Long-range properties and data validity for hydrogeological time series: the case of the Paglia river

TL;DR

This study analyzes 20 years of Paglia river streamflow data, revealing long-term persistence, periodicities linked to oceanic phenomena, antipersistent behavior, and data validity issues, using DFA and Benford's law assessments.

Contribution

It applies detrended fluctuation analysis to a large hydrogeological dataset and explores data validity through Benford's law, providing new insights into long-range properties and complex system behavior.

Findings

River streamflow shows periodicities linked to oceanic phenomena.

Data exhibits antipersistent behavior.

Measurement data does not follow Benford's law.

Abstract

This paper explores a large collection of about 377,000 observations, spanning more than 20 years with a frequency of 30 minutes, of the streamflow of the Paglia river, in central Italy. We analyze the long-term persistence properties of the series by computing the Hurst exponent, not only in its original form but also under an evolutionary point of view by analyzing the Hurst exponents over a rolling windows basis. The methodological tool adopted for the persistence is the detrended fluctuation analysis (DFA), which is classically known as suitable for our purpose. As an ancillary exploration, we implement a control on the data validity by assessing if the data exhibit the regularity stated by Benford's law. Results are interesting under different viewpoints. First, we show that the Paglia river streamflow exhibits periodicities which broadly suggest the existence of some common behaviour with El Nino and the North Atlantic Oscillations: this specifically points to a (not necessarily direct) effect of these oceanic phenomena on the hydrogeological equilibria of very far geographical zones: however, such an hypothesis needs further analyses to be validated. Second, the series of streamflows shows an antipersistent behaviour. Third, data are not consistent with Benford's law: this suggests that the measurement criteria should be opportunely revised. Fourth, the streamflow distribution is well approximated by a discrete generalized Beta distribution: this is well in accordance with the measured streamflows being the outcome of a complex system.