Some very old temperature records in the San Joaquin Valley that have stood for more than a century have recently been stricken from the record books.
It’s exceedingly rare, but it occasionally happens here and elsewhere, said meteorologists at the National Weather Service's Hanford station.
Hot-weather records set in the Kings County town of Hanford in 1917, 1918 and 1919 were recently investigated by NWS officials after meteorologists and climatologists began noticing apparent inconsistencies in the data collected during this period around World War I.
A record set in Bakersfield during that same time period is also being looked at.
Meteorologist Colin McKellar, who investigated the apparent inconsistencies, said he compared the high and low temperatures recorded at each valley station for each day, month and year in question.
"It stands to reason that temperatures should be roughly similar," he said. "That's how we found errant data in Hanford."
All of the records in question involve a tally that area residents both dread and yet are often curious about: the total number of days in a calendar year when the daily high temperature reaches 100 degrees or hotter.
For 102 years, Hanford has held the dubious distinction of enduring an incredible 84 days in 1919 when the daily high temperature reached triple digits.
In his weather summary for September, NWS Climatologist Brian Ochs made it clear that the 1919 record and two others have been expunged.
"Please note," Ochs wrote in the summary, "the old record of 84 days for the calendar year of 1919, as well as 76 days per year during both 1917 (and) 1918 are no longer valid."
In other words, the changes are now official.
"The records at Hanford for 1917-1919 are now invalidated," Ochs confirmed. "We compared data with nearby stations, including Fresno, Visalia, and Porterville.
"The National Center for Environment Information reviewed our findings and invalidated the records ... as we found that the maximum temperature data at Hanford was significantly higher (by 8 degrees or higher) than at the other three stations that were compared."
The weather watchers in Hanford are also taking a deeper look at Bakersfield's record of 76 days of triple-digit temps set in 1917.
Recording 76 days of 100-plus temperatures seems to be a statistical outlier as the next closest record in Bakersfield is 67 days (nine days fewer) recorded in three times, in 1939, 2017 and again this year. In addition, 65 days were recorded in 1906 and 1908. The annual average is 36.
In light of evidence pointing to problems in Hanford in the same time frame, in 1917, it raises questions about whether Bakersfield might have been experiencing similar issues.
"Although we're still investigating Bakersfield's data and other stations, it is unlikely that Bakersfield's records in 1917 will become invalidated," Ochs said in an email. "However, preliminary findings show Bakersfield's highs were mainly consistent (within 7 degrees difference) with Porterville and Maricopa, but not with Wasco."
How could these inconsistencies have occurred in the first place?
Jeffrey Barlow, lead forecaster at the weather service's Hanford station, said the old mercury thermometers that were used in those days had to be shaken or preferably spun to be properly reset each day. And unfortunately, repeating values — with the same high temperature repeated over several days — suggests that some were not resetting the equipment.
Mercury thermometers have certainly been subject to errors for various reasons, though they were the best technology available at the time, Ochs said.
Siting, or location of weather stations, was also a likely problem in those days, the climatologist said. Some may have been located near rooftops, trees and buildings versus open areas.
The Oct. 31, 1891 edition of The Weekly Californian included information about one of Bakersfield’s earliest weather bureaus, located at the home of W.A. Webster, 16 miles south of the city.
Not exactly central Bakersfield.
According to a Feb. 15, 1970 article in The Californian, weather instruments were set up at the Santa Fe Railroad station at 14th and F streets in 1889. Official temperature and rainfall readings were shifted to the Kern County Airport in 1937.
Differences in technology, changes in weather station locations, and the possibility of human error raise questions about the wisdom of comparing very early local weather information with data collected in more recent decades.
Should Bakersfield's record of 76 days of triple digits in 1917 stand as an all-time record, or should it include an asterisk?
When McKellar worked at the South Dakota State Climate office, the office did not use any records prior to 1960 for climate analysis. Besides instrumentation and observing errors, another concern is a station's metadata.
"For example," he said, "how often were the instruments calibrated to account for instrumental drift? Was the station ever relocated? Was it in an open field and now in a city? One major area of uncertainty is the surrounding surface types."
When he lived in South Dakota, their climate site was next to a research farm. When the farm planted corn vs. wheat, they recorded slightly higher dew points.
"The main point," he said, "is that weather stations now have a stringent set of guidelines and operating procedures to ensure data quality that simply didn't exist 100 years ago."
Even with these factors considered, Ochs said, it is rare to invalidate records, even those that are 100 years old.
"It's hard to say how often it actually happens, but it has occurred at other stations."
Some worldwide records have been in question recently, including Death Valley’s all-time record of 134 degrees set back on July 10, 1913. It has not yet been invalidated by the World Meteorological Organization, Ochs said. However, the organization did invalidate the 136-degree world record set in 1923 in Libya.