Fig 1.
Abridged definitions for local climate zones [18].
©American Meteorological Society. Used with permission.
Table 1.
Some of the parameter values associated with LCZ types from [18].
Columns represent the percentage of built (λB [%], ratio of building plan area to total plan area), impervious (λI [%], ratio of impervious plan area (paved, rock) to total plan area) and vegetated (λV [%], ratio of pervious plan area (bare soil, vegetation, water) to total plan area) land-cover and mean height of roughness elements (H [m], geometric average of building heights (LCZs 1–10) and tree/plant heights), sky view factor (SVF) and anthropogenic heat flux (AHF [W m−2]). The last column presents the total impervious surface density (IMD [%]), calculated as the sum of the outer ranges of λB and λI.
Table 2.
Available training areas for European cities.
Fig 2.
Colours depict the number of training areas per LCZ class and city. Values between brackets are the total number of training areas per city.
Fig 3.
Illustration of training areas: Toulouse (France, left panel), Birmingham (United Kingdom, middle panel), and Athens (Greece, right panel).
Colours denote the LCZ classes. Sources background map: Esri, DigitalGlobe, Earthstar Geographics, CNES/Airbus DS, GeoEye, USDA FSA, USGS, Aerogrid, IGN, IGP, and the GIS User Community.
Table 3.
Description of the input features.
All products are selected for the period 2015-01-01 to 2017-12-31 (except indicated otherwise), and are resampled to a horizontal resolution of 100 m.
Fig 4.
Input feature importance ranking.
The vertical dashed line separates the top-10 features. The error bars are derived from the bootstrapping procedure. Note that the spectral indices are also derived from Landsat 8; they are given a different color for clarity.
Fig 5.
Feature importance ranking per LCZ class.
Left panel: Built LCZs, right panel: Natural LCZs. Thick lines and shadings respectively present the importance mean and standard deviation from the bootstrapping.
Fig 6.
Classification accuracy assessment.
Results are based on All pixels and all input features (left panel), All pixels and the top 10 input features (middle panel), and the reduced pixels and all input features (right panel).
Fig 7.
European local climate zone map.
A selection of cities are depicted in more details in the Figures below.
Fig 8.
Binary urban maps extracted from ESA CCI and LCZ.
Black refers to urban, white to non-urban. For the LCZs, classes 1-10 with the exception of 9 (Sparsely built) are considered urban. Sources background maps: Esri, DigitalGlobe, Earthstar Geographics, CNES/Airbus DS, GeoEye, USDA FSA, USGS, Aerogrid, IGN, IGP, and the GIS User Community.
Fig 9.
Accuracy assessment urban land cover for all European countries.
Abbreviations refer to Large Scale International Boundary (LSIB) country codes: AL: Albania, AN: Andorra, AU: Austria, BE: Belgium, BK: Bosnia & Herzegovina, BO: Belarus, BU: Bulgaria, CY: Cyprus, DA: Denmark, EI: Ireland, EN: Estonia, EZ: Czech Republic, FI: Finland, FR: France, GM: Germany, GR: Greece, HR: Croatia, HU: Hungary, IC: Iceland, IT: Italy, KV: Kosovo, LG: Latvia, LH: Lithuania, LO: Slovakia, LS: Liechtenstein, LU: Luxembourg, MD: Moldova, MJ: Montenegro, MK: Macedonia, MN: Monaco, MT: Malta, NL: The Netherlands, NO: Norway, PL: Poland, PO: Portugal, RI: Serbia, RO: Romania, SI: Slovenia, SM: San Marino, SP: Spain, SW: Sweden, SZ: Switzerland, TU: Turkey, UK: United Kingdom, UP: Ukraine, VT: Vatican City.
Fig 10.
LCZ map (left panel) and EU Urban Atlas building height (right panel) for Budapest (Hungary).
Table 4.
Mean building height (m) per built LCZ class for the Urban Atlas cities.
Reference values from [18] are provided in the top row, mean and standard deviation (St. Dev.) across all Urban Atlas cities are provided on top of the individual city listings.
Fig 11.
LCZ map, impervious density (IMD, %) and mean annual anthropogenic heat flux (AHF, W/m2) for Germany.
Table 5.
Impervious density (IMD) and mean annual anthropogenic heat flux (AHF) urban canopy parameters.
Mean and standard deviation are calculated over the entire European domain (LCZ map) and are stratified per LCZ class. Reference values from [18] are provided as well.
Table 6.
Mean sky view factor (%) per built LCZ class.
Reference values from [18] are provided in the top row, mean and standard deviation (St. Dev.) across all selected cities are provided on top of the individual city listings.
Fig 12.
Sky view factors for 15 European cities.
Sources background map: Esri, DigitalGlobe, Earthstar Geographics, CNES/Airbus DS, GeoEye, USDA FSA, USGS, Aerogrid, IGN, IGP, and the GIS User Community.
Fig 13.
Local climate zones for 15 European cities.
Sources background map: Esri, DigitalGlobe, Earthstar Geographics, CNES/Airbus DS, GeoEye, USDA FSA, USGS, Aerogrid, IGN, IGP, and the GIS User Community.