Europe’s greenest small cities refuse to follow one winning formula

The cities that win the European Green Leaf Award, the European Union’s prize for small and mid-sized towns with outstanding environmental performance, turn out to be strikingly different from one another. A new analysis of the 18 award-winning cities between 2015 and 2026 finds no single model of a green city. Instead, the winners cluster into distinct environmental profiles, each shaped by different policy priorities and local conditions, according to a study published in Computational Urban Science.

Dalma Schmeller of the study set out to answer two questions: what patterns of environmental characteristics can be identified among the award-winning cities, and do they converge toward a common environmental model or follow multiple sustainability pathways? The answer, based on a multivariate statistical analysis of indicators spanning the award’s seven thematic areas, is clear. The winning cities do not converge toward a single model but exhibit distinct environmental profiles, the study reports, with results that may inform municipalities pursuing the green transition, especially in small and medium-sized cities.

The European Green Leaf Award, launched by the European Commission in 2015, targets settlements with populations between 20,000 and 99,999, a bracket that contains 3,730 potential applicant settlements across EU member states, candidate countries and European Economic Area members. Its sister award, the European Green Capital Award, established in 2008, covers larger cities above 100,000 people. Between 2015 and 2026, 86 settlements applied for the Green Leaf prize and 19 were awarded it, a maximum of two winners per year, representing 22 percent of applicants. Applications came mostly from Southern Europe and the Balkan Peninsula, yet it was generally settlements from Western and Northern Europe that advanced to the final round or won, and no settlements from Austria, the Czech Republic, Iceland, Liechtenstein, Norway, Slovakia, Switzerland and several Balkan countries applied at all. Low visibility may partly explain this: in a survey of 49 Norwegian municipal and regional council members cited in the study, only 8 percent had heard of the award.

To compare the winners systematically, Schmeller processed the cities’ award application documents, supplementing them with data from European, national, regional and municipal databases. Priority was given to the most recent post-pandemic data from 2021 to 2023, and values reported by statistical offices were cross-checked against application forms. For six municipalities whose applications were not publicly available, the author contacted city authorities; two, Lappeenranta and Winterswijk, provided documents, while for Mechelen, Limerick, Gabrovo and Valongo all indicator values came from external databases and urban development documents. Membership in international networks such as the Covenant of Mayors and the Green City Accord was combined into a single composite indicator counting each city’s memberships and signatory agreements.

The statistical core of the study is Multiple Factor Analysis, a dimension-reduction technique built on principal component analysis principles that groups correlated variables into new common dimensions while balancing the unequal numbers of variables across the award’s thematic groups. Four dimensions were retained, explaining 65.5 percent of the total variance, with the number chosen by the Kaiser criterion of eigenvalues above one. The dimensions are empirical combinations of indicators rather than theoretically coherent constructs. The first dimension loads heavily on drinking water consumption, the presence of noise maps, population density, energy consumption, waste quantity, green area size, PM10, bicycle path length and PM2.5. The second is strongly characterized by organizational memberships, with additional contributions from CO2 emissions, recycling rate, drinking water consumption, bicycle path length and the proportion of land under Natura 2000 protection. The third is defined mainly by waste generation, recycling rate, memberships, noise maps, Natura 2000 share and CO2 emissions, while the fourth is dominated by energy and drinking water consumption, noise maps, the proportion of cyclists and bicycle path length.

The interpretation of these dimensions reveals their ambivalent character. A high score on the first dimension indicates a city that generally lacks a noise map, has high per capita energy and drinking water consumption and generates high per capita waste, all unfavorable for environmental sustainability, yet the same dimension is also associated with larger per capita green areas and longer bicycle path lengths, which are beneficial. High scores on the second dimension correspond to high per capita CO2 emissions and fewer organizational memberships, alongside lower drinking water consumption and less Natura 2000 area, but also longer bicycle paths and higher recycling rates. This ambivalence, the study notes, complicates interpretation, which is why the author additionally constructed thematic rankings of the cities based on scale-type variables, assigning rank one to the most favorable raw value for each indicator and averaging ranks within each thematic category.

K-means clustering applied to the standardized MFA dimensions then grouped the 18 cities, with the optimal number of clusters determined by the Silhouette method. The seven-cluster solution achieved the highest silhouette value, and it produced strikingly uneven groups. Cluster 3 contains only Vaasa in Finland, which shares general characteristics with the Portuguese cities of Águeda and Torres Vedras but differs in magnitude, with fewer memberships, higher energy consumption, higher per capita municipal waste, lower annual PM10 values, a higher proportion of Natura 2000 areas, a higher share of bicycle users and greater bicycle path length. Cluster 7 likewise consists of a single city, Horst aan de Maas in the Netherlands, a low-density municipality characterized by high CO2 emissions, a high share of car users, long bicycle paths, substantial green area, high waste generation, high recycling rates and high energy consumption. Cluster 2 groups Leuven, Mechelen and Winterswijk, whose common features include the number of cars, modal shares, bicycle path length, green area coverage, Natura 2000 proportion, waste generation, recycling rates, noise maps and certain memberships. Cluster 5 pairs the Finnish city Lappeenranta with Sweden’s Växjö, two northern inland settlements with similar climates whose similarities are most evident in low population density, favorable air quality, high energy consumption, modal share, car ownership and memberships.

Geography emerges as a powerful organizing force in some regions but not others. Portuguese cities and those in the Benelux countries form clearly defined clusters, and all cities in Cluster 6, which comprises Cornellà de Llobregat, Elsinore, Mollet del Vallès and Valongo, are coastal. Cluster 4, containing Galway, Limerick, Treviso, Velenje and Viladecans, is the most geographically heterogeneous group, spanning coastal, inland, mediterranean and oceanic climates. The study attributes the regional clustering among Portuguese, Belgian and Dutch cities to national and regional development regulations and similar climatic conditions, yet Scandinavian cities with comparable geography and climate did not cluster together, a divergence the discussion links to local political ideologies, since Scandinavian urban development is described as prioritizing nature-based solutions and environmental protection over economic interests.

Bicycle infrastructure sharply separates the groups. All cities in Clusters 1, 4 and 6 of the seven-cluster model have bicycle path lengths below 1.50 meters per capita, and except for Torres Vedras these cities have roughly 13 square meters of green space per capita, with annual per capita energy consumption between 4.2 and 10.4 megawatt-hours. In the alternative four-cluster model, which avoids single-city clusters, the Portuguese trio of Águeda, Torres Vedras and Valongo stands out for low CO2 emissions, shorter bicycle paths below 0.70 meters per capita, low particulate matter concentrations, low municipal waste generation and the highest daily water consumption, while the cluster of Lappeenranta, Leuven, Växjö and Winterswijk, all cooler-climate cities, shows the highest per capita annual energy use. Cities in another cluster all have population densities exceeding 1,300 inhabitants per square kilometer and the most prominent walking rates, while Vaasa and Horst aan de Maas have among the lowest public transport shares, 5 percent and 3 percent respectively, complemented by a relatively high share of cyclists and top-three per capita green area.

Some results appear contradictory at first glance. In Lappeenranta and Torres Vedras, high car usage above 60 percent coincides with NO2 annual mean values below the World Health Organization guideline of 10 micrograms per cubic meter. The study offers several possible explanations, including a high proportion of commuters that could leave minimal traffic outside peak hours, a high share of electric vehicles, proximity to highways, efficient traffic management, low-emission zones, and coastal locations where stronger wind activity disperses pollutants faster. Monitoring stations are not always situated along the busiest roads, the author cautions, potentially biasing annual mean values. The reverse pattern appears in Viladecans and Cornellà de Llobregat, where low car usage below 50 percent corresponds to high NO2 annual means above 10 micrograms per cubic meter, possibly reflecting internal combustion engine traffic, industrial activity, fossil fuel use and biomass burning, with both cities located in the Barcelona metropolitan area where traffic concentrates along specific routes.

The study’s limitations are candidly stated. The sample is small because the award has existed only since 2015, applicants are unevenly distributed across space, and MFA cannot handle missing data, which forced the exclusion of the Bulgarian city Gabrovo and further reduced the sample. Not all winning cities make their applications publicly available. Yet even within this limited dataset, the conclusion holds: significant differences exist in environmental performance among the winners, and the practical implementation of the green city concept can take multiple forms, with some cities excelling in specific thematic areas while performing less strongly in others. At the same time, the study observes signs of convergence in knowledge transfer, adoption of policy frameworks and standardization. The broader lesson for urban policy is that assessing environmental performance requires complex interpretation of indicators that considers multiple factors, because examining individual indicators in isolation may be misleading when local conditions, regulatory frameworks or regional effects strongly influence outcomes.

Subject of Research: Environmental performance patterns and clustering of European Green Leaf Award-winning cities

Article Title: Urban sustainability through multiple pathways: insights from the European Green Leaf Award winning cities

Article References: Schmeller, D. (2026). Urban sustainability through multiple pathways: insights from the European Green Leaf Award winning cities. Computational Urban Science, 6(1), Article 46. https://doi.org/10.1007/s43762-026-00282-5

Image Credits: AI Generated

DOI: 10.1007/s43762-026-00282-5

Keywords: urban sustainability, European Green Leaf Award, green cities, Multiple Factor Analysis, k-means clustering, environmental indicators, small and medium-sized cities, benchmarking, Natura 2000, air quality, cycling infrastructure, European Commission