A Historical Perspective on U.S. Weather and Climate Extremes
WHAT HISTORY TELLS US ABOUT U.S. DAILY RAINFALL EXTREMES - METHODSBased on Hoerling et al. & Wolter et al. (2016), we utilize 987 meteorological stations in the contiguous US extracted from GHCN-D (Menne et al. 2012) with at least 100 years of non-missing daily observations during 1901–2014, as well as mostly complete data in 2015. The “RX1day” index (maximum 1-day precipitation), as defined by Sillmann et al. (2013), is computed at each station for all annual and bimonthly cases (base period 1901–80). We applied the generalized extreme value (GEV) distribution, known as the block or annual maxima approach for analysis of 20-yr precipitation events, using the Matlab NEVA package (Cheng et al. 2014). The lower confidence bounds (2.5th percentile) of the GEV-estimated return level for 20-yr events are applied in order to include all cases that might be considered of that intensity. We validated these results against the empirical estimates of the 20-yr events by ranking the annual and seasonal maxima at each station.
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for September-October 2017
DiscussionSeptember-October 2017 had the 7th highest national count of 20-yr daily extreme precipitation events since 1901 (17.5%). Combined with an even higher total in 2016 (18.3%), this tally increased the overall trend from 3.5% per century through 2015 (p=99.9%) to 4.0% per century through 2017 (p=99.99%). The national count was anchored by diverse regional totals with 33.3% in the Southeast (4th highest), 31.6% in the Pacific Northwest (tied for 4th place), and 23.4% in the Northeast (only 15th highest). Coming on the heels of 22.3% in 2016, the Southeast now shows a significant upward trend through 2017 (+5.2% per century; p=96%), compared to an insignificant upward trend through 2015 (p=87%). On the other hand, the Pacific Northwest is still not showing a significant upward trend, despite a record-breaking total of 45.1% in 2016 followed by this year's 31.6%. Third, the Northeast region count confirms a strong upward trend that was already +9.3% per century through 2015 (p=99.5%), only to rise to +10.2% per century through 2017 ((p=99.8%) after recording 28.8% in 2016 as well. In contrast, the West only reported 2.9%, not nearly as bad as those 30 prior cases with only 0.0% (since 1901), but a far cry from 36.6% in 2016. There were at least two "Billion-Dollar Disasters" ( https://www.ncdc.noaa.gov/billions/events) in this season: Hurricanes Irma and Maria, but this tally does not include the California wildfires of October just yet. According to Munich Reinsurance ( https://www.munichre.com/en/media-relations/publications/press-releases/2018/2018-01-04-press-release), Hurricane Irma in early September caused at least 128 fatalities and $67B in damages, but that includes Caribbean losses as well. It contributed significantly to the high Southeast 20-yr extremes total. Similarly, Hurricane Maria in late September caused at least 108 deaths and $63B in damages, with both counts still increasing as Puerto Rico is not back to 'normal' by a long shot.
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for July-August 2017
DiscussionJuly-August 2017 registered 13.5% of all stations with 20-yr daily extreme precipitation events (5th highest since 1901), cementing a significant upward trend for the lower 48 states through 2017 (+2.5% per century, p=99.99%). There were two regions with total counts in the highest decile: the South (22.0%; 3rd highest) and the Ohio Valley (16.2%; 11th highest). While the South has not shown a significant trend yet, the Ohio Valley had seen a significant upward trend through 2016 with +3.1% per century (p=99.2%). In contrast, the Pacific Northwest recorded not a single 20-yr daily extreme, tied for lowest rank with 14 other years, but not in support or contrast with any significant trends. There was only one "Billion-Dollar Disaster" ( https://www.ncdc.noaa.gov/billions/events) during this season, but that one was huge: Hurricane Harvey dropped record amounts of rain on Texas in late August, leading to the highest number of multiple 20-yr daily extreme counts for the year: 23. The total number of fatalities is at least 84 (see link above), but may have been higher still (Munich Reinsurance attributes 88 deaths to Harvey: https://www.munichre.com/en/media-relations/publications/press-releases/2018/2018-01-04-press-release). The same source gives an $85 Billion loss estimate that makes Hurricane Harvey the biggest global economic disaster in 2017, but also the biggest US disaster since at least Katrina (2005).
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for May-June 2017
DiscussionMay-June 2017 had a national count of 9.5% of all available stations, well below the median of more than 13% during the preceding decade, associated with a highly significant upward trend since 1901 (p=99.99%). The highest regional totals were 17.3% in the Ohio Valley (19th highest) and 14.3% in the Northeast (27th highest). The Ohio Valley tally is supported by a strong upward trend of more than 5% per century (p=99.9%), while the upward trend in the Northeast fails to reach the 90% significance level. In contrast, the Southwest confirmed its almost 95%-significant DOWNWARD trend with a measly total of 2.9% (23rd lowest). Two other regions came in under 5%: the Northern Rockies (3.5%, tied for 7th lowest) and the Pacific Northwest (3.9%, 32nd lowest). The dry spring in the northwestern portion of the US set it up for an active fire season. Four "Billion-Dollar Disasters" ( https://www.ncdc.noaa.gov/billions/events) include an extreme hailstorm in Denver (May 10th) that caused more than $2B in damages, but no fatalities, Midwestern severe weather (two events in June, also no fatalities), and flooding in Missouri and Arkansas that took 20 lives without leaving much of a footprint in regional extreme tallies.
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for March-April 2017
DiscussionMarch-April 2017 had the 3rd highest count of 20-yr daily precipitation events on record (16.5%). This confirms a highly significant upward trend of +3% per century (p=99.99%) that was already in place through 2016. This outcome was anchored by upper decile regional totals in the South (26.6% = 4th), Ohio Valley (25.7% = 6th), and Pacific Northwest (19.6% = 11th). Two extreme March-April seasons in a row in the South (2016 was highest on record) anchor a statistically significant upward trend through 2017 (p=98%), while the upward trend through 2015 was only "significant" at p=85%. Both the Ohio Valley and the Pacific Northwest have not shown significant changes in their extreme tallies for this season since 1901. Four "Billion-Dollar Disasters" ( https://www.ncdc.noaa.gov/billions/events ) focused on the South and Ohio Valley, causing 8 fatalities. Three of these events were associated with severe weather, while the fourth was due to a hard freeze after unusually warm weather preceding it (in the Southeast).
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for January-February 2017
DiscussionJanuary-February 2017 had the 21st highest national count of 20-yr daily precipitation events since 1901 (12.9%). This total was anchored by three regional totals at or above 20%: 25.5% in the Pacific Northwest (4th highest), 23.0% in the South (11th), and 20.0% in the Southwest (tied for 11th). Only the South showed a marginally significant upward trend preceding 2017 (p=93%), while the other two regions show little trend. There were two "Billion-Dollar Disasters" ( https://www.ncdc.noaa.gov/billions/events) during this season. The first one was a tornado outbreak in the South that killed 24 people, also causing some of the high precipitation totals in that region. The second one was due to flooding rains in California, especially in February when the dam at Lake Oroville was almost breached and the city of San Jose suffered extensive flooding that also took 5 lives. While the West region ended up with 16.7% coverage of 20-yr daily extremes, this was "only" tied for 20th highest place.
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for September-October 2016
DiscussionSeptember-October 2016 had the 4th highest national count of 20-yr daily precipitation events since 1901 (18.3% of all available stations). This count was boosted by regional totals in the Northwest (45.1%, highest on record)) and West (36.6%. 4th highest). Both of these regions show little trend over the last 116 years in all seasons of the calendar year, not just in September-October. Hurricane Matthew skirted the southeastern seaboard around the 10th of October, leading to flooding rains that caused 49 fatalities and around $10B in damages (Southeast coverage: 22.3%, 12th highest since 1901). Source: https://www.ncdc.noaa.gov/billions/events
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for July-August 2016
DiscussionJuly-August 2016 had the 6th highest count of 20-yr events since 1901 (13.1% of all available stations), highest in the Midwest climate region (Fig. 2) where 27.4% of the stations were flagged in this manner, 2nd only to 2007. The Ohio valley tallied 19.3% of its stations with such daily extremes, 3rd highest on record. These two regions are the only two to show significant upward trends in this season (5.3% and 3.5% per century, respectively). In the South climate region (Fig. 2), flooding rains around August 12-15 caused 13 deaths in Louisiana along with estimated damages around $10B. Source: https://www.ncdc.noaa.gov/billions/events
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for March-April 2016
DiscussionMarch-April 2016 tallied the highest national count on record for this season’s 20-yr events since 1901 (20.0% of all available stations). This was anchored by record totals in the Northern Rockies (33.6%; mainly in Nebraska) and South (41.3%). These two regions show significant upward trends in this season (just over 4% per century each), although the upward trend in the South did not even reach the 90% significance threshold before 2015. In contrast, the Southwest has the only significant downward trend during this season (-4.9% per century) in the country. Texas and Louisiana suffered through multiple flooding events, most notably in the Sabine River basin on March 12th (5 fatalities; $2.3B) and around Houston on April 17-18 ($2.7; 8 fatalities). Source: https://www.ncdc.noaa.gov/billions/events
Figure 1: Bimonthly daily extremes in excess of the lower estimate for 20-yr events for November-December 2015
DiscussionNovember-December 2015 saw more 20-yr events in this season than in any other year on record (29.0%). These extreme daily totals were mostly associated with a powerful storm system that swept northeastward from Texas through the Upper Midwest in late December, causing at least 50 deaths and $2B in damages. Regional record counts for the season were set in the Midwest (50.4%), Ohio Valley (39.9%), and South (36.6%). It is noteworthy that these regional records were NOT preceded by significant upward trends over the last century. Source: https://www.ncdc.noaa.gov/billions/events
U.S. Climate Regions
Click on a region to display data for that region, or here for Continental U.S.
Continental U.S. 20yr Precipitation ExtremesExtreme Weather and Climate datasets are compiled for the US Climate Regions and the continental US. Raw data can be accessed here.
Totals by time period:
Annual Totals by period
Values in the table are the percent of available stations that exceeded the 20yr Precipitation for each time period.
- Cheng, L., A. AghaKouchak, E. Gilleland, and R. W. Katz, 2014: Non-stationary extreme value analysis in a changing climate. Climatic Change, 127, 353–369. doi:10.1007/s10584-014-1254-5
- Hoerling, M., J. Eischeid, J. Perlwitz, X. Quan, and K. Wolter, 2016: Characterizing Recent Trends in U.S. Heavy Precipitation. J. Climate, 29, 2313-2332. doi:10.1175/JCLI-D-15-0441.1
- Menne, M. J., I. Durre, R. S. Vose, B. E. Gleason, and T. G. Houston, 2012: An overview of the Global Historical Climatology Network-Daily Database. J. Atmos. Oceanic Tech., 29, 897–910. doi:10.1175/JTECH-D-11-00103.1
- Sillmann, J., V. V. Kharin, X. Zhang, F. W. Zwiers, and D. Bronaugh, 2013: Climate extreme indices in the CMIP5 multimodel ensemble: Part 1. Model evaluation in the present climate. J. Geophys. Res. Atmos., 118, 1716–1733. doi:10.1002/jgrd.50203
- Wolter, K., M. Hoerling, J.K. Eischeid, and L. Cheng, 2016: What History Tells us about 2015 US Daily Rainfall Extremes (in "Explaining Extremes of 2015 from a Climate Perspective"). Bull. Amer. Meteor. Soc., 97, S9-S13. doi:10.1175/BAMS-D-16-0166.1 PDF
- Explaining extreme events in 2015 from a climate perspective - Press Conference
NOAA's Martin Hoerling explains the context for precipitation extremes