This project generates global, high-resolution plant trait maps based on crowdsourced biodiversity monitoring and Earth observation. The maps cover 31 plant functional traits as defined in the TRY Plant Trait Database at 1 km resolution with global extent.
Plant functional traits are predicted as community-weighted means based on a synthesis of crowdsourced biodiversity data (GBIF species observations, sPlot vegetation surveys, and in situ trait measurements from the TRY trait database) modeled as a function of Earth observation data, including surface reflectance, climatic variables, soil properties, canopy height, and vegetation optical depth.
├── LICENSE
├── README.md <- The top-level README for developers using this project
├── data
│ ├── external <- Data from third party sources
│ ├── features <- Features for modeling
│ │ └── Shrub_Tree_Grass <- PFT-specific features at different resolutions
│ ├── interim <- Intermediate data that has been transformed
│ │ ├── biomes <- Biome data
│ │ ├── eo_data <- Earth Observation data
│ │ ├── gbif <- GBIF processed data
│ │ ├── other_trait_maps <- Other published trait maps for comparison
│ │ └── splot <- sPlot processed data
│ ├── processed <- The final, canonical data sets for modeling
│ │ └── Shrub_Tree_Grass <- PFT-specific processed data at different resolutions
│ └── raw <- The original, immutable data dump
│
├── models <- Trained and serialized models
│ └── Shrub_Tree_Grass <- PFT-specific models at different resolutions
│
├── notebooks <- Jupyter notebooks organized by analysis stage
│ ├── 0-0_earth_observation_data <- Earth observation data exploration
│ ├── 0-1_global_trait_data <- Trait data exploration and preparation
│ ├── 1-0_featurize_train <- Feature engineering and model training
│ ├── 2-0_analysis_prep <- Analysis preparation
│ ├── 2-1_model_analysis <- Model analysis notebooks
│ └── misc <- Miscellaneous notebooks
│
├── reference <- Data dictionaries, manuals, and explanatory materials
│
├── results <- Generated analysis results
│ └── figures <- Generated graphics and figures
│
├── src <- Source code for use in this project
│ ├── __init__.py <- Makes src a Python module
│ ├── analysis <- Analysis code
│ ├── conf <- Configuration management
│ ├── data <- Scripts to download or generate data
│ ├── features <- Scripts to turn raw data into features
│ ├── io <- Input/Output utilities
│ ├── models <- Model training and evaluation
│ ├── utils <- Utility functions
│ └── visualization <- Data visualization scripts
│
├── tests <- Unit tests for use with `pytest`
│
├── tmp <- Temporary files directory
│
├── pyproject.toml <- Human-readable project dependencies managed with Poetry
├── poetry.lock <- File used by Poetry to install dependencies
├── conda-linux-64.lock <- File used by conda-lock for 64-bit Linux systems
├── environment.yml <- File used by conda-lock to specify dependencies
├── dvc.yml <- DVC pipeline definitions
├── params.yml <- DVC parameter definitions and project configuration
├── dvc.lock <- DVC file tracking changes to data
├── pyrightconfig.json <- Python type checker configuration
└── .pre-commit-config.yaml <- pre-commit Git hooks
This project uses Data Version Control (DVC) to manage data dependencies and pipeline execution. To reproduce the full workflow:
-
Clone the repository
git clone https://github.com/GeoSense-Freiburg/cit-sci-traits.git cd cit-sci-traits -
Install dependencies (using conda or poetry as preferred)
# Using conda conda env create -f environment.yml conda activate cit-sci-traits # OR using poetry poetry install
-
Install DVC for pipeline reproduction
pipx install dvc
-
Pull the data dependencies using DVC (requires collaborative setup: contact daniel.lusk[at]geosense.uni-freiburg.de for more information)
dvc pull
-
Reproduce the entire pipeline
dvc repro
Or run specific pipeline stages:
dvc repro <stage-name>
For more detailed information on DVC commands, see the DVC documentation.
The predictor data used for this experiment is all open source. See https://github.com/GeoSense-Freiburg/panops-data-registry for the exact Google Earth Engine scripts used to retrieve the predictor data.
Similarly, the TRY trait data is also available for download via https://trydb.org. See params.yaml or the table below for a list of traits used.
GBIF observations are also freely available. See https://github.com/GeoSense-Freiburg/panops-data-registry/references/gbif/query_all_tracheophyta.json for the exact query used in this experiment, as well as https://github.com/GeoSense-Freiburg/panops-data-registry/src/gbif/get_gbif_data.py for the download script.
sPlot vegetation data was obtained with express permission from the sPlot consortium. For an open-access version of sPlot data, sPlotOpen can be used.
| Trait | TRY trait name | TRY ID | Unit |
|---|---|---|---|
| Conduit element length | Wood vessel element length; stem conduit (vessel and tracheids) element length | 282 | µm |
| Dispersal unit length | Dispersal unit length | 237 | mm |
| LDMC | Leaf dry mass per leaf fresh mass (leaf dry matter content, LDMC) | 47 | g g-1 |
| Leaf area | Leaf area (in case of compound leaves: leaflet, undefined if petiole is in- or excluded) | 3113 | mm² |
| Leaf C | Leaf carbon (C) content per leaf dry mass | 13 | mg g-1 |
| Leaf C/N ratio | Leaf carbon/nitrogen (C/N) ratio | 146 | g g-1 |
| Leaf delta 15N | Leaf nitrogen (N) isotope signature (delta 15N) | 78 | ppm |
| Leaf dry mass | Leaf dry mass (single leaf) | 55 | g |
| Leaf fresh mass | Leaf fresh mass | 163 | g |
| Leaf length | Leaf length | 144 | mm |
| Leaf N (area) | Leaf nitrogen (N) content per leaf area | 50 | g m-2 |
| Leaf N (mass) | Leaf nitrogen (N) content per leaf dry mass | 14 | mg g-1 |
| Leaf P | Leaf phosphorus (P) content per leaf dry mass | 15 | mg g-1 |
| Leaf thickness | Leaf thickness | 46 | mm |
| Leaf water content | Leaf water content per leaf dry mass (not saturated) | 3120 | g g-1 |
| Leaf width | Leaf width | 145 | mm |
| Plant height | Plant height (vegetative) | 3106 | m |
| Rooting depth | Root rooting depth | 6 | m |
| Seed germination rate | Seed germination rate (germination efficiency) | 95 | days |
| Seed length | Seed length | 27 | mm |
| Seed mass | Seed dry mass | 26 | mg |
| Seed number | Seed number per reproduction unit | 138 | - |
| SLA | Leaf area per leaf dry mass (specific leaf area, SLA or 1/LMA): undefined if petiole is in- or excluded) | 3117 | m² kg-1 |
| SRL | Root length per root dry mass (specific root length, SRL) | 1080 | cm g-1 |
| SRL (fine) | Fine root length per fine root dry mass (specific fine root length, SRL) | 614 | cm g-1 |
| SSD | Stem specific density (SSD) or wood density (stem dry mass per stem fresh volume) | 4 | g cm-3 |
| Stem conduit density | Stem conduit density (vessels and tracheids) | 169 | mm-2 |
| Stem conduit diameter | Stem conduit diameter (vessels, tracheids) | 281 | µm |
| Stem diameter | Stem diameter | 21 | m |
| Wood fiber lengths | Wood fiber lengths | 289 | µm |
| Wood ray density | Wood rays per millimetre (wood ray density) | 297 | mm-1 |
More information on the PANOPS project is available at: