Mapping of compositional properties of coal using isometric log-ratio transformation and sequential Gaussian...

Area of Science:

• Geochemistry
• Data Science
• Mining Engineering

Background:

• Accurate mapping of coal chemical properties is crucial for safe and efficient mining and power generation.
• Ultimate analysis quantifies key elements (C, H, N, S, O) and moisture/ash content, vital for predicting coal behavior.
• Compositional data analysis presents unique statistical challenges, particularly in spatial mapping.

Purpose of the Study:

• To accurately map the ultimate analysis components of Springfield coal from the Illinois Basin.
• To compare the effectiveness of sequential Gaussian simulation on isometric log-ratio transformed compositions versus direct simulation of compositional parts.
• To evaluate the impact of different mapping approaches on predicting other coal properties via correlations.

Main Methods:

• Application of sequential Gaussian simulation to isometric log-ratio (ILR) transformed compositional data.
• Direct simulation of compositional parts for comparative analysis.
• Correlation analysis to assess the implications of mapping methods on derived coal properties.

Main Results:

• The study demonstrates a robust method for mapping compositional data in coal.
• Comparison reveals differences in mapping accuracy and implications for property prediction between ILR-transformed and direct simulation methods.
• The chosen compositional data treatment significantly impacts the reliability of spatial predictions.

Conclusions:

• Advanced compositional data analysis, specifically ILR transformation with sequential Gaussian simulation, provides a superior approach for mapping coal's ultimate analysis.
• Accurate spatial mapping of chemical properties is essential for reliable quantitative predictions in coal resource management.
• The presented methodology is broadly applicable to spatial mapping of any compositional dataset.