Tree-based approach for exploring marine spatial patterns with raster datasets

Area of Science:

• Environmental Science
• Data Mining
• Geospatial Analysis

Background:

• Extracting spatial association patterns from multiple raster datasets is a complex data-mining challenge.
• Existing methods struggle with the spatial pattern exploration subtask.

Purpose of the Study:

• To propose a novel data-mining technique, TAXMarSP, for exploring marine spatial patterns from multiple raster datasets.
• To address the challenge of spatial pattern exploration in marine environmental data.

Main Methods:

• Developed a Tree-based Approach for eXploring Marine Spatial Patterns (TAXMarSP) comprising two models: Tree-based Cascading Organization Model (TCOM) and Spatial Neighborhood-based CAlculation Model (SNCAM).
• TCOM organizes frequent patterns, while SNCAM calculates pattern-matching degrees considering spatial neighborhood contributions.
• Utilized the Apriori algorithm for pattern quantification and applied it to a real remote sensing dataset (1998-2014).

Main Results:

• Successfully applied TAXMarSP to explore marine spatial patterns in the Pacific Ocean.
• The method identified both well-known and previously undiscovered marine spatial patterns.
• Demonstrated the effectiveness of TCOM and SNCAM in analyzing interrelated marine environmental parameters.

Conclusions:

• TAXMarSP provides an effective framework for discovering complex marine spatial patterns from raster data.
• The approach enhances our understanding of marine environments by revealing novel spatial associations.
• This method offers valuable insights for Earth scientists studying oceanic phenomena.