Ecosystem Services 50 (2021) 101314 Contents lists available at ScienceDirect Ecosystem Services journal homepage: www.elsevier.com/locate/ecoser Interactive spatial planning of urban green infrastructure – Retrofitting green roofs where ecosystem services are most needed in Oslo Zander S. Venter a, David N. Barton a, Laura Martinez-Izquierdo b, Johannes Langemeyer c, d, Francesc Baró c, h, i, Timon McPhearson e, f, g a Norwegian Institute for Nature Research - NINA, Sognsveien 68, 0855 Oslo, Norway b Nabolagshager, 0191 Oslo, Norway c Institute of Environmental Science and Technology (ICTA), Universitat Autònoma de Barcelona, Spain (UAB), Edifici Z (ICTA-ICP), Carrer de les Columnes s/n, Campus de la, Spain UAB, 08193 Cerdanyola del Vallès, Spain d Department of Geography, Humboldt Universität zu Berlin, Germany e Urban Systems Lab, The New School, New York, NY, USA f Stockholm Resilience Centre, Stockholm University, Stockholm, Sweden g Cary Institute of Ecosystem Studies, Millbrook, NY, USA h Vrije Universiteit Brussel (VUB), Geography Department, Pleinlaan 2, B-1050, Brussels, Belgium i Vrije Universiteit Brussel (VUB), Sociology Department, Pleinlaan 2, B-1050, Brussels, Belgium A R T I C L E I N F O A B S T R A C T Keywords: Spatial multi-criteria decision analysis (MCDA) is increasingly being used to inform urban green infrastructure Green roofs planning. We explore the use of modern cloud computing technologies (Google Earth Engine) to facilitate public Ecosystem services access to spatial MCDA of ecosystem services from green infrastructure. Using the spatial prioritization of green Participatory mapping roof retrofitting in Oslo, Norway, as a case study, we present a web application that is a generalizable tool for Urban planning engaging stakeholders in spatial planning of ecosystem restoration and nature-based solutions. In our applica­ Green space tion, green roof designers, owners and operators identified the relative importance of a suite of potential ecosystem services (ES) gained from retrofitting of green roofs, conditional on preference profiles expected by users of different building functional types. The ES assessed included temperature regulation, stormwater runoff mitigation, habitat for biodiversity, aesthetic value, and noise reduction. In Oslo we found high spatial corre­ lation in ES deficits, implying that even large differences in stakeholder preferences for individual ES will lead MCDA to identify common interests in the spatial targeting of green roofs. Nevertheless, we found the interactive spatial MCDA web application to have potential for improving planning process efficiency in engaging stake­ holders. In more heterogenous urban landscapes, with lower spatial correlation of individual ES, spatial MCDA also has scope to improve the output efficiency of spatial targeting of nature-based solutions such as green roofs. Link to web application: https://nina.earthengine.app/view/green-roof-mcda. 1. Introduction that can deliver multiple benefits is critical, especially in the context of limited financial and other resources. Green infrastructure, precisely Green infrastructure is expanding rapidly in cities around the world because of the multiple benefits it can provide from cooling, to storm­ as a nature-based solution for meeting multiple sustainability and water absorption to sites for urban agriculture and recreation, has begun resilience goals (Andersson et al., 2019; Frantzeskaki et al., 2019; Keeler to go mainstream in policies, plans, and management strategies to et al. 2019). Green infrastructure has been described as a form of urban improve urban livability (Andersson et al. 2019). Green infrastructure ecological infrastructure Childers et al., 2019 that can provide multiple can have many components, including parks, rain gardens, community ecosystem services (ES), or benefits to human health and well-being and allotment gardens, and green roofs. (Gomez-Baggethun and Barton, 2013). Given that cities face many Green roofs, like other forms of green infrastructure, are often hybrid challenges to achieve normative goals for more sustainable and resilient green-grey infrastructure that integrate urban ecosystem components futures (Elmqvist et al. 2019), including climate driven extreme events into the built form of the city (Depietri and McPhearson 2017). Benefits (McPhillips et al., 2018; Depietri et al., 2018) and historical legacies of of green roofs can be many, but are also subject to the social, ecological, inequality (Grove et al., 2018; Locke et al., 2020), investing in solutions and technological contexts they are embedded in. A social-ecological- https://doi.org/10.1016/j.ecoser.2021.101314 Received 2 October 2020; Received in revised form 20 May 2021; Accepted 6 June 2021 Available online 15 June 2021 2212-0416/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 technological systems (SETS) framework has been used in urban ecology structure level resolution (Daniels et al., 2018; Kuller et al., 2019; Lan­ (McPhearson et al., 2016; Keeler et al., 2019) to understand that solu­ gemeyer and Connolly, 2020). The evaluation of green roof imple­ tions such as green roofs depend on many aspects of the urban envi­ mentations has recently been objective of spatial MCDA approaches ronment including: the structural integrity of building roofs, the based on the assessment of ES (cf. Langemeyer et al., 2020). These maintenance and stewardship aspects, the local and regional climate, spatially explicit MCDA approaches indicate promising pathways for the the type of green roof ecosystem designed and installed, the legal selection of optimized green roof types and optimal geographical loca­ frameworks to enable green roof installation, and many other SETS tion for their implementation. Langemeyer et al. (2020) follow the contexts that interact to impact the benefits that green roofs can provide, innovative approach to target ES needs for effective green roof imple­ including different perceptions of and preferences for benefits. mentations, considering both accessibility of benefit providing green Green roofs provide both public and privately appropriated ES in infrastructure as well as spatial distributions of social vulnerabilities. urban areas, including improved storm-water management, better Yet, the two most advanced approaches (Langemeyer et al., 2020, regulation of building temperatures and reduction of CO2 emissions, focusing on Madrid and Barcelona, Spain, respectively), which we build reduced urban heat-island effects, increased urban wildlife habitat, on in this study are limited in the level of benefits and the amount of reduced noise pollution due to the absorption of sound waves through stakeholder objectives that have been considered, and thus provide a soil and plants, enhanced community safety and quality of life, provide limited understanding of the tradeoffs and synergies between GI bene­ areas for recreation and opportunities to enjoy outdoor living spaces, fits/ES and applicability for urban green roof planners. contribution to local food production and boost the local economy This study demonstrates the implementation of spatial MCDA to creating jobs in many sectors (Susca et al., 2011; Turner et al., 2011; prioritize green roof locations at city scale with building resolution. In Whittinghill and Rowe, 2012; Berardi et al., 2014; Van der Horst et al., the Methods section, we describe the Oslo case study context, priority ES 2013). The retrofitting of green roofs on buildings that are often unused and the policy need for an application that identifies building sites can also contribute to solving challenges related to climate change and where transforming existing roof space to green roofs would make the rapid urbanization (Benedict & McMahon, 2002; Bugliarello, 2006). greatest potential contribution to priority ES in the city. Next we Multi-criteria decision analysis (MCDA) has been advocated as a tool describe a survey of green roof owners, designers and operators used to to integrate ES into green infrastructure planning (Saarikoski et al., identify potential variation in ES demand associated with different 2016). MCDA theory is rooted in operational research (e.g. Mendoza and building functional types. We describe how this data provides the basis Martins, 2006) and provides a framework that supports decision-making for importance weights attributed to ES in the MCDA. In the following in complex social-ecological systems Garmendia & Gamboa, 2012; section we describe the ES mapping data input to the MCDA. Next we Munda, 2008; Saarikoski et al., 2019; Zia et al., 2011; EF). A common describe the implementation of the MCDA in the Google Earth Engine approach for developing an ES-based MCDA framework consists of (GEE) online web application, aimed at providing public user- several consecutive steps, which have been summarized by Langemeyer customized access to the mapped ES input layers, valuation, and re­ et al. (2016) as: (a) problem definition, (b) definition of alternatives sults. In the results sections we present the spatial analysis of ES deficits, (consisting, for example, of alternative land-use options), (c) selection of and evaluate the trade-offs between different ES and the spatial priori­ ES as evaluation criteria (and corresponding indicators to assess them), tizations produced by different stakeholders in the MCDA model. We (d) scoring of criteria with regard to each alternative, (e) weighting of present interactive features of the GEE application aimed at promoting criteria (although the weighting is not necessarily made explicit), and (f) interaction with the data and results of the MCDA. Finally, we compare prioritization of alternatives through the application of an aggregation the GEE implementation of MCDA for green roofs with recent state-of- model. The latter follows different approaches to aggregate or compare the-art applications. We discuss pros and cons in relation to the the alternative, with regard to the evaluation criteria, such as un­ research application’s support for municipal planning and imple­ weighted and weighted summation (e.g. Grêt-Regamey et al., 2013), mentation of green roofs. Analytic Hierarchy Process (AHP) (Saaty, 1980), ideal point approaches, where a specific optimal target values for the criteria are defined (e.g. 2. Methods and materials Opricovic & Tzeng, 2007; Sanon et al., 2012), and pair-wise-comparison approaches (Oikonomouet al., 2011) among others. 2.1. Study area and context Integrated assessments of ES based on MCDA provide multiple entry points (a-f) for the engagement with stakeholders (Langemeyer et al., Oslo has been one of the fastest growing cities in Europe since 2000 2016, 2018). The inclusion of stakeholders in MCDA is most relevant (Oslo Kommune, 2013a, 2013b), attributed to high birth rates, intra- (and most commonly applied) for the elicitation of criteria weights national migration, and international migration. Extensive residential (Mascarenhas et al., 2016; Allain et al., 2017). Different stakeholder development is limited by the Marka peri-urban forest greenbelt. The perceptions of the relative benefits from each ES are expected to be Municipal Plan to 2030–2050 proposes residential densification around important in understanding trade-offs and conflicts of interest regarding transport nodes, and transformation of brownfields. Oslo’s street level planning priorities for green roofs. Drawing on Funtowicz & Ravetz greenview index is 28,8% (Treepedia, http://senseable.mit.edu/tr (1994) and Fish et al. (2016), participatory approaches to establish eepedia/cities/oslo). Oslo’s built zone has 47% green space cover, weights are recommended in practical application of MCDA in land-use with 60 m2 of regulated green space per inhabitant (Oslo Kommune planning. However, the elicitation of criteria weights is often limited to a 2018). Within the built zone additional surface area for green infra­ small number of stakeholders, through individual surveys or delibera­ structure at ground level is severely limited. However, Oslo’s first Green tive group exercises (e.g., Karjalainen et al., 2013; Srdjevic et al., 2013; Roofs Strategy proposal identified 14 million m2 of existing flat roof Zhang and Lu, 2010; Zia et al., 2011), with potential implications of space potentially suited for transformation to green roofs (Oslo Kom­ injustice (Langemeyer & Connolly, 2020; Venter et al., 2020c). The mune 2018). The first Green Roof Strategy identified the following consideration of a wider range of stakeholder interests - especially in the ecosystem service (ES) objectives in the city’s planning of green roofs to MCDA weighting step - promotes a more differentiated green infra­ 2030: better stormwater management, conservation of the city’s biodi­ structure decision-aid, and thus enhances the inclusiveness and rele­ versity; spaces for recreation, socialisation, learning and experience; vance of integrated green infrastructure assessments. local food production; aesthetics; temperature regulation in and be­ While ES assessments are often applied at landscape scale with rec­ tween buildings; improved air quality and noise mitigation; and CO2 ommendations on spatial targeting of GI at the level of administrative sequestration. The most recent Green Roof Strategy proposal simplifies units (e.g. García et al., 2020; Ramyar et al., 2020), applications to green the objectives of green roofs to include “nature, water, energy and roof spatial planning requires targeting with street, property and health” (Oslo Kommune 2020). 2 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 The strategy is accompanied by a new norm introduced in 2019 for building and green roof the interviewee had experience with, (2) bar­ blue green factor (BGF) scoring of building permits (Oslo Kommune riers and opportunities to transformation of existing roofs in Oslo, (3) 2019). The BGF norm scores different types of green surfaces and relative importance of private and public benefits from ES for building qualities at property level, including green roofs, in terms of their functional types, and ecosystem profiles of basic green roof designs. The relative contribution to: stormwater regulation and greenviews for ES benefit weights from part 3 were used to generate the default pref­ recreation, further indicating the city’s priority ES in implementing erence profiles in the App for the private and public benefits from green infrastructure. The BGF implements a simple spatial targeting at ecosystem services. Here, “private” refers to benefits enjoyed on prop­ city landscape level (Supplementary Material S1). Developers are erty, and “public” refers to benefits enjoyed off-property/in the neigh­ required to achieve a minimum BGF score which is lower in the city core borhood around the property with the green roof. The full interview due to space limitations, and higher in mostly residential areas between guide can be found in Supplementary Material S1 and the resulting the dense core and the peri-urban greenbelt. From its inception in 2014, MCDA weight profiles in Supplementary Material S2. The resulting Oslos BGF has undergone simplifications in response to property de­ default preference profiles for ecosystem services from the different velopers’ claim that criteria and scoring are too complicated to building functional types in the App are shown in Table 1. implement. Based on the above analysis of municipal strategies, plans and norms, 2.3. Ecosystem service criteria, scaling and deficit mapping in MCDA we identified a need for an assessment tool for green roofs that is suf­ ficiently detailed to differentiate properties’ ES potential, while acces­ The spatial prioritization of green roofs was based on a multi-criteria sible enough to understand the input data, and model function. Between decision approach where criteria are defined by a number of GIS map­ (i) the city wide zoning of residences around transport nodes, (ii) the ping layers (Table 2) that identify urban ecosystem conditions at broad two-zone differentiation in minimum BGF requirements and (iii) building level resolution across the city. Layers identify ecosystem the detailed design of blue- green infrastructure at property level, there condition in service areas around each roof as proxy indicators of the is potential for ‘meso-level’ spatial targeting of BGI that can differentiate regulating, provisioning, cultural and supporting service deficit and priorities at neighbourhood and even street level. The Green roof MCDA potential of green roofs (Fig. 1). Low ecosystem condition areas are App is designed to fill this ‘planning space’, while raising public defined as ES deficit areas (Langemeyer et al. 2020). These are awareness and demand for BGI through the visualisation of ES supply normalized by census district population density to identify areas with across the cityscape. the highest population weighted exposure to ES deficit. Population density has been used a proxy for potential demand (Vallecillo et al., 2.2. Stakeholder ecosystem service preferences 2018). The MCDA assumes a utility function in municipal green roof planning where buildings in areas with highest population weighted ES The App allows users to specify their personal ES preference weights, deficit are priorities for transformation to green roofs. We used the and to explore the implications of different weightings on spatial tar­ Google Earth Engine (GEE) JavaScript API (Gorelick et al., 2017) to geting of green roofs in Oslo. In order to provide users with a starting generate the mapping layers and implement the MCDA. GEE is a cloud- point for this exploration, we elicited ES weights from 12 green roof based platform for geospatial analysis that leverages Google’s compu­ owners, managers and designers. ES weights were obtained in an in- tational infrastructure to make GIS and remote sensing analyses more depth interview lasting 1–1.5 hours in May-July 2019. Interviewees efficient and scalable. were selected based on their experience with different building func­ We used a map layer of about 53,000 roofs (>10 m2 surface area and tional types. Building functional types have different occupants with < 30 degree slope) identified in the Oslo Green Roof Strategy as feasible different preferences, and we selected interviewees to reflect these dif­ roof area for green roof installation (Oslo Kommune 2018). Following ferences. The building functional types represented included urban the logic of a spatial MCDA (Langemeyer et al. 2020), each roof is an farming, shopping mall, private residence, housing association, offices, alternative subject to a targeting decision. community centre, protected/historic, theatre, hospital and hotel. Land surface temperature data was used to map the distribution of The interview was divided into three parts addressing (1) the type of heat over the city which defines areas in need of green infrastructure to Table 1 Ecosystem service importance weights by stakeholders building functional types. *Note: on private property ecosystem services have the same default importance weights (see Supplementary Material S1). (CICES v5.1 codes for each ecosystem service https://cices.eu/resources/. 3 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 Table 2 GIS mapping layers of ecosystem condition used in the MCDA of green roof targeting in Oslo. Related ecosystem service (ES) deficit categories are highlighted in color including regulating (R), provisioning (P), cultural (C), supporting (S) services. All layers are scaled by population density to represent relative population exposure to ES deficits. Fig. 1. Scaling of ecosystem condition to represent ecosystem service deficit and potential supply by green roofs. 4 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 deliver heat regulation services. We calculated satellite-derived land Apart from its utility to perform computation and analysis of large surface temperatures from Landsat 8 OLI/TIRS sensors using the single- datasets, GEE allows for the programming of an interactive user inter­ channel algorithm developed by Jiménez-Muñoz et al. (2008) and face known as a GEE web app. These online web applications allow one elaborated on in Venter et al., (2020b). Data were collected over July to deliver spatial data and results in an interactive map to end users who 2018 during which Norway experienced a severe heatwave. have no GEE credentials or scientific background. The web apps are To define areas in need of stormwater runoff regulation, and are developed in the GEE JavaScript API and are designed to facilitate sci­ therefore priorities for green roofs, we employed a mechanistic hydro­ ence communication and act as “sandbox” environments to develop logical model based on the rational formula to map annual runoff at the prototype web applications that can later be developed into more property level. The primary model inputs include terrain slope, property complex applications using the Google App Engine and Python API or dimensions, and surface land cover with associated runoff coefficients. similar. We programmed a web app to deliver the green roof MCDA as a We used the Norwegian Digital Terrain Model data to measure property tool to support municipal zoning in the implementation of the Green slope and a Sentinel-2 land cover product (http://urban.nina.no/) to Roof Strategy of Oslo and for increasing public interest and engagement define grass, tree and impervious surface area. These were ingested into with spatial planning of green infrastructure. We coded for a user the model to estimate runoff assuming a mean annual rainfall of 800 mm interface with widgets that allow users to toggle different GIS input and a one in 20-year storm event. For details about the runoff model layers, assign various criteria weightings and then render and compare please refer to Sælthun et al. (2021). the resulting prioritization in real-time. The background MCDA Green roofs offer habitat for urban biodiversity as provisioning and computation (weighted average of input criteria) takes place in raster supporting ES. We used pollinator habitat suitability as a proxy for format at 1 m spatial resolution. The raster format is a matrix/gridded biodiversity distribution within the city. Roofs that are located in areas geospatial data format that can be used for efficient computation at any with low pollinator habitat suitability are therefore priorities for spatial scale. All inputs and outputs are clipped to the building roof planting green roofs. We used a pollinator suitability map produced by geometries in Oslo. expert weighting of an urban land cover map (Stange et al., 2017; Zulian A beta-version of the App was presented to researchers and stake­ et al., 2013)) to calculate pollinator habitat within 250 m of each roof in holders during a workshop (October 2019) to evaluate and suggest the city. We used 250 m because this is the typical foraging distance of improvements to app functionality. Three of the participants had been wild bees which are keystone species in urban biodiversity assemblages. interviewed to obtain default preference profiles (urban farming, private To identify properties with very little existing green space, and residence, office) for the MCDA. therefore priorities for green roofs, we mapped green space at three levels: on-property, on-street and neighborhood. On-property green 3. Results space was calculated as the percentage vegetation (tree, shrub and grass) cover within each city property. The vegetation cover was derived from 3.1. Ecosystem service deficits the same Sentinel-2 land cover product as used in the stormwater runoff model. We calculated on-street greenness using the green view index There were 53,000 roofs within the Oslo built-up zone that were used data for Oslo provided by the Treepedia project (Seiferling et al., 2017). in the spatial prioritization exercise. The ES deficit scores vary sub­ The green view index quantifies the amount of green space perceived by stantially over the city with large ES deficits for roofs in the city center a person walking along a street using Google Street View imagery. To and low ES deficits for roofs in the city outskirts (Fig. 2). ES deficits characterize distal or neighborhood green space, we calculated the appear largest in areas with the highest population density which is due percentage of public green space within 250 m of each city roof. Here to the associated building density and lack of surrounding green space. public green space was defined as parks and green corridors digitized by High density population areas are used as a proxy indicator for demand the Oslo Bymiljøtaten. for ES. By this measure ES supply is not adequately matched to demand Green roofs may act as outdoor refuges from street level noise over space. pollution. We used a high resolution city noise map for Oslo (Oslo Kommune, 2013a, 2013b) to identify roofs located in noisy areas that 3.2. Ecosystem service trade-offs would most benefit from green roofs. Finally, to estimate the spatial distribution of potential exposure to We found that there were no clear spatial trade-offs between ES in ES deficits and potential demand for ES from green roofs, we used a map Oslo (i.e. one ES increases while another decreases). Instead, many ES of population density at 250 meter resolution supplied by the Statistics were correlated over space. A cluster of ES that are positively correlated Norway (https://kart.ssb.no/). Population density is multiplied by the include habitat for pollinators, greenspace amenities, and street green­ weighted sum of ES deficit, and can be switched on/off to assess dif­ view (Fig. 3). Another cluster includes on property greenview, storm­ ferences in priorities with and without a proxy indicator of demand. water regulation, and temperature regulation. None of the ES were With a 250 m grid resolution population density is a proxy for residential negatively correlated over space and therefore we did not expect to see and neighborhood ES exposure, but not exposure associated with significant trade-offs between stakeholder preferences. Therefore dif­ commuting, workplace, shopping and recreation away from home. ferences in MCDA criteria weightings were not expected to produce strong contrasts in spatial targeting of green roofs in that most stake­ 2.4. MCDA implementation and online web application holder profiles will result in high priority scores for roofs in the city center with relatively little vegetation ground cover. To maintain computational efficiency we used a simple weighted average of GIS criteria layers to produce the final MCDA prioritization 3.3. MCDA spatial prioritization outputs layer. All GIS input layers were first normalized between zero and one using the 5th and 95th percentile values to define the data range. This As stated, many ES criteria layers are already strongly correlated was done to prevent outlier values from skewing the normalization over the city with no strong trade-offs (Fig. 3). After calculating the process. Stakeholder weightings were used to assign relative weights to weighted-average ES deficit scores (i.e. green roof priority scores) based each GIS input layer and then a weighted average was calculated. This on the stakeholder-specific weightings in our MCDA, as expected pri­ approach assumes that planners would also use a linear additive utility oritizations were strongly correlated to one another also across stake­ function to prioritize green roof locations. The resulting map defined the holder preference profiles (Fig. 4). prioritization of areas for green roofs ranging from zero (low priority) to The one outlier was the community stakeholder profile which was one (high priority). negatively correlated to the hospital, hotel and office profiles (Fig. 4). 5 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 Fig. 2. Maps showing spatial distribution of GIS map layers of ES deficit used in the MCDA of green roof targeting in Oslo. ES deficit layers are based on rescaled ecosystem condition so that they represent a proxy for ecosystem service deficit ranging from low to high deficit (left to right of color bars; see Fig. 1 for explanation). Please refer to Table 1 for layer descriptions. The community stakeholder profile attributed higher priority to storm­ 4. Discussion water runoff mitigation and biodiversity ES criteria relative to hospital, hotel and office stakeholders (Table 1). Stormwater runoff and biodi­ The study presents an MCDA web application to engage with versity ES layers are more diffuse over the city and therefore this results stakeholders in GI design in a way that is intuitive for people. The in a more diffuse prioritization of roofs for green roof establishment. approach opens new pathways for capturing ‘plural values’ (Arias- This relative comparison is illustrated in Fig. 5. Community centre Arévalo et al., 2018; Cooper et al., 2016; Kenter, 2016), considering stakeholders weighted biodiversity and stormwater regulation as higher diverse societal needs and preferences at city scale, which potentially priorities compared to hotel stakeholders, reflected in the higher priority allows to overcome core shortcomings of deliberative group valuation in placed on roofs with a deficit of green infrastructure surrounding them attendance, such as limited stakeholder representation, dominant value (blue coloured buildings in Fig. 5). framings and power dynamics (Kenter et al., 2016; Lo & Spash, 2013; Wilson & Howarth, 2002), while information about trade-offs both be­ tween ES and between different stakeholder perspectives is interactively 3.4. Web application functionality enabling stakeholder engagement available within the valuation phase. The primary output of the green roof MCDA web application is a map of the weighted sum ES deficits indicating green roof priority for 4.1. An informed valuation approach different stakeholder profiles. Users are able to toggle between profiles and explore the distribution of green roof priorities by zooming in and A deliberative valuation process is especially useful when the de­ out and panning across the city. Users can also input their own hypo­ cisions at stake demand compromise solutions among a limited number thetical ES criteria weightings and render the MCDA result on-the-fly. of stakeholders (Langemeyer et al., 2018), By contrast our spatial MCDA We also catered for the ability to compare priority maps from con­ might be called an informed valuation approach, which has its strength in trasting stakeholder profiles using a screen slider widget (see center emphasizing societal differences in ES preferences and inequalities in panel in Fig. 5). Users can slide over the zoomable map and explore roofs the access of benefits across larger numbers of stakeholders and larger that are under- or over-prioritized relative to a contrasting stakeholder scales. An informed valuation approach enhances stakeholder knowl­ profile. This could help with stakeholder dialogue about the trade-offs of edge and enables them to make better informed choices on criteria prioritizing a specific building for green roof development. weights. Although similar web-based MCDA tools have been developed A further functionality of the web application is the ability to click on before (e.g. Padró et al., 2019), these are so far lacking a broader dif­ individual building roofs to extract ES criteria weightings specific to that ferentiation in objectives in order to become useful tools for supporting roof (Fig. 6). In this way, users can begin to investigate why some roofs ES trade-off analysis from a stakeholder perspective. In this vain, it is are prioritized over other roofs and how this varies across space at both important to highlight that even those MCDA approaches that put the neighborhood and city scale. emphasis on the analysis of trade-offs between ES (e.g. Grêt-Regamey Population density is applied as a scaling factor to ES deficit (Fig. 7) et al., 2013) or between different stakeholder groups (e.g. Munda 2008), providing a proxy indicator for areas of highest demand defined as are generally targeted to inform decision makers and show limitations in highest aggregate population deficit exposure. When applying the establishing informed criteria weights. However, the informed valuation population density scaling the MCDA identifies a band of optimal green approach applied here does not only inform stakeholders, but also roof locations in central residential neighbourhoods, but not in the widens the evaluative space for planners and policy-makers allowing for commercial city centre. a deeper understanding of contextual and perceptual factors that 6 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 Fig. 3. Correlation matrix between ecosystem service scores for roofs (n = 53,000) in Oslo. All ecosystem service scores are scaled between 0 and 1 and therefore no X- and Y-axis text is shown. The lower part of the matrix shows the scatter plots for ecosystem services combinations where each point represents an individual roof in the city. The upper half of the matrix shows the Pearson correlation coefficient and the level of statistical significance. The variable distribution is shown on the diagonal. Abbreviations: GV (greenview; mit. (mitigation); amen. (amenities); pollin. (pollinators); reg. (regulation). enhance ES benefits. expect to see. While the satellite image of landcover has high resolution, the ES maps have lower resolution, and because of smoothing functions employed across service areas, display little variation when zoomed to 4.2. Stakeholder feedback and uptake potential street or building level. Some participants expressed disappointment that individual buildings did not present discrete ES values differenti­ A workshop to evaluate the beta-version of the app revealed a ating them from adjacent buildings. An explanation offered by the differentiated picture of pros and cons of an interactive tool to raise model developers was that the tool does not represent the effects of awareness about ecosystem services (ES). Workshop participants potential green roof designs on the building. Users were disappointed appreciated the overview of ES gradients at city level. The clearly that building functional types were not identified for each building. delineated ES deficit of the city centre was commented on as relevant Consequently, the app does not provide guidance on what green roof input to municipal planning. Workshop participants quickly, and design is recommended for a building that is identified as being in a high without further technical instruction, engaged the different ES map deficit area. A further weakness in the GEE functionality that was layers. They started to use the tool to zoom in to familiar neighbour­ observed concerned the inability to save, store and aggregate preference hoods and buildings. This interactive functionality with map scale and weightings across many users. The results can be saved only by doing a resolution revealed a counter-intuitive drawback of the tool. One screen copy. The purpose of the GEE app was explained as providing a participant remarked that he had expected using the zoom function to be “sandbox” to discuss decision-support tool features desired by like “parachuting” - the closer you get to the ground the more detail you 7 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 stakeholders. More advanced user interface and database features would need to be programmed in more flexible environments. 4.3. Ecosystem service valuation assumptions From a technical MCDA modeling perspective we acknowledge that the linear utility function assumed by the weighted summation of normalized ES scores is a simplification. It also assumes that all the value functions scaling ES are linear and deterministic. It was implemented here due to its simplicity in obtaining preference profiles and its computational efficiency in providing summary maps of different pref­ erence profiles in real-time. More flexible non-linear valuation functions with uncertainty ranges (Beinat 1997) have been implemented in a spatially explicit Bayesian network model (Langemeyer et al. 2020). The Bayesian network model was not possible to implement in the GEE modeling environment. 4.4. Decision-support constraints and opportunities A possible drawback of the modeling process we adopted was an apparent research-driven definition of the decision objective and criteria. We identified the spatial targeting purpose of the model based on a desktop study of municipal plans, strategy and norms, supple­ Fig. 4. Correlation matrix for weighted-average ES deficit scores (i.e. spatial mented by individual interviews. ES included in the app were compat­ prioritization scores) under different stakeholder profiles. Correlations are ible with documented municipal policy priorities, but the desktop-based based on prioritization scores for all roofs in Oslo (n = 53,000). Pearson cor­ approach did not open up for possible innovations in e.g. new criteria for relation coefficients form the basis for the size and the color of each square. priority-setting that could have emerged from a group-based co-creation approach (Langemeyer et al. 2020). A defence of the modeling approach was that a consultative co-design process was followed in development Fig. 5. Maps showing outputs of the MCDA producing a map of the weighted sum of ecosystem service deficits, and conse­ quent green roof priority (left-hand maps). The difference in prioritization for two contrasting stakeholder profiles (com­ munity, user 1 vs hotel, user 2) is mapped in right-hand maps to illustrate potential trade-offs and synergies. Blue roofs are those that have a higher priority under the community centre relative to the hotel profile. White bounding boxes identify zoomed extents for borough (center row) and neighborhood (bottom row) scales. 8 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 Fig. 6. Screenshot of the MCDA web application highlighting the ability to select individual roofs (blue border) and extract statistics of ecosystem service deficit scores. Fig. 7. Screenshot highlighting ecosystem service deficit areas scaled by population density as a proxy for areas of highest demand. of Oslo’s Green Roof Strategy. The criteria we included in the app fol­ preference profiles. In our Oslo case, ES deficits combined with simi­ lowed the main priorities identified in a prior co-design process. larity in stakeholder preference profiles did not reveal large differences in the spatial targeting of green roofs across different stakeholders. The spatial gradient of vegetation from city core to periphery strongly cor­ 4.5. Spatial and temporal correlation of ES deficits and spatial relates with ES. The high correlation of ES deficits with population prioritization density is a pattern observed in other cities (Langemeyer et al. 2020). The general implication for spatial MCDA is that basic spatial analysis of An advantage of the app in relation to a co-creation objective to ecosystem service correlation may reveal coincidences of stakeholders’ promote learning, is the ease with which users can test different 9 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 interests despite even strong differences in preferences for individual ES. the use of online spatial MCDA tools can be expected to improve on both In our Oslo case, differences between stakeholder preference happened process and outcome efficiency in spatial planning of nature-based so­ to be moderate. A more differentiated spatial targeting map is obtained lutions and urban nature restoration. when engaging the population density criteria scaling ES deficit. The resulting map revealed priority areas in a residential belt in the urban Declaration of Competing Interest core of the city, but excluding the downtown commercial area with low population density. Workshop participants observed that this would be a The authors declare that they have no known competing financial significant modification to the zoning identified in the bluegreen factor interests or personal relationships that could have appeared to influence norm (Supplementary Material S1) which identifies the urban core as a the work reported in this paper. homogenous area. A weakness of population density as a proxy indicator for exposure is that it does not represent ES exposure for inhabitants Acknowledgements during daily activities outside of their residential neighbourhood. This modifies current planning incentives for blue-green infrastructure pro­ This research was funded through the 2015–2016 BiodivERsA moted by Oslo Municipality, which include the whole city centre (Oslo COFUND call for research proposals, with the national funders the Kommune 2019). Ideally, future iterations would include spatial data on Swedish Research Council for Environment, Agricultural Sciences, and population mobility in order to incorporate the temporal nature of Spatial Planning; Swedish Environmental Protection Agency; German population density through the day and year. Aerospace Center (DLR); National Science Centre (Poland); the Research Council of Norway; and the Spanish Ministry of Science, Innovation and 4.6. Transferability of the approach to other cities Universities. DNB acknowledges support from the Research Council of Norway to the ENABLE project through the BiodivERsA COFUND 2015- Despite the Oslo-specific pros and cons of the MCDA web application, 2016 call for research proposals. JL acknowledges support from the ERC there remains scope for testing similar MCDA approaches in other cities. Consolidator Grant 818002-URBAG awarded to Gara Villalba. FB ac­ Land use/cover data is the primary input into the ES surplus/deficit knowledges support from the EU’s Horizon 2020 framework program layers used in the Oslo MCDA. While we acknowledge that not all cities for research and innovation (project NATURVATION, grant agreement will have access to detailed spatial data (e.g. LiDAR digital terrain ID: 730243). TM was supported by US National Science Foundation model, population density etc.), high resolution land cover data can be grants (#1444755, #1927167, and #1934933). Thanks to Zofie Cim­ generated from open-access satellite data such as Copernicus Sentinel burova, Diego Wedgewood and Nabolagshager for help with data data (e.g. Baamonde et al., 2019; Qiu et al., 2019; Venter et al., 2020a) collection and analysis. for any location globally. Further, the GEE platform used to deploy the web application is available free of charge for research purposes. Appendix A. Supplementary data Therefore the approach presented here is broadly transferable and policy-relevant in other cities, especially those with dense and compact Supplementary data to this article can be found online at https://doi. neighborhoods where green infrastructure is scarce. In Barcelona, for org/10.1016/j.ecoser.2021.101314. instance, it could support the priority-setting process of current municipal funding incentives for the implementation of green roofs. References Similarly, in New York City, which recently passed legislation requiring all new buildings to have solar or green roofs, the approach here could Allain, S., Plumecocq, G., Leenhardt, D., 2017. How do multi-criteria assessments address be used to help prioritize where to focus limited funds to support new landscape-level problems? A review of studies and practices. Ecol. Econ. 136, 282–295. green roof installations. Andersson, E., Langemeyer, J., Borgström, S., McPhearson, T., Haase, D., Kronenberg, J., Barton, D.N., Davis, McKenna, Naumann, S., Röschel, L., Baró, F., 2019. Enabling 5. Conclusion urban green and blue infrastructure to improve contributions to human well-being and equity in urban systems. Bioscience. https://doi.org/10.1093/biosci/biz058. Arias-Arévalo, P., Gómez-Baggethun, E., Martín-López, B., Pérez-Rincón, M., 2018. In this study we evaluate prioritization green roof locations at city Widening the evaluative space for ecosystem services: a taxonomy of plural values scale with building-level resolution, using an interactive multi-criteria and valuation methods. Environ. Values 27 (1), 29–53. Baamonde, S., Cabana, M., Sillero, N., Penedo, M.G., Naveira, H., Novo, J., 2019. Fully decision analysis implemented in a GEE platform. Based on a desktop automatic multi-temporal land cover classification using Sentinel-2 image data. study of municipal plans, strategies and norms we identified a gap in Procedia Comput. Sci. 159, 650–657. mapping tools at area planning level. The app was developed as a ‘meso- Beinat, E., 1997. Value Functions for Environmental Management. Kluwer Academic level’ spatial tool for targeting green roofs at neighbourhood and street Publishers. Benedict, M.A., McMahon, E.T., 2002. Green infrastructure: smart conservation for the level. The main objective of the app is to provide a model “sandbox” 21st century. Renewable Resour. J. 20 (3), 12–17. where stakeholders and researchers can identify further planning pri­ Berardi, U., Ghaffarianhoseini, A., Ghaffarianhoseini, A., 2014. State-of-the-art analysis orities and functionalities desired in an operative decision-support tool. of the environmental benefits of green roofs. Appl. Energy 115, 411–428. Bugliarello, G., 2006. Urban sustainability: dilemmas, challenges and paradigms. For the Oslo application our analysis revealed high spatial correlation in Technol. Soc. 28 (1), 19–26. ES deficits. Under these conditions even large differences in stakeholder Childers, Daniel, Bois, Paul, Hartnett, Hilary, McPhearson, Timon, Metson, Genevieve, preferences for individual ES in MCDA will lead to common interests in Sanchez, Christopher A., 2019. Urban Ecological Infrastructure: An inclusive concept for the non-built urban environment. Elem. Sci. Anth. 7 (1), 46. https://doi.org/ spatial targeting of green roofs. The non-technical and interactive 10.1525/elementa.385. functionality of the online spatial MCDA was perceived as a useful tool Cooper, N., Brady, E., Steen, H., Bryce, R., 2016. Aesthetic and spiritual values of for the process efficiency of engagement with stakeholders, but MCDA ecosystems: recognising the ontological and axiological plurality of cultural ecosystem ‘services’. Ecosyst. Serv. 21, 218–229. itself was not crucial for the spatial targeting results. In our particular Daniels, B., Zaunbrecher, B.S., Paas, B., Ottermanns, R., Ziefle, M., Roß-Nickoll, M., 2018. Oslo case, simple spatial analysis of residential population exposure to Assessment of urban green space structures and their quality from a ES deficits was sufficient to improve on current zoning and permitting multidimensional perspective. Sci. Total Environ. 615, 1364–1378. Depietri, Yaella and Timon McPhearson. 2017. Integrating the grey, green, and blue in tools. Current green roof projects are largely focused on new office and cities: Nature-based solutions for climate change adaptation and risk reduction, commercial buildings in the city centre. We identified a residential belt (pp91-109) in: Kabisch, N., Korn, H., Stadler, J., Bonn, A. (Eds), Nature-based within the inner city as the area with highest potential for green roof Solutions to Climate Change in Urban Areas: Linkages Between Science, Policy, and retrofitting to satisfy public ES demands in Oslo. In other cities with Practice, Springer. Depietri, Yaella, Dahal, Khila, McPhearson, Timon, 2018. Multi-hazard risk in a coastal more heterogeneous distribution of green infrastructure, lower spatial megacity. Nat. Hazards Earth Syst. Sci. 18, 3363–3381. https://doi.org/10.5194/ correlation of ES and greater differences in stakeholder ES preferences, nhess-18-3363-2018. 10 Z.S. Venter et al. Ecosystem Services 50 (2021) 101314 Elmqvist, Thomas, Andersson, Erik, Frantzeskaki, Niki, McPhearson, Timon, Folke, Carle, Mendoza, G.A., Martins, H., 2006. Multi-criteria decision analysis in natural resource Olsson, Per, Takeuchi, Kazuhiko, 2019. Sustainability and resilience for urban management: a critical review of methods and new modeling paradigms. For. Ecol. transformations. Nat. Sustainability 2, 267–273. https://doi.org/10.1038/s41893- Manage. 230, 1–22. 019-0250-1. Oikonomou, V., Dimitrakopoulos, P.G., Troumbis, A.Y., 2011. Incorporating ecosystem Fish, R., Saratsi, E., Reed, M., Keune, H., 2016. Stakeholder participation in ecosystem function concept in environmental planning and decision-making by means of multi- service decision making. in: Potschin (Ed.), Routledge Handbook of Ecosystem criteria evaluation: the case-study of Kalloni Lesbos, Greece. Environ. Manage. 47 Services. Routlegde, pp. 256–270. (1), 77–92. Frantzeskaki, Niki, McPhearson, Timon, Collier, Marcus J., Kendal, Dave, Opricovic, S., Tzeng, G.H., 2007. Extended VIKOR method in comparison with Bulkeley, Harriet, Dumitru, Adina, Walsh, Claire, Noble, Kate, van Wyk, Ernita, outranking methods. Eur. J. Operat. Res. 178 (2), 514–529. Ordóñez, Camilo, Oke, Cathy, Pinter, Laszlo, 2019. Nature-based solutions for urban Oslo Kommune, 2013. Oslo Og Green! Application to become European Green Capital climate change adaptation: linking the science, policy, and practice community for 2016. Submitted 31.10.2013. evidence-based decision-making. Bioscience. https://doi.org/10.1093/biosci/ Oslo Kommune, 2013. STØYKARTLEGGING 2012 Oslo, Asker, Bærum, Skedsmo, biz042. Lørenskog, Rælingen og Oppegård Støysituasjon for året 2011. Bymiljøetaten. Funtowicz, S.O., Ravetz, J.R., 1994. The worth of a songbird: ecological economics as a Oslo Kommune, 2018. GRØNNE FRAMTIDSUTSIKTER Strategi for grønne tak og fasader post-normal science. Ecol. Econ. 10 (3), 197–207. i Oslo Bygningsintegrerte grønnstrukturer 2019-2030. Bymiljøetaten. García, A.M., Santé, I., Loureiro, X., Miranda, D., 2020. Green infrastructure spatial Oslo Kommune, 2019. Blågrønn faktor for boliger i Oslo—Norm. 11.12.2019. planning considering ecosystem services assessment and trade-off analysis. Oslo Kommune 2020. Høring på Forslag til strategi for grønne tak og fasader (Sak: Application at landscape scale in Galicia region (NW Spain). Ecosyst. Serv. 43, 201502863 – 22). Plan og Bygningsetaten. 101115. Padró, R., La Rota-Aguilera,M., J., Giocoli, A., Cirera, J., Coll, F., Pons, M., Pino, J., Pili, Garmendia, E., Gamboa, G., 2012. Weighting social preferences in participatory multi- S., Serrano, T., Villalba, G., Marull, J., 2019. Assessing the sustainability of criteria evaluations: a case study on sustainable natural resource management. Ecol. contrasting land use scenarios through the Socioecological Integrated Analysis (SIA) Econ. 84, 110–120. of the metropolitan green infrastructure of Barcelona, Metropolitan Laboratory of Gomez-Baggethun, E., Barton, D.N., 2013. Classifying and valuing ecosystem services for Ecology and Territory of Barcelona. urban planning. Ecol. Econ. 86, 235–245. https://doi.org/10.1016/J. Qiu, C., Mou, L., Schmitt, M., Zhu, X.X., 2019. Local climate zone-based urban land cover Ecolecon.2012.08.019. classification from multi-seasonal Sentinel-2 images with a recurrent residual Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., Moore, R., 2017. network. ISPRS J. Photogrammetry Remote Sensing 154, 151–162. Google Earth Engine: planetary-scale geospatial analysis for everyone. Remote Sens. Ramyar, R., Saeedi, S., Bryant, M., Davatgar, A., Hedjri, G.M., 2020. Ecosystem services Environ. 202, 18–27. mapping for green infrastructure planning–the case of Tehran. Sci. Total Environ. Grêt-Regamey, A., Celio, E., Klein, T.M., Wissen Hayek, U., 2013. Understanding ESs 703, 135466. trade-offs with interactive procedural modeling for sustainable urban planning. Saarikoski, H., Mustajoki, J., Barton, D., Geneletti, D., Langemeyer, J., Gómez- Landscape Urban Plann. 109 (1), 107–116. https://doi.org/10.1016/j. Baggethun, E., Marttunen, M., Antunes, P., Keune, H., Santos, R., 2016. Multicriteria Landurbplan.2012.10.011. decision analysis and cost-benefit analysis: comparing alternative frameworks for Grove, M.J., Ogden, L., Pickett, S., et al., 2018. The Legacy Effect: Understanding How integrated valuation of ecosystem services. Ecosyst. Serv. 22, 238–249. Segregation and Environmental Injustice Unfold over Time in Baltimore. Annals of Saarikoski, H., Mustajoki, J., Hjerppe, T., Aapala, K., 2019. Participatory multi-criteria the Association of American Geographers, 108: 524–37. decision analysis in valuing peatland ecosystem services—trade-offs related to peat Jiménez-Muñoz, J.C., Cristóbal, J., Sobrino, J.A., Sòria, G., Ninyerola, M., Pons, X., 2008. extraction vs. pristine peatlands in Southern Finland. Ecol. Econ. 162, 17–28. Revision of the single-channel algorithm for land surface temperature retrieval from Saaty, T.L., 1980. A scaling method for priorities in hierarchy structures. J. Math. Landsat thermal-infrared data. IEEE Trans. Geosci. Remote Sens. 47 (1), 339–349. Psychol. 15, 234–281. Karjalainen, T.P., Marttunen, M., Sarkki, S., Rytkönen, A.-M., 2013. Integrating ESs into Sælthun, N.R., Barton, D.N., Venter, Z.S., 2021. REO: estimering av overflateavrenning environmental impact assessment: an analytic–deliberative approach. Environ. fra urbane felt. Beregningsgrunnlag for et arealdifferensiert overvannsgebyr Impact Assess. Rev. 40, 54–64. (revidert utgave). Oslo: Norsk institutt for naturforskning (NINA) 2021 (ISBN 978- Keeler, Bonnie L., Hamel, Perrine, McPhearson, Timon, Hamann, Maike, 82-426-4723-8) 33 s. NINA rapport 1851b. Donahue, Marie, Prado, Kelly Meza, Arkema, Katie, Bratman, Gregory, Sanon, S., Hein, T., Douven, W., Winkler, P., 2012. Quantifying ES trade-offs: the case of Brauman, Kate, Finlay, Jacques, Guerry, Anne, Hobbie, Sarah, Johnson, Justin, an urban floodplain in Vienna, Austria. J. Environ. Manage. 111, 159–172. https:// MacDonald, Graham, McDonald, Robert, Neverisky, Nick, Wood, Spencer, 2019. doi.org/10.1016/j.jenvman.2012.06.008. Social-ecological and technological factors moderate the value of urban nature. Nat. Seiferling, I., Naik, N., Ratti, C., Proulx, R., 2017. Green streets− quantifying and Sustainability 2, 29–38. mapping urban trees with street-level imagery and computer vision. Landscape Kenter, J.O., 2016. Shared, plural and cultural values. Ecosyst. Serv. 21 (B), 175–183. Urban Plann. 165, 93–101. Kenter, J.O., Jobstvogt, N., Watson, V., Irvine, K.N., Christie, M., Bryce, R., 2016. The Srdjevic, Z., Lakicevic, M., Srdjevic, B., 2013. Approach of decision making based on the impact of information, value-deliberation and group-based decision-making on analytic hierarchy process for urban landscape management. Environ. Manage. 51 values for ecosystem services: integrating deliberative monetary valuation and (3), 777–785. storytelling. Ecosyst. Serv. 21, 270–290. Stange, E., Zulian, G., Rusch, G., Barton, D., Nowell, M., 2017. Ecosystem services Kuller, M., Bach, P.M., Roberts, S., Browne, D., Deletic, A., 2019. A planning-support tool mapping for municipal policy: ESTIMAP and zoning for urban beekeeping. One for spatial suitability assessment of green urban stormwater infrastructure. Sci. Total Ecosyst. 2. Environ. 686, 856–868. Susca, T., Gaffin, S.R., Dell’Osso, G.R., 2011. Positive effects of vegetation: Urban heat Langemeyer, J., Connolly, J.J.T., 2020. Weaving notion of justice into urban ecosystem island and green roofs. Environ. Pollut. 159 (8–9), 2119–2126. https://doi.org/ service research and practice. Environ. Sci. Policy 109 (2020), 1–14. https://doi.org/ 10.1016/J.Envpol.2011.03.007. 10.1016/j.envsci.2020.03.021. Turner, B., Henryks, J., Pearson, D., 2011. Community gardens: sustainability, health Langemeyer, J., Baraibar, S., Palomo, I., Gómez-Baggethun, E., 2018. Participatory and inclusion in the city. Local Environ. 16 (6), 489–492. Multi-Criteria Decision Aid: A way to operationalize ecosystem services. Ecosystem Vallecillo, S., La Notte, A., Polce, C., Zulian, G., Alexandris, N., Ferrini, S. and Maes, J., Services, 30, Part A, 49-60. DOI:10.1016/j.ecoser.2018.01.012. 2018. Ecosystem services accounting: part I–Outdoor recreation and crop Langemeyer, J., Gómez-Baggethun, E., Haase, D., Scheuer, S., Elmqvist, T., 2016. pollination. Publications Office of the European Union, Luxembourg. Bridging the gap between ecosystem service assessments and land-use planning Van der Horst, L., Townsend, L., Jacobs, A., 2013. Merging man with nature: Restoring through multi-criteria decision analysis (MCDA). Environ. Sci. Policy 62, 45–56. the environmental functions and processes of Sørhellinga and the regional biotope Langemeyer, J., Wedgwood, D., Barton, D., Baró, F., McPhearson, T., 2020. Creating diversity by creating a green roof. Term paper. INA Department. Norwegian nature-based solutions where they are needed – a spatial ecosystem service-based University of Life Sciences. decision analysis of green roofs in Barcelona. Sci. Total Environ. 707 (10), 135487 Venter, Z.S., Brousse, O., Esau, I., Meier, F., 2020a. Hyperlocal mapping of urban air https://doi.org/10.1016/j.scitotenv.2019.135487. temperature using remote sensing and crowdsourced weather data. Remote Sens. Lo, A.Y., Spash, C.L., 2013. Deliberative monetary valuation: in search of a democratic Environ. 242, 111791. and value plural approach to environmental policy. J. Econ. Surveys 27 (4), Venter, Z.S., Krog, N.H., Barton, D.N., 2020b. Linking green infrastructure to urban heat 768–789. and human health risk mitigation in Oslo, Norway. Sci. Total Environ. 709, 136193. Locke, D.H., Hall, B., Grove, J.M., Pickett, S.T.A., Ogden, L.A., Aoki, C., Boone, C.G., Venter, Z.S., Shackleton, C.M., Van Staden, F., Selomane, O., Masterson, V.A., 2020c. O’Neil-Dunne, J.P.M.. [in review]. Residential housing segregation and urban tree Green Apartheid: Urban green infrastructure remains unequally distributed across canopy in 37 US Cities. Submitted to Urban Sustainability. Preprint available here: income and race geographies in South Africa. Landscape Urban Plann. 203, 103889. https://osf.io/preprints/socarxiv/97zcs. Whittinghill, L.J., Rowe, D.B., 2012. The role of green roof technology in urban Mascarenhas, A., Ramos, T.B., Haase, D., Santos, R., 2016. Participatory selection of agriculture. Renewable Agric. Food Syst. 27 (04), 314–322. ecosystem services for spatial planning: Insights from the Lisbon Metropolitan Area, Wilson, M.A., Howarth, R.B., 2002. Discourse-based valuation of ecosystem services: Portugal. Ecosyst. Serv. 18, 87–99. establishing fair outcomes through group deliberation. Ecol. Econ. 41 (3), 431–443. McPhearson, T., Pickett, S.T.A., Grimm, N., Niemelä, J., Alberti, M., Elmqvist, T., Zhang, X., Lu, X., 2010. Multiple criteria evaluation of ecosystem services for the Weber, C., Haase, D., Breuste, J., Qureshi, S., 2016. Advancing urban ecology toward Ruoergai Plateau Marshes in southwest China. Ecol. Econ. 69 (7), 1463–1470. a science of cities. Bioscience 66 (3), 198–212. https://doi.org/10.1093/biosci/ Zia, A., Hirsch, P., Songorwa, A., Mutekanga, D.R., O’Connor, S., McShane, T., biw002. Brosius, P., Norton, B., 2011. Cross-scale value trade-offs in managing social- McPhillips, L., Chang, H., Chester, M.V., Depietri, Y., Friedman, E., Grimm, N.B., ecological systems: the politics of scale in Ruaha National Park, Tanzania. Ecol. Kominoski, J.S., McPhearson, T., Méndez-Lázaro, P., Rosi, E.J., Shiva, J.S., 2018. Society 16 (4). Defining extreme events: a cross-disciplinary review. Earth’s Future. https://doi. Zulian, G., Maes, J., Paracchini, M., 2013. Linking land cover data and crop yields for org/10.1002/2017EF000686. mapping and assessment of pollination services in Europe. Land 2, 472–492. 11