Plant species cover-abundance and presence observed in multi-scale plots. Plant species and associated percent cover in 1m2 subplots and plant species presence in 10m2 and 100m2 subplots are reported from 400m2 plots. Archived plant vouchers and foliar tissue support the data and additional analyses.

Abstract

The NeonTreeCrowns dataset is a set of individual level crown estimates for 100 million trees at 37 geographic sites across the United States surveyed by the National Ecological Observation Network’s Airborne Observation Platform. Each rectangular bounding box crown prediction includes height, crown area, and spatial location.

How can I see the data?

A web server to look through predictions is available through idtrees.org

Dataset Organization

The shapefiles.zip contains 11,000 shapefiles, each corresponding to a 1km^2 RGB tile from NEON (ID: DP3.30010.001). For example "2019_SOAP_4_302000_4100000_image.shp" are the predictions from "2019_SOAP_4_302000_4100000_image.tif" available from the NEON data portal: https://data.neonscience.org/data-products/explore?search=camera. NEON's file convention refers to the year of data collection (2019), the four letter site code (SOAP), the sampling event (4), and the utm coordinate of the top left corner (302000_4100000). For NEON site abbreviations and utm zones see https://www.neonscience.org/field-sites/field-sites-map.

The predictions are also available as a single csv for each file. All available tiles for that site and year are combined into one large site. These data are not projected, but contain the utm coordinates for each bounding box (left, bottom, right, top). For both file types the following fields are available:

Height: The crown height measured in meters. Crown height is defined as the 99th quartile of all canopy height pixels from a LiDAR height model (ID: DP3.30015.001)

Area: The crown area in m² of the rectangular bounding box.

Label: All data in this release are "Tree".

Score: The confidence score from the DeepForest deep learning algorithm. The score ranges from 0 (low confidence) to 1 (high confidence)

How were predictions made?

The DeepForest algorithm is available as a python package: https://deepforest.readthedocs.io/. Predictions were overlaid on the LiDAR-derived canopy height model. Predictions with heights less than 3m were removed.

How were predictions validated?

Please see

Weinstein, B. G., Marconi, S., Bohlman, S. A., Zare, A., & White, E. P. (2020). Cross-site learning in deep learning RGB tree crown detection. Ecological Informatics, 56, 101061.

Weinstein, B., Marconi, S., Aubry-Kientz, M., Vincent, G., Senyondo, H., & White, E. (2020). DeepForest: A Python package for RGB deep learning tree crown delineation. bioRxiv.

Weinstein, Ben G., et al. "Individual tree-crown detection in RGB imagery using semi-supervised deep learning neural networks." Remote Sensing 11.11 (2019): 1309.

Were any sites removed?

Several sites were removed due to poor NEON data quality. GRSM and PUUM both had lower quality RGB data that made them unsuitable for prediction. NEON surveys are updated annually and we expect future flights to correct these errors. We removed the GUIL puerto rico site due to its very steep topography and poor sunangle during data collection. The DeepForest algorithm responded poorly to predicting crowns in intensely shaded areas where there was very little sun penetration. We are happy to make these data are available upon request.

# Contact

We welcome questions, ideas and general inquiries. The data can be used for many applications and we look forward to hearing from you. Contact ben.weinstein@weecology.org.

Bare earth elevation surface (DTM) and actual surface (DSM) given in meters in the NAVD88 (Geoid12A realization) vertical reference frame. Horizontal coordinates referenced to appropriate UTM zone. Bare earth is created by classifying and removing vegetation and man-made structures from lidar point cloud prior to surface generation. Both the DSM and DTM are mosaicked onto a spatially uniform grid at 1 m spatial resolution in 1 km by 1 km tiles provided in a geotiff format.

These shapefiles display the watershed boundaries (area), drainage lines, and pour points for NEON Aquatic Field Sites. For most of the sites, NEON's 1 meter Elevation-LiDAR Digital Terrain Model (DTM) was used to derive the watersheds. In cases where NEON data did not provide complete watershed coverage, a 1/3 arc-second (10 meter) resolution Digital Elevation Model (DEM) raster, available from the U.S. Geological Survey (USGS) website, was utilized to provide full coverage of the watershed extent. The watershed boundary defines the perimeter of drainage areas formed by the terrain and other landscape characteristics. The pour point was selected from nearest the downstream most sensor set, primarily NEON’s S2 sensor in wadeable streams, and S1 sensor or water gauge in non-wadeable rivers, and the outlet sensor in lakes.Included in the attribute information are landcover percentages within the watershed boundary from the 2016 (CONUS and AK) and 2001 (PR) National Land Cover Database (NLCD), and soil classification percentages from the Soil Survey Geographic Database (SSURGO) and the State Soil Geographic Database (STATSGO). Soil data were not available or unclassified at the same resolution for every site and, where applicable, are marked as "NoSoilData". A stand-alone csv of NLCD area is also included. Details of the Watershed Delineation for NEON Aquatic Sites can be found in NEON.DOC.005246.

Digital models of the surface (DSM) and terrain (DTM) derived from NEON LiDAR data. DSM: Surface features (topographic information with vegetation and man-made structures present). DTM: Bare earth elevation (topographic information with vegetation and man-made structures removed). Images are given in meters above mean sea level and mosaicked onto a …

This collection contains unsorted bulk benthic macroinvertebrate samples collected in the field (NEON sample class: inv_fielddata_in.geneticSampleID). Benthic macroinvertebrate samples are collected three times per year at wadeable stream, river, and lake sites during aquatic biology bout windows, roughly in spring, summer, and fall. Samples are collected using the most appropriate sampler for the habitat type, including Surber, Hess, hand corer, modified kicknet, D-frame sweep, and petite ponar samplers. Macroinvertebrates are sampled using a percent-based macrohabitat approach. In wadeable streams, samples are collected in the two most dominant habitat types (e.g. riffles, runs, pools, step pools) within each 1 km-long wadeable stream site. In lakes, samples are collected near the buoy, inlet, and outlet sensors using a petite ponar, and in littoral areas using a D-frame sweep. In rivers, samples are collected near the buoy and two other deep-water locations using a petite ponar sampler, and in littoral areas using a D-frame sweep or large-woody debris sampler. Field protocols differ depending on the habitat and substrate being sampled, but all samples are collected from the surface of the natural substratum in each habitat. Samples are preserved in ethanol in the field, and returned to the domain support facility for a preservative change prior to shipping. The DNA samples from bout 2 (summer) are sent to an external laboratory for metabarcoding. The unsorted bulk DNA samples from bout 1 (spring) and bout 3 (autumn) are archived for future research. Unsorted bulk macroinvertebrate samples are preserved in 70-95% ethanol and archived in a temperature (17-18°C) and humidity controlled environment. See related links below for protocols and NEON related data products.

This collection contains vials of identified mosquitoes stored in bulk (not pathogen tested) from CO2 trapping at terrestrial sites (NEON sample class: mos_archivepooling_in.archiveVialIDList). When adult mosquitoes are active, sampling occurs (via CDC light traps) every two weeks at core sites and every four weeks at gradient sites. A sampling bout consists of one trapping night and the following day for up to ten plots per site. Following collection, samples are sent to a professional taxonomist where a subsample of each catch generated from each trap is identified to species and sex. Starting with samples collected in field year 2025, bloodfed mosquitos are archived seperately, but samples collected prior to 2025 may contain bloodfed mosquitos. Vials containing identified mosquitoes are archived in 2, 5, 10, or 15 mL cryovials in liquid nitrogen, or at -80 for samples collected before 2020. See related links below for protocols and NEON related data products.

The National Ecological Observatory Network (NEON) is a continental-scale ecological observation facility sponsored by the National Science Foundation to gather and synthesize data on the impacts of climate change, land use change and invasive species on natural resources and biodiversity. The mission of NEON is to enable understanding and forecasting by providing infrastructure and consistent methodologies to support research and education in these areas.

Structure measurements of individual woody and non-woody plants, mapped positions of qualifying woody and non-woody plants, and metadata required to draw inference from individual measurements at the plot scale.

This collection contains subsamples of Carabid adults from pitfall sampling from a single trap (NEON sample class: bet_sorting_in.subsampleID.bet). Ground beetles are sampled using pitfall traps (16 oz deli containers filled with 150 or 250 mL of propylene glycol). Four (pre-2018) or 3 (2018 and beyond) traps are deployed in each of 10 (pre-2023) or 6 (2023 and beyond) plots at each terrestrial NEON site, with traps arrayed approximately 20 meters from the center of the plot. Sampling occurs biweekly throughout the growing season (when temperatures are above 4 degrees C). Following trap collection, all beetles from the family Carabidae are sorted by NEON technicians and identified to species or morphospecies. A subset of collected Carabidae are pointed or pinned, while other specimens (non-pinned/non-pointed carabids, invertebrate bycatch, and vertebrate bycatch) are stored in 95% ethanol for archiving. Regardless of storage method, all collections data are reported at a per trap resolution. A subset of pinned ground beetles (up to 467 per site per year) are sent to an expert taxonomist for secondary identification. Identifications performed on these individuals may be used to estimate uncertainty in parataxonomist identification by NEON technicians. See related links below for protocols and NEON related data products.

NEON operates 81 field sites strategically located across 20 eco-climatic domains across the United States, including 47 terrestrial field sites and 34 freshwater aquatic field sites. When logistically possible, aquatic and terrestrial sites are colocated (i.e. in close proximity) to support understanding of linkages across terrestrial and aquatic ecosystems and their interactions with the atmosphere. Core sites The spatial design of the NEON program includes one core terrestrial and one core aquatic site in each of the 20 ecoclimate Domains (with the exception of D20, in Hawaii, which only has a core terrestrial site). These core sites were selected to represent wildlands or more pristine ecosystems with relatively limited human influence within each Domain. As such, many of these sites are in conservation areas or national parks. Gradient sites The gradient sites were selected to provide contrasts with the core sites and enable exploration of scientific questions of cause and effect specific to each Domain. A "gradient," in this case, is a range in a driving variable of ecological change that can be measured across a Domain, such as a gradient of nitrogen and dust deposition, permafrost, invasive species, precipitation, or land use. The gradient sites allow us to evaluate how these differences impact ecosystems within a Domain by comparison with the core wildland site.Aquatic instrument and observation systems are virtually identical between core sites and gradient sites. There are some differences in terrestrial instrumentation between sites (e.g., primary precipitation using a Double Fence Intercomparison Reference (DFIR), shortwave radiation, water vapor isotopes, and sun photometers), which are documented in each of the Data Product Page Descriptions on the NEON Data Portal. More site-specific details are included in the Sensor Position files associated with the data download expanded package. Some terrestrial observational sampling designs (e.g., mammal and mosquito sampling) differ between the core and gradient sites, which are documented within the associated data products' Science Designs and Protocols and Procedures, located on the data product landing pages of the NEON Data Portal.

Input data, evaluation data, and results from integrating National Ecological Observatory Network (NEON) measurements into single point simulations using the Community Terrestrial Systems Model (CTSM). The input data atmosphere data (datm) includes NEON meteorological measurements that provide boundary conditions for CTSM simulations. The evaluation data (eval) includes NEON eddy covariance flux data measured at each tower site, which includes energy, water vapor, and CO2 fluxes that at are time regularized with quality assurance and control flags applied. Finally, CTSM model results (ctsm) include monthly and daily history files of select variables simulated by the model at each NEON site. Additional miscellaneous data products (misc) with site information is also provided. These v2 data were created because we found errors in the values for latitude and longitude of NEON sites tower sites that were being used in CTSM simulations at the following four site: ONAQ, BLAN, ORNL, UNDE. These errors were corrected and sites rerun, with corrected data published here.

This includes teaching data subsets that contains spatio-temporal data for the National Ecological Observatory Network's Harvard Forest and San Joaquin Experimental Range (SJER) field sites.The tutorials using these data subset can be found on the NEON Data Skills portal . CONTENTS:NEON-DS-Met-Time-Series.zip: Precipitation, temperature and other variables collected from a flux tower at the NEON Harvard Forest site. NOTE: these data are not collected from the NEON tower.NEON-DS-Site-Layout-Files.zip: A set of shapefiles for the NEON's Harvard Forest field site and US and (some) state boundary layers.NEON-DS-Landsat-NDVI.zip: 2011 NDVI data product provided by USGS cropped to NEON's Harvard Forest and San Joaquin Experimental Range field sites.NEON-DS-Airborne-RemoteSensing.zip: LiDAR data including a canopy height model, digital elevation model and digital surface model for NEON's Harvard Forest and San Joaquin Experimental Range field sites.

This dataset contains an inventory for the paper entitled "ecocomDP: A flexible data design pattern for ecological community survey data" (O'Brien et al), submitted to Ecological Informatics. The paper describes an approach for harmonizing and reformatting community survey data such as organism abundance or cover measurements. Data currently using this data model and workflow approach are from the repository of the Environmental Data Initiative (EDI), the Long Term Ecological Research (LTER) Network, and the National Ecological Observatory Network (NEON). Data were assembled for this analysis in late 2020. The inventory is composed of two tables, describing data from EDI (including LTER) and data from NEON. The EDI inventory includes information for 70 datasets: identifiers for both the original and converted datasets, and basic coverage information such as temporal coverage (range of years and a measurement of sampling evenness), spatial coverage (maximum bounding coordinates and area of the "bounding box"), and taxonomic coverage (taxonomic classes). The NEON inventory contains information from 11 continent-wide NEON data products, divided into individual field sites to be more spatially compatible with EDI and LTER data. Taxonomic coverage is by group (e.g., algae, birds) rather than explicit taxonomic classes. Spatial coverage is the area of a field sampling site polygon. Temporal coverage includes the same minimum and maximum sampling years and temporal evenness measures as for the EDI data plus a count of months during that period when sampling occurred. At the time of data download, NEON data was considered provisional, however identifiers are persistent and now deliver final, "released" data.

 Also included in the data package is a script to reformat inventory data and create Figure 3 of the paper.

This collection contains fish vouchers collected from wadeable streams and lakes (NEON sample class: fsh_perFish_in.voucherSampleID). Fish are sampled twice per year at lakes and wadeable stream sites, during spring and fall. Ten sampling reaches or segments are established at each site; with 3 fixed reaches sampled during every sampling bout and a random subset of 3 additional reaches or segments selected for sampling each year. Fish are sampled using a combination of electrofishing, gill-nets and mini-fyke nets. Field technicians identify fish to the lowest practical taxonomic level and then weigh and measure a subset of captured individuals before releasing. Only mortalities and individuals that require euthanasia due to injuries are vouchered. Vouchers that are in good condition are fixed in 10% buffered formalin preservative and then transferred to 70% ethanol or isopropanol for long-term storage. Vouchers that are in poor condition are not fixed but preserved in 95% ethanol. Specimens are archived in a temperature (17-18°C) and humidity controlled environment. See related links below for protocols and NEON related data products.

The points in this file have been remotely and field validated by NEON specialists and have been determined to meet NEON criteria. To download published data collected as these locations visit https://data.neonscience.org/home. The sampling locations were created by stratifying vegetation types from the National Land Cover Database (NLCD). Points were distributed in each vegetation type using a spatially balanced system implemented in GIS called the Reversed Randomized Quadrant Recursive Raster (RRQRR) technique. Subsequently, the points went through a remote sensing and ground-truth procedures to validate the vegetation type and NEON criteria. For more information please contact the Permitting Department at NEON, contact information is available at www.neonscience.org. This layer is current as of July 5, 2025.

This collection contains small mammal vouchers collected during small mammal sampling (NEON sample classes: mam_pertrapnight_in.voucherSampleID). Small mammal sampling is based on the lunar calendar, with timing of sampling constrained to occur within 10 days before or after the new moon. Typically, core sites are sampled 6 times per year, and gradient sites 4 times per year. Small mammals are sampled using box traps (models LFA, XLK, H.B. Sherman Traps, Inc., Tallahassee, FL, USA). Box traps are arrayed in three to eight (depending on the size of the site) 10 x 10 grids with 10m spacing between traps at all sites. Small mammal trapping bouts are comprised of one or three nights of trapping, depending on whether a grid is designated for pathogen sample collection (3 nights) or not (1 night). Only mortalities and individuals that require euthanasia due to injuries are vouchered. The NEON Biorepository receives whole frozen specimens and prepares vouchers as either study skins with skulls (or full skeletons) or in 70-95% ethanol. Standard mammalian measurements are taken during specimen preparation (in mm; total length, tail length, hind foot length, ear length; and in g: mass) and are accessible in downloaded records (note: field measurements are listed in parentheses after preparation measurements, when available). Additional notes about parasites and reproductive condition are also accessible in downloaded records. See related links below for protocols and NEON related data products.

This collection contains genetic extracts from homogenized zooplankton samples collected during zooplankton sampling in lakes (NEON sample class: zoo_dnaExtraction_in.dnaSampleID). Zooplankton DNA metabarcoding samples are collected and sequenced once per year in mid-summer. Samples are collected using either a tow net (water deeper than 4 m) or a Schindler-Patalas trap (water shallower than 4 m) depending on the depth at the sampling location. Samples are collected near the NEON profiling buoy as well as the littoral sensor sets. Zooplankton are preserved in ethanol in the field. The samples are shipped to an external facility for homogenization and high-throughput sequencing (metabarcoding). Genetic extracts in 2 mL cryovials are archived in the NEON Biorepository at -80 degrees Celsius. See related links below for protocols and NEON related data products. The metabarcoding analysis protocols used by external facilities are available in the NEON document library (https://data.neonscience.org/documents).

This file is an example dataset from the National Ecological Observatory Network (NEON) Aerial Observation Platform (AOP) high resolution RGB camera with a nominal resolution of 10cm. This image is from 2017 near the Smithsonian Ecological Research Center (SERC) site in Maryland, USA and has the filename "2017_SERC_2_368000_4306000_image.tif"Citation Information: National Ecological Observatory Network. 2021. Data Product DP3.30010.001, High-resolution orthorectified camera imagery mosaic. Provisional data downloaded from http://data.neonscience.org on April 6, 2021. Battelle, Boulder, CO, USA NEON. 2021.

Height of the top of canopy above bare earth (Canopy Height Model (CHM)). Data are mosaicked over the AOP footprint; mosaicked onto a spatially uniform grid at 1 m spatial resolution, and delivered as 1 km by 1 km tiles. Data are provided in GeoTIFF (.tif) format. Associated metadata files include QA reports (.pdf, .md, .html), shapefile boundaries (.shp, .shx, .prj, .dbf), and .kml boundary files; shapefile and kml boundary files may be zipped (.zip).