DivShift: Exploring Domain-Specific Distribution Shift in Volunteer-Collected Biodiversity Datasets
Paper • 2410.19816 • Published
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
DivShift Paper | Extended Version | Code
https://doi.org/10.1609/aaai.v39i27.35060
Highlighting biases through partitions of volunteer-collected biodiversity data and ecologically relevant context for plant photographs.

Sourced from iNaturalist open data on AWS, WorldClim, the Harmonized World Soil Database, and the Human Footprint Index.
License: CC-BY-NC-SA
| Column Name | Description |
|---|---|
| state_name | State name |
| observation_uuid | ID of the observation |
| observer_id | ID of the iNaturalist observer |
| latitude | latitude (°) |
| longitude | longitude (°) |
| positional_accuracy | Accuracy of latitude and longitude (m) |
| taxon_id | Taxon ID of plant |
| quality_grade | Quality of observation |
| observed_on | Date of observation |
| ancestry | Higher-order identifiers of species taxonomy (e.g. genus, family) |
| rank_level | Name of rank ID of taxon |
| rank | Rank ID of taxon (e.g. 10 for species, 9 for genus) |
| name | Species name |
| date | Date of observation in Datetime form |
| state | State name |
| photo_uuid | Unique photo identifier on iNaturalist website |
| photo_id | Unique photo identifier on iNaturalist Open Data S3 bucket |
| extension | The photo extension |
| license | What creative commons license the photo is shared under |
| width | Width of image |
| height | Height of image |
| position | For a given observation with more than one associated photo, what is the order of the given image in the observation |
| l3_ecoregion | L3 ecoregion name |
| l2_ecoregion | L2 ecoregion name |
| l1_ecoregion | L1 ecoregion name |
| l4_ecoregion | L4 ecoregion name |
| bioclim_14 | Precipitation of Driest Month |
| bioclim_13 | Precipitation of Wettest Month |
| bioclim_8 | Mean Temperature of Wettest Quarter |
| bioclim_1 | Annual Mean Temperature |
| bioclim_6 | Min Temperature of Coldest Month |
| bioclim_12 | Annual Precipitation |
| bioclim_15 | Precipitation Seasonality (Coefficient of Variation) |
| bioclim_7 | Temperature Annual Range (BIO5-BIO6) |
| bioclim_9 | Mean Temperature of Driest Quarter |
| bioclim_3 | Isothermality (BIO2/BIO7) (×100) |
| bioclim_4 | Temperature Seasonality (standard deviation ×100) |
| bioclim_16 | Precipitation of Wettest Quarter |
| bioclim_11 | Mean Temperature of Coldest Quarter |
| bioclim_18 | Precipitation of Warmest Quarter |
| bioclim_5 | Max Temperature of Warmest Month |
| bioclim_2 | Mean Diurnal Range (Mean of monthly (max temp - min temp)) |
| bioclim_19 | Precipitation of Coldest Quarter |
| bioclim_10 | Mean Temperature of Warmest Quarter |
| bioclim_17 | Precipitation of Driest Quarter |
| land_use | North American Environmental Atlas Land Cover |
| soil_type | Harmonized World Soil Database V 2.0 |
| supervised | Whether the image is labeled |
| single_image | Whether this image is a unique observation |
| year | Year image taken |
| human_footprint | Human Footprint index (Wilderness to Modified) |
| not_city_nature | Train and Test images taken outside of the city nature challenge |
| city_nature | Train and Test images taken in the city nature challenge |
| alaska_socioeco | Train and Test images in Alaska |
| arizona_socioeco | Train and Test images in Arizona |
| baja_california_socioeco | Train and Test images in Baja California |
| baja_california_sur_socioeco | Train and Test images in Baja California Sur |
| british_columbia_socioeco | Train and Test images in British Columbia |
| california_socioeco | Train and Test images in California |
| nevada_socioeco | Train and Test images in Nevada |
| oregon_socioeco | Train and Test images in Oregon |
| sonora_socioeco | Train and Test images in Sonora |
| washington_socioeco | Train and Test images in Washington |
| yukon_socioeco | Train and Test images in Yukon |
| obs_engaged | Train and Test images taken by Engaged Observers (users with more than 1,000 unique observations) |
| obs_casual | Train and Test images taken by Casual Observers (users with less than 50 unique observations) |
| spatial_wilderness | Train and Test images from Wilderness regions |
| spatial_modified | Train and Test images from Human Modified regions |
| inat2021 | Train and Test images from iNat21 |
| inat2021mini | Train and Test images from iNat21 Mini |
| imagenet | Train and Test images from ImageNet |
| spatial_split | Spatially stratified sampled Train and Test images |
| taxonomic_balanced | Taxonomically balanced Train and Test images |
| taxonomic_unbalanced | Taxonomicall unbalance Train and Test images |
@article{Sierra_Gillespie_Soltani_Exposito-Alonso_Kattenborn_2025,
author = {Sierra, Elena and Gillespie, Lauren E. and Soltani, Salim and Exposito-Alonso, Moises and Kattenborn, Teja},
doi = {10.1609/aaai.v39i27.35060},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
month = {Apr.},
number = {27},
pages = {28386-28396},
title = {DivShift: Exploring Domain-Specific Distribution Shifts in Large-Scale, Volunteer-Collected Biodiversity Datasets},
url = {https://ojs.aaai.org/index.php/AAAI/article/view/35060},
volume = {39},
year = {2025},
bdsk-url-1 = {https://ojs.aaai.org/index.php/AAAI/article/view/35060},
bdsk-url-2 = {https://doi.org/10.1609/aaai.v39i27.35060}
}