Development and study of a parallel algorithm of iteratively forming latent functionally-determined structures for classification and analysis of meteorological data

Vasilii A. Sorokin, Yu V. Volkov, A. I. Sherstneva, I. A. Botygin

Research output: Contribution to journalArticle

Abstract

This paper overviews a method of generating climate regions based on an analytic signal theory. When applied to atmospheric surface layer temperature data sets, the method allows forming climatic structures with the corresponding changes in the temperature to make conclusions on the uniformity of climate in an area and to trace the climate changes in time by analyzing the type group shifts. The algorithm is based on the fact that the frequency spectrum of the thermal oscillation process is narrow-banded and has only one mode for most weather stations. This allows using the analytic signal theory, causality conditions and introducing an oscillation phase. The annual component of the phase, being a linear function, was removed by the least squares method. The remaining phase fluctuations allow consistent studying of their coordinated behavior and timing, using the Pearson correlation coefficient for dependence evaluation. This study includes program experiments to evaluate the calculation efficiency in the phase grouping task. The paper also overviews some single-threaded and multi-threaded computing models. It is shown that the phase grouping algorithm for meteorological data can be parallelized and that a multi-threaded implementation leads to a 25-30% increase in the performance.

Original languageEnglish
Article number012012
JournalIOP Conference Series: Earth and Environmental Science
Volume48
Issue number1
DOIs
Publication statusPublished - 26 Nov 2016

Fingerprint

oscillation
climate
least squares method
weather station
surface layer
temperature
climate change
experiment
analysis
method
programme
calculation
evaluation

ASJC Scopus subject areas

  • Environmental Science(all)
  • Earth and Planetary Sciences(all)

Cite this

@article{b68352c4110743f1b4e65a7816e1167d,
title = "Development and study of a parallel algorithm of iteratively forming latent functionally-determined structures for classification and analysis of meteorological data",
abstract = "This paper overviews a method of generating climate regions based on an analytic signal theory. When applied to atmospheric surface layer temperature data sets, the method allows forming climatic structures with the corresponding changes in the temperature to make conclusions on the uniformity of climate in an area and to trace the climate changes in time by analyzing the type group shifts. The algorithm is based on the fact that the frequency spectrum of the thermal oscillation process is narrow-banded and has only one mode for most weather stations. This allows using the analytic signal theory, causality conditions and introducing an oscillation phase. The annual component of the phase, being a linear function, was removed by the least squares method. The remaining phase fluctuations allow consistent studying of their coordinated behavior and timing, using the Pearson correlation coefficient for dependence evaluation. This study includes program experiments to evaluate the calculation efficiency in the phase grouping task. The paper also overviews some single-threaded and multi-threaded computing models. It is shown that the phase grouping algorithm for meteorological data can be parallelized and that a multi-threaded implementation leads to a 25-30{\%} increase in the performance.",
author = "Sorokin, {Vasilii A.} and Volkov, {Yu V.} and Sherstneva, {A. I.} and Botygin, {I. A.}",
year = "2016",
month = "11",
day = "26",
doi = "10.1088/1755-1315/48/1/012012",
language = "English",
volume = "48",
journal = "IOP Conference Series: Earth and Environmental Science",
issn = "1755-1307",
publisher = "IOP Publishing Ltd.",
number = "1",

}

TY - JOUR

T1 - Development and study of a parallel algorithm of iteratively forming latent functionally-determined structures for classification and analysis of meteorological data

AU - Sorokin, Vasilii A.

AU - Volkov, Yu V.

AU - Sherstneva, A. I.

AU - Botygin, I. A.

PY - 2016/11/26

Y1 - 2016/11/26

N2 - This paper overviews a method of generating climate regions based on an analytic signal theory. When applied to atmospheric surface layer temperature data sets, the method allows forming climatic structures with the corresponding changes in the temperature to make conclusions on the uniformity of climate in an area and to trace the climate changes in time by analyzing the type group shifts. The algorithm is based on the fact that the frequency spectrum of the thermal oscillation process is narrow-banded and has only one mode for most weather stations. This allows using the analytic signal theory, causality conditions and introducing an oscillation phase. The annual component of the phase, being a linear function, was removed by the least squares method. The remaining phase fluctuations allow consistent studying of their coordinated behavior and timing, using the Pearson correlation coefficient for dependence evaluation. This study includes program experiments to evaluate the calculation efficiency in the phase grouping task. The paper also overviews some single-threaded and multi-threaded computing models. It is shown that the phase grouping algorithm for meteorological data can be parallelized and that a multi-threaded implementation leads to a 25-30% increase in the performance.

AB - This paper overviews a method of generating climate regions based on an analytic signal theory. When applied to atmospheric surface layer temperature data sets, the method allows forming climatic structures with the corresponding changes in the temperature to make conclusions on the uniformity of climate in an area and to trace the climate changes in time by analyzing the type group shifts. The algorithm is based on the fact that the frequency spectrum of the thermal oscillation process is narrow-banded and has only one mode for most weather stations. This allows using the analytic signal theory, causality conditions and introducing an oscillation phase. The annual component of the phase, being a linear function, was removed by the least squares method. The remaining phase fluctuations allow consistent studying of their coordinated behavior and timing, using the Pearson correlation coefficient for dependence evaluation. This study includes program experiments to evaluate the calculation efficiency in the phase grouping task. The paper also overviews some single-threaded and multi-threaded computing models. It is shown that the phase grouping algorithm for meteorological data can be parallelized and that a multi-threaded implementation leads to a 25-30% increase in the performance.

UR - http://www.scopus.com/inward/record.url?scp=85009383161&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85009383161&partnerID=8YFLogxK

U2 - 10.1088/1755-1315/48/1/012012

DO - 10.1088/1755-1315/48/1/012012

M3 - Article

AN - SCOPUS:85009383161

VL - 48

JO - IOP Conference Series: Earth and Environmental Science

JF - IOP Conference Series: Earth and Environmental Science

SN - 1755-1307

IS - 1

M1 - 012012

ER -