Biases in Estimating Long‐Term Recurrence Intervals of Extreme Events Due To Regionalized Sampling.

Preparing for environmental risks requires estimating the frequencies of extreme events, often from data records that are too short to confirm them directly. This requires fitting a statistical distribution to the data. To improve precision, investigators often pool data from neighboring sites into single samples, referred to as "superstations," before fitting. We demonstrate that this technique can introduce unexpected biases in typical situations, using wind and rainfall extremes as case studies. When the combined locations have even small differences in the underlying statistics, the regionalization approach gives a fit that may tend toward the highest levels suggested by any of the individual sites. This bias may be large or small compared to the sampling error, for realistic record lengths, depending on the distribution of the quantity analyzed. The results of this analysis indicate that previous analyses could potentially have overestimated the likelihood of extreme events arising from natural weather variability. Plain Language Summary: We report a previously unknown bias in a common method for estimating how often extremely rare events such as extreme wind bursts or rain events will occur, when return periods are longer than the available data record. The method analyzed is one where an investigator combines data from nearby locations to reduce sampling error. We find by looking at new, high‐resolution data that variations in behavior across sites can sometimes produce biases much larger than the sampling error. The implication is that some observed extreme events are even less likely to have occurred than previously thought, assuming the underlying distribution hasn't changed over the period of observation. Key Points: Grouping data of nearby locations into one larger sample or "superstation" can induce biases at long recurrence intervalsThe superstation fit tends to the highest levels suggested by any of the pooled locationsThe bias may be large or small compared to random uncertainty, depending on the distribution of the extreme event analyzed [ABSTRACT FROM AUTHOR]