Three-phase hierarchical model-based and hybrid inference.

Area of Science:

• Ecology
• Remote Sensing
• Statistical Modeling

Background:

• Global environmental challenges necessitate accurate Earth ecosystem data.
• Integrating multi-source remote sensing and field data presents significant challenges in data fusion and uncertainty quantification.
• Existing methods often involve two-step modeling, limiting applications.

Purpose of the Study:

• To extend hierarchical modeling to three-step processes for enhanced ecosystem data analysis.
• To develop methods for deriving predictors and their variances from complex, nested data structures.
• To introduce new inference approaches for wall-to-wall and sample-based data in the final modeling phase.

Main Methods:

• Application of hierarchically nested models for predictor and variance derivation.
• Extension of two-step modeling to three-step processes, including intermediate prediction models.
• Development of three-phase hierarchical model-based inference (3pHMB) for wall-to-wall final data.
• Development of three-phase hierarchical hybrid inference (3pHHY) for probability sample final data.

Main Results:

• Demonstrated a framework for deriving predictors and variances in three-step hierarchical models.
• Successfully extended existing methodologies to accommodate more complex data integration scenarios.
• Established two distinct, named approaches (3pHMB and 3pHHY) for different final data availability.

Conclusions:

• The proposed three-step hierarchical modeling enhances the integration of remote sensing and field data for ecosystem assessment.
• The 3pHMB and 3pHHY methods provide robust frameworks for uncertainty assessment in environmental monitoring.
• These advancements are crucial for informing global commitments to climate change mitigation and biodiversity conservation.