In July 2016, the lower portion of a valley glacier in the Aru Range of Tibet detached and barreled into a nearby valley, killing nine people and hundreds of animals. The huge avalanche, one of the largest scientists had ever seen, sent a tongue of debris spreading across 9 square kilometers (3 square miles). With debris reaching speeds of 140 kilometers (90 miles) per hour, the avalanche was remarkably fast for its size.
(NASA Earth Observatory image by Joshua Stevens, using modified Copernicus Sentinel 2 data processed by the European Space Agency. Image collected on July 21, 2016.)
Researchers were initially baffled about how it had happened. The glacier was on a nearly flat slope that was too shallow to cause avalanches, especially fast-moving ones. What’s more, the collapse happened at an elevation where permafrost was widespread; it should have securely anchored the glacier to the surface.
Two months later, it happened again — this time to a glacier just a few kilometers away. One gigantic avalanche was unusual; two in a row was unprecedented. The second collapse raised even more questions. Had an earthquake played a role in triggering them? Did climate change play a role? Should we expect more of these mega-avalanches?
(NASA Earth Observatory image by Joshua Stevens and Jesse Allen, using ASTER data from NASA/GSFC/METI/ERSDAC/JAROS, and U.S./Japan ASTER Science Team. Image collected on October 4, 2016.)
Now scientists have answers about how these unusual avalanches happened. There were four factors that came together and triggered the collapses, an international team of researchers reported in Nature Geoscience. The scientists analyzed many types of satellite, meteorological, and seismic data to reach their conclusions. They also sent teams of researchers to investigate the avalanches in the field.
First, increasing snowfall since the mid-1990s caused snow to pile up and start working its way toward the front edge of the glaciers (a process known as surging). Second, a great deal of rain fell in the summer of 2016. As a result, water worked its way through crevasses on the surface and lubricated the undersides of the glaciers. Third, water pooled up underneath the glaciers, even as the edges remained frozen to the ground. Fourth, the glaciers sat on a fine-grained layer of siltstone and clay that became extremely slippery.
Notice the large amounts of silt and clay in the path of the first avalanche. (Photo taken on July 15, 2017, by Adrien Gilbert/University of Oslo)
Earth Observatory checked in with Andreas Kääb (University of Oslo), lead author of the study, to find out more about how the avalanche happened and what it means.
These glaciers were not on a steep slope, but the avalanche moved quite quickly. How did that happen? Strong resistance by the frozen margins and tongues of the glaciers allowed the pressure to build instead of enabling them to adjust. The glaciers were loading up more and more pressure until the frozen margins suddenly failed. Local people reported a load bang. Once the margins failed, there was nothing at the glacier bed to hold it back, just water-soaked clay.
Your study notes that there was “undestroyed grassy vegetation on the lee side of the hills, suggesting that the fast-moving mass had partially jumped over it.” Are you saying the avalanche was airborne? If so, is that unusual? Yes, for a small part of the avalanche path. We see this for other large-volume, high-speed avalanches, and it really illustrates the massive amount of energy released. You need quite high speeds in order for debris to jump. For us, the phenomenon is important as validation for the speeds obtained from the seismic signals the avalanches triggered and the avalanche modeling that we did.
Would you say these collapses were a product of climate change? Climate change was necessary, but other factors that had nothing to do with climate were also critical. The increasing mass of the glaciers since the 1990s and the heavy rains and meltwater in 2016 are connected to climate change. The type of bedrock and the way the edges were frozen to the ground had nothing to do with climate change.
Can we expect to see more big glacial collapses as the world gets warmer? It’s not clear. Climate change could increase or, maybe even more likely, decrease the probability of such massive collapses. Most glaciers on Earth are actually losing mass (not gaining, like the two glaciers in Tibet were). Also, if permafrost becomes less widespread over time and glacier margins melt, it is less likely that pressure will build up in that way that it did in this case.
I know you used several types of satellite data as part of this analysis. Can you mention a few that yielded particularly useful information? We used a lot of different sources of data: Sentinel 1 and 2, TerraSAR-X/TanDEM-X, Planet Labs, and DigitalGlobe WorldView. Landsat 8 was absolutely critical because it gave the first and critical indication of the soft-bed characteristics. The entire Landsat series was instrumental for checking the glacier history since the 1980s. We also used declassified Corona data back to the 1960s.
Are these sorts of avalanches likely to happen in other parts of the world? Honestly, I have no clue at the moment, but we would be much less surprised next time. We know now that this type of collapse can happen under special circumstances. (It happened once before in the Caucasus at Kolka Glacier.) One thing that should be investigated is whether there are other glaciers—especially polythermal ones—with these very fine-grained materials underneath them.
Three dimensional CNES Pléiades image of the avalanches. Processed by Etienne Berthier. Via Twitter.
Atmospheric rivers stretched from Asia to North America in October 2017. Learn more.
If you live on the West Coast of North America, you have probably heard meteorologists talk about “atmospheric rivers” — the narrow, low-level plumes of moisture that often accompany extratropical storms and transport large volumes of water vapor across long distances. When atmospheric rivers encounter land, they can drop tremendous amounts of rain and snow. That can be good for replenishing reservoirs and for quenching droughts, but these remarkable meteorological features can also trigger destructive floods, landslides, and wind storms.
During the past decade, atmospheric rivers have fueled a flood of another type: scientific research papers. Prior to 2004, fewer than 10 studies mentioned atmospheric rivers in any given year; in 2015, about 200 studies were published on the matter. The availability of increasingly sophisticated satellite and aircraft data has fueled the trend, according to a recent article in the Bulletin of the American Meteorological Society. Here’s a sampling of what scientists have learned about these rivers in the sky.
They Can Bring Rains, Winds, And Lots of Damage
In a study led by Duane Waliser of NASA’s Jet Propulsion Laboratory and published in Nature Geoscience, researchers showed that atmospheric rivers are among the most damaging storm types in the middle latitudes. Of the wettest and windiest storms (those ranked in the top 2 percent), atmospheric rivers were associated with nearly half of them. Waliser and colleagues found that atmospheric rivers were associated with a doubling of wind speed compared to all storm conditions.
They Shift With The Seasons
During the winter, atmospheric rivers in the Pacific generally shift northward and westward, Bryan Mundhenk of Colorado State University and colleagues concluded in a study. They also found that the El Niño/Southern Oscillation (ENSO) cycle can affect the frequency of atmospheric river events and shift where they occur. The research was based on data processed by MERRA, a NASA reanalysis of meteorological data from satellites.
They Aren’t Just a West Coast Thing
Atmospheric rivers are a global phenomenon and responsible for about 22 percent of all water runoff. One recent study from a University of Georgia team underscored that the U.S. Southeast sees a steady stream of atmospheric rivers. “They are more common than we thought in the Southeast, and it is important to properly understand their contributions to rainfall given our dependence on agriculture and the hazards excessive rainfall can pose,” said Marshall Shepherd of the University of Georgia. Other studies note that atmospheric rivers have contributed to anomalous snow accumulation in East Antarctica and extreme rainfall in the Bay of Bengal.
Climate Change Could Alter Them
A recent study led by Christine Shields of the National Center for Atmospheric Research suggests that climate change could push atmospheric rivers in the Pacific toward the equator and bring more intense rains to southern California. The modeling calls for smaller increases in rain rates in the Pacific Northwest. Another ensemble of models shows a 35 percent increase in the number of days with landfalling atmospheric rivers in western North America.
Satellites Are Key to Studying Their Precipitation
While there are few ground-based weather stations in the open ocean to tally how much rain falls, satellites such as those included in the Global Precipitation Measurement (GPM) mission can estimate precipitation rates from above. “Satellites have proven valuable over both the ocean and land, though uncertainties are often larger over land because of complicating factors like the terrain and the presence of snow on the surface,” said Ali Behrangi, the author of a study that assessed the skill of different satellite-derived measurements of precipitation rates.
The Gulf of Mexico, like any sea, is rich in dissolved salts. Unlike most seas, the Gulf also sits atop a big mound of salt. Left behind by an ancient ocean, salt deposits lie beneath the Gulf seafloor and get pressed and squeezed and bulged by the heavy sediments laying on top of them. The result is pock-marked, almost lunar-looking seafloor. The many mounds and depressions came into clearer relief this spring with the release of a new seafloor bathymetry map compiled from oil and gas industry surveys and assembled by the U.S. Bureau of Ocean Energy Management.
3D Water Babies
NASA’s Scientific Visualization Studio took a look back at conditions in the Pacific Ocean in 2015-16, which included the arrival and departure of both El Nino and La Nina. The 3D visualizations were derived from NASA’s Modern-Era Retrospective Analysis for Research and Applications (MERRA) dataset, a global climate modeling effort that is built from remote sensing data.
In other Nino news, a research team led by NASA Langley scientists found that the strong 2015-2016 El Niño lofted abnormal amounts of cloud ice and water vapor unusually high into the atmosphere, creating conditions similar to what could happen on a larger scale in a warming world.
Not the Kind of Brightening You Want to See
For past few years, warm ocean temperatures in the western Pacific Ocean have wrecked havoc on the Great Barrier Reef. Extreme water temperatures can disrupt the symbiotic partnership between corals and the algae that live inside their tissues. This leads the colorful algae to wash out of the coral, leaving them bright white in what scientists refer to as “bleaching” events. The health of coral reefs is usually monitored by airborne and diver-based surveys, but the European Space Agency recently reported that scientists have been able to use Sentinel 2 data to identify a bleaching event on the Great Barrier Reef. Such satellite monitoring could prove especially useful for monitoring reefs that are more remote and not as well studied as those around Australia.
Eyeing the Fuel for Hurricane Season
On June 1, the beginning of Atlantic Hurricane Season, the National Oceanic and Atmospheric Administration released a map of sea surface temperatures in the Caribbean, the Gulf of Mexico, and the tropical North Atlantic Ocean. The darkest orange areas indicate water temperatures of 26.5°C (80°F) and higher — the temperatures required for the formation and growth of hurricanes. Forecasters are expecting a hurricane season that is a bit more active than average.
(Finger)Prints of Tides
In a new comprehensive analysis published in Geophysical Research Letters, a French-led research team found that global mean sea level is rising 25 percent faster now than it did during the late 20th century. The increase is mostly due to increased melting of the Greenland Ice Sheet. A big part of the study was a reanalysis and recalibration of data acquired by satellites over the past 25 years, which are now better correlated to surface-based measurements. The study found that mean sea level has been increasing by 3 millimeters (0.1 inches) per year. The American Geophysical Union published a popular summary of the study.
Tropical forests, such as those in Gabon, Africa, are an important reservoir of carbon. (Photography courtesy of Sassan Saatchi, NASA/JPL-Caltech.)
Old-growth forests are vital because they capture large amounts of carbon and provide homes to hundreds of species. In the Eastern United States, trees in these minimally disturbed ecosystems tend to be more than 120 years old.
Can satellites help pinpoint this “old-growth” and quantify its value? That was the question Joan Maloof posed to a group of researchers during a talk at NASA Goddard Space Flight Center in May 2017. As head of the Old-Growth Forest Network and a professor at Salisbury University, Maloof aims to identify stands of old-growth forests for conservation. A large part of her job involves explaining why these areas are important—something satellite data can help show.
As it turns out, satellites have already told us much about trees. A 2012 story from NASA Earth Observatory described some of the remote sensing methods researchers use:
Scientists have used a variety of methods to survey the world’s forests and their biomass. […] With satellites, they have collected regional and global measurements of the “greenness” of the land surface and assessed the presence or absence of vegetation, while looking for signals to distinguish trees from shrubs from ground cover.
In January 2017, a paper in Science Advances tracked intact forest landscapes between 2000 and 2013. (Intact forest landscapes were defined as areas larger than 500 square kilometers with no signs of human activity in Landsat imagery). This new research underscores the importance of such landscapes. The study’s authors identified several key findings:
Dividing up forest landscapes with roads and development can hinder their ability to store carbon
Forest wildlands (forests least affected by human activity) have the highest conservation value
Large forest wildlands store more carbon than small forest wildlands; they are at risk of deforestation
The global extent of intact forests declined by 7 percent 2000 and 2013.
Rivers on three planetary bodies: the dry Parana Valles on Mars (left), the Nile River on Earth (middle), and Vid Flumina on Titan (right). Image by Benjamin Black using NASA data.
One of the more distinctive things about Earth among the planets is that we have plate tectonics. In other words, the hard, outer shell of the planet (called the lithosphere) is divided into several cool, rigid plates that float atop a hotter, more fluid layer of rock (the asthenosphere). These rigid surface plates do not float placidly: their grinding, colliding, shifting, and diving causes earthquakes, fuels volcanoes, builds mountains, tears open oceans, and constantly remodels and resurfaces the planet.
That is a far cry from what is happening on Mars and Titan, according to a recent study published in Science. Researchers came to that conclusion by carefully analyzing the way rivers cut through each of these planetary bodies. On Earth, countless rivers and streams snake their way across the surface. On Mars, rivers dried up long ago, but evidence of their presence remains etched into the arid surface. On Titan, Saturn’s largest moon, rivers of liquid ethane and methane still flow into lakes.
Artist’s cross section illustrating the main types of plate boundaries on Earth. (Cross section by José F. Vigil from This Dynamic Planet—a wall map produced jointly by the U.S. Geological Survey, the Smithsonian Institution, and the U.S. Naval Research Laboratory.)
By comparing imagery and data from all three planetary bodies, researchers noticed distinctive bends in the courses of rivers on Earth; these were formed as rivers were forced to wind around mountain ranges. These bends were absent in river networks on Mars and Titan. In an MIT press release, Benjamin Black, a geologist at the City College of New York, explained:
“Titan might have broad-scale highs and lows, which might have formed some time ago, and the rivers have been eroding into that topography ever since, as opposed to having new mountain ranges popping up all the time, with rivers constantly fighting against them.”
It is no secret that many diesel cars and trucks emit more pollution under real-world driving conditions than during laboratory certification testing. Many lab tests, for instance, are run with perfectly maintained vehicles on flat surfaces in ideal conditions. In the real world, drivers chug up hills or sit in traffic in bad weather in vehicles well past their prime.
Until this month, nobody had tallied the health effects of all the excess diesel air pollution entering the atmosphere through real-world driving conditions. According to a new study published in Nature, vehicles in eleven major markets (Australia, Brazil, Canada, China, Europe, India, Japan, Mexico, Russia, South Korea, and the United States) emitted about 4.6 million more tons of nitrogen oxides (NOx) in 2015 than official laboratory tests suggested they would. NOx contributes to the accumulation of both ground-level ozone (O3) and fine particulate matter (PM2.5) in the atmosphere.
According to the research team, nearly one-third of heavy-duty diesel vehicle emissions and over half of light-duty diesel vehicle emissions are above the certification limits. On average, light-duty diesel vehicles produce 2.3 times more NOx than the limit; heavy-duty diesel vehicles emit more than 1.45 times the limit.
The authors of the study calculated the health effects for current and future levels of this excess diesel NOx by running a global atmospheric chemistry model that simulates the distribution of PM2.5 and O3. The bottom line: excess NOx caused 38,000 premature deaths in 2015. It could cause as many as 183,600 premature deaths by 2040 as the use of diesel increases.
China suffered the largest health burden from diesel NOx emissions—31,400 deaths, of which 10,700 are attributed to excess NOx—followed by the European Union (28,500 total; 11,500 excess) and India (26,700 total; 9,400 excess).
Light-duty diesel vehicles in the European Union accounted for 6 out of every 10 deaths related to excess diesel NOx.
In the United States, heavy-duty diesel vehicles caused 10 times the impact of light-duty diesel cars.
This map, based on previous research, shows a model estimate of the average number of deaths per 1,000 square kilometers (386 square miles) per year due to fine particulate matter (PM2.5), a type of outdoor air pollution. Pollution from diesel exhaust is one contributer to PM2.5. Earth Observatory image by Robert Simmon based on data provided by Jason West. Learn more about this map here.
Model simulation of the hydroxyl radical concentration in the atmosphere. Image by Angharad Stell, University of Bristol.
A mystery about global methane trends just got more muddled. Two studies published in April 2017 suggest that recent increases in atmospheric concentrations of methane may not be caused by increasing emissions. Instead, the culprit may be the reduced availability of highly reactive “detergent” molecules called hydroxyl radicals (OH) that break methane down.
Understanding how globally-averaged methane concentrations have fluctuated in the past few decades—and particularly why they have increased significantly since 2007—has proven puzzling to researchers. As we reported last year:
“If you focus on just the past five decades—when modern scientific tools have been available to detect atmospheric methane—there have been fluctuations in methane levels that are harder to explain. Since 2007, methane has been on the rise, and no one is quite sure why. Some scientists think tropical wetlands have gotten a bit wetter and are releasing more gas. Others point to the natural gas fracking boom in North America and its sometimes leaky infrastructure. Others wonder if changes in agriculture may be playing a role.”
The new studies suggest that such theories may be off the mark. Both of them find that OH levels may have decreased by 7 to 8 percent since the early 2000s. That is enough to make methane concentrations increase by simply leaving the gas to linger in the atmosphere longer than before.
Atmospheric methane has continued to increase, though the rate of the increase has varied considerably over time and puzzled experts. (NASA Earth Observatory image by Joshua Stevens, using data from NOAA. Learn more about the image.)
As a press release from the Jet Propulsion Laboratory (JPL) noted: “Think of the atmosphere like a kitchen sink with the faucet running,” said Christian Frankenberg, an associate professor of environmental science and engineering at Caltech and a JPL researcher. “When the water level inside the sink rises, that can mean that you’ve opened up the faucet more. Or it can mean that the drain is blocking up. You have to look at both.”
Unfortunately, neither of the new studies is definitive. The authors of both papers caution that high degrees of uncertainty remain, and future work is required to reduce those uncertainties. “Basically these studies are opening a new can of worms, and there was no shortage of worms,” Stefan Schwietzke, a NOAA atmospheric scientist, told Science News.
You can find the full studies here and here. The University of Bristol has also published a press release.
Methane emissions related to human activity are on the rise. (NASA Earth Observatory image by Joshua Stevens, using data from CDIAC. Learn more.)
NASA/EO-1/ALI/Jesse Allen and Robert Simmon. More info about this image here.
Four years ago today, one of the largest non-volcanic landslides in U.S. history began high on the northern wall of Utah’s Bingham Canyon mine. During two main bursts of activity that day, several million cubic meters of rock and soil careened deep into the mine’s pit.
This was a case where science and the right preparation saved lives. The company that runs the mine had installed an interferometric radar system months before the slide, and it prevented miners from being blindsided. With the radar system in place, mine operators detected changes in the stability of the pit’s walls well before the landslide occurred. When the slope finally gave way, all the workers in the pit had already been evacuated. Not one person was injured.
The image above was acquired by the Advanced Land Imager (ALI) on the Earth Observing-1 satellite on May 2, 2013. For more detailed and recent images of the mine pit, see the montage below. Photos A, B, and C — from Rio Tinto Kennecott — show an overview of the pit, the source area, and slide debris in the immediate aftermath of the event. D is an aerial image from the state government of Utah that shows the mine nine months before the landslide. E is part of the ALI satellite image above. F is a second aerial image taken in February 2014. By then, much of the debris had been removed and a new access road had been built.
On April 10, at 9:30 p.m. and again at 11:05 p.m., the slope gave way and thundered down into the pit, filling in part of what had been the largest man-made excavation in the world. Later analysis estimated that the landslide was at the time the largest non-volcanic slide in recorded North American history. Now, University of Utah geoscientists have revisited the slide with a combined analysis of aerial photos, computer modeling, and seismic data to pick apart the details. The total volume of rock that fell during the slide was 52 million cubic meters, they report, enough to cover Central Park with 50 feet of rock and dirt. The slide occurred in two main phases, but researchers used infrasound recordings and seismic data to discover 11 additional landslides that occurred between the two main events. Modeling and further seismic analysis revealed the average speeds at which the hillsides fell: 81 miles per hour for the first main slide and 92 mph for the second, with peak speeds well over 150 mph.
The interferometric radar system is not the only safety technology in place at Bingham Canyon. Drones, GPS, and trained experts keep a vigilant eye out for signs of landslides at this mine. Technology and tactics like this mean landslides cause very few injuries and deaths in the United States even though significant landslide potential exists in many parts of the country. As we recently reported, many other parts of the world (notably Africa and South America) are not nearly so fortunate.
Decades or even centuries from now, when enough time has passed for historians to write definitive accounts of global warming and climate change, two names are likely to make it into history books: Syukuro Manabe and Richard Wetherald.
In the late-1960s, these two scientists from the Geophysical Fluid Dynamics Laboratory developed one of the very first climate models. In 1967, they published results showing that global temperatures would increase by 2.0 degrees Celsius (3.6 degrees Fahrenheit) if the carbon dioxide content of the atmosphere doubled.
The top map shows temperature changes in one of Manabe’s coupled atmosphere–ocean models. It depicts what would happen by the 70th year of a global warming experiment when atmospheric concentrations of carbon dioxide are doubled. The model’s predictions match closely with levels of warming observed in the real world. The lower image shows observed change from a 30-year base period around 1961–1990 and 1991–2015. The map was obtained using the historical surface temperature dataset HadCRUT4. Figure published in Stouffer & Manabe, 2017.
As Ethan Siegel pointed out in Forbes, carbon dioxide content has risen by roughly 50 percent since the pre-industrial era. Observed temperatures, meanwhile, have increased by 1°C (1.8°F). For a pair of scientists working in a time when computer instructions were compiled on printed punch cards and processing was thousands of times slower than today, they created a remarkably accurate model. That is not to say that Manabe claims his 1967 model is perfect. In fact, he is quick to point out that there are some aspects that his and other climate models still get wrong. In an interview with CarbonBrief, he put it this way:
Models have been very effective in predicting climate change, but have not been as effective in predicting its impact on ecosystem[s] and human society. The distinction between the two has not been stated clearly. For this reason, major effort should be made to monitor globally not only climate change, but also its impact on ecosystem[s] through remote sensing from satellites as well as in-situ observation.
This line plot shows yearly temperature anomalies from 1880 to 2014 as recorded by NASA, NOAA, the Japan Meteorological Agency, and the Met Office Hadley Centre (United Kingdom). Though there are minor variations from year to year, all four records show peaks and valleys in sync with each other. All show rapid warming in the past few decades, and all show the last decade as the warmest. Read more about this chart.
Most NASA stories about sea ice records tend to start with the year 1979, when consistent observations started to be collected regularly by satellites. But we do know a bit about what happened before then based on aerial, ocean, and ground-based data. And now some recently recovered observations and new modeling suggest that Arctic sea ice grew for a period between 1950 and 1975.
The Sun hangs low on the horizon above solidified pancake ice in the Arctic Ocean. (Photograph courtesy Andy Mahoney, NSIDC.)
A new analysis led by Marie-Ève Gagné of Environment and Climate Change Canada offers an explanation for the rise: air pollution. Gagné and colleagues showed that sulfate aerosol particles, which are released by the burning of fossil fuels, may have disguised the impact of greenhouse gases on Arctic sea ice.
In a press release, the American Geophysical Union explained more details:
These particles, called sulfate aerosols, reflect sunlight back into space and cool the surface. This cooling effect may have disguised the influence of global warming on Arctic sea ice and may have resulted in sea ice growth recorded by Russian aerial surveys in the region from 1950 through 1975, according to the new research.
“The cooling impact from increasing aerosols more than masked the warming impact from increasing greenhouse gases,” said John Fyfe, a senior scientist at Environment and Climate Change Canada and a co-author of the new study accepted for publication in Geophysical Research Letters.
To test the aerosol idea, researchers used computer modeling to simulate sulfate aerosols in the Arctic from 1950 through 1975. Concentrations of sulfate aerosols were especially high during these years before regulations like the Clean Air Act limited sulfur dioxide emissions that produce sulfate aerosols.
The study’s authors then matched the sulfate aerosol simulations to Russian observational data that suggested a substantial amount of sea ice growth during those years in the eastern Arctic. The resulting simulations show the cooling contribution of aerosols offset the ongoing warming effect of increasing greenhouse gases over the mid-twentieth century in that part of the Arctic. This would explain the expansion of the Arctic sea ice cover in those years, according to the new study.
Aerosols spend only days or weeks in the atmosphere so their effects are short-lived. The weak aerosol cooling effect diminished after 1980, following the enactment of clean air regulations. In the absence of this cooling effect, the warming effect of long-lived greenhouse gases like carbon dioxide has prevailed, leading to Arctic sea ice loss, according to the study’s authors.
This chart, from Gagné et al, shows the area-averaged annual mean sea ice concentration anomaly between 1950 and 2005. The red line reflects Arctic and Antarctic Research Institute (AARI) data, which is based on historical sea ice charts from several sources (aircraft, ship, and satellite observations). The blue Walsh & Chapman data includes additional historical sources and uses slightly different techniques for merging various data streams. The black line is a simulated mean sea ice concentration from the CanESM2 large ensemble, a group of models developed at the Canadian Center for Climate Modelling and Analysis. Anomaliesarerelativetoclimatologicalaveragesbetween1975and 2005.Thetime serieshavebeenlightlysmoothedwith5-yearrunningmean.Thegrayshading showsthe5-95 percentrangeovertheindividualsimulationsofthemodelensemble.Thedottedlines arethe piecewise linearapproximatedtrends.