Complexity and chaotic behavior of the U.S. rivers and estimation of their prediction horizon

TL;DR

This study analyzes the complexity and chaotic dynamics of US rivers' streamflow data using information-theoretic and chaos theory measures to estimate their predictability horizons.

Contribution

It applies Kolmogorov complexity and Lyapunov exponent analyses to a large dataset of US rivers, providing new insights into their predictability based on natural and human factors.

Findings

Streamflow predictability varies with environmental and human influences.

Chaotic behavior influences the length of the rivers' prediction horizons.

Quantitative estimates of predictability horizons for 1879 US rivers.

Abstract

A streamflow time series encompasses a large amount of hidden information and reliable prediction of its behavior in the future remains a challenge. It seems that the use of information measures can significantly contribute to determining the time horizon of rivers and improving predictability. Using the Kolmogorov complexity (KC) and its derivatives (KC spectrum and its highest value), and Lyapunov exponent (LE), it has previously been shown that the degree of streamflow predictability depends on human activities, environmental factors, and natural characteristics. This paper applied the KC and LE measures to investigate the randomness and chaotic behavior of monthly streamflow of 1879 rivers from the United States for a period from 1950 to 2015 and evaluated their time horizons via the Lyapunov and Kolmogorov time (LT and KT, respectively).