Semi-Automatic Method to Evaluate Ecological Value of Urban Settlements with the Biotope Area Factor Index: Sources and Logical Framework

Abstract

As the number of people living in cities continues to increase and as their needs continue rapidly to evolve, planners and scholars have been encouraged to define what constitutes high levels of quality of life in urban settlements. The relationship of an area’s inhabitants with natural and green resources increases urban environmental value, which is one of the most relevant aspects in the determination of the quality of life in built-up contexts. Moreover, it is fundamental to find quantitative parameters that can monitor the development of planning processes, working together with natural systems. The authors present a comparative method that can be used to analyze and evaluate the ecological value of urban settlements, using a semi-automatic process that is based on calculating the biotope area factor (BAF) using different open-access databases (a cartographic dataset, aerial imagery, and Sentinel-2 images). Two different Italian case studies that are set in the Milan metropolitan area are presented. In this paper, the authors describe the two settlements using the city-planning parameters of physical structure and morphology; they show the ecological differences and similarities throughout the various remote sensing sources and data. Finally, the authors indicate how the research can be developed, highlighting the weaknesses, the potentiality, the replicability process, and the urban planning implications of the methodology.

1. Introduction

In a constantly changing urban context, where the traditional, new, and always varied needs of inhabitants impose flexible and resilient physical structures and relationships in urban and territorial areas, it is vital to find a dynamic balance with natural systems: thus, it is mandatory to improve the quality of the urban environment and of urban development processes through the conscious use of resources (land use, energy, and environmental peculiarities).

With the most recent population projections seeing an increasing trend until 2030, with 60.4% living in cities and in urban agglomerations [1,2], the urbanization process is one of the strongest forces that drive land use and land-use changes and, consequently, impacts climate change, climate warming, and microclimate degradation [3].

Considering that the ecological environment is closely related to and influences human health and life, it is important to analyze as well as assess urban environmental value, which represents one of the most relevant aspects of quality of life [4]. Monitoring the urban ecological environment, its quality, and its value is a complex task because there are many interrelated parameters that are involved and that are linked to natural, social, economic, and environmental factors: housing, jobs, level of education, civic engagement, life satisfaction, work-life balance, health, safety, vegetation density, greenness, leaf area index, heat island, temperature, population density, building density, noise, air pollution, land use, land cover, water quality, land quality, solid waste management, green and open spaces, accessibility to roads, etc. [5,6].

In particular, public urban green spaces (municipal parks, public playgrounds, boulevards, areas for walking dogs, etc.) are an important factor for urban quality of life, as they provide various ecosystem services: mitigation of the urban heat-island effect, increase in urban resilience in the case of natural hazards (i.e., floods), accessibility to open spaces for citizen activities, and the improvement of people’s well-being and mental health [7,8,9,10,11,12,13,14,15,16].

Moreover, city planners need to have the necessary information about the location, qualities, and ecological efficiency of urban green spaces: mapping and definition of the real location, dimensions, and characteristics of those areas are fundamental. Regional and municipal datasets focusing on green spaces usually only contain areas that are owned and maintained by the city; private green spaces that are accessible to residents or small groups of people (e.g., private gardens belonging to apartment buildings) are usually not included in these datasets and indicators [17,18].

1.1. Aim of the Research and Current Paper Structure

The main goal of the present research is to assess the specific results that emerge from various open data sources when defining the ecological value of urban settlements, using a semi-automatic process that is based on the biotope area factor (BAF) calculation. The different sources of data that are used in this calculation are official cartographic datasets, aerial imagery, and Sentinel-2 images. The final result aims to highlight the variations in BAF value among the selected sources; each source represents a territory with specific features highlighted, depending on the survey techniques.

The paper is structured into five distinct conceptual parts: after a brief general introduction, the authors define the main aim, discuss the state of the art, and describe the study areas in which the research takes place. Subsequently, the authors describe the main methodology and the different datasets that are used for the analysis. Later, the authors present the different ways in which the BAF can be calculated in relation to the different datasets adopted. Section 4 shows the results that were obtained from the BAF calculation, providing tables and figures to better explain the relationship between the acquired numbers and the actual situation in the study areas. Finally, after the first results have been observed, the authors discuss the accuracy of the method (ranges of error due to database errors and the single operation performed, and the reliability of data) and the possibility of replicability in different contexts and at different territorial scales (from neighborhood, to municipality, to regional scales) while also considering the implications that the method has on urban planning.

1.2. State of the Art

Among the many international databases on land uses and urban functions, there are few that cover all of the typologies of green areas. For example, the European CORINE Land Cover data set contains “Green Urban Areas” but only includes green spaces larger than 25 ha [19,20], or the US-based dataset, the Trust for Public Land’s Park Serve dataset [21], which contains higher resolution land use information, but only for selected cities.

The importance of ecological efficiency in urban contexts has become clear over many years: the problem of excessive soil sealing, and all of its consequences require the implementation of specific measures, so the introduction of standards and indexes that have been finalized for the regulation of soil permeability allow flexible and resilient solutions to integrate urban planning issues and the ecological aspects [22,23,24,25,26].

Among these solutions is the biotope area factor calculation, which was defined in the Landscape Plan of Berlin [27].

In the early 1980s, the Landscape Program for West Berlin was introduced in the Berlin area and aimed to protect nature, natural resources, the landscape, and collective green spaces in urban areas. The Landscape Program attempted to identify new ways in which special urban planning could be implemented and could also increase urban environmental performance without losing the characteristics of the existing city. In the 1990s, the Berlin government developed the biotope area factor (BAF) model through the Landscape Plans as a landscape planning index, in order to reduce ecological deterioration in urban areas, enhance urban ecosystems, increase the sustainability of city development, preserve the local microclimate, give due attention to land and water use, improve the quality of plant life and animal habitats, and improve urban living spaces for human beings [27,28]. Moreover, the BAF index was also defined to control and regulate the construction and renovation of buildings in densely built-up areas; it was specifically developed to be used at a very detailed scale, such as building, parcel, and urban district scales, in areas with various land uses (commercial, residential, infrastructure, industrial).

The BAF is the ratio of ecologically effective surface area to building site area (1):

BAF = (ecologically effective surface area)/(building site area)

(1)

The ecologically effective surface area is the sum of the areas with different surface types, multiplied by their associated ecological weights, which are assigned according to the specific characteristics of those surfaces (

Table 1. Types of surfaces and weighting factors per sq. m. [27].

Types of Surfaces	Weighting Factors
Sealed surfaces	0.0
Partially sealed surfaces (e.g., clinker brick, mosaic paving, etc.)	0.1
Semi-open surfaces (e.g., sand, gravel, etc.)	0.2
Greened surfaces (e.g., gravel with grass, wooden cobbles, etc.)	0.4
Surfaces with vegetation, unconnected to the soil below, shallow substrate thickness (20–40 cm of soil coverage)	0.5
Surfaces with vegetation, unconnected to the soil below, medium substrate thickness (41–80 cm of soil coverage)	0.6
Surfaces with vegetation, unconnected to the soil below, deep substrate thickness (81–150 cm of soil covering)	0.7
Surfaces with vegetation, unconnected to the soil below, very deep substrate thickness (>150 cm of soil covering)	0.9
Surfaces with vegetation, connected to the soil below	1
Rainwater infiltration per m2 of roof area	0.2
Water surface (rainwater-fed water surface. Through the establishment of vegetation, the BAF can increase to 0.6)	0.5
Vertical greenery with connection to the ground	0.5
Vertical greenery without connection to the ground	0.7

).

The main criteria that are used to assign the ecological weights are the “high efficiency of evapotranspiration, ability to capture water from soil and its storage, ability of powder fixation with a reduction of suspended dusts, conservation and long-term development of soil functions (filtering, buffering and transformation of pollutants and hazardous substances), availability of suitable habitats for plants and animals” [29].

The BAF index ranges from 0 (completely impermeable surfaces or waterproof surfaces, such as buildings, streets, or parking lots) to 1 (complete permeable surfaces, i.e., green lawns or agricultural land) and includes 16 classes (Table 1). Finally, the intermediate classes of the BAF index refer to vegetation areas that have more or less connection with the underlying soil.

The definition of the BAF is connected to the current situation of urban settlements as an analysis parameter, as well as the planned intervention of requalification or new constructions as an ecological target that has to be reached, in order to guarantee the sites’ sustainable performance regarding different urban functions [30].

On a worldwide scale, there are various parameters that municipalities and regions might use to assess the environmental quality of their territory. From building an ecological network to nature-based solutions (NBS) [31,32,33,34,35], in the last few decades, many policies and practical indexes emerged. In this study, we cite the Seattle green factor [36] and the Bologna RIE—riduzione impatto edilizio (building impact reduction index) [37]. Both models are useful to improve the environmental quality of existing cities and new settlements, based on building ecological characteristics and water management. On the one hand, the building scale is useful to keep track of the small-scale dimension of city construction; on the other hand, the small details are not easily captured in satellite and aerial images. The BAF can be considered a good approximation of land use typologies, both on a block scale and for wider areas of the territory [38].

Moreover, since land-use maps describe the geographic distribution of natural resources and can be used as important input data to support decisions, their accuracy is crucial in planning processes. Territorial areas, landscapes, and, especially, urban agglomerations present unevenly distributed land uses and land cover types across a specific area. This spatial heterogeneity creates a complex problem that is related to the creation and interpretation of satellite images [39,40,41].

As such, in order to perform more specific, automatic, objective, and precise evaluations of urban environmental quality and value and to analyze all the green spaces in an urban context, remote sensing has become an effective and efficient tool that is able to record a variety of spatial and temporal data on land surfaces, provide real-time data and large-scale monitoring, and recognize spatial-temporal changes in regional eco-environments [42,43,44]. Due to the complexity of the ecological system, through the use of remote sensing data, several ecological indexes have been proposed: the normalized difference vegetation index (NDVI), which was proposed to monitor environmental change; the land surface temperature (LST) model, which was proposed to assess the urban heat island effect; the normalized difference built-up area index (NDBI); and the normalized difference water index (NDWI) and modified NDWI (MNDWI), which were proposed to extract data regarding water bodies, etc. [7,45,46,47,48,49,50].

In the last decade, the availability of free satellite images with a high spatial and temporal resolution creates new opportunities: the Copernicus Sentinel-2 mission, operated by the European Space Agency, provides satellite imagery that is designed for land monitoring applications, such as agricultural activity analysis, vegetation mapping, water body detection, built-up area mapping, urban green spaces analysis, and so on [51,52,53,54].

2. Methodology

The authors have defined a method that is based on the application of different data sources, which can be used to calculate the ecological index of the biotope area factor (BAF).

Technically, as shown in Figure 1, the first step of the method necessitates BAF calculation using existing cartography and aerial imagery, in order to define the land use maps that are necessary to calculate the BAF index in detail. Generally, BAF calculation is carried out following this data-crossing procedure, using cartography, aerial imagery, site inspections, surveys, and panoramic images. Then, soil classification usually takes place manually, an operation that is very demanding in terms of time.

To overcome the rigidity of each operation, as well as the lack of uniformity and objectivity, and to reduce execution times, the authors have used other data sources, such as Sentinel satellite images and aerial imagery, which, with semi-automatic methods, make it possible to calculate the BAF in a simplified way. The low resolution of the images does not allow the types of surfaces to be classified in detail, but the data processing highlights an ecologically effective system. To follow the same method, the real BAF (developed in the classical way) was translated into a simplified BAF.

The last step provides verification of the ability to measure the BAF correctly: the simplified BAF maps that were produced were compared to a real BAF map.

Figure 1. Flow chart methodology.

Figure 1. Flow chart methodology.

2.1. Study Area

The study area covers the metropolitan area of the city of Milan (Lombardy region), which is located in the north of Italy and specifically covers the municipality of Abbiategrasso and the municipality of Segrate (Figure 2).

Abbiategrasso is a city located in the south-western quadrant of the metropolitan area of Milan.

The municipal area, one of the largest in the Province of Milano, covers 47.778 sq. km. and had a population of 32,855 inhabitants in 2020 [55].

This is a mainly agricultural area; only a small part of the territory (about 30%) is urbanized [56]. Among the farmsteads that are scattered around the municipality, there is an urban center that has developed over time, due to expansion bands.

The city has Roman origins, but it only began to consolidate and develop in the late Middle Ages. The urban structure of the city is clearly recognizable, as the historic center is rectangular in shape and is bounded by the walls of Visconti’s castle. Outside the medieval walls, there are the 19th’s districts and the more recent residential and industrial settlements.

For the analysis of the data, we considered an area of 1 sq. km. for both municipalities.

In Abbiategrasso, we considered the north-east quadrant of the city, including the urban fabric of the historic center (Figure 3).

The municipality of Segrate is to the east of the city of Milan.

The entire territory has a total area of 17.494 sq. km. and is largely urbanized, with a population of 36,579 inhabitants [55].

The city of Segrate, begun as a set of different agricultural settlements, developed in parts during the 20th century as a mainly residential center supporting the metropolitan city of Milan.

The urban fabric of the area is highly fragmented: Segrate is made up of many separate and recognizable urban centers that make up the districts of the city.

Our study is focused on the district of Milano Due (Figure 4). It is located in the north portion of the municipality and is a residential neighborhood that was built in the 1970s. It is a good example of a modern garden city and has a medium-high density that is characterized by high buildings and wide green open spaces [57].

The study area in Abbiategrasso covers part of the consolidated urban fabric of the city center and part of the mixed fabric in the northeast quadrant of the city (Figure 5).

To begin, we considered different urban fabric types. The compact fabric is enclosed in the quadrilateral of the historic center and is bordered by a green system that starts from the moat of Visconti’s castle and continues along a linear path to the station. In the northern part, the urban fabric tends to open progressively, starting from the center. Beyond the railway line, urban expansion can be observed in the east. It is a medium-density urban fabric that is characterized by different building types and functions, ranging from residential to industrial.

On the other hand, Milano Due is more recent, and the district demonstrates a different developmental structure than the other districts. It is characterized by an open urban fabric, where the buildings are surrounded by green areas (Figure 6).

The district develops along a sinuous central road, where the main services can be found. The fabric is mainly residential and is characterized by high buildings. Productive and commercial settlements are only found in the southern part of the district and border the road.

Abbiategrasso and Milano Due have very similar urban parameters despite having a different urban morphology (

Table 2. Main urban planning data about the study areas of Abbiategrasso and Milano Due.

Urban Planning Parameters	Abbiategrasso	Milano Due
Buildings	237,363 sq. m.	217,042 sq. m.
Primary urbanization	180,937 sq. m.	268,211 sq. m.
Green areas	307,081 sq. m.	439,321 sq. m.
Remaining areas	274,618 sq. m.	75,426 sq. m.
Territorial surface	1,000,000 sq. m.	1,000,000 sq. m.
Built-up area	237,363 sq. m.	217,042 sq. m.
Coverage ratio	0.24 sq. m./sq. m.	0.22 sq. m./sq. m.
Volume	1,380,127 cu. m.	2,905,755 cu. m.
Volumetric density	1.38 cu. m./sq. m.	2.91 cm/sq. m.

).

These similarities allow these different urban areas to be compared for BAF calculation.

The only parameter that is different is the volumetric density, which, however, does not affect BAF calculation.

2.2. Topographic Database

The topographic database (DBT) is a geographic database that is made up of various types of digital territorial information that represent basic cartography and are used for urban and environmental planning processes.

The database derives from the directives of the Lombardy region, starting in 2005 with regional law LR 12/2005, which defined the abacus (of grammar and semantics) for all municipal urban plans [58].

The DBT was created using a photogrammetric method. The natural and artificial elements of the territory that were taken from aerial photographs were digitized according to structures conforming to points, lines, and polygons, and each one was classified in a detailed way, according to alphanumeric attribute tables [59].

The tree structure that characterizes the DBT is a major source of information for the study and analysis of urbanized areas. The main macro-categories of DBT concern viability, hydrography, buildings, and vegetation cover, which is subdivided into woods, urban green areas, and agricultural land.

For some categories, there is a secondary level that can be used to better specify the nature of each feature: for example, for the buildings category (020102), the secondary level includes the different types of buildings (considering their morphology and their constructive system) (EDIFC_TY) and can take a tenth of different values; moreover, for the vehicle circulation area category (010101), its secondary level describes the materials used for each element (AC_VEI_FON): asphalt, cement, stones, gravel, slabs of stone, ground, grass, opus incertum, interlocking paving bricks, etc. [29].

The DBT is the reference system that is used to categorize this large amount of information and geographic data. Access to the DBT is free, making access to this data free for sharing and consultation purposes.

In our case, the data were downloaded through the Lombardy Region Geoportal [60].

Nevertheless, some issues can arise when using the DBT that are the result of the cartographic processing that must be carried out.

For our work, the amount of information that is present in the DBT map was excessive in some respects, thus leading to the need for data-cleaning operations.

However, certain limitations exist when using the DBT that are linked to the availability, completeness, and accuracy of the data. It is inevitable that databases are not always updated, and this could present gaps [61]. To overcome this issue, we used panoramic aerial images (referenced to July 2020) to verify and implement the DBT data.

Some of the DBT data that were retained for our study are classified at a lower level of detail than some of the other features. For instance, private courtyards are not classified in the paving category (sand, gravel, asphalt, etc.); to provide an appropriate assessment, the data were implemented and updated by means of site inspections and other open-source information sources (visual analysis of aerial imagery).

2.3. Sentinel-2 Satellite Data

The Sentinel-2 mission, which was developed and is operated by the European Space Agency (ESA) as part of the Copernicus Program, consists of two polar-orbiting satellites: Sentinel 2A and Sentinel 2B, which were launched in 2015 and 2017, respectively.

The Sentinel satellites can acquire 13 spectral bands that range from the visible and near-infrared (NIR) to the shortwave infrared (SWIR) bands.

They capture multispectral images that are used to monitor land surface conditions, with a spatial resolution from 10 m to 60 m, depending on wavelength.

Monitoring land surface usage is important to detect changes over time: the two satellites are designed to have a high revisit time, spending 5 days at the Equator (2–3 days at mid-latitudes) [62].

Satellite data must be processed to provide information, for example, by combining bands to produce false-color images, which enhance land cover such as vegetation.

The Sentinel-2 data products are available free of charge from the Copernicus Open Access Hub [63].

When choosing which satellite images to download, it is necessary to consider the number of clouds in each image, known as cloud cover. Sentinel-2 images are sensitive to cloud cover: clouds can disturb the reflectance signal and affect the data. It is important to analyze and process satellite imagery on days with a low cloud cover or, if that is not possible, there are several methods that can be implemented to digitally remove clouds [64,65,66,67,68].

In our study, it was not necessary to remove clouds from multispectral Sentinel-2 images because, in each month analyzed, we chose those satellite images that have a low cloud cover. The number of clouds is irrelevant in the chosen images.

2.4. Aerial Imagery

In addition to satellite images, the use of aerial imagery allows one to have a complete view of the territory, directly distinguishing between green and non-green areas.

Aerial imagery refers to all imagery taken from an airborne craft (which can include drones, balloons, or airplanes); aerial photography can be classified into three different types, according to the camera axis (vertical, low-oblique, and high-oblique images), the scale of the image, and the type of sensor used. When the camera axis is in a vertical position, it is called a vertical (or nadiral) aerial photo: this is the most commonly used type of photo in mapping applications. The oblique ones are taken at an inclination of the camera axis with respect to the vertical, so they show the objects in perspective.

This research was focused on the analysis of planimetric aerial images, which have a vertical camera axis. All the aerial images analyzed have a spatial resolution of 50 cm and they include only the visible light bands (RGB).

One of the most significant benefits of aerial imagery is its ability to integrate as a base layer into GIS, CAD, and other applications, to work with it at scale [69,70]. Therefore, open-source aerial imagery was used as a basis for simplified BAF calculation.

Moreover, one of the advantages of using orthophotos is the possibility of improving the quality of the image by changing the contrast and brightness levels. In our case, the green color has been emphasized, and through visual analysis and pixel selection, we have isolated all the surfaces that are covered by vegetation.

2.5. Simplified BAF

It is possible to introduce a “simplified BAF” that only uses a binary 0–1 system: all of the areas with the indexes between 0 and 0.49 stand in class 0 while class 1 includes all of the surfaces that have weighting factors of between 0.5 and 1 [46].

The simplified BAF is an approximation of the real values that are defined by ground truth. The following describes how Sentinel images can only produce binary values.

The use of unconventional sources prevents the traditional BAF method from being applied. The degree of detail of this kind of data is not sufficient to precisely classify the surface types; therefore, it is also difficult to assign the right weighting factors. However, it is possible to recognize environmental and settlement systems using Sentinel images and aerial imagery.

3. Analysis

3.1. BAF Calculation

3.1.1. BAF Calculation from a DBT Source

The procedure for the analysis began with BAF calculations for the areas that were 1 sq. km. in size and that had been identified in the municipalities of Abbiategrasso and Segrate (see Figure 5 and Figure 6).

The topographic database (DBT) was used as the basis for data processing that was carried out for the municipalities of Abbiategrasso and Segrate; the data were downloaded from the free online database of the Lombardy Region [60].

The downloaded DBT data, updated to 2018 (to be precise, 8 January 2018, the last year of data available at the time of data processing), were processed in a GIS environment using QGIS version 3.18.

First, we cleaned the DBT of all of the excess information that was not necessary for BAF calculation. Later, we implemented cartography with the data that were missing using orthophotos and verified them during a direct survey in situ.

The final product consisted of a series of non-overlapping polygons in the GIS, covering the entire study area that was considered in the present research. The polygons were classified using the weighting factors for BAF calculation (Table 1).

In general, a factor of 0 was assigned to the road network and to the buildings, and a factor of 1 was assigned to the environmental systems. The weights for the BAF were detailed after site inspections and after visual interpretation. We checked the types of surfaces that comprised the streets and courtyards, distinguishing asphalt from gravel, and we identified any underground parking lots in which there were gardens.

For example, some streets in the center of Abbiategrasso are cobblestone streets, to which we attributed a weight factor of 0.1, while flowerbeds in the middle of roads were classified with a weight of 0.5, a decision that was reached by considering the shallow soil thickness (

Table 3. Example of correspondence between DBT categories and BAF weighting factors.

DBT Categories		DBT Specifications	BAF Weighting Factors
Level 1	Level 2
010101	0201	Vehicle circulation area	0
020102	0101	Buildings	0
020204	0103	Tennis court	0.2
050393	0105	Gravel	0.2
020104	0104	Flat roof (green)	0.5
060401	0104	Flowerbed (in roads)	0.5
060401	0101	Green urban areas	0.9
060105	010102	Wild grassland	1

).

The first evaluation was the “real BAF”, which could be derived from the DBT information system, together with the precise knowledge of the land use types that were present in the analyzed territory. In fact, it is impossible to verify certain soil characteristics using DBT information, such as underground parking areas with green roofs, private areas with mixed uses (which interfere with the surface typology), green flowerbeds in roundabouts, etc.

As such, the calculation of the “real BAF” is a semi-automatic procedure in which the DBT is not sufficient for a precise evaluation.

Then, we converted the real BAF into the simplified BAF: each surface with a weighting factor of between 0 and 0.49 was assigned a factor of 0; other surfaces with weighting factors of between 0.5 and 1 were assigned a factor of 1.

Because of its binary system, the simplified BAF represents the basis for comparison with the BAF that was calculated from the Sentinel images and aerial imagery, a calculation that was based on the classification of the areas of urbanized or natural surfaces (0–1).

Using this methodology, it is inevitable that there will be errors compared to when traditional BAF calculation is used, but it is necessary to validate the use of the semi-automatic method that has been proposed herein.

We recognize that there is a 10–12% error between the real BAF and the simplified BAF.

3.1.2. BAF Calculation from Sentinel Image Sources

The purpose of a satellite image is to provide information. For our analysis, which was focused on analyzing green spaces, we combined near-infrared, red, and green bands (B08, B04, B03) to produce false-color images. This combination emphasizes the amount of healthy and unhealthy vegetation. In fact, plants reflect more near-infra-red, so in false-color images, they appear as red [62].

Certainly, the use of satellite images, even with a spatial resolution of 10 m, does not provide a sufficient amount of information that can be used to classify urban surfaces in a precise way. Therefore, it is difficult to assign weighting factors for the calculation of the BAF.

However, the use of false-color images allows the environmental systems (shades of red) and the settlement systems to be distinguished. For this, the simplified BAF is useful, as it considers only two types of surfaces: the natural ones with vegetation, which have an ecological weight of 1, and the urbanized ones, which correspond to a weighting factor of 0.

To produce thematic maps of land cover, we used the Semi-Automatic classification plugin, a free open-source plugin for QGIS 3.18 that allows the supervised classification of remote sensing images, using several classification algorithms [71].

In our work, we detected the vegetation changes that took place during the year 2020 by considering those seasons in which there is a greater amount of vegetative activity (spring, summer, and autumn).

The satellite images that have been acquired by Sentinel-2 A were downloaded for the months of April, July, and October of 2020, and images were taken from those days in each month that had less cloud cover.

For each study area, the satellite images were processed in false color using a combination of near-infrared, red, and green bands, with spatial resolutions of 10 m (Figure 7).

This band combination is usually used to assess plant health over time [72,73]. Plants tend to reflect the near-infrared more, so land surfaces that are covered by vegetation appear in different shades of red, depending on the types and the condition of the plants, while buildings, roads, and bare soil can appear in shades of blue or gray, depending on their composition.

After this, we produced land cover maps using the semi-automatic classification plugin in QGIS 3.18.

For our purposes, the simplified BAF calculation was used, and we only defined two macro classes: the urbanized system and the environmental system. The built-up area system corresponds to a weighting factor of 0; the environmental system corresponds to a weighting factor of 1.

Therefore, we selected those areas in shades of blue for the first land cover class, and in shades of red for the second one, as training areas on the false-color images. Then, we used the maximum likelihood algorithm, one of the most common algorithms for supervised classification, to classify the whole study area: each pixel with similar spectral characteristics corresponded to a land cover class; these were developed on the training areas. Following this procedure, we had two color-themed maps (Figure 8), specifically BAF maps, that can be compared to a simplified BAF.

3.1.3. BAF Calculation from an Aerial Imagery Source

Finally, we considered aerial imagery for the calculation of the BAF.

In QGIS 3.18, it is possible to insert background maps, such as Google satellite images, using the Quick Map Services plugin. Therefore, we downloaded the aerial images that had been updated to include 2020 (July 2020).

From the aerial photos, it is possible to calculate the BAF using the simplified method by classifying the types of surfaces into green (vegetation) and non-green (urbanized system) areas.

Later, the aerial photo was modified by stressing the green color: this made it easier to distinguish between the green areas and the areas that did not include any green.

Then, we selected all of the shades of green in the image using a minimum tolerance criterion, and the green pixels were automatically counted. To define the effective ecological surface, we created a proportion between the number of total pixels present in the orthophoto, which corresponded to an area of 1 sq. m., and the number of pixels corresponding to the green areas.

For the BAF calculation, we assigned a weighting factor of 1 to all of the green pixels and a factor of 0 to the other pixels. the shadows represented an obstacle that impeded the detection of actual green areas [74,75,76] (see the Discussion section).

4. Results

In this section, the results that were obtained using the proposed methodology are reported. The results follow the same scheme. For Abbiategrasso we have the calculation of the real BAF (Table 4), the calculation of the simplified BAF (Table 5), the calculation of BAFs from the Sentinel satellite images (Table 6), the average of the Sentinel BAF (Table 7) to compare to the simplified BAF, and the calculation of BAFs from the aerial imagery (Table 8). For Milano Due we have the calculation of the real BAF (Table 9), the calculation of the simplified BAF (Table 10), the calculation of BAFs from the Sentinel satellite images (Table 11), the average of the Sentinel BAF (Table 12) to compare to the simplified BAF, and the calculation of BAFs from the aerial imagery (Table 13).

The BAF map that was calculated using the traditional method employs a color scale that ranges from red, representing sealed surfaces, to green, representing unsealed surfaces (for Abbiategrasso please see Figure 9; for Milano Due please see Figure 13).

On the other hand, the simplified BAF maps only show two colors (for Abbiategrasso please see Figures 10 and 11; for Milano Due please see Figures 14 and 15): these are light blue for categories 0–0.49 and green for categories 0.5–1.

The BAFs that were calculated based on aerial imagery have green areas that are more easily distinguishable (for Abbiategrasso please see Figure 12 and for Milano Due please see Figure 16).

4.1. Abbiategrasso

First, we will present the results for the study area of Abbiategrasso (Table 4, Table 5, Table 6, Table 7 and Table 8, Figure 9, Figure 10, Figure 11 and Figure 12).

Table 4. Calculation of the real BAF of the study area of Abbiategrasso, based on DBT and visual interpretation.

Type of Surface	Area	Weight Factor	Ecologically Effective Surface
Sealed surface	657,841 sq.m.	0	0 sq.m.
Partially sealed surface	24,230 sq.m.	0.1	2423 sq.m.
Semi-open surface	5472 sq.m.	0.2	1094 sq.m.
Surfaces with vegetation unconnected to soil below, shallow substrate	12,424 sq.m.	0.5	6212 sq.m.
Extensive roof greening	19,465 sq.m.	0.5	9733 sq.m
Surfaces with vegetation unconnected to soil below, deep substrate	245,225 sq.m.	0.9	220,703 sq.m.
Surfaces with vegetation connected to soil below	35,343 sq.m.	1	35,343 sq.m.
TOTAL	1,000,000 sq.m.		275,507 sq.m.
BAF	0.28

Table 4. Calculation of the real BAF of the study area of Abbiategrasso, based on DBT and visual interpretation.

Type of Surface	Area	Weight Factor	Ecologically Effective Surface
Sealed surface	657,841 sq.m.	0	0 sq.m.
Partially sealed surface	24,230 sq.m.	0.1	2423 sq.m.
Semi-open surface	5472 sq.m.	0.2	1094 sq.m.
Surfaces with vegetation unconnected to soil below, shallow substrate	12,424 sq.m.	0.5	6212 sq.m.
Extensive roof greening	19,465 sq.m.	0.5	9733 sq.m
Surfaces with vegetation unconnected to soil below, deep substrate	245,225 sq.m.	0.9	220,703 sq.m.
Surfaces with vegetation connected to soil below	35,343 sq.m.	1	35,343 sq.m.
TOTAL	1,000,000 sq.m.		275,507 sq.m.
BAF	0.28

Figure 9. Real BAF map of Abbiategrasso, based on DBT and visual interpretation.

Figure 9. Real BAF map of Abbiategrasso, based on DBT and visual interpretation.

Table 5. Calculation of simplified BAFs of the study area of Abbiategrasso, with a translation into the binary system of the real BAF.

Category	Area	Weight Factor	Ecologically Effective Surface
0–0.49	687,543 sq. m.	0	0 sq. m.
0.5–1	312,457 sq. m.	1	312,457 sq. m.
TOTAL	1,000,000 sq. m.		312,457 sq. m.
BAF	0.31

Table 5. Calculation of simplified BAFs of the study area of Abbiategrasso, with a translation into the binary system of the real BAF.

Category	Area	Weight Factor	Ecologically Effective Surface
0–0.49	687,543 sq. m.	0	0 sq. m.
0.5–1	312,457 sq. m.	1	312,457 sq. m.
TOTAL	1,000,000 sq. m.		312,457 sq. m.
BAF	0.31

Figure 10. Simplified BAF map of Abbiategrasso.

Figure 10. Simplified BAF map of Abbiategrasso.

Table 6. Calculation of Sentinel BAF of Abbiategrasso in April, July, and October 2020.

Month	Category	Area	Weight Factor	Ecologically Effective Surface
April 2020	0–0.49	689,637 sq. m.	0	0 sq. m.
	0.5–1	310,363 sq. m.	1	310,363 sq. m.
	TOTAL	1,000,000 sq. m.		310,363 sq. m.
	BAF	0.31
July 2020	0–0.49	762,998 sq. m.	0	0 sq. m.
	0.5–1	237,002 sq. m.	1	237,002 sq. m.
	TOTAL	1,000,000 sq. m		237,002 sq. m.
	BAF	0.24
October 2020	0–0.49	557,211 sq. m.	0	0 sq. m.
	0.5–1	442,789 sq. m.	1	442,789 sq. m.
	TOTAL	1,000,000 sq. m		442,789 sq. m.
	BAF	0.44

Table 6. Calculation of Sentinel BAF of Abbiategrasso in April, July, and October 2020.

Month	Category	Area	Weight Factor	Ecologically Effective Surface
April 2020	0–0.49	689,637 sq. m.	0	0 sq. m.
	0.5–1	310,363 sq. m.	1	310,363 sq. m.
	TOTAL	1,000,000 sq. m.		310,363 sq. m.
	BAF	0.31
July 2020	0–0.49	762,998 sq. m.	0	0 sq. m.
	0.5–1	237,002 sq. m.	1	237,002 sq. m.
	TOTAL	1,000,000 sq. m		237,002 sq. m.
	BAF	0.24
October 2020	0–0.49	557,211 sq. m.	0	0 sq. m.
	0.5–1	442,789 sq. m.	1	442,789 sq. m.
	TOTAL	1,000,000 sq. m		442,789 sq. m.
	BAF	0.44

Figure 11. Sentinel BAF maps of different seasons in Abbiategrasso, derived from semi-automatic classification: (a) April 2020; (b) July 2020; (c) October 2020.

Figure 11. Sentinel BAF maps of different seasons in Abbiategrasso, derived from semi-automatic classification: (a) April 2020; (b) July 2020; (c) October 2020.

Table 7. Calculation of the average Sentinel BAF.

	April 2020	July 2020	October 2020
Sentinel BAF	0.31	0.24	0.44
Average	0.33

Table 7. Calculation of the average Sentinel BAF.

	April 2020	July 2020	October 2020
Sentinel BAF	0.31	0.24	0.44
Average	0.33

Table 8. BAF calculation using aerial imagery of the study area of Abbiategrasso.

Category	Pixel	Area	Weight Factor	Ecologically Effective Surface
No green	52,028	633,129 sq. m.	0	0 sq. m.
Green	30,148	366,871 sq. m.	1	366,871 sq. m.
TOTAL	82,176	1,000.00 sq. m.		366,871 sq. m.
BAF	0.37

Table 8. BAF calculation using aerial imagery of the study area of Abbiategrasso.

Category	Pixel	Area	Weight Factor	Ecologically Effective Surface
No green	52,028	633,129 sq. m.	0	0 sq. m.
Green	30,148	366,871 sq. m.	1	366,871 sq. m.
TOTAL	82,176	1,000.00 sq. m.		366,871 sq. m.
BAF	0.37

Figure 12. Aerial imagery BAF map of study area of Abbiategrasso, showing the green areas.

Figure 12. Aerial imagery BAF map of study area of Abbiategrasso, showing the green areas.

4.2. Segrate

The same procedure was followed for the Milano Due district in the municipality of Segrate (Table 9, Table 10, Table 11, Table 12 and Table 13, Figure 13, Figure 14, Figure 15 and Figure 16).

Table 9. Calculation of the real BAF for the study area of Milano Due, based on DBT and visual interpretation.

Type of Surface	Area	Weight Factor	Ecologically Effective Surface
Sealed surface	516,515 sq. m.	0	0 sq. m.
Partially sealed surface	32,851 sq. m.	0.1	3285 sq. m.
Semi open surface	2850 sq. m.	0.2	570 sq. m.
Surfaces with vegetation unconnected to the soil below, shallow substrate	18,085 sq. m.	0.5	9042 sq. m.
Extensive roof greening	31,561 sq. m.	0.5	15,781 sq. m.
Surfaces with vegetation unconnected to the soil below, deep substrate	261,524 sq. m.	0.9	235,372 sq. m.
Surfaces with vegetation connected to soil below	136,614 sq. m.	1	136,614 sq. m.
TOTAL	1,000,000 sq. m.		400,664 sq. m.
BAF	0.4

Table 9. Calculation of the real BAF for the study area of Milano Due, based on DBT and visual interpretation.

Type of Surface	Area	Weight Factor	Ecologically Effective Surface
Sealed surface	516,515 sq. m.	0	0 sq. m.
Partially sealed surface	32,851 sq. m.	0.1	3285 sq. m.
Semi open surface	2850 sq. m.	0.2	570 sq. m.
Surfaces with vegetation unconnected to the soil below, shallow substrate	18,085 sq. m.	0.5	9042 sq. m.
Extensive roof greening	31,561 sq. m.	0.5	15,781 sq. m.
Surfaces with vegetation unconnected to the soil below, deep substrate	261,524 sq. m.	0.9	235,372 sq. m.
Surfaces with vegetation connected to soil below	136,614 sq. m.	1	136,614 sq. m.
TOTAL	1,000,000 sq. m.		400,664 sq. m.
BAF	0.4

Figure 13. Real BAF map of Milano Due, based on the DBT and visual interpretation.

Figure 13. Real BAF map of Milano Due, based on the DBT and visual interpretation.

Table 10. Calculation of the simplified BAF of the study area of Milano Due, with a translation into the binary system of the real BAF.

Category	Area	Weight Factor	Ecologically Effective Surface
0–0.49	552,216 sq. m.	0	0 sq. m.
0.5–1	447,784 sq. m.	1	447,784 sq. m.
TOTAL	1,000,000 sq. m.		447,784 sq. m.
BAF	0.45

Table 10. Calculation of the simplified BAF of the study area of Milano Due, with a translation into the binary system of the real BAF.

Category	Area	Weight Factor	Ecologically Effective Surface
0–0.49	552,216 sq. m.	0	0 sq. m.
0.5–1	447,784 sq. m.	1	447,784 sq. m.
TOTAL	1,000,000 sq. m.		447,784 sq. m.
BAF	0.45

Figure 14. Simplified BAF map of Milano Due.

Figure 14. Simplified BAF map of Milano Due.

Table 11. Calculation of Sentinel BAF of Milano Due in April, July, and October 2020.

Month	Category	Area	Weight Factor	Ecologically Effective Surface
April 2020	0–0.49	699,451 sq. m	0	0 sq. m.
	0.5–1	300,549 sq. m.	1	300,549 sq. m.
	TOTAL	1,000,000 sq. m.		300,549 sq. m.
	BAF	0.30
July 2020	0–0.49	572,533 sq. m.	0	0 sq. m.
	0.5–1	427,467 sq. m.	1	427,467 sq. m.
	TOTAL	1,000,000 sq. m.		427,467 sq. m
	BAF	0.43
October 2020	0–0.49	586,225 sq. m.	0	0 sq. m.
	0.5–1	413,775 sq. m.	1	413,775 sq. m.
	TOTAL	1,000,000 sq. m.		413,775 sq. m.
	BAF	0.41

Table 11. Calculation of Sentinel BAF of Milano Due in April, July, and October 2020.

Month	Category	Area	Weight Factor	Ecologically Effective Surface
April 2020	0–0.49	699,451 sq. m	0	0 sq. m.
	0.5–1	300,549 sq. m.	1	300,549 sq. m.
	TOTAL	1,000,000 sq. m.		300,549 sq. m.
	BAF	0.30
July 2020	0–0.49	572,533 sq. m.	0	0 sq. m.
	0.5–1	427,467 sq. m.	1	427,467 sq. m.
	TOTAL	1,000,000 sq. m.		427,467 sq. m
	BAF	0.43
October 2020	0–0.49	586,225 sq. m.	0	0 sq. m.
	0.5–1	413,775 sq. m.	1	413,775 sq. m.
	TOTAL	1,000,000 sq. m.		413,775 sq. m.
	BAF	0.41

Figure 15. Sentinel BAF maps of each different season of Milano Due, derived from semi-automatic classification: (a) April 2020; (b) July 2020; (c) October 2020.

Figure 15. Sentinel BAF maps of each different season of Milano Due, derived from semi-automatic classification: (a) April 2020; (b) July 2020; (c) October 2020.

Table 12. Calculation of average Sentinel BAF.

	April 2020	July 2020	October 2020
Sentinel BAF	0.30	0.43	0.41
Average	0.38

Table 12. Calculation of average Sentinel BAF.

	April 2020	July 2020	October 2020
Sentinel BAF	0.30	0.43	0.41
Average	0.38

Table 13. BAF calculation using aerial imagery in the study area of Milano Due.

Category	Pixel	Area	Weight Factor	Ecologically Effective Surface
No green	47,558	541,157 sq. m.	0	0 sq. m.
Green	40,324	458,843 sq. m.	1	458,843 sq. m.
TOTAL	87,984	1,000.00 sq. m.		458,843 sq. m.
BAF	0.46

Table 13. BAF calculation using aerial imagery in the study area of Milano Due.

Category	Pixel	Area	Weight Factor	Ecologically Effective Surface
No green	47,558	541,157 sq. m.	0	0 sq. m.
Green	40,324	458,843 sq. m.	1	458,843 sq. m.
TOTAL	87,984	1,000.00 sq. m.		458,843 sq. m.
BAF	0.46

Figure 16. Aerial imagery BAF map of study area of Milano Due, showing green areas.

Figure 16. Aerial imagery BAF map of study area of Milano Due, showing green areas.

4.3. Results Comparison

To verify the efficiency of the methodology, we compared the results that were obtained in each study area (Figure 17 and Figure 18).

To better compare the ecological value deriving from the different procedures with the real BAF, we calculated the average of the BAF that was derived from the Sentinel images and from the aerial imagery. The BAF average refers to a BAF range of 0–1: the average was achieved when comparing the average of the Sentinel BAF and the average of the BAF that was obtained using aerial imagery (2):

BAF average = (Sentinel BAF average + aerial imagery BAF)/2

(2)

Figure 17. BAF comparison for Abbiategrasso.

Figure 17. BAF comparison for Abbiategrasso.

Figure 18. BAF comparison for Milano Due.

Figure 18. BAF comparison for Milano Due.

The BAF values that were obtained were related to the real BAF and were calculated in the traditional way; the average of the BAF was derived using a semi-automatic method. The standard error was reported for each standard value and an estimated standard deviation was reported (it is important to underline that standard error is a statistical term that measures the accuracy at which a sample distribution represents a population—or a group of items—using standard deviation. Given a sample size, the standard error equals the standard deviation, divided by the square root of the sample size [77]).

5. Discussions

The results presented here permit quite a wide range of comments to be highlighted, starting from the various reliabilities of the different open-source data management systems. The first outcome is evidence that comparison among similar procedures with different datasets is not sufficient to obtain real BAF measures. In fact, it varies in a non-linear way with respect to classic city planning data, such as the density and coverage ratio. As such, the settlement morphology can only measure the order of magnitude of the ecological value but not its precise measurement.

Settlement morphology and urban fabric are the two main features of artificial land that strongly influence the BAF value. The land-use categories that are useful for BAF calculation are: buildings, streets, parking (as impermeable areas); parks, green roofs, green strings (as a permeable or partially permeable area in which evapotranspiration has only a certain level of effectiveness). The different urban functions, like residential, industrial, commercial, tertiary, and public facilities, have the same relevance for the BAF calculation.

In city planning activity, the urban functional mix mainly determines the number of public services, the accessibility needs (e.g., the road system), and the parking area. So, to obtain certain environmental performance targets, a public bureau can fix a certain BAF value as minimum implementation value (for example, in Berlin, it is fixed to be at least 0.6 for residential, tertiary, and public services areas). The BAF value can be performed at the ground level by using the quantity of green and the use of permeable surfaces at the building scale from green roof setups to green facades.

In the assessment phase, a public bureau can use the BAF to assess the level of evapotranspiration and to define policies and technical targets to be implemented in the regional and city plans.

Many studies have highlighted the temporal dimension of remote sensing as a valuable element that can be used to study ecological systems, as well as a source of information on variables such as the timing and monitoring of vegetative phenological events, specifically, vegetation growth [51,78]

It is important to underline the fact that plants interact with sunlight differently, depending on the wavelength with which they are observed. The electromagnetic radiation that is emitted by the sun and then reflected by plants contains information about the biophysical composition and physiological status of different plants. The chlorophyll pigments that are present in green leaves and that provide the necessary energy for photosynthesis are absorbed in the visible (VIS) region of the spectrum (400–700 nm), specifically in the blue and red wavelengths. In the near-infrared region (NIR), which ranges from approximately 700 to 1300 nm, plant leaves exhibit high reflectance values, absorbing and transmitting less radiation. Regarding the shortwave infrared region (SWIR), which ranges from 1300 to 2500 nm, the radiation absorption is largely dominated by the water that is present in leaves [79].

In recent decades, the fundamental role of agroecosystems in food production, their role in environmental conservation, and their role in ecological status control have encouraged the diffusion of remote sensing technologies, to provide crucial contributions to agriculture monitoring and management, and to improve the understanding of changes resulting from environmental impacts. Sentinel-2 was first designed and implemented to specifically meet the needs of the worldwide agricultural community [79].

For these reasons, it is commonly used for the analysis of large-scale portions of territories that are only characterized by natural (i.e., woods) and agriculture-related elements.

When used to study urbanized areas (from small settlements spread in the rural areas to important urban centers such as cities, or even megalopolis), the potential of Sentinel-2 imagery is limited because its spatial resolution of 10 m fails to capture fine-grained urban structures such as trees, shrubs, or small buildings, and to distinguish public from private green spaces [80].

In addition, atmospheric distortions may cause further uncertainties in the recognition of land use objects [81].

Considering that the availability and accessibility of this data are central, Sentinel-2 images are freely and openly downloadable from the web in several easy ways: as a direct download from ESA’s website or through Copernicus; third-party tools are also available to download these data, such as the open-source software QGIS, by using various plug-ins that provide tools that can be used to download the data in a simple manner, and the Google Earth Engine updates all of the available copies of the Sentinel-2 data daily [79].

Furthermore, considering the importance of always having up-to-date data available, whether in raw form or in the form of pre-calibrated and pre-processed satellite images, the continuous image-updating program that is provided by Sentinel-2 defines it as a precious tool for the analysis of current trends, for the definition of a historic series, and for the implementation of future projection studies.

In the European context, the Strategic Environmental Assessment (SEA) program has been in place since 2001 and is obligated to address environmental and ecological issues throughout the entire planning process [82]. In all planning procedures (both at territorial and urban scales) that aim to integrate urban planning and ecology, environmental indicators are the starting point for the creation of composite indices that contain urban parameters and environmental elements.

There are many examples of new urban-ecological standards that aim to achieve the correct sizing of urban area and settlement weights, as well as those of ecological and environmental facilities. Although not all of the new standards are codified by the law yet, they have been introduced into more innovative plans in order to assess and define soil permeability, allowable (sustainable) urban load, the environmental carrying capacity of transformation areas, and so on.

An effective index should have a solid scientific base, should have tested different and complex case studies, should be able to direct a structural choice and should not tamper with the design phase, and should also refer to the regulatory phases of planning processes: the BAF model presents all of these characteristics [29].

The possibility of integrating the BAF model during planning processes (in urban and territorial contexts) is thanks to the use of satellite and aerial images that can streamline the standard procedure for calculating the index (see paragraph 3.1), allowing a semi-automatic method to be adopted that is both more objective and faster, loses less information, and, as we have seen, demonstrates good value approximation.

Given that the calculation procedure for BAF is standardized, one of the most interesting results of the research is the dependability of the different open-source data. The results strengthen the nature of semi-automatic methods and the necessity to maintain the interaction between the tool and the professional: in fact, a completely automatic methodology offers many possible outcomes, but no one is 100% correct.

5.1. Method Evaluation: Range of Error

It is recognized that the proposed method presents possible errors, starting from the simplification of BAFs, as already mentioned above. The simplified BAF represents an approximation of the BAF calculation into a 0–1 binary system, producing an error of 10–12%, which can be considered as minimal, especially seeing that the error rate becomes less relevant as the scale increases. However, its use is essential for the verification of the reliability of the semi-automatic calculation method that is proposed herein.

Furthermore, the use of different and unconventional sources to calculate the BAF can lead to errors. For example, when automatically selecting green surfaces in aerial imagery, the portions of vegetation that are in shaded conditions are not considered. Therefore, the calculation of ecologically useful surface area using aerial imagery is not reliable (Table 8). For this, the BAF was also calculated for the aerial images by selecting the shadows created by trees (

Table 14. Aerial image and shadow analysis: calculation of BAFs in the study area of Abbiategrasso.

Typology	N° Pixel	Area	Coefficients	Ecologically Effective Surface
No green	35,564	432,778 sq.m.	0	0 sq.m.
Green	46,612	567,222 sq.m.	1	567,222 sq.m.
TOTAL	82,176	1,000,00 sq.m.		567,222 sq.m.
BAF	0.57

).

By adding shadows to the selection, however, the shadows of all of the buildings are automatically selected, and thus also consider the non-green portions that are in shaded conditions (Figure 19).

Another type of error derives from the use of Sentinel images, both because of the presence of clouds, which can disturb the images, and because of the choice of the images themselves. The different seasons obviously reproduce different vegetation states, so the green coverage areas change for each image.

We overlapped the ecologically effective areas for Sentinel BAF calculation for each month considered (April, July, October) to highlight the parts that coincide (

Table 15. Percentage of the overlapping area with a factor of 1 for the Sentinel BAF map for the study area of Abbiategrasso.

Sentinel BAF Map
April	July		55%
October	April		60%
July	October		51%
April	July	October	40%

).

For this, we considered the average of the Sentinel BAF values, which was more reliable.

Knowledge of the range of error is extremely useful when approaching existing regions or cities for which the ground truth is unknown. With the proposed methodology and open-source data, it is possible to evaluate the environmental performance of territories to obtain at least the order of magnitude of this value. The error evaluation allows scholars, public administrators, official bureaus, and private stakeholders to acquire basic knowledge to better address further analysis through on-site surveys or the acquisition of cartography for a fee.

Mixed Pixel Errors

To supervise the image classification process, we applied the maximum likelihood algorithm using a QGIS plug-in: this is an image classification method that allows pixel selection for each class to be better controlled (in our research, the different classes are the categories of land use and soil coverage). To improve classification accuracy and to reduce misclassifications, post-classification enhancement was used. To enhance the classification accuracy, visual interpretation, as well as local knowledge, was very important. The problem of mixed pixels derives from visual interpretation.

The proportion of mixed pixels in an image is a function of the sensor’s spatial resolution, and it is not unusual for classification analyses to map the land cover that is determined from remotely sensed data [81,82]. This means that a pixel does not belong to a single class. Usually, the proportion of mixed pixels tends to increase along with the pixel size but mixing problems may occur at fine spatial resolutions. This problem is often evident at the geographical boundaries between classes, or when the landscape mosaic is highly fragmented, which is often the case in an urban environment.

5.2. Replicability in Different Contexts

The presented method can be easily applied in other urban contexts: the ecological and urban implications are strictly related to the qualitative relevance of the parameter. It is important that the range of values in which the entire surface, with certain characteristics, fits and not the exact size of the ecological surface. Therefore, several variables are fundamental: the availability of the different sources, such as the topographic and cartographic databases (updated, formally corrected, and downloadable in a simple and direct way); aerial images that depict that portion of the territory (not all areas are investigated with the same resolution and at the same moments or time breaks, therefore, it could be difficult to compare the historical series, even in the same reference year); and images from Sentinel-2 or other satellite instruments that work above the different continents.

Moreover, it is also important that information is directly and freely available; in this regard, the Lombardy Region (where the analyses reported here were conducted) facilitates the work of researchers, public administrations, and professionals by providing a huge and well-structured database that includes basic cartographic charts, shapefiles that include all of the basic information that is necessary to read and analyze the territories (from a physical and planning point of view), and the territorial government plans of different municipalities.

Furthermore, the method, and especially the BAF tool, is valid in cases where the natural component is mainly composed of grasslands and trees. For example, desert areas present a permeability and an evapotranspiration potential that does not depend on green structures: therefore, studies that use the presented methodology are only appropriate where the natural area presents green components.

5.3. Scale Variation: From Local to Regional Dimensions

The two case studies that were presented here have dimensions and morphological characteristics that are comparable to those of specific urban neighborhoods (such as Milano Due). The structure of the urban fabric is easily recognizable (thanks to the presence of regular blocks and roads), and the presence of clearly identifiable urban functions and surfaces of about 1,000,000 sq. m. in size allows the classic (BAF calculation from DBT) and innovative semi-automatic (BAF with the use of aerial and satellite images) procedures to be managed in a precise, manageable and controllable way (mainly for ex post evaluation and the identification of potential errors).

As already mentioned, the highly complex mosaic of the urban fabric (especially in the context of Abbiategrasso, as the study area insists on a consolidated historical fabric that is characterized by overlapping elements, due to the historical evolution of the settlement) makes it difficult to correctly identify and associate each area with an ecologically effective surface class. However, this is not due to the dimension scale of the object to be investigated but is instead due to the resolution of the subject being investigated.

At the urban and extra-urban levels (and consequently at the municipal, provincial, or regional levels), the DBT always provides precise information and values (the use of databases such as DUSAF or CORINNE Land Cover or other similar tools that are present in other areas, if updated, give the exact land uses); aerial images convey information with the same precision; the use of Sentinel-2 is, as seen here, closely linked to the acquisition of images and the types of green areas that are present (including during periods of the year where the vegetation growth rate is highly different).

6. Conclusions

The presented method finds its natural outlet as a support tool for urban planning and strategic assessment processes.

The correlation among urban/territorial transformations, the functional mix of land uses (built fabrics, the relationship between built and undeveloped spaces, functional relationships in fringe areas, strategic functions, infrastructural systems, public and private green space systems) in urban and in rural–urban contexts, the variation of anthropic loads, and the variation and distribution of pollutants assume a fundamental role from a sustainable territorial planning perspective. The relevance of the ecological element within urban fabrics is therefore of essential importance to maintain and restore a high level of quality of life from the perspective of the healthy city approach.

As an index for the calculation of the percentage of ecologically effective areas, the BAF not only allows both built and consolidated contexts (as an ex post verification) to be analyzed, it can also be used in urban planning applications as a performance objective that must be reached when new buildings or redevelopments of dismissed areas are introduced. These measures allow the transposition of general sustainable strategies into physical actions because it is easy to measure the BAF value before and after the intervention. In existing settlements, improving the BAF can be associated with volume incentives or tax cuts to reduce the cost of positive green actions and to encourage the sustainable transformation of the existing city.

Moreover, the simultaneous use of cartography, satellite images, and aerial images provides a global understanding of all the soil features that are necessary for a correct and complete definition of the index itself. It also reduces the analysis time and produces more objective evaluations since considering data from multiple sources increases the possibility of confusing the real characteristics of the selected areas.

It cannot be denied that there are objective difficulties that are linked to the analysis of the data sources themselves: first of all, they must always be updated, and open-source data should always be available; then, the resolutions of the satellite and aerial images require particular attention in highly complex situations, such as in situations with a diverse urban fabric, in order to identify the correct parameters that can be used to refine the sensitivity of the instrument and of the analysis methods being used.

Moreover, the described methodology could also be used in a reverse way: it is able to establish which requirements are necessary for a map (cartography or aerial/satellite imagery) in order to fully support BAF calculation and BAF assessment in urban contexts.

This research demonstrates the level of accuracy that different sources guarantee in measuring the evapotranspiration capacity of urban settlements throughout BAF calculations. A perfect algorithm does not exist, but by combining the different data sources, it is possible to describe these land features with a good amount of approximation. For BAF calculation, open-source data are fundamental for the assessment of the performance of different regions, basins, and entire countries. The error percentage is not constant for each source, but the average value among the different calculations is able to furnish the correct order of magnitude of the BAF value in the selected area. The wider the area is, the more useful the semi-automatic methodology for planners to better define environmental targets for existing settlements.

Finally, it is interesting to note that the two case studies, Abbiategrasso and Milan Due, have very different volumetric building densities (the first, 1.38 cu. m./sq. m., the latter, 2.91 cu. m./sq. m.), deriving from completely different type-morphological implants but this does not lead to an inverse variation in the BAF. On the contrary, it stood at 0.28 (real BAF) for Abbiategrasso and at 0.40 for Milano Due: the most important parameters are the coverage ratio, green areas, and types of permeable surfaces present in the territory. Future research should analyze different urban settlements with different urban parameters (buildings area; built-up area; coverage ratio) in order to define whether there is a direct (or indirect) correlation among the urban parameters (considering the range of values) and BAF final values.

As was demonstrated in Section 2.3 and Section 2.4, the aerial images can furnish a high level of accuracy in terms of land description. At the moment, it is not possible to define the specific land use from the images but as we stated at the beginning of Section 5, for BAF calculation, it is not relevant to define the specific destination of every building. On the other hand, the Sentinel images have a granularity of about 10 × 10 m; this means that buildings like small villas or small streets are difficult to capture. In the same way, the Sentinel satellites are not able to register green areas like green strips (for example, green verges along the roads) for their small dimensions. From the Sentinel images, in small green areas, only the most effective chlorophyll production systems become relevant.

Starting from these suggestions, it is possible to define urban planning guidelines and architectural parameters that could be imposed on transformation areas and/or for renovation activities that take place in an urban environment, which are promoted by public administrations and private citizens. Some innovative urban plans have already incorporated this technique (or use a similar tool, for example, see Seattle Green Factor and RIE Bologna [36,37]) as being a necessary parameter for the development of the urban context and as an instrument for enhancing the ecological value of the urban environment, improving urban life quality (the regulation of hygrothermal comfort, the number of green areas, etc.) and reducing the risks that are associated with sudden and significant atmospheric events.