Detecting persistence of meteorological drought via the Hurst exponent

Abstract

This study proposes the Hurst exponent (H) to detect persistence in the Palmer Drought Severity Index (PDSI) over Turkey. Since a fractal structure admits the behaviour of global determinism and local randomness, the H exponent values could be used to detect self-similar statistical structure of the time series. Additionally, the H value >0.5 and near 1 shows the intensity level of persistence. The term persistence may be assessed as a criterion to be used as a measure of predictability. Moreover, the predictability index, fractal dimension and autocorrelation function of the PDSI values can also be obtained from the H values. Nevertheless, it is not easy to find a spatial meaning of information content of the results obtained by the methods used in the present study. For that reason, the Mann-Kendall (MK) trend test was also used, and the results showed that significant negative trends are widespread throughout the country. As expected, the H values were close to 1 in places where statistically significant trends exist. The high values of the H exponent in all the regions can be explained by the droughts observed there that might be caused due to the possible associations with the large-scale atmospheric circulations. In this context, it is suggested that the droughts can be predictable by constructing the appropriate general climate circulation models.

Drought Persistence Errors in Global Climate Models

“…Measures of persistence (or memory) of drought events and related climate phenomena have been used to analyze drought variability in both models and observations. Commonly used measures include power spectra of the time series (Ault et al, 2014;Pelletier & Turcotte, 1997) or the Hurst exponent (Mandelbrot & Wallis, 1969) which have been used, for example, to measure long-term persistence in time series of drought indices (Tatli, 2015) or precipitation (Bunde et al, 2013;Kumar et al, 2013). By transforming continuous time series of drought indicators into categorical time series of dry and wet spells, the persistence of drought can be effectively estimated through the parameters of a binary first-order Markov chain model (Sericola, 2013).…”

Multifractal analysis of the drought area in seven large regions of China from 1961 to 2012

“…Recently, the Hurst exponent has been used to detect the complexity and persistence in meteorological or climatic variables (Tatli 2014(Tatli , 2015. The advantages of this method are that it finds the long-term correlations of non-stationary time series.…”

Spatio‐temporal drought patterns of multiple drought indices based on precipitation and soil moisture: A case study in South Korea

“…In addition to the above drought features, an understanding of drought persistence, which has an impact on water resources management, is also of great interest for hydrologists (AghaKouchak, ; Ganguli and Ganguly, ; Meng et al , ). Drought persistence can be computed by the length of a dry spell for a certain threshold or by using temporal autocorrelations (Tatli, ).…”

“…In addition to the above drought features, an understanding of drought persistence, which has an impact on water resources management, is also of great interest for hydrologists (AghaKouchak, 2015;Ganguli and Ganguly, 2016;Meng et al, 2017). Drought persistence can be computed by the length of a dry spell for a certain threshold or by using temporal autocorrelations (Tatli, 2015). We further explore a monthly variation of the correlation coefficient between the SPI-3 and SSI-3 for the 55 weather stations, and the lagged relationships between the drought indices are additionally examined to capture any possible delayed response.…”