Species Outlier Detection Using Random Forest

This repository provides two Python implementations for outlier detection from species observation data using species distribution modeling (SDM) and machine learning models (Random Forest & three unsupervised models). The workflow includes data preparation, spatial sampling for background data, environmental enrichment, model training, evaluation, and visualization of results. The model integrates re-classified CORINE data, raster-based environmental variables, and occurrence data to predict species probabilities.

NOTE: Raster datasets are not in this repository as they were too large (> 1 GB). You can download them from the data download services by SYKE, NLS-FI, FMI and LUKE.

How to use

Dependencies

Install the dependencies using:

pip install -r requirements.txt

Enviromental variables

And store the following credentials in a .env file for API access:

VIRVA_ACCESS_TOKEN: Access token for querying sensitive occurrence data from api.laji.fi.
ACCESS_TOKEN: Access token for querying open data from api.laji.fi.
ACCESS_EMAIL: Email for sensitive data queries from api.laji.fi.

Usage

1. Modify the taxon_id parameter for your target species or taxon
2. Select features you want to use. All features are not always suitable for all species. For example, tree_vol can be suitable for flying squirrels but not for fishes. Edit the row:

feature_columns = ['x', 'y', 'Urban', 'Park', 'Rural', 'Forest', 'Open forest', 'Fjell', 'Open area', 'Wetland', 'Open bog', 'Freshwater', 'Marine', 'dem', 'nvdi', 'rain', 'temp', 'tree_vol', 'gathering.conversions.dayOfYearBegin']

Models' outputs

1. Predicted probabilities [0,1] and classifidications (0 for outliers, 1 for inliers) for each observation. Note that all models report false-positive outliers, meaning that models always find outliers even from the data that doesn't have any real outliers.
2. You can visualise predicted probabilities to the map or store them as a GeoPackage file.