We thank the reviewer for his review and thought-provoking comments.

• General comment:

We agree with the reviewer that the model we develop for duricrust formation by fluctuations of the water table might adequately represent other types of duricrusts than just ferricretes. We will modify the manuscript to ensure that our model is presented such that the ferricretes appear as one of the types of duricrusts that can form in this way. 

• Regarding the four general aspects mentioned by the reviewer:
  1. The model is based on sound depiction of physical processes, but it does not have any chemical component: 

     From this reviewer’s comment, it appears that our main objective, stated at the end of the introduction, i.e., “to present a simple, yet predictive numerical model to simulate the geometry and timing of duricrust formation on geological time scales, to predict their effect on surface processes and to compare them to observations”, was not explicit enough. We will improve this paragraph and move it to the start of the introduction.  We will also add a short paragraph justifying this approach and comparing it to more sophisticated models (as those mentioned by the reviewer) that are based on complex chemical processes but that require numerous parameters that are difficult to constrain and, most importantly, to know state variables characterizing the system (water chemistry, source rock and adjacent rock lithologies, etc.) that are very difficult to track over large temporal and spatial time scales.

     As the reviewer states it, our model is based on sound (yet simplified) physics that represent chemical weathering processes, which, as in many other models operating on similar temporal and spatial scales, do not contain explicit chemical process description. For example, the most-widely used model for soil (and regolith formation) assumes a simple exponential rate of propagation of the weathering as a function of the overlying regolith thickness (see Carretier et al, Geomorphology, 2014 for a nice review of this type of approximation, for example). Braun et al (2016)’s model is an attempt to improve on this simple relationship and is based on a well established relationship between water flow and weathering rate (Maher, 2010 and many others). As explained in Braun et al (2016), the constant of proportionality can be made a function of rock type or temperature, for example, but, for a more “generic” use of the model, i.e., over long temporal and spatial scales or to investigate its behavior under varying tectonic forcings and climates, it is adequate to consider this factor as representing a time and space averaged or integrated value.

     The duricrust formation model that we present in the current manuscript is another improvement of this representation of the weathering process, in which we include the fact that the product of the weathering can undergo hardening and in doing so affect its resistance to erosion, modify the landscape response and, as we show in our discussion, potentially lead to a feedback into the weathering process. This is the main justification for characterizing the duricrust by a single parameter, kappa, that we call the hardening parameter. As stated earlier, we will improve the introduction to cover these points more clearly.

  2. It is based on a misunderstanding of the mechanisms underlying the formation of iron duricrusts: 

     As discussed at length in our manuscript, it is well known that there exists much debate around various hypotheses for the formation of duricrusts. We agree on this point with the reviewer. The model we propose in this manuscript is based on the hypothesis that the formation of duricrusts relates to lateral transport of an element/ion that precipitates during fluctuations of the water table height, most likely in topographic lows, and thus, is similar to the lateral ferricrete hypothesis described by Bourman (1996) and Bourman et al. (2020) and other authors. Other proposed mechanisms include the in-situ hypothesis, as named by Bourman (1996), that we will test in another manuscript that is in preparation and the other type of lateral transport hypothesis of Bourman (1996) that duricrust form from the fragments of an older, dismantled duricrust (to quote Bournam (1996) describing his lateral ferricrete mechanism: “… lateral transport of physical particles or chemical precipitates derived from lateral sources.”).

     Consequently, we will improve the section of the manuscript describing the various hypotheses, using the elements brought by the reviewer. We will also make clear that these different hypotheses are not restricted to the formation of ferricretes. We will also improve the section that describes the most commonly accepted theories for each type of duricrust. In this way we think we will respond to the reviewer’s concern on this matter.

  3. Literature review on the topic ignores relevant information demonstrating that iron duricrusts are indeed long-lived and slowly eroding components of cratonal landscapes: 

     According to literature, duricrusts can become very old and this aspect has been mentioned in the article (section 2.3). The suggested literature about duricrust dating will be incorporated to improve the data. The suggestion to better differentiate between dating data according to different measuring categories is noted and modifications will be made accordingly.

     We fully agree that model results should not be taken at face value but confronted with observational constraints. The point we made about duricrusts being less efficient than stated in many studies in “protecting” landscapes is not at all dependent on the details of the model but is a simple result that can be derived from any hillslope/diffusion erosion model: introducing a thin hard layer is not an efficient way to significantly increase the time it takes for hillslope transport to erode a hill, regardless of the process that created the hardened layer. It does not mean that duricrust are not hard, but that even if they are hard (100 or more times than the rest of the hill) and thick (20% of the height of the hill) they can only retard the erosion of the hill by a factor 2. This conclusion is not inconsistent with the fact that duricrust can be very old: a duricrust can be billions of years old, until it is brought to the surface. And even when it is brought to the surface, it can resist erosion for billions of years until it is subjected to a noticeable base level fall driving erosion. We will make this point clearer in the revised version of the manuscript.

  4. When the model is applied it produces results that are not substantiated by other models or by observations and measurements of physical reality: 

     We assume that the reviewer makes a reference here to our simple computation showing that hard layers are inefficient at reducing the erosional timescale of a hill (see our point above) and to his criticism of the weathering model by Braun et al (2016). Regarding this last point, we do not believe that we should “defend” the physical soundness of the model that has been previously published. Furthermore, Braun et al (2016) addresses the concerns raised by the reviewer concerning the relationship between predicted regolith thickness and surface topography. As explained in Braun et al (2016), their model reproduces the observations of Rempe and Dietrich (2014) of a thicker regolith beneath hill tops in actively uplifting/eroding landscapes. It also predicts that in these situations, the water table must be very close to the regolith-bedrock interface, which is not a prediction but an assumption of Rempe and Dietrich (2014), which regard the bedrock as fractured and thus having a finite hydraulic conductivity. It is also worth noting that Rempe and Dietrich (2014) cannot predict any regolith thickness beneath the base level because, contrary to Braun et al (2016), their model assumes that it is nil there. It is also well accepted - see for example concerns expressed in Pelletier et al (2016) global model for regolith thickness – that Rempe and Dietrich (2014)’s model only applies to “uplands”, i.e., tectonically active areas. In non-actively uplifting/eroding areas, Braun et al (2016)’s model predicts indeed thinner regolith thickness beneath ridge tops. As argued in Braun et al (2016) this is true at many sites in Africa and India, where geophysical sounding evidenced that regolith thickness is thinner under hilltops or is relatively uniform beneath the topography (see data from Beauvais et al (1999) in Southern Senegal and Braun et al (2009) in India; see exact references in Braun et al (2016)). We recognize, however, that it is difficult to correlate regolith thickness and surface features in low relief slowly eroding terrains, where the position of channels may have evolved since or during the time of formation of the regolith.

• Regarding the line-by-line and section-by-section review:

Precisions and modifications in descriptions and possible confusions due to the chosen vocabulary will be addressed and adjustments will be made accordingly.

• Summary:

We thank the reviewer again for his thorough review. In summary, we believe that all of his concerns can be addressed (1) by improving the introduction of our manuscript to indicate more clearly what the objectives of the models are, (2) by not focusing its range of applicability to the formation of ferricretes, (3) by including a wider range of observations concerning the age and rate of formation of duricrusts as suggested by the reviewer, and (4) by better describing the implications of our findings for duricrust longevity and age, so that they do not appear to conflict with basic observational constraints.

We will therefore proceed with the preparation of a revised version.