A study of the just-concluded extremely warm weather in February concludes that climate change due to human-influenced greenhouse gas emissions was a big influence -- with more such occurrences to come. The study's authors are: Geer Jan van Oldenborgh of the Royal Netherlands Meteorological Institute (KNMI); Andrew King of the University of Melbourne in Australia; Friederike Otto of the University of Oxford in England; Gabriel Vecchi of Princeton University in New Jersey; Claudia Tebaldi of the National Center for Atmospheric Research in Boulder, Colorado; and Heidi Cullen of the science and news organization Climate Central, headquartered in Princeton, New Jersey.
Highlights of the study are presented in this blog item. The full study and array of accompanying graphics are at this link: https://goo.gl/…
-- Bryce Anderson
The contiguous United States (CONUS) was exceedingly warm east of the Rockies in February 2017 ... monthly mean temperatures (were) up to 6 degrees Celsius (about 11 degrees Fahrenheit) above normal ... One of the most noteworthy aspects of the persistent February warmth in the U.S. can be seen in the ratio of record high to record low temperatures. According to the National Centers for Environmental Information (NCEI), more than 6,300 record daily highs were tied or broken compared to less than 130 record lows. As a result, February will go down as the 27th month in a row with more record highs than lows. For every low temperature record set there were 49 high temperature records set, the highest such monthly ratio since January 1920.
Averaging the monthly mean temperatures over the CONUS, 2017 was found to have the second-warmest February on record, behind that of 1954, with a temperature of 4.9 C (41 F). This was 3.1 Celsius (5.6 Fahrenheit) warmer than the 1981 to 2010 normal.
Two questions tied to this apparent warming are: 1) is it distinct from changes one would expect from natural weather fluctuations ("detection")? and 2) if so, can the causes behind the trend be ascertained ("attribution")? A combined analysis of the historical observations and climate models can help answer these questions.
The analysis of observations provides us with an estimate of how rare the event was, and how its likelihood has changed since the 1900s, a time for which we have good temperature observations but global warming had only just started. The causes of the observed changes can be assessed with climate model experiments with which one can isolate the effects of external factors known to influence the climate: human-caused greenhouse gas and aerosol emissions, and also solar variability and volcanic eruptions.
TREND IN THE OBSERVATIONS
We use the NOAA/NCEI series of CONUS temperatures derived from many thousands of weather stations. Earlier research has shown that this series is not affected materially by urbanization and siting; using only rural, well-sited stations actually gives slightly higher trends than the full database. Similar results are found if the analysis is repeated using other surface temperature data. These temperature anomalies occur mainly in the lowest kilometer of the atmosphere, with satellite observations of the lower troposphere showing anomalies that are about half the magnitude of the near-surface observations. Besides this well-known difference with surface observations, the satellite record is also too short to robustly determine trends in February CONUS temperatures or to assess changes since 1900. We have therefore not included it in this analysis.
The upper tail of February CONUS temperatures is described well by a Normal distribution. We assume that it varies to first-order proportional to the global mean temperature, GMST (smoothed), but allow a slower or quicker pace (this method was first used in van Oldenborgh, 2007 and is described fully in van Oldenborgh et al, 2015). The remaining variability is due to the random weather, with virtually no low-frequency variability left over. El Nino-Southern Oscillation (ENSO), the Pacific Decadal Oscillation (PDO) and Atlantic Decadal Variability (AMV) do not influence the CONUS temperature in February, simplifying the analysis.
This fit, excluding 2017, shows that the CONUS temperature in February has risen a bit faster than the GMST ... the chance of getting a temperature this high or higher is around 8% each year, with 95% certainty within the range 2% to 30%. Thus, in the current climate this was not really an exceptional event any more.
Around 1900, the return period would have been about 160 years, i.e., a probability of 0.6% , with an uncertainty from 0.05% to 1.4% each year. Thus, on the basis of the observed trend we find that the probability has increased by roughly a factor of 13. We have 95% confidence that the increase in likelihood of events like February 2017 is at least a factor of four. However, it is important to note that this increase isn't all attributable to climate change. For this attribution we rely on climate models.
A similar fit can also give the return period of the 1954 temperature. At that time it was a very unusual situation with a return period of roughly 1 in 200 year (about 0.5% chance). That also means, however, that the odds of finding such an event in 120 years of data are about 50-50, just like throwing a 6 in three throws of dice. The weather maps also confirm that February 1954 had much more unusual weather than February 2017, while the circulation for this year is not as unusual, adding support to the fact that the heating trend has made high temperatures like last month more commonplace than they were in the 1950s.
The next question is what caused the heating trend in the February CONUS observations over the last century. This question cannot be directly assessed using the observational record, with which we can only assess correlation and not causality, and must be assessed using climate models, in which the various external "forcings" (such as changes in solar radiation, volcanoes, greenhouse gas concentrations, etc.) can be controlled and their impact understood. We use a collection of coordinated experiments from climate models run at centers across the world under the "5th Coupled Model Intercomparison Project", or CMIP5, which were assessed in the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report, the data for which are publicly available. In addition to CMIP5 experiments, two ensembles of atmosphere-only models are used. Fifteen of the CMIP5 models and the two sea surface temperature (SST)-forced models have been run at least three times with and without human-caused emissions of greenhouse gases and aerosols, allowing us to isolate the effect of the influence of greenhouse gas changes on the likelihood of February warm spells within the climate models.
Historical and projected increases in greenhouse gases lead to increased incidence of warm CONUS Februaries, and decreased incidence of cold CONUS Februaries. The climate model simulations show an 18-year-return period (95% confidence interval between 12 to 37 years) for Februaries over the CONUS as warm or warmer than 2017, similar to the observational estimate. Without the historical human-induced increases in greenhouse gases, the "best estimate" of the return period in the models is around 60 years (95% CI between 27 to 184 years), or a tripling of its probability -- the lower bound on the 95% range of the probability change is 1.8x now from past greenhouse gas increases. The model estimate of the greenhouse-driven increase in warm February probability is within the uncertainty range of the observed increase in probability, though on the lower range of the observational estimates. Around 2050 temperatures like this are projected to be completely normal, occurring approximately every three years on average (95% CI between 2.6 to 4.5 years).
Of note is that even in the 2050 ensemble, the probability of "cold" winters over the CONUS remains: natural fluctuations in weather and climate are expected to continue generating cool conditions, though at a reduced rate. That is, long-term warming and increased incidence of warm events does not obviate cool events -- it just reduces their probability.
These modeling and observational results are not very surprising given the broad observational, theoretical and modeling background on the connection between increasing greenhouse gases, global and regional temperature changes, and temperature extremes. The character of mechanisms behind the observed rise of the global mean temperature has been studied extensively in the scientific literature, and has been assessed in the Fifth IPCC Assessment Report, leading to the conclusion that 'it is extremely likely that human influence has been the dominant cause of the observed warming since the mid-20th century'. For most of the world, local enhanced likelihood of warm spells can already be related to this global change, including warming over the contiguous United States in winter.
Results from global climate models are inconsistent with the hypothesis that the increase in odds of warm Februaries was caused by natural forcing agents such as solar activity, which has declined since the 1960s, and volcanic eruptions. Meanwhile the model results indicate that past historical increases in greenhouse gases have raised the odds of warm Februaries in the CONUS considerably. The observed trend is compatible with the effects of human-induced emissions of greenhouse gases. Since past and projected future greenhouse gas increases will continue to raise the temperatures, the frequency of winter months like February 2017 should be expected to increase over the coming decades.
Overall, we find that the chances of seeing a February as warm as the one experienced across the Lower 48 has increased more than threefold because of human-caused climate change. The record-warm February of 1954 was, at the time, a very rare event (probability about 0.5% per year) but similar events should now be expected every few years.
Bryce Anderson can be reached at email@example.com
Follow Bryce Anderson on Twitter @BAndersonDTN
© Copyright 2017 DTN/The Progressive Farmer. All rights reserved.