WunderBlog Archive » Category 6™
Category 6 has moved! See the latest from Dr. Jeff Masters and Bob Henson here.The Rains of May and the Science of Recurrence Intervals
As this memorable month sloshes toward its final weekend, rainfall records have continued to accumulate across the Southern Plains. May is already the wettest month on record in the statewide averages for both Oklahoma (14.43” as of 9:30 am CDT May 30, crushing the 10.75” from October 1941) and Texas (7.54” as of May 27, beating out the ominous 6.66” from June 2004). This is a startling feat when you consider that the two states’ combined area is almost the size of Egypt. Several individual locations have also scored their wettest months on record, including Wichita Falls, TX; Oklahoma City, OK; and Colorado Springs, CO (more on these and other local records next week). The onslaught of wetness is now winding down just as May comes to a close. On the plus side, Oklahoma state climatologist Gary McManus has dubbed this wet period “the death of the Southern Plains drought,” one that had dug its claws into the region for most of the 2010s. However, the associated flooding has killed at least 25 people, with 14 missing, and damaged or destroyed many hundreds of structures.
Figure 1. Motorists are stranded along I-45 in north Houston on Tuesday, May 26, after overnight storms flooded the area. Image credit: Cody Duty/Houston Chronicle via AP.
When an event like this happens, it’s human nature to wonder just how unusual it is. Climate change has added urgency to such questions, but even in a constant (“stationary”) climate, precise answers on the rarity of heavy rain at a given location can be hard to come by. One of the most common strategies is to calculate the “average recurrence interval” (ARI), or the typical amount of time that would separate events of a certain magnitude when considered over a much longer time period. At heart, this is simply a different name for what was long called the return period—i.e., a “100-year flood.” The problem with the “XXX-year flood” wording is that it’s wide-open to misinterpretation. People may think that a 100-year flood won’t be followed by a similarly sized event until a century has passed, when in fact it could recur any time. (You might also describe this as a “1% probability flood,” meaning there’s a 1% chance of getting it in a given year.)
NOAA has decades of return-period estimates under its belt. From the 1950s into the 2000s, these were published as hard-copy maps (see PDFs). More recently, NOAA has undertaken a comprehensive effort called Atlas 14 that is updating the return periods (now called ARIs) and putting them into a fully digital, gridded, GIS-compatible dataset. Since we’re now well into the 21st century, the project has more than a hundred years of rain-gauge measurements to draw on, from thousands of U.S. sites. The Atlas 14 results for Texas won’t be ready until at least 2018, but the Oklahoma numbers are already online, and they provide an useful window on the rarity of the rains that befell the Sooner State this month.
Figure 2. 30-day rainfall totals for Oklahoma, valid from 10:00 am CDT April 30 through May 30. Most of the heavy rain occurred from May 5 onward. Image credit: Oklahoma Mesonet.
Take long-suffering Norman, OK, which has racked up one of the state’s highest totals for the past 30 days: 23.39” as of 10:00 am CDT Saturday, May 30. (Even higher totals can be found in southeast Oklahoma, but that region also gets more rain in a typical year, so its ARIs for this event aren’t as impressive.) For a 30-day period in Norman, the best estimate of ARIs from Atlas 14 is 500 years for a rainfall of 22.4,” and 1000 years for 24.0.” When a 90% confidence interval is employed, the 500-year range is 16.3” to 29.2”, and the 1000-year range is 16.9” to 32.0”. No matter how you slice it, this goes well beyond a once-in-a-lifetime event for Normanites when considered for the month as a whole. The 24-hour numbers for Norman are less spectacular: the biggest daily rain of the month so far, 4.67” on May 19, falls very near the 5-year ARI value of 4.66”. So it’s the full month of rain that’s much more noteworthy.
Figure 3. Maximum 24-hour rainfall totals for the Houston area during the period from 1200 GMT June 5 to 11, 2001 (top), as Tropical Storm/Depression Allison dawdled over eastern Texas, and preliminary 24-hour totals from 1200 GMT May 25 to 26, 2015 (bottom), the day that brought massive flooding to western parts of the Houston metro area. The two maps are at the same scale. The analyses are based on data from more than 450 raingauges. Image credit: Applied Weather Associates and MetStat.
Even though Atlas 14 data aren’t yet available for Texas, we can gain some ARI-based insight on this storm thanks to the work of Tye Parzybok. His company, MetStat, has calculated new ARIs for 6- and 24-hour rainfall under a subcontract with Applied Weather Associates for the Texas Water Development Board. MetStat produces near-real-time rainfall estimates and predictions and ARI maps for clients and the public, drawing on radar data as well as more than 20,000 North American rain guages, including many compiled through the Citizen Weather Observer Program. (See the MetStat Extreme Precipitation Blog for writeups on recent events.)
Examining the rainfall in and around Houston on Monday, May 25, MetStat found 24-hour ARIs of close to 1000 years for parts of the southwest Houston area (see Figure 4 below), where the gridded analysis indicates a 24-hour maximum of 13.48”. The water then flowed downstream to produce major flooding in central Houston, where the actual rainfall was much less. During Tropical Storm Allison in June 2001, extremely heavy rains covered a much larger part of southeast Texas, with several days of downpours encompassing some truly extreme totals (see Figure 3 above). One rain gauge at Greens Bayou and Mount Houston Parkway measured a 6-hour total of 20.64” during Allison, and the MetStat gridded analysis suggests a nearby maximum as high as 20.98”.
Figure 4. A preliminary map of average recurrence intervals (ARIs) for the 24-hour rains observed in and around Houston from 7:00 am CDT May 25 to 7:00 am May 26, 2015. The map was produced using 820 rain-gauge reports and the MetStorm analysis system. The highest ARI at a given point was 932 years. Image credit: MetStat, Weather Decision Technologies, and Synoptic.
It’s the weeks-long persistence that seems to be most unusual about this month’s rain in both Texas and Oklahoma. There’s an obvious parallel with this past February, which was marked by an unprecedented month-long barrage of cold and snow in much of New England and record winter warmth in and near California. These events raise questions (beyond the scope of this blog post!) about why the steering flow across North America this year has been so prone to multiweek blocking episodes.
And now, the caveats
ARIs play a huge role in public policy. As Parzybok notes in an EarthZine article, urban water systems are widely designed for ARI rainfalls of 1 to 10 years, while highways are generally built to reflect 10- to 25-year ARIs. While policymakers and the public need and want some kind of yardstick for extreme rain events—and ARIs are the best one we have—they’re also an imperfect tool and shouldn’t be taken as gospel. Some reasons:
It’s not surprising to have rare events over small areas. Very localized extremes may have very long recurrence intervals at a specific point. If you live in Oklahoma, the ARI for a major tornado at your house might be hundreds of years, yet dozens of twisters occur across the state in a typical spring.
Rainfall can’t be measured everywhere at once. Outside of urban areas, century-long observation points may be separated by tens of miles, which means some huge but localized downpours may be omitted from the long-term record.
Rare events are hard to contextualize, precisely because they’re so rare. Scientists can use certain statistical techniques to calculate ARIs of up to 200 years from a regional cluster of century-long raingauge records. Beyond that threshold, there’s a lot more uncertainty in the ARIs, as illustrated by the example above from Norman.
Our climate isn’t stationary. Climate science tells us that a warming planet will intensify rains as well as drought impacts, due largely to the increased evaporation of water from both oceans and land. More than 15 years of observational study backs this up, showing that the heaviest precipitation events are indeed becoming more intense in many parts of the world.
A new analysis released on Wednesday by Climate Central highlights the increasing risk of one-day downpours, defined as the wettest 1% of all wet days at a given spot for the period 1950 through 2014. Since precipitation is so variable in time and space, the state-by-state analysis is a patchwork. Most states show increases (as high as 104% in Rhode Island), while six states show minor declines (the largest being -6% in California). Overall, if we assume that downpours at various time scales are becoming more common in most parts of the nation, then it would follow that ARIs from Atlas 14 will tend to be on the high side—i.e., in many locations, we would expect an event diagnosed with a 100-year ARI to occur a bit more often as climate change proceeds.
The interactive below shows how the frequency of heavy downpours is evolving over time across the nation. For more background, see the accompanying research report from Climate Central (the interactive is also available at that report site].
Bob Henson
Figure 5. Heavy rains soaked parts of the drought-stricken Northeast on Thursday, May 28, including Portland, Maine. Image credit: wunderphotographer Alphaholik.
Figure 1. Motorists are stranded along I-45 in north Houston on Tuesday, May 26, after overnight storms flooded the area. Image credit: Cody Duty/Houston Chronicle via AP.
When an event like this happens, it’s human nature to wonder just how unusual it is. Climate change has added urgency to such questions, but even in a constant (“stationary”) climate, precise answers on the rarity of heavy rain at a given location can be hard to come by. One of the most common strategies is to calculate the “average recurrence interval” (ARI), or the typical amount of time that would separate events of a certain magnitude when considered over a much longer time period. At heart, this is simply a different name for what was long called the return period—i.e., a “100-year flood.” The problem with the “XXX-year flood” wording is that it’s wide-open to misinterpretation. People may think that a 100-year flood won’t be followed by a similarly sized event until a century has passed, when in fact it could recur any time. (You might also describe this as a “1% probability flood,” meaning there’s a 1% chance of getting it in a given year.)
NOAA has decades of return-period estimates under its belt. From the 1950s into the 2000s, these were published as hard-copy maps (see PDFs). More recently, NOAA has undertaken a comprehensive effort called Atlas 14 that is updating the return periods (now called ARIs) and putting them into a fully digital, gridded, GIS-compatible dataset. Since we’re now well into the 21st century, the project has more than a hundred years of rain-gauge measurements to draw on, from thousands of U.S. sites. The Atlas 14 results for Texas won’t be ready until at least 2018, but the Oklahoma numbers are already online, and they provide an useful window on the rarity of the rains that befell the Sooner State this month.
Figure 2. 30-day rainfall totals for Oklahoma, valid from 10:00 am CDT April 30 through May 30. Most of the heavy rain occurred from May 5 onward. Image credit: Oklahoma Mesonet.
Take long-suffering Norman, OK, which has racked up one of the state’s highest totals for the past 30 days: 23.39” as of 10:00 am CDT Saturday, May 30. (Even higher totals can be found in southeast Oklahoma, but that region also gets more rain in a typical year, so its ARIs for this event aren’t as impressive.) For a 30-day period in Norman, the best estimate of ARIs from Atlas 14 is 500 years for a rainfall of 22.4,” and 1000 years for 24.0.” When a 90% confidence interval is employed, the 500-year range is 16.3” to 29.2”, and the 1000-year range is 16.9” to 32.0”. No matter how you slice it, this goes well beyond a once-in-a-lifetime event for Normanites when considered for the month as a whole. The 24-hour numbers for Norman are less spectacular: the biggest daily rain of the month so far, 4.67” on May 19, falls very near the 5-year ARI value of 4.66”. So it’s the full month of rain that’s much more noteworthy.
Figure 3. Maximum 24-hour rainfall totals for the Houston area during the period from 1200 GMT June 5 to 11, 2001 (top), as Tropical Storm/Depression Allison dawdled over eastern Texas, and preliminary 24-hour totals from 1200 GMT May 25 to 26, 2015 (bottom), the day that brought massive flooding to western parts of the Houston metro area. The two maps are at the same scale. The analyses are based on data from more than 450 raingauges. Image credit: Applied Weather Associates and MetStat.
Even though Atlas 14 data aren’t yet available for Texas, we can gain some ARI-based insight on this storm thanks to the work of Tye Parzybok. His company, MetStat, has calculated new ARIs for 6- and 24-hour rainfall under a subcontract with Applied Weather Associates for the Texas Water Development Board. MetStat produces near-real-time rainfall estimates and predictions and ARI maps for clients and the public, drawing on radar data as well as more than 20,000 North American rain guages, including many compiled through the Citizen Weather Observer Program. (See the MetStat Extreme Precipitation Blog for writeups on recent events.)
Examining the rainfall in and around Houston on Monday, May 25, MetStat found 24-hour ARIs of close to 1000 years for parts of the southwest Houston area (see Figure 4 below), where the gridded analysis indicates a 24-hour maximum of 13.48”. The water then flowed downstream to produce major flooding in central Houston, where the actual rainfall was much less. During Tropical Storm Allison in June 2001, extremely heavy rains covered a much larger part of southeast Texas, with several days of downpours encompassing some truly extreme totals (see Figure 3 above). One rain gauge at Greens Bayou and Mount Houston Parkway measured a 6-hour total of 20.64” during Allison, and the MetStat gridded analysis suggests a nearby maximum as high as 20.98”.
Figure 4. A preliminary map of average recurrence intervals (ARIs) for the 24-hour rains observed in and around Houston from 7:00 am CDT May 25 to 7:00 am May 26, 2015. The map was produced using 820 rain-gauge reports and the MetStorm analysis system. The highest ARI at a given point was 932 years. Image credit: MetStat, Weather Decision Technologies, and Synoptic.
It’s the weeks-long persistence that seems to be most unusual about this month’s rain in both Texas and Oklahoma. There’s an obvious parallel with this past February, which was marked by an unprecedented month-long barrage of cold and snow in much of New England and record winter warmth in and near California. These events raise questions (beyond the scope of this blog post!) about why the steering flow across North America this year has been so prone to multiweek blocking episodes.
And now, the caveats
ARIs play a huge role in public policy. As Parzybok notes in an EarthZine article, urban water systems are widely designed for ARI rainfalls of 1 to 10 years, while highways are generally built to reflect 10- to 25-year ARIs. While policymakers and the public need and want some kind of yardstick for extreme rain events—and ARIs are the best one we have—they’re also an imperfect tool and shouldn’t be taken as gospel. Some reasons:
It’s not surprising to have rare events over small areas. Very localized extremes may have very long recurrence intervals at a specific point. If you live in Oklahoma, the ARI for a major tornado at your house might be hundreds of years, yet dozens of twisters occur across the state in a typical spring.
Rainfall can’t be measured everywhere at once. Outside of urban areas, century-long observation points may be separated by tens of miles, which means some huge but localized downpours may be omitted from the long-term record.
Rare events are hard to contextualize, precisely because they’re so rare. Scientists can use certain statistical techniques to calculate ARIs of up to 200 years from a regional cluster of century-long raingauge records. Beyond that threshold, there’s a lot more uncertainty in the ARIs, as illustrated by the example above from Norman.
Our climate isn’t stationary. Climate science tells us that a warming planet will intensify rains as well as drought impacts, due largely to the increased evaporation of water from both oceans and land. More than 15 years of observational study backs this up, showing that the heaviest precipitation events are indeed becoming more intense in many parts of the world.
A new analysis released on Wednesday by Climate Central highlights the increasing risk of one-day downpours, defined as the wettest 1% of all wet days at a given spot for the period 1950 through 2014. Since precipitation is so variable in time and space, the state-by-state analysis is a patchwork. Most states show increases (as high as 104% in Rhode Island), while six states show minor declines (the largest being -6% in California). Overall, if we assume that downpours at various time scales are becoming more common in most parts of the nation, then it would follow that ARIs from Atlas 14 will tend to be on the high side—i.e., in many locations, we would expect an event diagnosed with a 100-year ARI to occur a bit more often as climate change proceeds.
The interactive below shows how the frequency of heavy downpours is evolving over time across the nation. For more background, see the accompanying research report from Climate Central (the interactive is also available at that report site].
Bob Henson
Figure 5. Heavy rains soaked parts of the drought-stricken Northeast on Thursday, May 28, including Portland, Maine. Image credit: wunderphotographer Alphaholik.
The views of the author are his/her own and do not necessarily represent the position of The Weather Company or its parent, IBM.