A Historical Perspective on Arctic Warming: Part One
Posted on 28 January 2014 by robert way
During her most recent Senate testimony, Dr. Judith Curry (Georgia Tech) repeated one of the most common misconceptions found in the blogosphere, that the Arctic was warmer than present during the 1940s. This period - known as the Early Century Warm Period (ECWP) - coincides with observations of reduced Arctic sea ice cover and allowed for more widespread ship navigation than during the late 1800s and early 1900s (Johanessen et al. 2004).
There are two elements to the contrarian views on the ECWP in the Arctic. First, they argue that during the ECWP the Arctic was warmer than present. Secondly they have used the ECWP as a means of casting doubt on the main drivers of global warming. These contrarians argue that internal climate variability caused the ECWP and that this internal variability may have contributed to recent Arctic warming, thereby suggesting that climate sensitivity to greenhouse gases may be lower than current estimates. Some of these discussion points have also somehow found themselves in the IPCC AR5's Chapter 10 where the following claim is made.
"Arctic temperature anomalies in the 1930s were apparently as large as those in the 1990s and 2000s. There is still considerable discussion of the ultimate causes of the warm temperature anomalies that occurred in the Arctic in the 1920s and 1930s."
Based on previous examination of the surface temperature record and also reading the literature on the topic, I found myself skeptical of this IPCC claim and by extension the contrarian views. Tamino expressed a similar sentiment in a recent article. In this post I will be examining the first element of the discussion and will evaluate whether "Arctic temperature anomalies in the 1930s were apparently as large as those in the 1990s and 2000s" is an accurate statement.
The challenge with describing Arctic surface air temperature changes is that the observational network is sparse, something we noted and corrected for in Cowtan and Way (In Press). Using a single observational network therefore has the potential to mislead - particularly on short timescales. However, comparison of all available long surface temperature records for the Arctic (here defined as regions North of 60°N) shows relatively strong agreement amongst the various products (Figure 1).
Figure 1: Arctic annual surface air temperature changes from ~1900 to ~2013 relative to the 1901-2000 baseline. (Top Panel) Combined Land & Ocean air temperatures; (Bottom Panel) Land only air temperatures.
From the above graph it is also apparent that some temperature anomalies between 1930 and 1950 were well above the 20th century average, but they do not match the magnitude of those observed over the past decade for any complete record. Over longer timescales (120 months/10 years) this difference becomes more apparent with no records showing similar warmth to present in the Arctic during any previous period (Figure 2). One record (NansenSAT; Johannsen et al, 2008) shows greater mid-century warmth and less recent warmth relative to the other datasets; this dataset includes 20th century data from both Russian drift stations and Argo buoys, but the land station data comes from CRUTEM2v which has limited Arctic coverage. It should be noted that the record terminates in 2008 before several of the warmest years in the Arctic.
Figure 2: Centered rolling mean (120-month/10-year) of Arctic surface air temperatures from ~1900 to ~2008 (truncated at both ends) relative to the 1901-2000 baseline. (Top Panel) Combined Land & Ocean air temperatures; (Bottom Panel) Land only air temperatures.
The Arctic surface temperature record presented by CW2014 is the most complete spatially due to its incorporation of interpolation (e.g. kriging) and has been validated against both Arctic buoys and satellite records during the recent period. It also shows a greater warming during the ECWP than other Land+Ocean records therefore it is retained for comparing the ECWP and recent Arctic warming.
To compare the relative distributions of monthly anomalies we take the warmest 120 month period (10-year) during the ECWP and compare it to the warmest 120 month period during the recent warm period (Figure 3). Comparison of these two periods reveal a clear shift in the average air temperatures and also an increase in the probabilities of warm months over the past decade. Recently, it has also become increasinly rare for even a single month to have a below normal average temperature in contrast with the ECWP where this was common.
Figure 3: Comparison of warmest 120-month (10-year) periods in Arctic surface air temperatures during the early century warm period and recent warm period using the CW2013 dataset. (Left Panel) Density plots showing the frequency of temperature anomalies during both periods; (Right Panel) Boxplots showing the minimum/maximum values, lower/upper quartiles and medians for both periods. This figure was updated due to mistakenly using HadCRUTv4 in the original.
Based on the data presented above there is virtually no evidence that Arctic air temperatures were greater than present during any previous period of the last century. This is clearly a case where the IPCC should consider amending its text to provide a more accurate picture of Arctic temperature changes. In Part Two the Early Century Warm Period will be discussed in the context of its causes and origins.
CW2013 = Cowtan and Way Long-Kriged Global Temperature Product (After Cowtan and Way, In Press). Discussed here.
GISS = Goddard Institute for Space Studies (Hansen et al. 2010).
HadCRUT/CRU = Hadley Climate Research Unit (Jones et al. 2012; Morice et al. 2012).
NOAA/NCDC = National Oceanic and Atmospheric Administration's National Climate Data Center (Lawrimore et al. 2011; Vose et al. 2012;).
BEST = Berkeley Earth Surface Temperature Project (Rohde et al. 2013).
NansenSAT = Nansen Environmental and Remote Sensing Center (Johanessen et al. 2004; Kuzmina et al. 2008).