Octria Adi Prasojo, Anna van Yperen, Trevor B Hoey, Amanda Owen, Richard David Williams

Quantifying paleodischarge from geological field observations have been for decades, and remains, a key research challenge. Several paleodischarge scaling relationships have been developed for fluvial environments, such as BQART, Fulcrum and regional hydraulic geometry or for river deltas by precluding the role of wave and tide. In deltas where marine (wave, tide) energy causes bidirectional flow, the available paleodischarge scaling relationships are not applicable. Here, the spatial variability of distributary channel widths from a database of 114 global modern river deltas is assessed to understand the limit of marine influence on distributary channel widths. Compiling 4459 distributary channel width measurements enables improvements to distributary channel width-discharge scaling relationships specifically for river-, tide- and wave-dominated deltas. By bootstrapping the channel widths measured from modern deltas, the minimum number of width measurements needed to apply width-discharge scaling relationships to ancient deltaic deposits is estimated as 3 and 30 for upstream and downstream river-dominated deltas, consecutively, 6 for upstream part of tide-dominated deltas and 4 for wave-dominated deltas. This estimate will guide sedimentologists who often have limited numbers of distributary channel widths exposed in the rock record. Statistically significant width-discharge scaling relationships are derived for river- and wave-dominated deltas, with no significant relationships identified for tide-dominated deltas. To test the reliability of these improved width-discharge scaling relationships in the rock record, paleodischarges were estimated for the well-studied Cretaceous lower Mesa Rica Formation, USA. Comparison of these results with the more complex Fulcrum method suggests that these new scaling relationships are accurate. Hence these scaling relationships obtained from modern deltas can be applied to the rock record, and this approach requires less, and easier to measure, data inputs than previously published methods.