariate technique used in plemented in the ENVI 3.4 software package; accuracy assessments, that is a measure of RSI, 2000), of which about 70% as training ar- agreement or accuracy (Congalton, 1991). The eas, and 30% to further classification testing. evaluation procedure applied to the data gave The ROI were chosen to represent adequately an overall accuracy equal to 80.69% and a K the vegetation classes shown in table III, per- coefficient equal to 0.7899. taining to forest, shrub and herbaceous forma- The classification produced good results in tions for a total of 17 vegetation classes; they the discrimination of forest and shrub formations were selected within several morphologic con- and herbaceous formations. In particular, be- texts to better represent the spectral class vari- tween forest formations the methodology well ability. discriminated conifers from broadleaves (e.g., The classification procedure is based on the Pinaceae from Fagaceae), and between broad- Maximum Likelihood (ML) algorithm applied leaves, evergreens from deciduous ones (e.g., to the first 28 bands (spectral range from 0.43- Quercus ilex woods from Castanea sativa 1.55 nm). The ML classifier was chosen be- woods), mainly in those areas where broad- cause allows a better recognition of those spec- leaves form pure and well developed forests. tral classes, like the vegetation species, which Less accurate results were obtained for the class- present very correlated spectra. This algorithm es that form mixed woods, among them: mixed assumes that the statistics for each class in each woods with dominant presence of Quercus cer- band are normally distributed and calculates the ris; mixed woods with dominant presence of probability that a given pixel belongs to a spe- Quercus pubescens and mixed woods with dom- cific class (Richards, 1999). Each pixel is as- inant presence of Ostrya carpinifolia. The large signed to the class that shows the highest prob- vegetation biodiversity was highlighted locating ability. and mapping also vegetation classes with limited The cartographic restitution of the final the- extension and circumscribed within certain areas, matic product was obtained by mosaiking the or local prevailing species inside more extended three sectors into a GIS-compatible (GeoTIFF) formations. From this point of view, the location image format, using the Gauss-Boaga East time and mapping of Quercus suber woods and Cercis zone reference system (fig. 4). siliquastrum thickets, of high naturalistic and en- vironmental value, located in the Catillo Mt. Nat- ural Reserve, also represented an important re- 5. Results and discussion sult. It was possible to localize Quercus robur communities within into Quercus cerris woods, The ÇMap of Prevailing vegetation classesÈ present in the Nomentum and Macchia di Gat- (fig. 4) also reproduces other classes pertaining to taceca Natural Reserves. The identification of ar- the anthropic formations defined in table II. The eas with ÇanomalousÈ spectral responses (in the surface percentages of the cover classes found the sector between the villages of S. Vito Romano study area are as follows: forest formations and Capranica Prenestina, Prenestini Mt.), inter- 35.10%, shrub formations 15.23%, herbaceous preted as Castanea sativa woods deteriorated by formations 14.18 %, bare surfaces and rock out- parasites’ attack, proved to be interesting as a fu- crops 0.65%, Waters bodies 0.11%, arable lands ture application hypothesis. Moreover it was pos- 30.71%, urban areas 3.92%, quarries 0.11%; con- sible to map complex vegetation covers (mainly sequently the natural and semi-natural area cov- shrubby formations), otherwise difficult to locate ers about 65% of the entire study area. with traditional techniques. The integrated ap-

194 The natural areas of Rome Province detected by airborne remotely sensed data l Park Map of prevailing vegetation classes. A Ð Natural Reserve of Nomentum; B Ð Natural Reserve of Macchia di Gattaceca; C Ð Regiona of Nomentum; B Ð Natural Reserve A Ð Natural Reserve classes. vegetation Map of prevailing Fig. 4. of Catillo Mt. of Simbruini Mts.; E Ð Natural Reserve Park of Lucretili Mts.; D Ð Regional

195 Rosa Maria Cavalli et al.

Table IV. Confusion matrix.

Ground Truth (pixels) Confusion Sparse Crataegus sp., Deciduous, Rock Mesoph. Shrubs with Thermoph. Thickets Mountain matrix veget. Spartium evergreen outcrops pastures Juniperus sp. pastures in evolution pastures junceum, shrubs Cytisus sp. Class Unclassif. 6 23 28 34 12 15 39 107 12 Sparse veg. 373 00000001 Crataegus sp., 0 125 500 9 0590 Spartium junceum, Cytisus sp. Deciduous, 0 0 280 0 1 12 0 27 19 evergreen shrubs Rock 0 0 0 305 00200 outcrops Mesoph. 0 2 0 0 442 17 0 0 11 pastures Shrubs with 0 3 2 0 6 258 0016 Juniperus sp. Thermoph. 4 0 0 0 0 2 400 00 pastures Thickets 0 10 3 0 0 0 0 473 0 in evolution Mountain 0 0 6 0 7 11 0 1 533 pastures Quercus 02 300 0 080 ilex Quercus 00 000 0 000 cerris Ostrya 00 000 1 020 carpinifolia Castanea 00 001 0 040 sativa Fagus 00 000 0 000 sylvatica Quercus 00 802 6 090 pubescens Riparian veg. 0 3 0 0 0 1 0 0 0 Conifers 0 0 0 0 0 0 0 0 0 Total 383 168 335 339 469 331 441 687 592 proach between image processing and photoint- chiefly characterized by covers deriving from an- erpretation disclosed 22 land cover classes. thropic activities, the areas devoted to sowing In particular, the analysis of the map shows cultivations, woody cultivations (with a large dif- that in the western sector (to west of Tivoli city), fusion of olive trees) and pastures prevail. It is al-

196 The natural areas of Rome Province detected by airborne remotely sensed data

Quercus Quercus Ostrya Castanea Fagus Quercus Riparian Conifers Total Confusion ilex cerris carpinifolia sativa sylvatica pubescens vegetation matrix

Class 34 101 164 26 39 67 1 4 710 Unclassif. 0 0 0 0 0 0 0 0 374 Sparse veg. 2 0 0 10 0 1 1 0 212 Crataegus sp., Spartium junceum, Cytisus sp. 1 1 8 0 0 12 0 1 360 Deciduous, evergreen shrubs 0 0 0 0 0 0 0 0 307 Rock outcrops 0 0 0 5 0 1 0 0 477 Mesoph. pastures 0 0 0 4 0 0 2 0 290 Shrubs with Juniperus sp. 0 0 0 0 0 0 1 0 407 Thermoph. pastures 13 1 2 2 0 4 0 0 508 Thickets in evolution 0 0 0 0 1 0 0 0 559 Mountain pastures 451 0 17 1 0 70 6 18 575 Quercus ilex 3 485 92 0 114 8 0 0 702 Quercus cerris 126422 0272200501Ostrya carpinifolia 11 0668 2 0 0 0 676 Castanea sativa 026660733 0 0 0 825 Fagus sylvatica 87 10 81 4 0 327 0 5 538 Quercus pubescens 28 0 0 0 0 1 197 2 231 Riparian veg. 14 0 0 0 0 4 0 492 510 Conifers 636 649 853 720 916 515 208 521 8762 Total so possible to note an evident building develop- most significant and extended natural areas are ment of urban centers, mainly along the most im- located in or very close to the Natural Reserves portant roads, and a diffuse and progressive ur- (Nomentum and Macchia di Gattaceca), last evi- banization of the surrounding countryside. The dence of a habitat, once more extended, which

197 Rosa Maria Cavalli et al. characterized the so called Çagro romanoÈ, environmental recovery plans of the areas Mt. namely the countryside around Rome. Soratte, Mt. Guadagnolo, Travertini Acque Al- The central and eastern sectors (to east of bule (Tivoli), pertaining in the area study, indi- Tivoli city), mainly mountainous, are mostly cated to the European Community as area SIC characterized by relatively intact habitat, with and ZPS. wide continuous natural areas or in progress of The additional maps show other land cover re-naturalization, extended and compact forest types, such as the different agricultural crops (e.g., formations which are well developed, diffused corn, wheat, beetroot, etc), artificial surfaces in prairies and large shrubby formations. Shrubby urban areas (brick, asbestos-cement, etc.). formations present a broader variety in their The MIVIS image processing procedures floristic composition and spatial distribution, applied in this project are characterized by the forming, together with some pastures and sparse use of the correction of the atmospheric contri- or grouped trees, vegetation mosaics character- bution due to the path radiance, a geometric ized by a great lack of cover homogeneity, and correction procedure (including the land orog- giving the relevant land a wide variability of the raphy) and particularly for the application of distributive pattern of herbaceous, shrubby and selective criteria between informative layers of arboreal components. Such variability especial- the main land cover classes (first level classes ly occurs where agricultural activities are being in the classification hierarchy). This procedure progressively deserted (a phenomenon common then enabled us to process many adjacent enough in this area) and uncultivated soil is scenes in a fast and practical manner, to man- tending to re-naturalize. Anthropic activites are age classes statistically homogeneous, and as a mainly concentrated along valleys (Aniene Val- consequence to produce thematic maps at dif- ley) and plain areas. ferent scales characterized by relatively high K values. The use of the above described procedures 6. Conclusions allows us to develop a follow-on phase which will likely include classification refinements in The multidisciplinary formulation given to areas presenting a strong dynamism, and the the research activity planning, with the constant better diversification of some vegetation types contribution of expertise from different disci- (e.g., shrubby and herbaceous communities). plines, like botanists, naturalists and remote The availability of a DEM will also permit us to sensing experts, has certainly aided in the good carry out a morphological study of the area with outcome of the project. reference to the relations between morphologi- The thematic maps produced have been cal classes and vegetation classes, as already helpful in the preliminary drawing up of the made on the area of Mt. Soratte. Management Plans of the Natural Reserves, Based on the encouraging results obtained which should constitute the reference for the so far by means of the airborne remotely sensed future land management policy and the en- data on the area of Rome Province, interesting hancement of biocompatible activities, in re- scenarios can be hypothesized for a future de- spect and safeguard of the environment. velopment of common activities (CNR - Local The developed methodology enabled us to Authorities) finalized to plan and implement of rapidly and practically process MIVIS data effective actions for the safeguard and manage- covering wide land extensions, and to attain the ment of vegetation and forest resources. map of the main vegetation classes with a good confidence. The application of such procedure also permitted us to produce maps with ancil- Acknowledgements lary information of interest for the environmen- tal planners of the Province Administration. The authors are grateful to Maurizio Poscol- With reference to this the Rome Province Ad- ieri for valuable discussions and to Angela Mi- ministration is completing the management and rabelli for help in improving the paper.

198 The natural areas of Rome Province detected by airborne remotely sensed data

REFERENCES spettrali telerilevati per la gestione di riserve naturali della Provincia di Roma, in Atti della V Conferenza ABBATE, G., G.C AVENA, C. BLASI and L. VERI (1981): Car- nazionale ASITA, vol. I, 477-482. ta della vegetazione del M.te Soratte 1:10000, in Stu- CENTRO REGIONALE PER LA DOCUMENTAZIONE DEI BENI dio delle Tipologie Fitosociologiche del Monte Soratte CULTURALI ED AMBIENTALI (1993): Carta del Paesaggio () e loro Contributo nella Definizione Fito- vegetale della Valle del Tevere (scala 1:50 000), (Ass.to geografica dei Complessi Vegetazionali Centro-Appen- alla Cultura - Regione Lazio). ninici (P.F. «Promozione della qualità dell’ambiente» CONGALTON, R.G. (1991): A review of assessing the accura- CNR, Roma), AQ/1/125, 1-41. cy of classifications of remotely sensed data, Remote AVANZI, G., C.M. MARINO and S. PIGNATTI (1997): Geocod- Sensing Environ., 37, 35-46. ifica e mosaico su modello altimetrico digitale di im- CONGALTON, R.G., K. GREEN and J. TEPLY (1993): Mapping magini MIVIS, in Atti della I Conferenza Nazionale old-growth forest on National Forests and park lands in ASITA, p. 179. the Pacific Northwest from remotely sensed data, Pho- AVANZI, G., A. PALOMBO and S. PIGNATTI (2006): MIVIS togramm. Eng. Remote Sensing, 59, 529-535. image geocoding experience on merging position atti- GAO, B.-C. (1996): NDWI Ð A Normalized Difference Wa- tude system data and public domain GPS stream (ASI- ter Index for remote sensing of vegetation liquid water GeoDAF), Ann. Geophysics, 49 (1), 11-20 (this vol- from space, Remote Sensing Environ., 58, 257-266. ume). HILL, J., J. MEGIER and W. MEHL (1995): Land degradation, AVENA, G.C., L. BONIFAZI, S. FASCETTI and L. RUBECA soil erosion, and desertification monitoring in Mediter- (1990): Carta della Vegetazione del Territorio della IX ranean ecosystems, Remote Sensing Rev., 12, 107-130. Comunita’Montana del Lazio, Compreso nei Limiti del LAMBIN, E.F. (1998): Modeling and monitoring land-cover Parco dei M.ti Lucretili (scala 1:25000), (Dipartimen- change pro-cesses in tropical Regions, Prog. Phys. Ge- to di Biologia Vegetale, Università degli Studi di Roma ogr., 21 (3), 375-393. ÇLa SapienzaÈ). MONTELUCCI, G. (1972): Considerazioni sul componente BERK, A., L. BERNSTEIN and D. ROBERTSON (1989): MOD- orientale nelle foreste della Penisola, Ann. Accad. Ital. TRAN: a Moderate Resolution Model for LOW- Sci. For., XXI, 121-169. TRAN7, Final Report, GL-TR-0122, AFGL, Hanscom MONTELUCCI, G. (1978)(1976-1977): Lineamenti della veg- AFB, p. 42. etazione nel Lazio, Ann. Bot. (Roma), XXXL-VI,1- BLASI, C. (1994): Fitoclimatologia del Lazio, Fitosociol., 107. 27, 151-175. MONTELUCCI, G. (1984): I monti di Tivoli dal punto di vista BORAK, J.S., E.F. LAMBIN and A.H. STRAHLER (2000): The botanico, Natura e Montagna, 3, 37-48. use of temporal metrics for landcover change detection MORENO, J.M. and W.C. OECHEL (Editors) (1995): Global at coarse spatial scales, Int. J. Remote Sensing, 21 (6- Change and Mediterranean-Type Ecosystems (New 7), 1415-1432. York, Springer-Verlag), p. 527. BRANDIZZI, G., A. ANGELUCCI,A .BIASINI, G. CASELLA,R. NEMANI, R. and S.W. RUNNING (1996): Global vegetation FUNICIELLO, M. GIANNINI, G. GIOVAGNOLI, L. RAMPI- cover changes from coarse resolution satellite data, J. CHINI, A. PANASCé and G. BIGI (1974): Carta Agrofore- Geophys. Res., 101, 7157-7162. stale della Provincia di Roma (Camera di Commercio, RICHARDS, J.A. (1999): Remote Sensing Digital Image Ana- Industria, Artigianato e Agricoltura di Roma). lysis (Springer-Verlag), p. 240. CAVALLI, R.M., L. FUSILLI, A. GUIDI, S. PANZARASA, S.PIG- RSI (2000): ENVI User’s Guide (Research Systems Inc.), NATTI, A. VINCI and C.M. MARINO (2001): Dati iper- 261-288.

199