Next-Generation Rainfall IDF Curves for the Virginian Drainage Area of Chesapeake Bay

Xixi Wang | Old Dominion University



Infrastructure (e.g., port locks and storm sewers of naval bases) involving hydrologic flows are conventionally sized in terms of historic rainfall intensity-duration-frequency (IDF) curves with an assumption that climate is stationary. However, as climate has shown significant changes in rainfall characteristics in many regions (particularly in coastal regions such as Chesapeake Bay), the stationary assumption will likely become invalid and thus the adequacy of existing deterministic IDF curves may be questionable. That is, next-generation IDF curves reflecting “the non-stationarity at temporal and spatial scales relevant to improving future Department of Defense (DoD) infrastructure planning processes” is needed for DoD to construct and/or manage its hydrology-influenced infrastructure in changing climate in guarding against over- or under-committing resources.

The objectives of this project are to: 1) develop methodologies for creating IDF curves under non-stationary conditions using appropriate historic time series as base data, with investigation of appropriate down-scaling approaches for localized application; and 2) use these methodologies to create probability-based rainfall IDF curves for the Virginian drainage area of Chesapeake Bay, within which numerous DoD infrastructure have been, and/or will be, constructed.

Back to Top

Technical Approach

Besides National Elevation Dataset (NED) and National Hydrography Dataset (NHD), the observed 15-min rainfall time series at 57 rain gauge stations for the historic (i.e., pre-2013) period and projected 3-h 50-km precipitation time series by eleven pairs of General Circulation Model (GCM) and Regional Climate Model (RCM) for both the historic and future (i.e., 2039 ~ 2070) periods will be acquired and used. The observed time series will be used to downscale the projected 3-h 50-km to 3-h station-level precipitation using a statistical approach, which in turn will be further dis-aggregated into 15-min precipitation using a modified stochastic method. In addition, each of the observed and projected 15-min station-level time series will be aggregated into ten more datasets of 30 min, 1, 2, 3, 4, 6, 12, 24, 48, and 72 h durations. For a given rain gauge, this will result in: eleven time series of observed durational precipitation, eleven time series of projected pre-2013 durational precipitation, and eleven time series of projected future durational precipitation. Each of these thirty-three time series will in turn be used to generate a dataset of annual maximum durational precipitation. Subsequently, each of the datasets will be used for trend analysis, and it may be further subdivided into two or more subdatasets at years when significant step changes occur and/ or trends start. In terms of the datasets or subdatasets, empirical exceedance probabilities will be computed and then fitted by distributions of Fréchet, Weibull, and Gumbel to select a distribution with best goodness-of-fit. In turn, the selected best distribution will be used to create existing IDF curves, projected historic IDF curves, and projected future IDF curves, for return periods of 2, 5, 10, 50, and 100 year. For a given return period, the projected historic versus existing IDF curves will be compared to determine a range (i.e., lower and upper bounds) of bias correction factor, and this range will be taken as the lower and upper limits of the future IDF curve. The most-probable future IDF curve will be determined as the average of the eleven curves that correspond to the GCM-RCM models. Further, for a given duration and a given return period, the eleven values that correspond to the GCM-RCM models will be used to create a probability curve of the future IDF curve of this duration and this return period. Moreover, the NED and NHD will be used in ArcGIS® to delineate boundaries of watersheds, and define Thiessen polygons for estimating areal precipitation of a watershed. As with the time series for the individual rain gauge stations, the time series of areal precipitation will be used for trend and frequency analyses to generate watershedlevel future probability-based IDF curves.

Back to Top


This project will develop: 1) an innovative approach for creating next-generation IDF curves that take into account non-stationarity and uncertainty of rainfall; and 2) probability-based future IDF curves for the Virginian drainage area of Chesapeake Bay. The results will advance the understanding of the non-stationarity resulting from climate change, improving the planning, design, and management of DoD infrastructure. Such IDF curves can be a usable tool for increasing resilience of the infrastructure.

Back to Top

Points of Contact

Principal Investigator

Xixi Wang

Old Dominion University

Phone: 757-683-4882

Program Manager

Resource Conservation and Resiliency