Quantification of areas covered by vegetation in urban centres of major Italian cities using aerial images (orthophotos)

Objectives and Context

The presence of vegetation in an urban area is particularly important for the quality of life in cities. Numerous studies have shown that access to green spaces reduces stress, improves mood and promotes a more active lifestyle.

The statistical measurement of its distribution across the territory can help assess whether these benefits are accessible to the entire population. Information on the distribution of vegetation is also useful for environmental monitoring, another factor that affects quality of life, as vegetation influences air quality, helps mitigate urban heat, and contributes to the absorption of fine particulate matter. Finally, this information is also valuable for the implementation of sustainable urban planning, helping decision-makers to pursue, for example, the goal of an equitable distribution of green areas across neighbourhoods. All these considerations have led to focusing the analyses on the built-up areas of the 14 metropolitan cities (in order to exclude agricultural and forested areas), as they represent the main territorial domains where people live and where the presence of vegetation can have an important impact on the quality of life by mitigating the consequences of high urbanisation.

The proposed statistics on the quantification and distribution of vegetation in the main urban centres were obtained by integrating the digital cartography available at Istat with remotely sensed imagery, a type of data source particularly rich in information but still scarcely used within National Statistical Institutes. The images used are high resolution georeferenced orthophotos (20 cm/pixel) acquired by AGEA and available for each region on a three-year basis from 2005 to 2023.

This data source was preferred to satellite imagery due to its high spatial resolution, a key aspect in the context of urban areas. The applied procedure classifies the pixels of the orthophoto images as either ‘vegetation’ or ‘non-vegetation’. The classification relies on machine learning techniques that process the radiometric indicator NDVI (Normalized Difference Vegetation Index), based on the spectral behaviour of chlorophyll. The proposed procedure, described in detail in the methodological notes, is one of the main outcomes of this study, as it has the not negligible advantage, particularly relevant in the context of official statistics, of being easily replicable, thus allowing the production of indicators that can be analysed over time.

For the quality assessment of the produced statistics, a sample of pixels was selected on which the classifications obtained through the proposed algorithm were compared with the photo-interpretation carried out by a domain expert through visual assessment. The two classifications showed a high degree of agreement, providing reassurance about the reliability of the obtained results.

Based on the estimated classification, it is possible to calculate indicators describing the amount and distribution of vegetation in urban areas. The indicators presented in the tables below are described as follows:

1. Total green percentage: this is the percentage ratio between the total green surface area and the surface area of the built-up area. It represents an indicator of the territorial ‘density’ of vegetation within the urban area.
2. Total green percentage for areas ≥100 m²: this is the percentage ratio between the total green surface of areas larger than 100 m² and the surface area of the built-up area of the main urban centre. This indicator is useful for quantifying green areas excluding very small surfaces, which are negligible from a statistical point of view.
3. Variation in total green area between two flights: this indicates the change in the percentage of total green area for surfaces larger than 100 m². It is important for understanding temporal changes in green coverage within the main urban area. The time interval between two flights is three years.

Quartiles of the distribution of green area sizes: for all green areas with a surface larger than 100 m², the distribution of their size (in m²) is calculated, and the first, second, and third quartiles (Q1, Q2, Q3) are reported. This indicator is useful for quantifying the size of green areas within the main urban centre and provides a measure of their fragmentation