Detecting Filamentarity in Climate and Galactic Spatial Point Processes

TL;DR

This paper introduces a new statistical method to detect filamentary structures in spatial point data, applied to climate and galactic datasets, demonstrating effective identification of filamentarity and outliers.

Contribution

The paper develops a diagnostic test based on aligned triads and tetrads, and proposes a Poisson filament process model with arc search and ABC inference, advancing filament detection techniques.

Findings

Strong evidence of filamentarity in both applications

Method successfully identifies outlying precipitation data sets

Simulations show good performance of the proposed approach

Abstract

Evidence of excess filamentarity is considered for two spatial point process applications: local minima in whole earth precipitation modelling and locations of cold clumps in the Milky Way. A diagnostic test using the number of aligned triads and tetrads is developed. A Poisson filament process is proposed based on a parent Poisson process with correlated random walk offspring locations. Filaments are initially identified using an arc search method, with ABC for subsequent inference. Simulations indicate good performance. In both applications there is strong evidence of filamentarity. The method successfully identifies two outlying precipitation data sets.