Spatially and thematically detailed land use maps are of special importance to study and manage populated mountain regions. Due to the complex terrain, high elevational gradients as well as differences in land demand, these regions are characterized by a high density of different land uses that form heterogeneous landscapes. Here, we present a new highly detailed land use/landcover map for the areas included in the European Strategy for the Alpine Region. The map has a spatial resolution of up to 5 m and a temporal extent from 2015 to 2020. It was created by aggregating 15 high-resolution layers resulting in 65 land use/cover classes. The overall map accuracy was assessed at 88.8%. The large number of land use classes and the high spatial resolution allow an easy customization of the map for research and management purposes, making it useable by a broad audience for various applications. Our map shows that by combining theme specific “high-resolution” land use products to build a comprehensive land use/land cover map, a high thematic and spatial detail can be achieved.
Background & Summary
Land use/land cover (LULC) maps present information on the physical land types that characterize the surface of the earth (i.e., land cover) and describe how humans use this land (i.e., land use)1. These maps allow to monitor land cover changes and land allocation for agriculture, urban development, nature conservation et cetera, and to assess the provision of ecosystem services and habitats2,3. The use of high resolution LULC maps is particularly important in those areas that are characterized by complex landscapes and unique geo-topographic conditions, such as mountain ranges. These areas face multiple challenges, such as biodiversity loss, a high vulnerability to climate change, and negative demographic trends, and are therefore in need of accurate and updated LULC information for their effective management4,5,6.
The European Alps represent a unique environment characterized by a great variety of ecosystems and landscapes that are increasingly threatened by different pressures7. Land use intensification in the valley bottoms is affecting the presence of green infrastructure elements such as hedgerows and riparian areas, leading to the isolation of natural habitats and a decrease in ecological connectivity8. The increase in temperatures caused by climate change is progressively opening to agriculture new areas at higher elevations, causing the upward shift of economically valuable crops9 as well as a natural shift in habitats10. Rural abandonment is causing the progressive marginalization of large areas, while urban areas are experiencing intensive urbanization with a significantly growing number of inhabitants11. To tackle these challenges, it is important to develop specific tools and data that inform policymaking, research, land planning and resource management2.
The availability of LULC maps of the European Alps that have both, a high thematic and spatial detail (i.e., maps characterized by a high spatial resolution and many LULC classes) is, however, limited. Indeed, even if the increased accessibility of “high-resolution” satellite imagery, of powerful computing capabilities, and of new computing techniques (e.g., deep learning) has brought new opportunities for the automated mapping of land cover3, LULC maps of the Alps still usually only fulfill one of the two desired characteristics. An example of a thematically very detailed LULC map is the Corine Land Cover map (CLC12 that includes 44 LULC classes13. However, from the spatial point of view, CLC has only a medium resolution (100 m, with a minimum mapping unit (MMU) of 25 ha), which limits its usability in mountain areas. Conversely, the map recently developed by Malinowski et al. 2020 has a high spatial resolution (10 m) but only 13 LULC classes14. The same holds true for other recent LULC maps that include the European Alps15,16,17. To improve both the spatial and the thematic detail of existent LULC maps, various methodologies have been developed by researchers: Rosina et al.18, for example, used a CLC refinement approach by integrating multiple datasets with higher spatial resolution and decreased the MMU from 25 to 1 ha, Pigaiani & Batista e Silva 202119 applied a similar methodology increasing the spatial resolution to 50 m. Using similar procedures many other LULC maps have been produced, mostly focusing at the national and subnational level20,21,22,23. However, there has been no attempt to create a specific LULC map focused on the entire Alps with both a high spatial and thematic resolution.
Here, we present the first spatially and thematically highly detailed LULC map for the European Alps. We collected, harmonized and combined freely available datasets from 11 different sources to build a high-resolution map that includes 65 different LULC classes. By including small LULC features, this map is intended to support a wide range of analyses spanning from research to land management and decision making. For example, the spatial impact of linear elements such as roads, rivers and hedges can be analyzed and included in ecological connectivity mapping models or ecosystem service assessments. Local administrations can also benefit from the high resolution of the map, which can support landscape planning and resource-efficient management.
As a reference to define the extent of the European Alps we used the area included in the European Strategy for the Alpine Region (EUSALP). This area covers a total surface of more than 440,000 km², including 7 nations and 48 administrative regions (Fig. 1).
The creation of the EUSALP map included the following main steps: firstly, we selected freely available datasets that covered our area of interest. Secondly, we adapted the retrieved datasets with minor alterations in order to combine high-resolution datasets from different sources. Thirdly, we harmonized all the layers using the same spatial reference system and resolution. As a last step we mosaicked the layers using a specific hierarchy based on codes given to each LULC class (Fig. 2). Finally, we validated the resulting map using an area-weighted confusion matrix approach.
In the first step, we collected all openly available LULC datasets that cover the whole EUSALP macro region. The following collection criteria were applied: a reference year between 2015 and 2020, a thematic accuracy higher than 80%, and a high spatial resolution (10 m). The selected data are presented in Table 1 (the area covered by the single datasets is shown in Figure S1).
For certain data layers (i.e., OSM Roads & Railways, EU Hydro, HRL Grassland) some adaptations were necessary prior to harmonization. Linear features (i.e., roads, railways) from the OSM were converted into polygon features by assigning the width defined by the OSM specifications (6 m width for secondary and tertiary roads as well as tracks and field roads, 10 m width for primary roads and railways, 20 m width for motorways and trunks), all tunnels were excluded. The EU Hydro River polylines were converted into polygon features using a width according to the Strahler Stream Order24. To characterize the use intensity of grasslands, that in the HRL Grassland dataset25 are defined using only a binary grassland/non-grassland classification, we divided them into three LULC classes based on elevation and slope. The classification was based on the following criteria: managed grassland (<2000m elevation and <26° slope), seminatural grassland (<2000 m elevation and >26° slope), Alpine natural grassland (>2000 m elevation)26,27,28. For the calculation we used the European Digital Elevation Model (EU-DEM), version 1.129.
We harmonized all the layers using the same reference system and resolution to ensure the geographical consistency of the final dataset. We projected the selected raster datasets into the same spatial reference system (EPSG:3035 ETRS89/ETRS-LAEA) and then resampled them to a resolution of 5 m using the nearest neighbor algorithm to ensure that the original pixel values are preserved, and no interpolated values are created. We also projected the vector-based datasets to EPSG:3035 and rasterized them at 5 m resolution. Next, we snapped all the layers to the same reference raster layer to ensure cell alignment. Resolution: We did not perform resampling to improve the resolution of the input data, but to allow an increase in the thematic detail so that landscape features smaller than 100 m2 and 10 m width (e.g., buildings, roads, hedgerows, small streams) can be represented on the final map. Therefore, only in and near buildings, roads and linear elements, a map resolution of 5 m can be expected (which corresponds to approximately 15–20% of the map area).
Data structuring and classification
We used the ESRI Land Cover Map 202030 as a base layer to build our LULC map, as it is the only selected land cover dataset with complete geographical coverage for the whole research area. We added land use information to this dataset using the data presented in Table 1. To combine the layers, we first assigned specific codes to each LULC class value in all datasets (Table 2). Reoccurring LULC types across different datasets were assigned the same code (since MMU is very small and mostly pixel-based no further harmonization steps of land use types were necessary). We then overlayed the data by applying a specific layer hierarchy (Table 3) following a decision tree based on data accuracy (i.e., level of thematic and spatial detail). By assigning the value of the highest-ranking layer, we could decide which information to show on the final map, to control the uncertainties built in specific layers (e.g., presence of green linear elements in cultivated areas and grassland) and to include small LULC features (e.g. roads, single buildings, small streams in forests or grassland). All the work was done using ArcGIS Desktop 10.8.
We present an easily accessible and freely available high resolution LULC map of the EUSALP region that can be used to support researchers and practitioners in the field of landscape planning and management. The data is freely available through the Figshare data publisher31.
It includes two raster geospatial files that contain the EUSALP high resolution LULC map and a reference to the source dataset used to define each of the pixel values. The file has a pre-built color palette included to classify the 65 classes of the LULC map. The files included are:
EUSALP_LULC_05 m_2020.tif: a .tiff file that includes the classification of the EUSALP area based on 65 LULC classes.
EUSALP_LULC_data_sources.tif: a .tiff file that includes the reference information about the dataset used to define each pixel of the map (dataset name, publication year, reference year).
EUSALP_LULC_classes.csv: a .csv file that includes the code and description of the 65 classes of the LULC map.
The primary purpose of the present validation procedure is not to assess the individual LULC classes, but to ensure that the harmonization steps and hierarchy in combining the data are still capable of producing accurate LULC information, given that the map is built upon already validated and published input data. For more details on the validation and accuracy of the input data, see Table S2.
We applied a stratified random sampling design using the Eurostat LUCAS 2018 survey data points as the reference dataset34. In total, 32,227 LUCAS 2018 survey points are located within the EUSALP map extent. From these, a random selection of survey sites was made using the subset feature analysis tool in ArcGIS. The number of sites to be allocated to each LULC class was calculated as a function of their area proportion in the EUSALP map. In this way, the sampling design is not only systematic but also stratified. A minimum number of 20 sample units per LULC class was defined to ensure that even small strata were represented in the sample. However, for some strata there were no reference points available (41200, 42200). In the end, 2300 LUCAS 2018 points were randomly selected (see Figure S2).
An initial blind interpretation was performed, which consists in constructing the validation data without any knowledge of the map layer being evaluated. This was done by evaluating LULC on the reference points using EUSALPs’ LULC map classification codes. ESRI World Imagery (https://services.arcgisonline.com/ArcGIS/rest/services/World_Imagery/MapServer) and LUCAS 2018 thematic information were used for this first round of classification. As this method may underestimate the accuracy for complex and heterogeneous land use classes and potential land use changes (especially on arable land) or class definitions, we then used a plausibility approach, which is applied on all sample units that result in disagreement with the EUSALP LULC Map. This step consists in checking both classified values (blind validation and EUSALP map) for plausibility within the accepted product specifications, without knowing the corresponding classification source.
The overall map accuracy was assessed using an error matrix approach35. The producer accuracy (PA) and the user accuracy (UA) for each LULC class were evaluated in an area-weighted confusion matrix with 95% confidence interval. We obtained an overall accuracy (OA) of 88.8% ± 1.8 for the plausibility approach (Tables 4, 6, S3), which is a good result that meets validation standards, even though the blind evaluation showed substantially lower overall accuracy (64.8% ± 3.7) (Tables 5, S4).
For classes 41200, 42200, 52100 and 32200 there were too few sample points available. Therefore, these classes could not be properly validated35. However, this is of little concern as these LULC classes cover only 0.06% of the total map area. Only 17 reference points could not be classified.
The OA of the EUSALP LULC map is very similar to the OA of the various input datasets and it would be very unlikely that the output is better than the input. Therefore, we are confident that the map creation approach was successful and that the created dataset meets accuracy standards.
Insight into the temporal extent of the LULC data is given by using the EUSALP_LULC_data_sources.tif raster31, which shows the reference year of each map cell. Information on the reference year exists for each input data layers except for Open Street Map.
Logical and format consistency of our map is ensured by the harmonization steps each data input file has undergone (the MMU is pixel based, the Coordinate Reference System is set to EPSG 3035, Pixel size is set to 5 m). Overlap cannot occur due to the final data format.
Positional accuracy could not be assessed due to missing reference data with sufficient spatial accuracy. However, all of the input data used have been evaluated for positional accuracy during the validation process.
The EUSALP LULC map has a high potential for customization as the regrouping of the 65 LULC classes allows for interest-specific reclassifications in any GIS program. Due to the high level of detail, our map can be used even at the local scale, having a level of detail near artificial structures and settlements comparable to maps at 1:5,000 scale.
However, the EUSALP LULC map still holds some limits and improvement potential. Indeed, the time dimension of different data layers needs to be carefully considered when using the map: in fact, although corresponding to the newest available high-resolution data layers, the combined data are from different years. If time specificity is required, the user needs to refer to the Datasource layer (Fig. 3).
No custom code has been used during the generation and processing of this dataset.
Fisher, P. F. & Wadsworth, R. Land use and Land cover: Contradiction or Complement. Chapter 6 in Re-Presenting GIS, Wiley (2005).
Saah, D. et al. Land Cover Mapping in Data Scarce Environments: Challenges and Opportunities. Front. Environ. Sci. 7, 150 (2019).
Szantoi, Z. et al. Addressing the need for improved land cover map products for policy support. Environmental Science & Policy 112, 28–35 (2020).
Hock, R. et al. High Mountain Areas. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. Cambridge University Press, Cambridge, UK and New York, NY, USA, pp. 131–202. https://doi.org/10.1017/9781009157964.004 (2019).
Körner, C. Mountain biodiversity, its causes and function. Ambio 66, 11–1343 (2004).
Martín-López, B. et al. Nature’s contributions to people in mountains: A review. PLOS ONE 14, e0217847 (2019).
Egarter Vigl, L. et al. A multi-pressure analysis of ecosystem services for conservation planning in the Alps. Ecosystem Services 47 (2021).
Bartol, B. et al. Safeguarding open spaces in the Alpine region. Positionpaper ARL- Academy for Territorial Development in the Leibniz Association 133 https://www.arl-net.de/system/files/media-shop/pdf/pospapier/pospapier_133.pdf (2022).
Egarter Vigl, L. et al. Upward shifts in elevation – a winning strategy for mountain viticulture in the context of climate change? E3S Web Conf. 50, 02006 (2018).
Walther, G.-R., Beißner, S. & Pott, R. Climate Change and High Mountain Vegetation Shifts. in Mountain Ecosystems: Studies in Treeline Ecology (eds. Broll, G. & Keplin, B.) 77–96, https://doi.org/10.1007/3-540-27365-4_3 (Springer, 2005).
Tappeiner, U., Borsdorf, A. and Tasser, E.-I. Alpenatlas: Society, Economy, Environment = Atlas des Alpes = Atlante delle Alpi = Atlas Alp = Mapping the Alps. Spektrum, Akad. Verl. (2008).
European Union, Copernicus Land Monitoring Service (2022), European Environment Agency (EEA) -Dataset: CORINE Land Cover 1990-2018 https://land.copernicus.eu/pan-european/corine-land-cover.
Bielecka & Jenerowicz Intellectual Structure of CORINE Land Cover Research Applications in Web of Science: A Europe-Wide Review. Remote Sensing 11, 2017 (2019).
Malinowski, R. et al, Automated Production of a Land Cover/Use Map of Europe Based on Sentinel-2 Imagery, https://doi.org/10.3390/rs12213523 (2020).
Pflugmacher, D., Rabe, A., Peters, M. & Hostert, P. Mapping pan-European land cover using Landsat spectral-temporal metrics and the European LUCAS survey. Remote Sensing of Environment 221, 583–595 (2019).
Venter, Z. S. & Sydenham, M. A. K. Continental-Scale Land Cover Mapping at 10 m Resolution Over Europe (ELC10). Remote Sensing 13, 2301 (2021).
Brown, C. F. et al. Dynamic World, Near real-time global 10 m land use land cover mapping. Sci Data 9, 251 (2022).
Rosina, K. et al. Increasing the detail of European land use/cover data by combining heterogeneous data sets. International Journal of Digital Earth 13, 602–626 (2020).
Pigaiani, C. & Batista e Silva, F., The LUISA Base Map 2018 - A geospatial data fusion approach to increase the detail of European land use/land cover data, EUR 30663 EN, Publications Office of the European Union, Luxembourg, JRC124621, https://doi.org/10.2760/503006 (2021).
Rüdisser, J. & Tasser, E. Landbedeckung Österreichs – Datenintegration und Modellierung. in Beiträge zum 23. AGIT-Symposium Salzburg. in (ed. Strobl, J.) Wichmann (2011).
Efthimiou, N. et al. Developing a high-resolution land use/land cover map by upgrading CORINE’s agricultural components using detailed national and pan-European datasets. Geocarto International 1–36, https://doi.org/10.1080/10106049.2022.2041107 (2022).
Fonte, C. C. et al. Generating Up-to-Date and Detailed Land Use and Land Cover Maps Using OpenStreetMap and GlobeLand30. ISPRS International Journal of Geo-Information 6, 125 (2017).
Pazúr, R. & Bolliger, J. “Enhanced Land use Datasets and Future Scenarios of Land Change for Slovakia.”. Data in Brief 14, 483–488 (2017).
Downing, J. et al. Global abundance and size distribution of streams and rivers. Inland Waters 2, 229–236, https://doi.org/10.5268/IW-2.4.502 (2012).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) - User Manual v.1.2 Grassland 2018 and Grassland change 2015–2018 (2020).
BMLFUW (Bundesministerium für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft): Deckungsbeiträge und Daten für die Betriebsplanung 2008. Wien. (2008).
Bosshard, A. and Häusler, L. Gezieltere Erschwernisbeiträge zugunsten einer flächendeckenden Bewirtschaftung. Vision Landwirtschaft/Ö + L, Oberwil-Lieli (2012).
Peratoner, G., Figl, U., Florian, C. and Mairhofer, F. Arbeitszeitbedarf bei der Futterproduktion in Südtirol. Laimburg Journal 3 (2021).
European Union, Copernicus Land Monitoring Service. European Digital Elevation Model (EU-DEM), Version 1.1. https://land.copernicus.eu/pan-european/satellite-derived-products/eu-dem/eu-dem-v1.1/ (2016).
Karra, K. et al Global land use/land cover with Sentinel 2 and deep learning In 2021 IEEE international geoscience and remote sensing symposium IGARSS (pp. 4704–4707). [Data file reference year 2020] retrieved from https://livingatlas.arcgis.com/landcoverexplorer (accessed 08/2021).
Marsoner, T., Simion, H., Giombini, V., Egarter Vigl, L., Candiago, S. A detailed land use/land cover map for the European Alps macro region. figshare https://doi.org/10.6084/m9.figshare.c.6357056.v1 (2023).
GMES Initial Operations/Copernicus Land monitoring services – Validation of products. Riparian zones land cover land use extension to Strahler 2 validation report. https://land.copernicus.eu/user-corner/technical-library/riparian-zones-land-cover-land-use-validation-report (2018).
GMES Initial Operations/Copernicus Land monitoring services – Validation of products. CLC2018/CLCC1218 validation report. https://land.copernicus.eu/user-corner/technical-library/clc-2018-and-clc-change-2012-2018-validation-report (2021).
d’Andrimont, R. et al. Harmonised LUCAS in-situ land cover and use database for field surveys from 2006 to 2018 in the European Union. Sci Data 7, 352 (2020).
Olofsson, P., Foody, G. M., Stehman, S. V. & Woodcock, C. E. Making better use of accuracy data in land change studies: Estimating accuracy and area and quantifying uncertainty using stratified estimation. Remote Sensing of Environment 129, 122–131 (2013).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: High Resolution Layer: Imperviousness Density (IMD) 2018, IMD-2018-010m- https://land.copernicus.eu/pan-european/high-resolution-layers/imperviousness/status-maps/imperviousness-density-2018 (2020).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: High Resolution Layer: Imperviousness Built-up (IBU) 2018, IBU-2018-010m-https://land.copernicus.eu/pan-european/high-resolution-layers/imperviousness/status-maps/impervious-built-up-2018 (2020).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: Grassland (GRA) 2018 - GRA-2018-010m- https://land.copernicus.eu/pan-european/high-resolution-layers/grassland/status-maps/grassland-2018 (2020).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: High Resolution Layer: Forest Type (FTY) 2018, FTY-2018-010m- https://land.copernicus.eu/pan-european/high-resolution-layers/forests/forest-type-1/status-maps/forest-type-2018 (2020).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: High Resolution Layer: Tree Cover Density (TCD) 2018, TCD-2018-010m- https://land.copernicus.eu/pan-european/high-resolution-layers/forests/tree-cover-density/status-maps/tree-cover-density-2018 (2020).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Datasets: High Resolution Layer: Water & Wetness (WAW) 2018, WAW-2018-010m- https://land.copernicus.eu/pan-european/high-resolution-layers/water-wetness/status-maps/water-wetness-2018 (2020).
d’Andrimont, R. et al. EUCROPMAP 2018. European Commission, Joint Research Centre (JRC) [Dataset] PID: http://data.europa.eu/89h/15f86c84-eae1-4723-8e00-c1b35c8f56b9 (2021).
OpenStreetMap contributors. Country shapefiles [Data files from 01.2020]. Retrieved from https://download.geofabrik.de (2020)
El-Ashmawy, K. Testing the positional accuracy of OpenStreetMap data for mapping applications. Geodesy and cartography 42, 25–30 (2016).
Tasking manager/validating data. Tasking Manager/Validating data - OpenStreetMap Wiki. (n.d.). https://wiki.openstreetmap.org/wiki/Tasking_Manager/Validating_data accessed December (2022).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: Urban Atlas LCLU 2018 https://land.copernicus.eu/local/urban-atlas/urban-atlas-2018 (2021).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: High Resolution Layer: Small Woody Features (SWF) 2015 v. 1.2, swf-2015-vec- https://land.copernicus.eu/pan-european/high-resolution-layers/small-woody-features/small-woody-features-2015 (2021).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: Riparian Zones status 2018 https://land.copernicus.eu/local/riparian-zones/riparian-zones-2018 (2021).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: Riparian Zones Green Linear Elements http://land.copernicus.eu/local/riparian-zones/green-linear-elements-gle-image/view (2015).
European Union, Copernicus Land Monitoring Service. European Environment Agency (EEA) -Dataset: EU-Hydro – River Network Database, Version 1.3 https://land.copernicus.eu/imagery-in-situ/eu-hydro/eu-hydro-river-network-database (2020).
GMES Initial Operations/Copernicus Land monitoring services – Validation of products. EU-HYDRO VALIDATION REPORT 1.1. (2017).
This work was supported by the European Regional Development Fund through the Interreg Alpine Space Programme (‘LUIGI | Linking Urban and Inner-alpine Green Infrastructure’, project number ASP 863). Sebastian Candiago’s PhD grant was co-financed by the European Regional Development Fund through the Interreg Alpine Space Programme (‘AlpES | Alpine Ecosystem Services – mapping, maintenance, management’, project number ASP 183), and the Interreg V-A ITA-AUT programme (REBECKA, project number ITAT1002). The authors thank the Department of Innovation, Research and University of the Autonomous Province of Bozen/Bolzano for covering the Open Access publication costs.
The authors declare no competing interests.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Marsoner, T., Simion, H., Giombini, V. et al. A detailed land use/land cover map for the European Alps macro region. Sci Data 10, 468 (2023). https://doi.org/10.1038/s41597-023-02344-3