Skip to main content
MethodsX logoLink to MethodsX
. 2020 Nov 19;7:101150. doi: 10.1016/j.mex.2020.101150

Local climate zones classification method from Copernicus land monitoring service datasets: An ArcGIS-based toolbox

Ana Oliveira a,, António Lopes b, Samuel Niza a
PMCID: PMC7718173  PMID: 33304834

Abstract

Local Climate Zones (LCZ) have become a worldwide standard for identifying land cover classes, according to their climate-relevant morphological parameters. The LCZ's are mostly used to evaluate urban climate performance, particularly the relationship between the urban heat island effect (UHI) and the characteristics of the built-up environment. The World Urban Database and Access Portal Tools (WUDAPT) has provided a supervised LCZ classification method based only on moderate resolution free satellite imagery, mostly Landsat 7 or 8 (30 m pixel size, in the visible spectrum brands); however, its’ results are less accurate for European cities. Conversely, alternative geographic information system (GIS)-based methods developed so far require information that is hardly available to all, such as building footprints or heights. Here, the ArcGIS based LCZ from Copernicus Toolbox (LCZC) provides an alternative classification method that uses only freely accessible information from the Copernicus Land Monitoring Service (CLMS), being possible to replicate it in 800 European urban locations. The method combines Urban Atlas (UA) and Corine Land Cover (CLC) with Tree Cover Density, Dominant Leaf Type and Grassland information, to produce a higher-resolution baseline shapefile that is classified according to each feature's dominant characteristics. The LCZC toolbox output is a LCZ raster map. It has been validated in five European cities: Athens, Barcelona, Lisbon, Marseille, and Naples.

  • The LCZC toolbox provides an alternative LCZ GIS-based classification, based on freely accessible CLMS datasets.

  • The use of CLMS shapefile higher-resolution inputs, particularly the UA and CLC datasets, ensures an output LCZ map that has greater detail and higher accuracy.

  • The availability of CLMS information in 800 European urban areas guarantees that the method can be replicated in those locations.

Keywords: Local climate zones, ArcGIS toolbox, Land cover classification, Copernicus, Urban atlas, Corine land cover, Urban climate

Graphical abstract

Image, graphical abstract


Specifications table

Subject Area: Environmental Science
More specific subject area: Urban Climate
Method name: Local Climate Zones from Copernicus ArcGIS Toolbox (LCZC)
Name and reference of original method: The LCZC toolbox delivers classified maps of the Local Climate Zones scheme by Stewart et al. [1,2] and is an alternative open access method to the World Urban Database and Access Portal Tools (WUDAPT) [3]
Resource availability: Requirements: ArcGIS Desktop version10.0 (or later), with Advanced License (http://desktop.arcgis.com/en/arcmap/10.3/get-started/system-requirements/arcgis-desktop-system-requirements.htm)
The LCZC toolbox is provided with this article, as well as its’ python scripts retrieved through the ArcGIS export functionality.

Method details

The Local Climate Zones (LCZ) from Copernicus Toolbox (LCZC) method entails a sequence of steps to reclassify several Copernicus Land Monitoring Service (CLMS) layers into a LCZ-based [1,2] classification, that can be used in urban climate-related studies, in 800 European urban regions. The method aims to provide an alternative solution to the satellite-based World Urban Database and Access Portal Tools (WUDAPT) supervised classification process [3], ensuring the greater accuracy and higher spatial resolution of Geographic Information Systems (GIS)-based methods [4], [5], [6] while preserving the ability to be freely reproducible. To process the LCZC tool, ArcGIS software with Advanced License is necessary. The list of inputs the LCZC requires is available in Table 1; all are mandatory, except Building Height (BH) which is only available from CLMS for capital cities. The toolbox is provided as supplementary material to this article in ArcGIS Toolbox format (.tbx) (Appendix 1), as well as the corresponding python scripts (Appendix 2), as exported through the Model Builder Arc-GIS functionalities.

Table 1.

List of land use and land cover datasets used in this study.

Dataset Version Release date Spatial resolution Original format Reference Overall accuracy (%)
Urban Atlas (UA) 2012 2016–08–04 10 m Vector Shapefile [7] 88.03- 94.19 3
UA Building Height (BH)1 2012 2018–04–19 10m Raster GeoTiff [8] n.a.
Corine Land Cover (CLC) 2012 2016–09–19 100 m Vector Shapefile [9] 83.6 - 89.7 4
Imperviousness Density (IMD) 2015 2018–03–22 20 m Raster GeoTiff [10] 89.63 - 98.37 3
Tree Cover Density (TCD) 2015 2018–03–22 20 m Raster GeoTiff [11] 85.03 - 92.27 3
Dominant Leaf Type (DLT) 2015 2018–04–13 20 m Raster GeoTiff [12] 85.30 - 95.63 3
Grassland (GRA) 2015 2018–09–09 20 m Raster GeoTiff [13] 85.50 - 92.67 3
OpenStreetMap (OSM) n/a n/a n/a Vector Shapefile [14] n.a.
1

optional dataset

2shapefile extracted from OSM, including land-use field column.

3

as per [15].

4

as per [16].

Due to its’ spatial resolution, a combination of the Urban Atlas (UA) and Corine Land Cover (CLC) shapefile datasets was chosen to establish the LCZ baseline vector layer for the procedure, and its’ shapefile format was preserved throughout the process. Additional layers related to vegetation were also combined in the model to better distinguish non-urban classes – Tree Cover Density (TCD), Dominant Leaf Type (DLT) and Grassland (GRA). These were all converted from raster to vector shapefiles, classified according to equivalent LCZ based classes, and resulting maps were merged into the LCZ baseline layer, to quantify the dominant class in each polygon. Most built-up LCZ classes (LCZ's 1–10) were reclassified directly from the UA classes, by comparing both classifications’ specifications in terms of built-up density, imperviousness degree (IMD) and typical land use/cover. As UA methodology does not allow to distinguish LCZ's 8 and 10, thus additional land use information from OpenStreetMap (OSM) was used to solve that specific gap. On the other hand, non-built-up land cover types (LCZ's A-G) were reclassified according to conditional sentences that filter through the CLC and the vegetation-based classes. The reclassification process was subject to several iterations between algorithm testing and correction, and it was found that combining CLC classes with the High-Resolution Layers related to Tree Cover (i.e. TCD, DLT and GRA) contributed to greater accuracy. Figure 1 summarizes the GIS-based workflow and Figure 2 illustrates the principal correspondence flows between UA and LCZ classes, from the example of Lisbon, where the flow thickness represent the proportion of each class. The method was implemented in ArcGIS software, using the Model Builder functionalities to produce the LCZC custom toolbox.

Fig. 1.

Fig 1

LCZ GIS-based classification workflow from Copernicus Land Monitoring Service (CLMS) datasets. Detailed diagram based on Oliveira et al. [5].

Fig. 2.

Fig 2

Sankey Diagram of the LCZ reclassification algorithm – an example of Lisbon's LCZ_v1 map. The flows represent the correspondence between UA (left codes) and LCZ (right codes) classes, and their thickness depicts the proportion of the surface area. The text in boxes corresponds to the LCZ reclassification syntax. Detailed diagram based on Oliveira et al. [5].

The method involves five steps to complete the LCZ classification, from pre-processing of the datasets to the final LCZ maps. If several CMLS layer tiles are needed to contain a study region, they need to be merged into a new mosaic raster previously. The OSM information also must be retrieved previously in shapefile format, with the field column land-use.

Besides using ArcGIS-based tools for data conversion and merging, the LCZC workflow involves several python-based reclassification steps. Due to its’ extension, these steps are provided as supplementary material Appendix 2, in python-format, for easier access to the algorithm conditional sentences details.

The LCZC tool includes 11 separate custom models, for more proficient use of computational capabilities. These models process the input data in 5 workflow steps, as follows:

  • STEP 1.
    Input raster layers are preprocessed, which includes:
    • STEP 1.1.
      Pre-Processing: Clips the Copernicus High-Resolution raster Layers TDC, DLT, GRA and IMD according to the UA Boundary, and converts the clipped result to polygon shapefile (see Fig. 3). The raster to polygon conversion is based on a regular squared polygon grid, in which the grid cells have the same size of the original raster pixels, ensuring that values are unchanged (i.e., each polygon feature corresponds to a pixel of equivalent size and value).
    • STEP 1.2.
      Pre-Processing: Polygons are reclassified into relevant LCZ-based classes, according to conditional algorithms applied to the raster's gridcode values (see Fig. 4).
  • STEP 2.
    LCZ baseline shapefile is assembled:
    • STEP 2.1.
      LCZ Baseline: Assembles the LCZ baseline polygon layer, based on conditional selections and merging functions, applied to the UA, CLC and OSM inputs (see Fig. 5). UA features are chosen to represent urban classes, while CLC features are preferred in non-urban land cover typologies. Selected features of both layers are merged through the Identity tool. The OSM land use information is added through the Spatial Join function.
    • STEP 2.2.
      LCZ Baseline: Calculates area per IMD class, adding the resulting field to the LCZ baseline polygon layer (obtained in the previous STEP 2.1) (see Fig. 6).
    • STEP 2.3.
      LCZ Baseline: Calculates area per TCD, DLT and GRA classes, adding the resulting field to the LCZ baseline polygon layer, from previous 2.2 (see Fig. 7).
  • STEP 3.

    LCZ classification: LCZ baseline dataset from previous STEP2.3 is reclassified as two LCZ classification fields (LCVZv1 and LCZv1_leaf), both without building height information (i.e. Urban LCZ classes are merged by density, e.g. LCZ's 1, 2 and 3 are classified as LCZ '123′). The LCZ classification fields created are named LCZ (containing classes 123, 456, 8, 9, 10, A, B, C, D, E, F and G), and LCZ_leaf (containing the same classes but separating LCZ's A and B according to DLT, deciduous or coniferous) (see Figure 8).

  • STEP 4.

    LCZ with Building Height: converts, classifies and merges the BH raster layer with the STEP3 LCZ classification output; LCZ's 123 and 456 are reclassified according to dominant BH: LCZ 1 Compact high-rise LCZ 2 Compact midrise LCZ 3 Compact low-rise LCZ 4 Open high-rise LCZ 5 Open midrise LCZ 6 Open low-rise. Urban features without BH information remain as LCZ 123 and LCZ 456. The two fields created, LCZ_BH and LCZ_leaf_BH, correspond the reclassification of the two fields from the previous STEP3 (see Figure 9).

  • STEP 5.

    LCZ conversion to Raster: converts the shapefile LCZ classification into a raster, according to user's pixel size specifications. STEP5 has 4 alternative models, according to the LCZ desired content: (a) LCZv1 - classification without DLT or BH; (b) LCZv1_leaf - classification with DLT but without BH; (c) LCZv1_BH classification with BH, but without DLT; and (d) LCZv1_leaf_BH classification with BH and DLT. The resulting raster uses a numerical codification for LCZ's classes, and the corresponding attribute table also contains 2 string fields with LCZ's names and description, and a numerical field with the corresponding area (m2) (see Figure 10 and Table 2 ).

Fig. 3.

Fig 3

ArcGIS toolbox model, STEP1.1. Pre-Processing: Copernicus High-resolution layers GRA, DLT, TCD, IMD are clipped and converted to shapefile.

Fig. 4.

Fig 4

ArcGIS toolbox model, STEP1.2. Pre-Processing: Copernicus High-resolution layers GRA, DLT, TCD and IMD are classified into LCZ-based classes.

Fig. 5.

Fig 5

ArcGIS toolbox model, STEP2.1. LCZ Baseline: UA and CLC layers are subject to class selection and merged into an LCZ baseline feature layer. OSM land-use field is added through the Spatial Join function.

Fig. 6.

Fig 6

ArcGIS toolbox model, STEP2.2. LCZ Baseline: LCZ baseline is subject to adding fields that account the area per IMD LCZ-based class.

Fig. 7.

Fig 7

ArcGIS toolbox model, STEP2.3. LCZ Baseline: the same procedure from STEP2.2 is applied to the TCD, DLT and GRA shapefiles, by adding fields to the LCZ baseline shapefile and calculating the area of each LCZ-based class.

Fig. 8.

Fig 8

ArcGIS toolbox model, STEP3. LCZ Classification: LCZ baseline is reclassified according to the dominant LCZ class, without building height information. Two classification fields are created in the attribute table: LCZ and LCZ_leaf (LCZ with added DLT classification).

Fig. 9.

Fig 9

ArcGIS toolbox model, STEP4. LCZ with Building Height: reclassifies the LCZ baseline LCZ and LCZ_leaf fields, by decoding LCZ's 123 and 456 polygon features according to dominant BH class.

Fig. 10.

Fig 10

ArcGIS toolbox model, STEP5a. LCZ conversion to Rater: the classified LCZ baseline shapefile is converted to Raster format, according to the user's chosen spatial resolution. There are 4 STEP5 models, one per each alternative LCZ field.

Table 2.

Alternative LCZ classifications and the corresponding list of LCZ classes contained in the attribute tables1.

Numerical code LCZ class Description LCZv1.tif
Without BH
Without DLT
LCZv1_leaf.tif
With DLT
Without BH
LCZv1_BH
With BH
Without DLT
LCZv1_BH_leaf
With DLT
Without BH
100,100 1 Compact high-rise X X
100,200 2 Compact midrise X X
100,300 3 Compact low-rise X X
100,123 123 Compact mix-rise X X X X
100,400 4 Open high-rise X X
100,500 5 Open midrise X X
100,600 6 Open low-rise X X
100,456 456 Open mix-rise X X X X
100,800 8 Large low-rise X X X X
100,900 9 Sparsely built X X X X
101,000 10 Heavy industry X X X X
110,100 A Dense trees X X X X
110,110 A coniferous Dense trees coniferous X X
110,120 A deciduous Dense trees deciduous X X
110,200 B Scattered trees X X X X
110,210 B coniferous Scattered trees coniferous X X
110,220 B deciduous Scattered trees deciduous X X
110,300 C Bush scrub X X X X
110,400 D Low plants X X X X
110,500 E Bare rock or paved X X X X
110,600 F Bare soil or sand X X X X
110,700 G Water X X X X
1

X = Class contained in the dataset.

The presented method was tested in 5 Southern European cities [5]: Athens, Barcelona, Lisbon, Marseille and Naples [4]. Even though Stewart and Oke describe 300 m as a reasonable minimum radius for the LCZ classification [2,17], the resulting LCZ shapefile datasets were converted to a 50 m pixel raster format, where each pixel value depicts the LCZ class that has the greatest area. Each city's LCZ classification was subject to accuracy assessment analysis, by randomly selecting samples of pixels per class, stratified by surface area coverage. About 550 samples per each city were classified according to the dominant LCZ type (based on satellite true color imagery and 3D information from Google Earth). The sample's classification was compared with the LCZC toolbox output, and results re-arranged into a confusion matrix. Average overall accuracy (OA) was 81% and Kappa coefficient 0.79. Correctly classified pixels varied according to LCZ class, as built-up LCZ classes 1–10 revealed, on average, 90,0% agreement, but non-built-up LCZ classes A-G had fewer correctly classified sample pixels. The lowest accuracies occur in low-density vegetation types since only the dense trees class (LCZ A) proved to have 80% correct results, on average. LCZ 123 (compact urban fabric), LCZ 8 and LCZ 10 revealed the most noteworthy agreement, being correct in approximately 95% samples. LCZ 456 (open urban fabric) and LCZ 9 (sparsely built) were found to be less accurate, even though above the 80% threshold. This agrees with the great diversity of suburban neighbourhood typologies, more difficult to group in one class (an issue also present in the UA dataset). It should be noted that, while the LCZC toolbox uses the high-resolution CLMS layers to improve the Urban Atlas and Corine Land Cover miss-classification rates, particularly in rural areas, it is still limited by its inputs overall accuracy. This agrees with the fact that the resulting LCZ's maps revealed an overall accuracy that is slightly lower than that of its input datasets. Nonetheless, the LCZC toolbox aims to provide a readily available tool for LCZ mapping in European cities, improving the accuracy reached through satellite-based alternatives, and to be useful on a metropolitan scale.

Acknowledgments

AO acknowledeges “Fundação para a Ciência e Tecnologia (FCT) – Portugal” for providing financial support to the research, through the individual Ph.D. grant number PD/BD/52304/2013.

The Copernicus Land Monitoring Service datasets used in this study were produced “with funding by the European Union”.

Map data copyrighted OpenStreetMap contributors and available from https://www.openstreetmap.org.

We acknowledge the use of imagery provided by services from the Global Imagery Browse Services (GIBS), operated by the NASA/GSFC/Earth Science Data and Information System (ESDIS, https://earthdata.nasa.gov) with funding provided by NASA/HQ.

AO would like to thank all the Zephyrus research group for guidance, particularly Professors António Lopes and Ezequiel Correia (CEG, IGOT, Universidade de Lisboa) for providing helpful supervision throughout the development of the GIS-based tool.

AO gratefully acknowledges the editors and reviewers by their helpful suggestions to improve an earlier version of the manuscript.

Declaration of Competing Interest

The Authors confirm that there are no conflicts of interest.

Footnotes

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.mex.2020.101150.

Contributor Information

Ana Oliveira, Email: anappmoliveira@tecnico.ulisboa.pt.

António Lopes, Email: antonio.lopes@campus.ul.pt.

Samuel Niza, Email: samuel.niza@tecnico.ulisboa.pt.

Appendix. Supplementary materials

mmc1.zip (129.1KB, zip)
mmc2.zip (16.6KB, zip)

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

mmc1.zip (129.1KB, zip)
mmc2.zip (16.6KB, zip)

Articles from MethodsX are provided here courtesy of Elsevier

RESOURCES