Forecasting deforestation and carbon loss across New Guinea using machine learning and cellular automata

Area of Science:

• Ecology
• Environmental Science
• Remote Sensing

Background:

• New Guinea's ecosystems are highly biodiverse and endemic.
• Land-use change poses a significant threat to biodiversity and carbon stocks.
• Understanding deforestation drivers and future risks is crucial for conservation.

Purpose of the Study:

• Compare deforestation drivers between Indonesian New Guinea and Papua New Guinea.
• Identify high-risk deforestation areas under various development scenarios.
• Evaluate the impact of deforestation on carbon pools.

Main Methods:

• Integrated machine learning and cellular automata for deforestation modeling.
• Forecasted deforestation and assessed carbon stock loss for four scenarios (4.8%–28% forest loss).
• Analyzed deforestation risk across different altitudes and regions.

Main Results:

• High deforestation risk consistently forecasted in lowland regions.
• Deforestation in Indonesian New Guinea primarily below 380m a.s.l., higher in Papua New Guinea (<750m a.s.l.).
• Land-use change could result in 156–918 Mt carbon loss in Indonesian New Guinea and 223–1082 Mt in Papua New Guinea.

Conclusions:

• Machine learning and cellular automata effectively forecast high-resolution deforestation.
• New Guinea's lowlands are highly vulnerable to future deforestation.
• Prioritize protection of key areas where deforestation conflicts with carbon, ecosystem, and biodiversity conservation.