Data-Model Comparisons of the Tropical Hydroclimate Response to the 8.2 ka Event with an Isotope Enabled Climate Model
Abstract. The 8.2 ka Event, a prominent climate anomaly that occurred approximately 8,200 years before present (8.2 ka), has been the subject of extensive research due to its potential implications for understanding the characteristics and mechanisms of abrupt climate change events. We characterize the tropical hydroclimate response to the 8.2 ka Event based on a multiproxy compilation of 61 tropical hydroclimate records and assess the consistency between the reconstructed hydroclimate changes and those simulated by a new isotope-enabled climate model simulation of the 8.2 ka Event with iCESM. The timing and duration of the hydroclimate anomalies is calculated using two event detention methods, one of which uses a new changepoint detection algorithm to account for age uncertainty. When age uncertainties are explicitly accounted for, significant hydroclimate anomalies associated with the 8.2 ka Event are detected in 30 % of the records in the compilation, with a mean onset age of 8.28±0.12 ka (1σ), mean termination age of 8.11±0.09 ka (1σ), and mean duration of 152±70 years (1σ; with a range of 50–289 years), comparing well with previous estimates, and lending support to a regionally-variable tropical hydroclimate response to the 8.2 ka Event, with events that span decadal to multi-centennial timescales in the proxy record. Notably, the hydroclimate anomalies are not hemispherically uniform, but rather display rich regional structure. Anomalous conditions are characterized by pronounced isotopic enrichment across East Asia, South Asia, and the Arabian Peninsula. In the Americas, drying and isotopic enrichment occurred in southern Central America, contrasting with isotopic depletion in central/eastern Brazil. In contrast, no robust signatures of the 8.2 ka Event were found over the Maritime Continent. Many of these regional patterns generally agree with the new set of iCESM simulations of the 8.2 ka Event. In iCESM, the North Atlantic meltwater forcing leads to a broad southward shift in tropical rainfall, resulting in a generally drier Northern Hemisphere and wetter Southern Hemisphere, but with large regional variations in precipitation amount and the isotopic composition of precipitation. Over the oceans, the precipitation δ18O anomalies are generally consistent with the ”amount effect”, wherein areas characterized by drying have more isotopically enriched precipitation and areas of wetting have more isotopically depleted precipitation. However, the precipitation δ18O anomalies are more decoupled from changes in precipitation amount over land. iCESM captures many of the regional hydroclimate responses observed in the reconstructions, including the large-scale isotopic enrichment pattern in precipitation δ18O in South and East Asia and the Arabian Peninsula, drying and isotopic enrichment in precipitation δ18O in southern Central America, isotopic depletion in parts of northeastern South America, and a muted hydroclimate response in the Maritime Continent. Overall, this study provides new insights into the tropical hydroclimate response to the 8.2 ka Event, emphasizing the importance of accounting for age uncertainty in the hydroclimate reconstructions and the value of using isotope-enabled model simulations for data-model intercomparison.
General comments
In this paper, Moore and colleagues have thoroughly analyzed the signal of the Holocene 8.2 ka event in hydroclimatic proxy records from the tropics. The proxy-based signal is compared to simulation results obtained with the isotope-enabled iCESM global climate model. In the compilation, proxy records from different environments are considered: marine sediment records, lake sediment records and speleothems. As such this work extends the compilation recently published by Parker and Harrison (2022), who focused on the 8.2 ka signal in speleothem records. Innovative aspects of this manuscript are the extensive analysis of the impact of age uncertainties and the comparison of the proxy-based hydroclimatic anomaly with a state-of-the-art isotope-enabled model simulation of the 8.2 ka event. The paper is well-written, with a clear structure, and includes high-quality figures. The main result is a convincing proxy-based overview of the hydroclimatic response to the 8.2 ka event in the tropics, showing clear regional variability. In my view, the manuscript requires some minor revisions, mostly associated with the detection methods used and the climate model experiment. I provide details below.
Main comments
Introduction, Line 59. “This period occurred during the otherwise stable Holocene epoch (11,700 years ago to present) and was driven by the discharge of around 163,000 km3 of meltwater from proglacial Lakes Ojibway and Agassiz (remnants of the Laurentide Ice Sheet) into the North Atlantic”. The volume and source of the freshwater perturbation are actually still under discussion. For instance, the volume of the proglacial lakes is under discussion (Törnqvist and Hijma, 2012). In addition, Gregoire et al. (2012) argue that a collapse of the “ice saddle” over Hudson Bay could have played an important role in forcing the 8.2 ka event. It is important to make the reader aware of this uncertainty. See also the recent discussion by Aguiar et al. (2021). Besides, it is confusing to refer to Lakes Ojibway and Agassiz as “remnants of the Laurentide Ice Sheet”, so I suggest removing this part of the sentence.
Section 2.3.1. MM method. I propose including a schematic figure explaining how the MM method of detection works in practice, using two examples: one example with multiple events of the same sign and one example with multiple events of different signs. I am thinking of an extended version of Morrill’s Figure 2. It is not clear to me what the actual modification is to the Morrill method. Please specify. If feasible, including such a schematic for the actR method would also be helpful for readers not familiar with this type of changepoint analysis.
Section 2.4. Please make clear that the experimental setup for the iCESM simulation of the 8.2 ka event does only partly follow the PMIP4 design reported by Otto-Bliesner et al. (2017). In the PMIP4 design, the 8.2 ka simulation starts from an experiment with fixed forcing for 9.5 ka instead of 9 ka as is done here. In addition, in the PMIP4 design the freshwater hosing is applied in the Labrador Sea instead of across the entire northern North Atlantic.
Discussion. In some regions, for example the Caribbean and SE Asia, there are records without any detected change located close to records with a clear signal for the 8.2 ka event (see for instance Figure 2). How to interpret the absence of a signal in these records without significant change? It would be insightful to discuss the possible reasons for the absence of a signal.
Section 4.2. Simulated 8.2 ka event. The results of the last 50 years of the “hose” experiment are taken to represent the modelled 8.2 ka event. The mapped difference between “hose” and “ctrl” is used, for instance in Figure A2 and in the model-data comparison. It would be informative to know how the climate evolved through time in “hose” relative to “ctrl” (for instance related to detectability in the model result, see next point), so I propose including additional figures with modelled time-series of hydroclimate for the key regions shown in Figures 4 to 7. These additional figures could show 100-year time-series of both “ctrl” and “hose”.
Detectability of the simulated 8.2 ka event. The question of detectability of the 8.2 ka event is not only relevant for proxy records, but also for the model results. The difference between “hose” and “ctrl” includes anomalies produced by both internal variability and forced variability (i.e. by hosing). Ideally, one should not perform just one simulation for the 8.2 ka event as is done here, but rather an ensemble experiment in which the members differ only in initial conditions and have identical freshwater forcing (hosing). The ensemble mean allows then to analyze the forced response and to separate it from the internal variability. I refer to Wiersma et al. (2011), who performed this analysis for the temperature response associated with the 8.2 ka event. I realize that it is probably not feasible within the framework of this study to perform several additional ensemble members with iCESM, but in my view the aspect of detectability in the model results should also be discussed in the paper. In addition, I suggest performing a statistical test to determine what results of “hose” are significantly different from the “ctrl” climate, and to include only those significantly different results in the model-data comparison.
Other comments
Line 87. Why is it a critical tool? Please elaborate.
Line 104. How is this sensitivity to hydroclimate determined?
Line 151. Why 10 years? On what is this based?
Line 166. Why 50 or 100 years? Why not taking just 50 years?
Line 183. In the actR method, the time window for detection (7.9 to 8.3 ka) is different from the window used in the MM method (7.9 to 8.5 ka). Why this difference and what is the consequence? In addition, is there also a minimum duration for detection in the actR method?
Line 212. Are the forcings for the “ctrl” simulation identical to the 400-year-long 9ka simulation? If not, what are the differences? And is the climate stable in “ctrl” or is there is there still a trend in the surface conditions, signifying adjustment to the different forcings?
4.3 Data-Model comparisons. It is noteworthy that the regional structure of the 8.2 ka event found in this study resembles the hydroclimatic anomaly in a simulation of the Younger Dryas cold event in which freshwater forcing applied in the North Atlantic also plays an important role (Renssen et al. 2018). The hydroclimatic response for the Younger Dryas shows wetter conditions in the Caribbean, SE South America, Southern Africa and Madagascar, but drier conditions in S Central America, the Arabian Peninsula and SE Asia, broadly consistent with the proxy-based signal of the 8.2 ka event provided in the present paper in Figure 2.
Line 238. “… a lack of agreement of the sign or presence of an event”. Do you mean a lack of agreement between the two detection methods? Please clarify.
Figures 2-7. The grey symbols are very hard to see, so I suggest improving their visibility.
Figures 3-7. What do the grey isolines represent? Please explain.
Figures 4-7. In several cases, there are no proxy records of a specific type in a region (for example marine records in East Asia, Figures 4c and f). It is not very meaningful to show the simulation results in these cases, as there is no data-model comparison possible. So, I propose to only show the maps of key regions if there are proxy records available for the data-model comparison.
Figure 5. Legend is missing.
Figure A2a and b. According to the figure caption, the contours in these figures indicate the “range of temperatures in the “ctrl” simulation over the full 100 years”. It is not clear what temperatures these isotherms represent, so I do not see how these contours are useful, and I suggest removing them from the figure.
Additional references
Gregoire, L. J., et al. (2012) Deglacial rapid sea level rises caused by ice-sheet saddle collapses. Nature 487:597 219-222. https://doi.org/10.1038/nature11257.
Renssen, H., et al. (2018) The global hydroclimate response during the Younger Dryas event. Quaternary Science Reviews 193, 84-97. https://doi.org/10.1016/j.quascirev.2018.05.033
Törnqvist, T.E., Hijma, M.P. (2012) Links between early Holocene ice-sheet decay, sea-level rise and abrupt climate change. Nature Geoscience 5, 601-606. https://doi.org/10.1038/NGEO1536`.
Wiersma, A.P., et al. (2011) Fingerprinting the 8.2 ka event climate response in a coupled climate model. Journal of Quaternary Science 26, 118-127, https://doi.org/10.1002/jqs.1439.