# Generating realistic synthetic meteoroid orbits

## TL;DR

This paper introduces a new Kernel Density Estimation-based method for generating synthetic meteoroid orbits that better reproduces observed data structures and statistics, aiding in more accurate meteor shower analysis.

## Contribution

A novel synthetic meteoroid orbit generation method using Kernel Density Estimation that balances data structure preservation and statistical accuracy.

## Key findings

- Improved synthetic orbit generation over existing methods.
- Demonstrated ability to reproduce observed data structure and statistics.
- Provided visualization tools for parameter influence assessment.

## Abstract

Context. Generating a synthetic dataset of meteoroid orbits is a crucial step in analysing the probabilities of random grouping of meteoroid orbits in automated meteor shower surveys. Recent works have shown the importance of choosing a low similarity threshold value of meteoroid orbits, some pointing out that the recent meteor shower surveys produced false positives due to similarity thresholds which were too high. On the other hand, the methods of synthetic meteoroid orbit generation introduce additional biases into the data, thus making the final decision on an appropriate threshold value uncertain. Aims. As a part of the ongoing effort to determine the nature of meteor showers and improve automated methods, it was decided to tackle the problem of synthetic meteoroid orbit generation, the main goal being to reproduce the underlying structure and the statistics of the observed data in the synthetic orbits. Methods. A new method of generating synthetic meteoroid orbits using the Kernel Density Estimation method is presented. Several types of approaches are recommended, depending on whether one strives to preserve the data structure, the data statistics or to have a compromise between the two. Results. The improvements over the existing methods of synthetic orbit generation are demonstrated. The comparison between the previous and newly developed methods are given, as well as the visualization tools one can use to estimate the influence of different input parameters on the final data.