Predicting Seabed Sand Content Across the Australian Margin Using Machine Learning

18 spatial interpolation methods were compared to predict seabed sand content across the Australian Exclusive Economic Zone. The study, based on samples from Geoscience Australia's Marine Samples Database extracted in August 2010, found that RFIDS and RFOK were among the most accurate methods across three tested regions. Model averaging further improved prediction accuracy, with the most accurate methods reducing error by up to 7%.

Use Cases

• Comparing machine learning and geostatistical methods for spatial interpolation of marine data.
• Predicting seabed sand content across the Australian Exclusive Economic Zone (AEEZ).
• Evaluating the impact of secondary variables and search window size on prediction accuracy.
• Applying model averaging techniques to improve the accuracy of environmental predictions.

Strengths

• Compares 18 distinct spatial interpolation methods for a specific prediction task.
• Reports a reduction in prediction error of up to 7% for the most accurate methods.
• Based on samples from Geoscience Australia's authoritative Marine Samples Database.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data freshness should be verified as the core samples are from a single month in 2010.

Provenance

Source

Geoscience Australia Data

Collection Method

Samples extracted from Geoscience Australia's Marine Samples Database.

Time Range

Samples extracted August 2010.

Freshness

Last updated 2026-04-30 13:02:39.124837; freshness should be verified

Geography

Australian Exclusive Economic Zone (AEEZ).