ABSTRACT:

The dataset includes daily, weekly, monthly, quarterly, and yearly streamflow records with fractal and chaos characteristics for 2,899 USGS gauging stations in the U.S. and Puerto Rico (1970–2023) for three flow regimes: highest, average, and lowest. The series are available in various formats, including versions with incorrect or missing data that have been interpolated and filled, as well as the raw format. he available fractal and chaos metrics include Hurst exponents (calculated using rescaled range and detrended fluctuation analysis), multifractality dimension, entropy, Morlet wavelets (which consist of max, mean, and L2-norm modulus across three bands: low, mid, and high), Lyapunov exponents, and RQA analysis (which covers recurrence rate, determinism, laminarity, and trapping time). Fuzzy C-means clustering categorized the gauges into three dynamic classes with membership probabilities included. The dataset includes raw and processed series, dynamics metrics, and cluster probabilities matrix. The dataset supports hydrological modeling, regional classification, benchmarking, and machine learning applications. A visual map-based proxy is available.