Long term climate deviations: An alternative approach and application on the Palmer drought severity index in Hungary

Research output: Contribution to journalArticle

15 Citations (Scopus)

Abstract

A new statistical test (Makra-test), applicable for long time series, is introduced to identify extended sub-periods, namely, "breaks", average of which is significantly higher or lower than the mean of the entire time series. In order to apply this test, normal distribution of the time series being examined is a sufficient condition. In another case, if the number of elements increases in the time series, its distribution is near normal and the test can be applied, as well (according to the Central Limit Theorem). The method is demonstrated on monthly Palmer drought severity index (PDSI) data sets, computed for five stations of East Hungary in 1901-1999. Due to strongly recursive (auto-correlative) nature of PDSI every second month of the warm season (April, June, August and October) is analysed and treated as independent samples. Normality of the time series, which is a sufficient condition of the Makra-test, is validated by Kolmogorov-Smirnov test and X2-test. Analysis of the PDSI time series indicates that separate treatment of the months is important not only to ensure the normality, but also to consider the existing slight seasonal differences in standard deviation and skewness of the index in the East-Hungarian region. The Makra-test delimits one or more (maximum 4) significant sub-periods of the PDSI in every station and month (not considering the sub-intervals within the significant breaks, although most of them are also significant). Since the PDSI is based on monthly temperature and precipitation data that exhibit considerable inhomogeneities (Szentimry, 1999), the test is applied both for the original and the homogeneous time series. Effect of the inhomogeneity on long term variations of PDSI is strong: about the half of the significant breaks in the original time series disappear or become totally different from the time series based on homogenised data. All negative (dry) breaks of each month and station, however, occurred in more recent decades of the 20th century, according to both homogeneous and original series.

Original languageEnglish
Pages (from-to)1063-1071
Number of pages9
JournalPhysics and Chemistry of the Earth
Volume27
Issue number23-24
DOIs
Publication statusPublished - 2002

Fingerprint

Hungary
drought
Drought
climate
Time series
time series
deviation
normality
stations
inhomogeneity
Kolmogorov-Smirnov test
index
test
statistical tests
skewness
Statistical tests
Precipitation (meteorology)
Normal distribution
normal density functions
standard deviation

ASJC Scopus subject areas

  • Geochemistry and Petrology
  • Geophysics

Cite this

@article{74b267cf053740d983cbb38c033d66e4,
title = "Long term climate deviations: An alternative approach and application on the Palmer drought severity index in Hungary",
abstract = "A new statistical test (Makra-test), applicable for long time series, is introduced to identify extended sub-periods, namely, {"}breaks{"}, average of which is significantly higher or lower than the mean of the entire time series. In order to apply this test, normal distribution of the time series being examined is a sufficient condition. In another case, if the number of elements increases in the time series, its distribution is near normal and the test can be applied, as well (according to the Central Limit Theorem). The method is demonstrated on monthly Palmer drought severity index (PDSI) data sets, computed for five stations of East Hungary in 1901-1999. Due to strongly recursive (auto-correlative) nature of PDSI every second month of the warm season (April, June, August and October) is analysed and treated as independent samples. Normality of the time series, which is a sufficient condition of the Makra-test, is validated by Kolmogorov-Smirnov test and X2-test. Analysis of the PDSI time series indicates that separate treatment of the months is important not only to ensure the normality, but also to consider the existing slight seasonal differences in standard deviation and skewness of the index in the East-Hungarian region. The Makra-test delimits one or more (maximum 4) significant sub-periods of the PDSI in every station and month (not considering the sub-intervals within the significant breaks, although most of them are also significant). Since the PDSI is based on monthly temperature and precipitation data that exhibit considerable inhomogeneities (Szentimry, 1999), the test is applied both for the original and the homogeneous time series. Effect of the inhomogeneity on long term variations of PDSI is strong: about the half of the significant breaks in the original time series disappear or become totally different from the time series based on homogenised data. All negative (dry) breaks of each month and station, however, occurred in more recent decades of the 20th century, according to both homogeneous and original series.",
author = "L. Makra and Sz Horv{\'a}th and R. Pongr{\'a}cz and J. Mika",
year = "2002",
doi = "10.1016/S1474-7065(02)00142-0",
language = "English",
volume = "27",
pages = "1063--1071",
journal = "Physics and Chemistry of the Earth",
issn = "1474-7065",
publisher = "Elsevier Limited",
number = "23-24",

}

TY - JOUR

T1 - Long term climate deviations

T2 - An alternative approach and application on the Palmer drought severity index in Hungary

AU - Makra, L.

AU - Horváth, Sz

AU - Pongrácz, R.

AU - Mika, J.

PY - 2002

Y1 - 2002

N2 - A new statistical test (Makra-test), applicable for long time series, is introduced to identify extended sub-periods, namely, "breaks", average of which is significantly higher or lower than the mean of the entire time series. In order to apply this test, normal distribution of the time series being examined is a sufficient condition. In another case, if the number of elements increases in the time series, its distribution is near normal and the test can be applied, as well (according to the Central Limit Theorem). The method is demonstrated on monthly Palmer drought severity index (PDSI) data sets, computed for five stations of East Hungary in 1901-1999. Due to strongly recursive (auto-correlative) nature of PDSI every second month of the warm season (April, June, August and October) is analysed and treated as independent samples. Normality of the time series, which is a sufficient condition of the Makra-test, is validated by Kolmogorov-Smirnov test and X2-test. Analysis of the PDSI time series indicates that separate treatment of the months is important not only to ensure the normality, but also to consider the existing slight seasonal differences in standard deviation and skewness of the index in the East-Hungarian region. The Makra-test delimits one or more (maximum 4) significant sub-periods of the PDSI in every station and month (not considering the sub-intervals within the significant breaks, although most of them are also significant). Since the PDSI is based on monthly temperature and precipitation data that exhibit considerable inhomogeneities (Szentimry, 1999), the test is applied both for the original and the homogeneous time series. Effect of the inhomogeneity on long term variations of PDSI is strong: about the half of the significant breaks in the original time series disappear or become totally different from the time series based on homogenised data. All negative (dry) breaks of each month and station, however, occurred in more recent decades of the 20th century, according to both homogeneous and original series.

AB - A new statistical test (Makra-test), applicable for long time series, is introduced to identify extended sub-periods, namely, "breaks", average of which is significantly higher or lower than the mean of the entire time series. In order to apply this test, normal distribution of the time series being examined is a sufficient condition. In another case, if the number of elements increases in the time series, its distribution is near normal and the test can be applied, as well (according to the Central Limit Theorem). The method is demonstrated on monthly Palmer drought severity index (PDSI) data sets, computed for five stations of East Hungary in 1901-1999. Due to strongly recursive (auto-correlative) nature of PDSI every second month of the warm season (April, June, August and October) is analysed and treated as independent samples. Normality of the time series, which is a sufficient condition of the Makra-test, is validated by Kolmogorov-Smirnov test and X2-test. Analysis of the PDSI time series indicates that separate treatment of the months is important not only to ensure the normality, but also to consider the existing slight seasonal differences in standard deviation and skewness of the index in the East-Hungarian region. The Makra-test delimits one or more (maximum 4) significant sub-periods of the PDSI in every station and month (not considering the sub-intervals within the significant breaks, although most of them are also significant). Since the PDSI is based on monthly temperature and precipitation data that exhibit considerable inhomogeneities (Szentimry, 1999), the test is applied both for the original and the homogeneous time series. Effect of the inhomogeneity on long term variations of PDSI is strong: about the half of the significant breaks in the original time series disappear or become totally different from the time series based on homogenised data. All negative (dry) breaks of each month and station, however, occurred in more recent decades of the 20th century, according to both homogeneous and original series.

UR - http://www.scopus.com/inward/record.url?scp=0036430134&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0036430134&partnerID=8YFLogxK

U2 - 10.1016/S1474-7065(02)00142-0

DO - 10.1016/S1474-7065(02)00142-0

M3 - Article

AN - SCOPUS:0036430134

VL - 27

SP - 1063

EP - 1071

JO - Physics and Chemistry of the Earth

JF - Physics and Chemistry of the Earth

SN - 1474-7065

IS - 23-24

ER -