Insect-Based Biodiversity Intactness (IBI)

A reproducible workflow for modelling insect diversity and computing biodiversity intactness across Africa

This repository contains all scripts, workflows, and resources used for the study:

“An insect- and EO-based indicator reveals microhabitat-driven biodiversity intactness across Africa.”

The repository provides a complete implementation of the insect-based Biodiversity Intactness (IBI) indicator, including:

• Preprocessing of insect occurrence data
• Derivation of environmental predictor layers
• Random Forest species distribution modelling
• Per-pixel biodiversity intactness computation
• Uncertainty quantification
• Visualizations and statistical comparisons

---

📁 Repository Structure

---

🔧 Requirements

Software

• R ≥ 4.2
• Python ≥ 3.9
• QGIS ≥ 3.22
• Google Earth Engine account

Key R packages

• terra, raster, sf, dismo
• spThin
• ggplot2, dplyr

Key Python packages

• scikit-learn
• rasterio, geopandas
• numpy, pandas

---

📘 Project Overview

This repository implements the full workflow for creating the Insect-Based Biodiversity Intactness (IBI) indicator.

1. Current insect diversity (D)

Modelled using a Random Forest classifier trained on:

• Coleoptera, Lepidoptera, Odonata occurrences
• Sentinel-2 spectral metrics
• GEDI canopy height
• SRTM elevation
• Pseudo-absence/background points

Spatial block cross-validation ensures no spatial leakage.

---

2. Potential insect diversity (P)

Baseline values for pre-human potential diversity, assigned per habitat class.

Two scenarios included:

• High P
• Low P

Used for sensitivity analysis.

---

3. Human footprint (hF)

Derived from the Global Human Modification (GHM) dataset, normalized 0–1.

---

4. Biodiversity Intactness Indicator (IBI)

[ \text{IBI} = \frac{D}{P} \times (1 - hF) ]

Outputs range from 0 (degraded) to 1 (intact).

---

5. Uncertainty

Combined from:

• Distance-to-occurrence
• Predicted suitability
• Spatial block CV performance

Used to identify data-poor or model-uncertain regions.

---

🚀 Running the Workflow

Step 1 — Prepare Occurrence Data

Rscript 1_data_preprocessing/occurrence_cleaning.R
Rscript 1_data_preprocessing/spatial_thinning.R
Rscript 1_data_preprocessing/pseudo_absence_generation.R