Review of “Understanding the Mid-Pleistocene transition with a simple physical model” by Perez-Montero et al 

This paper uses a recently published simple model, the Physical Adimensional Climate Cryosphere Model (Perez-Montero et al., 2025), to investigate several hypotheses about the Mid-Pleistocene Transition (MPT). They find that removal of North American regolith caused by repeated glaciations throughout the Pleistocene (i.e., the regolith hypothesis) is capable of reproducing the increase in ice volume amplitudes and switch to longer ~100-kyr cycle lengths in the mid-Pleistocene. The model also simulates increases in glacial-interglacial amplitudes (i.e., decreasing glacial minima) for CO2 and regional surface temperatures as a consequence of ice volume changes, which the manuscript argues demonstrates that a change in CO2 forcing is not necessary to drive these changes. The manuscript also includes many sensitivity tests to explore how the model responds to different parameter choices, insolation forcing, different constant CO2 levels, and other potential climate system changes in order to investigate mechanisms that are and are not associated with transitions in the model.

This manuscript is well written and thoroughly explores the dynamics of the model associated with MPT-like glacial cycle changes. The work is novel, well-executed and likely to be of interest to many paleoclimate researchers. However, there are a couple issues that should be addressed before its publication in Climate of the Past.

Major concerns

For comparing the model results to paleoclimate observations of the MPT, the manuscript (e.g., Figure 4) mainly relies on results from Bintanja & van de Wal (2008) and Berends et al (2021), both of which use forward inverse modeling to infer temperature and ice volume from benthic d18O. This is a concerning limitation of current manuscript because it relies on the accuracy of assumptions in these inverse models. While there are no direct observations of CO2 across and before the MPT, there are direct observations of regional temperatures to which the authors could compare their model results and there are estimates of ice volume that don’t rely on the same inverse modeling assumptions. The authors should compare their model results to some of these more observationally based estimates, such as

NH extra-tropical SST records:

Clark, P. U., Shakun, J. D., Rosenthal, Y., Köhler, P., and Bartlein, P. J.: Global and regional temperature change over the last 4.5 million years, Science, 383, 884–890, 2024.

McClymont et al. , Earth Sci. Rev. 123, 173–193 (2013).

Lawrence et al, North Atlantic climate evolution through the Plio-Pleistocene climate transitions, EPSL, 300, 329–342, 2010.

Ice volume estimates:

Elderfield, et al (2012) Evolution of ocean temperature and ice volume through the mid-Pleistocene climate transition, Science, 337, 704–709, https://doi.org/10.1126/science.1221294.

Rohling et al (2014) Sea-level and deep-sea-temperature variability over the past 5.3 million years, Nature, 508, 477–482.

Clark et al (2025), Mean ocean temperature change and decomposition of the benthic d18O record over the past 4.5 million years Clim. Past, 21, 973–1000, https://doi.org/10.5194/cp-21-973-2025

In describing these comparisons, it would also be appropriate for the manuscript to clarify that regional SST records would be expected to systematically differ from temperature over the ice sheet because of different sensitivities to ice sheet height.

Because the new pre-MPT ice volume estimates of Clark et al (2025) differ substantially from most other study’s reconstructions (and this model’s output), I recommend that the authors add some discussion of this new study somewhere in the manuscript. More generally, the model seems to produce very small ice volume estimates right up until the MPT (e.g., see 1500-1200 kyr BP in Figure 4f). Do the authors have any comments about this aspect of the model response? Is the model specifically producing estimates of Laurentide ice volume or all NH ice sheets?

Minor comments

• Lines 205-207: The sentence at the bottom of page 16 is somewhat difficult to parse. One way to clarify it (I think) would be “This improvement … trigger the MPT, whereas the decreasing trends … occur as consequences of changes in the cryosphere in this model.”
• Lines 223-236: The authors show model results for different CO2 trends in Figure 14 and on line 228 summarize the results by saying the carbon cycle affects the amplitude of GIV but “has little influence on the periodicities.” However, it is difficult to discern the periodicities of glacial cycles in Figure 14c, particularly because results from different model runs are overlain. The manuscript should provide more specific documentation of constant periodicity between the early and late Pleistocene in these simulations, perhaps by plotting NSP for a couple of the TREND-C experiments (e.g., similar to Fig. 15d).
• Line 266: Please specify which “two parameters” are referred to here.
• Line 289: The sentence “We cannot say which is the physical mechanism behind because of the simplicity of its implementation” is unclear. Perhaps one or more words is missing after “behind.”
• The citation for Perez-Montero et al. (2025) should be updated to the final, published version instead of the pre-print.

Review of Pérez-Montero et al., “Understanding the Mid-Pleistocene transition with a simple physical model” 



This study investigates the MPT using the conceptual but physically based model PACCO (Physical Adimensional Climate Cryosphere Model). The model incorporates insolation forcing, CO₂, ice dynamics, and the evolution of regolith layers. Modelling simulations suggest that the progressive removal of regolith slowed ice-sheet flow, allowing larger ice sheets that match the timing and amplitude of the MPT seen in proxy records. Simulations including CO2 trend but constant sediment layers are not conclusive for simulating MPT. Overall, the findings support the regolith hypothesis as a plausible mechanism for the MPT, but also that hydrological sensitivity may have contributed as well.

The approach addresses a crucial question in paleoclimate research: the trigger of the MPT. The choice of model to test the regolith hypothesis, related to the spatial distribution of regolith through time beneath the ice sheet is unconventional, given that the study relies on a 0-D physical model. The study effectively leverages the model flexibility by testing several hypothesis and conducting sensitivity tests. However, the results of the model require more thorough explanation to support the conclusions, in my opinion.

The manuscript is generally well written and structured. However, the discussion section lacks sufficient contextualisation with respect to previous studies, does not provide a critical assessment of the authors work, and suffers from a lack of clear structure.



Main comments: 



1. The regolith hypothesis is inherently geographically-based and has already been tested using 2D models (e.g., Willeit et al. 2019). In particular, the spatial evolution of regolith patches beneath the ice sheets is expected to play a critical role in their stability. For this reason, investigating the hypothesis with a spatially adimensional model is, in my view, not an obvious or straightforward approach. At a minimum, a section describing the implications and nuances of this approach choice in the discussion section is necessary. However, the manuscript currently lacks any explicit acknowledgment of model limitations or a critical appraisal of the robustness of the results.



2. The BASE simulation, which serves as the reference for sensitivity tests and comparisons across different hypotheses, does not adequately reproduce the amplitude of the 40 kyr world (Fig. 4f), but only approximately a quarter of its amplitude. As a result, the climate state preceding the MPT is not well simulated in the model. Indeed, the too low amplitude of the climatic 40 ka cycles likely preserve a large quantity of regolith until the start of the MPT.  How does this limitation affect the simulation of the MPT under the various hypotheses tested? A related question is the choice of insolation forcing. Using an alternative insolation metric (CST or ISI instead of SSI) allows the model to better capture the 40 kyr amplitude, yet in doing so the MPT itself is simulated much earlier than in the paleoclimate records, which is likely due to the larger quantity of regolith removed during the 40 ka cycles. The choice of insolation metric in the BASE model (SSI) versus alternatives such as CSI or ISI is not clearly justified and likely has a larger impact on the study results than the authors suggest.



3. This study is in line with a broader set of modeling efforts aimed at simulating the MPT over the past decade. Several of these studies, however, have reached conclusions that differ from those presented here. I strongly encourage the authors to expand the discussion by explicitly comparing their results with previous work. A concrete example, though not the only relevant one, is the recent publication by Scherrenberg et al. (2025) in this journal. It would be valuable to discuss whether the differences arise from the type of model employed, the assumptions underlying the hypotheses, or the specific formulation of the model.



4. I am quite surprised by the absence of effect of a decreasing trend in CO2, with concentrations starting as high as 700 (!) ppm down to the late Pleistocene values, on the triggering of the MPT. In another hand, the model is able to switch from the 40 to 100 kyr world only by changing hydrological component (i.e. accumulation sensitivity to temperature). Both of these results are quite different from the rest of the literature and would deserve further discussion and comparison with other studies in the discussion section.



Minor comments:



Title: Consider using “Exploring” instead of “Understanding” to reflect a less ambitious and more accurate scope.

Abstract: In my opinion, too much space is given to the introduction, with too little devoted to summarizing the study methods and results. The long sentence that carries all results (line 9-11) should be split into at least two shorter sentences. The last sentence is overly descriptive and does not capture the main conclusion.

Line 21: The citation of Chalk 2017 is not the most appropriate reference to introduce the concept of the MPT, go for a more historical one.

Line 23: This sentence should be moved up, as the described feature is an inherent part of the MPT itself, not an additional aspect.

Line 31: References are need to support the existence of the regolith. Currently, all cited works relate only to the second part of the sentence.

Line 24: The paragraph is quite dense. it would be better to start a new paragraph line 31 to improve clarity.

Line 50: The sentence is quite reductive of the work done, as the authors also test the CO₂ decrease hypothesis.

Line 51 and following: I would avoid describing the paper section by section. The current structure is standard and the descriptions here are not helpful because too vague.

Line 114: It should be mentioned that while there is a change in amplitude and frequency, the model does not reproduce the amplitude of 40 kyr cycles. It seems the BASE model transitions from a quasi-stable climate directly to 100 kyr cycles, which is not equivalent to reproducing the MPT. This is nuanced by the fact that the 41 kyr cycles are better captured using other insolation metrics. However, why were these other insolation not applied across all simulations, instead of SSI?

Line 115 and following: At this stage, it may be premature to draw such a conclusion. In my view, it is the constant sediment simulation that provides stronger evidence for that.

Figure 8: How is defined the Tmpt value ? I guess it is an arbitrary choice, but it needs to be justified in the text.

Figure 9: It would be interesting to do the same plot but with sea level value (as in fig. 4f) to see how it vary compared to the sea level curves.

Line 200: Interesting observation. However, the authors note that with CSI and ISIn, 40 kyr cycles are larger and thus remove sediments earlier. This relates to a key criticism of the regolith hypothesis: much of the regolith would already be removed during the initial 40 kyr cycles. Could the good match between modeled regolith removal and MPT timing be due to the BASE experiment’s failure to produce realistic 40 kyr cycles compared to data?

Line 205-207: I disagree with this conclusion. The change in frequency and amplitude occurs very early compared to the “real” timing of the MPT.

Line 224: The absence of an impact on the trend in frequency and amplitude is quite surprising, especially in light of previous modeling studies on the MPT (e.g. Scherrenberg et al., 2025, CP; Willeit et al., 2019; Science Advances).

Line 234: This section and the results here are quite surprising, and to my knowledge, the first modelling study that proposes the hydrological cycle as a trigger of the MPT. Line 244: The sentence here, will describing accurately the results of this study, sounds very surprising.  See main comment 4 for more details. 

Line 284 and after. The paragraph should be reworked, as is it poorly written: e.g.  “Of course, it is difficult to come to a final conclusion without proxies with a more accurate time resolution covering the Early Pleistocene.”



Typos:

Line 8: “sediment layers of sediments above the continents” 

Ligne 105 : “beyon the regoliths” 

Ligne 225 : allows

Ligne 255 : removes



References:

Willeit, M., Ganopolski, A., Calov, R., & Brovkin, V. (2019). Mid-Pleistocene transition in glacial cycles explained by declining CO2 and regolith removal. Science Advances, 5(4), eaav7337.

Scherrenberg, M. D., Berends, C. J., & van de Wal, R. S. (2025). CO 2 and summer insolation as drivers for the Mid-Pleistocene Transition. Climate of the Past, 21(6), 1061-1077.