Multi-scale habitat modelling and predicting change in the distribution of tiger and leopard using random forest...

Area of Science:

• Ecology
• Conservation Biology
• Machine Learning Applications

Background:

• Tiger and leopard populations are declining due to habitat loss and fragmentation.
• Traditional statistical models struggle with limited ecological data for accurate species distribution mapping.
• Understanding multi-scale habitat relationships is crucial for effective conservation planning.

Purpose of the Study:

• To investigate multi-scale habitat relationships of tigers and leopards.
• To predict the impact of future climate change on their habitat suitability.
• To assess niche overlap and identify critical conservation areas using advanced machine learning.

Main Methods:

• Collected species occurrence data (camera traps, scats) over two years.
• Applied the Random Forest (RF) machine learning algorithm for habitat suitability modeling.
• Developed niche overlap models to assess species similarity.

Main Results:

• Tigers and leopards utilize habitat resources at broad spatial scales (up to 28,000 m).
• A 23% loss in suitable tiger habitat is predicted under the RCP 8.5 scenario by 2050.
• Identified potential refugee habitats for large carnivores in disturbed landscapes.

Conclusions:

• Multi-scale habitat suitability modeling using RF provides accurate predictions even with limited ecological knowledge.
• Conservation efforts should focus on identified potential refugee habitats to protect tigers and leopards.
• The study offers a methodological framework for multi-scale, multi-species monitoring using machine learning.