Causal machine learning methods for understanding land use and land cover change

Area of Science:

• Environmental Science
• Data Science
• Machine Learning

Background:

• Understanding land use and land cover change (LULCC) drivers is a critical research challenge due to complex socio-ecological processes.
• Causal modeling is essential for effective LULCC policy formulation, but current approaches face difficulties.
• The LULCC community has limited understanding and application of causal machine learning (ML) methods.

Purpose of the Study:

• To introduce the state-of-the-art in causal ML methods for LULCC research.
• To outline the limitations and potential applications of causal ML in understanding LULCC dynamics.
• To bridge the gap between causal ML advancements and LULCC community needs.

Main Methods:

• Conducted two workshops to identify promising ML methods for LULCC.
• Focused on methods applicable to understanding complex LULCC dynamics.
• Synthesized insights from domain experts and ML practitioners.

Main Results:

• Identified key causal ML approaches relevant to LULCC challenges.
• Provided an overview of the limitations associated with these causal ML methods.
• Illustrated challenges in causal modeling of LULCC with a simple example.

Conclusions:

• Causal ML methods show significant potential for enhancing causal modeling of LULCC.
• Effective policy design requires combining causal ML with deep domain understanding and qualitative insights.
• Addressing the complexity of LULCC dynamics is crucial for successful application of causal ML.