the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Dec 2022
Unraveling the burial and exhumation history of foreland basins using the spread of apatite (U-Th-Sm)/He single grain ages
Abstract. Reconstructing the evolution of foreland basins that experienced late exhumation is challenging due to an incomplete sedimentary record. Thermochronometry has been applied successfully to reconstruct basin evolution, but the method is subject to uncertainties. For the Swiss Molasse Basin, a wide range of exhumation magnitude and timing has been proposed based on thermochronometry. We aim to reduce uncertainty by dating larger numbers of grains and samples, to obtain statistically more robust data. New apatite (U-Th-Sm)/He (AHe) data from a single borehole shows ages of 4 to 30 Ma in the upper 500 meters and ages of 3 to 80 Ma below 1300 meters. This is counterintuitive as a total reset is expected at depths exceeding approximately 600 m. To arrive at a single consistent thermal history including our and previously published data, we conduct thermal modeling with different software. In particular we test the influence of different provenance histories and distinguish between cooling associated with changes in heat flow vs changes in exhumation.
We determine 1050 m +/- 100 m of exhumation, starting slowly at 13 Ma and accelerating at 9 Ma. Coinciding with exhumation, heat flow begins to rise sharply, causing heating until 5 Ma, despite ongoing exhumation. We show that this discrepancy between start of exhumation and start of cooling is the main reason for differing estimates for the burial and exhumation history of the basin. We suggest that the remaining misfit between modeled and measured Molasse AHe ages can be explained by post-Miocene hydrothermal flux in the Neogene sediment fill above a sealing layer, potentially the Opalinus Clay or Triassic evaporites.
In summary, we show that a single consistent model for basin exhumation relies on large sets of grains and samples, as well as inclusion of provenance ages in the models. With timing of the main exhumation phase constrained to start at 9 Ma, we can rule out a 5 Ma climatic event as exhumation driver. As the region is not affected by extensive faulting, deep seated processes related to mantle dynamics remain as exhumation driving process.
- Preprint
(3568 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2022-1323', Anonymous Referee #1, 12 Jan 2023
Dear Authors
This paper potentially provides new thermochronological data that could help in the clarification of thermal history of Swiss Molasse. I could appreciate the efforts made to explain the complexity of the dataset, including the use of different modelling approaches. However, I found some important issues that need to be properly addressed before considering this paper for publication. Detailed comments are on the annotated pdf but they can be summarized as follows.
- Your discussion on possible causes for overdispersion of AHe data is rational but you did not consider implantation from surrounding host U-Th rich minerals. As it has been well demonstrated (e.g. Spiegel et al., 2009, EPSL; Murray et al., 2014, Chem. Geol.), implantation results in anomalous “old” ages. This could be the case of your BST and RL samples, above all given that you describe the presence of iron coating. Other possible causes (radiation damage, zonation) cannot be discarded a priori but their effect on the AHe system is less important. Thermal history of BST and RL is characterized by a very long permanence at shallow depths (i.e. at temperatures mostly lower than 100°C) and this enhances different kinetics (whatever is the cause).
- Discussion of HeFTy and QTQt results is not always correct, especially when you refer to the lack of sensitivity to heat flow variations. These programs are not designed to take into account petrophysical properties and heat flow as they only consider mineral kinetic and time-temperature data as inputs.
- Vitrinite data are not properly shown and discussed. They represent usually the most reliable tool for paleotemperature reconstructions so I think that they give a strong constraint to the thermal history. I have not the exact values but I played a bit with another software and other solutions are possible. By the way, both HeFTy and QTQt give the possibility to model VR data. Why did you not use them? Finally, the idea of possible inherited VR values is plausible but it should be verified by analysis of raw data.
I stopped my reading at chapter 5.3 as I think that a serious re-evaluation of data is necessary before going into discussion of geological implications.
-
RC2: 'Comment on egusphere-2022-1323', Peter van der Beek, 31 Jan 2023
First of all, I should apologise for the time it has taken to get this review back to you. The last two months have been exceptionally packed, but that should not be an excuse for making a PhD candidate wait so long for a manuscript review …
General comments
This paper presents new apatite (U-Th-Sm)/He (AHe) data from a well in the northern part of the North Alpine Foreland Basin (NAFB) and uses three different modelling approaches (HeFTy on single samples, QTQt on the ensemble of well data and PyBasin also on the combined well data) to infer an internally consistent thermal and burial/exhumation history for these data. A surprising aspect of the data is that AHe dates from Permian (Buntsandstein and Rotliegendes) sandstones at >1300 m depth in the well show a much wider spread and thus appear less reset than those from the Miocene molasse at depths <500 m. Unfortunately, the intervening 800 m of Mesozoic stratigraphy consists mainly of carbonates and is not amenable to AHe analysis. Nevertheless, this interesting new observation could potentially provide strong constraints on permissible thermal histories but is not fully exploited in the manuscript (see below).
The question of the timing, spatial pattern and particularly the drivers of late-stage exhumation in the NAFB has been the subject of discussion and controversy since Cederbom et al. (2004) demonstrated its widespread nature and suggested a climatic driver. Several groups have presented new data and alternative interpretations. At the same time, how to interpret often quite complex patterns of dispersed AHe dates (from both well- and outcrop samples) to infer an internally consistent thermal history is an important research question in the thermochronological community. This manuscript thus addresses two important issues and is therefore potentially very suitable for a journal like Solid Earth.
However, I do feel that the current manuscript has some weaknesses that need to be addressed before it can be published. Some of the implicit messages of the manuscript (even though these are not explicitly written up as conclusions) are that: (1) pre-depositional exhumation history and variable thermal conductivity/heat capacity of the stratigraphic units in the well are more important in explaining dispersion of AHe dates than kinetic differences (due to variation in eU and/or grain size); (2) PyBasin does a better job of modelling these well data than either HeFTy or QTQt. However, neither of these inferences are robustly substantiated and (2) is really based on a bit of a strawman argument.
Concerning (1): the potential influence of kinetic variation in the data is insufficiently explored. The manuscript starts out by inferring a lack of clear correlation of AHe dates with either grain size or chemistry (section 4.2, Fig. 4), and also argues that data inversions using HeFTy and QTQt did not show significant differences when using either the standard Farley (2000) model or the Flowers et al. (2009) model that incorporates radiation-damage control on He-diffusivity, characterised by eU. However, an apparent lack of correlation of AHe dates with either grain size or eU does not mean the dates are not influenced by these parameters. The problem is that they are not independent; for instance, a large grain with low eU or a small grain with high eU would both result in intermediate dates that would not contribute to a clear correlation with either parameter separately. It would make more sense to show the variation in ages as a function of both parameters jointly and to perform a multiple correlation rather than separate correlations for each parameter. Secondly, from the HeFTy and QTQt modelling result presented, the reader cannot assess the potential influence of kinetic variability. More importantly, it appears from Figs. 6 and 8 that inverted time-temperature histories using these codes do not start at temperatures above the nominal AHe closure temperature – it may be that the figures simply do not show the full modelled thermal history but if that is the case it should be made clear. If models were started below the AHe closure temperature that fundamentally affects the predicted AHe ages, in particular when using radiation-damage-dependent kinetic models. See Ketcham et al. (2018a) as well as very recent reviews by Abbey et al. (2023) and Murray et al. (2022) for a more complete discussion of these issues. In short, if the inversions were started below the AHe closure temperature they are incorrect and should be redone.
Concerning (2), the manuscript makes some incorrect statements about both HeFTy and QTQt when building its case for using PyBasin instead. For instance, it is argued that HeFTy cannot jointly model different samples from a well suite, although this has been incorporated in more recent versions of the code (see Ketcham et al., 2018b). Similarly, the manuscript states that QTQt cannot incorporate variable pre-depositional exhumation histories, while this is done in Wildman et al. (2021). Overall, these oversights give an impression of a lack of acquaintance with the recent literature on the subject, which should be corrected.
There is a more profound and philosophical issue concerning the use of these modelling tools: both HeFTy and QTQt are used in inverse mode and are fundamentally “data driven”, i.e., the inputs to these models are solidly known (AHe dates, grain sizes and compositions, present-day temperature, geological constraints on when parent rocks were at the surface, etc.). In contrast, PyBasin is a forward model and is much more “scenario driven”, i.e., based on what we think we know. Specifically, in PyBasin, besides the above inputs, the heat-flow history, surface-temperature history and pre-depositional exhumation history are given as inputs and thus considered known. However, what are the real constraints on these? Some numbers are given but no justification or critical discussion of them is provided. These fundamental inputs and their uncertainties should be discussed in much greater detail than what is currently done in the manuscript. Also, the “robustness” (or more precisely the resolution) of the forward PyBasin modelling approach should be assessed more in depth; the sensitivity to small changes in some of the input parameters is discussed but there may be significant trade-offs and I have the impression that the parameter space is not fully explored. This is an aspect where formal inversion techniques provide a clear advantage.
Specific comments (tied to line number)
76: I don’t think anyone specifically argued for (rock or surface?) uplift of the NAFB and in any case this cannot be inferred from the thermochronology data.
81-85: this section comes a bit too early here. It would be better to discuss the orogen and basin evolution first, and then go into details on the different interpretations of the burial, heat flow and exhumation history (see also comment on sub-section 2.3 below).
176-188 (sub-section 2.3): this sub-section is a bit awkward. I am not sure it is in the right place – it might be better placed after sub-section 2.4 – and, more importantly, it just provides some numbers without any context or justification. How did the earlier studies come up with these paleo-heat-flow values? Why present the values as heat-flow in the first place when this is always an inferred quantity (I’m pretty sure the referenced studies quantified paleo-temperatures and -depths, not a paleo-heat-flow)? What are the constraints and uncertainties? What is the link to the geological history? For instance, heating in the Late Jurassic/Early Cretaceous could potentially be linked to Valaisan rifting, as recently also argued for the western Alps (Célini et al., 2023). Did previous authors suggest a mechanism for the inferred Miocene-recent increase in heat-flow? I would suggest to combine this sub-section with the previous paragraph on estimates of basin erosion/exhumation at the end of the “Geological setting” section and expand it significantly, describing and discussing the constraints, uncertainties and mechanisms.
202: What is meant by “tectonic overprint” here? Is this a fault? If so, can its offset be estimated? This is important as later in the manuscript there are some conflicting statements about the level of tectonic disturbance in this part of the basin.
206-207: it would be useful to provide an estimated thickness of the locally eroded sediments here, i.e., how much additional OMM/OSM stratigraphy is preserved in the hills surrounding the drill site?
222: it would be honest to acknowledge here that 800 of these 1370 m were not sampled …
233-234: this is the first time I see this treatment mentioned in any thermochronology paper. This may need a bit more detail and justification. Are you sure this treatment does not affect the apatite grains (etching or removal of an outer layer)? Even affecting a few mm would have a significant effect on the a-correction!
242-247: the description of criteria for determining the robustness of AHe dates is unclear. First, how were analytical errors on single-grain ages determined (in general, single-grain-age error in AHe analyses is determined by repeat age determination of a standard and the relative error is thus the same for all grains)? Second, how is “improbably old” defined? Without a rigorous definition, this seems like a fairly subjective criterion. Third, do grains have to fulfil all criteria or only one of the criteria to be dubbed “questionable” or “unreliable”? The text says “and” twice but is it “and” or “or”? Finally, what is the problem with U-concentrations <15 ppm? This is fairly common in apatite. Unless your instrument is not very sensitive the U-concentration in itself should not be a criterion for dubbing an AHe date “questionable”.
269-270: this statement is incorrect; see general comment above.
281: it is unclear what you extrapolate and how to obtain this estimate of full AHe resetting below 600 m.
287-288: the downward decrease in age spread for the BST-RL samples is based on just 4 or 5 individual ages from the BST and RL1 samples, whereas a fairly large proportion of dates from the RL2 sample are dubbed unreliable or questionable. I’m not sure how robust this inference is…
302: what is the physical basis for using logarithmic fits to the date-eU and date-volume trends? Also, why use volume rather than equivalent-sphere radius? The two should be highly correlated but the propagated uncertainty is much higher for the volume.
302-320: since AHe dates depend on BOTH eU and grain size, discussing the correlations separately is somewhat misleading. A better approach is to show the variation as a function of both parameters and use multiple rather than single correlation (see also general comment above).
321-328: this change in Th/U signature between lower and upper molasse samples is interesting and could do with a bit more discussion.
383-392 (and Fig. 6): these models appear not to start above the nominal closure temperature for AHe (i.e. with a fully reset sample). If this is the case, they are incorrect and should be redone.
429: have you tested whether 10.000 model iterations is sufficient? How many of these were burn-in models?
431-437: are all these additional constraints really needed and are you sure they do not over-constrain the model outputs? You really only need the depositional constraint box at 15-30 Ma. Also, were these models started with fully reset AHe (do they go back beyond 70 Ma)? If not, they should be redone.
446-451: this paragraph is puzzling. Why would a history with converging cooling paths to the present day (fig. 7b) be “consistent” but not one with diverging cooling paths (Fig. 7a)? You later argue (in the PyBasin runs) for an increased geothermal gradient between the mid-Miocene and the present day. This is exactly what is predicted by the model in Fig. 7a, and would be expected when transitioning from deposition to exhumation in the basin.
452-453 and Fig. 8: are you showing the full model time-span here? If so, these models do not start above the nominal AHe closure temperature and should be redone. If not, show the full model (you can shade pre-depositional histories if you prefer).
457: but QTQt implicitly includes heat-flow changes over time (by allowing variable temperature differences between modelled samples) and can also include variable provenance (cf. Wildman et al., 2021).
471-475: some strong modelling choices are made here, without much justification. How important is it to include detailed rock-physical parameters, when you choose to ignore radiation-damage effects? There really should be some sensitivity analysis done here to justify these choices.
479-481: this is the only description of the constraints on provenance histories I could find, in the table legend. This is really not acceptable as these pre-depositional exhumation histories will strongly influence the modelled post-depositional history in the basin. This should be much better explained and justified, including uncertainties and how they are handled. As it is, this seems very pre-determined.
485-488: this is also a bit mystifying. How do you “adjust” surface temperatures for water depth? How well known are these surface-temperature histories in the first place? What effect do they have on the predicted burial/exhumation history? This seems to be putting a lot of emphasis (and probably too much confidence) in a second-order parameter while ignoring (uncertainties on) first-order parameters.
501-506: this paragraph is unclear because no uncertainties are provided and the fit to the data is not quantified. It is stated that the “best fit” is obtained for the numbers reported here but how is that defined? What is the fit parameter and how does it vary with changing input parameters?
510: pre-depositional histories are “defined”, but what do we really know about these? These should be treated as uncertain and some variability in the pre-depositional history should be allowed.
515: how were heat-flow histories “adjusted”; how much is really known; what are the uncertainties? It appears the model has many knobs that can be turned to eventually fit the data, but the problem is that we cannot assess whether the final best-fit parameters are reasonable. It would be good to show some constraint boxes on the heat-flow history (and potentially also the surface-temperature history) in Fig. 9.
532: What are the real constraints on surface heat-flow between the Miocene and the present day?
536: this is very hard to see in Fig 9e/9.2 (see comments on Fig. 9 below). The panel should zoom much more on the data and find a better way to show the modelled age span. This is a quite unsatisfactory result.
540-541: the problem of course is that this model will not fit the Permian data. You come back to this issue in the discussion, but it should probably be stated upfront here.
548-549: is the smaller magnitude of exhumation an input or output of this model?
574-575: I fail to see how a model with less total exhumation predicts AHe ages that are about 20 Ma older than the data. Panel 11c appears to show a better fit to the BST/RL samples than 11a (but the plots are so small it is hard to see). Again, you would need some quantitative fit criterion to make these statements.
590: Similarly, the age predictions of a model with a later start of exhumation do not look qualitatively worse than those of the preferred model.
601: again, plot 11g looks like a pretty good fit to the BST/RL data to me. It is unclear why these predictions are deemed worse fits than the preferred model.
613: and again, I fail to see why the results in panel 11i provide a “worse fit” to the Molasse data than those in panel 11a. You really need to state how you quantify the fit when making these statements.
629: similar to the above 4 comments.
632: given all of the above, the estimated uncertainty on the exhumation magnitude (“100 m at maximum”) appears overly optimistic. One can really only make such a statement with respect to a clearly defined target function for the misfit.
636-645 and Fig. 12: it is unclear how you can reason on the “gradient” of the VR data while Fig. 12e really just shows a small cloud of data. No obvious gradient appears from this data and it appears there may be significant degrees of freedom in modelling it. It may be useful to show the predictions of the base model of Fig. 9 as well here.
654-658: note that HeFTy can also be used on an ensemble of well data (Ketcham et al., 2018b).
659-660: how important are variable rock properties with respect to all the other unknowns in the model? There should be a discussion of this.
661: but is that applicable to the current study? There are no evaporites or coal in the modelled NAFB section right (although there is salt under the NAFB locally)?
663-664: this is a bit misleading, as it is very well possible to group samples in HeFTy according to age, kinetic parameter or inferred pre-depositional history. See Ketcham et al. (2018b) for an example.
671-674: I do not see the problem here (see also comment on lines 446-451). Non-parallel time-temperature paths between samples at different depths indicate a change in geothermal gradient through time, similar to what input in the PyBasin models (but here it is an outcome of the model, not an input, and therefore potentially more credible if it can be shown that these variations are consistent with independent data).
676-677: is this “broad range of equally viable possibilities” a problem or the reality? This just shows that the data in themselves are not able to constrain the pre-Cenozoic history better than indicated. Tighter constraints can be achieved by adding more data, but the robustness of and uncertainties in those data need to be carefully assessed.
682-683: argument (3) is incorrect; see Wildman et al. (2021).
683-686: argument (4) is also incorrect in my view. See comments on lines 446-451 and 671-674 above. The sentences in lines 684-686 are technically correct but QTQt allows variable temperature gradients over time, which is the same as varying heat flow.
687-695: the comparison between PyBasin, HeFTy and QTQt would be more balanced if it was also acknowledged that the forward-modelling approach in PyBasin requires making strong assumptions on for instance the heat-flow history and the pre-depositional exhumation histories of samples. This introduces significant non-uniqueness in the model predictions that can only be partly explored using sensitivity tests. Also, the approach carries the danger of overestimating what we really know.
698-706: the inferred cause for the high VR values may be possible, and other studies have pointed to late Jurassic – early Cretaceous heating due to Valaisan rifting (e.g., Célini et al., 2023). However, it is not clear how the gradient in VR data was defined (see also comment lines 636-645) and the statement that “the AHe system is not affected by these hydrothermal shocks as it is not sensitive enough to short temperature peaks” should be backed up with some quantitative evidence (see, for instance, Reiners et al., 2007 for an example of “kinetic crossover” between AHe and AFT but for ultra-short high-temperature bursts related to wildfires; not sure this can be extrapolated to fluid-flow events). The possibility that the Permian sequences were not influenced by Mesozoic fluid-flow events is viable but would also be more convincing if it were backed up by some independent data. Finally, yes there is the possibility that the VR is detrital but I would imagine an expert in this technique should be able to discriminate between detrital and diagenetic organic matter?
716: Not sure this is the ONLY explanation. You have shown that the BST/RL apatites are overall somewhat larger than the MOL apatites and also have overall somewhat higher eU (Fig. 4). As argued in the comments above, kinetic differences between these groups of apatites have been insufficiently explored.
726-727: this is not very clear from Fig. 9 – there may be a better way to show the fit to the data.
729-730: Modelling algorithms do not “expect” anything – they turn an input number into an output number … the following phrase and reference appear to imply that the importance of radiation damage in AHe kinetics has only been appreciated recently is a bit misleading – this has been known and worked on for at least 15 years (Shuster et al., 2006).
731-736: this is a strange (and I’m pretty sure incorrect) argument. Since the standard Farley (2000) He-diffusion algorithm was used in PyBasin, it is impossible to account for any radiation-damage effects, independent of whether you model (old) pre-depositional histories or not. You will include any He that might have been retained from the pre-depositional history in your models, but NOT any radiation-damage effects. I would strongly suggest you take this argument out.
744-751: this paragraph is not useful. Better skip it and go straight to the potential explanations in the NAFB context.
754: Where is the Hegau volcanic system? How important is it? This needs to be shown on Fig. 1 and discussed in the “Geological context” section.
756-758: How does basement fault reactivation lead to elevated heat flow? This is all quite unclear… there is a potentially simpler mechanism that is not discussed here: as the basin transitions from deposition to exhumation, downward advection is replaced by upward advection of rocks with respect to the surface, which would lead to an increased geothermal gradient. You could model the potential importance of this effect, given characteristic depositon and exhumation rates for the region.
774-777: OK, there is a possibility that some impermeable layers in the stratigraphy set up a fluid-flow system in the higher part of the section that was not felt by the Permian rocks. But there should ideally be some independent evidence for this in the well – I would suspect somebody has looked at indicators of fluid-rock interaction on these cores and cuttings? Also, it is not clear what the fact that some layers are a structural decoupling horizon has to do with the argument.
782: here you clearly state that the area is undeformed. How does that fit with the inference of a “tectonic overprint” on line 202?
807-810: in contrast to what was written in line 782, now an argument for exhumation along structures associated with Jura folding is made. This should be made consistent. Also, try to avoid the term “tectonic exhumation” which is strictly applicable only to exhumation due to extension along normal faults, without surface erosion. Where does the maximum amount of 100 m of exhumation related to tectonics come from? Note that Cederbom et al. (2011) already argued that any exhumation related to Jura deformation can only be minimal as the entire external NAFB was translated over a very low-dip decollement surface.
812: isostatic rebound after glaciation does not lead to exhumation.
821: While I am fine with the revised onset of exhumation proposed here compared to Cederbom et al. (2011 – on which I was a co-author) and actually somewhat relieved to see arguments against a potential “climatic driver” (not clear what it would have been in any case), I am not sure the argument that a later onset of exhumation would have necessarily led to complete resetting of the BST-RL samples is fully supported by the results shown here; I don’t think this has been demonstrated in this manuscript.
831-834: it is not clear what “geodynamic processes” would have led to the main exhumation phase starting at 9 Ma; this could be made more explicit and specific. Also the link with the Hegau volcanics (which are a pretty minor occurrence, right) appears somewhat overstated.
835-837: this final phrase on potential glacial influence is a bit mystifying and should be rephrased or taken out.
842: The Permo-Triassic samples are reset because they are all younger than the depositional age. They only show older ages than the MOL samples.
850-851: as you will have noted from many comments above, you have not convinced at least this reader that the results of this study are robust. This will need much more sensitivity testing and a proper incorporation of kinetic effects.
863: reorganisation of the Rhine River drainage system has not been discussed previously and comes out of the blue here. Either introduce this properly in the “geologic setting” section or drop this argument.
Minor comments on writing style etc. are in the annotated pdf file of the manuscript.
Comments on Figures
Figures 5-8: all representations of model predictions should also show the fit to the data.
Figure 9: Add constraint boxes (what is known?) to panels a and c. Panel e is superfluous and practically unreadable. Call panels 9.1 and 9.2 rather 9e and 9f. Zoom in on the depth range 0-1300 m in panel 9.2 so we can better see the fit to the data.
Figure 10: also show constraint boxes on heat-flow (and surface-temperature) history, and zoom in on the modelled data in panel c.
References (other than those in the manuscript):
Abbey, A. L., Wildman, M., Goddard, A. L. S. and Murray, K. E.: Thermal history modeling techniques and interpretation strategies: Applications using QTQt, Geosphere, doi:10.1130/ges02528.1, 2023.
Célini, N., Mouthereau, F., Lahfid, A., Gout, C. and Callot, J.-P.: Rift thermal inheritance in the SW Alps (France): insights from RSCM thermometry and 1D thermal numerical modelling, Solid Earth, 14(1), 1–16, doi:10.5194/se-14-1-2023, 2023.
Ketcham, R. A., van der Beek, P., Barbarand, J., Bernet, M. and Gautheron, C.: Reproducibility of thermal history reconstruction from apatite fission-track and (U-Th)/He data, Geochemistry, Geophysics, Geosystems, 19, 2411–2436, doi:10.1029/2018gc007555, 2018a.
Ketcham, R. A., Mora, A. and Parra, M.: Deciphering exhumation and burial history with multi-sample down-well thermochronometric inverse modelling, Basin Research, 30(Suppl. 1), 48–64, doi:10.1111/bre.12207, 2018b.
Murray, K. E., Goddard, A. L. S., Abbey, A. L. and Wildman, M.: Thermal history modeling techniques and interpretation strategies: Applications using HeFTy, Geosphere, 18(5), 1622–1642, doi:10.1130/ges02500.1, 2022.
Reiners, P. W., Thomson, S. N., McPhillips, D., Donelick, R. A. and Roering, J. J.: Wildfire thermochronology and the fate and transport of apatite in hillslope and fluvial environments, Journal of Geophysical Research, 112(F4), F04001-29, doi:10.1029/2007jf000759, 2007.
Wildman, M., Gallagher, K., Chew, D. and Carter, A.: From sink to source: Using offshore thermochronometric data to extract onshore erosion signals in Namibia, Basin Res, 33(2), 1580–1602, doi:10.1111/bre.12527, 2021.
Status: closed
-
RC1: 'Comment on egusphere-2022-1323', Anonymous Referee #1, 12 Jan 2023
Dear Authors
This paper potentially provides new thermochronological data that could help in the clarification of thermal history of Swiss Molasse. I could appreciate the efforts made to explain the complexity of the dataset, including the use of different modelling approaches. However, I found some important issues that need to be properly addressed before considering this paper for publication. Detailed comments are on the annotated pdf but they can be summarized as follows.
- Your discussion on possible causes for overdispersion of AHe data is rational but you did not consider implantation from surrounding host U-Th rich minerals. As it has been well demonstrated (e.g. Spiegel et al., 2009, EPSL; Murray et al., 2014, Chem. Geol.), implantation results in anomalous “old” ages. This could be the case of your BST and RL samples, above all given that you describe the presence of iron coating. Other possible causes (radiation damage, zonation) cannot be discarded a priori but their effect on the AHe system is less important. Thermal history of BST and RL is characterized by a very long permanence at shallow depths (i.e. at temperatures mostly lower than 100°C) and this enhances different kinetics (whatever is the cause).
- Discussion of HeFTy and QTQt results is not always correct, especially when you refer to the lack of sensitivity to heat flow variations. These programs are not designed to take into account petrophysical properties and heat flow as they only consider mineral kinetic and time-temperature data as inputs.
- Vitrinite data are not properly shown and discussed. They represent usually the most reliable tool for paleotemperature reconstructions so I think that they give a strong constraint to the thermal history. I have not the exact values but I played a bit with another software and other solutions are possible. By the way, both HeFTy and QTQt give the possibility to model VR data. Why did you not use them? Finally, the idea of possible inherited VR values is plausible but it should be verified by analysis of raw data.
I stopped my reading at chapter 5.3 as I think that a serious re-evaluation of data is necessary before going into discussion of geological implications.
-
RC2: 'Comment on egusphere-2022-1323', Peter van der Beek, 31 Jan 2023
First of all, I should apologise for the time it has taken to get this review back to you. The last two months have been exceptionally packed, but that should not be an excuse for making a PhD candidate wait so long for a manuscript review …
General comments
This paper presents new apatite (U-Th-Sm)/He (AHe) data from a well in the northern part of the North Alpine Foreland Basin (NAFB) and uses three different modelling approaches (HeFTy on single samples, QTQt on the ensemble of well data and PyBasin also on the combined well data) to infer an internally consistent thermal and burial/exhumation history for these data. A surprising aspect of the data is that AHe dates from Permian (Buntsandstein and Rotliegendes) sandstones at >1300 m depth in the well show a much wider spread and thus appear less reset than those from the Miocene molasse at depths <500 m. Unfortunately, the intervening 800 m of Mesozoic stratigraphy consists mainly of carbonates and is not amenable to AHe analysis. Nevertheless, this interesting new observation could potentially provide strong constraints on permissible thermal histories but is not fully exploited in the manuscript (see below).
The question of the timing, spatial pattern and particularly the drivers of late-stage exhumation in the NAFB has been the subject of discussion and controversy since Cederbom et al. (2004) demonstrated its widespread nature and suggested a climatic driver. Several groups have presented new data and alternative interpretations. At the same time, how to interpret often quite complex patterns of dispersed AHe dates (from both well- and outcrop samples) to infer an internally consistent thermal history is an important research question in the thermochronological community. This manuscript thus addresses two important issues and is therefore potentially very suitable for a journal like Solid Earth.
However, I do feel that the current manuscript has some weaknesses that need to be addressed before it can be published. Some of the implicit messages of the manuscript (even though these are not explicitly written up as conclusions) are that: (1) pre-depositional exhumation history and variable thermal conductivity/heat capacity of the stratigraphic units in the well are more important in explaining dispersion of AHe dates than kinetic differences (due to variation in eU and/or grain size); (2) PyBasin does a better job of modelling these well data than either HeFTy or QTQt. However, neither of these inferences are robustly substantiated and (2) is really based on a bit of a strawman argument.
Concerning (1): the potential influence of kinetic variation in the data is insufficiently explored. The manuscript starts out by inferring a lack of clear correlation of AHe dates with either grain size or chemistry (section 4.2, Fig. 4), and also argues that data inversions using HeFTy and QTQt did not show significant differences when using either the standard Farley (2000) model or the Flowers et al. (2009) model that incorporates radiation-damage control on He-diffusivity, characterised by eU. However, an apparent lack of correlation of AHe dates with either grain size or eU does not mean the dates are not influenced by these parameters. The problem is that they are not independent; for instance, a large grain with low eU or a small grain with high eU would both result in intermediate dates that would not contribute to a clear correlation with either parameter separately. It would make more sense to show the variation in ages as a function of both parameters jointly and to perform a multiple correlation rather than separate correlations for each parameter. Secondly, from the HeFTy and QTQt modelling result presented, the reader cannot assess the potential influence of kinetic variability. More importantly, it appears from Figs. 6 and 8 that inverted time-temperature histories using these codes do not start at temperatures above the nominal AHe closure temperature – it may be that the figures simply do not show the full modelled thermal history but if that is the case it should be made clear. If models were started below the AHe closure temperature that fundamentally affects the predicted AHe ages, in particular when using radiation-damage-dependent kinetic models. See Ketcham et al. (2018a) as well as very recent reviews by Abbey et al. (2023) and Murray et al. (2022) for a more complete discussion of these issues. In short, if the inversions were started below the AHe closure temperature they are incorrect and should be redone.
Concerning (2), the manuscript makes some incorrect statements about both HeFTy and QTQt when building its case for using PyBasin instead. For instance, it is argued that HeFTy cannot jointly model different samples from a well suite, although this has been incorporated in more recent versions of the code (see Ketcham et al., 2018b). Similarly, the manuscript states that QTQt cannot incorporate variable pre-depositional exhumation histories, while this is done in Wildman et al. (2021). Overall, these oversights give an impression of a lack of acquaintance with the recent literature on the subject, which should be corrected.
There is a more profound and philosophical issue concerning the use of these modelling tools: both HeFTy and QTQt are used in inverse mode and are fundamentally “data driven”, i.e., the inputs to these models are solidly known (AHe dates, grain sizes and compositions, present-day temperature, geological constraints on when parent rocks were at the surface, etc.). In contrast, PyBasin is a forward model and is much more “scenario driven”, i.e., based on what we think we know. Specifically, in PyBasin, besides the above inputs, the heat-flow history, surface-temperature history and pre-depositional exhumation history are given as inputs and thus considered known. However, what are the real constraints on these? Some numbers are given but no justification or critical discussion of them is provided. These fundamental inputs and their uncertainties should be discussed in much greater detail than what is currently done in the manuscript. Also, the “robustness” (or more precisely the resolution) of the forward PyBasin modelling approach should be assessed more in depth; the sensitivity to small changes in some of the input parameters is discussed but there may be significant trade-offs and I have the impression that the parameter space is not fully explored. This is an aspect where formal inversion techniques provide a clear advantage.
Specific comments (tied to line number)
76: I don’t think anyone specifically argued for (rock or surface?) uplift of the NAFB and in any case this cannot be inferred from the thermochronology data.
81-85: this section comes a bit too early here. It would be better to discuss the orogen and basin evolution first, and then go into details on the different interpretations of the burial, heat flow and exhumation history (see also comment on sub-section 2.3 below).
176-188 (sub-section 2.3): this sub-section is a bit awkward. I am not sure it is in the right place – it might be better placed after sub-section 2.4 – and, more importantly, it just provides some numbers without any context or justification. How did the earlier studies come up with these paleo-heat-flow values? Why present the values as heat-flow in the first place when this is always an inferred quantity (I’m pretty sure the referenced studies quantified paleo-temperatures and -depths, not a paleo-heat-flow)? What are the constraints and uncertainties? What is the link to the geological history? For instance, heating in the Late Jurassic/Early Cretaceous could potentially be linked to Valaisan rifting, as recently also argued for the western Alps (Célini et al., 2023). Did previous authors suggest a mechanism for the inferred Miocene-recent increase in heat-flow? I would suggest to combine this sub-section with the previous paragraph on estimates of basin erosion/exhumation at the end of the “Geological setting” section and expand it significantly, describing and discussing the constraints, uncertainties and mechanisms.
202: What is meant by “tectonic overprint” here? Is this a fault? If so, can its offset be estimated? This is important as later in the manuscript there are some conflicting statements about the level of tectonic disturbance in this part of the basin.
206-207: it would be useful to provide an estimated thickness of the locally eroded sediments here, i.e., how much additional OMM/OSM stratigraphy is preserved in the hills surrounding the drill site?
222: it would be honest to acknowledge here that 800 of these 1370 m were not sampled …
233-234: this is the first time I see this treatment mentioned in any thermochronology paper. This may need a bit more detail and justification. Are you sure this treatment does not affect the apatite grains (etching or removal of an outer layer)? Even affecting a few mm would have a significant effect on the a-correction!
242-247: the description of criteria for determining the robustness of AHe dates is unclear. First, how were analytical errors on single-grain ages determined (in general, single-grain-age error in AHe analyses is determined by repeat age determination of a standard and the relative error is thus the same for all grains)? Second, how is “improbably old” defined? Without a rigorous definition, this seems like a fairly subjective criterion. Third, do grains have to fulfil all criteria or only one of the criteria to be dubbed “questionable” or “unreliable”? The text says “and” twice but is it “and” or “or”? Finally, what is the problem with U-concentrations <15 ppm? This is fairly common in apatite. Unless your instrument is not very sensitive the U-concentration in itself should not be a criterion for dubbing an AHe date “questionable”.
269-270: this statement is incorrect; see general comment above.
281: it is unclear what you extrapolate and how to obtain this estimate of full AHe resetting below 600 m.
287-288: the downward decrease in age spread for the BST-RL samples is based on just 4 or 5 individual ages from the BST and RL1 samples, whereas a fairly large proportion of dates from the RL2 sample are dubbed unreliable or questionable. I’m not sure how robust this inference is…
302: what is the physical basis for using logarithmic fits to the date-eU and date-volume trends? Also, why use volume rather than equivalent-sphere radius? The two should be highly correlated but the propagated uncertainty is much higher for the volume.
302-320: since AHe dates depend on BOTH eU and grain size, discussing the correlations separately is somewhat misleading. A better approach is to show the variation as a function of both parameters and use multiple rather than single correlation (see also general comment above).
321-328: this change in Th/U signature between lower and upper molasse samples is interesting and could do with a bit more discussion.
383-392 (and Fig. 6): these models appear not to start above the nominal closure temperature for AHe (i.e. with a fully reset sample). If this is the case, they are incorrect and should be redone.
429: have you tested whether 10.000 model iterations is sufficient? How many of these were burn-in models?
431-437: are all these additional constraints really needed and are you sure they do not over-constrain the model outputs? You really only need the depositional constraint box at 15-30 Ma. Also, were these models started with fully reset AHe (do they go back beyond 70 Ma)? If not, they should be redone.
446-451: this paragraph is puzzling. Why would a history with converging cooling paths to the present day (fig. 7b) be “consistent” but not one with diverging cooling paths (Fig. 7a)? You later argue (in the PyBasin runs) for an increased geothermal gradient between the mid-Miocene and the present day. This is exactly what is predicted by the model in Fig. 7a, and would be expected when transitioning from deposition to exhumation in the basin.
452-453 and Fig. 8: are you showing the full model time-span here? If so, these models do not start above the nominal AHe closure temperature and should be redone. If not, show the full model (you can shade pre-depositional histories if you prefer).
457: but QTQt implicitly includes heat-flow changes over time (by allowing variable temperature differences between modelled samples) and can also include variable provenance (cf. Wildman et al., 2021).
471-475: some strong modelling choices are made here, without much justification. How important is it to include detailed rock-physical parameters, when you choose to ignore radiation-damage effects? There really should be some sensitivity analysis done here to justify these choices.
479-481: this is the only description of the constraints on provenance histories I could find, in the table legend. This is really not acceptable as these pre-depositional exhumation histories will strongly influence the modelled post-depositional history in the basin. This should be much better explained and justified, including uncertainties and how they are handled. As it is, this seems very pre-determined.
485-488: this is also a bit mystifying. How do you “adjust” surface temperatures for water depth? How well known are these surface-temperature histories in the first place? What effect do they have on the predicted burial/exhumation history? This seems to be putting a lot of emphasis (and probably too much confidence) in a second-order parameter while ignoring (uncertainties on) first-order parameters.
501-506: this paragraph is unclear because no uncertainties are provided and the fit to the data is not quantified. It is stated that the “best fit” is obtained for the numbers reported here but how is that defined? What is the fit parameter and how does it vary with changing input parameters?
510: pre-depositional histories are “defined”, but what do we really know about these? These should be treated as uncertain and some variability in the pre-depositional history should be allowed.
515: how were heat-flow histories “adjusted”; how much is really known; what are the uncertainties? It appears the model has many knobs that can be turned to eventually fit the data, but the problem is that we cannot assess whether the final best-fit parameters are reasonable. It would be good to show some constraint boxes on the heat-flow history (and potentially also the surface-temperature history) in Fig. 9.
532: What are the real constraints on surface heat-flow between the Miocene and the present day?
536: this is very hard to see in Fig 9e/9.2 (see comments on Fig. 9 below). The panel should zoom much more on the data and find a better way to show the modelled age span. This is a quite unsatisfactory result.
540-541: the problem of course is that this model will not fit the Permian data. You come back to this issue in the discussion, but it should probably be stated upfront here.
548-549: is the smaller magnitude of exhumation an input or output of this model?
574-575: I fail to see how a model with less total exhumation predicts AHe ages that are about 20 Ma older than the data. Panel 11c appears to show a better fit to the BST/RL samples than 11a (but the plots are so small it is hard to see). Again, you would need some quantitative fit criterion to make these statements.
590: Similarly, the age predictions of a model with a later start of exhumation do not look qualitatively worse than those of the preferred model.
601: again, plot 11g looks like a pretty good fit to the BST/RL data to me. It is unclear why these predictions are deemed worse fits than the preferred model.
613: and again, I fail to see why the results in panel 11i provide a “worse fit” to the Molasse data than those in panel 11a. You really need to state how you quantify the fit when making these statements.
629: similar to the above 4 comments.
632: given all of the above, the estimated uncertainty on the exhumation magnitude (“100 m at maximum”) appears overly optimistic. One can really only make such a statement with respect to a clearly defined target function for the misfit.
636-645 and Fig. 12: it is unclear how you can reason on the “gradient” of the VR data while Fig. 12e really just shows a small cloud of data. No obvious gradient appears from this data and it appears there may be significant degrees of freedom in modelling it. It may be useful to show the predictions of the base model of Fig. 9 as well here.
654-658: note that HeFTy can also be used on an ensemble of well data (Ketcham et al., 2018b).
659-660: how important are variable rock properties with respect to all the other unknowns in the model? There should be a discussion of this.
661: but is that applicable to the current study? There are no evaporites or coal in the modelled NAFB section right (although there is salt under the NAFB locally)?
663-664: this is a bit misleading, as it is very well possible to group samples in HeFTy according to age, kinetic parameter or inferred pre-depositional history. See Ketcham et al. (2018b) for an example.
671-674: I do not see the problem here (see also comment on lines 446-451). Non-parallel time-temperature paths between samples at different depths indicate a change in geothermal gradient through time, similar to what input in the PyBasin models (but here it is an outcome of the model, not an input, and therefore potentially more credible if it can be shown that these variations are consistent with independent data).
676-677: is this “broad range of equally viable possibilities” a problem or the reality? This just shows that the data in themselves are not able to constrain the pre-Cenozoic history better than indicated. Tighter constraints can be achieved by adding more data, but the robustness of and uncertainties in those data need to be carefully assessed.
682-683: argument (3) is incorrect; see Wildman et al. (2021).
683-686: argument (4) is also incorrect in my view. See comments on lines 446-451 and 671-674 above. The sentences in lines 684-686 are technically correct but QTQt allows variable temperature gradients over time, which is the same as varying heat flow.
687-695: the comparison between PyBasin, HeFTy and QTQt would be more balanced if it was also acknowledged that the forward-modelling approach in PyBasin requires making strong assumptions on for instance the heat-flow history and the pre-depositional exhumation histories of samples. This introduces significant non-uniqueness in the model predictions that can only be partly explored using sensitivity tests. Also, the approach carries the danger of overestimating what we really know.
698-706: the inferred cause for the high VR values may be possible, and other studies have pointed to late Jurassic – early Cretaceous heating due to Valaisan rifting (e.g., Célini et al., 2023). However, it is not clear how the gradient in VR data was defined (see also comment lines 636-645) and the statement that “the AHe system is not affected by these hydrothermal shocks as it is not sensitive enough to short temperature peaks” should be backed up with some quantitative evidence (see, for instance, Reiners et al., 2007 for an example of “kinetic crossover” between AHe and AFT but for ultra-short high-temperature bursts related to wildfires; not sure this can be extrapolated to fluid-flow events). The possibility that the Permian sequences were not influenced by Mesozoic fluid-flow events is viable but would also be more convincing if it were backed up by some independent data. Finally, yes there is the possibility that the VR is detrital but I would imagine an expert in this technique should be able to discriminate between detrital and diagenetic organic matter?
716: Not sure this is the ONLY explanation. You have shown that the BST/RL apatites are overall somewhat larger than the MOL apatites and also have overall somewhat higher eU (Fig. 4). As argued in the comments above, kinetic differences between these groups of apatites have been insufficiently explored.
726-727: this is not very clear from Fig. 9 – there may be a better way to show the fit to the data.
729-730: Modelling algorithms do not “expect” anything – they turn an input number into an output number … the following phrase and reference appear to imply that the importance of radiation damage in AHe kinetics has only been appreciated recently is a bit misleading – this has been known and worked on for at least 15 years (Shuster et al., 2006).
731-736: this is a strange (and I’m pretty sure incorrect) argument. Since the standard Farley (2000) He-diffusion algorithm was used in PyBasin, it is impossible to account for any radiation-damage effects, independent of whether you model (old) pre-depositional histories or not. You will include any He that might have been retained from the pre-depositional history in your models, but NOT any radiation-damage effects. I would strongly suggest you take this argument out.
744-751: this paragraph is not useful. Better skip it and go straight to the potential explanations in the NAFB context.
754: Where is the Hegau volcanic system? How important is it? This needs to be shown on Fig. 1 and discussed in the “Geological context” section.
756-758: How does basement fault reactivation lead to elevated heat flow? This is all quite unclear… there is a potentially simpler mechanism that is not discussed here: as the basin transitions from deposition to exhumation, downward advection is replaced by upward advection of rocks with respect to the surface, which would lead to an increased geothermal gradient. You could model the potential importance of this effect, given characteristic depositon and exhumation rates for the region.
774-777: OK, there is a possibility that some impermeable layers in the stratigraphy set up a fluid-flow system in the higher part of the section that was not felt by the Permian rocks. But there should ideally be some independent evidence for this in the well – I would suspect somebody has looked at indicators of fluid-rock interaction on these cores and cuttings? Also, it is not clear what the fact that some layers are a structural decoupling horizon has to do with the argument.
782: here you clearly state that the area is undeformed. How does that fit with the inference of a “tectonic overprint” on line 202?
807-810: in contrast to what was written in line 782, now an argument for exhumation along structures associated with Jura folding is made. This should be made consistent. Also, try to avoid the term “tectonic exhumation” which is strictly applicable only to exhumation due to extension along normal faults, without surface erosion. Where does the maximum amount of 100 m of exhumation related to tectonics come from? Note that Cederbom et al. (2011) already argued that any exhumation related to Jura deformation can only be minimal as the entire external NAFB was translated over a very low-dip decollement surface.
812: isostatic rebound after glaciation does not lead to exhumation.
821: While I am fine with the revised onset of exhumation proposed here compared to Cederbom et al. (2011 – on which I was a co-author) and actually somewhat relieved to see arguments against a potential “climatic driver” (not clear what it would have been in any case), I am not sure the argument that a later onset of exhumation would have necessarily led to complete resetting of the BST-RL samples is fully supported by the results shown here; I don’t think this has been demonstrated in this manuscript.
831-834: it is not clear what “geodynamic processes” would have led to the main exhumation phase starting at 9 Ma; this could be made more explicit and specific. Also the link with the Hegau volcanics (which are a pretty minor occurrence, right) appears somewhat overstated.
835-837: this final phrase on potential glacial influence is a bit mystifying and should be rephrased or taken out.
842: The Permo-Triassic samples are reset because they are all younger than the depositional age. They only show older ages than the MOL samples.
850-851: as you will have noted from many comments above, you have not convinced at least this reader that the results of this study are robust. This will need much more sensitivity testing and a proper incorporation of kinetic effects.
863: reorganisation of the Rhine River drainage system has not been discussed previously and comes out of the blue here. Either introduce this properly in the “geologic setting” section or drop this argument.
Minor comments on writing style etc. are in the annotated pdf file of the manuscript.
Comments on Figures
Figures 5-8: all representations of model predictions should also show the fit to the data.
Figure 9: Add constraint boxes (what is known?) to panels a and c. Panel e is superfluous and practically unreadable. Call panels 9.1 and 9.2 rather 9e and 9f. Zoom in on the depth range 0-1300 m in panel 9.2 so we can better see the fit to the data.
Figure 10: also show constraint boxes on heat-flow (and surface-temperature) history, and zoom in on the modelled data in panel c.
References (other than those in the manuscript):
Abbey, A. L., Wildman, M., Goddard, A. L. S. and Murray, K. E.: Thermal history modeling techniques and interpretation strategies: Applications using QTQt, Geosphere, doi:10.1130/ges02528.1, 2023.
Célini, N., Mouthereau, F., Lahfid, A., Gout, C. and Callot, J.-P.: Rift thermal inheritance in the SW Alps (France): insights from RSCM thermometry and 1D thermal numerical modelling, Solid Earth, 14(1), 1–16, doi:10.5194/se-14-1-2023, 2023.
Ketcham, R. A., van der Beek, P., Barbarand, J., Bernet, M. and Gautheron, C.: Reproducibility of thermal history reconstruction from apatite fission-track and (U-Th)/He data, Geochemistry, Geophysics, Geosystems, 19, 2411–2436, doi:10.1029/2018gc007555, 2018a.
Ketcham, R. A., Mora, A. and Parra, M.: Deciphering exhumation and burial history with multi-sample down-well thermochronometric inverse modelling, Basin Research, 30(Suppl. 1), 48–64, doi:10.1111/bre.12207, 2018b.
Murray, K. E., Goddard, A. L. S., Abbey, A. L. and Wildman, M.: Thermal history modeling techniques and interpretation strategies: Applications using HeFTy, Geosphere, 18(5), 1622–1642, doi:10.1130/ges02500.1, 2022.
Reiners, P. W., Thomson, S. N., McPhillips, D., Donelick, R. A. and Roering, J. J.: Wildfire thermochronology and the fate and transport of apatite in hillslope and fluvial environments, Journal of Geophysical Research, 112(F4), F04001-29, doi:10.1029/2007jf000759, 2007.
Wildman, M., Gallagher, K., Chew, D. and Carter, A.: From sink to source: Using offshore thermochronometric data to extract onshore erosion signals in Namibia, Basin Res, 33(2), 1580–1602, doi:10.1111/bre.12527, 2021.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 377 | 167 | 22 | 566 | 11 | 14 |
- HTML: 377
- PDF: 167
- XML: 22
- Total: 566
- BibTeX: 11
- EndNote: 14
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1