The vegetation mask classifies land cover into 4 macro-classes: (0) no vegetation, (1) trees, (2) shrubs, and (3) grassland. The mask, with a spatial resolution of 10m, was obtained through the integration of different data from the European Space Agency (ESA) Copernicus satellite observation programme, and the Italian Institute for Environmental Protection and Research (ISPRA – Istituto Suoeriore per la Protezione e Ricerca Ambientale)

• National land cover map ISPRA 2018 (COP) [1]
• Copernicus Urban Atlas 2018 (UA) [2]
• Copernicus Street Tree layer 2018 (STL) [3]
• Copernicus High Resolution Layer: Tree Cover Density 2018 (TCD) [4]
• Copernicus High Resolution Layer: Grassland 2018 (GR) [5]
• Copernicus Small Woody Features 2018 (SWF) [6]

To obtain the four categories, each input map was reclassified in a 0-3 range according to the following scheme:

Furthermore, the various maps (which ranged from 0-3 or 0-1), were combined through the logic OR (||) according to the following order:

1. ISPRA Land cover [0-3] || Copernicus Urban Atlas [0-3] = a [0-3]
2. Copernicus HRL Grassland [0-3] || a [0-3] = b[0-3]
3. Copernicus HRL Tree Cover Density [0-1] || b [0-3] = c [0-3]
4. Copernicus Street Tree Layer [0-1] || Copernicus Small Woody feature [0-1] = d [0-1]
5. d [0-1]|| c [0-3] = vegetation map [0-3]

For each city, the vegetation mask was computed on the boundaries of the “city” as identified in DEGURBA classification [9]. Here, the urban core was identified according to the National Institute of Statistics (ISTAT) as “as an ‘aggregate of contiguous or nearby houses […] that constitutes an autonomous form of social life and, generally, also a gathering place for the residents of the surrounding areas to demonstrate the existence of a coordinated form of social life centered around it'” [10] [10].

Land Surface Temperature (LST) maps constitute a fundamental element of the Italian Observatory for Urban Afforestation Policies, managed by the Foundation for the Future of Cities. These maps allow for the identification and long-term monitoring of urban heat islands, as well as quantifying the contribution of vegetation and trees in mitigating temperature.

The Landsat satellite mission provides LST maps at a spatial resolution of 30 meters.

However, this level of detail is often insufficient in urban environments, where green areas are frequently dispersed and of small dimensions. To address this limitation, we leveraged data from the Sentinel-2 satellite mission operated by the European Space Agency (ESA), which provides images at a higher spatial resolution of 10 meters and a revisit time of 3-5 days Unfortunately, the sensors on the two Sentinel-2 satellites do not directly measure temperature. Using a machine learning algorithm called Random Forests[11], we developed a model to estimate land surface temperature using Sentinel-2 images, effectively enhancing the spatial and temporal resolution of our product

The temperature data displayed in the interactive maps correspond to the average temperature during the period of June-August 2022.

In more detail, the procedure involves the following steps, which remain consistent for each analyzed city:

1. Selection of all Sentinel-2 images(https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S2_SR_HARMONIZED) and Landsat 8 images (https://developers.google.com/earth-engine/datasets/catalog/LANDSAT_LC08_C02_T1_L2) with cloud cover below 70% [12] and acquired between June 1, 2022, and August 31, 2022.
2. Production of a summer composite for Landsat 8 [13,14] with a spatial resolution of 30 meters, indicating the median temperature for each pixel.
3. Using Sentinel-2 data, production of a summer composite[14,15]with a spatial resolution of 10 meters, indicating the median reflectance values for the four available Sentinel-2 bands at 10 meters spatial resolution: blue (B2), red (B3), green (B4), and near-infrared (B8).
4. Simple random sampling and selection of 10,000 Sentinel-2 pixels. Subsequently, the construction of a training database consisting of 10,000 examples with available Sentinel-2 bands (dependent variables) and Landsat 8 temperature (independent variable).
5. Training of the Random Forests algorithm[11]implemented on Google Earth Engine(https://developers.google.com/earth-engine/apidocs/ee-classifier-smilerandomforest)
6. Application of the model to each Sentinel-2 image (refer to point 1) to estimate temperature based on the four bands
7. Calculation of the average of the predicted temperatures from N images and estimation of the average temperature for each 10-meter pixel.
8. For each one-hectare hexagon, calculation of the average temperature of included pixels.

This procedure is implemented on a cloud computing platform called Google Earth Engine [16]. The codes and data necessary to replicate the entire procedure can be provided upon request (saverio.francini@unifi.it).

One of the primary ecosystem services provided by trees and vegetation in urban contexts is the sequestration of carbon from the atmosphere. To monitor this aspect, we utilized data from the National Forest Inventory (2015) and Sentinel-2 satellite data acquired between April 2017 – when the second Sentinel-2 satellite became operational – and March 2018, to have a complete year of observations. Using Sentinel-2 time series data and implementing the procedure introduced by Bozzini et al. [17] [17], we calculated a wide range of predictors. Then, as demonstrated by Chirici et al. [18], Vangi et al. [19], and Giannetti et al. [20] we used Random Forests [11] to build a model capable of estimating ground-measured carbon within the inventory based on the predictors calculated from Sentinel-2 images.

In more detail, the procedure involves the following steps, which remain consistent for each analyzed city:

1. Selection of all Sentinel-2 images (https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S2_SR_HARMONIZED) with cloud cover below 50% (White et al., 2014) and acquired between April 1, 2017, and March 31, 2018.
2. Calculation of harmonic predictors using the selected Sentinel-2 data and implementing the procedure demonstrated by Bozzini et al., [17] and Parisi et al., [21].
3. Calculation of a summer composite using Sentinel-2 data acquired between June 1, 2017, and August 31, 2017, and implementing the Medoid procedure as shown by Flood [22], Kennedy et al., [23], and Francini et al., [13].
4. Calculation of average temperature for the period 2010-2020 using Copernicus Climate Change Service data(https://developers.google.com/earth-engine/datasets/catalog/ECMWF_ERA5_MONTHLY)
5. Calculation of median precipitation for the period 2010-2020 using Copernicus Climate Change Service data(https://developers.google.com/earth-engine/datasets/catalog/ECMWF_ERA5_MONTHLY)
6. The variables produced in steps 2, 3, 4, and 5 are combined with the 2005 forest provision map [19] and a digital elevation model(https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_DEM_GLO30) and form the predictors (dependent variables) of the model that will be implemented in Step 8.
7. Selection of ground-measured carbon for approximately 7,000 points within the Italian National Forest Inventory. This data serves as the independent variable for the model.
8. Training of the “Random Forests” algorithm [11] implemented on Google Earth Engine (https://developers.google.com/earth-engine/apidocs/ee-classifier-smilerandomforest)
9. As demonstrated by Chirici et al. [18], Vangi et al. [19], and Giannetti et al. [20], application of the model to obtain a carbon map with a spatial resolution of 10 meters for each pixel on the vegetation map
10. For each one-hectare hexagon, calculation of the total carbon, i.e., the sum of carbon contained in each 10-meter pixel.

This procedure is implemented on the cloud computing platform Google Earth Engine [24]. The codes and data necessary to replicate the entire procedure can be provided upon request (saverio.francini@unifi.it).

Elevated concentrations of particulate matter, specifically PM10, are associated with cardiovascular and respiratory problems [25]. Trees and urban green infrastructure play a crucial role in reducing air pollution and absorbing particulate matter.

This concept, along with the associated strategic, tactical, and operational tools, is gaining increasing attention in innovative urban planning processes for cities and urban regions [26]. The presence of trees and forest cover in cities, urban centers, and surrounding suburbs contributes to improving the quality of life for urban communities and helps local governments achieve their environmental, social, and economic sustainability goals.

To monitor this aspect, we implemented the procedure shown by Bottalico et al. [27]. Using satellite data, it is possible to estimate the leaf surface area of trees, which is directly related to their ability to remove PM10 from the air.

In more detail, the procedure involves the following steps, which remain consistent for each analyzed city:

1. Selection of all Sentinel-2 images(https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S2_SR_HARMONIZED) with cloud cover below 50% [12] and acquired between June 1, 2022, and August 31, 2022
2. Production of a Medoid summer composite [13,22,23]
3. Calculation of the Leaf Area Index(LAI) ([27], Eq.5).
4. Calculation of the PM10 removal capacity ([27], Eq. 4). To apply this equation, we used the PM10 value in the air in Florence in 2013[27]. This value may vary significantly depending on the city or year, and the data we produce is therefore not the actual quantity of PM10 absorbed but an index showing the vegetation’s capacity to absorb PM10. The calculation is performed for each pixel on the vegetation map.

This procedure is implemented on the cloud computing platform Google Earth Engine [16]. The codes and data necessary to replicate the entire procedure can be provided upon request (saverio.francini@unifi.it).

1. ISPRA Carta Nazionale Di Copertura Del Suolo Available online: https://www.isprambiente.gov.it/it/attivita/suolo-e-territorio/suolo/copertura-del-suolo/carta-nazionale-di-copertura-del-suolo.
2. ESA Copernicus Urban Atlas Available online: https://land.copernicus.eu/local/urban-atlas/urban-atlas-2018 (accessed on July 13, 2023).
3. ESA Copernicus Street Tree Layer Available online: https://land.copernicus.eu/local/urban-atlas/street-tree-layer-stl-2018 (accessed on July 13, 2023).
4. ESA Copernicus HRL: Tree Cover Density Available online: https://land.copernicus.eu/pan-european/high-resolution-layers/forests/tree-cover-density/status-maps/tree-cover-density-2018 (accessed on July 13, 2023).
5. ESA Copernicus HRL: Grassland Available online: https://land.copernicus.eu/pan-european/high-resolution-layers/grassland/status-maps/grassland-2018 (accessed on July 13, 2023).
6. ESA Copernicus Small Woody Features Available online: https://land.copernicus.eu/pan-european/high-resolution-layers/small-woody-features/small-woody-features-2018 (accessed on July 13, 2023).
7. ISPRA Copertura Suolo 2018 Available online: https://groupware.sinanet.isprambiente.it/uso-copertura-e-consumo-di-suolo/library/copertura-del-suolo/carta-di-copertura-del-suolo/copertura-del-suolo-2018 (accessed on July 13, 2023).
8. European Commission Mapping Guide for a European Urban Atlas Regional Policy. 2006, 21-22.
9. Eurostat Applying the Degree of Urbanisation; 2021; ISBN 9789276203063.
10. ISTAT Descrizione Dei Dati Geografici e Delle Variabili Censuarie Delle Basi Territoriali per I Censimenti Anni 1991, 2001, 2011; 2016;
11. Breiman, L. Random Forests. MachLearn2001, 45, 5-32, doi:https://doi.org/10.1023/A:1010933404324.
12. White, J.C.; Wulder, M.A.; Hobart, G.W.; Luther, J.E.; Hermosilla, T.; Griffiths, P.; Coops, N.C.; Hall, R.J.; Hostert, P.; Dyk, A.; et al. Pixel-Based Image Compositing for Large-Area Dense Time Series Applications and Science. Canadian Journal of RemoteSensing2014, 40, 192-212, doi:10.1080/07038992.2014.945827.
13. Francini, S.; Hermosilla, T.; Coops, N.C.; Wulder, M.A.; White, J.C.; Chirici, G. An Assessment Approach for Pixel-Based Image Composites. ISPRS Journal of Photogrammetry and RemoteSensing2023, 202, 1-12, doi:10.1016/j.isprsjprs.2023.06.002.
14. Qiu, S.; Zhu, Z.; Olofsson, P.; Woodcock, C.E.; Jin, S. Evaluation of Landsat Image Compositing Algorithms. Remote SensEnviron2023, 285, 113375, doi:10.1016/j.rse.2022.113375.
15. Francini, S.; Hermosilla, T.; Coops, N.C.; Wulder, M.A.; White, J.C.; Chirici, G. An Assessment Approach for Pixel-Based Image Composites. ISPRS Journal of Photogrammetry and RemoteSensing2023, 202, 1-12, doi:10.1016/j.isprsjprs.2023.06.002.
16. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google Earth Engine: Planetary-Scale Geospatial Analysis for Everyone. Remote SensEnviron2017, 202, 18-27, doi:10.1016/j.rse.2017.06.031.
17. Bozzini, A.; Francini, S.; Chirici, G.; Battisti, A.; Faccoli, M. Spruce Bark Beetle Outbreak Prediction through Automatic Classification of Sentinel-2 Imagery. Forests2023, 14, 1116, doi:10.3390/f14061116.
18. Chirici, G.; Giannetti, F.; McRoberts, R.E.; Travaglini, D.; Pecchi, M.; Maselli, F.; Chiesi, M.; Corona, P. Wall-to-Wall Spatial Prediction of Growing Stock Volume Based on Italian National Forest Inventory Plots and Remotely Sensed Data. International Journal of Applied Earth Observation andGeoinformation2020, 84, 101959, doi:10.1016/j.jag.2019.101959.
19. Vangi, E.; D’Amico, G.; Francini, S.; Borghi, C.; Giannetti, F.; Corona, P.; Marchetti, M.; Travaglini, D.; Pellis, G.; Vitullo, M.; et al. Large-Scale High-Resolution Yearly Modeling of Forest Growing Stock Volume and above-Ground Carbon Pool. Environmental Modelling &Software2023, 159, 105580, doi:10.1016/j.envsoft.2022.105580.
20. Giannetti, F.; Laschi, A.; Zorzi, I.; Foderi, C.; Cenni, E.; Guadagnino, C.; Pinzani, G.; Ermini, F.; Bottalico, F.; Milazzo, G.; et al. Forest Sharing® as an Innovative Facility for Sustainable Forest Management of Fragmented Forest Properties: First Results of Its Implementation. Land (Basel)2023, 12, 521, doi:10.3390/land12030521.
21. Parisi, F.; Vangi, E.; Francini, S.; D’Amico, G.; Chirici, G.; Marchetti, M.; Lombardi, F.; Travaglini, D.; Ravera, S.; De Santis, E.; et al. Sentinel-2 Time Series Analysis for Monitoring Multi-Taxon Biodiversity in Mountain Beech Forests. Frontiers in Forests and GlobalChange2023, 6, doi:10.3389/ffgc.2023.1020477.
22. Flood, N. Seasonal Composite Landsat TM/ETM+ Images Using the Medoid (a Multi-Dimensional Median). Remote Sens (Basel)2013, 5, 6481-6500, doi:10.3390/rs5126481.
23. Kennedy, R.; Yang, Z.; Gorelick, N.; Braaten, J.; Cavalcante, L.; Cohen, W.; Healey, S. Implementation of the LandTrendr Algorithm on Google Earth Engine. Remote Sens (Basel)2018, 10, 691, doi:10.3390/rs10050691.
24. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google Earth Engine: Planetary-Scale Geospatial Analysis for Everyone. Remote SensEnviron2017, 202, 18-27, doi:10.1016/j.rse.2017.06.031.
25. Anenberg, S.C.; Horowitz, L.W.; Tong, D.Q.; West, J.J. An Estimate of the Global Burden of Anthropogenic Ozone and Fine Particulate Matter on Premature Human Mortality Using Atmospheric Modeling. Environ HealthPerspect2010, 118, 1189-1195, doi:10.1289/ehp.0901220.
26. Pauleit, S.; Liu, L.; Ahern, J.; Kazmierczak, A. Multifunctional Green Infrastructure Planning to Promote Ecological Services in the City. In Urban Ecology; Oxford University Press, 2011; pp. 272-285.

27. Bottalico, F.; Chirici, G.; Giannetti, F.; De Marco, A.; Nocentini, S.; Paoletti, E.; Salbitano, F.; Sanesi, G.; Serenelli, C.; Travaglini, D. Air Pollution Removal by Green Infrastructures and Urban Forests in the City of Florence. Agriculture and Agricultural Science Procedia 2016, 8, 243-251, doi:10.1016/j.aaspro.2016.02.099.