# GeoLifeCLEF 2023 - LifeCLEF 2023 x FGVC10 — Data

来源：https://www.kaggle.com/competitions/geolifeclef-2023-lifeclef-2023-x-fgvc10/data

采集时间：2026-10-24T13:46:48Z

The training data is composed of the
species observations
(output) and
environmental data
(input). We explain below their precise structure and data sources.
Links
[Seafile repository](https://lab.plantnet.org/seafile/d/936fe4298a5a4f4c8dbd/)
: repository containing all the data. See the
data downloading
section below where we provide recommendations to optimize download time.
[Github GLC](https://github.com/plantnet/GLC)
: Useful codes to manipulate data with simple dataloaders, examples and sample data. More dataloaders can be added after the challenge starts.
[GBIF.org (08 November 2022) GBIF Occurrence Download](https://doi.org/10.15468/dl.4ysfh4)
: Occurrence data provided by Global Biodiversity Information Facility (GBIF). DOI:
[https://doi.org/10.15468/dl.4ysfh4](https://doi.org/10.15468/dl.4ysfh4)
Observations data
The
training species observation data
is composed on one hand of more than 5 millions of plant species
presence-only occurrences
(species name, geolocation, time, dataset, observer, etc. ) collected since 2017 in Europe (38 countries), and on the other hand of around 5 thousands presence-absence surveys (inventory of all species in a small plot of 10 to 400m²). This data covers overall around 10,000 species, namely most of the European flora. This data should be used to inform the model outputs keeping in mind that the presence-only data (PO) only partially represent the local species composition and it is subject to various sampling biases. The local absence of a species among PO data doesn't mean it is truly absent: An observer might not have reported it because it was difficult to detect at this time of the year, difficult to identify, not a monitoring target or just not attractive. Indeed, most of the PO data come from citizen science programms without sampling protocol. The presence-absence data (PA) is provided to compensate the problem of false-absences of PO data and calibrate models in order to avoid associated biases.
Species observation details
. In more details, there are two occurrence CSV for the training species observation data, one containing the PO data (/Presence_only_occurrences/Presences_only_train.csv) and the other the PA data (/Presence_Absence_surveys/Presences_Absences_train.csv). Both CSV include notably the columns speciesID (species are anonymised), lon (longitude WGS84), lat (latitude), geoUncertaintyInM, year, dayOfYear (1 to 366), glcID (observation id), gbifID (for gbif data only, matches with www.gbif.org), patchID (see below), timeSerieID (see below), datasetName (string) and observer (string). Hence, with PA data, there is generally many speciesID for a same combination of geolocation, year and dayOfYear, and potentially several observers or datasetNames.
Environmental data
We also provide spatialized geographic and
environmental data
to be used as input variables. For each species observation, we provide a four-band 128x128
satellite image
at 10m resolution around the occurrence location and
time series
of the past values for six satellite bands at the point location. Besides we provide various
environmental rasters
at the European scale (e.g., climatic and pedologic, land use and human footprint variables) from which you may extract local 2D images and build 3D tensors using the dataloader provided on our Github. These three types of input environmental data are illustrated in the Figure below. Note that you may also extract time series of climatic variables for any observation, since we provide monthly climatic rasters from 2000 to 2019.
![](https://filedn.com/ltHjVlFrSfSJL1S4hXqqTzy/Figure.jpg)
Figure.
Illustration of of the environmental data for an occurrence (glcID=4859165) collected in northern Switzerland (lon=8.5744;lat=47.7704) in 2021. A. The 1280x1280m satellite image patches sampled in 2021 around the observation. B. Quarterly time series of six satellite bands at the point location since winter 1999-2000. C. Three example bioclimatic images (~65x65km) around the observation, extracted from the provided environmental rasters.
In more details:
Satellite imagery
: 1280mx1280m RGB and NIR patches (four bands) centered at the observation geolocation and taken the same year. The patches are compressed in two zip files (patchs_rgb.zip, patchs_nir.zip) accessible in folder /SatelliteImages/.
Format: 128x128 JPEG images, a color JPEG file for RGB data and a grayscale one for Near-Infrared.
Resolution: 10 meter per pixel
Source: Sentinel2 remote sensing data pre-processed by the
[Ecodatacube platform](https://stac.ecodatacube.eu/)
Access: One first needs to download and decompress the files patchs_rgb.zip and patchs_nir.zip (located in /SatelliteImages/), generating the folders ./rgb and ./nir respectively. Each contains around 4 millions JPEG images organised in a folder tree of depth two. Each JPEG file is named with a numeric identifier called the "patchID" representing a unique combination of geolocation and year. To recover the RGB (resp. NIR) patch of a species observation, the patchID column of the occurrence CSV is the key to the file path with the rule '.../CD/AB/XXXXABCD.jpeg'. For instance, the path to RGB image of patchID=3018575 is "./rgb/75/85/3018575.jpeg". In the case where the patchID has less than 4 decimals, just add zeros before to get the folder path.
Time series
Satellite time series:  Each observation is associated with the time series of the satellite median point values over each season since winter 1999 for six satellite bands (RGB, NIR, SWIR1 and 2). This data carry a high resolution local signature of the past 20 years succession of seasonal vegetation changes, potential extreme natural events (fires) or land use changes.
Format: Six CSV files, one per band. The rows are the timeSerieIDs (identifying a combination of location and day) and the columns are the 84 seasons from winter 2000 until automn 2020.
Resolution: The original satellite data has a resolution of 30m per pixel
Source:  Landsat remote sensing data pre-processed by the
[Ecodatacube platform](https://stac.ecodatacube.eu/)
Access: /SatelliteTimeSeries/
Monthly climatic rasters: Four climatic variables computed monthly (mean, minimum and maximum temperature and total precipitation) from january 2000 to december 2019, yielding 960 low-resolution rasters covering Europe.
Format:  GeoTIFF files with compression
Resolution: ~1 kilometer
Source:
[Chelsa](https://chelsa-climate.org/)
Access: EnvironmentalRasters/Climate/Climatic_Monthly_2000-2019
Environmental rasters
all provided in longitude/latitude coordinates (WGS84)
Bioclimatic rasters:  19 low-resolution rasters covering Europe. These 19 variables are classically used in species distribution modeling.
Format: GeoTIFF files with compression
Resolution: 30 arcsec (~ 1 kilometer)
Source:  Landsat remote sensing data pre-processed by the
[Ecodatacube platform](https://stac.ecodatacube.eu/)
Access: /EnvironmentalRasters/Climate/BioClimatic_Average_1981-2010
Soil rasters: Nine pedologic low-resolution rasters covering Europe. These variables (e.g., pH, clay, organic carbon and nitrogen contents, check definition.txt) describe the soil properties from 5 to 15cm depth and are determinant of plant species distributions.
Format:  GeoTIFF files with compression
Resolution: ~1 kilometer
Source:
[Soilgrids](https://soilgrids.org/)
Access: /EnvironmentalRasters/Soilgrids
Elevation: High-resolution raster covering Europe.
Format:  GeoTIFF file with compression, Int16 numeric storage (13.2GB).
Resolution: 1 arc second (~30 meter)
Source:
[ASTER Global Digital Elevation Model V3](https://lpdaac.usgs.gov/products/astgtmv003/)
Access: EnvironmentalRasters/Elevation
Land Cover: A medium-resolution multi-band land cover raster covering Europe. Each band describes either the land cover class prediction or its confidence under various classifications. We recommend the use of IGBP (17 classes) or LCCS (43 classes) layers, often used in species distribution modeling.
Format:  GeoTIFF file with compression.
Resolution: ~500m
Source:
[MODIS Terra+Aqua 500m](https://ladsweb.modaps.eosdis.nasa.gov/missions-and-measurements/products/MCD12Q1/#overview)
Access: EnvironmentalRasters/LandCover/
Humand footprint: Several low-resolution rasters describing human footprint, encapsulating seven pressures on the environment (e.g., nighlight level, population density) induced by human presence and activity, are provided for two time periods, the early 90's (~1993) and late 2000' (~2009). We provide two summary rasters combining all human pressures and two detailed rasters per pressure which avoid an arbitrary degradation of the original data.
Format:  GeoTIFF files with compression.
Resolution: ~1km
Source:
[Venter et al., 2016](https://datadryad.org/stash/dataset/doi:10.5061/dryad.052q5)
Access: The folder EnvironmentalRasters/HumanFootprint/  contain a readme file with the detailed information about this data and two subfolders: summarized/ contains the two summary rasters (for 1993 and 2009) and detaild/ contains the 2*7 single pressure raster.
Data downloading
We noticed that the download of data folders in
seaFile
is long when they contain large files. This is due to a compression of the folders prior to download. We will soon share a few archives containing the whole dataset for direct download. Note that you can still directly download the files without compression by using the download button on the right of each file row.
To download the full dataset, we recommend to download the following (groups of) folders independently and according to your needs (pre-compression time on seaFile in brackets):
Presence_Absence_surveys (seconds)
For_submission (seconds)
EnvironmentalRasters/Climate/BioClimatic_Average_1981-2010 (~1min)
Presence_only_occurrences (~2min)
EnvironmentalRasters/LandCover + HumanFootprint + SoilGrids (~3min)
SatelliteTimeSeries (~10min)
EnvironmentalRasters/Climate/Climatic_Monthly_2000_2019 (~25min)
Then, directly download the following files (to avoid the pre-zipping) by clicking on the direct download button on the right of the file name in the
seaFile
repository:
SatelliteImages/patchs_nir.zip AND patchs_rgb.zip (~17Go)
EnvironmentalRasters/Elevation/ASTER_Elevation.tif (~13Go)