An Evaluation of NOAA Atlas 14 for Extreme Rainstorms in Colorado and the United States

In the past decade, there has been an increase in the number of reported 1,000-year storms in the United States (e.g., 2013 Front Range of CO; 2016 Harvey in Houston, TX, and two 1,000-year storms in 2017 and 2018 in Ellicott City, MD; 2018 in Flagstaff, AZ). Certainly, some 1,000-year storms are to be expected each year in the United States, but the dramatic increase in the rate of such reports is questionable. This apparent increase in more 1,000-year storms has been attributed, in part, to the effects of global climate change, and also the availability of NOAA Atlas 14 (https://hdsc.nws.noaa.gov/hdsc/pfds/pfds_map_cont.html) in about the past ten years. This presentation first provides a systematic assessment of extreme flood-producing rainstorm in Colorado using NOAA Atlas 14. NOAA Atlas 14 was derived from an analysis of point-rainfall data obtained at precipitation gages and is used to estimate the frequency of rainstorms for various durations. In September 2013, up to 510 mm of rainfall over seven days (but most rain occurred in four days) produced record flooding over much of the Colorado Front Range. According to NOAA Atlas 14, the rainfall recurrence interval at many locations was at least 1,000 years. The September 2013 rainstorm produced widespread flooding with maximum flood recurrence intervals of about 700 years in the areas of maximum rainfall. In addition to the 2013 storm, many such rainfall amounts have occurred in eastern Colorado, which motivated this comprehensive analysis of the frequency of extreme rainstorms.

The evaluation of NOAA Atlas 14 was first used to estimate point rainfall frequency values for extreme storms compiled in Eastern Colorado through 1997 (McKee and Doesken, 1997) with the September 2013 Storm added. Most of these extreme storm data were from rainfall bucket surveys at miscellaneous (or ungaged) sites. The analysis indicates more than a dozen storms exceeded not only a 1,000-year recurrence interval, but many storms exceeded 10,000 to 100,000-year recurrence intervals. The 2013 storm of 510 mm has about a 100,000-year recurrence interval. For comparison, the July 1976 Big Thompson Canyon storm, which produced Colorado’s deadliest flood, had a point rainfall of about 355 mm in 4 hours with an associated recurrence interval far exceeding one million years. The envelope curve of maximum rainfall for various storm durations in eastern Colorado has a recurrence interval of about 500 million years. These results are implausible and strongly suggests NOAA Atlas 14 overestimates storm frequency for “rare events” in eastern Colorado. The evaluation of NOAA Atlas 14 was then expanded to include selected record flood-producing rainstorms in the United States (O’Connor and Costa, 2004; Costa and Jarrett, 2008), which had similar results as for eastern Colorado. The most likely explanation for these excessive rainfall frequencies is because NOAA Atlas 14 is only based on analysis of rainfall data at gaged sites. Improvements to NOAA Atlas 14 would benefit from use of the availability of over a thousand rainfall bucket data at ungaged sites compiled for the United States (unpublished data, U.S. Bureau of Reclamation, Denver, Colorado), as well as validated NEXRAD rainfall estimates. Additional comprehensive assessments of NOAA Atlas 14 for estimating infrequent rainstorms for possible NOAA Atlas 14 revision appears warranted, particularly in areas where the most extreme rainfall data are comprised of bucket data obtained at miscellaneous sites (e.g., rural areas where gage records are sparse).