Evidence for intensifying rainfall extremes has not translated into “actionable” information needed by engineers and risk analysts, who are often concerned with very rare events such as “100‐year storms.” Low signal‐to‐noise associated with such events makes trend detection nearly impossible using conventional methods. We use a regional aggregation approach to boost this signal‐to‐noise, showing that such storms have increased in frequency over much of the conterminous United States since 1950, a period characterized by widespread hydrologic infrastructure development. Most of these increases can be attributed to secular climate change rather than climate variability, and we demonstrate potentially serious implications for the reliability of existing and planned hydrologic infrastructure and analyses. Though trends in rainfall extremes have not yet translated into observable increases in flood risks, these results nonetheless point to the need for prompt updating of hydrologic design standards, taking into consideration recent changes in extreme rainfall properties.

Impacts modelers and stakeholders use publicly available data sets of downscaled climate projections to assess and design infrastructure for changes in future rainfall extremes. If differences across data sets exist, infrastructure resilience decisions could change depending on which single data set is used. We assess changes in U.S. rainfall extremes from 2044-2099 compared with 1951-2005 based on 227 projections under RCP4.5 and RCP8.5 from five widely used data sets. We show there are large differences in the change magnitude and its spatial structure between data sets. At the continental scale, the data sets show different increases, with high-end extremes (e.g. 100-year event) generally increasing more (between 10% and 50%) than low-end extremes (e.g. 5-year). These differences largely contribute to the overall uncertainty for small average recurrence intervals (ARIs) extremes (2-to 10-year), while uncertainties due to short record length dominate large ARIs (25-to 100-year). The results indicate that robust infrastructure planning should consider these uncertainties to enable resilient infrastructure under climate change. Plain Language SummaryObserved extreme rainfall magnitudes have increased since 1950, and climate model projections indicate that these increases will continue throughout the 21st century in many areas in the US. Adapting and designing infrastructure for climate change requires future extreme rainfall projections at high resolutions. Global climate model (GCM) output resolution is generally much coarser, and methods have been developed to create relevant climate information at regional scales. Multiple open data sets exist that provide downscaled projections of future rainfall extremes. We analyze changes in future daily extremes using five widely used data sets in climate change impacts assessments and decision making. We found large differences between data sets in how much and where extreme events will intensify by the end of the 21st century. This and other sources of uncertainty need to be considered when designing resilient infrastructure.

No abstract

Designing infrastructure for a changing climate remains a major challenge for engineers. In popular discourse a narrative has emerged that infrastructures are likely undersigned for the future. Weather-related hazards are directly embedded in the infrastructure design process. Yet the codes and standards that engineers use for this risk analysis have been changing for decades, sometimes increasing and other times decreasing design values. Further complicating the issue is that climate projections show increasing or decreasing intensities depending on the hazard and region. Thus, it is not clear that infrastructure is universally underdesigned. Here, analyses are developed at both regional and national scales using precipitation and roadway drainage systems to answer this question. First, it is shown that modeling uncertainty can pose challenges for using future projections to update regionspecific standards. Second, the results show that depending on the historical design conditions and the direction of projections, roadway drainage infrastructure may be designed

Capsule:Updating extreme rainfall information in a changing climate is essential for communities and infrastructure and requires an inclusive, science-driven process.