The GeoSat Project: Using Remote Sensing to Keep Pace with Urban Dynamics

T. Santos, S. Freire, J.A. Tenedório
e-GEO, Research Centre for Geography and Regional Planning, Faculdade de Ciências Sociais e Humanas, FCSH, Universidade Nova de Lisboa, Portugal

A. Fonseca, N. Afonso
National Laboratory for Civil Engineering (LNEC), Lisbon, Portugal

A. Navarro, F. Soares
University of Lisbon, Faculty of Sciences, LATTEX-IDL, Portugal


This work was conducted in the framework of project GeoSat – Methodologies to extract large scale GEOgraphical information from very high resolution SATellite images. The purpose of the project was to develop methods to expedite the production of geographic information for municipal planning and land monitoring.

A. The Problem – Motivation for GeoSat

Urban dynamics are induced by such activities as new construction (buildings and roads), demolition of unwanted structures, tree plantings along roads, vacant lots being transformed into urban agricultural sites or green areas becoming parking lots. Maintaining updated cartographic datasets in such environments is a challenging task.

The characterization of urban dynamics considers two key elements – the spatial, or the available space in which growth occurs, and the temporal, or the period during which this development takes place [1].

The temporal dynamics can be characterized as fast, medium or slow, depending on the time frame of the transformation. Temporal urban databases have immediate applications in the monitoring of urban sprawl, watershed analysis, environmental assessment, hydrologic modeling, land surface degradation and the development of predictive modeling techniques to forecast future areas of urban growth.

The spatial dynamics can occur in two ways – through the addition of new areas, e.g., formerly rural areas converted to new urban uses, or through the more intensive, but same use of a site, e.g., changing a residential area from a lower to a higher density. The former leads to urban expansion, and the latter to intensification.

The urban environment is a physical representation of human activities and as such is subject to measurement. A number of data sources help characterize this environment and describe its variability. Census data identifies buildings and their uses, cadastral maps show properties subject to different taxes and maps of urban infrastructure identify water, sewer or high voltage power lines. Geographic information for urban planning is formalized through cartographic representations such as topographic and thematic maps.

Presently in Portugal, the cartographic framework for municipal land planning requires compliance with demanding and complex technical specifications that are mandatory for the production of detailed maps of the required quality. To obtain such large-scale topographic maps, municipalities must devote the necessary human and financial resources. Consequently, detailed spatial information regarding municipal land use is only produced when the Master Plan is prepared. In the Portuguese land planning system, plan revision takes place every 10 years (the legal term for local plans), but longer periods are common.

Such update periodicity does not reflect the dynamic nature of the land use, and hampers the work of the municipal departments that deal with geographic information on a daily basis.

In fact, many situations that occur in the municipal context – such as updating cadastral databases, management of urban areas, street maintenance and construction or planning for such potential disasters as earthquakes or floods – require expedited production of digital maps at large scale.

B. The Solution – Using Earth Observations to Support the Planning Process

Aerial imagery has been the most common data source for mapping human activities in the urban environment [2]. Only recently, satellite images have gained interest as alternative data sources for mapping urban areas, mainly due to their higher spatial detail. This development coincides with the emergence of new integrated methodologies (e.g., GEographic Object-Based Image Analysis – GEOBIA) and new application fields that had previously been the domain of airborne remote sensing and can now be tackled by satellite remote sensing [3].

Very High Resolution (VHR) satellite imagery (i.e., images with spatial resolution equal or greater than 1 m) can now be used as a data source for extracting geographic information at the local scale and on a regular basis. Its acquisition and processing is much easier and quicker when compared with the process based on aerial imagery. The detail and quality of the extracted geographic information, however, is still inferior to that obtained by photogrammetric methods. Nevertheless, for local applications where temporal resolution is paramount, the use of satellite imagery allows users to monitor changes in the urban status, making urban remote sensing a valuable contribution to research in urban geography and planning. Mapping land-use change provides an historical perspective and an assessment of the spatial patterns, rates, correlation, trends and impacts of that change.

C. Direct Mapping vs. Updating of Existing Outdated Cartography

Land status can be assessed through direct mapping or by updating already existing cartography. In fact when cartographic information already exists, but is outdated, a change-detection procedure using recent geographic data can be applied for map updating. The aim of this analysis is to highlight those areas where changes have most likely occurred. Effort is focused in those changed areas, and the remaining ones, i.e., the unchanged areas, keep the geometry and attribute stored in the database.

Large-scale topographic mapping, however, usually has to conform to legal technical specifications and quality standards. Buildings are a major urban element and one of the main feature classes of interest for a municipality, and the “correct” automatic extraction of buildings information from imagery remains a challenging task, even with the advent of high spatial resolution [4]. Difficulties include scene complexity, building occlusions (trees, shadows) and the internal heterogeneity of the feature class [5], and these increase with refinement of image resolution [6].

The scientific literature on the use of VHR images data for mapping purposes suggests that this kind of image can potentially be used for feature extraction and large-scale cartographic updating [7]. Regarding the extraction of specific topographic features from VHR images, several tests on road networks (e.g., [8]) and buildings detection (e.g., [9]) have been reported in the literature, but very few compare the mapping accuracy that can be obtained from high-resolution satellite images to the actual requirements for large-scale mapping in well-mapped countries. In fact, most of the challenge in obtaining a cartographic product from VHR imagery using feature extraction results from the interplay of several factors – the object and its context, the nature of the imagery and the mapping requirements and constraints [4]. Despite the many methodologies proposed for feature extraction, none has so far proved to be effective in all conditions and for all types of data. At present, the quality assessment of extracted buildings is still a complex endeavor for which there is no optimum, consensus or standard approach [4].

Under the GeoSat framework, the potential of VHR satellite imagery and GEOBIA for detection and mapping of urban features and their integration into operational urban planning and management activities was investigated. Santos et al. [10] have explored and proposed detailed vector-based metrics for accuracy assessment of QuickBird-derived buildings, but without taking map standards into account. Recently, Freire et al. [4] presented a methodology that incorporates existing scale-based mapping constraints from official specifications in the process of quality assessment of building polygons extracted semi-automatically from VHR imagery.

D. The City of Lisbon

As of 2001, the city of Lisbon, Portugal, had 556,797 residents, and occupied an area of 84 km2. Lisbon is a typical European capital city, with very diverse land use dynamics, varying from consolidated historical neighborhoods where the street network is dense and most of the area is built up, to modern residential areas with ongoing construction of roads and multi-family buildings. Between these two situations, there are more heterogeneous places with varied land uses such as residential, parks, agriculture, vacant land, industrial, utilities and schools.

Because a new Master Plan has not yet been approved, Lisbon’s official cartography in use in 2011 dates from 1998. In areas subject to strong urban pressures, the frequency of map updating is not compatible with the high rate of change. In these cases, the municipal cartography fails to represent the current reality, thus hindering decision-making on land planning, land-use management and land conservation, as well as compromising policy delineation of human and economic activities, or even limiting efficient law enforcement.

The GeoSat project, which took place between 2008 and 2010 and involved the Lisbon City Hall, developed several sample applications based on VHR imagery to expedite the production of geographic information for municipal planning and land monitoring.

Thematic mapping was addressed in several works. Freire et al. [11] investigated the potential of VHR satellite imagery and GEOBIA for detecting, mapping and characterizing agricultural areas in Lisbon. Dinis et al. [12] applied a methodology based on a multi-temporal satellite imagery dataset and LiDAR (Light Detection And Ranging) data to overcome the problem of shadows in urban areas. Santos et al. [13] tested the contribution of LiDAR data when extracting urban features using VHR imagery. Freire et al. [5] tested the semi-automated extraction of different building types from areas with diverse characteristics, and investigated the impact of the heterogeneity of these features and the urban context in the extraction process.

Land planning also requires information for analytical purposes. Santos et al. [14] analyzed the solar potential of rooftops in an urban context. Santos et al. [15] demonstrated that VHR imagery can be used for quick updating of detailed land-cover information. Based on this recent information several applications can be implemented. Indicators of land-sealing areas and the quantification of green areas and available vacant land in the city are ecological measures that can be used as tools for cities to assess and communicate different environmental risks, and to promote strategies and measures of sustainable urban development and disaster risk management.


This section presents three major GeoSat project applications – the updating of existing maps, the mapping of impervious surfaces and the analysis of rooftop solar potential using LiDAR data.

Figure showing study area location in the city of Lisbon.

Figure 1 – Study area location in the city of Lisbon.

A. Updating of Existing Maps

A multi-temporal strategy for updating a map using existing cartography, a satellite image and an altimetric dataset was applied in a study area located in the eastern part part of Lisbon [16]. The aim of this analysis is to highlight those areas where changes have most likely occurred, thus rendering the existing map outdated.

The selected area occupies 64 ha (800 m X 800 m), and is characterized by several building typologies including industrial properties, schools, apartments and single-family housing (Figure 1).

The spatial database explored in this case study included cartography, satellite imagery and altimetric data. The map to be updated is the Lisbon’s Municipal Cartography from 1998, at a 1:1 000 scale. The altimetric data are compiled by the normalized Digital Surface Model (nDSM). An nDSM is a spatial dataset that depicts the elevation of all objects above the ground. The nDSM of the study area is from 2006, and has 1 m resolution. The imagery includes a pansharp QuickBird image acquired in 2005, with a 0.6 m pixel size.

Figure showing Map updated to 2006 showing the type of changes occurring in the study area.

Figure 2 – Map updated to 2005 showing the type of changes occurring in the study area.

The goal of this application is to produce updated information for the following classes present in the 1998 Municipal Cartography: “Buildings”, “Annexes” and “Shacks”. The extraction methodology was applied to this dataset using a feature extraction software to produce a map of the buildings present in the image. After building extraction, a post-processing stage was conducted to enhance the geometric quality of the elements.

Once the 2005 building map was developed, the next step was to produce a changed map using the 1998 municipal map at the 1:1 000 scale. The change detection process was able to identify missing structures and to detect new ones. For objects with an area larger than 20 m2, the detection quality had an overall accuracy of 99%. Figure 2 shows the updated map with detected changes and their type. In the period under analysis, the main types of change identified in the study area were shack eradication and building demolitions (industrial properties), newly built industrial sites (e.g., the wastewater treatment plant, located in the bottom left corner of the map) and new residential housing (e.g., two multi-family buildings).

Municipal technicians can use this new product to decide, based on analysis of the image and related information, if the marked spot is in fact a change area (a new urbanization or a built-up object that was demolished), and if so, to update the map by adding new buildings or eliminating demolished buildings in the old cartography. Technicians can eliminate any spots considered to be false detections.

Such methodology can be used by the municipality to keep its cartographic database of urban areas up to date in the period between the development of official maps, and can achieve high thematic and positional accuracy.

Figure showing the city of Lisbon and the IKONOS-2 image used for imperviousness mapping.

Figure 3 – The city of Lisbon and the IKONOS-2 image used for imperviousness mapping.

B. Mapping of Impervious Surfaces

Impervious surfaces can generally be defined as anthropogenic features, such as roads, buildings, sidewalks and parking lots, through which water cannot infiltrate into the soil. The artificial surface coverage can be used to evaluate the quality of urban streams, and to study the effects of runoff. Impervious surface is increasingly recognized as a key indicator for assessing the sustainability of land-use changes due to urban growth [17].

An updated and detailed map of imperviousness for the whole city of Lisbon was produced using IKONOS-2 satellite imagery and the nDSM from 2006 [15] (Figure 3).

The imagery classification aims at extracting the three main components of land cover: “Vegetation”, “Impervious Surfaces” and “Soil” [18]. The map of impervious areas includes a wide range of materials, some of which have very different spectral properties (e.g., pavement, concrete and roof tiles). The first level class, “Impervious Surfaces”, corresponds to the land surface after the “Vegetation”, “Soil” and “Shadow and Water” classes are masked out. Based on the pansharp image and the nDSM, six classes were distinguished on the second level of the nomenclature: “Trees”, “Low Vegetation”, “Buildings”, “Roads”, “Other impervious surfaces”, “Soil” and “Shadows and Water”, (Figure 4). The thematic accuracy of the map was investigated and returned an overall accuracy of 89%.

Figure showing land cover map of 2008 derived from IKONOS imagery and LiDAR data for the city of Lisbon.

Figure 4 – Land cover map of 2008 derived from IKONOS imagery from 2008 and LiDAR data for the city of Lisbon.

This application demonstrates that an automated classification of VHR images can produce fast updating of detailed land cover information and can be used to support land planning decisions or to aid in the response to a crisis situation where official maps are generally outdated.

C. Analysis of Rooftop Solar Potential Using LiDAR Data

This application consists of a methodology that applies altimetric data to the evaluation of the potential for incorporating solar power systems into buildings in a city neighborhood. The use of LiDAR data can play an important role in analyzing the suitability of buildings for receiving solar systems. Solar mapping takes advantage of Geographic Information System (GIS) and visualization technologies, and offers a solid knowledge base on solar resources and best practices. Solar maps also offer a comprehensive planning tool to evaluate energy reduction opportunities for new and existing buildings, to plan future energy consumption and supply or to monitor compliance with energy and greenhouse gas goals.

This work, conducted in an area of 625 ha located in the heart of the city, analyzed the suitability of rooftop areas for the installation of solar energy systems [14], and performed a brief technical analysis that considered the optimal location for solar Photovoltaic (PV) systems (Figure 5).

Identifying the incoming solar energy at rooftop level entails the modeling of solar radiation incident in each location. Two inputs are required – a DSM and the buildings’ footprints. With these data, modeling the solar radiation can be done in a GIS environment.

Figure showing study area for solar potential analysis in Lisbon and the Digital Surface Model from 2006.

Figure 5 – Study area for solar potential analysis in Lisbon and the Digital Surface Model from 2006.

The dataset used in this application thus included cartographic and altimetric data. The cartographic data that represent the buildings’ footprints are the Buildings layer of the Municipal Cartography. To characterize the altimetry, the DSM for 2006 was used.

A four-step methodology was applied: 1) calculating the solar energy for the whole surface; 2) assessing the solar energy at the rooftop level; 3) locating the best sites for the installation of PV panels; and 4) quantifying the energy that could be produced (Figure 6). A map of the solar potential of rooftops located in the study area was produced.

This LiDAR-based solar resource map helps rate buildings by the solar resources available, and provides unique information on which parts of the buildings’ roofs are more suitable for solar applications when all critical factors are considered. This information can be used to develop detailed solar generation potential maps.

Figure showing solar potential analysis in the city of Lisbon.

Figure 6 – Solar potential analysis in the city of Lisbon.


Detailed and updated geographic information is essential to effective urban planning and monitoring. Our understanding of nearly every aspect of the changing environment depends upon regular updates of land use/cover status and land-cover conditions, and we need new sources of spatial data and innovative approaches to understand and manage dynamic urban areas. In the remote sensing of cities, VHR satellite imagery offers the opportunity to characterize and monitor the intra-urban environment by enabling discrimination among the land-cover objects that compose this environment.

We have distinguished three areas of application – large-scale map updating, imperviousness mapping and developing indicators of rooftop potential for solar systems – each of which requires its own level of accuracy of geographical information.

For municipal planning according to the technical specifications of large-scale cartography (1: 5 000 scale and higher), map production based on VHR images and photogrammetry is still necessary to guarantee that each uniquely identified feature is well delineated and stored in the database as a geometric entity with a list of attributes. For large-scale analytical applications, however, the current VHR images constitute a valuable source of geographic information, and can play an important role in municipal planning. The three applications presented are a good demonstration of these capabilities.
Based on the premise that a product derived from less accurate images can be effective for land monitoring, the first application proposed an alarm system obtained through satellite and altimetric data processing. The goal was not to provide cartographic data ready for integration into the municipal databases but to assist the process of map updating.

The second application demonstrated that the automatic classification of remote sensing data can expedite the creation of useful spatial knowledge that can support decision-making. The mapping methodology ensures that urban planners have updated land cover data on a regular basis. This tool can be used to monitor the incidence of land cover change within the city, to decide on areas of priority intervention or to assess natural resource sites for preservation and restoration.

Premised on the idea that the wide adoption of solar technologies will depend upon detailed solar suitability information on every building in a community, a map of the solar potential of rooftops located in a study area was produced. This LiDAR-based solar resource map helps rate buildings by the solar resources available, and provides unique information on which parts of the buildings’ roofs are more suitable for solar applications when all critical factors are considered. This information can be used to develop detailed solar generation potential maps. The next step will be making solar maps publicly available. In fact, interactive Web-based urban solar maps are already available [19].

Very High Resolution remote sensing data can contribute to better monitoring, modeling and understanding of urban dynamics and their impacts on the urban and suburban environment, and can enhance the analytical tools available for land-use planners. Our experience, however, suggests that extracting features for large-scale applications still requires much human intervention. Nevertheless, new VHR sensors with high-spectral resolution constitute a new opportunity for urban mapping. The development of object-based algorithms allow the introduction of information such as color, shape, adjacency or context in the classification process, and improve the mapping of urban elements.


[1] M. Batty, Y. Xie, and Z. Sun, “The dynamics of urban sprawl,” Working papers series, Centre for Advanced Spatial Analysis (CASA), University College London, 1999.

[2] T.T. Vu, F. Yamazaki, and M. Matsuoka, “Multi-scale solution for building extraction from LiDAR and image data,” International Journal of Applied Earth Observation and Geoinformation, 11: 281–289, 2009.

[3] T. Blaschke, “Object based image analysis for remote sensing,” ISPRS Journal of
Photogrammetry and Remote Sensing, 65: 2–16, 2010.

[4] S. Freire, T. Santos, A. Navarro, J. Silva, F. Soares, and A. Fonseca, “Extraction of buildings from QuickBird imagery for municipal planning purposes: Quality assessment considering existing mapping standards,” in Proc. GEOBIA – GEOgraphic Object-Based Image Analysis, 2010.

[5] S. Freire, T. Santos, N. Gomes, A. Fonseca, and J.A. Tenedório, “Extraction of buildings from QuickBird imagery – What is the relevance of urban context and heterogeneity?” ASPRS/CaGIS, 2010.

[6] M. Awrangjeb, M. Ravanbakhsh, and C. Fraser, “Automatic detection of residential buildings using LIDAR data and multispectral imagery,” ISPRS Journal of Photogrammetry and Remote Sensing, 65(5), 457–467, 2010.

[7] D.A. Holland, D.S. Boyd, and P. Marshall, “Updating topographic mapping in Great Britain using imagery from high-resolution satellite sensors,” ISPRS Journal of Photogrammetry and Remote Sensing, 60: 212–223, 2006.

[8] U. Bacher and H. Mayer, “Automatic road extraction from multispectral high resolution satellite images,” In U. Stilla, F. Rottensteiner, and S. Hinz (eds), CMRT05. IAPRS, Vol. XXXVI, Part 3/W24, 29-30 August, Vienna, Austria, 2005.

[9] S. Lefèvre, J. Weber, and D. Sheeren, “Automatic building extraction in VHR images using advanced morphological operators,” IEEE/ISPRS Joint Workshop on Remote Sensing and Data Fusion over Urban Areas (URBAN), Paris, 2007.

[10] T. Santos, S. Freire, I. Boavida-Portugal, A. Fonseca, and J.A. Tenedório, “Accuracy assessment of features extracted from QuickBird imagery for urban management purposes,” in Proc. of the 33rd International Symposium on Remote Sensing of Environment, ISRSE, Stresa, Italy, 2009.

[11] S. Freire, T. Santos, and J.A. Tenedório, “Agriculture and food availability – Cultivating the city: Mapping and characterizing urban agriculture with satellite imagery,” Earthzine (www., IEEE Committee on Earth Observation, November 26, 2009.

[12] J. Dinis, A. Navarro, F. Soares, T. Santos, S. Freire, and A. Fonseca, “Hierarchical object-based classification of dense urban areas by integrating high spatial resolution satellite images and LIDAR elevation data,” in Proc. GEOBIA – GEOgraphic Object-Based Image Analysis, 2010.

[13] T. Santos, S. Freire, A. Navarro, F. Soares, J. Dinis, N. Afonso, A. Fonseca, and J.A. Tenedório, “Extracting buildings in the city of Lisbon using QuickBird images and LIDAR data,” in Proc. of GEOBIA – GEOgraphic Object-Based Image Analysis, 2010.

[14] T. Santos, N. Gomes, M. Brito, S. Freire, A. Fonseca, and J.A. Tenedório, “Solar potential analysis in Lisbon using LiDAR data,” in Proc. of 31st EARSeL Symposium 2011, 2011.

[15] T. Santos, S. Freire, J.A. Tenedório, and A. Fonseca, “Using satellite imagery to develop a detailed and updated map of imperviousness to improve flood risk management in the city of Lisbon”. Computiational Vision and Medical Image Processing, Tavares & Natal Jorge (eds), Taylor&Francis Group, London.

[16] T. Santos, S. Freire, A. Fonseca, and J.A. Tenedório, “Producing a building change map for urban management purposes,” in Proc. of 30th EARSeL Symposium 2010, 2010.

[17] T. Esch, C. Conrad, G. Schorcht, M. Thiel, T. Wehrmann, and S. Dech, “Model-based estimation of impervious surface by application of support vector machines,” in Proc. of the International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences ISPRS, Part B8, 2008.

[18] M.K. Ridd, “Exploring a V-I-S (Vegetation-Imperious Surface-Soil) model or urban ecosystem analysis through remote sensing: Comparative anatomy of cities,” International Journal of Remote Sensing, 16: 2165–2185, 1995.

[19] T. Santos, S. Freire, A.M. Rodrigues, N. Gomes, M. Brito, and J.A. Tenedório, “Creating Web-based solar maps: Mapping applications based on LiDAR data,” in Proc. of 7th Virtual Cities and Territories, 2011.


This work was conducted in the framework of project GeoSat – Methodologies to extract large scale GEOgraphical information from very high resolution SATellite images, funded by the Portuguese Foundation for Science and Technology (PTDC/GEO/64826/2006). The authors would like to thank Logica for the opportunity of using the LiDAR dataset.

Teresa Santos is a doctoral researcher at e-GEO, Centre of Geographical and Regional Planning Studies, New University of Lisbon, Portugal. She holds a Masters in GIS from the Technical University of Lisbon (Portugal), and recently completed her PhD in Geography. Her current research activities include urban feature extraction from very high resolution satellite imagery and LiDAR data.

Sérgio Freire is a geographer working as a research assistant at e-GEO, Centre of Geographical and Regional Planning Studies, New University of Lisbon, Portugal. His current research activities include feature extraction from very high resolution satellite imagery, modeling of disaggregated spatio-temporal population distribution, and risk mapping and assessment.

J. A. Tenedório is associate professor at the Dept. of Geography and Regional Planning, New University of Lisbon, Portugal. His research interests are in urban remote sensing, urbanization, and geographic information systems.

Ana Fonseca is a geographical engineer and senior research officer at the National Laboratory of Civil Engineering. Her research interests are urban remote sensing, remote sensing for emergency management, RADAR remote sensing and INSAR.

Nuno Afonso is a geographical engineer, with a Masters in GIS from the Technical University of Lisbon (Portugal). He participates in the research and development of technological projects in remote sensing and GIS domains.

Ana Navarro is an assistant professor in the Department of Geographical Engineering, Geophysics and Energy at the University of Lisbon. She holds a diploma in Geographical Engineering from the University of Lisbon (1995), a MSc degree in Geographical Information Systems from the Technical University of Lisbon (1999) and a PhD in Geographical Engineering and Geoinformatics (Geodesy) from the University of Lisbon (2006). Her main areas of interest are the estimation of crustal deformation from GPS observations, the establishment of relationships between volcanic structures and active tectonics from remote sensing data, land use/cover classification from remote sensing data, and seabottom characterization from backscatter data.

Fernando Soares is an assistant professor at the Department of Geographical Engineering, Geophysics and Energy at the University of Lisbon. He holds a diploma in Geographical Engineering from the University of Lisbon (1994), a MSc degree in Geo-resources from the Technical University of Lisbon (1998) and a PhD in Sciences of Engineering from the Technical University of Lisbon (2006). His main area of interest is morphological image processing applied to optical and SAR remote sensing imagery aiming at the extraction of land features.