the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Summer Greenland Blocking in observations and in SEAS5.1 seasonal forecasts: robust trend or natural variability?
Abstract. Given its impact on enhanced melting of the Greenland ice sheet, it is crucial to assess changes in frequency and characteristics of summer Greenland blocking. Indeed, the occurrence of such atmospheric pattern has seen a marked increase in recent decades: however, the observed trend is not captured by any simulation from state-of-the-art global climate models. It is therefore paramount to determine whether the lack of trend is caused by a misrepresentation of key physical mechanisms in climate models or whether such trend is mainly attributable to decadal variability, or both. Here we investigate Greenland blocking characteristics in reanalysis (ERA5) and ECMWF seasonal forecasts (SEAS5.1), showing that about 10 % of the 1000 permutations of SEAS5.1 runs can simulate a 43-year trend equal or larger to the ERA5 one: this suggests that the initialization and the higher model resolution contribute to a more realistic representation of the blocking dynamics than in freely-evolving climate runs. To further investigate these aspects, we apply the Peter and Clark momentary conditional independence (PCMCI) algorithm to assess monthly causal pathways. Results show that while the relationship among Arctic temperature, snow cover, Atlantic multidecadal variability and Greenland blocking is consistent both in ERA5 and SEAS5.1, the effect of early snow melt over North America on Greenland blocking is mostly absent in SEAS5.1. Therefore, while it is possible that the observed trend is due to internal decadal variability, the misrepresentation of the snow cover processes may explain the difficulty that SEAS5.1 has in reproducing the observed trend. This deficit in representing the snow impact on the atmospheric circulation might also be the culprit of the missing trend in climate models, raising the question whether long-term projections underestimate a future increase in Greenland blocking and ice melt.
- Preprint
(3753 KB) - Metadata XML
-
Supplement
(3987 KB) - BibTeX
- EndNote
Status: open (until 25 Feb 2025)
-
RC1: 'Comment on egusphere-2024-3998', Anonymous Referee #1, 14 Feb 2025
reply
Review of “Summer Greenland Blocking in observations and in SEAS5.1 seasonal forecasts: robust trend or natural variability?” by Beckmann et al.
Summary: This manuscript compares Greenland blocking in ERA5 and the ECMWF seasonal forecast model SEAS5.1. Causal networks and causal inference are used to compare blocking dynamics, as hypothesized by Preece et al 2023, between ERA5 and SEAS5.1. They determine that the effect of North American snowmelt is lacking in the seasonal forecast model, and suggest this could also be missing from climate models, and suggest this as a reason for the inability of climate models to capture the recent high GB period.
I think the use of seasonal forecast models to understand natural variability is interesting, and the causal networks are a nice application here. I particularly like the break down of blocking in to northern and southern components. However, there are some flaws with the manuscript and I have a few issues I’d like addressed before this manuscript is ready for publication
Major comments
1. Observations and reanalysis are repeatedly conflated in the paper. Reanalysis is still a model-derived product, and its snow cover is biased (e.g. Mudryk et al 2015,Mortimer et al 2020) when comparing to in-situ and observation-derived gridded products. This makes me wonder how dependent the results in this paper are on the use of ERA5 as ‘observations’, and I’d recommend first that the authors are more careful about their use of the word observations, and second that some discussion around how ERA5’s biases could be impacting the results. I also wonder whether other reanalysis products would be able to reproduce the same causality? Or whether a different metric for GB would yield similar results, both for causality and for how unusual the reanalysis trend is. I wonder as well where there is a state-dependence and how that might come in to play, for example a non-linearity when future snow cover over North America is much lower on average?
2. Despite the title, quite a lot more time is spent on the idea that there is a forced positive trend in GB, driven by the Preece et al 2023 mechanism, rather than the idea that natural variability (in particular anything other than the AMV), or even a forced increase in variability, has caused the trend in reanalysis. Evidence from CMIP6 is that the forced trend is negative with a lot of variability super-imposed, and so even if the Preece et al 2023 mechanism is correct and is missing from models, it’s not obvious to me that that means the models are wrong in the direction of their trend. Perhaps the forced trend for GB is not driven from the pole, but rather from the lower latitudes (on balance) and that’s the source of the decline in future GB? I do agree, however, that a missing mechanism that increases GB variability on an interannual timescales could still be important for future Greenland melt, and I do think that the results here are useful science, I’m just not sure about the way it has been framed.
3. The intro and the conclusions are both long and meandering at times between forcing of GB between the tropics, midlatitudes and poles, and between climate models and observations. Please consider re-writing to make it clearer.
4. I don’t think using T2m-Arctic as an indicator for Arctic amplification is sufficient. A difference between the Arctic and some mid-latitude band would probably be better, as a year with high T2m Arctic could also have high temperatures in general, i.e. T2m Arctic is highly correlated with T2m global. In general, I think the term Arctic amplification is used when the authors intend to say Arctic warming, so I’d recommend more careful wording.
Minor comments
L143: Is the mean of each month for the entire period removed from that month? Following sentence is obvious and need not be included.
L150 Why isn’t April one of the initialisations for SEAS5.1?
L155: Everything after ‘Liner correlation should be moved to the section 2.2
Figure 2: It’s interesting that there’s a reversal in the positions of ERA-40 and ERA-81 in terms of their percentile between GBI and GGI. The red lines do not look to be correlated in (c) and (d), as in Figure 1(c). Is there is a mistake in the plot or in the caption? Why is GHGS and GHGN written on panels (c) & (d)?
Figure 3: I wonder if a difference plot of (b)-(a) would be helpful for visualising where ERA5 and SEAS5.1 differ
Figure 4: (j) It’s interesting that all the members are so tightly constrained for Snow-Nam compared to other fields, and I wonder why that might be, and if its showing a related issues, whereby the seasonal model is not simulating variability in snow cover properly?
Paragraph L 396: non-significant correlations can’t support a relationship, the only thing that’s been shown there is that Arctic temp and GB are correlated.
L429: Why does a seasonal forecast model have lower signal-to-noise ratios?
Technical comments:
There are quite a few typos, missing words, and instances of poor grammar throughout the paper. I will highlight a few examples here but there are far too many and I would recommend a more thorough edit and grammar check before re-submission.
L15 incorrect use of colon
L21 climate runs -> climate model runs
L29 ice melt -> ice sheet melt
L100 of representing blocking -> to represent blocking
L115 casual -> causal
L162 their identification -> its identification
L174 I think the use of ‘condition’ isn’t the correct word, as those are the three equations after L180, what is here is a definition.
L225 Need to define s.d.
Section 2.3 doesn’t need its own subsection
L251 yearly – seasonal
L253 variability -> spatial variability
Figure 2 has (a) & (b) have (9095) squished together.
L339-340: ‘above its own 1 s.d.’ is awkward wordings
L279 & elsewhere: running mean -> running mean trend.
L 475 ‘higher’ -> lower? As the sign is negative? Same with snow anomalies below
L509: keep -> keeping
L530 – 545: issues with font, I think arising each time ‘beta’ is written.
References
Mudryk, L. R., C. Derksen, P. J. Kushner, and R. Brown, 2015: Characterization of Northern Hemisphere Snow Water Equivalent Datasets, 1981–2010. J. Climate, 28, 8037–8051, https://doi.org/10.1175/JCLI-D-15-0229.1.
Mortimer, C., Mudryk, L., Derksen, C., Luojus, K., Brown, R., Kelly, R., and Tedesco, M.: Evaluation of long-term Northern Hemisphere snow water equivalent products, The Cryosphere, 14, 1579–1594, https://doi.org/10.5194/tc-14-1579-2020, 2020.
Citation: https://doi.org/10.5194/egusphere-2024-3998-RC1 -
RC2: 'Comment on egusphere-2024-3998', Anonymous Referee #2, 17 Feb 2025
reply
General comments
This paper represents a crucial effort in identifying potential shortcomings in model representation of recent Greenland blocking trends and is therefore an important contribution to the literature. I found the authors’ methods to be well-suited to the objectives of the paper and thought that the conclusions were mostly well-founded. My main critique is that more careful consideration and in-depth discussion of the implications of the authors’ choice of blocking detection method are warranted. Given that much of the work demonstrating a positive trend in Greenland blocking – including the Preece et al. 2023 hypothesis that this work tested – was based on a field-departure-based blocking index, how might the application of a reversal-based detection method in this study impact the results presented herein and how they compare to previous works?
The authors’ use of the deconstructed components of the reversal-based blocking index in their causal discovery method was particularly novel and interesting; however, I do wonder if too much emphasis was placed on the GHGN criterion in distinguishing between anticyclonic conditions and Greenland blocking. For example, L441 states, “Thus, in ERA5 our findings generally support the first part of Preece23’s hypothesis, showing that T2M-Arctic and May snow cover may influence MSLP over North America, which in turn favours pressure highs over Greenland (GHGS>0). However, this does not consistently lead to blocking, as MSLP also contributes to GHGN>0, reducing the likelihood of blocking.” I’m not sure that this distinction is quite so definitive. For example, Tyrlis et al. (2021) argue that high-latitude blocks such as those that impact Greenland are distinct in that they shift the jet stream to the south and, consequently, requiring strong westerly flow to the north may not be appropriate. They argue that the poleward geopotential height gradient criterion should be relaxed to 0 m per degree latitude for locations north of 60◦N latitude. How might this argument impact the interpretation of the seemingly contradictory links with the GHGN index revealed by the authors?
Tyrlis, E., Bader, J., Manzini, E., & Matei, D. (2021). Reconciling different methods of high-latitude blocking detection. Quarterly Journal of the Royal Meteorological Society, 147(735), 1070–1096. https://doi.org/10.1002/qj.3960
Specific comments
L38: Add a comma after “pattern”
L43: Add a closing en dash between the in-text citation and “seem”
L47: I think this would read clearer as “Greenland blocking is a large-scale atmospheric high-pressure, low-vorticity system located over Greenland that is associated with the negative phase of the North Atlantic Oscillation (Woollings and Hoskins, 2008).
L71: I suggest rewriting this as “There is accumulating evidence that the frequency of summer Greenland blocking has increased over the last two decades…”
L79: Rewrite as, “leaving open the possibility that the increase is a consequence of natural variability.”
L92: replace “contribute to inhibiting” with “inhibit”
L100: replace “ability of representing” with “representation of”
L113: replace “identifying” with “identify”
L116: replace “identifying” with “identify”
L152: should “till” be “until”?
L184: What is the reason for extending the domain as far east as the prime meridian? Why start the southern bound of the domain at 67N?
L185: The Greenland Blocking Index, or GBI, has already been well established with a specific definition of the average 500 hPa height within the domain of 60-80N and 20-80W. I strongly suggest that the name here is altered to distinguish the index defined herein from the established GBI. Perhaps something as simple as the reversal-based Greenland blocking index (rGBI).
L255: replace “irrespectively” with “irrespective”
L322: Here you note that 33% of the GGI>1 s.d. T2m-G fall above the 90th climatological quantile in SEAS5.1-03; however, Figure 3f indicates 35.3% fall above the 90th climatological quantile. Which is correct?
Figure 3: The meaning of the red shading and the text annotations in panels (e) and (f) should be noted in the figure caption.
L347: replace “months is summer” with “summer months”
L371: I believe AVM should be AMV here
Figure 4: The caption title is a bit confusing. Do the time series in the right column show the 11-year running mean of index values or an 11-year moving window trend of monthly-mean index values? The units at the top of each plot would suggest the latter, but the caption title suggests the former.
L408: GBI is repeated here. Should one of these be GGI?
L418-421: The stationary wave response should increase as the background westerly flow weakens due to Arctic amplification. This could explain why the relationship with NA snow cover anomalies is stronger in the ERA5-81 record.
Hoskins, B., & Woollings, T. (2015). Persistent Extratropical Regimes and Climate Extremes. Current Climate Change Reports, 1(3), 115–124. https://doi.org/10.1007/s40641-015-0020-8
Coumou, D., Di Capua, G., Vavrus, S., Wang, L., & Wang, S. (2018). The influence of Arctic amplification on mid-latitude summer circulation. Nature Communications, 9(1), 2959. https://doi.org/10.1038/s41467-018-05256-8
L462: FDR has not been defined
L475: I believe this should be eastern US, not western US
Figure 6: More explanation is needed in the figure caption. What is the meaning of the numbers on the linkage arrows? Why do some connecting lines not include an arrow head? Why are there two color bars included at the bottom of the figure (i.e., what does each bar correspond to?) I see that this information is given on L232, but it would be helpful to have it in the caption as well.
L530: replace “snow cover on North America” with “North American snow cover”
Figure 7a: Where does this CEN diagram come from? Is this based on the analysis summarized in Figure 6? If so, why is the lag-0 linkage b
Citation: https://doi.org/10.5194/egusphere-2024-3998-RC2
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
132 | 26 | 6 | 164 | 17 | 2 | 2 |
- HTML: 132
- PDF: 26
- XML: 6
- Total: 164
- Supplement: 17
- BibTeX: 2
- EndNote: 2
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1