Back

Predicting near-future deforestation in West African Key Biodiversity Areas to inform conservation urgency

Trew, B. T.; Edwards, F. A.; Buchanan, G. M.; Sanderson, F. J.

2024-10-08 ecology
10.1101/2024.10.07.616969 bioRxiv
Show abstract

Site-based protection is a cornerstone of 21st century conservation and a core component of global biodiversity conservation targets. However, loss of tropical forests, the most biodiverse of habitats, is a major threat to such sites. Here, we predict near-future deforestation risk in 113 Key Biodiversity Areas (KBAs) - sites of objectively defined global conservation importance - in the Guinean Forest biodiversity hotspot and identify those factors associated with loss. We find that (i) KBAs in the Guinean Forests have lost over 265,000 hectares of forest cover over the past decade, with those in Cote dIvoire experiencing the highest forest loss; (ii) future deforestation risk is highest among KBAs in Liberia and Cote dIvoire (on average 10% predicted loss across KBAs), where some are predicted to lose over a fifth of remaining forest cover by 2033. Models indicate that deforestation is highly contagious, with historical forest loss effectively predicting further adjacent loss, and that forest fragmentation and ease of human access also increased the localised deforestation risk. Conversely, predicted forest loss was lower in sites under some form of conservation protection. Our methods to predict near-future hotspots of deforestation risk in KBAs are reproducible and therefore applicable to other biodiversity hotspots. In the Guinean forests, our results highlight where conservation interventions to mitigate forest loss should be urgently prioritised.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.