Crowther_Nature_Files.zip This description pertains to the original download. Details on revised (newer) versions of the datasets are listed below. When more than one version of a file exists in Figshare, the original DOI will take users to the latest version, though each version technically has its own DOI. -- Two global maps (raster files) of tree density. These maps highlight how the number of trees varies across the world. One map was generated using biome-level models of tree density, and applied at the biome scale. The other map was generated using ecoregion-level models of tree density, and applied at the ecoregion scale. For this reason, transitions between biomes or between ecoregions may be unrealistically harsh, but large-scale estimates are robust (see Crowther et al 2015 and Glick et al 2016). At the outset, this study was intended to generate reliable estimates at broad spatial scales, which inherently comes at the cost of fine-scale precision. For this reason, country-scale (or larger) estimates are generally more robust than individual pixel-level estimates. Additionally, due to data limitations, estimates for Mangroves and Tropical coniferous forest (as identified by WWF and TNC) were generated using models constructed from Topical moist broadleaf forest data and Temperate coniferous forest data, respectively. Because we used ecological analogy, the estimates for these two biomes should be considered less reliable than those of other biomes . These two maps initially appeared in Crowther et al (2015), with the biome map being featured more prominently. Explicit publication of the data is associated with Glick et al (2016). As they are produced, updated versions of these datasets, as well as alternative formats, will be made available under Additional Versions (see below).

Methods: We collected over 420,000 ground-sources estimates of tree density from around the world. We then constructed linear regression models using vegetative, climatic, topographic, and anthropogenic variables to produce forest tree density estimates for all locations globally. All modeling was done in R. Mapping was done using R and ArcGIS 10.1.

Viewing Instructions: Load the files into an appropriate geographic information system (GIS). For the original download (ArcGIS geodatabase files), load the files into ArcGIS to view or export the data to other formats. Because these datasets are large and have a unique coordinate system that is not read by many GIS, we suggest loading them into an ArcGIS dataframe whose coordinate system matches that of the data (see File Format). For GeoTiff files (see Additional Versions), load them into any compatible GIS or image management program.

Comments: The original download provides a zipped folder that contains (1) an ArcGIS File Geodatabase (.gdb) containing one raster file for each of the two global models of tree density – one based on biomes and one based on ecoregions; (2) a layer file (.lyr) for each of the global models with the symbology used for each respective model in Crowther et al (2015); and an ArcGIS Map Document (.mxd) that contains the layers and symbology for each map in the paper. The data is delivered in the Goode homolosine interrupted projected coordinate system that was used to compute biome, ecoregion, and global estimates of the number and density of trees presented in Crowther et al (2015). To obtain maps like those presented in the official publication, raster files will need to be reprojected to the Eckert III projected coordinate system. Details on subsequent revisions and alternative file formats are list below under Additional Versions.----------

Additional Versions: Crowther_Nature_Files_Revision_01.zip contains tree density predictions for small islands that are not included in the data available in the original dataset. These predictions were not taken into consideration in production of maps and figures presented in Crowther et al (2015), with the exception of the values presented in Supplemental Table 2. The file structure follows that of the original data and includes both biome- and ecoregion-level models.

Crowther_Nature_Files_Revision_01_WGS84_GeoTiff.zip contains Revision_01 of the biome-level model, but stored in WGS84 and GeoTiff format. This file was produced by reprojecting the original Goode homolosine files to WGS84 using nearest neighbor resampling in ArcMap. All areal computations presented in the manuscript were computed using the Goode homolosine projection. This means that comparable computations made with projected versions of this WGS84 data are likely to differ (substantially at greater latitudes) as a product of the resampling. Included in this .zip file are the primary .tif and its visualization support files.

References:

Crowther, T. W., Glick, H. B., Covey, K. R., Bettigole, C., Maynard, D. S., Thomas, S. M., Smith, J. R., Hintler, G., Duguid, M. C., Amatulli, G., Tuanmu, M. N., Jetz, W., Salas, C., Stam, C., Piotto, D., Tavani, R., Green, S., Bruce, G., Williams, S. J., Wiser, S. K., Huber, M. O., Hengeveld, G. M., Nabuurs, G. J., Tikhonova, E., Borchardt, P., Li, C. F., Powrie, L. W., Fischer, M., Hemp, A., Homeier, J., Cho, P., Vibrans, A. C., Umunay, P. M., Piao, S. L., Rowe, C. W., Ashton, M. S., Crane, P. R., and Bradford, M. A. 2015. Mapping tree density at a global scale. Nature, 525(7568): 201-205. DOI: http://doi.org/10.1038/nature14967Glick, H. B., Bettigole, C. B., Maynard, D. S., Covey, K. R., Smith, J. R., and Crowther, T. W. 2016. Spatially explicit models of global tree density. Scientific Data, 3(160069), doi:10.1038/sdata.2016.69.

The England species map was funded by DEFRA’s Natural Capital and Ecosystem Assessment (NCEA) programme. The map was created using satellite remote sensing data (Sentinel-2) and machine learning to classify common tree species in England. The model was trained to distinguish 35 common tree species, with minority species grouped into “Other broadleaf” or “Other conifer” classes for better classification performance. The final product comprises a species classification and confidence raster output.

The species map represents a predicted distribution of common tree species in England, produced using a time series of multispectral satellite remote sensing data (Sentinel-2) and machine learning. A classifier based on the XGBoost algorithm was trained to distinguish tree species, utilising a time-series of surface reflectance data and labelled training samples from the sub-compartment database (SCDB). To enhance classification performance, minority species with fewer than 1,000 training samples were grouped into broader categories, resulting in a total of 35 classes. Given the significant class imbalances, a sample weighting strategy was employed to guard against significant underfitting of the minority classes. Model evaluation demonstrated strong classification performance, with an overall accuracy of 89% and balanced class accuracy of 90%. Predictions were made at the pixel level and used to generate a species classification and confidence raster output. Field validation for Norway spruce within the Ips typographus demarcated area, confirmed a precision of 69%, aligning with test data results for this class. Additional validation using National Forest Inventory (NFI) data further reinforced model reliability, though accuracy was observed to be worse for underrepresented species.

The INTERPNT method can be used to produce accurate maps of trees based solely on tree diameter and tree-to-tree distance measurements. For additional details on the technique please see the published paper (Boose, E. R., E. F. Boose and A. L. Lezberg. 1998. A practical method for mapping trees using distance measurements. Ecology 79: 819-827). Additional information is contained in the documentation that accompanies the program. The Abstract from the paper is reproduced below. "Accurate maps of the locations of trees are useful for many ecological studies but are often difficult to obtain with traditional surveying methods because the trees hinder line of sight measurements. An alternative method, inspired by earlier work of F. Rohlf and J. Archie, is presented. This "Interpoint method" is based solely on tree diameter and tree-to-tree distance measurements. A computer performs the necessary triangulation and detects gross errors. The Interpoint method was used to map trees in seven long-term study plots at the Harvard Forest, ranging from 0.25 ha (200 trees) to 0.80 ha (889 trees). The question of accumulation of error was addressed though a computer simulation designed to model field conditions as closely as possible. The simulation showed that the technique is highly accurate and that errors accumulate quite slowly if measurements are made with reasonable care (e.g., average predicted location errors after 1,000 trees and after 10,000 trees were 9 cm and 15 cm, respectively, for measurement errors comparable to field conditions; similar values were obtained in an independent survey of one of the field plots). The technique requires only measuring tapes, a computer, and two or three field personnel. Previous field experience is not required. The Interpoint method is a good choice for mapping trees where a high level of accuracy is desired, especially where expensive surveying equipment and trained personnel are not available."

This web map depicts tree and canopy cover for the Chugach National Forest. These maps were prepared for the Chugach National Forest to provide up-to-date and more complete information about vegetative communities, structure, and patterns across the Forest. The Copper River Delta vegetation structure attributes were added in 2021; the Kenai Peninsula data products were completed in 2017; Cordova was completed in 2021.Nearly 11 million terrestrial acres were mapped through a partnership between the Geospatial Technology and Applications Center (GTAC), Chugach National Forest, the Alaska Regional Office, and other State, Tribal and Federal agencies. The Chugach National Forest and their partners prepared the regional classification system and identified the desired map units (map classes) that characterized the existing vegetation. GTAC served as the technical lead for developing the mapping methodology that produced the final data products. A combination of field and image interpreted reference data were used to inform the map models. Federal, State, and contracted staff collected plot data on the ground, while Ducks Unlimited and GTAC personnel collected reference information from a helicopter. Classification and regression models were used to characterize modeling units (mapping polygons) with the following vegetation attributes: 1) vegetation type; 2) tree canopy cover; 3) tree size; and 4) tall shrub canopy cover. The minimum map feature depicted is 0.25 acres. All map products were designed according to National Forest Service vegetation mapping standards and are stored in Federal databases.

Estimated tree species richness per hectare.
This map can be downloaded in two formats. One is a geoTIFF file (S_mean_raster.tif) containing the fully geo-referenced map of tree species richness worldwide at a 0.025°×0.025° resolution. The other is a comma-separated file (S_mean_grid.csv) with the following attributes: S is local average tree species richness per hectare x, y are centroid coordinates of all 0.025°×0.025° pixels;

The distribution and local abundance of tree species constitute basic information about our forest ecosystems that is relevant to understanding their ecology, diversity, and relationship to people. The US Forest Service conducts a forest inventory across all forest lands in the United States. We developed geospatial models of forest attributes using this sample-based inventory which make this information available for an even wider variety of applications. From these modeled datasets, we created a series of maps for 24 US states in an effort to connect more people to trees, the datasets, and the scientific research behind them. Presenting these maps in an attractive way invites engagement. The sidebar text is presented in accessible scientific language that clearly defines terms, guides readers in interpreting the maps and histograms, and provides source details and links. The resulting maps are inviting, informative, and accessible to a broad range of people of different ages and backgrounds.

Urban Tree Canopy Assessment. This was created using the Urban Tree Canopy Syracuse 2010 (All Layers) file HERE.The data for this map was created using LIDAR and other spatial analysis tools to identify and measure tree canopy in the landscape. This was a collaboration between the US Forest Service Northern Research Station (USFS), the University of Vermont Spatial Laboratory, and SUNY ESF. Because the full map is too large to be viewed in ArcGIS Online, this has been reduced to a vector tile layer to allow it to be viewed online. To download and view the shapefiles and all of the layers, you can download the data HERE and view this in either ArcGIS Pro or QGIS.Data DictionaryDescription source  USDA Forest ServiceList of values  Value 1 Description Tree CanopyValue 2 Description Grass/ShrubValue 3 Description Bare SoilValue 4 Description WaterValue 5 Description BuildingsValue 6 Description Roads/RailroadsValue 7 Description Other PavedField Class Alias Class Data type String Width 20Geometric objects  Feature class name landcover_2010_syracusecity Object type  complex Object count 7ArcGIS Feature Class Properties Feature class name landcover_2010_syracusecity Feature type  Simple Geometry type Polygon Has topology FALSE Feature count 7 Spatial index TRUE Linear referencing  FALSEDistributionAvailable format  Name ShapefileTransfer options  Transfer size 163.805Description Downloadable DataFieldsDetails for object landcover_2010_syracusecityType Feature Class Row count  7 Definition  UTCField FIDAlias FID Data type OID Width  4 Precision 0 Scale 0Field descriptionInternal feature number.Description source ESRIDescription of valueSequential unique whole numbers that are automatically generated.Field ShapeAlias Shape Data type Geometry Width 0 Precision 0 Scale 0Field description Feature geometry.Description source  ESRIDescription of values Coordinates defining the features.Field CodeAlias Code Data type Number Width 4Overview Description  Metadata DetailsMetadata language  English Metadata character set utf8 - 8 bit UCS Transfer FormatScope of the data described by the metadata  dataset Scope name  datasetLast update 2011-06-02ArcGIS metadata properties Metadata format ArcGIS 1.0 Metadata style North American Profile of ISO19115 2003Created in ArcGIS for the item 2011-06-02 16:48:35 Last modified in ArcGIS for the item 2011-06-02 16:44:43Automatic updates Have been performed Yes Last update 2011-06-02 16:44:43Item location history  Item copied or moved 2011-06-02 16:48:35 From T:\TestSites\NY\Syracuse\Temp\landcover_2010_syracusecity To \T7500\F$\Export\LandCover_2010_SyracuseCity\landcover_2010_syracusecity

TreeMap 2016 provides a tree-level model of the forests of the conterminous United States. We matched forest plot data from Forest Inventory and Analysis (FIA) to a 30x30 meter (m) grid. TreeMap 2016 is being used in both the private and public sectors for projects including fuel treatment planning, snag hazard mapping, and estimation of terrestrial carbon resources. We used a random forests machine-learning algorithm to impute the forest plot data to a set of target rasters provided by Landscape Fire and Resource Management Planning Tools (LANDFIRE: https://landfire.gov). Predictor variables consisted of percent forest cover, height, and vegetation type, as well as topography (slope, elevation, and aspect), location (latitude and longitude), biophysical variables (photosynthetically active radiation, precipitation, maximum temperature, minimum temperature, relative humidity, and vapour pressure deficit), and disturbance history (time since disturbance and disturbance type) for the landscape circa 2016.

The main output of this project (the GeoTIFF included in this data publication) is a raster map of imputed plot identifiers at 30×30 m spatial resolution for the conterminous U.S. for landscape conditions circa 2016. In the attribute table of this raster, we also present a set of attributes drawn from the FIA databases, including forest type and live basal area. The raster map of plot identifiers can be linked to the FIA databases available through the FIA DataMart (https://doi.org/10.2737/RDS-2001-FIADB) or to the text and SQL files included in this data publication to produce tree-level maps or to map other plot attributes. The accompanying database files included in this publication also contain attributes regarding the FIA plot CN (or control number, a unique identifier for each time a plot is measured), the subplot number, the tree record number, and for each tree: the status (live or dead), species, diameter, height, actual height (where broken), crown ratio, number of trees per acre, and a code for cause of death where applicable. The dataset has been validated for applications including percent live tree cover, height of the dominant trees, forest type, species of trees with most basal area, aboveground biomass, fuel treatment planning, and snag hazard. Because falling snags cause hazard to firefighting personnel and other forest users, in response to requests from the field, we provide a separate map that provides a rating of the severity of snag hazard based on the density and height of snags. Application of the dataset to research questions other than those for which it has been validated should be investigated by the researcher before proceeding. The dataset may be suitable for other applications and for use across various scales (stand, landscape, and region), however, the researcher should test the dataset's applicability to a particular research question before proceeding.Geospatial data describing tree species or forest structure are required for many analyses and models of forest landscape dynamics. Forest data must have resolution and continuity sufficient to reflect site gradients in mountainous terrain and stand boundaries imposed by historical events, such as wildland fire and timber harvest. The TreeMap 2014 dataset (Riley et al. 2019) was the first of its kind to provide such detailed forest structure data across the forests of the conterminous United States. The TreeMap 2016 dataset updates the TreeMap 2014 dataset to landscape conditions c2016. Prior to this imputed forest data, assessments relied largely on forest inventory at fixed plot locations at sparse densities.See the Entity and Attributes section for details regarding the relationship between the data files included in this publication and the FIA DataMart (https://doi.org/10.2737/RDS-2001-FIADB).

These data were published on 08/26/2021. On 02/01/2024, the metadata was updated to include reference to a recently published article and update URLs for Forest Service websites.

For more information about these data, see Riley et al. (2022).

The files linked to this reference are the geospatial data created as part of the completion of the baseline vegetation inventory project for the NPS park unit. Current format is ArcGIS file geodatabase but older formats may exist as shapefiles. Inc. (AIS) out of Redlands, CA. The mapping effort began in 1996 and by 2004 they had produced a vegetation map (referred to as the 2005 version of the map), along with two reports (see Appendix F and G) titled, Photo-Interpretation Report, USGS-NPS Vegetation and Inventory and Mapping Program, Joshua Tree National Park and USGS-NPS Vegetation Mapping Program, Joshua Tree National Park Mapping Classification. AIS was hired again in 2009-2010 to assist in updating the map; they hosted the meeting in August 2009, then proceeded to make changes to the map as discussed at the meeting. For the most part, this involved revisiting aerial photos and reevaluating the map class assigned to each problematic polygon, as well as correcting any global recodes and minor edits to the nomenclature. Aerial imagery used for the project was from 1998, including the revisits in 2009, and the minimum mapping unit was defined as 0.50 hectares. For more detail on methods used by AIS to produce the map and a summary of the project pre-2005, refer to the reports mentioned above.

The Prince of Wales Existing Vegetation mapping project encompasses over 4.2 million acres of Southeastern Alaska—2.3 million of which are terrestrial. This map was designed to be consistent with the standards established in the Existing Vegetation Classification and Technical Guide (Nelson et al. 2015), and to provide baseline information to support project planning and inform land management of the Prince of Wales and surrounding islands. The final map comprises seven distinct, integrated feature layers: 1) vegetation type; 2) tree canopy cover; 3) trees per acre (TPA) for trees ≥ 1’ tall; 4) trees per acre for trees ≥ 6” diameter at breast height (dbh); 5) quadratic mean diameter (QMD) for trees ≥ 2” dbh; 6) quadratic mean diameter for trees ≥ 9” dbh; and 7) thematic tree size. The dominance type map consists of 18 classes, including 15 vegetation classes and 3 other land cover types. Continuous tree canopy cover, TPA, QMD, and thematic tree size was developed for areas classified as forest on the final vegetation type map layer. Geospatial data, including remotely sensed imagery, topographic data, and climate information, were assembled to classify vegetation and produce the maps. A semi-automated image segmentation process was used to develop the modeling units (mapping polygons), which delineate homogeneous areas of land cover. Field plots containing thematic vegetation type and tree size information were used as reference for random forest prediction models. Important model drivers included 30 cm orthoimagery collected during the height of the 2019 growing season, in addition to Sentinel 2 and Landsat 8 satellite imagery, for vegetation type prediction. Additionally, detailed tree inventory data were collected at precise field locations to develop forest metrics for Quality Level 1 (QL1) Light Detection and Ranging (LiDAR) data. LiDAR information was acquired across approximately 80% of the project’s land area. Continuous tree canopy cover and 2^nd order forest metrics (TPA and QMD) were modeled across the LiDAR coverage area, and subsequently, extrapolated to the full project extent using Interferometric Synthetic Aperture Radar (IfSAR) as the primary topographic data source.

The National Trees Outside Woodland (TOW) V1 map is a vector product funded by DEFRA’s Natural Capital and Ecosystem Assessment (NCEA) programme produced under Forest Research’s Earth Observation for Trees and Woodlands (EOTW) project.

The TOW map identifies canopy cover over 3m tall and 5m2 area which exists outside the National Forest Inventory (National Forest Inventory - Forest Research). Canopy cover is categorised into the following woodland types - lone trees, groups of trees and small woodlands.

The data set was derived from the Vegetation Object Model (VOM) (Environment Agency, EA), the National Lidar Survey (EA), and Sentinel-2 (European Space Agency) imagery using spatial algorithms. The method is fully automated with no manual manipulation or editing. The map and its production method has been quality assured by DEFRA science assurance protocols and assessed for accuracy using ground truth data.

Because the process classifies objects based on proximity to features within OS mapping, there could be some misclassifications of those objects not included in the OS (specifically: static caravans, shipping containers, large tents, marquees, coastal cliffs and solar farms).

This is a first release of this dataset, the quality of the production methods will be reviewed over the next year, and improvements will be made where possible.

The TOW map is available under open government licence and free to download from the Forestry Commission open data download website (Forestry Commission) and view online on the NCEA ArcGIS Online web portal (Trees Outside Woodland). A full report containing details on methodology, accuracy and user guide is available.

TOW map web portal link : ncea.maps.arcgis.com/apps/instant/sidebar/index.html?appid=cf571f455b444e588aa94bbd22021cd3

FR TOW map web page : https://www.forestresearch.gov.uk/tools-and-resources/fthr/trees-outside-woodland-map/

This data publication contains raster maps of live tree basal area for each tree species along with corresponding assessment data. An efficient approach for mapping multiple individual tree species over large spatial domains was used to develop these raster datasets. The method integrates vegetation phenology derived from MODIS imagery and raster data describing relevant environmental parameters with extensive field plot data of tree species basal area to create maps of tree species abundance and distribution at a 250-meter (m) pixel size for the contiguous United States. The approach uses the modeling techniques of k-nearest neighbors and canonical correspondence analysis, where model predictions are calculated using a weighting of nearest neighbors based on proximity in a feature space derived from the model. The approach also utilizes a stratification derived from the 2001 National Land-Cover Database tree canopy cover layer.The mapping methodology and resultant datasets were intended to address three major issues. 1) Land use policy decisions are often made at the landscape scale because landscape processes, like risk of forest pests or fire, occur over large areas. 2) Distribution and abundance information is often needed for individual species as opposed to forest types because individual species can play significant roles in natural systems, may have high economic impact, or may be indicators for ecosystem health. 3) The maintenance of a realistic species covariance structure across a set of maps of individual species is important because species assemblage information is used in coarse scale modeling of ecosystem processes like response to disturbance, urbanization, and climate change.Original metadata date was 09/09/2013. Minor metadata updates on 12/15/2016.

Dominant tree species map of Switzerland We created a tree species map of Switzerland for the dominant tree species in the forested areas. The spatial resolution of the map is 10 m and the coordinate system is ETRS89-extended / LAEA Europe (EPSG 3035). The map comprises Sentinel-2 index time series from the year 2020, a digital elevation model and species reference data from the Swiss National Forest Inventory. The map is available as raster (.tif) or vector dataset (.gpkg). Access will be granted upon request. In total, the following 15 species were mapped: Abies alba, Acer pseudoplatanus, Alnus glutinosa, Alnus incana, Betula pendula, Castanea sativa, Fagus sylvatica, Fraxinus excelsior, Picea abies, Pinus cembra, Pinus mugo arborea, Pinus sylvestris, Quercus petraea, Quercus robur, Sorbus aucuparia. -br/--br/- Approach -br/--br/- Data - Swiss National Forest Inventory Data (stand species with - 60 % dominance in upper canopy; on at least more than 9 plots dominant) - Sentinel-2 time series (2020, Indices: CCI, CIRE, NDMI, EVI, NDVI) - Digital elevation model (DEM) (swissalti3d, 5 m) - Biogeographical regions (Federal Office for the Environment FOEN) - Forest mask 2017 (Approach: Waser et al., 2015) -br/--br/- Modeling approach We identified the most meaningful variables that led to separation of the respective groups by using random forest models with a forward feature selection (Meyer et al., 2018; Ververidis & Kotropoulos, 2005). In this approach, the final random forest model is solely built from the selected meaningful variables. By identifying meaningful variables, we can determine which variables might influence the grouping. Further, to avoid overfitting and overly optimistic results, we applied 10-fold spatial cross-validation and put all pixels from a plot in the same spatial fold. The modeling was realized using the CAST package in R (Meyer et al., 2022), based on the well-known caret package (Kuhn, 2022). We used the ranger package in R (Wright & Ziegler, 2017) to implement the random forest models, due to its short computation time. -br/--br/- Training data for modeling - 295 Sentinel-2, DEM & Biogeographical variables - 10525 tree species pixels -br/--br/- Selected variables for final model 1. EVI of 2020.05.16 2. NDMI of 2020.03.12 3. CIRE of 2020.04.16 4. NDMI of 2020.07.05 5. CCI of 2020.05.11 6. dem 7. CCI of 2020.08.14 8. NDMI of 2020.08.24 9. CCI of 2020.12.22 10. NDMI of 2020.04.21 11. NDMI of 2020.11.17 12. NDMI of 2020.08.09 13. CIRE of 2020.03.22 14. CIRE of 2020.08.09 14. CCI of 2020.11.02 15. CIRE of 2020.06.10 -br/--br/- Overall Accuracy of final model - 0.759 -br/--br/- Nationwide prediction - Predicted throughout forest mask 2017 (Approach: Waser et al., 2015) - Not applied on incomplete Sentinel-2 time series (own category in final map: incomplete_ts) - Applied the Area of Applicability (Meyer 2022) to sort out pixels outside of the feature space; basically where the model had not the same values for pixels as in the available training data -br/--br/- -br/--br/- Be aware that the map is only validated with the training data itself, an independent validation with other data sources remains missing -br/--br/- -br/--br/- References - Kuhn, M. (2022). Classification and Regression Training. 6.0-93. - Meyer, H., Reudenbach, C., Hengl, T., Katurji, M., & Nauss, T. (2018). Improving performance of spatio-temporal machine learning models using forward feature selection and target-oriented validation. Environmental Modelling and Software, 101, 1-9. https://doi.org/10.1016/j.envsoft.2017.12.001 - Meyer, H., Milà, C., & Ludwig, M. (2022). CAST: 'caret' Applications for Spatial-Temporal Models. 0.7.0. - Ververidis, D., & Kotropoulos, C. (2005). Sequential forward feature selection with low computational cost. 2005 13th European Signal Processing Conference. - Waser, L., Fischer, C.,Wang, Z., & Ginzler, C. (2015). Wall-to-Wall Forest Mapping Based on Digital Surface Models from Image-Based Point Clouds and a NFI Forest Definition. Forests, 6, 12, 4510–4528. - Wright, M. N., & Ziegler, A. (2017). ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R. Journal of Statistical Software, 77(1), 1-17. https://doi.org/doi:10.18637/jss.v077.i01

This reference contains the imagery data used in the completion of the baseline vegetation inventory project for the NPS park unit. Orthophotos, raw imagery, and scanned aerial photos are common files held here. The vegetation mapping portion of the project was contracted to Aerial Information Systems, Inc. (AIS) out of Redlands, CA. The mapping effort began in 1996 and by 2004 they had produced a vegetation map (referred to as the 2005 version of the map), along with two reports (see Appendix F and G) titled, Photo-Interpretation Report, USGS-NPS Vegetation and Inventory and Mapping Program, Joshua Tree National Park and USGS-NPS Vegetation Mapping Program, Joshua Tree National Park Mapping Classification. AIS was hired again in 2009-2010 to assist in updating the map; they hosted the meeting in August 2009, then proceeded to make changes to the map as discussed at the meeting. For the most part, this involved revisiting aerial photos and reevaluating the map class assigned to each problematic polygon, as well as correcting any global recodes and minor edits to the nomenclature. Aerial imagery used for the project was from 1998, including the revisits in 2009, and the minimum mapping unit was defined as 0.50 hectares. For more detail on methods used by AIS to produce the map and a summary of the project pre-2005, refer to the reports mentioned above.

City of Austin Open Data Terms of Use https://data.austintexas.gov/stories/s/ranj-cccq This dataset was created to depict approximate tree canopy cover for all land within the City of Austin's "full watershed regulation area." Intended for planning purposes and measuring citywide percent canopy. Definition: Tree canopy is defined as the layer of leaves, branches, and stems of trees that cover the ground when viewed from above. Methods: The 2022 tree canopy layer was derived from satellite imagery (Maxar) and aerial imagery (NAIP). Images were used to extract tree canopy into GIS vector features. First, a “visual recognition engine” generated the vector features. The engine used machine learning algorithms to detect and label image pixels as tree canopy. Then using prior knowledge of feature geometries, more modeling algorithms were used to predict and transform probability maps of labeled pixels into finished vector polygons depicting tree canopy. The resulting features were reviewed and edited through manual interpretation by GIS professionals. When appropriate, NAIP 2022 aerial imagery supplemented satellite images that had cloud cover, and a manual editing process made sure tree canopy represented 2022 conditions. Finally, an independent accuracy assessment was performed by the City of Austin and the Texas A&M Forest Service for quality assurance. GIS professionals assessed agreement between the tree canopy data and its source satellite imagery. An overall accuracy of 98% was found. Only 23 errors were found out of a total 1,000 locations reviewed. These were mostly omission errors (e.g. not including canopy in this dataset when canopy is shown in the satellite or aerial image). Best efforts were made to ensure ground-truth locations contained a tree on the ground. To ensure this, location data were used from City of Austin and Texas A&M Forest Service databases. Analysis: The City of Austin measures tree canopy using the calculation: acres of tree canopy divided by acres of land. The area of interest for the land acres is evaluated at the City of Austin's jurisdiction including Full Purpose, Limited Purpose, and Extraterritorial jurisdictions as of May 2023. New data show, in 2022, tree canopy covered 41% of the total land area within Austin's city limits (using city limit boundaries May 2023 and included in the download as layer name "city_of_austin_2023"). 160,046.50 canopy acres (2022) / 395,037.53 land acres = 40.51% ~41%. This compares to 36% last measured in 2018, and a historical average that’s also hovered around 36%. The time period between 2018 and 2022 saw a 5 percentage point change resulting in over 19K acres of canopy gained (estimated). Data Disclaimer: It's possible changes in percent canopy over the years is due to annexation and improved data methods (e.g. higher resolution imagery, AI, software used, etc.) in addition to actual in changes in tree canopy cover on the ground. For planning purposes only. Dataset does not account for individual trees, tree species nor any metric for tree canopy height. Tree canopy data is provided in vector GIS format housed in a Geodatabase. Download and unzip the folder to get started. Please note, errors may exist in this dataset due to the variation in species composition and land use found across the study area. This product is for informational purposes and may not have been prepared for or be suitable for legal, engineering, or surveying purposes. It does not represent an on-the-ground survey and represents only the approximate relative location of property boundaries. This product has been produced by the City of Austin for the sole purpose of geographic reference. No warranty is made by the City of Austin regarding specific accuracy or completeness. Data Provider: Ecopia AI Tech Corporation and PlanIT Geo, Inc. Data derived from Maxar Technologies, Inc. and USDA NAIP imagery

This database contains distribution maps of 135 eastern U.S. tree species based on Importance Values (IV) derived from Forest Inventory Analysis (FIA) data and a geographical information system (GIS) database of Elbert L. Jr. Little's published ranges. Between 1971 and 1977, Elbert L. Jr. Little, Chief Dendrologist with the U.S. Forest Service, published a series of maps of tree species ranges based on botanical lists, forest surveys, field notes ad herbarium specimens. These published maps have become the standard reference for most U.S. and Canadian tree species ranges.

The USDA Forest Service's FIA units are charged with periodically assessing the extent, timber potential, and health of the trees in the United States. The investigators have created a spatial database of individual species IV based on the number of stems and basal area of understory and overstory trees using FIA data from more than 100,000 plots in the eastern United States. The IV ranges from 0 to 100 and gives a measure of the abundance of the species. (See the investigator's Climate Change Atlas for 80 Forest Tree Species of the Eastern United States at [http://www.fs.fed.us/ne/delaware/atlas/web_atlas.html] for details). The investigators have aggregated the plot-level IV to 20km cells.

Both sets of maps (Little's ranges and IV based on FIA data) are available for download. The Little's range maps (little.shp) are vector based and are provided as shape files. These maps can span United States or United States and Canada in extent depending on the species. The Importance Value (IV) are raster maps (asciigrid) in asciigrid format. This is an ascii file with header information that can be used to import data into ArcInfo GRID or ArcView's Spatial Analyst GIS software. The spatial resolution is 20km. These raster maps span the eastern U.S. (east of the 100th meridian) in extent.

This EnviroAtlas dataset addresses the tree buffer along heavily traveled roads. The roads are interstates, arterials, and collectors within the EnviroAtlas community boundary. In this community, tree cover is defined as Trees & Forest, Orchards, and Woody Wetlands. Sufficient tree bufferage is defined as 25% coverage within the circular moving window with a radius of 14.5m at any given point along the roadway. There are potential negative health effects for those living in a location without a sufficient tree buffer. This dataset was produced by the US EPA to support research and online mapping activities related to EnviroAtlas. EnviroAtlas (https://www.epa.gov/enviroatlas) allows the user to interact with a web-based, easy-to-use, mapping application to view and analyze multiple ecosystem services for the contiguous United States. The dataset is available as downloadable data (https://edg.epa.gov/data/Public/ORD/EnviroAtlas) or as an EnviroAtlas map service. Additional descriptive information about each attribute in this dataset can be found in its associated EnviroAtlas Fact Sheet (https://www.epa.gov/enviroatlas/enviroatlas-fact-sheets).

The Prince of Wales Existing Vegetation mapping project encompasses over 4.2 million acres of Southeastern Alaska—2.3 million of which are terrestrial. This map was designed to be consistent with the standards established in the Existing Vegetation Classification and Technical Guide (Nelson et al. 2015), and to provide baseline information to support project planning and inform land management of the Prince of Wales and surrounding islands. The final map comprises seven distinct, integrated feature layers: 1) vegetation type; 2) tree canopy cover; 3) trees per acre (TPA) for trees ≥ 1’ tall; 4) trees per acre for trees ≥ 6” diameter at breast height (dbh); 5) quadratic mean diameter (QMD) for trees ≥ 2” dbh; 6) quadratic mean diameter for trees ≥ 9” dbh; and 7) thematic tree size. The dominance type map consists of 18 classes, including 15 vegetation classes and 3 other land cover types. Continuous tree canopy cover, TPA, QMD, and thematic tree size was developed for areas classified as forest on the final vegetation type map layer. Geospatial data, including remotely sensed imagery, topographic data, and climate information, were assembled to classify vegetation and produce the maps. A semi-automated image segmentation process was used to develop the modeling units (mapping polygons), which delineate homogeneous areas of land cover. Field plots containing thematic vegetation type and tree size information were used as reference for random forest prediction models. Important model drivers included 30 cm orthoimagery collected during the height of the 2019 growing season, in addition to Sentinel 2 and Landsat 8 satellite imagery, for vegetation type prediction. Additionally, detailed tree inventory data were collected at precise field locations to develop forest metrics for Quality Level 1 (QL1) Light Detection and Ranging (LiDAR) data. LiDAR information was acquired across approximately 80% of the project’s land area. Continuous tree canopy cover and 2^nd order forest metrics (TPA and QMD) were modeled across the LiDAR coverage area, and subsequently, extrapolated to the full project extent using Interferometric Synthetic Aperture Radar (IfSAR) as the primary topographic data source.

Data description

This file is a 100 m resolution map of % tree canopy cover per pixel for Africa in 2019. Tree cover values range from 0-100 % and nodata areas are masked as -1.

The map is derived from predictions of tree cover using 3 m PlanetScope imagery, which were resampled to 1 m, and then aggregated to 100 m. Predictions at 1 m resolution (or anything between 1 and 100 m) are available on request.

The image data was processed from raw scenes into merged 1x1 degree mosaics. Differences in the quality of raw scenes and processed mosaics can lead to artifacts in the predicted tree cover, visible as seamlines along scene or mosaic edges. Therefore the current version of this map should not be seen as a final product representing the exact tree cover at each location, but as a proof of concept of what is possible with PlanetScope data.

The main benefit of this map is the inclusion of cover from scattered single trees in sparse cover areas such as savannahs. Here it can serve as a complement to existing forest cover maps to extend the mapping to include non-forest trees.

License

This tree cover map is made freely available for non-commercial purposes. All usage of the data must be attributed and should be cited with the paper citation.

Please see the NICFI license for full terms of usage, available at: https://assets.planet.com/docs/Planet_ParticipantLicenseAgreement_NICFI.pdf

Version differences

We recommend to always use the latest version.

We are continuously improving the quality and consistency of the tree cover mapped, and are expecting to release regular version updates, based on an improved model and mosaics. We also plan to release different years in the future.

Version 0.1: ps_africa_treecover_2019_100m_v0.1

This version was used for the analyses in the paper. It used the first version of the mosaics, which suffered from the inclusion of lower-quality scenes and missing data.

Known issues

- missing data coverage: incomplete mosaics and mosaics missing scenes

- inconsistent predictions between mosaics: visible mosaic edges, see Zimbabwe

- underprediction in very sparse cover areas

- overprediction of cover in denser shrublands

- overprediction (artifacts) in mountains, desert dunes, croplands and some urban areas

- flowering trees in closed forests are mapped as gaps

- occasional confusion between understory or shrubs and trees in wood- and shrublands

Version 1.0: ps_africa_treecover_2019_100m_v1.0

This is the latest version available as of March 2023. It is based on improved mosaics and a revised model, leading to better consistency both between and within mosaics.

However the current model underpredicts solid tree cover in shrublands and woodlands, leading to an overall underprediction in lower rainfall areas.

Improvements over v0.1:

- improved mosaic quality

- reduced artifacts in croplands and urban areas (though some remain)

- better consistency of forest area mapping

- better detection rate of small trees in dry areas

Known issues

- underprediction of tree clusters in shrublands

- some (few) false predictions of small trees in dry areas

- occasional inconsistent predictions within mosaics: seamlines along scene edges

- areas of underprediction in dense tropical forest due to lower quality scenes, see DRC

- overprediction (artifacts) in mountains, and occasionally desert dunes

- flowering trees in closed forests are mapped as gaps

- occasional confusion between understory or shrubs and trees in wood- and shrublands

- trees without leaves may not be mapped correctly

The Young Trees map was funded by DEFRA through the Natural Capital and Ecosystem Assessment (NCEA) programme. The young trees mapping project developed a machine learning methodology using remote sensing to identify restocked stands where saplings persist in healthy numbers. The approach uses an eight-year timeframe since planting, crucial for verifying government grant compliance. Automating this methodology ensures easy replication and model transferability across years by training on multi-year data, making it resilient to climatic variations. Validation has confirmed the model’s accuracy, recommending high-confidence thresholds for restock classification. In the future, integration with the National Forest Inventory will enhance woodland mapping, accelerating updates and improving national indicators for forest extent and connectivity.

The aim of the young trees mapping project was to develop a machine learning methodology using remote sensing data, to identify stands where trees have been planted and saplings persist in healthy numbers. This was conducted within restock contexts across a specific timeframe, currently eight years since planting. This timeframe is significant because funding provided by government grants for planting can be reclaimed if it can be demonstrated that the funding has not been utilised by the landowner. Furthermore, the restock status of clearfell polygons has the potential to improve the accuracy of extent and connectivity environmental indicators developed as part of the Tree Health Resilience Strategy (THRS). The aim of this part of the project was to automate the methodology in such a way that it can be easily replicated, and to make the model transferable across years. Specifically, to train the model using multiple years of data, which makes the model agnostic to variable annual climactic conditions. The model is both robust and accurate, as demonstrated by the validation. It is recommended that only polygons with over 95% and under 5% confidence are treated as restocked or not restocked with any certainty. Outside of these limits confidence scores are only indicative of the restock status. In the future, the model is likely to be implemented as part of the National Forest Inventory (NFI) woodland map creation procedure. This will result in accelerated turnover of polygon labels from clearfell to young trees, where appropriate and will provide an important improvement to a national indicator for woodland extent and connectivity.