EUVIMEDEUVIMED entdecken
Crossref

Latent Precipitation Component Analysis Using Latent Dirichlet Allocation

Asami Yamashita, Toru Shiina, Mamoru Ota

Journal of Atmospheric and Oceanic Technology · 2026-11 · 43 (11) · 1357-1369

DOI: 10.1175/jtech-d-25-0088.1 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Atmospheric and Oceanic Technology, 2026-11. Verfasst von Asami Yamashita und 2 weiteren Autor:innen. Die Arbeit nennt 16 Literaturangaben. Gefördert von: The First Bank of Toyama, Ltd., TAKEUCHI Scholarship Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

An accurate understanding of precipitation microphysics is essential for improving quantitative precipitation estimation, weather forecasting, and disaster risk reduction. Electromagnetic remote sensing provides wide-area observations of precipitation; however, differences in the scattering characteristics of precipitation particles complicate the estimation of precipitation intensity and phase. In this study, latent Dirichlet allocation (LDA) was applied to particle size–velocity distribution (PSVD) observations obtained by a Parsivel disdrometer. Unlike conventional clustering approaches, the proposed framework directly models the count structure of PSVD observations and determines the optimal number of latent precipitation components using the singular Bayesian information criterion (sBIC). Application to approximately 1 year of Parsivel observations identified two dominant latent precipitation components with rain-like and snow-like characteristics. Each 1-min PSVD is represented as a probabilistic mixture of these components, providing a continuous description of transitional precipitation conditions rather than a hard classification. Because the proposed framework does not require a prescribed continuous parametric form, such as a Gaussian distribution, for the PSVD, it provides a statistically principled, data-driven approach to precipitation characterization and establishes a foundation for future retrieval of precipitation composition and Doppler spectra from vertically pointing radar observations.

Förderung

  • The First Bank of Toyama, Ltd.
  • TAKEUCHI Scholarship Foundation

Zitieren

APA

Yamashita, A., Shiina, T., & Ota, M. (2026). Latent Precipitation Component Analysis Using Latent Dirichlet Allocation. Journal of Atmospheric and Oceanic Technology, 43(11), 1357-1369. https://doi.org/10.1175/jtech-d-25-0088.1

Vancouver

Yamashita A, Shiina T, Ota M. Latent Precipitation Component Analysis Using Latent Dirichlet Allocation. Journal of Atmospheric and Oceanic Technology. 2026;43(11):1357-1369. doi:10.1175/jtech-d-25-0088.1

BibTeX
@article{Yamashita2026,
  author = {Yamashita, Asami and Shiina, Toru and Ota, Mamoru},
  title = {Latent Precipitation Component Analysis Using Latent Dirichlet Allocation},
  journal = {Journal of Atmospheric and Oceanic Technology},
  year = {2026},
  volume = {43},
  number = {11},
  pages = {1357--1369},
  doi = {10.1175/jtech-d-25-0088.1}
}
Kostenlos starten →