HEAT ISLAND STUDY IN THE AREA OF ROME BY INTEGRATED USE OF ERS-SAR AND LANDSAT TM

ABSTRACT

Modifications of natural land cover together with localized industrialization and enormous increase in motor traffic greatly contribute to air pollution and degradation of environmental and climatological quality in urban areas. As concentration of world population in urban areas shows an increasing trend, there is no doubt that climatological elements (i.e. temperature, ventilation, sun/shading, relative humidity) have to be integrated in urban planning and building to improve quality of life. The present study is aimed at contributing to a better understanding of climate characteristics in the "heat and roughness island" of Rome and surrounding rural areas. The satellite point of view (time synchronized dense grid of data over the whole area) allows a double-sided approach: 1) the city as a whole, with its overall interactions with neighboring areas; 2) local features within the city. As for point 1), three Landsat TM images of a large area around Rome (including sea coast, Bracciano lake, Tevere valley and Castelli Romani) for different season and meteorological conditions have been compared for all bands. Effects of interaction with the regional circulation were observed. As for point 2), multitemporal radiometric and texture properties of various land cover types (different types of building, urban fabric and vegetation) have been analysed in nine ERS-1 SAR.PRI images (acquired during 1992 under different meteorological conditions) together with Landsat TM thermal bands and NDVI. The aim is to obtain a classification of urban land-cover to be compared with temperature patterns as derived from TM thermal band. Combination of SAR data with NDVI looks very promising to this purpose.

Keywords: Land remote sensing, urban areas, multisensor analysis, environmental monitoring

Introduction

Urban areas all over the world are facing an increasing concentration of almost all economic, finantial and cultural activities of humanity. Together with industrialization and vehicles traffic, this concentration is leading to a progressive degradation of living environment, in terms of chemical and noise pollution, biodiversity loss and climate change. Effects of urbanization on daily mean air temperature at ground (of the order of 0.1 - 0.4 °C in the last century) have been noticed even at larger scale in data from US Historical Climatological Network, as formerly rural stations were progressively approached by urbanization (Jones et al., 1990).

Judgement of the specific urban environmental characteristics which make quality of life good, less good or unattractive, certainly involves many subjective aspects (i. e. genetical, cultural, age, etc.), neverthless a fair agreement might be reached on apparently contradictory aspects, like: no change/new opportunities, high/low population densities, privacy/accessibility, order/no planning, energy consunption/pollution, etc.

Meteorological and climatic aspects in urban areas are widely recognised to be of major importance in this context (Bitan, 1992). They can be controlled/improved by a wise urban planning and management, based on careful understanding of related physical phenomena, and knowledge of real-time situation over the whole area (Nichol, 1994, 1996). Space remote sensing techniques show a great potential of operational applications in this field (Abbate et al., 1995; Parlow, 1996; Scherer et al., 1996). A typical climatic phenomenon of urban areas is "heat island" (Oke, 1995), with related features in local and regional atmospheric circulation, cloud cover and precipitation.

Urban heat islands were observed on meteorological satellites infrared images, since the beginning of the Seventies, with relatively poor spatial resolution (7-8 Km). Several studies based on AVHRR data (spatial resolution 1.1 Km) are reported in literature (Balling and Brazel, 1988; Gallo and Tarpley, 1996). At this spatial resolution, temperature variations across metropolitan areas became evident; direct correlation was found between satellite derived surface temperature values and incidence of residential, commercial and industrial land-use, while indirect correlation was observed with values of NDVI (Normalized Vegetation Index), that is with urban-green land-use.

HCMM satellite (Heat Capacity Mapping System, Goddard Space Flight Center, 1978) acquired thermal data near the time of diurnal maximum, at a spatial resolution of 500 m, which were proven useful for placement of air monitoring stations in cities and for spatial interpolation between such sites (Price, 1979).

Ten years ago it was commonly believed that satellite images were only adequate for comparing urban-rural temperature differences and that inner structure of urban heat islands could not be studied by means of this technology, at least at that stage of development. Thoretical modelling approaches and measurements campaigns by conventional instrumentation and remote sensing from ground and aircraft were considered more useful for understanding atmospheric boundary layer phenomena in urban areas (Plate, 1993), even if under very limiting conditions.

Just a few attempts were made afterwards of using more recently available Landsat TM data, having spatial resolution of 120 m in thermal infrared band, and 30 m in other bands (Kim, 1992). Very recently, these data have been processed in combination with other types of spatial information and accurate ground-truth measurements, by means of GIS techniques, to identify thermal characteristics of urban features down to the scale of a city block, a single row of trees, or an individual building (Nichol, 1996). ERS- SAR data (spatial resolution = 12.5 m) are being exploited to derive 3-dimensional and roughness characteristics of land (Parlow, 1996; Scherer, D., 1996), and soil moisture (Borgeaud et al., 1994), parameters which play a fundamental role in urban climate. More accurate surface temperature measurements can be obtained by ERS-ATSR, even if at spatial resolution of 1.1 Km. Digital techniques for fusion/merging of multi-sensor, multi-band, multi-temporal data allow to improve spatial resolution. Data from new sensors with improved performance characteristics will shortly be available. Thus, operational methodologies for urban planning and management, based on satellite data, are very likely to be finalized in the near future.

Many physical features can affect the climate of an urban area, including: location of the urban area within a given region, density of built-up areas, height distribution of buildings, orientation and width of streets, position and design details of green areas, design details of buildings which affect outdoor conditions (Givoni, 1992). All them can be analysed on the whole area of interest by means of techniques of Earth observation either presently available or planned to be operational in the near future. To resolve the temporal evolution of physical phenomenons - for forecasting purposes - parallel research on modelling and assimilation of satellite-derived parameters will have to be carried out.

Rome heat island evolution as observed by Landsat TM

Three Landsat TM images of a wide area around Rome (including sea coast, Bracciano lake, Tevere valley and Castelli Romani) have been compared for all bands:

Path/Row = 191/031, acquisition dates: 4 May 1994, 15 January 1995, 26 July 1995, selected for different seasons and wind conditions. In particular, regional wind was blowing from N-NW on 4 May 1994, from N-NE on 15 January 1995 (strong intensity), from SW on 26 July 1995 (pure sea breeze).

Significant variations were observed in bands 3 (red) and 4 (reflective infrared), due to vegetation species phenological cycle, and in band 6 (thermal infrared), due to changes in thermal emission. Fig. 1 (a), (b), (c), shows RGB = 6,4,3 band combination, for the study area. NDVI (Normalized Differential Vegetation Index) values were computed from bands 3 and 4:

( TM3 - TM4) / ( TM3 + TM4).

Vegetated areas appear relatively bright.

Brightness temperatures were computed from band 6, according to formula (Malaret et al., 1985):

T(K) = 206.127 + 1.0545 * TM6 - 0.00371 * TM62 + (6.606 * 10 -6) * TM63.

Accurate thermal emissivity measurements will have to be carried out in order to derive surface temperatures.

Different heat island patterns and effects of interaction with the regional circulation can be observed. Rome heat island appears someway "pushed" downwind, at least at an intuitive level of interpretation.

In particular, on 26 July image - Fig. 1(c) - reddish colours (warmer areas) appear as "pushed" from SW towards inland reliefs and entering the Tiber valley up to a distance of about 60 Kms. This is a well known characteristic summer behaviour of sea breeze in this area (Colacino and Dell' Osso, 1978).

In visual analysis of images, different vegetation conditions have to be taken into account. To this purpose RGB = 3,2,1 band combination and NDVI maps are very useful. For instance, it can be noticed that the N-NW facing slopes of Castelli Romani, are less vegetated in May than in July. This characteristic may account for this area being relatively warmer in May. CORINE 1:100.000 land cover map (Centro Cartografico Interregionale, Roma) is also helpful in the interpretation of temperature features in rural areas as related to land-cover. The effect of wind is noticeable in the image of May, Fig. 1 (a), as the SE sector of Rome looks generally warmer than the rest of the city. In the image of 15 January, Fig. 1 (b), water and highly vegetated areas appear warmer. City area is relatively cold, as Romans experience early morning in these cold winter days (satellite passage is at 8:59 GMT). Even if the time of observation is rather early in the morning, still the presence of an heat island can be observed.

SODAR measurements carried out in the area of Rome during 1992 spring and summer gave evidence that some convective activity is still present in the urban area during night and early morning hours (plume height about 50-100 m). Convection increases due to surface heating by the sun generally between 8:00 and 9:00 local solar time (Mastrantonio et al., 1994).

SAR multitemporal study of the urban area of Rome

Nine SAR.PRI images were selected for different meteorological conditions: Frame 2763, descending path, dates: 21 Jan., 17 Feb., 15 March, 26 May, 11 June, 16 July, 24 Sept., 29 Oct., 17 Nov. 1992.

ERS-1 SAR multitemporal composites

After co-registration (RMS error was generally about 0.5-0.6 pixels), various multitemporal combinations were produced and visually compared, i.e: RGB = 15 March/16 July/17 November, to look for annual variations; RGB = 11 June/16 July/24 September, summer variations; RGB = 17 February/15 March/26 May, spring-summer variations; RGB = 29 October/17 November/ 21 January, winter variations. To better locate various features, images were also registered to a 1:200.000 map. Variations in vegetated areas are seen very clearly (plant phenological cycle, irrigation / rain, ploughing, etc.). Agricultural fields appear as colour patterns, as SAR backscatter signal is very sensitive to surface water content and roughness. City appears generally bright. Structure and density features are well distinguishable at a closer look as it can be noticed in subset in Fig. 2 (a), (b), RGB = 16 July/15 March/ 17 November and 16 July/26 May/17 November, respectively. Note that the line bordering Tevere river at the left of images appears in different colour in the two combinations, this indicating highest backscattering on 15 March and, decreasingly, on 16 July, 26 May, 17 November. Interpretation requires ground-truth survey. Note also that this level of detail allows to distinguish single buildings. The impression of out-of-focus can be reduced by further co-registering subsets at this full detail.

ERS-1 SAR multitemporal analysis

To allow quantitative comparison among data sets, radar backscattering coefficients ° (dB) were derived, according to formula:

° = <I>/K() = (<I>/ K) * ( sin n / sen (23°)).

Statistics were computed for twenty three training areas, covering interesting types of surfaces (different types of buildings, urban fabric, vegetation, water). Here following, they are listed and generally described in order of increasing NDVI: 1-Bracciano lake (water area), 2-Albano lake (water area), 3-Historical centre (urban, old district with narrow streets, no greenery), 4-Termini station (urban, old district, concrete, asphalt, no greenery), 5-Bologna square (not high-rising buildings, no greenery), 6-Centocelle (modern urban district, high buildings, almost no greenery), 7-Mazzini (residential, large villas), 8-Garbatella (modern urban district, wide roads), 9- Mostacciano (new villas), 10-Parioli (villas and low-rising buildings), 11-Nuovo Salario (modern new district), 12-Colosseo (ancient ruins, bare soil and some greenery), 13- Fiumicino airport (international airport), 14-EUR (modern urban district with high buildings and wide roads), 15-Ponte Galeria (crops), 16-Appia Antica (grass, some trees), 17-Isola Farnese (grass), 18-Villa Borghese (urban park), 19-Settebagni (crops), 20-Monterotondo (crops), 21-Villa Ada (urban park), 22-Mentana forest (woods), 23-Albano forest (woods).

Results are summarized in Fig.3 and Fig.4.

Plot in Fig.3 shows mean, minimum and maximum ° values for each subset among the nine images. The lowest values are found for water areas. Some differences are observed between high density urban areas and low density urban and agricultural areas.

Plot in Fig. 4 shows for each subset, the difference between maximum and minimum values of standard deviation in the nine ERS-1 SAR images. It can be seen that this parameter better discriminates built-up areas (subsets 3-12); crops and deciduous vegetation are in fact characterized by high standard deviation due to plants phenological cycle and irrigation, while water has high standard deviation values due to waves on the surface. Grass areas (subset 17) and evergreen parks (subsets 18 and 21) have standard deviation ranges of the same order as urban areas. Consequently, while this parameter can provide useful information to discriminate and classify urban areas, it has to be used in combination with NDVI.

Conclusions

This study has allowed to gain insight in the heat island phenomenon of the area of Rome, and in opportunities offered by satellite data for this kind of studies. In particular, combination of ERS-SAR data with thermal and NDVI data from Landsat looks very promising.

Further work will include:

1) Further analysis of Landsat TM images to study temporal and spatial evolution of temperature patterns in the region around Roma;

2) Emissivity measurements and derivation of ground temperature maps;

3) Use of ERS ATSR data to derive temperature at ground, and comparison with the above;

4) For a subset (only urban area) co-registration with a high spatial resolution image and correlation analysis with ground cover;

5) Computation of the average ° image from the above mentioned 9 SAR.PRI images; texture analysis and maybe classification of average image (different structure of buildings and urban fabric, direction of main roads). Maybe also use of other algorithms - to be studied - based on ascending/descending SAR images, to obtain quantitative parameters for urban structure;

6) Classification of NDVI map (different plant cover and distribution in urban districts and close surroundings);

7) Comparison of thematic layers 5) and 6) with temperature maps;

8) Integration of texture and vegetation information into ground temperature maps;

9) Integration of ground-truth data, to be acquired synchronously with satellites passages.

Acknowledgements

The present work was carried out under ESA Project Ref. I 102/0, "Heat Island Study in the area of Rome by integrated use of remote sensing techniques".

I wish to thank M. Fea, J. Lichtenegger, A. Bellini, M Barbieri, A. Argentieri, at ESA-ESRIN, Frascati, for support and enlightening discussions. Advice by K.S. Rao (Indian Institute of Technology) and technical help by L. De Cecco and S. Martini at ENEA-Casaccia are also gratefully acknowledged.

References

Abbate, G., Kowalewska, H., Borfecchia, F., Della Rocca, B., 1995:

Land cover characteristics of Rome urban and countryside area as observed by Landsat TM (bands 1-4) and ERS-1 SAR, Proc. 7-th URSI Commission F Open Symposium, Wave propagation and remote sensing, Ahmedabad, India, 20-24 Nov. 1995.

Balling, R. C., Brazel, S. W., 1988:

High-resolution Surface Temperature Patterns in a complex Urban Terrain, Photogrammetric Engineering and Remote Sensing, Vol. 54, No. 9, pp. 1289-1293.

Bitan, A., 1992:

The high climatic city of the future", Atmospheric Environment Vol. 26B, No. 3, pp. 313-329.

Borgeaud, M., Noll, J., Bellini, A., 1994:

Use of ERS-1 SAR data for land applications, Proc. Second ERS-1 Symposium - Space at the Service of our Environment, Hamburg, Germany, 11-14 October 1993, ESA SP-361.

Colacino, M., Dell' Osso, L., 1978:

The local atmospheric circulation in the Rome area: surface observations, Boundary Layer Meteorology, Vol. 14, pp. 133-151.

Gallo, K. P., Tarpley, J, D., 1996:

The comparison of vegetation index and surface temperature composites for urban heat-island analysis, Int J. Remote Sensing, Vol.17, No. 15, pp. 3071-3076.

Givoni, B., 1992:

Climatic aspects of urban design in tropical regions", Atmospheric Environment Vol. 26B, No. 3, pp. 397-406.

Jones, P. D., Groisman, P. Y., Coughlan, M., Plummer, N., Wang, W. C., Karl, T. R., 1990:

Assessment of urbanization effects in time series of surface air temperature over land, Nature, Vol. 347, pp. 169-172.

Kim, H. H., 1992:

Urban heat island, Int. J. Remote Sensing, Vol. 13, No. 12, 2319-2336.

Malaret, E., Bartolucci, L. A., Lozano, D. F., Anuta, P. E., and McGillen C. D., 1985:

Thematic Mapper data quality analysis, Photogramm. Eng. Remote Sens., 51 (9), 1407-1416.

Mastrantonio, G.,Viola, A. P., Argentini, S. (CNR-IFA), Fiocco, G., Giannini, L., Rossini, L. (University of Rome), Abbate, G., Ocone, R. (ENEA), Casonato, M. (AMNU), 1994:

Sea breeze observation in the Roman area by a network of Doppler Sodars, Boundary Layer Meteorology, 71: 67-80, 1994.

Nichol, J. E., 1994:

A GIS-based approach to microclimate monitoring in Singapore's high-rise housing estates, Photogrammetric Engineering & Remote Sensing, Vol. 60, No. 10, October 1994, pp.1225-1232

Nichol, J. E., 1996:

High resolution surface temperature patterns related to urban morphology in a tropical city: a satellite-based study, Journal of Applied Meteorology, Vol. 35, No.1, pp.135-146.

Oke, T. R., 1995:

The heat island of the urban boundary layer: characteristics, causes and effects", Wind Climate in Cities, J. E. Cermak et al. (eds.), NATO ASI Series E, Vol. 277, Kluwer Academic Publishers.

Parlow, E., 1996:

Net radiation in the REKLIP area - A spatial approach using satellite data, Progress in Environmental Remote Sensing Research and Applications, Parlow (ed.), Balkema, Rotterdam, ISBN 90 5410 598 4.

Plate, E. J., 1993:

Urban Climates and Urban Climate Modelling: an introduction, lecture at NATO-ASI "Wind Climate in Cities", Waldbronn, Germany, July 1993.

Price, J. C., 1979:

Assessment of the urban Heat Island Effect through the use of satellite data, Monthly Weather Review, Vol. 107, pp. 1554-1557.

Scherer, D., Parlow, E., Beha, H. D., 1996:

Roughness parameter derivation from ERS-1 and Landsat-TM satellite data for the agglomeration of Basel, Switzerland, Progress in Environmental Remote Sensing Research and Applications, Parlow (ed.), Balkema, Rotterdam, ISBN 90 5410 598 4.

Fig. 3. ° (dB) mean, maximim and minimim values of training areas (9 ERS-1 SAR images, 1992). Subsets are in order of increasing NDVI.

Fig. 4. ° (dB) standard deviation range of training areas (9 ERS-1 SAR images, 1992). Subsets are in order of increasing NDVI.

Keywords: ESA European Space Agency - Agence spatiale europeenne, observation de la terre, earth observation, satellite remote sensing, teledetection, geophysique, altimetrie, radar, chimique atmospherique, geophysics, altimetry, radar, atmospheric chemistry