Unprecedented(?) Arctic warming. Part II

by Marcia Wyatt

UPDATE:  Addendum from Marcia Wyatt

UPDATE:  Giff Miller responds

Miller et al.’s 2013 paper – Unprecedented recent summer warmth in Arctic Canada – splashed into the public eye last week with the declaration that current average summer temperatures in the Eastern Canadian Arctic are warmer now than in any century in the past 44,000 years, and perhaps in the past 120,000 years.

The authors note that solar insolation today is 9% less than solar insolation during the early Holocene. They further reason that because current summer temperatures in the Eastern Canadian Arctic are greater than those during the early Holocene when solar insolation was greater, the only explanation for the current excess summer warmth can be found in anthropogenic greenhouse-gas emissions. They express confidence in ruling out natural climate variability as the reason for the warmer summers.

From a few samples of long-dead moss exposed by a couple of receding ancient ice caps atop Baffin Island to the conclusion of unprecedented global warming due to mankind, the reasoning is intricate. Many assumptions made.

Field research for this study was conducted on Baffin Island between 2005 and 2010. The first goal of the study was to determine the history of summer average temperatures in this Arctic region west of Greenland. Remains of ancient moss, recently exposed via current snowline retreat, reveal this history. Snowline is the elevation at which winter accumulation equals summer snow melt.

The reasoning behind using moss to determine the history of summer average temperatures begins with the fact that moss grows when the ground is ice-free. It grows until climate chills and the snowline advances. When the snow advances, it covers the moss, thereby killing it. Locked within the dead moss’s relics is carbon, some of which is C-14 – a radioactive isotope of carbon created from the interaction of galactic cosmic rays and nitrogen in the atmosphere. C-14’s half-life is 5730 years, allowing this chemical signature to be used to date the last ‘breath’ of the once living carbon-consumer. Complications adjusted for, reasonable age estimates to ~50,000 years can be obtained. A sample’s radiocarbon age would mark either the last time the snowline had retreated to the current snowline’s position or the maximum age identifiable by radiocarbon dating.

What does snowline elevation have to do with temperature? Miller et al. point to a study by Koerner 2005 that finds strong correlation between ice-mass balance and summer temperatures in the Canadian High Arctic between 1960 and 2003. Thus, Miller et al. use snowline as a proxy for summer average temperature in this study. When the snowline was last at the elevation recently exposed is determined from the moss’s age. That is the time at which temperatures are inferred to have been matched by, or exceeded by, today’s temperatures.

How does one know if the moss was only now uncovered or if it had been uncovered one or more times between when it was alive and now? The researchers note that erosion rapidly removes any newly exposed moss; therefore, such a previous unveiling would be unlikely.

Miller et al. took hundreds of samples. Many were of young moss – moss that had re-grown on the area left bare by the retreating snow mass. Of the older samples, most of them dated back to ~5,000 years ago, in the mid-Holocene. A few samples dated back to 44,000 years. Limitations of radiocarbon dating cannot go much beyond this date, so this is the moss’s minimum possible age. Forty-four thousand years ago, Earth was in the midst of a glacial period and nearby Greenland was buried in ice – an unlikely time for sprouting moss on nearby Baffin Island, so the moss is likely older than 44,000 years. Information captured in oxygen isotopes in the Greenland Ice Sheet reflects timing of the last interglacial, when conditions were warm enough for moss growth. This was ~120,000 years ago. Putting all of this together, researchers infer the likely age of the moss from Baffin Island dates to the last interglacial, 120,000 years ago. If this moss had not seen daylight for 120,000 years, Miller et al. reason, then summer temperatures during the last 10,000 years (the Holocene) of the current interglacial have not been as warm as now; otherwise the moss would have been previously exposed. One might argue that snow accumulation could have been so great that temperatures as warm as or warmer than today may not have been able to etch away the pile clear down to the moss-covered ground. Miller et al. have an answer for this. They argue that the ice caps from which the samples of old moss were taken could never have been thicker than 70 meters. This, they say, is due to topography – a flat summit surrounded by steep slopes – and snow dynamics – a snow accumulation greater than 70m atop this summit could not be physically constrained. They further argue that modeled scenarios show that a 70m accumulation of snow would melt if a one-hundred-year stretch of warm temperatures had prevailed. Because the pile remains, that one-hundred-year stretch must not have occurred prior to now.  And now must be warmer than the early-to-mid Holocene. The next question to answer is why.

The research team attempts to ascertain cause for temperature amplitudes, both for today’s temperatures and those of the early-to-mid Holocene. Primarily, the question is, could the excess of today’s temperatures be due to natural forcing? What Miller et al. call ‘natural’ variability or forcing is actually no more than direct solar insolation. They look to the Holocene summer surface temperature record for answers. This requires use of snowlines. 1) First they must estimate snowline elevation changes between the early Holocene and now. 2) Then they must estimate what factors determined the snowline-elevation changes – summer surface air temperature or solar insolation. And 3) once they calculate the contribution of solar, then they estimate the surface temperatures.

Many of the collected moss samples were around 5,000 years old. Their elevations can be plotted against their sample ages. This exercise reveals the evolution of snowline-elevation changes. Of course, nothing is so straight-forward. It must be determined first just what snowlines can be considered to be ‘regional’, i.e. representative. Some moss samples do not follow the average snowline.

Koerner explains why. Ice-mass balance at low elevations has been increasing or staying the same over the 1960 to 2003 interval. This is because of increasing exposure of the Arctic Ocean, allowing for fog formation, which creates conditions of ice growth and persistence at low elevations. Miller et al. dismiss the low-elevation values, identifying them as outliers. Using their established ‘regional’ snowline, they estimate snowline elevations lowered between the mid (~5,000 years ago) and late Holocene, on average, about 650 meters.

Once they determined what snowline elevation was ‘regional’, and how much that snowline elevation had changed over the last 5,000 years, the next step was to assign cause to elevation changes. How much of the snowline change between the mid Holocene and the late Holocene was due to changes in surface air temperature? How much was due to natural causes – i.e. to short wave radiation from solar insolation? Today the latter contribution is considered to be minimal. In the early Holocene (b/n ~12 and 10 kya), solar insolation in June at 65ºN is estimated to have been ~9% higher than it is today. To answer the question about cause of snowline-elevation changes, the Greenland Ice Sheet is once again invoked. Model estimates of solar contribution to changes in snowline elevation on Greenland during the last interglacial (> 120,000 years ago) are scaled to current-day anomalous short-wave radiation in the Arctic. From this, the researchers determine that 95 meters of the estimated 650-meter snowline lowering between 5,000 years ago and the mid-20^th century could be assigned to solar, independent of surface air temperature. They adjust their snowline elevations to reflect the elevation change not due to solar short-wave radiation effects so they can next focus on reconstructing summer air temperature of the early-Holocene.

More assumptions are brought into the calculations. Using the present-day moist adiabatic lapse rate of ~6ºC/km [Correction: Lapse rate used ~4.9ºC/km (+/- 0.4ºC/km), derived from measuring lapse rates on glacier surfaces in Arctic Canadian region in summer], they estimate the inferred snowline lowering between 5,000 years ago and the mid-20^th century represents a decrease in summer temperature of 2.7 +/-0.7ºC. This estimate is based upon the assumption that there have been no significant changes in precipitation patterns, despite the observations of Koerner regarding the increasingly open Arctic Ocean’s impact on moisture, at least at low elevations. They invoke support for their estimates via borehole temperature profiles through the Greenland Ice Sheet. These measurements reflect a similar magnitude of temperature decrease. On the other hand, models in the 5^th Coupled Model Intercomparison Project (CMIP5) show smaller temperature decreases, on the order of only 0.7ºC to 1.4ºC. The authors suggest the models underestimate Arctic amplification. Continuing with the same method and assumptions, they determine that since 1960, the summer temperature has increased over 3.7ºC. Thus, it is concluded that present-day summer temperatures in the Baffin Island region are higher than those of the warm mid-Holocene. And if some moss samples were just recently exposed, the temperatures today are likely warmer than any since the last interglacial. Furthermore, they argue that because solar insolation was substantially higher in the early Holocene than today, and because temperatures are higher now than then, this must mean today’s excess warmth cannot be accounted for by natural variability.

A few points to consider before accepting this line of reasoning and conclusions: First, they cite the work of Koerner 2005 regarding the summer air temperature relationship to ice-mass balance in the Canadian High Arctic – a region including the Miller et al. study area of Baffin Island. Koerner points out that between 1960 and 2003 summer warming has been slight, less than 1.0ºC. This stands in contrast to the 3.7ºC increase assigned to the Baffin Island region over the same time frame. Trends of ice ablation are not necessarily apples to apples in the Koerner study, as some include the persistence of snow accumulations at low elevations due to fog resulting from the exposed Arctic Ocean; some do not include this persistent snow cover. Thus some trends of ice-mass-balance decrease statistically linked to increasing summer temperature are artificially steep. In addition, trends are not temporally consistent. Negative trends were strongest between 1980 and 2001, but the last two years of the Koerner study showed a weakened negative trend. He goes on to note that a single very negative-balance year (e.g. 1962) can have a disproportionate effect on trends, cancelling out the effect of several years of increasing snow levels. To illustrate that point, he notes that three very warm summers on the northwest side of Devon Ice Cap (1962, 1998, and 2001), together, cancel out the combined impact of all the positive-balance years in the other 43 years! Koerner also notes that there appears to be no trend in winter balance, despite modeled prediction of increased precipitation with increased temperatures. This is why he ultimately concludes that with no identified winter trend in ice-mass balance, the net changes of the ice measurements in this particular region are due to variability in summer climate, with the caveats mentioned above taken into consideration.

Koerner further cautions that the summer balance is evaluated over a very short time span – a two-to-three-month period. This summer ice-mass balance does not necessarily follow any annual temperature trend. His point is that annual temperatures should not be used in attempt to gauge impact of climate on glaciers in the Canadian High Arctic. He also notes that in regions outside the Canadian High Arctic, the role of summer climate in determining the net ice-mass balance variability is not as strong. In many regions, parts of Alaska, for example, winter climate dominates the calculation. Furthermore, in regions such as Svalbard and northern Scandinavia the influence of summer climate on the variability in the net ice-mass balance is not strong. So, with the seasonal diversity of behavior seen in different parts of the Arctic, is it reasonable for Miller et al. to assert such bold conclusions about anthropogenic global warming based on summer-temperature impact on ice melt in a small region of the Arctic?

Re summer melt on Baffin Island, Koerner found that the maximum rates of ice thinning and glacial retreat occurred in the Penny and Barnes Ice Caps on Baffin Island. These are the ice caps in the Miller et al. study. Koerner suggests that the higher thinning rates there are atypical, most likely due to the fact that these caps are essentially Pleistocene relics of the last glacial (~20,000 years old) – remnants of the Laurentide ice sheet from the last glacial interval that covered much of Canada. They have been continually thinning throughout the Holocene (Fisher and Koerner, 2003). Koerner’s work concludes that the ice-mass balance in some of the study area has weakly decreased over the last 100 years, with a greater increase over the last 50, but not as much decrease as occurred during the early Holocene 10,000 years ago. While Miller et al. base their use of snowline elevation as a summer temperature proxy on Koerner’s work, much of Koerner’s conclusions appear to diverge from those of Miller et al.

What about the warming? The Holocene Thermal Maximum (HTM) is often assigned the age between ~10kya and 5kya; yet maximum temperatures occurred at different places at different times. Solar insolation played a large role in the HTM, but that role was modified temporally, spatially, and in magnitude depending on feedback responses related to land-surface changes, sea-ice distribution, ocean-current dynamics, and atmospheric circulation patterns.

During the early Holocene, at the end of the last glacial period, Earth’s axis tilt was at a maximum and precession positioned the Earth so that it was closest to the sun during the Northern Hemisphere’s summer. Total annual insolation b/n 12 and 10 kya was ~1W/m^2 higher at 60ºN and 5W/m^2 greater at the pole than it is today. During the month of June, that translates to 10% higher insolation at 60ºN, according to Kaufman et al. 2004. Miller et al. 2013 cite Kaufman et al. regarding the observation that insolation at the NH poles was strong and conditions were conducive to ice melt in summer in the early Holocene. But this is the general overview. Kaufman et al. go further than this. They note that the peak summer insolation occurred b/n 12 and 10 kya, not 5,000 years ago when many of the Baffin Island dead-moss samples last were exposed. Clearly, insolation effects were not felt at the same time in all regions of the Arctic. And clearly, the link between solar insolation and timing of the warmest summer temperatures over Baffin Island must be viewed within this more detailed context.

The Pacific sector of the Arctic was relatively in-phase with insolation changes in the Holocene; the Atlantic sector, including Baffin Island and the Canadian High Arctic, was not. Changes there in response to solar insolation were delayed by thousands of years.

According to Kaufman et al., whose research was based on numerous and diverse proxy data across the Arctic region within the western hemisphere, because much of the Arctic region of northwest North America and westward remained unglaciated during the last glacial maximum, when solar insolation peaked in the following interglacial, warmth was concentrated here, where land surfaces could absorb, rather than reflect, incoming radiation. Furthermore, no ocean exchange took place between the Arctic and the Pacific through the Bering Strait due to its elevation relative to the low sea level of the time. This absence of flow impacted ocean circulation in both the Pacific and Atlantic Oceans. Peak warmth occurred b/n 12 and 9 kya in the Pacific/northwest North American region, followed by continued, albeit lesser warming, b/n 10 and 8 kya, coinciding with resumption of flow through the Bering Strait. While warmth was pronounced in this region, with greater ice retreat in places like the Brooks Range and north-central Alaska than today, it was likely moderated over time by rising sea levels that converted continental interiors to maritime environments in the region. The rising waters also increased the moisture content of the troposphere over the region.

To the east, in the Canadian High Arctic, home of Baffin Island, influence of the residual Laurentide Ice Sheet, which was retreating slowly to the northeast, delayed peak warming there for another couple of millennia, peaking between 7 and 5 kya, and in some of the northern areas of this region, peak warmth did not occur until about 3.7 kya – a time when solar insolation had decreased considerably since earlier in the Holocene. Delayed response to that early Holocene occurrence of strong insolation is attributed to influence of the large residual ice masses (related to Laurentide Ice Sheet). They directly affected atmospheric advection and indirectly affected ocean circulation, mostly through impact on the North Atlantic/Arctic Ocean freshwater balance and exchange. General circulation models estimate that the effect of the ice-sheet remnants on atmospheric-circulation moderated the solar-insolation-induced warming by 2ºC over the northeastern North American Arctic and downstream. This modeled estimate may be low, as it does not consider the additional indirect effects on ocean circulation and the exchange of water between the Arctic and North Atlantic – a feature critical to multidecadal climate variability during the 20^th century. Where Atlantic inflow was blocked, as in the region of the Canadian High Arctic, temperatures remained cool. Where inflow was not blocked, temperatures increased by up to 5ºC (Koc et al. 1993). This natural variability related to feedback responses of the coupled ocean-ice-atmospheric system clearly is a strong factor in climate of the Arctic – then and now. It cannot be accurately assessed from estimates of solar insolation intensity – this latter quantity meaning little without specification of season, location, and distribution of that insolation, and meaning even less without consideration of the temporal and spatial variety of feedbacks to that insolation.

With no straight-forward pattern of uniform Holocene warming emergent, only a time-transgressive behavior of warming, paired with strong spatial diversity, especially longitudinally, Kaufman et al. note that the spatial pattern of warming observed in the early to mid Holocene resembles the modern-day pattern. The common denominator is the influx of the North Atlantic Ocean into the Arctic Ocean and its subsequent impact on sea ice, and in turn, on sea-level-pressure in the Arctic High, thereby affecting atmospheric circulation patterns. Regions tended to cool in the same order as they warmed. Could this be a long-term-scale version of a stadium-wave???  Kaufman et al. conclude with the following: ‘the longitudinally asymmetric pattern of warming during the early Holocene exemplifies the contrasting response of the Pacific and Atlantic sectors to symmetrical forcing’. On multidecadal timescales, the stadium-wave hypothesis attempts to put these asymmetries in context. How it might operate on longer term time scales is unknown. My interpretation of work by Kaufman et al. suggests it may be worth exploring.

Can Miller et al. reasonably support their conclusion dismissing natural forces as strongly contributing to the warmth of today’s temperatures, basing this conclusion on summer air temperatures inferred from snowline-elevation changes in one small area of the Canadian High Arctic, where maximum temperatures in the past followed maximum natural variability (i.e. insolation) in the area by thousands of years? This conclusion carries with it the implication that the varying feedbacks responding to a natural forcing play no additional natural role in scripting climate profiles. While anthropogenic greenhouse emissions, along with anthropogenic modifications of land surfaces and other related changes, no doubt influence surface air temperatures, the arguments laid forth in this study fall short of making a strong case for their dominant role, in my view.

 JC comments:  This is a guest post, please keep your comments on topic and civil.

UPDATE:  Giff Miller responds, sent via email:

Miller responds

Wyatt does an admirable job of describing our recent work.  However, there are several errors in her report that weaken her skepticism regarding our conclusions
1. We do not use the present day moist adiabatic lapse rate.  Rather we use a more conservative figure derived from surface measurements on glaciers over a wide elevation range in the Eastern Canadian Arctic, as described in our article.  This gives a significantly lower magnitude of late Holocene cooling.

2. Snowline may change due to increased accumulation or greater melt (warmer summers).  We minimize the effect of increased accumulation by noting that there is no trend in annual accumulation layers from adjacent Greenland over the past 8000 years, as summers have cooled, nor over the historical period when temperatures have been warming.  Hence summer temperature is the dominant determinant of snowline over our 5 ka record

3. Wyatt questions the magnitude of current summer warming that we derived from our snowline records and NASA repeat lidar altimetry that shows the nearby Penny Ice Cap is losing mass at all elevations.  Koerner’s 2005 paper did not have temperature records of the past decade that are much warmer than any other decade in the record. Furthermore, our calculation of recent summer temperature increase is similar to that independently derived from adjacent Greenland, as we note in our paper.

4. We never claim that our data demonstrates Arctic-wide unprecedented warming, despite what Wyatt writes.  Read the ms carefully, please.  Our current research is expanding this study to other Arctic regions to evaluate the spatial domain of the Baffin Island pattern

5. Wyatt writes that our study was done on the Penny and Barnes ice caps.  This is false.  None of our samples came from either of these ice caps, which are known to have persisted through the HTM. Look at the figures in the paper carefully, please.  Koerner wrote his paper more than a decade ago….at that time, he went on the best available data.  New data are now available.  That’s the way knowledge progresses

6. During the early Holocene Earth’s axial tilt was not at its maximum [Clarification: Axial tilt was greater during the early Holocene than it is today (although not at its maximum)].  But Earth was closest to the Sun during NH summer, which resulted in 9% more insolation in June and July than at present 11 ka.  By 5 ka, as we note in our paper, that difference was down to 5% as Earth moved farther from Sun in NH summer.  It was this slow, but steady reduction in insolation since 11 ka, that eventually lowered the snowline to where it intercepted the summits where mosses were growing; highest summits first, lowest summits last.

7. Most proxy records from the Eastern Canadian Arctic show peak warmth in the earliest Holocene consistent with peak insolation, and gradual, but irregular cooling subsequently, especially after 5 ka.  There is an extensive literature on this.  Few records show peak warmth after 5 ka.
8.  You are left with our observations that some small, thin ice caps did not melt during the early Holocene warm period but melted in 2010 (the year we collected the “old” samples).  Summer insolation is now 9% less than the peak warmth of the early Holocene.  Of the primary factors determining the planetary energy balance, GHG remain by far the most likely term to explain such unusual summer warmth.

Addendum from Marcia Wyatt: