Urbanisation is a complex territorial process that refers to changes in the population distribution and to a transformation of the environment (Mcgranahan and Satterthwaite 2014, United Nations 2019). It implies changes in dominant occupations of land, lifestyle, culture and behaviour and, thus, alters the demographic and social structure of both urban and rural areas (Montgomery et al. 2004). Urbanisation causes pressure on natural resources and can lead to critical environmental degradation, such as flooding, air pollution, microclimatic alteration, collapse of natural ecosystems and modifications in the structure of species communities (Marzluff et al. 2008, Villalobos-Jiménez et al. 2016, Ferreira et al. 2019, Kondratyeva et al. 2020). Urban environments are extensive and are growing. Europe's level of urbanisation is expected to increase to approximately 83.7% in 2050*1.

Urban ecosystems are defined as socio-ecological systems where most people live (Maes et al. 2020). The urban ecosystem includes abiotic spheres (the atmosphere, hydrosphere, lithosphere and soil or pedosphere) and biotic spheres (often viewed as an interacting biosphere of urban plants and animals, plus the socio-economic world of people, the anthroposphere. Urban ecosystems are largely artificial but they include vegetation, green patches (public and private), forests, lakes and rivers and agricultural areas, all strongly influenced by human activities (Maes et al. 2020).

The condition of urban ecosystems affects human well-being (Gascon et al. 2016, Dadvand et al. 2016, van den Berg et al. 2016, Tischer et al. 2017, Gascon et al. 2017, Sarkar and Webster 2017); biodiversity (Kowarik 2011, McPhearson et al. 2018, Kowarik et al. 2020) and the way cities impact their surroundings. “A better understanding of urban ecosystems will help us create more liveable cities that provide high-quality habitat for humans and non-humans alike” (Parris 2016). Nowadays, the importance of nature in cities is well recognised (La Notte and Zulian 2021). Urban green spaces provide several key regulating, cultural and provisioning ecosystem services, such as microclimate regulation, flood control, air quality regulation, noise pollution mitigation or nature-based recreation (Haase et al. 2014, Haines-Young and Potschin-Young 2018). In addition, urban ecosystems are also important for biodiversity conservation goals (Kowarik 2011). However, the capacity of urbanised areas to support biodiversity is strongly related to their configuration and to the structure and extent of their Urban Green Infrastructure (UGI) (Pellissier et al. 2012, Beninde et al. 2015), which is defined by the European Commission (European Commission 2013) as: "a strategically managed network of urban green spaces and natural and semi-natural ecosystems” situated within the boundary of an urban ecosystem.

The role of urban areas in addressing environmental issues is increasingly being recognised in European policies and international commitments, such as the post-2020 Global Biodiversity Framework*2. On 1 December 2019, the Commission published The European Green Deal (European Commision 2019), a set of policy initiatives with the overarching aim of making Europe climate-neutral in 2050. The EU Biodiversity Strategy for 2030 is one of the European Green Deal pillars (European Commission 2020). The Strategy aims to: “…ensure that Europe's biodiversity will be on the path to recovery by 2030 for the benefit of people, the planet, the climate and our economy...” (European Commission 2020). Specifically, the Strategy intends to protect and restore nature in the European Union by “improving and widening our network of protected areas and by developing an ambitious EU Nature Restoration Plan” (European Commission 2020).

Interestingly, cities take centre-stage in this strategy. Section 2.1 focuses on the development of "A coherent network of protected areas", expressing the need to enlarge and improve the "Trans-European Nature Network". In this context, UGI is acknowledged for its pivotal role in supporting Trans-European network connectivity, especially in dense urbanised areas. Section 2.2 outlines the "new EU Nature Restoration Plan" and provides ten spheres of action to "improve the health of existing and new protected areas, and bring diverse and resilient nature back to all landscapes and ecosystems". One of the areas of interests are cities. Section 2.2.8 specifically refers to “greening urban and peri-urban areas” to halt and reverse the loss of urban green . Cities with more than 20,000 inhabitants are called on to develop an Urban Greening Plan to increase urban biodiversity and improve UGI, such as forests, parks and gardens. To facilitate and support the process in 2021, the Commission initiated the “Green City Accord” (GCA)*3, an initiative of European cities committed to safeguarding the environment. The GCA aims to improve the quality of life and accelerate the implementation of the European Green Deal, relevant EU environmental directives and regulations, as well as the Sustainable Development Goals. By signing the GCA, cities commit to step up their efforts in five key areas by 2030: air, water, nature and biodiversity, circular economy, waste and noise.

For monitoring the nature and biodiversity section, signatory cities will report on five mandatory indicators which should be easy to assess and monitor. One of them is "changes in vegetation cover in UGI".

The analysis of changes in vegetation cover is a common methodology, applied at a national (Forkel et al. 2013, Novillo et al. 2019), regional (Han et al. 2013) or urban scale (Yu et al. 2017, Corbane et al. 2018, Jin et al. 2020). It provides a measure of vegetation presence, health and dynamics and complements area-based indicators.

Traditionally, UGI is analysed by reporting the extent of urban green, measured with reference to the city area (% of urban green) or as m² per inhabitants (Kabisch and Haase 2013, Carmen et al. 2020, Merino-Saum et al. 2020). Additionally, UGI is analysed with reference to the extent of urban green, measured in relative or absolute terms, in terms of accessibility or proximity to people (Morar et al. 2014, Kabisch et al. 2016, Maes et al. 2019, Dumitru and Wendling 2021), or considering the structure or the connectivity of green patches (Chan et al. 2014, Rusche et al. 2019) .

However, the above-mentioned indicators do not capture the presence, health and dynamics of vegetation that determine the delivery of ecosystem services and provide habitats to fauna. The analysis of changes in vegetation cover, therefore, complements other metrics that report on the share of restored and protected semi-natural and natural urban areas and proximity of public green spaces to citizens. The approach is implemented in several studies. To cite only a few examples: Wang et al. (2018) quantified the process of growth, shrinking or disappearance of patches, by implementing object-based image analysis and using 2.5 m resolution ALOS and SPOT image data captured in 2005 and 2009. The authors state that the approach provides a tool for planners and managers. Jin et al. (2020) analysed the green spaces trends within and amongst 16 major pan-Pacific cities. They evaluated long-term changes in vegetation cover focusing on the analysis of patterns along the urban-rural gradient. The work has enormous potential for policy support, but high scientific and technical skills are required. Zhang et al. (2020) explored changes in vegetation cover in both urban core and peripheral areas and the relationships between these changes and urban development in eights Chinese cities, whereas Yu et al. (2017) developed an approach that integrates trends with dynamic analysis to measure the vegetation changes from the process perspective, based on the time‐series Landsat imagery. However, the methods described above are quite complex, require advanced competences in geomatic, remote sensing and statistics and imply an additional effort to communicate the results to non-expert public and policy-makers.

Several review papers and real-world experiences demonstrated that simple indicators are preferred for policy support. To cite just a few examples: Merino-Saum et al. (2020) show that the green areas per capita (second ranked) and share of population with access to green areas (21^st ranked) are amongst the most frequently used indicators for monitoring and comparing urban sustainability performance. Michalina et al. (2021), in a recent review of urban sustainability indicator frameworks (USIFs), provide a list of key indicators that can be used by policy-makers to assess and monitor urban sustainable development. Amongst environmental indicators, the authors report:

• Share of built-up area, forest, water, agricultural land and other areas of the total city area (%)
• Share of protected nature areas of the total city area (%).

Amongst social indicators the authors report:

• Green area within the city (forests, parks, gardens etc.) per inhabitant (m²/inhabitant)
• Percentage of inhabitants living within 300 m or 15 min walk from public green space > 5000 m² (%).

Carmen et al. (2020) addressed why cities prefer certain ecological, economic and social indicators. The authors analysed policy documents produced in Coimbra, Vilnius, Leipzig and Genk. Central to their analysis was the actual use of certain indicators by city officials. Amongst other factors, such as accessibility and knowledge about the indicators, this choice was assumed to be related to the indicators’ relevance, feasibility, clarity and credibility. An indicator is more effective if it provides higher quality information on multiple benefits and is easier to implement (i.e. feasibility is high and/or it is easy to implement in practice). Amongst the ten most effective indicators, the authors listed: “Area of green space/capita (district wide)”, being measured and used in Leipzig and Vilnius and “Tree cover” being measured, but not used in Leipzig.

Other examples of indicators to measure urban green or urban biodiversity come from Manchester and Lisbon. Manchester uses “Percentage green infrastructure cover” and “Percentage of tree canopy cover” (Manchester City Council 2015). The City of Lisbon, a pioneer in establishing goals and strategies on urban green and urban biodiversity (Secretariat of the Convention on Biological Diversity 2012), has been one of the first cities in Europe to implement the City Biodiversity Index (CBI) (Kohsaka et al. 2013). The CBI, however, does not include an indicator related to vegetation trends, consequently, it was not included in the assessments used to frame the Lisbon biodiversity strategy*6.

Here, we argue that the analysis of long-term trends in vegetation cover should be part of an indicator framework to monitor urban sustainability and urban green. Yet, the long-term assessment of vegetation cover at the urban scale requires high scientific-technical skills and associated indicators may be complex, which has implications for their communication. Hence, this might explain why indicators that quantify long-term vegetation trends are not regularly included amongst the indicators used for urban policy-making and planning.

This study fills this gap by providing a framework for the analysis of changes in vegetation cover inside UGI. The framework was developed and implemented from a policy-support and decision-making perspective. Additionally, a java-script for the automatic computation of the trend analysis on the Google Earth Engine (GEE) platform is included.

The methodology allows users to monitor land use change and land compensation, as well as the availability and quality of different typologies of urban green. Thus, it could serve as a decision support tool for the design and implementation of "no net loss" policies (Ermgassen et al. 2019), enhancing biodiversity or restoring degraded habitats and ecosystems.

Since the analysis covers multiple governance levels and scales, from urban park to district, from municipality to European level, it further supports the harmonisation of methods and data needed to provide information for policy-making and for monitoring of implementation across scales (Scholes et al. 2013). For instance, at EU level, the methodology provides information for the European Green Deal (European Commision 2019), the EU Climate Adaptation Strategy (European Commission 2021) and the EU Biodiversity Strategy for 2030 (European Commission 2020).

The objectives of this paper are to:

1. Introduce the full framework for the analysis of changes in vegetation cover inside UGI;
2. Rationalise and optimise the analytical procedure and make it available on GEE;
3. Apply the methodology in a multi-scale perspective;
4. Discuss the potential of this indicator to support policy-making and implementation across governance levels.

The methodology was implemented in all European Functional Urban Areas (FUAs). For demonstrative reasons, an additional analysis was done at the metropolitan and sub-municipal level.

Study areas

A framework for the analysis of changes in vegetation cover of UGI

The aim of this study was to implement a replicable approach to assess the urban vegetation changes in a multi-scale perspective. For this reason, only consistent and open source data were used. Table 2 reports the data used in the study.

Land-cover data, used to establish the zonation (as described above), was derived from Corine Land Cover (CLC). CLC uses a Minimum Mapping Unit (MMU) of 25 ha for areal phenomena and a minimum width of 100 m for linear phenomena and is available at a 100 m resolution (Büttner and Kosztra 2017). At European level, Urban Atlas*4 is also available; it provides an inter-comparable, high-resolution land-use and land-cover data for FUA for 2012 and 2018.

In this application, we did not use the Urban Atlas for three reasons:

1. The Urban Atlas provides the extent and form of forest, semi-natural vegetation, herbaceous vegetation and urban green. It does not allow the detection of small green patches.

2. The Urban Atlas is available only for 2006 (but only 319 Large Urban Zones (LUZ) are available (see the technical*4 documentation), 2012 and (since 2019) 2018. This study was developed when data for 2018 were not published.

3. The Urban Atlas does not allow us to consider what is happening immediately outside a FUA's boundaries which is very important when assessing urban ecosystems. In Europe, many border FUAs exist, for instance, Vienna (Austria) and Bratislava (Slovakia). In this case, the LM approach may give wrong results if applied using Urban Atlas.

Data on urban green were derived from GEE. The Greenest_TOA (from now on called greenest) product was chosen. These composites are created for all the scenarios in each annual period beginning from the first to the last day of the year. All the images from each year are included in the composite, with the greenest pixel as the composite value, where the greenest pixel means the pixel with the highest NDVI value (Chander et al. 2009). The time-series ranges from 1996 to 2018 and the spatial resolution is 30 m.

The FUA limits were derived from the URBAN AUDIT catalogue, version 2018.

Data used for the multi-scale assessment of Padua (FUA and Municipality) were provided by the Municipality of Padua and the Italian Institute of Statistics (ISTAT 2011).

GIS and statistical analysis were carried out in GRASS-GIS 7.8 (GRASS Development Team 2020) and Python (Python 2020).

Results are reported in a multi-level perspective. Fig. 2 shows the workflow with a focus on the three key indicators: the average greenest value in 2010 (C.2); the change per decade (C.4); the green balance (C.5). For each level, the reporting units were chosen to fit the specific purpose of the analysis.

1. European level, 696 FUA (EU-28) were analysed;

1. Metropolitan level: the FUA of Padua, was analysed with reference to the core city and the 30 municipalities of the commuting zone.

1. Municipal level, results were reported at the district and the public parks level. Districts are important intermediate sub-administrative units for local level actions and activities (Colsaet et al. 2018). Public parks were included for demonstrative purposes.

Table 3 shows the percentage of pixels, in core cities and commuting zones, with significant trends. At the three levels, the share of significant pixels (i.e. pixels that observe a significant change) is consistent amongst the reporting units and in line with other authors; for instance, Teferi et al. (2015) and Novillo et al. (2019), who implemented similar methodologies. At the EU level, core cities had, in total, 49.56% of significant pixels. The average value amongst the 696 cities was 54.12%, with a maximum of 90% and a minimum of 6%. The commuting zones (608 because there are Functional Urban Areas which do not have a commuting zone) had in total 52.4% of significant pixels, with an average value of 54.18%, a maximum of 92.43% and a minimum of 10.18%.

At the public green spaces level, 1396 pixels (125.64 hectares) with significant values were extracted and evaluated.

Average Greenest 2010

Change in vegetation cover inside UGI

Balance between abrupt greening and browning

Public green spaces level

Three public green areas in Padua

The analysis confirms previous studies that demonstrated a decline of urban green with the increase of urbanisation (Fuller and Gaston 2009, Corbane et al. 2018). Dallimer et al. (2011) showed that in nine out of thirteen UK cities, a decline in urban green occurred since 2000, when policy reform favoured urban densification. A larger study analysed the trend of urban green in 202 European cities (Kabisch and Haase 2013) from 1990 to 2006, identifying an overall decline in urban green spaces in most Eastern cities, accompanied by an increase in residential areas.

We documented a general relative decrease in vegetation cover inside the most urbanised areas in Europe between 1996 and 2018. The vegetation cover remained, indeed, relatively stable in the long term, with a slight upward trend in not-densely built areas, either in core cities and commuting zones. However, when considering the difference between abrupt greening and browning, cities are characterised by a negative balance that can be interpreted in terms of an absence of clear green compensation policies and an imbalance between urban development at the expense of green spaces.

In addition, the loss of vegetation is part of a more complex dynamic, characterised by a progressive densification of settlements, the movement of population towards urbanised areas (European Environment Agency 2019, Maes et al. 2020) and a relatively low share of tree canopy coverage in European cities (average of 20% inside FUA and Core cities, with an extremely diverse distribution amongst European cities).

The FUA of Padua well represents the effects of the long urbanisation process that characterises north-east Italy. Urban expansion, developed since the 1950s (Romano and Zullo 2016, Romano et al. 2017, Rizzo et al. 2017, Romano et al. 2020), characterises the last twenty years. Besides the loss of traditional agricultural landscapes and forests, the area experienced a process of progressive densification with a loss of urban green spaces, especially in densely buit-up areas. Currently, 40% of the municipalities of the FUA of Padua miss vegetation cover in the densely built areas. When vegetation is present, it tends to be stable over time. Considering abrupt greening and browning, 93.5% of the municipalities show a negative balance, especially in densely built-up areas. The average values extracted to represent the balance between greening and browning in the FUA of Padua are sliglitly higher than the EU level average.

At the municipal level, urban districts show an uneven distribution of UGI in terms of size, greenness and direction of change in DB and NDB zones. Interestingly, in DB zones only District 3 presents a negative balance. On the other hand, in NDB zones, a negative balance characterises districts 2, 3 and 4 (see Suppl. material 3). District 2 is a very dense and populated residential area which experiences a surge in population growth (see Suppl. material 3). District 3 is characterised by a residential area, with high (and increasing) population density, (see Suppl. material 3 in census blocks bordered by districts 4 and 1) and by an economic area with a low population density, but an intense urban development. District 4 is a residential district still characterised by peri-urban agricultural land, which is experiencing an increase of population and urban development (see Suppl. material 3). This trend may explain the loss of green and may confirm other studies that found that changes in greenspace occurred in the urban-rural periphery, coincident with urban expansion (Seto et al. 2002, Yuan et al. 2005, Portillo-Quintero et al. 2012, Peng et al. 2016).

At the public green spaces level, the method allows us to monitor the management practices within public parks. The vegetation cover in public green is slightly increasing and, in 6.08% of the cases, this represents an abrupt greening. This is the case of the realisation of new parks or playgrounds as presented in the three local showcases, that improved the quality of local neighbourhoods.

Scholars have demonstrated that greenness in the proximity of residential areas is fundamental for human well-being (Gascon et al. 2016, Dadvand et al. 2016, van den Berg et al. 2016, Tischer et al. 2017, Ponjoan et al. 2021) and that the neighbourhoods are the most suitable spatial unit to analyse green space as this unit matters most to residents’ living quality (Colsaet et al. 2018).

There is an increasing body of evidence that urban ecosystems in good condition contribute to biodiversity conservation (Kowarik 2011, Baldock et al. 2015, Hall et al. 2016, Lepczyk et al. 2017, Egerer et al. 2020, Kowarik et al. 2020) and to human well-being (Gascon et al. 2016, Dadvand et al. 2016, van den Berg et al. 2016, Tischer et al. 2017). Cities and their surroundings are sources of threat for biodiversity (Salafsky et al. 2008, Bowler et al. 2020) and pressures for environmental quality (Environment Agency, Chief Scientist’s Group 2021). However, they can have a role in the policy agenda to halt biodiversity loss and improve ecosystems condition (European Commission 2020). This perspective implies the responsibility for local governments and planning departments towards integrated nature and biodiversity-positive urban planning. According to the EU Biodiversity Strategy for 2030, local actions should prioritise the preservation and expansion of existing urban green spaces, an increase in tree canopy cover (i.e. planting 3 billion trees in the EU by 2030) and restoration of degraded or destroyed urban habitats (i.e. to protect 30% of terrestrial and 30% of marine ecosystems). The analysis at European and municipal level suggested that cities still have a long way to go. The multi-level approach, proposed in this study, aims at measuring progress towards specific targets on urban green set at city or national level. The indicator "change in vegetation cover in UGI" shows great potential in several respects. It could provide information for city planning, policies and actions for enhancing biodiversity, restoring degraded habitats, down to management decisions at green infrastructure level. Since an analysis of greening performance is possible at the scale of a district, the shortage of urban green in specific districts and the potential to increase it could be identified to help achieve aggregated citywide targets.

One of the main challenges in protecting and expanding high quality urban green spaces is conflicting interests in land use and high competition over space, especially in densely built-up urban areas. Haaland and van den Bosch (2015) provide a examples of possible strategies for green space provision in compact city environments, amongst others:

• saving existing urban green space, in particular remnant semi-natural vegetation;
• enhancing quality of existing green space both from a recreational and a biodiversity perspective, especially when no further public green space can be provided;
• providing green space on redeveloped sites; green spaces plan for developments sites should be elaborated before the building plan.

Other strategies should include the preservation of mature existing trees (if in healthy status) (Erlwein and Pauleit 2021); the application of ecological principles when choosing vegetation type (Jim 2013) in public and private green spaces.

An overall rule should be the development of effective urban green strategies developed in coordination with strategic and holistic plans that comprise the entire region (Haaland and van den Bosch 2015.

The Greening-Browning-Balance provides valuable insight into the overall balance between green and grey areas (i.e. infrastructure, housing etc.) in the city. Showing progress towards the before-mentioned policies implies that the balance should be zero, i.e. no net loss of urban vegetation or positive, i.e. increase in urban vegetation. Thus, the indicator is suited to provide information for no net loss policies in terms of city performance and monitoring of distance to target. At the same time, it provides decision support with regards to future planning and development applications and policy instruments. The European Commission's Roadmap to a Resource Efficient Europe (European Commission 2011) introduced an important policy measure which was expected to affect the urban development in Europe: ‘no net land take by 2050' (Barbosa et al. 2017). Despite recent reductions in soil sealing, soil continues to be lost by land-take (Montanarella and Panagos 2021). No net loss of urban green could act as parallel strategy to achieve the ‘no net land-take by 2050’ objective. For instance, following (European Commission 2016) by:

• recycling (areas with uses that were once active and now exhibit no viable use should be recycled by either introducing new uses or through renaturation);
• compensating (compensation should be required when construction must take place on previously un-built land. This can take the form of renaturation projects or de-sealing measures in built areas, where soil sealing is no longer necessary.

Another application of the indicator is policy evaluation. To put nature on a path to recovery in European cities, the collaboration of private parties will be essential, considering that private space, on average, makes up to the largest part of cities. Private spaces play an important role as stepping stones in ecological corridors and providing a diversity of habitats for biodiversity to thrive in cities. The indicators capture both private and public space and could, therefore, be used to evaluate if and to what extent any policy-promoting green spaces creation has taken effect.

In this paper, we analysed the trend of urban green areas in European cities, through a multi-scale approach. This framework provides an indicator to report the status of UGI under the Green City Accord. Furthermore, the scripts implemented in GEE deliver a simple, flexible and accessible tool, easily available for researchers, administrators and stakeholders, which is suitable to provide a policy-orientated instrument. The results of the study highlighted criticalities in the status and trend of UGI in Europe, as urbanisation is continuously increasing and compensation measures are currently lacking. The current picture of the situation is showing an imbalance between green areas and impervious surfaces, with an increase in the latter at the expense of the first. UGI and nature-based solutions more in general, are being recognised as critical assets in mitigating the detrimental effects of increasing environmental stressors, such as climate change and in counteracting the ecological crisis. For the first time, the importance of urban ecosystems for nature protection is emerging, with a growing body of policies aimed to protect and implement UGI. Cities nowadays can have a constructive role for nature protection, ecosystem restoration, biodiversity and ecosystems services, rather than being exclusively considered as sources of pressure or threats. Therefore, in view of the present and future measures that need to be undertaken in the EU, aimed to systematically integrate UGI in cities, it is essential to monitor changes in vegetation within a city, with instruments capable of providing timely and accurate information for policy support. It is worth noting that UGI within cities are multi-functional systems capable of providing different ecosystem services. Therefore, an effective management of UGI must take into account their inherent complex nature and be able to identify the most appropriate strategy, but this is dependent on the availability of site-specific and spatially-explicit information.