Nicolle et al “Climate variability in the sub-arctic area for the last two millennia”
Summary: Data from 56 sites within the Arctic 2k database was aggregated into 3 regions (North Atlantic, Siberia and Alaska) and the composite curves analyzed for overall trends and cyclical variability. The data comprises a mix of archives and proxies (ice core, tree rings, lake cores, a speleothem and some marine records) at different temporal resolutions, although only the highest resolution data (35 sites) was analysed for cyclical variability.
The paper is generally well written and structured, with figures and tables that are clear and illustrative of the text. However, I can see a number of issues with the study that would tend to dilute the conclusions.
Main Issues
1. The authors combine archives/proxies (ice core, tree rings, lake cores, a speleothem and some marine records) that respond to different seasons (winter, spring, summer, annual) and different climate variables (temperature, precipitation, sst, ice cover, p-e etc). The authors acknowledge the weak relationship with mean annual temperature in many of these proxies, and consequently choose to use the raw uncalibrated data. This data is then normalized (see point 4 below) and then combined into 3 regional time-series. By mixing seasons and climate variables into one metric for ‘climate’, it is not clear what the resulting combined record represents, or how it should be interpreted. The authors appear to discuss the combined record as if it were comparable to mean annual temperature, but it seems more likely that the 3 composite records reflect the complex amalgamation of temporal, spatial and variable bias of the component records of each dataset. I agree that major impacts on the climate system such as greenhouse gas induced global warming can be found in many different elements of the climate system, over large regions and through different seasons, but this type of forcing is quite exceptional over the last 2000 years. This makes the last 100 years a poor analogue for justifying this type of approach over longer time scales in the past, and explains why records that show good coherence over the last 100 years show much greater variance over longer time periods.
2. The authors combine these sites/records into 3 regional records. It is not clear how and why these regions were chosen. A basic assumption is that all of the sites in the one area should be responding in the same way to the same climate phenomena. The problem is that this assumption may not hold true, especially if changes in atmospheric or ocean circulation are involved that involve positive and negative responses within the same region. For instance, the NAO is the primary mode of temperature variability in the authors ‘Atlantic’ region. A positive (negative) NAO would result in a positive (negative) temperature anomaly at sites located in Scandinavia, and a negative (positive) temperature anomaly for sites located in Greenland and NE Canada. By combing sites from both these areas into one region, the overall climate response when viewed through the composite of the records will be neutralized. The authors show a similar example for the late Holocene trend in their Siberian data, where sites show contrary trends.
3. Another problem is the varying length of the records, more especially the annual records (Figure 1). Combining sites of different record lengths across a spatial network will generate variability irrespective of any actual climate change, simply through variable spatial sampling. The authors do not appear to have addressed this problem, perhaps by monte-carlo sampling of the dataset to show the robustness of the reconstruction to the recombination of the sites from which it is composed. This problem may also have an impact on the variance (Figure 2). Averaging together more sites will generally decrease the variance, so therefore the number of site/records will influence the variance as well as the climate. Most of the time series appear to show increased variance in the past when fewer records were available.
4. The use of normalization. The authors first normalize the records before combining them. This allows proxies that reflect different climate parameters (temperature, precipitation etc) to be combined based on the same units, but at the same time this removes information about the magnitude of change. For instance, is it right to view (for instance) a 20mm increase in precipitation in one record as the same as a 1C increase in temperature in another record? Similarly, should one site that shows a 1C increase have the same weight as another record that shows a 10C increase for the same event? This also has consequences for when these records are combined, since the sign of the response may not be in the same direction. For instance if a climatic event causes warming and drying in a certain area, a temperature sensitive record will show a (positive) increase in temperature, but a precipitation sensitive record will show a (negative) decrease in precipitation, potentially cancelling each other out when these records are combined into the ‘regional’ record. As a result of these issues, I am not sure what if anything we can make of the variability recorded in the composite records.
More specific issues: [delete] {add}
Page 1, row 19-20: Cooling is visible in the Siberian region at two sites, and warming at the others.
Page 1, row 24-25: ..did not [a] show {a} relationship..
Page 1, row 29: [apparently] {probably}
Page 4, row 12: ..due to local effect{s}
Page 6, paragraph 1: Alkenone v Diatom records also show substantial divergence during the Holocene, other factors include sensitivity to different depths (Jansen et al 2009 The Early to Mid‐Holocene Thermal Optimum in the North Atlantic, Hessler et al 2014 Implication of methodological uncertainties for Mid-Holocene sea surface temperature reconstructions).
Page 7, row 15: ..difficulty [to] {in} distinguishing..
Page 7, row 17: ..to [the] {a} negative..
Page 7, row 19: ..in [a] palaeohydrological..
Page 7, row 27: ..in all region[s]..
Page 8, paragraph 2: The instrumental period barely covers a couple of oscillations of periodicity 50-90 years. This is the advantage of this type of study if it can actually demonstrate periodicities are persistent on longer timescales.
Page 8, row 18: ..which [an] {are} important..
Page 8, row 21: ..in [the] sea ice..
Page 8, row 22: ..in [the] sea ice..
Page 9, paragraph 3: I think you should make it clearer that the dataset is a mix of climate variables and seasonal sensitivities, and that while at time similar to the mean annual temperature trend, it cannot be said that it is a record of annual mean temperature.
Page 9, row 20: As far as I understand it, the AMO is a marine phenomena and it has not been shown to be recorded in the terrestrial record. |