Understanding past weather to improve forecasts
Today, weather instruments around the globe collect millions of observations daily, giving forecasters and others a fairly comprehensive view of the global atmosphere and upcoming weather. A century ago, a few hundred observers on land and sea—most in the Northern Hemisphere—recorded a few thousand observations a day, at best. Now, ESRL researchers and colleagues around the world have produced realistic guesses at historic atmospheric conditions, based on sparse observations and an understanding of the way the atmosphere behaves today. This is the kind of detailed information that climate and weather researchers and historians have longed for, but accurate information has not been available beyond the past 60 years. The 20th Century Reanalysis project will eventually provide global surface and lower atmosphere weather data from the 1870s to the present.
Led by ESRL researchers, an international research team used Department of Energy supercomputers to stitch together sparse historic observations into an image of the most-likely atmospheric conditions at the time. Their conclusions, they hope, will not only help historians understand how weather affected key events, but will help climate modelers understand Earth’s future.
The 20th Century Reanalysis Project started when NOAA’s National Climatic Data Center and the National Center for Atmospheric Research began digitizing and making available original manuscript weather observations from the past 100 years. Building on the availability of these data, ESRL Physical Sciences Division and CIRES researchers Gil Compo, Jeff Whitaker, Prashant Sardeshmukh, and Nobuki Matsui proved to skeptics that by analyzing barometric pressure observations it would be possible to figure out the atmospheric patterns that created past weather conditions.
This research team wanted to produce a more extensive and improved historical dataset by combining the digitized observations from the past 100 years. With the vast amount of historical observations now available, powerful computing resources would be required to generate this huge dataset. Compo and his colleagues won a series of three Department of Energy INCITE Awards, giving them access to supercomputers at Lawrence Berkeley National Laboratory. During more than one million processing hours in the last three years, the team integrated a reanalysis model over many possible weather scenarios, and combined the possibilities with the available pressure observations to produce 6-hourly historical weather maps. The first version of the data is freely available and covers the period 1908-1958.
With the data, researchers will be able to understand a broad range of past weather variations, including conditions that led to historically notable events such as the 1930s Dust Bowl. Some of the data are already in use by researchers in Canada studying atmospheric conditions associated with the fatal Mallory Expedition of 1924. British climbers George Mallory and Sandy Irvine were attempting the first documented summit of Mt. Everest when they disappeared in an unexpected storm.
A manual analysis of the atmosphere during the Tri-State Tornado Outbreak of 18 March 1925, the deadliest tornado in US history (courtesy of R Maddox, retired, Tucson, Ariz).
The ensemble mean sea level pressure analysis which did not use any of the observations shown in the manual analysis produces a nearly identical pattern.
G.W. Kent Moore of the University of Toronto’s Department of Physics has long been interested in high-altitude meteorology, especially around Mt. Everest, and in understanding the weather systems that give rise to really bad storms. “I was always interested in the most famous disappearance with the first serious attempt to summit Mt. Everest,” said Moore. “Mallory and Irvine disappeared into what was described as some sort of blizzard. It was never known if they made the summit.”
Digging around in original records from the Royal Geographical Society, Moore was surprised to learn that during the expedition, pressure and temperature data had been collected at the base camp. These observations showed a huge pressure drop of 18 mb—the largest drop that Moore had seen in the region’s recorded history—which would mean that this was a much more serious storm than people originally thought. Interestingly these observations were published in 1926, but no discussion or comments on this major pressure drop ever developed.
Moore then found out about the 20th Century Reanalysis Project, and asked Compo for a look at the 1908-1958 data, which had just been compiled. The pressure drop was actually due to a large-scale weather system called a “western disturbance,” Moore found. Western disturbances are now known to be responsible for much of the bad weather in the Himalayas. “What is remarkable about this incredibly rich data set is that it reconstructs the state of the atmosphere, and for the first time we have a three-dimensional look at it,” said Moore. “It’s a wonderful tool for looking deeper back in time.”
Reanalysis is also a powerful tool for researchers looking forward in time, Compo said. By better understanding historic variability in the atmosphere, researchers can both improve climate models and learn to discern changes in weather variability, which may be associated with climate change.