Peaks-over-Threshold (POT) method allows for the consideration of multiple flood events in a single year, offering several advantages over the annual maximum (AM) series, which exclusively retains a single maximum value per year. These advantages encompass increased data availability and the preservation of intra-year variations in the timing and magnitude of flood events. However, the POT method inherently lacks clarity in terms of ensuring the independence of events, often leading to its underutilization. In this research, we aim to address this limitation by employing various independence-securing algorithms. Specifically, we apply different POT methods to daily mean flow data from Hydro-Climatic Data Network (HCDN) sites, which are minimally impacted by human activity, over the coterminous United States (CONUS). With the generated POT series, we proceed to assess the spatial dependence of flood occurrences across various regions by quantifying the co-occurrence of floods at different locations. Subsequently, we compare the outcomes of these different POT methods and discuss the significance of selecting proper POT methods.
