Reliability-Constrained Statistical Analysis of NASA CNEOS Fireball Events: Implications for Bolide Population Studies

Shreyas Khobragade 

The NASA Center for Near-Earth Object Studies (CNEOS) Fireball Database is one of the most widely used resources for investigating global bolide populations. However, variations in event reliability, detection thresholds, and observational coverage introduce potential selection biases that may influence inferred population statistics. While the database has been extensively employed in studies of meteoroid fluxes, atmospheric entry processes, and planetary defense, the quantitative impact of reliability-based filtering on large-scale fireball statistics remains insufficiently characterized.

In this study, we analyze 1,064 fireball events recorded in the NASA CNEOS Fireball Database to quantify the influence of observational selection effects on inferred bolide population properties. The catalogue was partitioned into six subsets based on observation period, impact-energy thresholds, and reliability-related criteria, including full, post-2017, high-energy (>0.45 kt), reliable, combined-reliable, and low-reliability populations. Statistical analyses included power-law fitting of impact-energy distributions, temporal trend estimation, area-normalized spatial distribution assessment, correlation analysis of impact energy and peak-brightness altitude, and non-parametric distribution comparison using Kolmogorov--Smirnov and Mann--Whitney U tests.

Results demonstrate persistent heavy-tailed scaling behaviour across multiple independently filtered populations. The full dataset, high-energy subset, and combined-reliable subset converge toward a consistent power-law exponent of approximately α ≈ 1.94, suggesting robust underlying population characteristics despite substantial differences in event selection. In contrast, the low-reliability subset yields a significantly steeper exponent (α ≈ 2.55), indicating that unreliable detections can substantially bias inferred scaling relationships. Distribution comparison tests reveal highly significant differences between reliability-constrained populations (KS p-values ranging from 10⁻³⁴ to 10⁻¹¹⁹; MWU p-values ranging from 10⁻²³ to 10⁻⁷⁸). Spatial analyses indicate broadly uniform global distributions after area normalization, while high-energy events exhibit a weak negative relationship between impact energy and peak-brightness altitude (Spearman ρ ≈ -0.25).

These findings quantify the magnitude of reliability-selection bias within the CNEOS Fireball Database and demonstrate that reliability-aware filtering is essential for obtaining consistent estimates of bolide population statistics. The stability of α ≈ 1.94 across multiple independently filtered populations suggests a robust underlying scaling behaviour, while deviations observed in low-reliability events highlight the importance of dataset quality in meteoroid population modelling. The methodology presented here provides a reproducible framework for evaluating observational biases in heterogeneous fireball catalogues and supports improved future studies of meteoroid fluxes, atmospheric entry processes, and planetary defense applications.