Abstract While extreme rainfall is projected to intensify with rising global temperatures, natural variability can obscure the detectability of long‐term changes at local scales. Given current extreme rainfall assessments largely rely on historical observations, the relative severity of historical events can influence perceived present‐day risk. With a focus on New Zealand, here we use 3,226 years of high‐resolution (∼50 km) model simulations of the recent past climate to quantify uncertainty in perceived extreme rainfall risk due to internal variability. By repeatedly sampling 100‐year periods of initial‐condition model simulations and identifying the most severe rainfall events to occur at each grid cell, we examine the rarity and intensity of synthetic “worst‐in‐century” (WIC) events as a proxy for the worst rainfall event in “living memory.” Depending on the period, WIC intensity ranges from 200% of the “true” 1‐in‐100‐year rainfall event estimated from the full data set, with corresponding WIC return periods spanning 25 to >5,000 years. The co‐occurrence of significant events nationwide within the same century can also differ substantially. These findings highlight the consequences of sampling uncertainty when inferring the potential risks of record‐shattering rainfall from historical experiences alone, presenting challenges for flood management and resilience planning in a warming climate.
Sigid et al. (Fri,) studied this question.