Simulating sub-daily rainfall time series in the absence of sub-daily observations

This paper presents a novel framework for simulating sub-daily rainfall time series in the absence of sub-daily observations, which are typically essential for calibrating conventional approaches. The proposed approach combines two classes of models: a daily rainfall model that generates long synthetic daily time series and a disaggregation model based on multifractal theory to refine the temporal resolution to sub-daily scales. The implemented procedure is parsimonious and relies solely on the observed daily rainfall time series and the power law exponent n of the intensity–duration–frequency curves, information widely available to practitioners.The framework was tested on a challenging case study consisting of 70 rain gauges in the Arno River basin (Italy), each with 20 years of continuous 15-minute rainfall data. The performance was evaluated by comparing key statistical attributes estimated from the benchmark dataset and the simulated rainfall time series at 15-minute temporal resolution, such as the dry frequency, the autocorrelation at lags 1 and 10, and the dependence of rainfall intensity on the duration of spatial averaging and return period as embodied in the well-established notion of intensity–duration–frequency (IDF) curves.The results show promising agreement, despite the limited sample size, which introduces some calibration challenges. The relative errors of the selected attributes fall within ± 15% for most of the analyzed time series, indicating that the framework offers a valuable alternative for hydrological studies, particularly in contexts where sub-daily observations are scarce or entirely absent.