Applications of solar mapping in the urban environment

lable at ScienceDirect

Applied Geography 51 (2014) 48e57

Contents lists avai

Applied Geography

journal homepage: www.elsevier .com/locate/apgeog

Applications of solar mapping in the urban environment

T. Santos a,*, N. Gomes b, S. Freire a, M.C. Brito c, L. Santos b, J.A. Tenedório a

a e-GEO e Research Centre for Geography and Regional Planning, Faculdade de Ciências Sociais e Humanas (FCSH), Universidade Nova de Lisboa,Avenida de Berna, 26 C, 1069-061, Lisboa, Portugalb INþ, Center for Innovation, Technology and Policy Research, Instituto Superior Técnico, Oeiras, Portugalc IDL, Faculdade de Ciências da Universidade de Lisboa, Campo Grande, 1749-016, Lisboa, Portugal

Keywords:Photovoltaic potentialLiDARGISSolar radiationUrban

* Corresponding author.E-mail addresses: [email protected] (T. Sant

pt (N. Gomes), [email protected] (S. Freire), [email protected] (L. Santos), [email protected]

http://dx.doi.org/10.1016/j.apgeog.2014.03.0080143-6228/� 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

Promoting the use of solar energy in urban environments requires knowing the geographical distributionand characteristics of the best places to implement solar systems. In this context, buildings can be used tolocally generate electricity. Based on remote sensing data, the city’s surface can be modeled and the solarincome at each location can be estimated. The present study assesses photovoltaic (PV) potential ofresidential buildings. Two variables are modeled, the income of solar energy at the roof tops, and thepopulation at the building level. The model is tested in Lisbon, Portugal, using Geographic InformationSystems (GIS) based solar models and Light Detection and Ranging (LiDAR) data. The total PV potential isassessed and compared with the local electricity demand. The results constitute an initial assessment ofthe city’s solar potential that can be used to support management decisions regarding investments insolar systems.

� 2014 Elsevier Ltd. All rights reserved.

Introduction planning energy strategies, in line with sustainable development

Local governments are responsible for applying, in their cities,strategic guidelines to improve energy efficiency based on renew-able sources. In order to make informed policy decisions, officialsneed a higher level of detail than the standard numerical analysis(Lamie, Campbell, & Molnar, 2012). Geographic-based approachesare necessary to determine the most suitable locations, to estimaterevenues and expenses. In this context, remote sensing technolo-gies can be an effective source of updated geo-information aboutthe urban environment. Through the generation of energy demandand supply scenarios for the city, urban planners and city officialscan obtain accurate assessments and decide upon realistic sus-tainable goals.

Solar energy is one of the best renewable energy sources, havingthe least negative impacts on the environment (Solangi, Islam,Saidur, Rahim, & Fayaz, 2011). Assessing the city’s solar energypotential through solar mapping, constitutes a valuable analyticaltool that permits quantifying local capabilities for energy produc-tion and use those findings for designing and implementing urban

os), [email protected]@fc.ul.pt (M.C. Brito), luis.f.l.pt (J.A. Tenedório).

goals and aims.Furthermore, solar maps can be updated on a regular basis and

thus used for monitoring effects of policy application. Energy pro-duction and replacement of fossil fuels by renewable sources, alongwith energy savings on the demanding side (Lund, 2007), consti-tute the basis for sustainable energy policies that are concernedwith reducing dependence on those fuels, thus gaining in envi-ronmental benefits. Such policies can include new legislation andincentives to investment. Knowing the installed capacity for solarenergy generation, as well as the geographical distribution andcharacteristics of the best places for implementing solar systems,can lead to effective expansion of distributed generation ofrenewable energy in the city.

The residential sector plays an important role in cities’ elec-tricity consumption. In Lisbon, capital city of Portugal, about a thirdof the electric intake is absorbed by the residential sector (Ferreira& Pinheiro, 2011). Solar technology in Portugal is already beingvalued via the implementation of European Union Directives(Directive 2002/91/EC). This new awareness, associated with thefact that Portugal is one of the European countries with the highestlevels of annual solar irradiation (�Súri, Huld, Dunlop, & Ossenbrink,2007), contributes to a growing interest in the quantification ofenergy-based indicators at the city and building’s scale, in order toassess photovoltaic (PV) conversion and thermal solar potential. Inorder to propose technical and financial solutions, data on the solarpower generating potential of the city is required.

T. Santos et al. / Applied Geography 51 (2014) 48e57 49

Planning for energy investments on solar systems requires in-formation on 1) the electric power demand, and 2) on the gener-ation capabilities. Methodologies for energy planning have beenaddressed in the literature (Cellura, Di Gangi, Longo, & Orioli, 2012;Gadsden, Rylatt, Lomas, & Robinson, 2003; Izquierdo, Rodrigues, &Fueyo, 2008; Mourmouris & Potolias, 2013; Ordóñez, Jadraque,Alegre, & Martínez, 2010; Wiginton, Nguyen, & Pearce, 2010).These studies deal with energy consumption models and thecontribution of solar systems for fulfilling that demand. However,either the built structure was not modeled in detail (e.g., Izquierdoet al., 2008; Mourmouris & Potolias, 2013; Wiginton et al., 2010), orit wasmade through the use of statistical data, or 2Dmaps obtainedfrom imagery analysis (e.g., Cellura et al., 2012; Ordóñez et al.,2010). Consequently, the incoming solar irradiation at the rooftopwas not very accurately estimated.

As far as the generation capabilities of the solar systems, it isimportant to highlight that to adopt solar technologies, detailedsolar suitability information on every building in a communityshould be available for urban planners (Santos et al., 2011). Iden-tifying buildings that are suitable for solar panel installation re-quires modeling 1) the built environment, 2) the solar irradiationand, 3) the available area at the rooftops for panels’ installation. Inthe following paragraphs each of these issues is discussed in detail.

The accuracy of urban solar mapping depends, among others, onthe quality of the 3D city model. Modeling urban environments canbe a difficult task, and constitutes one of the active research topicsin Geography. In the last decades, photogrammetric techniquesallowed the production of precise 2Dmaps. But the ability to obtain3D information over large areas was limited. However, in order toperform accurate estimations of solar radiation, information onroof structure is essential. Several methods for data acquisition arecommonly used in this context, such as imagematching algorithms,image segmentation or integration of different data sources (e.g.,Bergamasco & Asinari, 2011; Izquierdo et al., 2008; Wiginton et al.,2010). Aerial LiDAR currently offers the possibility of producingdetailed and accurate 3D city models (Haala & Kada, 2010;Kaartinen et al., 2005; Moreira, Nex, Agugiaro, Remondinoc, &Lima, 2013; Ruiz-Arias, Tovar-Pescador, Pozo-Vázquez, & Alsa-mamra, 2009). Consequently, airborne LiDAR has become an ac-curate, cost-effective alternative to conventional technologies forthe creation of altimetric data at vertical accuracies that range from0.15 to 1m (Gamba & Houshmand, 2002; Hill et al., 2000; Kaartinen& Hyyppä, 2006). There are several approaches to interpolate andconstruct a 3D urban surfacemodel (incorporating the relief), basedon LiDAR and GIS buildings data (e.g., Brédif, Tournaire, Vallet, &Champion, 2013; Carneiro, Morello, & Desthieux, 2009;Rottensteiner, Trinder, Clode, & Kubik, 2003; Santos et al., 2011;Tack, Buyuksalih, & Goossens, 2012; Vögtle, Steinle, & Tóvári, 2005).

The incident solar radiation can be measured by ground-basedmeteorological stations or meteorological satellites and/or beestimated through models. There are several solar models availablein the literature. They vary in the detail of the input parameters and,consequently, in the output map. Solar Analyst (Fu & Rich, 1999)and Photovoltaic Geographical Information System (PVGIS) (�Súri,Huld, & Dunlop, 2005) are two examples of solar radiation models.

Manipulating the solar resources at the building level within aGIS is a straightforward way of identifying appropriate roof areasfor panel installation. Applying algorithms to automatically classifyand segment data, enables analyzing buildings’ roofs according totheir slope, azimuth, and shaded areas (Santos, 2011). Knowing theamount of incident solar radiance and the optimal roof areas forcapturing that energy, the solar potential of any roof plane can beeasily calculated (e.g., Brito, Gomes, Santos, & Tenedório, 2012;Hofierka & Ka�nuk, 2009; Jochem, Hofle, Hollaus, & Rutzinger,2009; Kassner, Koppe, Schüttenberg, & Bareth, 2008; Kodysh,

Omitaomu, Bhaduri, & Neish, 2013; Rottensteiner, Trinder, Clode,& Kubik, 2005). This analysis has also been extended to includesolar potential of facades and other vertical surfaces (Redweik,Catita, & Brito, 2013).

In the present study, we analyze the rooftops’ solar potentialcontribution for the residential electricity demand at the buildinglevel, thus linking technical solar capabilities with user needs. Theroofs’ PV generation capability is tested using a 3D model of thebuildings based on LiDAR data. This approach allows for a detailedestimation of the energy generation through PV since the shadowscast by surrounding buildings and topography, are calculated foreach building in the area. Furthermore, the methodology is basedon the projected roof area (retrieved from LiDAR data) and localatmospheric effects (obtained from PVGIS). The results are pre-sented in two rooftop exploration scenarios - considering the in-vestment in PV panels for fulfilling electricity demand at thebuilding level and, alternatively, by renting the roof space to thirdparties and receiving benefits from it.

It is pertinent to clarify that, according to the literature (e.g.,Izquierdo et al., 2008), solar potential may be characterized indifferent hierarchal classes: the total amount of energy receivedfrom the sun is the physical potential, while the geographic po-tential is restricted to the locations where this energy can becaptured; the technical solar potential takes into account thetechnical characteristics of the equipment, including its perfor-mance and losses in order to determine the power generated by thePVmodules. This workmainly focus on the technical solar potentialsince it takes into account the effect of the mutual shades betweenbuildings as well as the performance of the modules themselves,via PV conversion efficiency, including the minimum area restric-tion for viable roof top PV systems.

Methodology

The methodology is a two-stage process. Firstly, PV generationpotential of the rooftops is calculated based on the solar incidenceon each roof and the available area. Then, the resident population ineach building is estimated. Based on the combination of theseoutputs, twomodels of rooftop PV exploitation, for energy planningpurposes, are analyzed in a study area.

Study area and data set

The methodology is implemented in Alvalade, a parish locatedin the city of Lisbon, the capital of Portugal (Fig. 1). The parish in-cludes part of the Alvalade neighborhood, a planned zone from the1940se50s, characterized by a modern morphology, with residen-tial areas, avenues, squares and school facilities, designed to pro-mote pedestrian movement. The area includes mainly residentialfive-story buildings with commercial services in the lower floors,green areas (in the web version) and public buildings. According tothe 2001 Census (INE, 2001), 9620 people live in this parish.

Within the area, a total of 811 buildings were identified in themunicipal cartography. The average footprint is 221.6 m2.

For solar mapping, a spatial database including planimetric andaltimetric data was used (Fig. 2). The planimetric informationcomprises three sets: the building footprints, the census blockgroups, and the land use map. The building footprints were ob-tained from the Lisbon’s Municipal Cartography from 1998, at1:1000 scale, and updated for the year of the LiDAR flight. Thecensus block groups in Portugal are supported by cartography andare available in vector format for every decade. The block group isthe finest spatial unit for which the decadal Census reports popu-lation and housing data. Land use information was provided by theUrban Atlas, which maps land use for selected European cities,

Fig. 1. Study area for solar potential analysis in Lisbon.

Fig. 2. Spatial layers from the planimetric (A, B) and altimetric (C) dataset.

T. Santos et al. / Applied Geography 51 (2014) 48e5750

T. Santos et al. / Applied Geography 51 (2014) 48e57 51

including Lisbon, using 19 thematic classes and has a minimummapping unit of 0.25 ha (for ‘Artificial surfaces’) (EEA, 2013).

The altimetric data is derived from two sources: a LiDAR pointcloud, and digital cartography. From a flight with a LiDAR cameraperformed in 2006, a surface imagewas produced based on the 2ndreturn. This raster has a spatial resolution of 1m, and represents theDigital Surface Model (DSM) of the area. A Digital Terrain Model(DTM) for the city of Lisbon was produced from a set of elevationmass points and contours available in municipal cartography. Then,the normalized Digital Surface Model (nDSM) was obtained bysubtracting the DTM from the DSM image. This raster file stores theheight of all elements above the terrain.

Evaluating PV generation capacity

The goal is to evaluate the rooftop area suitable for installation ofsolar energy systems and select the optimal location for solar photo-voltaic systems. Therefore, when assessing the electrical supply of PVinstalled in residential rooftops, the following variables must be

Fig. 3. Workflow for sola

considered: the available solar radiation, the technical characteristicsof the solar panels, and the best places for installation (Fig. 3).

Identifying buildings that are suitable for solar panel installationrequiresmodeling the solar radiation incident in each location. Twoinputs are required: a DSM and the buildings’ footprints.

ArcGIS‘ Solar Analyst extension (ESRI) is suitable for local-scaleapplications. It allows characterizing the physical characteristicsof the study area, regarding insolation, using a surface model. Onthe other hand, PVGIS operates on a lower scale, and accounts forsky obstruction (shadowing) by local terrain features (hills ormountains), based on a digital elevationmodel (USGS Shuttle RadarTopography Mission). Booth tools are used to obtain the solar ra-diation over the study area.

The Area Solar Radiation tool, available in Solar Analyst exten-sion, derives the total amount of incoming solar radiation(direct þ diffuse) calculated for each location of the input rastersurface. The diffuse proportion is the fraction of global normal ra-diation flux that is diffuse. The transmissivity is the ratio of solarradiation outside the atmosphere to that reaching the Earth’s

r potential analysis.

T. Santos et al. / Applied Geography 51 (2014) 48e5752

surface. Taking into account a user specifiedmodel (a DTM or, moredesirable, a DSM), and further parameters such as transmissivityand diffuse radiation fraction, among others, the model accountsfor atmospheric effects, as well as site latitude and elevation,steepness (slope) and compass direction (aspect), daily and sea-sonal shifts of the Sun angle, and effects of shadows cast by sur-rounding topography (Fu & Rich, 1999).

The model requires as input, among others, the local annualaverage of beam and diffuse irradiation. However, the annual valuescalculated using the default values differ from the ones produced byPVGIS. PVGIS only produces daily values for the city scale, based onsurface model with 1 km resolution, derived from the USGS SRTMdata. Nevertheless, its estimations are reliable and were validatedwith values from local meteorological stations (JRC, 2010). Therefore,the reference values for radiation energy were estimated using thePVGIS model at Lisbon‘s latitude. To calibrate the solar parameters inArcGIS in order to obtain energy values closer to the ones available atPVGIS, a model of a standard building with 0� (terrace), 34� (roof)and 90� (façades) was created and used in ArcGIS (Gomes, 2011). Thediffusion and transmission where assessed on a monthly basis andthe model was applied for each month separately (Santos, 2011).Then, all 12 maps were summed up and the annual solar radiation atthe surface, in Wh/m2, was calculated. This approach reduces themodel monthly insolation variation with respect to the PVGIS data-base from about 20% to less than 1% (Gomes, 2011).

Therefore, by inputting the local DSM, the tool, after parameteri-zation, produces a solar map that accounts for local topographic in-fluences on solar radiation over the study area. This aspect isparticularly important in urban areas, where shadowing effects arevery common.

The specifications of the equipment used to convert the solarresource into photovoltaic energy are variables that contribute forthe technical potential of these systems. Two variables areconsidered: PV modules’ area and efficiency. The roof area for PVinstallation must be estimated. Due to the minimum requirementsfor economically viable solar system sizing, only areas above 24 m2

per rooftop are investigated. In order to proceed with the annualquantification of irradiation of each building it is necessary tocorrect the coverage area of the buildings considering their incli-nation, since a cell with some inclination has a higher areacompared to the same cell with no inclination (Fig. 4).

The solar radiation incident on the roofs of the buildings cannowbe determined using the actual coverage area thus dividing thevalue of this radiation on each cell by the cosine of the inclination.The sum of the values obtained from the cells in each building givesthe corrected value of solar irradiation by building (SIB):

Fig. 4. Schema of area calculation from the slope of the roof of the building.

SIB ¼Xn �

SIP�� x ��1

�

cos i

where SIP is Solar irradiation per cell, x corresponds to the spatialresolution of each cell with no inclination (1 m in the presentstudy), and cos i indicates the cosine of inclination of each cell.

The conversion efficiency of the PV systems, including PVmodule efficiency and system losses is estimated at 14% (�Súri et al.,2007). Furthermore, buildings with roofs with low solar radiation(less than 0.8 MWh/m2/year) will not be considered for analysis.Applying these constraints in the layer with the annual solar ra-diation available at each rooftop, the map with the energy pro-duced by PV panel is estimated.

Using a layer with the residential buildings footprints, and theroof’s solar characteristics calculated in the previous step, the solarpotential of each roof is calculated.

Estimating the residential population by building

The population distribution at the building level is used to es-timate two features. On one hand, the electricity demand of theresidential sector will be assessed using the local average electricitydemand per capita. On the other hand, the estimated number ofresidents will be used to characterize the type of building owner-ship, based on the average family size.

To model the electricity demand at the building level, severalmethodologies can be used. Swan and Ugursal (2009), whenreviewing modeling techniques for describing residential sector en-ergy demand, distinguished two approaches: top-down and bottom-up.The top-downapproachmodels the residential sectorasanenergysink, and no concern is given to the individual end-uses. The bottom-up approach, on the otherhand, extrapolates energy consumptions ofrepresentative individual houses to regional or national levels.

The methodology proposed is a function of data availability: 1)census population at the block group level, 2) building footprints invector format, and 3) land use information.

The population estimation approach is based on dasymetricmapping with areal interpolation. Dasymetric mapping is a carto-graphic technique that allows limiting the distribution of a variableto the zones where it is present by using related ancillary infor-mation in the process of areal interpolationxe “areal interpolation”(Eicher & Brewer, 2001).

Here the goal is to disaggregate the total resident populationfrom each census block group (source zones) to respective build-ings with residential use (target zones), using ‘residential volume’as the proxy variable. First, the Urban Atlas is used to characterizethe use of each building in the Municipal map as ‘Residential’ or‘Non-residential’. Then, the nDSM is used to impose an additionalrestrictive criterion: only buildings with at least one floor aboveground (i.e., mean height � 2.6 m) are deemed suitable for habi-tation and to receive resident population. Finally, this set ofbuildings receive a proportional (linear) share of the residentpopulation of the respective block group based on the ratio of theirvolume to the total resident volume in the block group. Estimationof residents per building is rounded to integer values.

Results

Solar potential at the rooftops

Following the workflow (Fig. 3), the solar incomewas calculatedfor all residential buildings in the study area (Fig. 5). The buildingswere classified into quartiles according to their PV potential.

Fig. 5. Electricity produced by PV panels at the residential rooftops.

T. Santos et al. / Applied Geography 51 (2014) 48e57 53

Note that artifacts like roof overhangs, chimneys, dormers, an-tennas, are not taken into account by our methodology. To be in-tegrated in this analysis, such identification requires more detailed3D data or additional spectral information.

Residential population by building

The adopted population modeling approach allowed estimatingthe residential population for each building.

Based on information from the Urban Atlas, 806 buildings wereclassified as ‘Residential’, while only 5 had other uses (‘Non-resi-dential’). When the size criterion was imposed, only 787 buildingswere deemed suitable for habitation and, thus, targets for residentialpopulation disaggregation. Of these, 760 received at least one resi-dent, with the maximum value being 137 residents (in one building)(Fig. 6). The mean estimated residential population by building was12.2, for a total of 9628 residents. This figure differs from the censustotal of 9620 residents due to the effects of rounding.

PV Potential and electricity demand of residential buildings

Rooftops solar potential and electricity demandConsidering the local average electricity demand per capita,

which for the Lisbon municipality is 1300 kWh/person/year(PORDATA, 2013), and the population distribution calculated in the

Fig. 6. Residential suitability of build

previous section, the electricity demand of residential buildings canbe estimated. By direct comparison with the PV potential (Fig. 5),the fraction of the electricity demand that can be satisfied by PVsystems can be determined - the electricity solar fraction (PV po-tential/electricity consumption) (Fig. 7). In the study area, the PVpotential corresponds to about ¼ of the total electricity demand.

Rooftops solar potential and type of buildingAmore detailed analysis regarding the relationship between the

solar fraction and the type of buildings highlights the distributionof buildings with high PV potential in the area. This analysis can beassessed through correlations between solar fraction and each ofthe following variables: volume, floor area ratio (FAR) and height.However, for this purpose, we consider that buildings with similarvolumes could have different typologies. Moreover, as FAR is lessheterogeneous in certain areas of the city, it implies that buildingswith similar areas but different heights could correspond todifferent typologies. This is explained by the gross floor area, whichtakes into account the entire area within the perimeter of theexterior walls of the building (Cheng, 2009). Therefore we choosethe building height as the variable that best describes the typol-ogies of the buildings, because it allows to more accuratelydistinguish between single family and multifamily buildings.Regarding the correlation mentioned above, Fig. 8 plots the solarfraction as a function of the height of the buildings.

ings and respective population.

Fig. 7. Solar fraction by residential building in the study area.

Fig. 8. Scatter plot of solar fraction with exponential fit and plot with groupings.

T. Santos et al. / Applied Geography 51 (2014) 48e5754

The data is grouped in three different sets of buildings, ac-cording to their height. Notice that buildings in the first quartile(Group 1, below 10 m, typically less than three-storey buildings)feature solar fractions above 100%, hence potentially generatingmore solar electricity than they consume, while buildings in groups2 and 3 are net electricity consumers. This was to be expected as, fora given footprint, taller buildings will have larger volumes henceare assumed to havemore residents and therefore higher electricitydemand. Table 1 summarizes the descriptive statistics for building

Table 1Descriptive statistics of building height and solar fraction.

Mean Variance Max. Min. 1stQuartile

2ndQuartile

3rdQuartile

Height ofbuildings(meters)

15.837 74.278 59 2 10 15 18

Solar fraction 0.661 0.606 3.885 0.027 0.198 0.317 0.698

height and for solar fraction, Table 2 the electricity demand and PVpotential of the different groups.

These results show that 25% of the buildings (Group 1) haveabout 25% of the solar potential but only 12% of the electricity de-mand. These buildings, with an average solar fraction of about 75%,are the best candidates for net-metering schemes, which areknown to be more popular among households achieving solarfractions closer to unity (Wallenborn, 2013).

The following model is estimated, reflecting the groupingsdefined above:

ln�solar_frac

�¼ b0 þ

X4 �b1;j � build_high � Dj

�þ vi; i

i

¼ 1; :::;N; j ¼ 1; :::; 4

where b1,j is the grouping specific effect of the height of thebuildings over the solar fraction.

Not surprisingly, the estimated results, described in Table 3,corroborate the idea of heterogeneity in our data. In other words,the estimated coefficients do not share the same sign acrossdifferent clusters: positive for Group 1 (single family buildings) andnegative for Group 2 and 3 (multifamily buildings). Also, in terms ofintensity, the estimated coefficient for the multifamily building,with possible physical influences is the highest, in absolute value,compared to the remaining groupings, which reflects the associ-ated inefficiencies.

Regarding the results described above, three hypotheses thatmight clarify if these estimated effects are somewhat affected bythe relative proximity of other type of buildings, are tested.We nowsummarize the hypothesis to test:

i) H1;20 : b1;1 ¼ b1;2

ii) H1;30 : b1;1 ¼ b1;3

iii) H2;30 : b1;2 ¼ b1;3

Table 2Energy production and consumption in the different groupings.

Production(MWh/year)

Consumption(MWh/year)

Average solarfraction

Number ofbuildings

Group 1 797 1064 0.749 179Group 2 1644 5970 0.275 353Group 3 761 5824 0.131 161

Table 3OLS regression with robust standard errors, groupings.

build_heightGroup 1

build_heightGroup 2

build_heightGroup 3

Constant R squared F test

ln(solar_frac) 0.046** �0.072*** �0.06*** �0.09 0.625 0.00(0.02) (0.009) (0.005) (0.133)

Robust standard errors in parentheses ***p < 0.01, **p < 0.05, *p < 0.1.

T. Santos et al. / Applied Geography 51 (2014) 48e57 55

and test statistic is:

F ¼ SSR* � SSRSSR

� n� kp

wFðp;n�kÞ

where SSR* is the sum squared residuals for the restricted model(under the null hypothesis), SSR is the sum squared residuals for theunrestrictedmodel, K is the number of regressors and p the numberof restrictions (one, for these cases).

The test results, presented in Table 4, are straightforward: theestimated coefficients for each grouping are all pairwise statisti-cally different.

Notwithstanding, the result for the test deserves some atten-tion. At 99% confidence level, the estimated effect for the multi-family buildings is statistically equal, so it might be reasonable totake into consideration that the relative proximity of such buildingsis not sufficiently relevant to determine bigger differences on theestimated effects. The same is not valid for the single familybuildings, whose solar fraction is likely to be affected by theirrelative position to higher buildings.

Rooftops solar potential and building ownershipAnother interesting result from themapping of the rooftop solar

potential arises when these data are crossed against the type ofbuilding ownership. According to INE (2011), the average familysize is approximately 3 individuals. This number is used to distin-guish buildingwith one owner, frombuildings with several owners.The results are summarized in Table 5. It is clear that, for the studyarea, a very significant share (75%) of the total solar potential ofresidential buildings is located in apartment blocks’ condominiums(i.e., buildings with multiple owners). Notice that in Portuguesecities this type of building ownership is quite common (INE, 2011)and therefore one can expect this situation to occur in the city as awhole, and in other urban areas.

These results lead to relevant implications regarding solar energyincentives policies. In fact and since its inception, the feed in tariff

Table 4F tests.

Null hypothesis Test statistic P-Value

H1;20 : b1;1 ¼ b1;2 84.36 0.000

H1;30 : b1;1 ¼ b1;3 42.27 0.000

H2;30 : b1;2 ¼ b1;3 5.28 0.022

Table 5Planning scenarios for PV implementation.

Planning options Buildings Population PV generationMWh/year

PV coverageper building(%)

All residential buildings 693a 9420 3202 23Buildings with multiple

owners514 8790 2405 17

Buildings with one owner 179 630 797 42

a 67 Residential buildings were excluded for not meeting the previously specifiedparameters.

scheme in Portugal has been overwhelmingly used by homeownersof single-family buildings, although the Law already allows for theincentive to be available for buildings with multiple owners (withdifferent eligibility criteria including the completion of an energyaudit to the common areas). The reasons that have beenput forwardto explain the low popularity of these incentives for condominiumsare mainly administrative barriers (since condominiums cannotrequest bank loans, and therefore individual owners would have torequest individual loans for a common purpose, which is notparticularly welcomed by lending institutions) and organizationalissues (since all the owners need to agreewith the investment). Theresults shown inTable 5 highlight that the large-scale disseminationof photovoltaic energy in the urban landscape requires the outreachto multiple-ownership buildings’ rooftops. This could be achievedby streamlining the procedures for the application to feed in tariffincentives by condominiums (including facilitating borrowingconditions for this type of investments) or, facilitating the lease ofthe commonly owned roof space to third parties.

Conclusions

Urban planners require detailed geo-referenced consistentspace-time data base on urban performances in order to performforecasting and policy analysis (Kourtit, Nijkamp, & Reid, 2014). Themethodology presented in this paper constitutes the first step to-wards the creation of such data base. Using remote sensing data asan alternative source of geospatial information for solar mappingconstitutes a fast and efficient approach to support and assist thedesign of renewable energy policies.

The motivation for this case study lies in the fact that incorpo-rating solar systems into buildings offers a mean to locally generateelectricity, based on a clean and renewable decentralized source ofenergy. In these circumstances, the use of LiDAR data can play animportant rolewhen analyzing the buildings capability for receivingthe solar systems.

The model results have shown that, for the study area, about25% of the local electricity demand in residential buildings can bemet by installing PV systems on their rooftops. Furthermore, thefirst quartile of buildings (according to their heights) feature rela-tively high solar fraction potential, making them particularly suit-able for net-metering schemes, which are known to be popularamong citizens with net zero annual consumption, hence with thepotential of reaching solar fractions of the order of unity. This targetgroup has about 25% of the total solar energy potential of the areaunder study, but only 12% of the electricity demand.

On the other hand, it was also shown that most of the solarpotential in the urban landscape is located on the rooftops ofmultiple-owner buildings. This leads to the conclusion that, for thelarge scale dissemination of solar power in cities, it is important tobroaden solar energy incentive policies, such as feed in tariffs, tothese types of buildings, by either facilitating condominium in-vestment in PV systems and/or promoting third party leases of roofspace for grid integrated PV systems.

These results recommend a simple approach that produces aninitial assessment of a city’s solar potential that can be used tosupport management decisions regarding investments in solarsystems. The spatial resolution of the DSM obtained from the LiDAR

T. Santos et al. / Applied Geography 51 (2014) 48e5756

data used in this study compromised more detailed analysis. If ahigher point density LiDAR cloud is used, the derived DSM candepict elements in the roof area, and optimal locations for panelinstallation can be identified. Furthermore, the upcoming systemsthat perform simultaneous collection of LiDAR together with highresolution digital imagery in the same flight potentiate the devel-opment of higher-scale solar mapping applications.

The methodology described in this paper could be furtherdeveloped to analyze the relevance of other interesting variablessuch as the income level of the households, if data was available. Infact, recent studies have shown that there is a strong correlationbetween high income and roof top photovoltaic systems(Macintosh & Wilkinson, 2011). Crossing solar irradiation potentialwith income data could therefore be used to identify residentialareas where, in spite of a strong solar potential, it is less likely thatbuilding owners or tenants would invest in solar energy systems,thus requiring particular attention from public policies.

One other variable that could be interesting to look at in thiscontext is the building age. In Portugal, the Thermal PerformanceBuilding Regulations (RCCTE, 2006) determines that new homes (orbuildings with significant remodeling works) should include a solarthermal collector (1 m2 per resident). If available, the correlationbetween building age and the solar potential would be useful totackle questions such as the impact of extending this policy tophotovoltaic systems.

As a general conclusion, the results discussed particularly inter-esting for supporting the adoption of solar energy planning in urbanenvironments (Rylatt, Gadsden, & Lomas, 2001). For planning pur-poses, is essential to assess the solar potential. This task requiresknowledge on the location of optimal areas and this information canbe obtained through LiDAR modeling. The methodologies for datamodeling, such as the one presented here, may be the first step topromote building projects (renovation, rehabilitation and urbanregeneration) and for the definition of policies for urban planning,including: mechanisms for accessing local grants for adoption of solartechnology in order to facilitate social equity of access; implementa-tionofurbanblockswhichareemission freeorhave reduced-emissionof CO2, or promoting urban design optimized for sun exposure.

Acknowledgments

The authors would like to thank Logica for the opportunity ofusing the LiDAR data set.

This paper presents research results of : 1) Strategic Project of e-GEO (PEst-OE/SADG/UIo161/2011) Research Centre for Geographyand Regional Planning funded by the Portuguese State Budgetthrough the Fundação para a Ciência e Tecnologia; and 2) researchproject “IntegerSum e Integrated geo-referenced model for sus-tainable urban metabolism” of IST/UTL (PTDC/SEN-ENR/121747/2010), funded by the Portuguese State Budget through the Funda-ção para a Ciência e Tecnologia.

The work presented in this communication was funded by theFundação para a Ciência e Tecnologia (Grant SFRH/BPD/76893/2011).

References

Bergamasco, L., & Asinari, P. (2011). Scalable methodology for the photovoltaic solarenergy potential assessment based on available roof surface area: further im-provements by ortho-image analysis and application to Turin (Italy). Solar En-ergy, 85(11), 2741e2756. http://dx.doi.org/10.1016/j.solener.2011.08.010.

Brédif, M., Tournaire, O., Vallet, B., & Champion, N. (2013). Extracting polygonalbuilding footprints from digital surface models: a fully-automatic global opti-mization framework. ISPRS Journal of Photogrammetry and Remote Sensing, 77,57e65. http://dx.doi.org/10.1016/j.isprsjprs.2012.11.007.

Brito, M. C., Gomes, N., Santos, T., & Tenedório, J. A. (2012). Photovoltaic potential ina Lisbon suburb using LiDAR data. Solar Energy, 86(1), 283e288. http://dx.doi.org/10.1016/j.solener.2011.09.031.

Carneiro, C., Morello, E., & Desthieux, G. (2009). Assessment of solar irradiance onthe urban fabric for the production of renewable energy using LIDAR data andimage processing techniques. In P. M. Sester, P. L. Bernard, & P. V. Paelke (Eds.),Advances in GIScience (pp. 83e112). Berlin Heidelberg: Springer. Retrievedfromhttp://link.springer.com/chapter/10.1007/978-3-642-00318-9_5.

Cellura, M., Di Gangi, A., Longo, S., & Orioli, A. (2012). Photovoltaic electricity sce-nario analysis in urban contests: an Italian case study. Renewable and Sustain-able Energy Reviews, 16(4), 2041e2052. http://dx.doi.org/10.1016/j.rser.2012.01.032.

Cheng, V. (2009). Understanding density and high density. In E. Ng (Ed.), Designinghigh-density cities for social and environmental sustainability (pp. 3e16). London:Earthscan.

EEA (European Environment Agency). (2013). Urban Atlas. Retrieved from http://www.eea.europa.eu/data-and-maps/data/urban-atlas.

Eicher, C. L., & Brewer, C. A. (2001). Dasymetric mapping and areal interpolation:implementation and evaluation. Cartography and Geographic Information Sci-ence, 28(2), 125e138. http://dx.doi.org/10.1559/152304001782173727.

Ferreira, J., & Pinheiro, M. (2011). In search of better energy performance in thePortuguese buildingsdthe case of the Portuguese regulation. Energy Policy,39(12), 7666e7683. http://dx.doi.org/10.1016/j.enpol.2011.08.062.

Fu, P., & Rich, P. M. (1999). Design and implementation of the Solar Analyst: an Arc-View extension for modeling solar radiation at landscape scales. In Presented atthe IX Annual ESRI User Conference, San Diego, California. Retrieved fromhttp://proceedings.esri.com/library/userconf/proc99/proceed/papers/pap867/p867.htm.

Gadsden, S., Rylatt, M., Lomas, K., & Robinson, D. (2003). Predicting the urban solarfraction: a methodology for energy advisers and planners based on GIS. Energyand Buildings, 35(1), 37e48. http://dx.doi.org/10.1016/S0378-7788(02)00078-6.

Gamba, P., & Houshmand, B. (2002). Joint analysis of SAR, LIDAR and aerial imageryfor simultaneous extraction of land cover, DTM and 3D shape of buildings. In-ternational Journal of Remote Sensing, 23(20), 4439e4450. http://dx.doi.org/10.1080/01431160110114952.

Gomes, N. (2011). Integração de Dados Lidar com Imagens de Muito Alta ResoluçãoEspacial para Determinação de Áreas Urbanas com Potencial Solar. MSc Thesis.Portugal: New University of Lisbon.

Haala, N., & Kada, M. (2010). An update on automatic 3D building reconstruction.ISPRS Journal of Photogrammetry and Remote Sensing, 65(6), 570e580. http://dx.doi.org/10.1016/j.isprsjprs.2010.09.006.

Hill, J. M., Graham, L. A., Henry, R. J., Cotter, D. M., Ding, A., & Young, D. (2000). Widearea topographic mapping and applications using airborne light detection andranging (LIDAR) technology. Photogrammetric Engineering & Remote Sensing,66(8), 908e914.

Hofierka, J., & Ka�nuk, J. (2009). Assessment of photovoltaic potential in urban areasusing open-source solar radiation tools. Renewable Energy, 34(10), 2206e2214.http://dx.doi.org/10.1016/j.renene.2009.02.021.

INE (Instituto Nacional de Estatística). (2001). Recenseamento Geral da População eda Habitação. Lisboa. Retrieved from http://www.ine.pt/.

INE (Instituto Nacional de Estatística). (2011). Censos 2011 Resultados Definitivos ePortugal. Retrieved from http://www.ine.pt/.

Izquierdo, S., Rodrigues, M., & Fueyo, N. (2008). A method for estimating thegeographical distribution of the available roof surface area for large-scalephotovoltaic energy-potential evaluations. Solar Energy, 82(10), 929e939.http://dx.doi.org/10.1016/j.solener.2008.03.007.

Jochem, A., Hofle, B., Hollaus, M., & Rutzinger, M. (2009). Object detection inairborne LIDAR data for improved solar radiation modeling in urban areas. In ,Vol. XXXVIII, Part 3/W8. The International Archives of the Photogrammetry, RemoteSensing and Spatial Information Science. Paris: France.

JRC, Joint Research Center, EU. (2010). Solar radiation (Europe) in PVGIS. Available athttp://iamest.jrc.it/pvgis/solres/solrespvgis.htm#Comparison. assessed inDecember, 2013.

Kaartinen, H., & Hyyppä, J. (2006). Evaluation of building extraction. In , 50.EuroSDR-European Spatial Data Research Official Publication (pp. 9e77).Retrieved from http://www.eurosdr.net/publications/50.pdf.

Kaartinen, H., Hyyppä, J., Gülch, E., Vosselman, G., Hyyppä, H., Matikainen, L., et al.(2005). Accuracy of 3D City Models: EuroSDR comparison. In , Vol. XXXVI. In-ternational Archives of the Photogrammetry, Remote Sensing and Spatial Infor-mation Sciences (pp. 227e232). Eschede, the Netherlands.

Kassner, R., Koppe, W., Schüttenberg, T., & Bareth, G. (2008). Analysis of the solarpotential of roofs by using official Lidar data. In , Vol. XXXVII, Part B4. The In-ternational Archives of the Photogrammetry, Remote Sensing and Spatial Infor-mation Science. Beijing, China.

Kodysh, J. B., Omitaomu, O. A., Bhaduri, B. L., & Neish, B. S. (2013). Methodology forestimating solar potential on multiple building rooftops for photovoltaic sys-tems. Sustainable Cities and Society, 8, 31e41. http://dx.doi.org/10.1016/j.scs.2013.01.002.

Kourtit, K., Nijkamp, P., & Reid, N. (2014). The new urban world: challenges andpolicy. Applied Geography, 49, 1e3. http://dx.doi.org/10.1016/j.apgeog.2014.01.007.

Lamie, R. D., Campbell, W., & Molnar, W. (2012). The fiscal-geographic nexus: anoverview of fiscal impact assessment in local policy development. Applied Ge-ography, 32(1), 54e60. http://dx.doi.org/10.1016/j.apgeog.2010.07.007.

T. Santos et al. / Applied Geography 51 (2014) 48e57 57

Lund, H. (2007). Renewable energy strategies for sustainable development. Energy,32(6), 912e919. http://dx.doi.org/10.1016/j.energy.2006.10.017.

Macintosh, A., & Wilkinson, D. (2011). Searching for public benefits in solar sub-sidies: a case study on the Australian government’s residential photovoltaicrebate program. Energy Policy, 39(6), 3199e3209. http://dx.doi.org/10.1016/j.enpol.2011.03.007.

Moreira, J. M., Nex, F., Agugiaro, G., Remondinoc, F., & Lima, N. J. (2013). From DSM to3D building models: a quantitative evaluation, Vol. XL-1/W1. International Archivesof the Photogrammetry, Remote Sensing and Spatial Information Sciences (pp.213e219). Hannover: Germany.

Mourmouris, J. C., & Potolias, C. (2013). A multi-criteria methodology for energyplanning and developing renewable energy sources at a regional level: a casestudy Thassos, Greece. Energy Policy, 52, 522e530. http://dx.doi.org/10.1016/j.enpol.2012.09.074.

Ordóñez, J., Jadraque, E., Alegre, J., & Martínez, G. (2010). Analysis of the photo-voltaic solar energy capacity of residential rooftops in Andalusia (Spain).Renewable and Sustainable Energy Reviews, 14(7), 2122e2130. http://dx.doi.org/10.1016/j.rser.2010.01.001.

PORDATA (Base de Dados Portugal Contemporâneo). (2013). Consumo de energiaelétrica. Retrieved from http://www.pordata.pt/.

RCCTE e Regulamento das Características de Comportamento Térmico dos Edifícios.(2006). Decreto-Lei 80/2006 de 4 Abril, inserted in the transposition of the EUdirective 2002/91/CE.

Redweik, P., Catita, C., & Brito, M. (2013). Solar energy potential on roofs and facadesin an urban landscape. Solar Energy, 97, 332e341. http://dx.doi.org/10.1016/j.solener.2013.08.036.

Rottensteiner, F., Trinder, J., Clode, S., & Kubik, K. (2003). Building detection usingLIDAR data and multi-spectral images. In C. Sun, H. Talbot, S. Ourselin, &T. Adriaansen (Eds.), Seventh digital image computing: techniques and applica-tions. Sydney: Australia.

Rottensteiner, F., Trinder, J., Clode, S., & Kubik, K. K. T. (2005). Automated delineationof roof planes from LIDAR data. In , Vol. XXXVI, Part 3. Proceedings of the ISPRSWorkshop Laser scanning 2005. Enschede, The Netherlands.

Ruiz-Arias, J. A., Tovar-Pescador, J., Pozo-Vázquez, D., & Alsamamra, H. (2009).A comparative analysis of DEM-based models to estimate the solar radiation inmountainous terrain. International Journal of Geographical Information Science,23(8), 1049e1076. http://dx.doi.org/10.1080/13658810802022806.

Rylatt, M., Gadsden, S., & Lomas, K. (2001). GIS-based decision support for solarenergy planning in urban environments. Computers, Environment and UrbanSystems, 25(6), 579e603. http://dx.doi.org/10.1016/S0198-9715(00)00032-6.

Santos, T. (2011). Producing geographical information for land planning using VHRData. Local Scale Applications. LAP LAMBERT Academic Publishing GmbH &Co.KG, ISBN 978-3-8473-0460-9.

Santos, T., Freire, S., Rodrigues,A.M., Gomes,N., Brito, & Tenedório, J. A. (2011). CreatingWeb-based solar maps: mapping applications based on LiDAR data. In N. N. Pinto,J. A. Tenedório, M. Santos, & R. Deus (Eds.), Proceedings of Seventh InternationalConference on Virtual Cities and Territories (pp. 555e560). Lisbon, Portugal.

Solangi, K. H., Islam, M. R., Saidur, R., Rahim, N. A., & Fayaz, H. (2011). A review onglobal solar energy policy. Renewable and Sustainable Energy Reviews, 15(4),2149e2163. http://dx.doi.org/10.1016/j.rser.2011.01.007.

�Súri, M., Huld, T. A., & Dunlop, E. D. (2005). PV-GIS: a web-based solar radiationdatabase for the calculation of PV potential in Europe. International Journal ofSustainable Energy, 24(2), 55e67. http://dx.doi.org/10.1080/14786450512331329556.

�Súri, M., Huld, T. A., Dunlop, E. D., & Ossenbrink, H. A. (2007). Potential of solarelectricity generation in the European Union member states and candidatecountries. Solar Energy, 81(10), 1295e1305. http://dx.doi.org/10.1016/j.solener.2006.12.007.

Swan, L. G., & Ugursal, V. I. (2009). Modeling of end-use energy consumption in theresidential sector: a review of modeling techniques. Renewable and SustainableEnergy Reviews, 13(8), 1819e1835. http://dx.doi.org/10.1016/j.rser.2008.09.033.

Tack, F., Buyuksalih, G., & Goossens, R. (2012). 3D building reconstruction based ongiven ground plan information and surface models extracted from spaceborneimagery. ISPRS Journal of Photogrammetry and Remote Sensing, 67, 52e64. http://dx.doi.org/10.1016/j.isprsjprs.2011.10.003.

Vögtle, T., Steinle, E., & Tóvári, D. (2005). Airborne Laserscanning Data for deter-mination of suitable areas for photovoltaics. In , Vol. XXXVI, Part 3. Proceedings ofthe ISPRS Workshop Laser scanning 2005. Enschede, The Netherlands.

Wallenborn, G. (2013). The enrolment of consumers in smart grids. In Proceedings ofConference STS Perspectives on Energy, Lisbon, Portugal.

Wiginton, L. K., Nguyen, H. T., & Pearce, J. M. (2010). Quantifying rooftop solarphotovoltaic potential for regional renewable energy policy. Computers, Envi-ronment and Urban Systems, 34(4), 345e357. http://dx.doi.org/10.1016/j.compenvurbsys.2010.01.001.