Assessing and managing design storm variability and projection uncertainty in a changing coastal environment.

Abstract Coastal urban infrastructure and water management programs are vulnerable to the impacts of long-term hydroclimatic changes and to the flooding and physical destruction of disruptive hurricanes and storm surge. Water resilience or, inversely, vulnerability depends on design specifications of the storm and inundation, against which water infrastructure and environmental assets are planned and operated. These design attributes are commonly derived from statistical modeling of historical measurements. Here we argue for the need to carefully examine the approach and associated design vulnerability in coastal areas because of the future hydroclimatic changes and large variability at local coastal watersheds. This study first shows significant spatiotemporal variations of design storm in the Chesapeake Bay of the eastern U.S. Atlantic coast, where the low-frequency high-intensity precipitations vary differently to the tropical cyclones and local orographic effects. Average and gust wind speed exhibited much greater spatial but far less temporal variability than the precipitation. It is noteworthy that these local variabilities are not fully described by the regional gridded precipitation used in CMIP5 climate downscaling and by NOAA's regional design guide Atlas-14. Up to 46.4% error in the gridded precipitation for the calibration period 1950–1999 is further exacerbated in the future design values by the ensemble of 132 CMIP5 projections. The total model projection error ( δ M ) up to −61.8% primarily comes from the precipitation regionalization ( δ 1 ), climate downscaling ( δ 2 ), and a fraction from empirical data modeling ( δ E ). Thus, a post-bias correction technique is necessary. The bias-corrected design wind speed for 10-yr to 30-yr storms has small changes