Document Type

Article

Publication Date

2026

DOI

10.3390/stats9030059

Publication Title

Stats

Volume

9

Issue

3

Pages

59

Abstract

Cross-basin tropical cyclone variability may exhibit complex, non-linear dependence structures influenced by large-scale climate modes and potential regime shifts. Reliance on traditional linear correlation measures without accounting for structural changes can therefore lead to misleading interpretations of global storm relationships. This study investigates the regional dependence structures of tropical cyclone counts across six major ocean basins (NA, ENP, WNP, NI, SI, and SP) from 1980 to 2024. We adopt a two-stage analytical framework integrating changepoint detection and copula modeling to address non-stationarity in both marginal distributions and dependence structures. First, we identify a significant structural break in the year 2000 via a penalised likelihood applied jointly to the d=6-variate Poisson series, with inter-basin dependence captured by a latent Gaussian process (the construction used by Lund et al. (2025). This is mathematically equivalent to a Gaussian copula with Poisson margins (Genest and Nešlehová (2007)). Then, we apply bivariate copula models separately to the pre- and post-2000 regimes using the randomized probability integral transform with results averaged over 500 replications of the auxiliary uniforms to mitigate randomization noise. The results reveal substantial non-stationarity, most notably a 59% increase in North Atlantic storm frequency and a fundamental reorganization of global dependence structures, while dependence structures evolved from primarily symmetric and weak (dominated by Gaussian and Clayton copulas) to more complex and stronger dependencies (increased Frank and Gumbel copulas). Notably, a statistically significant (p< 0.001) and strong negative dependence emerged between the Southern Pacific and Northern Indian basins (τ=-0.464) in the recent regime. The inclusion of changepoint detection significantly improves model fit and reveals a fundamental reorganization of global tropical cyclone teleconnections, with enhanced coordination between basins in the contemporary climate regime. Modeling these regimes separately, as opposed to a single stationary period, uncovers a shift towards more complex, tail-dependent copula families (Gumbel, Clayton) in the recent era. These findings have important implications for climate risk assessment, seasonal forecasting, and understanding the impacts of climate change on global storm patterns. The proportion of Gumbel copulas (capturing upper-tail dependence) increased from 7% to 20%, while Gaussian copulas decreased from 53% to 33%, indicating more complex, extreme-value-focused dependencies in the contemporary climate. Due to small sample sizes (n₁=20, n₂=25), copula and dependence estimates are exploratory, not confirmatory. Interpretations reflect this power constraint, utilizing Benjamini-Hochberg adjustments for significance.

Rights

© 2026 by the authors.

This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution 4.0 International (CC BY 4.0) License.

Data Availability

Article states: "The real data used for this research are from the National Centers for Environmental Information https://www.ncei.noaa.gov/ (accessed on 1 May 2026) and can be found at the website (https://www.ncei.noaa.gov/data/international-best-track-archive-for-climate-stewardship-ibtracs/ (accessed on 1 May 2026)). All analyses were performed in R version 4.5.1 using the packages copula, VineCopula, mvtnorm, mnormt, GA, tidyverse, gridExtra, ggplot2, and maps. The full set of R scripts are available at a GitHub repository: https://github.com/iqbal1012/CP-Copula-Model-to-Count-TS.git (accessed on 1 May 2026)."

Original Publication Citation

Hossain, M. I., & Diawara, N. (2026). A two-stage changepoint-copula framework for non-stationary count time series: Application to tropical cyclones. Stats, 9(3), Article 59. https://doi.org/10.3390/stats9030059

ORCID

0009-0000-7297-4953 (Hossain), 0000-0002-8403-6793 (Diawara)

Download

Share

COinS