Flash floods in arid regions represent a significant hydrological challenge due to their abrupt nature, spatial variability, and the scarcity of high-resolution data. We analyze twelve flash-flood events using a uniquely dense sub-basin network in the semi-arid to hyper-arid Ze’elim Basin (Israel). This study provides the first quantitative evaluation of broader assumptions about rainfall–runoff relationships at the sub-basin scale in a hyper-arid environment. The analysis shows that rainfall depth reliably predicts runoff depth, with relationships strengthening in larger sub-basins where local variability is being smoothed. In contrast to common expectations, peak discharge correlates more strongly with rain-core coverage (the spatial extent of intense rainfall) than with point-maximum rainfall intensity. These findings emphasize the value of intermediate-scale monitoring in arid basins and the use of spatial rainfall indices for capturing runoff responses and improving flash-flood assessment. More broadly, the results identify transferable predictors for runoff generation in data-scarce arid and semi-arid regions.