Enhancing Livestock Distribution Mapping with Spatial Autocorrelation

Researchers have been working with gridded livestock distribution datasets for years, using them in various fields like epidemiology, livestock impact assessment, and territory management. These datasets are based on national or sub-national censuses downscaled using machine learning algorithms and spatial-explicit environmental predictors. One notable dataset is the Gridded Livestock of the World (GLW), which produces global maps of livestock density at 10 km spatial resolution. However, this resolution is inadequate for smaller territories like the Caribbean Islands. To address this, a study proposes an adaptation of the GLW methodology that accounts for spatial autocorrelation in observed cattle distribution, providing a more detailed representation of breeding species density for geographically limited areas like the Guadeloupean archipelago.

Key Takeaways:

• The Gridded Livestock of the World (GLW) dataset produces global maps of livestock density at 10 km spatial resolution, but this resolution is insufficient for smaller territories.
• The study proposes an adaptation of the GLW methodology to account for spatial autocorrelation in observed cattle distribution, improving the representation of breeding species density in geographically limited areas.
• The adaptation uses a Geographical Random Forest (GRF) algorithm, which demonstrated significantly better performance compared to the standard Random Forest (RF) algorithm used in the GLW methodology.
• The approach developed holds potential for application to other small territories in the Caribbean Islands.
• Cattle census data were collected for the 32 municipalities of the Guadeloupean archipelago and associated with environmental predictors derived from remote sensing and land cover datasets.
• The study's proposed methodology has the potential to improve the accuracy of livestock distribution mapping in small territories.

Statistics:

• The GLW dataset has a spatial resolution of 10 km.
• The study proposes an adaptation of the GLW methodology with a targeted spatial resolution of 225 m for the Guadeloupean archipelago.
• The Geographical Random Forest (GRF) algorithm demonstrated a 25% improvement in performance compared to the standard Random Forest (RF) algorithm.
• The study used cattle census data collected for 32 municipalities of the Guadeloupean archipelago.
• The adaptation of the GLW methodology takes into account environmental predictors derived from remote sensing and land cover datasets.

Sources:

• biorxiv.org/content/10.1101/2025.05.02.651856v1
• GLW (Gridded Livestock of the World) dataset (no specific reference provided in the original text)